<DOCINFO AUTHOR ""TITLE "Academic Voices: Across languages and disciplines"SUBJECT "Pragmatics & Beyond, New Series, Volume

148"KEYWORDS ""SIZE HEIGHT "240"WIDTH "160"VOFFSET "4">

Academic Voices
Pragmatics & Beyond

Editor
Andreas H. Jucker
University of Zurich, English Department
Plattenstrasse 47, CH-8032 Zurich, Switzerland
e-mail: ahjucker@es.unizh.ch

Associate Editors
Jacob L. Mey
University of Southern Denmark
Herman Parret
Belgian National Science Foundation, Universities of Louvain and Antwerp
Jef Verschueren
Belgian National Science Foundation, University of Antwerp

Editorial Board
Shoshana Blum-Kulka Catherine Kerbrat-Orecchioni
Hebrew University of Jerusalem University of Lyon 2
Jean Caron Claudia de Lemos
Université de Poitiers University of Campinas, Brazil
Robyn Carston Marina Sbisà
University College London University of Trieste
Bruce Fraser Emanuel Schegloff
Boston University University of California at Los Angeles
Thorstein Fretheim Deborah Schiffrin
University of Trondheim Georgetown University
John Heritage Paul O. Takahara
University of California at Los Angeles Kobe City University of Foreign Studies
Susan Herring Sandra Thompson
University of Texas at Arlington University of California at Santa Barbara
Masako K. Hiraga Teun A. Van Dijk
St. Paul’s (Rikkyo) University Pompeu Fabra, Barcelona
David Holdcroft Richard J. Watts
University of Leeds University of Berne
Sachiko Ide
Japan Women’s University

Volume 148
Academic Voices: Across languages and disciplines
by Kjersti Fløttum, Trine Dahl and Torodd Kinn
Academic Voices
Across languages and disciplines

Kjersti Fløttum
University of Bergen

Trine Dahl
Norwegian School of Economics and Business Administration

Torodd Kinn
University of Bergen

John Benjamins Publishing Company

Amsterdam
/
Philadelphia
 TM The paper used in this publication meets the minimum requirements
8

of American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Fløttum, Kjersti.
Academic voices : across languages and disciplines / Kjersti Fløttum, Trine
Dahl and Torodd Kinn.
p. cm. (Pragmatics & Beyond, New Series, issn 0922-842X ; v. 148)
Includes bibliographical references and indexes.
1. Academic writing. I. Dahl, Trine. II. Kinn, Torodd. III. Title. IV.
Series: Pragmatics & beyond ; new ser., 148.

P301.5.A27F58 2006
808.042--dc22 2006047709
isbn 90 272 5391 9 (Hb; alk. paper)

© 2006 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or
any other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
 Table of contents

Preface and acknowledgements ix

chapter 1
Introduction 1
1.1 The KIAP project 1
1.2 The KIAP Corpus 7
1.2.1 Corpus statistics 7
1.2.2 Text selection 8
1.2.3 Representativity 9
1.2.4 Text formatting 11
1.2.5 Text encoding 12
1.2.6 Search options 14
1.3 Cultural identity 16
1.3.1 Introduction 16
1.3.2 National identity 18
1.3.3 Academic identity 19
1.3.4 Disciplinary identity 20
1.3.5 Genre and discourse community 22
1.3.6 Cultural identity in KIAP 23
1.4 Previous research 25

chapter 2
Theoretical and methodological frameworks 29
2.1 Theoretical framework 29
2.1.1 Main theoretical approaches 29
2.1.2 A polyphonic framework: ScaPoLine 35
2.2 Quantitative methods 40
2.2.1 Our independent variables 40
2.2.2 Article length and absolute frequency 40
2.2.3 Relative frequency and measures based on it 41
2.2.4 Cross-linguistic comparisons and relative frequency 42
2.2.5 Statistical tests 43
vi Academic Voices

2.2.6 Why non-parametric statistics? 45

2.3 Qualitative methods 46
2.3.1 An illustration of method 47
2.3.2 Our use of case studies 49

chapter 3
Quantitative results 51
3.0 Introduction 51
3.1 Cultural identities as tendencies in linguistic practices 51
3.2 Is discipline or language the most important factor? 53
3.3 Differences and similarities between disciplines 55
3.3.1 Which differences between disciplines are significant? 59
3.3.2 Similarities between disciplines 60
3.4 Differences and similarities between languages 61
3.4.1 Which differences between languages are significant? 63
3.4.2 Similarities between languages 64

chapter 4
Presence of the author 67
4.0 Introduction 67
4.1 Quantitative and comparative results 73
4.1.1 First person subjects 73
4.1.2 Indefinite pronoun subjects 79
4.2 Author roles 81
4.2.1 Verb groups 83
4.2.2 Author roles and context 87
4.3 First person plural subjects 95
4.3.1 The reference and functions of first person plurals 95
4.3.2 ‘We’ in combination with two verbs 101
4.3.3 Differences between languages and disciplines 104
4.3.4 ‘We’ and polyphony 108
4.4 “Indefinite” authors 110
4.4.1 Indefinite pronouns in English, French and Norwegian 110
4.4.2 Previous studies on the French indefinite pronoun on 113
4.4.3 Six values of on in French research articles 115
4.4.4 Distribution of on-values 124
4.4.5 Final remarks 125
4.5 Author roles and evaluation in abstracts 126
4.5.1 Introduction 126
4.5.2 Author roles 127
 Contents vii

4.5.3 Evaluative elements 134

4.5.4 Summing up 137
4.6 Case studies 138
4.6.1 Pronoun use in individual articles 138
4.6.2 Implicit and explicit references to ‘we’ in a Norwegian
medical article 147
4.6.3 A study of two English abstracts 153
4.7 Concluding remarks 156

chapter 5
Reader/writer interaction 159
5.0 Introduction 159
5.1 Quantitative and comparative results 161
5.2 Metatext 163
5.3 ‘Let us’-imperatives 169
5.3.1 Types of ‘let’-imperatives 170
5.3.2 A classificatory overview of different uses 173
5.3.3 Functions of text composition 177
5.3.4 Rhetorical functions 178
5.3.5 Variation between disciplines, languages and individual authors 179
5.4 Presentation of “results” 180
5.4.1 Introduction 180
5.4.2 Quantitative analysis of RESULT 183
5.4.3 Discussion 191
5.4.4 Analysis of three linguistics articles 192
5.5 Case studies 202
5.5.1 Metatext in three economics articles 202
5.5.2 ‘Let’-imperatives in a Norwegian economics article 205
5.5.3 Metatext and author manifestation in a French medical article 210
5.6 Concluding remarks 212

chapter 6
Presence of the others 215
6.0 Introduction 215
6.1 Quantitative and comparative results 218
6.1.1 Bibliographical references 219
6.1.2 Negation 222
6.1.3 Adversatives 223
6.2 Explicit presence: bibliographical references and citations 226
6.2.1 Four subtypes 226
viii Academic Voices

6.2.2 Quantitative distribution of the four subtypes 230

6.2.3 Bibliographical references as grammatical subjects 231
6.2.4 Discoursal and rhetorical functions of bibliographical references 236
6.3 Who can do what? Self, other and verb use 238
6.4 Implicit presence: negation and concessive constructions 242
6.4.1 Introduction 242
6.4.2 Quantitative comments 243
6.4.3 The case of negation 243
6.4.4 The case of concession 247
6.4.5 Summing up 250
6.5 Case study: Negation in individual articles 251
6.6 Concluding remarks 255

chapter 7
Conclusions 259
7.0 Introduction 259
7.1 Quantitative results: summing up and discussion 260
7.1.1 Overview 260
7.1.2 The disciplines 261
7.1.3 The languages 263
7.1.4 Gender 265
7.2 Community identity versus individual heterogeneity 266
7.3 Academic voices in harmony and contrast 269

References 273

KIAP publications 285

Appendix A The KIAP Corpus 289

Appendix B Statistical tests 291

Index 307
 Preface and acknowledgements

Academic voices are of course present in all research articles, but to a different
extent in different languages, in different disciplines and in different articles. The
aim of this book is to present some observations and interpretations related to this
issue through studies undertaken by the Norwegian research project KIAP, Kul-
turell Identitet i Akademisk Prosa: nasjonal versus disiplinavhengig (in English: Cul-
tural Identity in Academic Prose: national versus discipline-specific; see www.uib.
no/kiap/), financially supported by the Research Council of Norway 2002–2006.
The project has been located at the Department of Romance studies, University of
Bergen, and led by Kjersti Fløttum, professor of French linguistics. Project mem-
bers have been Kjersti Rongen Breivega, postdoctoral research fellow (with a back-
ground in Scandinavian studies and linguistics), Trine Dahl, associate professor of
English linguistics at the Norwegian School of Economics and Business Adminis-
tration (NHH), Torodd Kinn, associate professor of Scandinavian linguistics, Jorid
Skiple, research assistant (cand.philol. of French linguistics), Anje Müller Gjesdal,
research assistant (cand.philol. of French linguistics and currently PhD-candidate)
and Anders Alvsåker Didriksen (cand.philol. of French linguistics). The part of the
project related to the establishing of the electronic KIAP corpus has been based at
Aksis, Avdeling for språk, kultur og informasjonsteknologi (in English: Depart-
ment of Culture, Language and Information Technology; part of Unifob AS). We
would like to thank Jarle Bareksten, Anne Lindebjerg, Johan Poppe and Øystein
Reigem at Aksis for valuable help and advice.
Students who have related their master’s thesis to the KIAP project are Pia An-
dersen, Anders Alvsåker Didriksen, Anje Müller Gjesdal, Lisbeth Grinde, Ingrid
Kilskar Næss, Elena Mihaela Rusu and Jorid Skiple. Doctoral candidates Eva Thue
Vold and Anje Müller Gjesdal have also based their research on the KIAP corpus
and have shown great commitment to the project activities.
Further, the KIAP project has benefited from cooperation with the follow-
ing research groups: L’Equipe Sémantique des Textes (Paris), headed by professor
François Rastier; Prosjektmiljøet Norsk Sakprosa (‘Norwegian Factual Prose’, Oslo),
headed by professor Kjell Lars Berge; Representing Specialized Knowledge (Vaasa),
headed by professor Merja Koskela; Constitution automatique de ressources lexi-
cales et élaboration d’outils didactiques multimédia pour l’enseignement du lexique
x Academic Voices

académique en langue seconde (Grenoble), headed by professor Francis Gross-

mann; Språk i contrast or SPRIK (‘Languages in contrast’, Oslo), headed by profes-
sor Cathrine Fabricius-Hansen.
The KIAP members are especially grateful to Céline Poudat (Orléans), mem-
ber of the Equipe Sémantique des Textes, for valuable help and advice in the estab-
lishing of the electronic corpus, and to Prosjektmiljøet Norsk Sakprosa for continu-
ously reminding us of the importance of contextual factors in a linguistic study
of the present kind. We are also grateful to Kolbjørn Slethei for valuable advice
related to statistical methods, and to Sandra Halverson for improving our English.
Any remaining flaws are naturally our responsibility.
The authors of this book would like to thank two anonymous referees for a
number of valuable comments and suggestions. Their contributions have un-
doubtedly made this a better book. Furthermore, we are grateful to the editors for
believing in the book project and for assisting us towards its completion.
Finally, I would like to thank all the participants for valuable input to the KIAP
project, for fruitful discussions through several seminars and for showing a con-
tinuous enthusiasm for the questions raised by the project.
Kjersti Fløttum
Head of the KIAP project
 chapter 1

Introduction

1.1 The KIAP project

Ever since Charles Bazerman’s work on the genre and activity of the experimental
article in science (Bazerman 1988), considerable attention has been paid to the
special character of the language of science, and with John M. Swales’ classic genre
analysis (Swales 1990), extensive focus has been placed on the structure of the
research article and its components. By attaching importance to the rhetorical na-
ture of academic discourse, both of these authors, as well as many other research-
ers, have contested the traditional conception of academic discourse as neutral
and objective. This fundamental “state of the art” inspired us to start the KIAP
project, or, by its full title, Cultural Identity in Academic Prose: national versus dis-
cipline-specific (www.uib.no/kiap/). We wished to explore to what extent and in
which ways research articles display traces of authors and their interaction with
the reader and the whole discourse community, in different disciplines, in differ-
ent languages, and in individual texts, with a particular focus on the manifesta-
tion of different voices. Our aim has been to describe and explain similarities and
differences between different language and disciplinary communities. Our main
motivation has been the need to obtain a more nuanced picture of the research
article, going beyond the language perspective isolated from the discipline per-
spective, and vice versa.
We formulated our key research issue as follows: Can cultural identities be
identified in academic discourse, and, if so, what are the major factors constituting
these identities? To what extent are they language-specific or discipline-specific in
nature? We have attempted to answer these questions through a comparative study
of English, French and Norwegian research articles within the fields of economics,
linguistics and medicine, looking for similarities as well as differences.
As regards the language issue, we have attempted to include articles written by
authors having the respective language as their mother tongue. However, we are
well aware that this is a problematic criterion. In the first place, it is possible that
some authors, especially those writing in English, are not English native speakers.
Second, English has developed in different ways in different parts of the world
(British, American, Australian, Indian, South-African, etc.). The same holds for
2 Academic Voices

French to a certain extent. In fact, the internationalisation of the research world

complicates the language issue in many respects. Nevertheless, we think that it is
meaningful to investigate cultural differences related to the use of different lan-
guages. We will return to the selection of both languages and disciplines in sec-
tions 1.2 and 1.3.
Throughout the project, our main hypothesis has been that discipline is more
important than language as regards cultural identity (in the sense that there will
be more similarities, for example, between a French and a Norwegian medical re-
search article than between a French linguistics and a French medical article). Our
investigations have confirmed this hypothesis in most respects. However, impor-
tant differences between languages have also been revealed. In addition, the differ-
ences between individual articles turned out to be greater than we had expected.
This seems to us a clear indication that the research article is in many respects a
flexible genre.
In a very general manner, we can sum up our results as follows: There are a
number of similarities within disciplines across different languages (for example
a relatively low frequency of first person pronouns in medical articles); there are,
however, notable differences between languages (as regards for example the use of
pronouns, negation and metatext). It is furthermore difficult to postulate a “typical”
research article, even within one and the same discipline and language, because of
great individual differences.
Being well aware of the fact that cultural identity is an extremely complex no-
tion, we emphasise that our purpose is not to give a general and comprehensive
description of the nine potential identities one might imagine within the frame-
work of KIAP: English economics, English linguistics, English medicine, French
economics, French linguistics, French medicine, Norwegian economics, Norwe-
gian linguistics and Norwegian medicine. However, considering culture as a com-
munity having a number of common interests (as a scientific community) and
as a community of people viewed as forming a distinct segment of society (as a
language community), we look for linguistic features which might indicate com-
mon identity traits in the nine communities studied. Section 1.3 elaborates on
the notion of cultural identity. The development of a culture is considered as a
two-way process with individuals interacting in forming and being formed. Four
settings which may have an impact on the identity of the academic author are
mentioned: the national/native language culture the author belongs to, leading to
a national/language identity, the world of academia, providing the author with a
general academic identity, the author’s chosen discipline and a disciplinary iden-
tity, and finally the genre and the discourse community settings.
In order to attain our aim, i.e. to describe and explain similarities and differ-
ences between the nine potential communities, we have chosen to base our work
 Chapter 1. Introduction 3

on an electronic corpus, the KIAP Corpus, consisting of 450 articles with 50 in

each of the nine subcorpora (see section 1.2).
KIAP is a linguistically based project, and it relates identity to a selection of
linguistic features which may be linked to argumentative and rhetorical strategies.
More specifically, focus will be on the following three questions:
• How do article authors manifest themselves in the texts?
• How are the voices of other researchers reflected?
• How do the authors present and promote their own research?

With these questions we signal a clear interest in the persons behind the discourse
presented, more than in the research reported itself. These questions are also
the reason why we have chosen Academic Voices as the title of the present book.
Our main object of study can be summarised in the notion of person manifesta-
tion. Our purpose is to search for person manifestation as realised through voices
and roles, both those assumed by the author and those assigned by the author to
others. This approach may be justified by a rhetorical view of scientific discourse
as something which is created in a particular multivoiced communicative situa-
tion. Research articles are rhetorical in the sense that they represent a text created
to induce cooperative actions and attitudes. The rhetorical aim of the research
article is to create convincing effects. In this book, however, our purpose is not
to study the effects of this type of rhetorical discourse; we want to focus on texts
as the product of the author’s creation (the inventive dimension of rhetorical dis-
course). We study how the rhetorical discourse of research articles is realised in
text, how a selection of linguistic features are used, how different academic voices
act interpersonally and become more or less explicit in a discourse traditionally
considered as neutral and objective. These academic voices represent what we may
call the self- and the other-dimensions. By self we refer to first person and by other
to both second and third persons.
There is not always a clear-cut distinction between the self- and the other-
dimensions when taking into consideration relevant features which may realise
them (see the list of studied features below). However, first person pronouns be-
long primarily to the self-dimension and bibliographical references to the other-
dimension. Metatextual elements unite the two dimensions in a self- & other-di-
mension in that they constitute links between the author (or the author’s text) and
the reader. The indefinite pronouns are treated in the same chapter as first person
pronouns, and negations and adversatives are treated in the same chapter as bib-
liographical references, even if these features may also unite the two dimensions
of self and other.
Our theoretical framework has been chosen in order to serve our main pur-
pose, i.e. the search for person manifestation as realised through voices and roles,
4 Academic Voices

and is therefore based on a broad interpersonal and polyphonic perspective. This

framework will be presented in Chapter 2.
To illustrate the relation between rhetorical cooperation, multivoiced interac-
tion and linguistic features, we have selected a set of linguistic features for investi-
gation. The selection is primarily based on a pilot study undertaken in the begin-
ning of the project (see Breivega, Dahl & Fløttum 2002) and on various explorative
studies undertaken throughout the project. We are not claiming that the study of
these features will provide anything close to a complete picture of the research
article. However, the features are, in our view, essential indicators of our main ob-
ject of study, i.e. person manifestation within the self- and other-dimensions as
defined above. The main features which have been subjected to statistical analysis
are the following six:
1) First person pronouns used as grammatical subject:
English I and we; French je/j’ and nous; Norwegian Bokmål jeg / Nynorsk
eg and Bokmål vi / Nynorsk me.1
2) Indefinite pronouns:
English one; French on; Norwegian man, en/ein.
3) Metatext
such as section and above (see section 5.2 for the full list)
4) Negation:
English not; French ne … pas; Norwegian ikke/ikkje.
5) Adversatives:
English but; French mais; Norwegian men.
6) Bibliographical references:
References including the years 19xx and 20xx.

The statistical analyses made us aware of many other interesting and important
aspects which we have pursued in more qualitative studies. We have broadened
our perspective and looked at, for instance, some lexical items and verbs combined
with different pronouns.
In order to transgress the simple description of linguistic observations, we
have to take the context in which the research article is created into consideration.

1. Norwegian has two standard variants, Bokmål and Nynorsk. When listing Norwegian ex-
pressions, we typically mention the Bokmål variant(s) before the Nynorsk variant(s). In the
study, we have treated the Norwegian communities as monolingual. We have no reason to be-
lieve that the Bokmål–Nynorsk issue affects metatext, negation, adversatives, or references. As
for pronouns, it should be mentioned that Nynorsk usage is more restrictive as regards passives
and other deagentivising features. There are few Nynorsk texts in economics and medicine, but
many in linguistics. This allows us to expect a higher frequency of first person pronouns and a
lower one of the indefinite pronouns in linguistics. (See also Blom & Gumperz 1972.)
 Chapter 1. Introduction 5

In this matter we consider the genre itself as an important context. The research
article has been developed within socio-professional frames with their own norms
and rules. We consider genre as a repetitive social practice manifesting itself ver-
bally in texts often characterised by specific compositional and linguistic features
(see Chapter 2). This social and rhetorical practice will to a large extent explain the
linguistic observations made by KIAP.
We are, of course, indebted to many researchers and research groups that have
studied research articles in different perspectives. Some of the features we have
studied have been investigated by these researchers, and we will discuss their find-
ings with ours to the extent that they are comparable. This applies primarily to
studies of English, and in some cases to French. For Norwegian no similar and
comparable studies have been undertaken. Even for English, where there are many
cross-disciplinary studies, these are comparable only to a limited extent because
they focus on aspects or features which are different from ours.
We hope that our doubly comparative perspective, taking into consideration
both language and discipline, combined with a focus placed on person manifesta-
tion as realised through voices and roles, will contribute to a richer understanding
of the research article.
As regards our hypotheses concerning the use of the main features studied,
these have been formed both on the basis of previous researchers’ work and ex-
pectations coloured by our own culturally based conceptions. With a point of
departure in the general conception of differences between academic traditions,
especially between so-called soft and hard disciplines, we expected linguistics and
medicine to be clearly opposed to each other and economics to occupy a middle
position. Similarly, the language issue led us to expect English and French to be
very different and Norwegian to be more similar to English than to French. Our
assumption about Norwegian is based on the fact that English and Norwegian are
related languages and that Norwegian culture has been more strongly oriented
towards the Anglo-American cultural sphere than to the Romance sphere. In ad-
dition, Norwegian academic culture is strongly influenced by the Anglo-American
world of research. More specific hypotheses concerning the features we study in
this book will be formulated in the relevant chapters.
The book is organised as follows: In this introductory chapter, we present the
KIAP Corpus (section 1.2), discuss the notion of cultural identity (section 1.3) and
outline relevant previous research (section 1.4). In Chapter 2 we describe our main
theoretical and methodological frameworks. Chapter 3 provides an overview of
our main quantitative results. The rest of the book is organised more or less ac-
cording to our three main research questions listed above as well as to the self- and
other-dimensions.
6 Academic Voices

In Chapter 4, we study the presence of the author (self-dimension, related to

our first research question) as manifested through the use of first person and in-
definite pronouns (which may include the author in their reference) and the verbs
combined with these pronouns. The pronoun and verb combination constitutes the
basis for our proposal of different author roles. Chapter 4 also includes a section on
author roles in abstracts. Chapter 5 is dedicated to reader/writer interaction (self
& other) and the way the author presents his or her own research through studies
of metatext, ‘let us’-imperatives and a few lexical expressions (mainly related to
our third research question). In Chapter 6, we look at the presence of others and
their voices (other-dimension, related to our second research question) through
the study of bibliographical references as well as some polyphonic constructions
(such as polemic negation where the author’s voice refutes an external voice). We
also compare the self- and other-dimensions by studying which verbs self and other
are combined with. In addition to the main thematic sections, these three chap-
ters each contain an introductory section setting forth our hypotheses, relating our
study to previous research, a section presenting our main quantitative findings, one
presenting qualitative case studies which primarily are meant to illustrate how the
studied features are realised in individual articles, and finally a concluding section.
In Chapter 7 we sum up and discuss our main results. In that chapter we also
deal with an issue which is central to our main research question: If there exists
something which can be called cultural identity in the world of research, as realised
in research articles, a definition of it has to include reference to the significant het-
erogeneity we have observed in individual articles. Academic texts display a great
diversity that cannot be explained solely by reference to discipline or to language.
Even though the three authors have worked together as a team in the KIAP project
and with this book, each author has special responsibility for certain chapters and
sections:
Kjersti Fløttum: 1.1, 2.1, 4.2, 4.4, 4.6.1, 5.4, Chapter 6, 7.3.
Trine Dahl: 1.3, 1.4, 4.5, 4.6.3, 5.0, 5.2, 5.5.1, 5.6, 7.2.
Torodd Kinn: 1.2, 2.2, 2.3, Chapter 3, 4.0, 4.1, 4.3, 4.6.2, 4.7, 5.1, 5.3, 5.5.2, 5.5.3,
6.1 (with Fløttum), 7.0, 7.1, Appendices.
The references section of the book contains, in addition to the general refer-
ences, also a list of KIAP publications. Some of these, but not all, appear on the
general list as well, which explains why some entries may seem irregular; in the
general list, we find, e.g., Fløttum (2003b, e, g, i), but not (2003a, c, d, f, h). The
latter are then listed under KIAP publications.
 Chapter 1. Introduction 7

1.2 The KIAP Corpus

1.2.1 Corpus statistics

The KIAP Corpus is a finite-size, machine-readable corpus consisting of 450 re-

search articles from three different disciplines, written in three different languag-
es by presumed native speakers (see 1.2.2). The number of articles is distributed
evenly over languages and disciplines: Each language – English, French and Nor-
wegian – is represented by 150 articles, 50 from each of the disciplines of econom-
ics, linguistics and medicine. Thus, there is a total of nine subcorpora, each with
50 articles.2
In terms of the number of words, the subcorpora vary considerably. Table 1.2-
1 shows the number of words in the subcorpora as well as the KIAP Corpus as
a whole. It contains the figures for article bodies only. That is, the figures do not
include front matter (titles, abstracts, author names, epigraphs etc.), back matter
(acknowledgements, references, endnotes (and footnotes, see 1.2.5), appendices
etc.) or more or less text-external body matter (tables and figures and their cap-
tions, mathematical formulae and linguistic examples in a separate paragraph,
direct quotes and section headings). When all these non-body text parts are in-
cluded, the number of words in the corpus rises to 3,152,022. The medical articles
constitute less than half the length of the economics and linguistics articles. Also,
the English articles are, on average, longer than the ones in the other languages.

Table 1.2-1. Number of body words in the KIAP Corpus and its subcorpora

Economics Linguistics Medicine Total

English 298,319 437,798 163,663 899,780

French 295,859 231,296 138,510 665,665
Norwegian 312,850 269,913 102,660 685,423

Total 907,028 939,007 404,833 2,250,868

The subcorpora have been given labels based on combinations of abbreviations

for language (English: eng, French: fr, Norwegian: no) and discipline (economics:
econ, linguistics: ling, medicine: med); thus, for instance, frmed means French

2. Of the Norwegian articles, 1 economics article, 12 linguistics articles and 4 medical articles
are in Nynorsk, the lesser used variety of written Norwegian. The skewed representation of Ny-
norsk reflects the fact that linguists, especially those writing about Norwegian, use this variety
more than researchers in economics and medicine. Except in the case of pronoun use, we have
no reason to expect the two varieties to differ for the features that we investigate quantitatively.
8 Academic Voices

medicine. We sometimes use the abbreviation engall for the union of English dis-
cipline subcorpora, i.e. engecon, engling and engmed together, and similarly frall
for French and noall for Norwegian. In the same manner, we use the abbreviation
econall for the union of economics subcorpora, i.e. engecon, frecon and noecon
together, and similarly lingall for linguistics and medall for medicine. The indi-
vidual articles have been given alphanumerical labels composed of language/dis-
cipline subcorpus abbreviations plus a two-digit number in the range 01–50; for
example, frmed34 is French medicine article number 34.

1.2.2 Text selection

As far as possible, we have chosen texts that match the category of ‘research arti-
cle’, i.e. texts that primarily report and discuss empirical (including experimental)
or theoretical research that has been done by the authors themselves (cf. Breive-
ga 2003: 61–63). Texts that merely review the research of others have thus been
avoided. However, in the case of Norwegian, the limited number of available texts
in economics and linguistics has made it necessary to relax this criterion a little.
Further, especially in linguistics, but also to some extent in economics, there is no
clear division between review and other types of articles. Most medical articles
deal with research with a rather large empirical basis, but some texts are case stud-
ies dealing with one or a few patients. In economics, articles have been chosen
that contain reasonably large amounts of ordinary text in addition to the math-
ematical formulae that characterise most texts from this discipline. In general,
the Norwegian economics articles are rather less mathematically oriented than
their English and French counterparts. In linguistics, articles about the respective
languages (e.g. English articles about English) have been preferred to texts about
other or several languages, and to texts with a typically general-linguistic perspec-
tive. Especially in the case of the English subcorpus, however, there are a number
of exceptions to this.
As already indicated, there are great differences between the languages in terms
of the availability of texts in the three disciplines, something which is due to the
varying numbers of academic journals. To guarantee acceptable scientific stand-
ards, only journals with peer reviewing were used. There is a plethora of English
journals, while the number of French journals is considerably lower. In the case
of Norwegian, there exists only one economics and one medical journal, while in
linguistics there is one linguistics journal and one mixed linguistic-philological
journal. Electronic accessibility also varies, and journals that are published elec-
tronically (in addition to paper versions) have, in general, been preferred to jour-
nals that are only published on paper. The English and French journals are mostly
publications recommended to us by experts from the respective disciplines. The
 Chapter 1. Introduction 9

Norwegian economics and medical journals are less common channels of publica-
tion for researchers of their fields than international journals published in English.
With regard to our two Norwegian linguistics journals, this may be true for re-
searchers of general linguistics, while the journals are well-established outlets for
researchers in the field of Norwegian linguistics.
Table 1.2-2 shows the distribution of single- and multi-author articles in the
various subcorpora. We have attempted to ascertain that both single-author and
multi-author articles are represented, and, for the single-author ones, articles writ-
ten by women as well as by men. However, in this respect the disciplines are vastly
different. While in economics both multi- and single-author articles are common,
medical articles normally have several authors, and linguistics articles typically
have one author. The selections of linguistics and medical texts are therefore highly
skewed in terms of authorship. As far as possible, texts have been chosen where
the author (or at least one of the authors) appears to be a native speaker of the
language in which the article is written (on the basis of his or her name and other
information available). Appendix A is an overview of which journals the indi-
vidual articles come from. A complete list of corpus articles can be found at www.
uib.no/kiap/KIAPCorpus.htm.

Table 1.2-2. Number of articles with one author and with several authors in the KIAP
Corpus and its subcorpora

Economics Linguistics Medicine Total

one several one several one several one several

English 23 27 35 15 01 49 59 91
French 31 19 48 02 04 46 83 67
Norwegian 33 17 48 02 10 40 91 59

Total 87 63 1310 19 15 1350 2330 2170

1.2.3 Representativity

Empirical findings based on a sample can only be generalised to populations that

the sample is representative of (cf. McEnery & Wilson 2001: 77–81). In the selection
process for the KIAP Corpus, we have attempted to ensure that the samples, i.e. our
nine language–discipline subcorpora, are representative of the population of research
articles published in refereed journals in the respective languages and disciplines.
Several considerations have been discussed above, such as selection based on our
definition of the research article as well as some additional discipline-specific criteria,
distribution over several journals (if available) and over types of authorship.
10 Academic Voices

However, we wish to point out some limitations that may affect representativ-
ity (some of these have already been mentioned). The limited number of journals
in Norwegian in particular has restricted our possibility to draw the samples from
more than one. Further, most of the French economics articles come from one
electronically published journal (which is also published on paper). We would
have preferred a more varied article selection for this latter subcorpus, but within
the given limits of finances and time for the KIAP project, this was simply not pos-
sible, since photocopying and scanning articles from paper and especially correct-
ing the scanned files would have required an inordinate amount of time.
The low number of journals in Norwegian does not in itself reduce representa-
tivity – on the contrary, the articles in the samples are probably highly representa-
tive of what is published in Norwegian in the three disciplines. But this situation
might reduce the degree of comparability with the English and French samples.
As mentioned above, the texts in the Norwegian economics corpus are typically
less mathematically oriented than those in the corpora for the other languages. It
is hard to assess to what extent the fact that the French economics corpus mainly
contains articles from one journal has reduced its representativity. Quite the oppo-
site problem may exist for some of the other subcorpora, where articles have been
culled from several journals: It is possible that the journals represent too divergent
research orientations, so that it could be claimed that the samples are drawn from
more than one population.
There are three potential problems in connection with the language and na-
tionality of authors that we wish to point out. We have attempted to ascertain for
each article that the author, or at least one of the authors if there are several, is
a native speaker of the article language. But an author’s nationality and mother
tongue cannot be decided on with certainty on the basis of his or her name. Thus,
it is possible that there are a couple of corpus articles written (only) by non-native
authors. We do not believe, however, that the number of these can be large enough
to jeopardise representativity.
Second, articles with several authors of which one (or more) is a non-native
speaker of the article language may well have that very person as its main author.
The number of non-native authors is greatest in medicine and smallest in linguis-
tics, primarily because of the different relative shares of multi-author articles in
the three disciplines, and it is greatest in the English subcorpora and smallest in
the Norwegian ones. This possibly makes some of our subcorpora slightly more
culturally diverse than others, but we consider this a minor problem.
Third, while Norwegian is the language of one country only, French is used as
a mother tongue in several countries and English in even more. This means that
the French and especially the English subcorpora are geographically, and quite
possibly culturally, more diverse than the Norwegian ones. The problem of lan-
 Chapter 1. Introduction 11

guage versus national identity is a complex one, and we discuss the issue further
in 1.3. Here we wish to point out the necessity of taking this difference into con-
sideration in the interpretation of our quantitative results. Although we admit that
this diversity is problematic, we do not think that the corpus would have become
much better had we insisted on some kind of cultural purity.
Another source of interference which should be pointed out is the guidelines
for authors and editorial policies of the various journals. We do not regard the lan-
guage of the texts in our corpus simply as the authors’ own preferred language use.
Rather, we see it as the result of complex interactions between the authors, written
guidelines, referees and journal editors. The degree to which guidelines for the
writing of articles as well as the demands and changes made by referees and editors
have affected the language of the authors no doubt varies considerably between the
disciplines as well as between the individual journals. This holds for text structure
and style as well as for grammar and lexical choices. In general, medical writers
operate within narrower limits than do economists and linguists. It is important
to realise that the claims we make about the linguistic usage of the language–disci-
pline populations (that we study by means of our sample corpus) are claims about
the language that authors use in the context of publishing in research journals –
and that context includes the demands and censorship of referees and editors. This
also, unfortunately, adds to the complexity of the question about representativity,
for it is possible that our samples are skewed in different directions with respect to
the degree and nature of such editorial interference.
In all our investigations, we have been aware of the limitations that the above-
mentioned weaknesses impose on us with respect to generalisability. Of course,
when we make statements about differences between subcorpora, we want to
make those statements about the text populations that the corpora have been
drawn from. However, our claims always imply a reservation: They hold only if
our samples are representative. By pointing out these limitations, we hope to have
warned the readers against drawing too rash and wide-ranging conclusions based
on the quantitative investigations in this book. Despite all these caveats, we risk
the claim that our doubly contrastive perspective and the large corpus base of our
studies, combined with our extensive qualitative work, provide us and our field of
research with valuable new insights about the differences and similarities between
languages and disciplines, as well as those between individual authors.

1.2.4 Text formatting

The text documents that have served as input for the corpus were in txt-format.
This means that special features of formatting such as font (e.g. IPA or Greek al-
phabet), font size and font style (italics, bold type, underlining etc.) have not been
12 Academic Voices

preserved. Similarly, all non-text features such as the graphic lay-out of figures and
tables have been excluded, although in some cases the textual content has been
preserved (as non-body text). These omissions have been considered harmless for
our research purposes, as most of the excluded matter is extraneous to the texts.
Remaining parts of figures and tables have uniformly been encoded and excluded
from the text bodies for search purposes. The loss of special fonts mainly pertains
to linguistic examples and mathematical formulae in the economics articles. The
details of such features are of little importance to our research. The loss of font
styles is regrettable, as special styles are often used for emphasis. However, it does
not directly affect the research questions of the present work.

1.2.5 Text encoding

To encode the corpus texts, we have used XML (Extensible Markup Language, cf.
http://www.xml.com/). The encoding has many characteristics in common with
the system defined by the TEI (Text Encoding Initiative, cf. http://www.tei-c.org/),
but is not fully TEI-conformant. Roughly speaking, the annotations encode three
kinds of information, and the tags that we have used may conveniently be divided
into three classes: extratextual tags, structural tags and textual tags. Tags are or-
ganised in pairs of the general form <start tag>text</end tag>. Certain start tags
contain additional information encoded in the form attribute=”value”. As an ex-
ample, in the annotated text selection ”<title type=”native”> Optimal card pay-
ment systems </title>” from engecon29, the tag pair ”<title> … </title>” says that
Optimal card payment systems is the title of the article, and the attribute value pair
”type=”native””, that the title is in the same language as the article.
The extratextual information is delimited by the tag pair <meta> … </meta>
and placed in front of the actual article text. It contains elements with information
about the name(s) of the author(s) and the bibliographical source, as well as a clas-
sification of the article with respect to four parameters, viz. authorship, language,
discipline and article structure. The tags are: <filename> for the alphanumerical
label given to the article, <author> for the name(s) of the author(s), <authorship>
for the classification into articles written by one woman, by one man or by several
persons, <source> for the journal that the article is taken from (name, volume, is-
sue, pages), <language> for the language that the article is written in, <discipline>
for the article discipline and <article> for IMRAD- vs. non-IMRAD articles. (IM-
RAD, or IMRD, is the relatively fixed article format found in some disciplines,
with the sections of Introduction, Method/Materials, Results and Discussion; see
Swales 1990.)
The structural tags mark up the major sections of the articles, such as the body
and its subparts (introduction, methods etc.), abstracts, notes, bibliography etc. The
 Chapter 1. Introduction 13

information thus encoded has been relevant because certain linguistic features tend
to be concentrated in certain article sections. The structural encoding also contrib-
utes to the division between body matter and body-external text. An outline of the
structural encoding of a typical medical article is shown in Figure 1.2-1.

Figure 1.2-1. The structural encoding of a medical article

Typical front-matter structural tags are <title> and <abstract>. Some French and
Norwegian articles have an English title in addition to their native title. Hence, the
tag <title> contains the attribute type, whose value is either ”native” or ”foreign”.
Abstracts may be given in the native language or, if this is French or Norwegian, in
English – or both or none of them. (A few French medical articles also have an ab-
stract in Dutch.) Consequently, the tag <abstract>, too, contains the attribute type
with possible values ”native” and ”foreign”. Abstracts are sometimes placed after the
body of the article.
Body-matter structural tags are <body>, <intro>, <mmr>, <disc>, <mid> and
<concl>. All articles have a <body> part. If an introductory text part is structur-
ally identifiable on the basis of headings or other layout features, it is enclosed by
the <intro> tag pair. Otherwise, the beginning of the text is treated as part of the
mid section of the article. Medical articles are typically structured according to
the IMRAD format, which means that there are sections dealing with material
(and methods) and results followed by a discussion. The encoding recognises this
by dividing the middle into a part containing materials, methods and results (en-
closed by the <mmr> tag pair) and a discussion (<disc>). In some articles, no such
division has been feasible, and the whole mid part has been included in either the
<mmr> or the <disc> section depending on which label is the more appropriate.
Economics and linguistics articles are more variably structured, and a <mid> tag
pair replaces the two tag pairs for medical articles, i.e. <mmr> and <disc>. Thus,
14 Academic Voices

article bodies from these disciplines are maximally divided into <intro>, <mid>
and <concl>. Not all articles have any identifiable concluding section. This is more
often the case for medicine, but articles from the other disciplines, too, may lack a
conclusion. Therefore, the body may end with the middle part.
Typical back-matter structural tags are <notes> and <references>. After the
body, there are generally notes and a bibliography. If an article contains footnotes
(i.e. the note texts are found at the bottom of the page), these have been placed
together with the back matter. The values ”end” and ”page” for the attribute type
distinguishes between endnotes and footnotes. All articles have some kind of bib-
liography or reference section, enclosed by the <references> tag pair.
Front and back matter that cannot be classified as belonging to any of the
above parts have been enclosed by the <misc> tag pair (for miscellaneous). This
includes appendices, acknowledgements, contact addresses, epigraphs etc.
Four textual tag pairs have been used, viz. <subtitle>, <quote>, <example>
and <table>. The purpose of this part of the encoding has been to exclude text por-
tions that are more or less outside the text proper, or, in the case of the <quote> tag
pair, are not the author’s own words. The <subtitle> tag pair encloses section head-
ings. The <quote> pair encloses direct quotes of at least three words; the reference
to the quoted text is not included. In linguistics articles, linguistic examples that
have been put in separate paragraphs or consist of more than one word have been
encoded with the <example> tag pair. Tables, figures, mathematical formulae and
similar matter have been enclosed indiscriminately by the <table> tag pair. Figure
captions and similar texts belonging to such features have been included here (and
thus excluded from the body text).
In addition to XML-coding, a couple of types of annotations have been made
in the texts, notably ”nRRR” in front of numbers for bibliographical references in
medical articles (where n is the variable number of works referred to) and ”NNN”
in front of note numbers. Such annotations have been used in order for these text
features to be searchable.

1.2.6 Search options

Collaborators of the KIAP project in the Research Group for Language Technol-
ogy at the Department of Culture, Language and Information Technology (Aksis),
University of Bergen, generated the corpus from the encoded texts and developed a
custom-made search program and interface that can be accessed over the internet.
The corpus search interface allows searches for single words or collocations of
two or three words. Wild cards (non-specified letters) are allowed, and truncation
is possible in both ends of words. Thus, the search string ”analy.e” applied to the
English texts will return examples of both analyse and analyze, and ”.*regard.*”
 Chapter 1. Introduction 15

will yield examples with regard, regarding, regardless, disregards etc. There are fur-
ther options for case sensitivity and searches involving punctuation. In collocation
searches, the spans between the first and the second and between the second and
the third word can be specified in terms of the number of words. The context
before and after the search items presented in the search result can be specified
in terms of the number of characters. Different ways of alphabetising the search
results can also be specified. Searches can be restricted either to one particular
article or to sets of articles defined by language, discipline and type of authorship,
and it is possible to search in specified parts of the articles (such as the body, the
abstract or the discussion) or the whole articles.
Figure 1.2-2 shows the interface to the KIAP Corpus search program, specifi-
cally a specification that the program should look for the sequence we will argue,
with no intervening words, in the body of English linguistics articles written by sev-
eral authors. The program will return hits (if there are any) with a left context of 20
characters and a right context of 60 characters. The result of this search is the con-
cordance shown in Figure 1.2-3. There are eight hits, listed in the sequence in which
they can be found in the specified corpus section. The result is here presented in a
Java applet window. It is also possible to see it in an ordinary browser window.

Figure 1.2-2. An example of the use of the KIAP Corpus search program interface
16 Academic Voices

Figure 1.2-3. The result of the search specified in Figure 1.2-2.

The encoded KIAP Corpus texts can be accessed via hyperlinks in the browser
search result window, or alternatively from an internet page with links to each of
the 450 articles. The texts can then for instance be read in a browser window or
searched in by using the browser’s own search function.

1.3 Cultural identity

1.3.1 Introduction

Culture and identity are both concepts that may be defined in a number of ways
depending on the context in which they are used. The anthropologist Clifford
Geertz gives his view of culture in the following way:
Believing, with Max Weber, that man is an animal suspended in webs of signifi-
cance he himself has spun, I take culture to be those webs, and the analysis of it to
be therefore not an experimental science in search of law but an interpretive one
in search of meaning. (Geertz 1973, in Redding & Stening 2003: 5)

Identity most often refers to human beings, often individuals, but also groups of in-
dividuals. When used in combination, culture and identity indicate some kind of in-
terplay between their meanings. This implies that the identity of a person or group
will be influenced or even shaped by culture while at the same time the person/
group will influence and shape that culture. The development of culture is hence a
two-way process with individuals interacting in forming and being formed.
In this book our aim is, of course, not to outline and discuss all the possible
uses and contexts for either culture or identity; we shall restrict our discussion to
angles relevant for the context that we have established, viz. academic authors and
 Chapter 1. Introduction 17

their texts. Our object of study is person manifestation realised through voices and
roles. Texts represent the site where we look for linguistic manifestations of voices.
Having identified these manifestations in texts from different disciplines and in
different languages, we would like to contribute to explanations of why the texts
look the way they do. At that point we link up with culture and identity. The texts
have not been produced in a void; they are very clearly the products of a contextual
setting marked by a multitude of factors. The key role played by text in cultural
studies is reflected in the following quote from Anna Mauranen:
[T]exts are in themselves one of the main keys to understanding a culture. Texts
as cultural products act out relevant social relationships within the culture, and in
this way provide keys to understanding themselves as well as other aspects of the
culture. (Mauranen 2001a: 53)

The author is also in his or her turn a product of a contextual setting or, more pre-
cisely, many settings. We need to identify the factors which are likely to be relevant
in a discussion of how the author’s academic voice comes to sound like it does. The
use of the hedge likely in the previous sentence should be read as an acknowledge-
ment of the fact that we are leaving the − more or less − precise world of linguistic
analysis. How can we select factors in the author’s surroundings and justify that
they have made an impact on the way he or she behaves in the role of academic
author? The answer is that we can only offer our selection and hope that the reader
finds them relevant and useful. Geertz, in his 1973 paper, discusses how to make
sense of the data from anthropological fieldwork: ”Analysis […] is sorting out the
structures of significance” (Geertz 1973, in Redding & Stening 2003: 5). This is
what we attempt to do as well.
We are not purporting to present a comprehensive list of aspects that may be
relevant in the forming of an author’s academic voice. So many different variables
come into play that this would be an impossible task. In the rest of this section we
will introduce and briefly discuss what we consider to be the contextual factors
that are important in shaping the academic author. Going back to our key words
culture and identity, we see four settings, or cultures, that each in their own way
will have an impact on the identity of the academic author. The first is the national/
native language culture the author belongs to, leading to a national/language iden-
tity. This setting is dealt with in section 1.3.2. The second is the world of academia,
providing the author with a general academic identity (1.3.3). The third is the au-
thor’s chosen discipline and a disciplinary identity (1.3.4). Then, we suggest that
aspects related to genre and discourse community should be considered a fourth
setting (1.3.5). Finally, in section 1.3.6 we consider how the four settings may be
relevant to our study.
18 Academic Voices

1.3.2 National identity

The academic author obviously comes with a history as a member of a national

culture. Belonging to a particular culture with particular values reflected in, inter
alia, social hierarchies and educational systems, the individual is socialised into
that culture. As already stated, the language issue is a difficult one to handle. It is
of course not the case that all nations have their own unique language, and some
nations have more than one. In the case of the present study, two of the languages
represented, English and French, are the official language in several countries,
while the third language, Norwegian, is used as a native language in Norway only.
English is commonly accepted as the lingua franca in large parts of the world
today, and in the academic world it occupies a privileged position (cf. Montgomery
2004). It is the first language of Britain, the USA, (parts of) Canada, Australia, New
Zealand and various other countries around the globe. It thus becomes difficult
to assume that native users of English represent one and the same national cul-
ture. However, in terms of writing culture, especially in an academic setting, Eng-
lish-language texts seem to display similarities across national boundaries, based
primarily on Anglo-American norms. The same situation seems to hold for the
francophone world, with French writing culture as the norm. It is also often the case
that nations sharing a language share cultural values in general. Intercultural stud-
ies (Hall & Hall 1990; Triandis 1994; Hofstede 2001) provide evidence of this fact.
Hinds (1987), in a typological study of various languages with regard to the
communication situation, establishes the two categories of reader-responsible and
writer-responsible languages. As regards his own culture, he maintains that ”[t]he
desire to write or speak clearly in English permeates our culture” (Hinds 1987:
144), while for instance in Japan ”it is the responsibility of the listener (or reader)
to understand what it is the speaker or author had intended to say” (ibid.). Hinds
goes on to claim that ”[t]his difference in the way of looking at the act of com-
munication permeates the thoughts of anyone who operates as a functioning Japa-
nese-American bicultural” (ibid.). His ideas are reflected in Mauranen’s discussion
of metatext in Finnish and English:
The careful and explicit guidance practised by Anglo-American writers, together
with frequent signalling of the personal presence of the author […] conveys the
impression that the reader is invited to take a tour of the text, together with the
author, who acts as a guide. (Mauranen 1993b: 16)

Fredrickson & Swales (1994: 18) note that as concerns contrastive rhetoric studies,
”[o]ver the last decade this field has rightly moved beyond stylistics and textual rheto-
ric to include the investigation of such areas as educational and cultural traditions”.
 Chapter 1. Introduction 19

In addition to such aspects, it is likely that each language ”behaves” in ways

which are not always possible to explain by reference to cultural variables. Certain
lexemes or linguistic constructions may be highly frequent, while other, semanti-
cally very similar items are less used.

1.3.3 Academic identity

This identity is connected to scientific communication in general. It is an integral

part of the present work, since the texts studied represent academic discourse. The
author’s main task is to communicate new knowledge, but may also be to confirm
and support already existing knowledge. For a novice student writer, the proc-
ess of developing into a mature academic scholar is a long and challenging one,
involving a number of stages. Learning to take part in the continuous dialogue
which academic communication represents implies that the author must develop
a persona willing to claim authority as a scholar. Only then can his or her research
contribution be presented in a rhetorically convincing manner. The aspect of per-
suasion in scientific discourse is stressed by Prelli (1989), who claims that ”[l]ogic
and experimentation are not the fundamental means of securing scientific change.
They are efficacious only if applied persuasively” (Prelli 1989: 100; italics in origi-
nal). Along the same lines Berge (2003: 155) describes the scientific writer as ”a
rhetorician trying to establish the probability of his or her interpretation”.
The process of developing the identity of the academic writer is thoroughly
discussed in Ivanič (1998). From her social constructionist position, she posits
four, clearly interrelated, aspects which are relevant as the student’s identity as an
academic writer is being formed. The first, called the autobiographical self, is ”the
identity which people bring with them to any act of writing, shaped […] by their
prior social and discoursal history” (1998: 24); the second is termed the discoursal
self, defined as the impression a writer conveys, consciously or unconsciously, in
a particular written text, constructed through the discourse characteristics of the
text, which are said to relate to values, beliefs and power relations in the social
context in which they were written (1998: 25). The third aspect is called self as
author, and is described as a relative concept, since ”writers see themselves to a
greater or lesser extent as authors” (1998: 26). It is not easily distinguished from
the discoursal self, but relates particularly to the writer’s ability to assert his or her
authority in the text. Finally, she establishes an aspect termed possibilities for self-
hood, denoting the abstract prototypical identities available to writers in the social
context of writing (1998: 27).
Hyland (2002a) deals with the issue of writer authority in a study of Hong
Kong students’ use of personal pronouns, claiming that the connotations of au-
thority conveyed by ‘I’ are problematic for many users in the student group. In ad-
20 Academic Voices

dition, cultural factors come into play, as Asian cultures have different perceptions
of the notion of the individual from those projected by cultures where English is
the first language (cf. Hofstede 2001).

1.3.4 Disciplinary identity

Becher & Trowler (2001), taking an ethnographic approach to the investigation of

the interplay of academic cultures (called tribes) and disciplinary knowledge (their
territories), maintain that ”[t]he changing nature of knowledge domains over time
has its impact on the identities and cultural characteristics of disciplines” (2001:
43). As regards who or what is responsible for identity and cultural construction,
they refer to two kinds of approaches, viz. constructivist approaches, stressing the
role of the individual, and situated practice approaches (ibid.: 48), putting the main
focus on the community.
Becher & Trowler also discuss the globalisation process taking place in the
academic world, and find that despite national traits and traditions, ”we may ap-
propriately conceive of disciplines as having recognizable identities and particular
cultural attributes” (ibid.: 44). These global identities linked to academic disci-
plines must then be acquired by the student through a socialisation process. As
mentioned above, such a process involves not only mastering the knowledge base
of the discipline; as knowledge claims have to be communicated, mastering of the
rhetoric of a discipline (Prelli 1989) becomes a vital part of the identity building.
The three disciplines that we study here, medicine, economics and ling-
uistics, may be seen as representatives of the three branches of science, i.e. the
natural sciences, the social sciences and the humanities. The classification of disci-
plines into branches of science is of course always difficult. For example, linguistics
is often described as situated at the crossroads of the humanities and the social sci-
ences (or even as extending across all three branches of science). As regards medi-
cine, Kjørup (1996: 18) suggests that it may be considered a separate branch of
science. Medicine is today a field encompassing aspects from other natural science
disciplines, such as biology and chemistry, but also from disciplines within the
social sciences, such as psychology and anthropology. We have, as already stated,
chosen to follow the more traditional view here, and classify medicine as part of
natural science.
The three branches represent knowledge bases with different characteristics.
The natural sciences may be said to have a relatively homogeneous and objective
knowledge base, and within disciplines belonging there ”[s]cientists act as if they
see themselves as discovering the truth, not making it” (Hyland 2000: 33). The
social sciences occupy a middle position between natural science and the humani-
ties (see below), with some of the knowledge base rooted in more objective natural
 Chapter 1. Introduction 21

science matters, e.g. mathematics, while other parts represent more subjective, in-
terpretive aspects. Ekelund & Hébert (1997) point to an apparent struggle within
economics, illustrating this situation: ”Many ‘mainstream’ economists argue […]
that increased respect for economics as a separate, scientific discipline will only
ensue from the steady application of rigorous mathematical and statistical tools”
(1997: 576); however, ”[s]ome economists […] have voiced strong reservations
concerning this view. They argue that the nature of social science makes exact
formulation and verification impossible” (ibid., italics in original). A quote from
Bloor & Bloor (1993: 157) may serve to illustrate how this struggle is visible at
the textual level: ”Surprisingly, many of the economics articles do not include a
straightforward ‘results’ section of the type that is common in scientific papers”
(i.e. natural science papers). Bloor & Bloor refer to one of their corpus texts with a
section headed ‘Interpretation of the Results’, which is preceded by ”a detailed ex-
planation of why the interpretation of the results ‘is not as straightforward as may
appear at first sight’” (1993: 157–158).
The third branch, the humanities, is in many contexts considered to be herme-
neutical, with subjective interpretation as a prominent feature. It is, of course, not
the case that all knowledge claims made within the humanities are based solely
on subjective considerations by the individual researcher. The discipline we study
here, linguistics, is certainly also characterised by regularities and − up to a point
− common methodological tools. However, the humanities researcher may, to a
larger extent than in the other two branches, be said to create new knowledge
through his or her arguing for relations between facts and notions. Hence, the
”results” of research in the humanities may take the shape of the actual argumen-
tation presented in text (Breivega 2003: 36). As indicated in the classification of
the three branches of science above, it might be argued that linguistics is not a
prototypical humanities discipline, and that, say, history or literary theory and
analysis might have been better choices. Clearly, in addition to overlapping traits
with the social sciences, some fields within linguistics even share features with the
natural sciences. In our text selection process we have, however, primarily looked
for articles with subject matter that tended more towards the humanities than the
natural sciences.
This brief and somewhat simplistic attempt at describing the three branches
of science has only scratched the surface of the debate on which disciplines belong
to which branch of science. As e.g. Becher & Trowler (2001) clearly demonstrate,
the shape of the academic landscape is not given once and for all, but is very much
the product of changing traditions, caused by social and political trends as well as
financial factors.
22 Academic Voices

1.3.5 Genre and discourse community

A fourth setting influencing the academic author is created by the discourse format
and writing traditions used to present knowledge claims in concert with the dis-
course community representing the audience for the text. Mastering the relevant
discourse forms of a discipline may be said to be part of the socialisation process
into the academic world in general and the discipline in question in particular.
However, as there are several issues linked specifically to genre and discourse com-
munity, we feel that they warrant separate consideration here.
The choice of appropriate genre for presenting knowledge claims is to some
extent linked to the epistemology of the discipline. Within medicine, the experi-
mental research design is well handled by the IMRAD format, while for instance
philosophy is a discipline marked by personal reflection, something which the
essay format is well suited to render. However, text linguistic research has also
shown that there is no straightforward interdependence between a discipline’s
epistemology and the text traditions used to present knowledge claims (Breivega
2003: 248).
As regards the issue of audience, several recent genre studies have pointed
to factors relating to differences in discourse community. Fredrickson & Swales
(1994) present the hypothesis of discourse community pressure, and maintain that
this is a variable in text analysis. The greater the competition in a territory, the
greater the rhetorical effort authors have to spend in order to create a research
space for themselves (ibid: 11). Hence, the way knowledge claims are argued in
a crowded field may take a different textual shape than those argued in a more
sparsely populated one. A related issue is raised in Melander (1998). In a com-
parison of scientific article introductions in English and Swedish within the two
disciplines of medicine and biology, Melander finds that the medical articles in
Swedish are aimed at a different audience (practitioners) than those in English
(researchers).
Another cultural factor related to discourse community is also mentioned in
Fredrickson & Swales (1994). They refer to a study by Taylor & Tingguan (1991),
comparing article introductions by Anglo-American and Chinese researchers.
Taylor & Tingguan found that, in contrast to the Anglo-American writers, the
Chinese researchers (both when writing in English and in Chinese) tended to omit
the step of discussing previous literature in the field. Taylor & Tingguan suggest
that a possible explanation may be that the Chinese writers, in accordance with
their cultural values, try to avoid the potentially face-threatening act of pointing
to shortcomings of fellow researchers in the field.
The issue of discourse community may be said to have a bearing on the present
study as well. As Norwegian is a language used by less than five million people, it
 Chapter 1. Introduction 23

is rarely the sole language used by Norwegian researchers when mediating their
findings (Kyvik & Larsen 1997). Not surprisingly, English is the preferred foreign
language choice. Without going into language policy and domain loss issues here,
let us just note that Norwegian scholars frequently publish in English and that for
medicine and economics at least, this is the preferred language. A consequence
of this fact is that for the disciplines of medicine and economics there is only one
Norwegian journal, while for linguistics there are two. An informal survey car-
ried out among economists at the Norwegian School of Economics and Business
Administration indicated that the economics journal is considered to be a less
prestigious outlet for research than international journals (cf. section 1.2.2). The
contributors to this journal are also likely to be more junior researchers than those
represented in the English-language journals of our corpus. The subject matter
of the texts is also somewhat more general than that found in the international
journals. However, as the Norwegian journal is also read by more mature scholars,
we feel justified in claiming that the Norwegian articles belong to the same genre
as the English ones.

1.3.6 Cultural identity in KIAP

So, how can we make use of these reflections on cultural identity in our study?
As regards national identity, we believe that there exists a writing culture which
each of the three languages investigated here belongs to. This writing culture is
strongly related to cultural values as specified in intercultural studies, represented,
for instance, by the concepts of high and low context (Hall & Hall 1990), and is
reflected in the communicative ideals and aims promoted in educational settings.
In addition, it is likely that lexical preferences in each language may have an effect
on our study. Pilot investigations into our material, e.g. Dahl (2003), have shown
this to be a relevant issue.3
When it comes to academic identity, we see Prelli’s (1989) ideas as highly rel-
evant. His view of academic rhetoric as primarily persuasive in nature fits nicely
in with our discussion of author presence in text. The notion of writer authority,
discussed by Ivanič (1998) and Hyland (2002a), also represents an interesting vari-
able. However, the nature of our corpus (expert rather than student writers, and
native rather than non-native writers), calls for a somewhat different focus. We
will, however, briefly consider the gender factor, and in that context the notion
of writer authority is of interest. Studies focusing on gender issues have indicated
that women and men may differ in their discourse strategies with regard to, e.g.,

3. In that study it was shown that while English economists typically ‘report’, Norwegian
economists ‘discuss’.
24 Academic Voices

assertiveness and politeness (Lakoff 1975; O’Barr & Atkins 1980). In the present
study, however, we have found little evidence that there are gender differences with
respect to visibility in the texts.4
As concerns disciplinary identity, previous text linguistic research (e.g. Breive-
ga 2003) has indicated that a discussion of linguistic behaviour in academic texts
should take into account the knowledge base of the field in question. In addition,
descriptions of how scientists within a specific discipline perceive its epistemologi-
cal status (cf. Ekelund & Hébert’s reflections on economics, above) may add useful
insight into author behaviour. In our lexical investigations into, e.g., how research
findings are given a linguistic realisation in the three disciplines we will attempt to
draw on such epistemological aspects. We also take into account Becher & Trowl-
er’s reflections on the globalisation process observable in academia (1.3.4), but at
the same time we wish to focus on the writer as an individual.
Finally, there is the interpretative context represented by genre and discourse
community. Our corpus texts are all defined as belonging to the genre of the re-
search article. However, the structural realisation of the genre differs between the
three disciplines, with the IMRAD pattern for the prototypical medical text, and
some IMRAD-like texts and some less standardised texts for economics and lin-
guistics. Text structure will be relevant in discussions of some of the selected lin-
guistic features. The notion of discourse community will be considered when we
deal with issues relating to the differences in status enjoyed by the three languages
in an academic setting.
We would like to end this section with a quote from Bhatia (2001), illustrating
what is, as indicated above, an important point in our study, namely the fact that
the academic author, like any author, is also an individual, sometimes displaying
discoursal behaviour that goes against cultural norms and traditions:
[…] it is necessary to emphasise that discourse and genre analysis is an area of
enquiry that is and will always be fuzzy to some extent, even though we often find
it based on highly conventionalised and standardised social actions (Miller 1984;
Swales 1990). In analysing language use, we are invariably investigating human
behaviour, which is not entirely predictable, for the simple reason that most of us,
even in highly predictable settings, like to exploit conventions to express ‘private
intentions’ within the framework of socially constructed discourse forms (Bhatia
1995). (Bhatia 2001: 79–80)

The fact that we are describing human behaviour will serve as a natural backdrop
for our discussion of the academic voice throughout this book.

4. In a KIAP context, the issue of gender and scientific writing is discussed in Gjesdal
(2005).
 Chapter 1. Introduction 25

1.4 Previous research

The present study rests on and builds from what is by now a fairly large collection
of studies within the field known as academic discourse. The term refers to at least
two strands of research. The first involves studies which are pedagogically ori-
ented, focusing on student needs and competences. The proliferation of courses on
academic writing in general and English for academic purposes in particular has
entailed increased research activity into what language and communication tools
the students must acquire to become fully socialised into their research commu-
nity. In such contexts, the process of gaining entry into these communities is seen
as being dependent on awareness of and competence in the writing practices of the
relevant discourse community (Hyland & Hamp-Lyons 2002). The second strand
of research forms the basis for the first; it concerns studies on how expert writers
within a discourse community communicate with their peers. In many works the
two aspects, i.e. investigation of professional communication practices among ex-
perts and pedagogical issues relevant for novice communicators, go hand in hand.
Typical contributions may be found in Ventola & Mauranen (1996) and Candlin
& Hyland (1999).
As our study belongs primarily within the second strand of research, describ-
ing the practices of disciplinary professionals, this account of previous research
will primarily, but not exclusively, adopt the same angle. Academic discourse stud-
ies comprise several key concepts which, it might well be argued, merit separate
discussion. Cases in point are for instance ‘genre’ and ‘discourse community’. These
two concepts represent vital aspects of our work and have already been introduced
as representing a contextual setting for discussing our data (see 1.3). Genre is also
discussed in some detail in the presentation of our theoretical framework (Chapter
2). The work undertaken by Swales (primarily 1990, but also 2004) on the genre of
the research article today serves as a natural backdrop for the lion’s share of stud-
ies on academic discourse. As regards ‘discourse community’, this is a concept we
bring in at several points, but notably in our discussions of findings related to dif-
ferences between the languages involved (since the three languages operate in com-
munities which are very different both in terms of size and geographical spread).
Studies which are relevant to mention in the present context have mainly been
limited to those involving the three languages we deal with, viz. English, French
and Norwegian, and further, primarily those dealing with the research article.
Even though the section is intended as an account of previous research in the
field, we have chosen to focus on very recent studies; some are even reports from
currently ongoing projects. The reason for this is the sheer size of the field. The
privileged position of English in the world of science today has made that language
the natural reference point for contrastive studies within academic discourse. Ref-
26 Academic Voices

erences to specific studies will be provided in connection with our discussions of

the various issues dealt with throughout this book. In the present context we have
chosen to focus on a few authors who have published extensively on academic
discourse in studies similar to ours. For English, Ken Hyland is such an author.
During the past decade he has written books and papers on many features of the
research article, such as hedging (e.g. 1998a), metadiscourse (e.g. 1998b, 2005)
and authorial stance (e.g. 2002a). He has studied these features in research articles
across various disciplines. In contrast to the KIAP Corpus, Hyland’s corpora have
comprised texts produced by learners as well as professionals in the various fields.
Another researcher who deals with issues that border on our own is Marina Bondi.
She studies English research articles within the disciplines of economics and his-
tory (e.g. Bondi & Silver 2004; Bondi 2005), and compares, among other things,
how so-called textual voices are used in the establishing of knowledge claims in
the article introductions.
When it comes to studies involving English and one or several other languag-
es, we have particularly been inspired by works by, in alphabetical order, Anna
Mauranen, Françoise Salager-Meyer, Philip Shaw, Irena Vassileva and Eija Vento-
la. Mauranen’s and Ventola’s studies of written academic discourse (e.g. Mauranen
1993b; Ventola 1994) have included comparisons of English and a lesser used lan-
guage, Finnish, with particular focus on issues relating to rhetorical differences
between English and Finnish academic discourse within economics. In contrast
to our work, they have often looked at problems arising from writing in English
as a second language, i.e. how the Finns fare when communicating their research
findings to an international audience. Salager-Meyer is particularly interested in
practices within medical discourse (e.g. 1998, 1999). She often deals with several
languages, typically contrasting English, French and Spanish (e.g. Salager-Meyer,
Alcaraz Ariza & Zambrano 2003). Unlike the present synchronic study of aca-
demic discourse, Salager-Meyer often adopts a diachronic approach (e.g. Salager-
Meyer & Defives 1998; Salager-Meyer 1999). Shaw has undertaken several studies
involving rhetorical aspects of English and Danish academic discourse within the
realm of economics (e.g. 2003; Shaw & Vassileva forthcoming).
Finally in this part of the section, we would like to mention the work under-
taken by Vassileva (e.g. 1997, 2000, 2001). Her comprehensive study of authorial
presence in academic writing (2000) includes data on as many as four languages in
addition to English, viz. German, French, Russian and Bulgarian. To some extent
we may say that the Bulgarian perspective is at the heart of the study, as the other
four languages included have ”to varying degrees and at different times in history,
influenced and shaped the Bulgarian canon of academic writing” (2000: 12). In
addition to studying author presence through the use of first person pronouns in
single-author research articles in linguistics in the five languages, Vassileva also
 Chapter 1. Introduction 27

addresses the non-native aspect by including a corpus of linguistics articles writ-

ten by Bulgarians in the other four languages.
As regards relevant studies taking French academic discourse as a point of
departure, it becomes particularly interesting to look to the research environments
of L’Equipe Sémantique des Textes in Paris, Univérsité d’Orléans and Université
Stendhal in Grenoble. Only a few names will be mentioned here − François Rastier
(Paris), Céline Poudat (Orléans), Françis Grossmann, Cristelle Cavalla and Fanny
Rinck (all Grenoble) − as they link up with KIAP through studies which are very
similar to ours; some even make use of our corpus. Rastier (e.g. Malrieu & Rastier
2001) has undertaken a multifactor analysis of genres, including the research ar-
ticle. As for Poudat, she deals in her doctoral thesis, among other things, with
author presence as manifested through first person pronouns in linguistics articles
in French and English (2003, 2004, forthcoming). Grossmann has undertaken ci-
tation studies (e.g. Grossmann 2002), while Cavalla and Grossmann have studied
academic vocabulary (lexemes such as hypothèse, thèse and théorie) in French re-
search articles within various disciplines including the ones of the present study
(e.g. Cavalla & Grossmann 2005). Rinck in her doctoral thesis (2005, forthcoming)
deals with the research article as genre, investigating linguistic and socio-discour-
sal aspects of the writing process. She has also compared what she calls ”images of
scientific activity” in linguistics and literary articles (2005).
When it comes to the third, and least used, language represented in our study,
we find that studies of Norwegian academic discourse are thin on the ground.
However, a few important studies and projects dealing with how Norwegian pro-
fessionals communicate do exist, in addition to a more comprehensive literature
on academic writing for pedagogical purposes (e.g. Hertzberg 1995; Dysthe 2002).
A very important study for the KIAP project has been Kjersti Breivega’s (2003)
account of argumentation strategies in Norwegian scientific prose. Her study is a
comparative analysis of research articles from the disciplines of medicine, history
and linguistics. Another important source of inspiration has been the research
unit Norsk sakprosa (”Norwegian factual prose”), located at the University of
Oslo, with Kjell Lars Berge (e.g. 2003) and Johan L. Tønnesson (e.g. 2001, 2004)
as key professors. Due to the closeness and even mutual intelligibility of the three
Scandinavian languages, it might be argued that Swedish (including Finns with
Swedish as their native language), Danish and Norwegian researchers within the
various disciplines together constitute a common linguistic discourse community.
This can be illustrated by the fact that scholars from the three countries to some
extent publish in each other’s national journals in their own language (e.g. one
of our corpus sources, Norsk Lingvistisk Tidsskrift). Each of the other three Nor-
dic countries has one or two research environments in particular that have been
very active within the field of academic discourse. In Denmark, Lita Lundquist
28 Academic Voices

at the Copenhagen Business School (cf. e.g. Lundquist 2003) and Jan Engberg at
the Aarhus School of Business (e.g. Engberg 1997) have undertaken important
research into, inter alia, legal language. In Sweden, the FUMS research group at
Uppsala University headed by Britt-Louise Gunnarsson (e.g. 1994), has for several
decades published on Swedish academic discourse, in a diachronic as well as a
synchronic perspective. In Finland, at the University of Vaasa, Christer Laurén
and Marianne Nordman (e.g. Laurén & Nordman 1998), as well as Merja Koskela
(e.g. 1999, 2000) have for many years been describing various aspects of Swedish
for use in academic contexts.
This very incomplete survey of academic discourse research which has a clear
bearing on our own study will end with a few reflections on the direction the field
currently seems to be moving in. Studies of the kind mentioned above have often
been placed under the umbrella of contrastive rhetoric. This concept, dating back
to Robert Kaplan’s 1966 article ”Cultural thought patterns in intercultural educa-
tion”, has primarily been applied to work with a pedagogical aim, but has also been
used for studies of communication between experts. Contrastive rhetoric has, for
various reasons, lost ground in recent years, even though the idea that culture may
and should be used to explain differences in written texts and writing practices
(Atkinson 2004) is very much alive. One of the main problems of contrastive rhet-
oric has been the multidisciplinary nature of the field, something which represents
a daunting methodological challenge. This is a problem which has been voiced
by several researchers in the past couple of years (e.g. Melander 1998; Mauranen
2001a). In a special issue of Journal of English for Academic Purposes (Contrastive
rhetoric in EAP, vol. 4 (3)), the editor Ulla Connor in the introduction to the vol-
ume advocates the need for broadening the scope of contrastive rhetoric and the
need for a new name (2004a: 272). She proposes the term intercultural rhetoric,
intended to refer to a concept which is more dynamic than contrastive rhetoric
and intended to comprise oral as well as written texts (cf. also Connor 2004b). The
increased focus on oral texts witnessed in the past couple of years (e.g Carter-Tho-
mas & Rowley-Jolivet 2001; Mauranen 2001b, forthcoming; Swales 2001; Rowley-
Jolivet 2002) represents a clear trend within the field.
The widening of the scope of studies involving academic discourse will, how-
ever, not solve the methodological problems currently experienced. On the contrary,
we must expect the debate on how to accommodate all the possible variables in such
studies into a unified framework to become even more intense in the years ahead.
 chapter 2

Theoretical and methodological frameworks

2.1 Theoretical framework

2.1.1 Main theoretical approaches

As outlined in Chapter 1, our main object of study is person manifestation as re-

alised through voices and roles in the genre of the research article. We study lin-
guistic features constituting traces of the author(s) (the self-dimension) and the
roles they take on, the reader(s) and various third person constellations of other
researchers or discipline communities (the other-dimension), as well as the inter-
relation between these dimensions (self & other). This choice of study object has
influenced our choice of theoretical framework. Our main approach may be situ-
ated within the French enunciative approach, a tradition which takes as its point of
departure the utterance situation. The enunciative approach is based on the view
that an utterance necessarily contains traces of the act producing it, traces relating
to the context in which the utterance is situated (time and space) and, particularly
important for our study, traces left by the subject producing it.
These traces can be studied within the research article considered as a poly-
phonic “drama”, in the sense that different voices and roles are dramatised by the
author, interacting with readers and other researchers (see Ducrot 1984; Nølke
1994; Nølke, Fløttum & Norén 2004; Fløttum 2005b). In Bakhtinian terms, this
drama corresponds to a dialogical conception of discourse. Even if the text is for-
mally monological (like the research article), it may be dialogical in that the au-
thor gives the floor, explicitly or implicitly, to other voices.
Before going further into the theory of linguistic polyphony, we will present
aspects of the broader theoretical framework within which our studies are under-
taken. We profit from a number of theoretical perspectives, bringing together in-
sights from both Anglo-American, Anglo-Nordic (if we may use such a label) and
French traditions. The theoretical perspectives provide us with an understanding
of the linguistic means which indicate the presence of and the interrelation be-
tween persons or groups of persons involved in the research process reported.
We take a special interest in these persons or individuals and the argumentation
they are engaged in. In Halliday’s terms (Halliday 1994: 179), we are particularly
interested in the interpersonal metafunction, realised by structures expressing in-
30 Academic Voices

teractional meaning. We study what the utterance is doing as a verbal exchange

between author and audience. This means that we are more interested in the in-
dividuals behind the reported research than the research itself. However, even if
we do not focus on the representational utterance meaning, what the utterance is
about, we are not leaving out the ideational metafunction. For example, we take
it into account when studying the combination of first person pronoun and verb,
since this typically reflects a process associated with a particular participant. Fi-
nally, we also include Halliday’s textual metafunction in different contexts, for ex-
ample when studying metatextual expressions referring to the textual organisation
of the message.
Our focus on academic voices, how the author and other participants are
manifested in the text, has determined the orientation of our investigations, in
the sense that we undertake analyses in a mainly ascending way, going from the
micro-level of linguistic phenomena to a macro-level including a larger context.
Since one of our purposes is to contribute to a linguistic description of a particular
text genre, viz. the research article, the socio-professional context in which the
texts are produced is also taken into consideration.
In our view, one of the main features of the research article is that it is
rhetorical. It is rhetorical in the sense that it represents a discourse created in or-
der to induce cooperative attitudes and actions (see Prelli 1989), on the one hand,
and to give the possibility for the authors to position themselves in or in relation
to a particular discourse community, on the other. This implies the evident fact
that the research article is addressed, i.e. it is directed at someone.1 Even if the
manifestation of addressivity may be different in different traditions, this observa-
tion points to the simple but important fact that the research article constitutes a
piece of communication. The final rhetorical aim of a research article is to create
effects which convince the audience to such a degree that the article becomes an
integrated part of a particular field’s literature.
In his study, Prelli (1989) aims at describing the rhetorical dimension in the
way scientific discourse is both created and evaluated. Creation (or production)
and evaluation (or effect) are, of course, closely linked. However, in this book we
focus on the creation side. This is what we can say something about through the
study of the texts. Even if, in a few cases, we have asked the authors why they have
chosen a specific linguistic expression, we do not investigate the reception of the
texts. The reception of the texts is, of course, a sociologically interesting question,
but one which goes beyond the aims of our linguistic and genre-related study.

1. For a more comprehensive discussion of rhetorical discourse and of the notion of addres-
sivity in academic discourse, see Kjeldsen (2004) and Tønnesson (2003a, b) respectively.
 Chapter 2. Theoretical and methodological frameworks 31

If it is accepted that rhetoric and science or research are not notions in con-
flict, and that scientific discourse is rhetorical, one should look further into what
this rhetoric consists of. In this book, we will study a selection of linguistic features
contributing to the rhetorical reasoning in research articles.
We will now take a closer look at the particular text genre of the research
article. The genre context in which the text is produced is, in fact, crucial to our
explanations of the linguistic observations. As stated in 1.3, the text genre is con-
sidered a context in itself. The research article has evolved in different socio-profes-
sional communities with their own norms and rules (e.g. Bazerman 1988; Swales
1990; Berkenkotter & Huckin 1995; Berge 2003; Breivega 2003; for more general
perspectives on the notion of genre, see Berge 1990; Freedman & Medway 1994;
Adam 1999, 2005; Ledin 1999; Berge & Ledin 2001; Rastier 2001). Our study is
clearly more linguistically and less sociologically oriented than the North-Ameri-
can tradition of the New Rhetoric. However, we have also been inspired by Mill-
er’s observations (1994) in that we tend to consider the genre as a verbal practice
realised by texts. According to Miller, this practice has for historical or cultural
reasons acquired a name, which gradually has become conventional in the sense
that language users have a sufficiently clear conception of it to be able to interpret
it and possibly to produce it.
In this book, we opt for a simple definition of genre: We consider genre to rep-
resent a recurrent social practice linguistically realised through texts, according to
more or less fixed patterns. This means that the texts under study normally contain
particular linguistic and textual traits; however, these traits do not constitute abso-
lute patterns. Thus, our notion of genre corresponds to the well-known Bakthinian
conception of genre as a discourse object, being at the same time dynamic (a genre
can change over time and even disappear) and relatively stable (a genre is charac-
terised by a set of “rules” of variable nature, from linguistic to cultural) (Bakhtin
1986). This might imply that differences between research articles written within
different scientific disciplines motivate a distinction between, for example, IM-
RAD-structured articles (within medicine) and non-IMRAD-structured articles
(within economics and linguistics) as different sub-genres of the research article.
As already stated, the genre perspective is an important explanatory factor in
our study. We place ourselves in a tradition analysing linguistic phenomena in a
discoursal context (considering linguistics and text linguistics as integrated parts
of discourse analysis). The main contextual frame of our studies is the genre, and
the genre represents many interesting linguistic and text-linguistic dimensions.2 In

2. Dimensions such as coherence, composition, thematics, enunciation and illocution are all
relevant. For a modular model taking these into consideration, see Adam (1999).
32 Academic Voices

accordance with our research questions (see section 1.1), we limit ourselves main-
ly to the enunciative, polyphonic and semantic-pragmatic perspectives.
Among the large number of studies undertaken within academic discourse
during the last two decades, many different theoretical and methodological ap-
proaches are represented. One of these is John Swales’ genre theory presented in
his 1990 book, which has gained enormous success, documented through numer-
ous studies of academic genres, especially in English. We have chosen a differ-
ent approach as our main perspective, because of our focus on linguistic traces
of academic voices rather than article components and structure. In our view, the
enunciative and polyphonic approach accounts for most of the phenomena we are
studying. Thus it has the advantage of offering a comprehensive theoretical and
methodological approach. Further, taking as our point of departure the fact that
a research article is created in a particular multivoiced situation, we consider the
enunciative approach, with its focus on the utterance and the utterance situation,
as particularly pertinent. Language in use and language studied in relation to users
and context are essential dimensions in this approach.3
There is no one uniform and homogeneous enunciative theory. It is more ap-
propriate to speak about different variants of a common perspective. French enun-
ciative linguistics has developed over many years. It is quite common to point to
Charles Bally’s work as the origin and to Emile Benveniste as the one to elaborate the
enunciative perspective into a theory (see Benveniste 1966). Different orientations
have later been developed by for example André Culioli and Oswald Ducrot.4
Benveniste (1966) defines enunciation as ”la mise en fonctionnement de la
langue par un acte individuel d’utilisation”5 and thus attaches importance to enun-
ciation as an individual act (and a historical event) where the language as sys-
tem (langue) is put into function. The result of this act is the utterance (énoncé).6

3. A similar view is found in cognitive-functional linguistics, also emphasising the elabora-

tion of usage-based models for linguistic analysis.
4. For a more comprehensive presentation, see Marnette (2001). Most researchers working
within the enunciative tradition focus on language in use (an exception is Ducrot who has
always made it clear that his object of study is la langue). In this context we should point to
Roman Jakobson and Mikhaïl Bakhtin as important sources of inspiration. The fact that the
perspective is extended from the langue- to the parole- or discourse-level is probably one of the
reasons why many literary researchers have taken an interest in enunciative theory. There are
French traditions where the distinction between linguistics and literary studies is not considered
as important; representatives are Dominique Maingueneau, Jean-Michel Adam, François Ras-
tier and Alain Rabatel.
5. English translation: ‘putting langue into function by an individual act of use’.
6. The utterance (l’énoncé) is considered as a realisation of the abstract sentence or clause (phra-
se or proposition); the utterance can also be a realisation of a syntactically incomplete sentence.
 Chapter 2. Theoretical and methodological frameworks 33

This definition is not unproblematic since enunciation might include all factors
integrated in the production of an individual utterance – from physical, articula-
tory to context-related and situational, as well as psychological motivation factors
(see Maingueneau 1993). On this basis it might be argued that enunciation is not
of interest to the linguist – enunciation has to do with individual acts, and not
the abstract language system. However, we do not accept such a delimitation of
linguistic studies. Enunciation can be studied through the observable traces left
in the utterance. These specific traces are usually instances of general patterns and
can thus be related to the language system (for example personal pronouns, verb
tenses and adverbs). As Marnette says:
There is a crucial distinction to be made between each individual enunciation
(seen as a single historical event) and the general phenomenon of enunciation,
namely a stable system which emerges from the multiplicity of all the individual
acts of enunciation. To study enunciation is, thus, to study a set of specific mecha-
nisms through which the locutor converts the abstract system of langue into dis-
cours. These mechanisms can be studied via the traces they leave in their products,
the utterances. (Marnette 2001: 244)

Enunciation has three main dimensions: personal, temporal and spatial. In the
personal dimension, the two protagonists of the enunciation have been given dif-
ferent names according to different orientations, for example locuteur, énoncia-
teur vs. allocutaire, co-énonciateur. In the following, we will use both the anglified
terms locutor and allocutor as well as the more traditional sender/speaker and re-
ceiver/hearer.
For a simple illustration of how the three dimensions can be manifested, we
can look at the following constructed sentence: I am here now. This sentence is
full of traces of the utterance situation, viz. deictic expressions. The pronoun I
refers to the utterance locutor (personal dimension), the adverb here to the place
where the utterance is produced (spatial dimension), and the adverb now to the
moment in which the utterance is produced (temporal dimension, also expressed
by the simple present tense). Deictic expressions are, of course, not the only rel-
evant expressions in this context; a series of other linguistic expressions constitute
enunciative traces. As regards the personal dimension, it is important to note that
the allocutor is also a part of the enunciation. An utterance is usually part of a dis-
course or interaction, and the locutor tends to position him- or herself in relation
to the allocutor, in one way or another.
The locutor can position him- or herself or manifest an attitude regarding dif-
ferent dimensions: the utterance content, the enunciation itself, the allocutor, the
world outside the particular situation, previous or successive utterances. In this
way, traces of the author can be manifested through numerous linguistic phenom-
34 Academic Voices

ena, like speech acts (especially performatives), different kinds of modality, con-
nectives, verb tense and aspect, evaluative expressions, etc. An important reason
for choosing the enunciative approch is the “personal” orientation we have given to
our research questions (see 1.1).
In order to answer our first research question, about the manifestation of au-
thors (self-perspective), we study first person (singular and plural) and indefinite
pronouns, pertaining to the personal dimension, and the use of metatextual ex-
pressions, which may be related to the spatial dimension. In our study of pro-
nominal use, we take into account the tense of the finite verb combined with the
pronoun, i.e. the temporal dimension. As regards the determination of the role
that the author may take on in the text, we also study the semantic-pragmatic
meaning of verb constructions combined with first person pronouns. This part of
our analysis necessitates a semantic-pragmatic approach in addition to the enun-
ciative one.
As regards the second question, about the manifestation of others’ voices (the
other-perspective), we study bibliographical references, different forms of report-
ed speech and other polyphonic constructions. In this context we add the third
person, which we also consider part of the personal dimension.7 The third per-
son can be referred to explicitly, as in Johnsen (1998) claims that …, or implicitly
through polyphonic expressions, as in a construction with the connective however:
The water in Bergen tastes good; however, it is polluted. The connective however
indicates that the locutor considers the point of view reported in the proposition
introduced by however as his own and as the most important one. The source of
the point of view reported in the proposition preceding however (only conceded
by the locutor) may be a third person, identifiable by the context.
For the third question, how the authors present and promote their own re-
search, the personal dimension is also relevant. Some of the most relevant fea-
tures in this context are epistemic qualifications (expressing the author’s attitude
towards the propositional content) such as may in The colour of the water may
indicate pollution.8 This third question is the one that most clearly requires a step
out of the enunciative perspective and into a lexical-semantic one. In this context,
word choice and evaluative qualifications constitute an important dimension.
As indicated above, the research article may be considered as a polyphonic
drama where the author interacts with different parts; however, the author always

7. The personal dimension of the enunciation is commonly related to first and second per-
sons; these are the protagonists of the utterance situation.
8. Epistemic qualifications will not be discussed in this book, but are thoroughly studied in
the doctoral work of KIAP member Eva Thue Vold (see the KIAP publication list at the end of
the book.)
 Chapter 2. Theoretical and methodological frameworks 35

has a double function. He or she presents own points of view at the same time
as setting up and controlling the whole drama. This hierarchical conception of
the author (or the sender) corresponds to the main conception of the ScaPoLine
theory of linguistic polyphony, which constitutes an important elaboration of the
enunciative perspective which we will apply in this book. ScaPoLine is short for
”théorie SCAndinave de la POlyphonie LINguistiquE” (Scandinavian theory of
linguistic polyphony), see Nølke et al. (2004). The ScaPoLine theory refutes the
traditional postulate of the unique and indivisible speaking subject and claims the
possibility of the superposition of several voices or points of view in one and the
same utterance. In the next section, we will give a more detailed presentation of
this theory.

2.1.2 A polyphonic framework: ScaPoLine

The ScaPoLine theory is inspired by Bakhtin’s view of dialogism as fundamentally

constitutive of language use, i.e. there is a multitude of interacting voices influenc-
ing the speaker’s language use. The meaning of an isolated utterance, as well as of
a text, is not only the expression of a speaking subject’s idea, but rather a scene or
drama of interrelated voices or points of view attributed to more or less abstract
instances in addition to the speaker or writer (see Perrin 2004a: 7–12 for a more
detailed introduction). According to Bakhtin, the utterance meaning undergoes a
double influence, which might be called interdiscoursal and interlocutive, respec-
tively (see Bres 1999). On the one hand, the utterances enter into interdiscoursal
(or intertextual) resonance with what has already been said (by echoing and/or
reacting to other previous utterances). On the other hand, the utterances also an-
ticipate the reactions of a real or potential interlocutor:
Se constituant dans l’atmosphère du « déjà dit », le discours est déterminé en
même temps par la réplique non encore dite, mais sollicitée et déjà prévue. (Bak-
thine 1978: 103)
‘By constituting itself in the atmosphere of the “already said”, the discourse is si-
multaneously influenced by the reply, not yet said but solicited and already an-
ticipated.’

While we are clearly inspired by this dialogical view of language use, we do not use
the concept of polyphony in the same way as Bakthin in his studies of Dostoyevsky
(see Bakthine 1970). One of the main differences between the linguistic ScaPo-
Line theory and the Bakthinian concept is that the former considers the relation
between the speaker’s voice and the others’ voices as hierarchical (see above): the
speaker has the dominant voice. In the Bakthinian polyphonic conception, on the
other hand, the different voices are independent: there is a relation of equality
36 Academic Voices

between the speaker (or narrator) and the other voices. However, the contestation
of the well-established idea of the uniqueness of the speaking subject is common.
In what follows, we provide a very simplified introduction to the basic ideas of the
ScaPoLine theory (see Nølke et al. 2004).
With a polyphonic conception of meaning, it is essential to demonstrate how
the presence of several voices is signalled in discourse. In this linguistic version of
the theory, the object of study is what is expressed by the utterance out of context,
and the polyphonic structure is to be found at langue (or sentence) level. It is a
structure which is uncovered by an investigation of signals given in the utterance.
At the same time, the polyphonic structure gives us instructions as regards pos-
sible interpretations of the utterance in relation to its discoursal context. The aim
of the ScaPoLine theory is to explain linguistic polyphony and thereby anticipate
the influence of such phenomena on text interpretation (see Nølke et al. 2004: 15).
Thus, this theory distinguishes itself, for example, from Oswald Ducrot’s strictly
linguistic objective (Ducrot 1984) in its expressed aim to explain utterance mean-
ing, taking as its point of departure hypotheses concerning the semantic instruc-
tions embedded in linguistic expressions.
The relevance of the polyphonic perspective in the present context is that the
author may set up a polyphonic play or drama signalling the presence of both his
or her own voice and the voices of others. Different voices are given the floor, ex-
plicitly (for example by citation) or implicitly, by some distinctive mark signalling
polyphony. This is a play which the author creates in his or her own way and which
represents a subtle means of interaction, where the source of the different voices
or points of view is not necessarily explicit. Let us consider the classic example
(translated from French “Ce mur n’est pas blanc”; see Ducrot 1984):
(1) This wall is not white.

Two points of view are presented in this utterance, one saying that ‘this wall is white’
(point of view 1) and another qualifying this as ‘not valid’ – or as ‘false’ (point of
view 2). The isolated utterance does not indicate who is the source of the first point
of view; the source may or may not be identified by the context. For the interpreta-
tion of the utterance it is important to try to determine the different points of view
which are manifested and to identify the (group of) individual(s) corresponding
to the source of these points of view (the author, another researcher, the scientific
community, the doxa, etc.), in short, to identify who is responsible for them.
We will now look at an authentic example, a complex multivoiced sequence
taken from an English linguistics article (emphasis added):
(2) It is generally recognized that what Banfield (1982) calls represented speech
and thought may include expressions and constructions which, although they
 Chapter 2. Theoretical and methodological frameworks 37

can be attributed to the person whose thought is being represented, cannot be

easily analysed as contributing to something propositional. […] It is true that
the expressions and constructions which according to Banfield, characterize
free indirect speech cannot be indirectly quoted in embedded clauses as but
can. However, the phenomenon of free indirect speech or thought does raise
the question of what it means to say that a writer or speaker is representing a
thought. If we say […], then it seems we cannot construe thoughts simply in
terms of their (truth conditional) propositional content. (engling05)

This sequence is not only an open verbal interaction where the author explicitly
brings the voice of Banfield and a general (?) opinion (cf. It is generally recognized
…) into the discussion. There are other voices present, implicitly, signalled by dif-
ferent polyphonic markers. There are several examples of refutation indicated by
the negation not (in cannot). The interesting questions in this context are: To what
extent is there refutation and if refutation it is, then who is responsible for the
underlying positive point of view? Is it another researcher, a specific discourse
community, or a more or less vaguely defined doxa? In this example there is also a
concessive connective, however, which signals that, for the author in this context,
the preceding proposition is not the most important one: The one that follows
is. Who, then, is responsible for the content transmitted by the first proposition,
the author at another time, or an external voice? And how should the expression
according to Banfield be interpreted? It does not necessarily mean that the author
takes the same view as Banfield, even if she refers to her in this way. Finally, there
are two occurrences of the plural pronoun we (engling05 is a single-author arti-
cle). Who does this pronoun refer to? It seems reasonable to suggest that at least
the author and the reader are included; however, is the reader ready to accept this
inclusion, or does the author include the reader in a point of view the reader might
disagree with?
There are many questions to ask when faced with polyphonic constructions,
be they explicit or implicit. The ScaPoLine theory puts forward at least three types
of questions: How many points of view are present? What is the relation between
these and the locutor? Who is responsible for the different points of view? These
are just some of the questions that should be asked if one wants to reach a better
understanding of the polyphonic play typically taking place in research articles.
In order to analyse polyphonic markers in a more rigorous linguistic way, we will
now present in a simplified way some of the basic aspects and terminology of the
ScaPoLine theory (see Nølke et al. 2004).
As stated above, the polyphonic structure is identified at the langue level. This
abstract structure is given a concrete expression at the level of the polyphonic con-
figuration, the parole level, which is observable (the level at which the discussion of
38 Academic Voices

example (2) above takes place) and which is of particular interest here. Through its
instructions the polyphonic structure imposes constraints on the interpretation of
the configuration.
The configuration consists of four entities:
1. the locutor-constructor – LOC – responsible for the enunciation and its result
(i.e. the utterance), and three other elements, all constructed by LOC:
2. the point of view(pov), which is a semantic entity, related to a source
3. the discoursal beings, which are semantic entities that constitute the sources
4. the enunciative relations which relate the pov to the discoursal beings. There are
two main enunciative relations – one of responsibility and one of non-responsibility.
The responsibility relation is by far the most important. In the negation example
(1) above, the locutor is responsible for the negative point of view (pov2). There
are several subtypes of relations: for example, semantic-pragmatic relations (argu-
mentative, counter-argumentative, reformulative) and logical-semantic (epistemic,
refutative) (see Fløttum 2001b). If we go back to the negation example (1) again,
the locutor’s relation to the positive pov1 is one of non-responsibility, more precisely
a refutative relation. The formal representation of the polyphonic structure of the
negation goes as follows, where X represents the unknown source of pov1:
(1’) This wall is not white.
pov1: [X] TRUE (‘this wall is white’)
pov2: [l0] UNJUSTIFIED (pov1)
A further explanation of the LOC-entity is required. LOC is an abstract entity that
can create two images of itself: the utterance locutor, l0, i.e. the image correspond-
ing to the moment of enunciating or uttering; and the textual locutor, L, corre-
sponding to a general image of LOC, with all the properties of a complete being
except for the one responsible for the utterance. L can also correspond to an image
of LOC at another moment in its history. In the example
(3) I ask myself if ...

both l0 and L are explicitly present, in the pronouns I and myself respectively.
The main idea is that a LOC can construct different polyphonic plays which are
signalled by different linguistic expressions and in which other abstract individu-
als than LOC’s own images enter the scene. The individuals are the allocutor (in
French: allocutaire, or the receiver) and the third (in French: les tiers). The abstract
third, corresponding to a person or to a more or less defined group of persons, is
manifested by third person pronouns, indefinite pronouns, proper names or other
nominal phrases. There is a main distinction which has to be made between dif-
 Chapter 2. Theoretical and methodological frameworks 39

ferent thirds: the individual third and the collective third. The first is represented
by Tom in the following:
(4) Tom said that the weather had been fine.

In constructions with reported or represented speech, as in (4), we call the locutor

that is mentioned the represented locutor.
We are not going further into the problem of distinguishing different collec-
tive types here (see Nølke et al. 2004), but limit the presentation to an example of
what we may call the heterogeneous third, or the polyphonic one, which is a col-
lective whose members may be distinguished as individuals with the possibility to
take the floor, like many in the following:
(5) Many thought the game was over.

It should be noted that the different discoursal beings are not necessarily repre-
sented explicitly. This is especially true for the utterance locutor, who is typically
represented only in performative constructions.
Even though in this book we have chosen a rather non-technical application
of ScaPoLine, this perspective contributes to our understanding of how academic
discourse works. It helps us detect the implicit interaction interwoven in the ex-
plicit one, both of which contribute to the negotiation of socio-professional rela-
tions, represented by different voices, that is displayed in this kind of discourse.
As regards the study of the main features investigated in this book, the poly-
phonic perspective constitutes a theoretical refinement of the more general enuncia-
tive perspective. Some phenomena are more relevant than others in this perspective.
We return to the relevance of the polyphonic dimension in Chapters 4–6, treating
author presence, reader/writer interaction and other-presence respectively.
Before going on to the empirical part of this book, we will consider some
methodological questions related to our own study as well as to corpus-based lin-
guistic studies in general.

2.2 Quantitative methods

2.2.1 Our independent variables

The KIAP Corpus is a stratified collection of texts. The texts have been classified in
several ways, most importantly according to language, discipline and authorship.
Language and discipline are our two basic independent variables, with three pos-
sible values each. We compare samples (collections of texts) that have different val-
ues for either the language or the discipline variable. For instance, we ask whether
40 Academic Voices

there is a difference between articles from different disciplines in how often a cer-
tain kind of metatext is used, and whether there is a difference between articles
from different languages in how frequently first person grammatical subjects are
used. As regards authorship, there are three possible values: several authors, one
female author and one male author; the last two can be joined into one, so that a
binary variable is achieved. Questions about differences between disciplines or
languages are more central in our investigations than questions about differences
relating to authorship, i.e. between multi- and single-author articles or between
articles written by one man or by one woman. However, authorship is sometimes
criterial for the inclusion of a text in a sample. For example, multi-author articles
are of course not included in our study of the use of first person singular subjects,
since ‘I’ subjects do not occur in such texts. We have also performed secondary
studies where the gender of a single author is an independent variable (in order to
check whether female and male authors write differently, cf. 4.1, 5.1 and 6.1), and
one study where the number of authors is an independent variable (in order to see
if there is any difference in the use of first person plural subjects (‘we’), cf. 4.1).

2.2.2 Article length and absolute frequency

Underlying our quantitative analyses are data about dependent variables of two
types: information about the length of the individual texts and information about
the absolute frequency of selected linguistic features in those texts (i.e. number of
occurrences). Article length is measured by the number of text words. In order to
attain the highest possible level of comparability between texts, we have used only
the bodies of the articles as described in 1.2.1, and measurements regarding article
length employed in our analyses relate to the article bodies.
Frequencies have been calculated on the basis of automated corpus searches
and manual classification of the search hits. The amount of noise (i.e. the share
of search hits that are not examples of the features searched for and therefore not
included in the statistics) varies greatly among the features. For example, as part
of our investigation of pronoun use in Chapter 4, the French articles have been
searched for tokens of nous (‘we/us’). In this investigation, we are only interested
in nous used as a grammatical subject. However, since the corpus has not been an-
alysed syntactically, the search returns instances of nous as a subject, object, com-
plement of a preposition etc. Hence, to calculate the frequency of nous subjects,
we first had to categorise all the tokens of nous as subjects or non-subjects. Similar
procedures have been undertaken for all the linguistic features that we analyse.
When we use frequencies as the basis of our analyses, a simple assumption
is made: The prevalence of a feature is appropriately measured by counting how
many times it appears in the text. For example, a first person singular subject has
 Chapter 2. Theoretical and methodological frameworks 41

the same value regardless of its context, such as where in the text it is used and
what kind of verb it is used with. On the word level, this might be unproblematic.
On the text level, however, it is not. Factors such as text location and verb type are
obviously important aspects of self-representation. It is important to realise that
frequency-based studies need to be complemented by qualitative analyses.
Not all the features that we study are formally expressed as single words. For
instance, even though we search for single words, e.g. article, the actual metatex-
tual expression may be of the type in this article. Regardless of their length, such
expressions are counted as one instance of metatext.
Problems related to the comparison of linguistic features from different lan-
guages are discussed in 2.2.4.

2.2.3 Relative frequency and measures based on it

If all the articles in our corpus had been the same length, absolute frequency would
have been an appropriate measure for our studies. However, the number of words
varies between 796 for nomed26 (our shortest text) and 19,882 for engling30 (our
longest text), that is, a range of 19,086 words. The mean length is 5001.9 words,
while the median length is 4643.5 words. The frequencies therefore need to be
normalised (cf. Biber, Conrad & Reppen 1998: 263–264) for the statistical tests
that we employ. Relative frequencies are calculated by dividing the absolute fre-
quencies by the total number of body words. Since the relative frequencies are in
general very small numbers, they can be more conveniently presented as percent-
ages or per thousand words. We have opted for the former; the choice between
them has no theoretical significance.
The measure of relative frequency makes it possible to find the median and
mean relative frequencies for collections of articles, i.e. for our various subcor-
pora. Further, it is possible to rank the articles for their prevalence of a certain
linguistic feature. Article rankings are the basis for the various statistical tests that
we perform in order to test hypotheses about the populations of texts from which
our subcorpora are sampled. The statistical tests are the subject of 2.2.5.

2.2.4 Cross-linguistic comparisons and relative frequency

In our studies, we compare research articles written in three different languages.

The things we are interested in include such features as first person pronouns,
metatext, bibliographical references etc. As discussed in 2.2.3, we use relative fre-
quency to measure the prevalence of these features, and the calculation of this
measurement involves the use of the total number of words in the texts as well as
absolute frequencies of features. Neither of these arithmetical elements is unprob-
42 Academic Voices

lematic when comparing texts in different languages. Consequently, a fair amount

of caution needs to be exercised in any study that is based on them, and we would
like to mention a few points that pertain to our studies.
First, the number of words that are used to express a certain meaning varies
between languages. As an example, English we are going to assume might corre-
spond to French nous supposerons – with five words in English and two in French.
And the English phrase the theory corresponds to the Norwegian noun teorien.
Systematic differences of these and other kinds affect the length of texts measured
in words and, consequently, any comparison based on relative frequency.
Second, the frequency of some of the linguistic features that we study is af-
fected by grammatical differences between the languages. One example of such
a difference is the existence of constructions like French gerundive en supposant
(‘in assuming’) and similar participial constructions in English, where the verb
has no expressed grammatical subject that would show up in the frequency count;
Norwegian does not have such constructions, and one might instead write når vi
antar (‘when we assume’), with an expressed subject. In our case, this affects the
relative frequencies of first person (‘I’, ‘we’) and indefinite pronoun (‘one’) subjects
in Chapter 4.
Third, differences relating more to language use than to the grammatical
systems cause similar problems of comparability. One example is differences in
sentence length, which may affect various linguistic features both directly and
indirectly. Another is stylistic or register differences, e.g. although the English
indefinite pronoun one means approximately the same as its Norwegian closest
equivalents man and en/ein, its use is much more restricted due to its stylistic
markedness. As shown in Chapter 4, there is a very clear difference between Eng-
lish and Norwegian in the relative frequency of indefinite pronoun subjects.
Fourth, many lexical items of one language may have no clear equivalents in
the other languages, or there is a complex relation between lexical items in one
language and the other. Such issues are important for example in our study of
metatext (see 5.2), where we have attempted to select equivalent sets of metatex-
tual expressions in the three languages.

2.2.5 Statistical tests

We now turn to a brief presentation and discussion of the statistical tests that we
have opted for. We have chosen to employ only non-parametric statistical meth-
ods; our motivations for this decision are discussed in section 2.2.6. Our main
investigations concern the effect of the factors of discipline and language on the
relative frequency of a set of linguistic features, viz. first person subjects (‘I’, ‘we’),
indefinite pronoun subjects (‘one’), metatext (e.g. ‘section’), negation (‘not’), adver-
 Chapter 2. Theoretical and methodological frameworks 43

satives (‘but’) and bibliographical references (whether of the type James (1990) (or
similar) or (in medical articles) a reference to an item in the bibliography, e.g. [7]).
For each of these features, we have carried out a battery of tests. The investigations
can be divided into two levels. The first is a higher level which comprises, at the
same time, articles from more than one discipline and in more than one language.
We will refer to this as the superlevel. The second is a lower level, where articles
from different disciplines in one language are compared, or, conversely, articles
in different languages from one discipline. We will refer to this as the sublevel.
The statistical tests are also described, along with their results, in Appendix B. For
further details about the methods, the reader is referred to the statistics literature
(see references below).
At the superlevel, we first perform a two-way analysis of variance (ANOVA),
using the Scheirer–Ray–Hare extension of the Kruskal–Wallis test (cf. Sokal &
Rohlf 1995: 445–447). The input of the test is observations of relative frequen-
cies for the whole corpus, i.e. for 3 × 3 × 50 = 450 articles. With this analysis, we
achieve three kinds of results. First, we find out whether there is any significant
effect of the factors of discipline and language – whether there is significant vari-
ance that can be attributed to interdisciplinary differences or differences between
languages. For instance, does it matter for the relative frequency of negation if an
article comes from economics, linguistics or medicine? If the answer is yes, we
know that at least one pair of disciplines exhibits a significant difference between
them, but not which pair(s) this is. We also find out whether there is a significant
effect of the interaction between discipline and language, i.e. whether the effect of
the discipline factor depends on the language factor, and vice versa. Second, the
test tells us which of the two factors has the strongest effect with respect to the
feature in question. For instance, which factor (discipline or language) has the
greatest effect on the relative frequency of negation? Third, we can tell how much
of the variance can be accounted for with reference to each of the factors and their
interaction – and, hence, also how much of the variance is due to factors which we
cannot account for.
If significant effects of a factor are found in the two-way analysis of variance,
we can proceed (still at the superlevel) to pairwise comparisons of disciplines and
of languages, using the Mann–Whitney–Wilcoxon test (cf. Butler 1985: 98–102;
Siegel & Castellan 1988: 128–137). The input of each test is observations for 2 ×
(3 × 50) = 300 articles. Since there are three discipline pairs and three language
pairs, the number of such comparisons will be six for each feature. The outcome
of these comparisons is a decision, for each feature with respect to each discipline
pair or language pair, as to whether there is a significant difference in the relative
frequency of the feature between that pair. For instance, are the relative frequencies
of negation in economics articles and in medical articles significantly different?
44 Academic Voices

At the sublevel, we first perform a one-way analysis of variance (ANOVA),

using the Kruskal–Wallis test (cf. Siegel & Castellan 1988: 206–216; Sokal & Rohlf
1995: 423–427). The input of each test is observations for 3 × 50 = 150 articles.
There will be two analyses for each feature: one for variance among disciplines and
one for variance among languages. With this analysis, we find out whether there
is significant variance among the disciplines (in each language) or among the lan-
guages (in each discipline). For instance, does it matter for the relative frequency
of negation in English if an article belongs to economics, linguistics or medicine?
If the answer is yes, we know that at least one pair of disciplines exhibits a signifi-
cant difference between them, but not which pair(s) this is.
If significant variance is found in the one-way analysis of variance, we can
proceed (also at the sublevel) to pairwise comparisons of disciplines (in each lan-
guage) and of languages (in each discipline), using the Mann–Whitney–Wilcoxon
test. The input of each test is observations for 2 × 50 = 100 articles. Since there are
three discipline pairs in each language and three language pairs in each discipline,
the number of such comparisons will be up to 2 × (3 × 3) = 18 for each feature.
The result of these comparisons is a decision, for each feature with respect to each
discipline pair or language pair, as to whether there is a significant difference in
the relative frequency of the feature between that pair. For instance, are the rela-
tive frequencies of negation in English economics articles and in English medical
articles significantly different?
In addition to tests where language and discipline are the independent vari-
ables and where whole language/discipline subcorpora are involved, we perform a
number of other supplementary tests. In Chapter 4, we break down the dependent
variable of first person subjects into singulars and plurals and divide the subcor-
pora into smaller subcorpora of single- and multi-author articles, and this distinc-
tion in authorship is in one case also used as an independent variable. Throughout
the study, we also test differences between articles written by one man and articles
written by one woman, i.e. use gender as an independent variable.

2.2.6 Why non-parametric statistics?

Our reliance on non-parametrics may need some justification. In general, paramet-

ric methods should be preferred to non-parametric ones whenever this is appropri-
ate, because they are typically more powerful and provide us with better possibili-
ties of correctly rejecting our null hypotheses of no difference between populations.
Unfortunately, the more powerful the test, the more assumptions need to be made
about the compared populations. Non-parametric tests are more robust than para-
metric ones against non-normal distributions and differences in variance.
 Chapter 2. Theoretical and methodological frameworks 45

Specifically, two assumptions on which parametric tests are generally based

are potentially problematic for studies based on our data: The populations that the
samples are drawn from are normally distributed, and they have the same vari-
ances with respect to the variables under investigation (cf. Siegel & Castellan 1988:
20). Neither the distribution nor the variance of our populations of research arti-
cles is known. To the extent that our samples are representative of the populations,
however, we have a certain idea what the populations are like, and we are led to
believe that with respect to most of our variables, the assumptions do not do jus-
tice to the actual properties of the populations.
This is the case, for instance, with the use of first person singular subjects (see
4.1). In the Norwegian economics articles with one author (N = 33), the mean
relative frequency is 0.11 %, the median is 0.04 %, and the variance is 0.026 %. The
corresponding numbers for French (N = 31) are 0.01 %, exactly 0 and 0.003 %. A
comparison of the mean and median relative frequencies for Norwegian indicates
a great difference: The mean is almost three times the size of the median. Thus, the
distribution is highly positively skewed (towards 0). This difference is smaller for
French, but only because almost all values in the sample are 0 – the distribution
here is even more extreme. And if we look at the variance, we see that it is more
than eight times as large for Norwegian as for French, which can hardly be said to
support an assumption that the populations have the same variance.
Admittedly, this example is extreme, but the point remains: In our view, the
assumptions of normal distribution and equal variances are not generally valid for
our studies. This speaks against using parametric methods. It is often observed
(e.g. Butler 1985: 84) that parametric tests such as the t-test are rather robust,
tolerating even quite large deviations from normality of distribution and equality
of variances. And when samples number at least 30, it is sometimes assumed that
parametric tests are generally applicable. For instance, Butler (1985: 93) has a deci-
sion schema where the z-test is chosen for large samples regardless of distribution
and variances. On the other hand, Oakes (1998: 14) states categorically about the
t-test that ”[i]f the distribution is skewed, […] a non-parametric test should be
employed […]”. Given this apparent lack of agreement among experts on statistics
for linguistics and our reasons not to make the assumptions about normal distri-
bution and equal variances that parametric tests presuppose, we opt for a conserv-
ative decision to restrict our quantitative methods to non-parametric ones. This
choice may have reduced our possibilities of rejecting null hypotheses of no dif-
ferences and, conversely, of corroborating alternative hypotheses that there are in-
deed differences between our populations. However, the non-parametric tests that
we employ do approach the parametric alternatives in power when the samples are
large, so we do not believe that our studies have suffered much from this decision,
even in the cases where parametric tests might have been more defendable.
46 Academic Voices

2.3 Qualitative methods

In the previous section, we presented the statistical methods that we have em-
ployed to analyse the large amount of data that we have produced using the KIAP
Corpus. These large-scale quantitative methods have been combined with qualita-
tive analyses and discussions, sometimes supplemented with more limited quanti-
tative studies, and case studies.
Several of our investigations include aspects which might be called explora-
tory, corpus-based contrastive linguistics (primarily in the linguistic subfield of
pragmatics (cf. Nølke 1991)). Contrastive linguistics is a large field where many
theoretical and methodological approaches co-exist (see e.g. Krzeszowski 1990),
and corpus-based contrastive studies (often with a translational perspective) are
becoming progressively more common (see e.g. Fabricius-Hansen 1998, 2004; Jo-
hansson 1998; other contributions in Johansson & Oksefjell 1998).
Contrastive-linguistic and corpus-linguistic theoretisation has not been
among the aims of our investigations. Rather, our efforts have been concentrated
around specific analyses, trying to structure some of the material that we have
extracted from our corpus and to generate new empirical findings. Needless to
say, we have not performed these analyses in a theoretical void, but employed the
theoretical framework described in 2.1, supplemented eclectically with ideas from
related approaches as needed.
In some of our investigations, we have been able to draw on previous research.
In other cases, we have developed new categories and conceptual systems in order
to deal with the phenomena in question. What kinds of categories and concepts
we have established has depended to a high degree on the specific aims of our
inquiries. These aims include some that relate specifically to questions about cross-
linguistic differences and similarities, i.e. contrastive-linguistic ones. Equally cen-
tral, however, are aims relating to cross-disciplinary comparison. Our aims also
encompass questions about the research article genre in general as well as matters
relating to individual variation.
Throughout this book, the study of individual examples in their paradigmatic,
textual and rhetorical contexts is our hermeneutic method for creating descrip-
tions (as well as some partial explanations) of the huge amounts of data made
available through our corpus. The nature of the analyses differs from section to
section, and we do not everywhere present detailed methodological descriptions.
In several cases, the work leading up to our final analyses has been processes of
trial and error, and the published analyses are merely instances of the famous ”tip
of the iceberg”. What we proffer in 2.3.1 is an illustration, viz. our methodological
approach that led to our analysis of ‘let us’-imperatives in section 5.3. This descrip-
tion gives a fairly good picture of how we have typically carried out our studies.
 Chapter 2. Theoretical and methodological frameworks 47

In addition, 2.3.2 contains a brief discussion of the purpose of including case

studies and our reasons for subjecting particular articles to closer scrutiny.

2.3.1 An illustration of method

In 5.3 we describe eight functional types of ‘let us’-imperatives. These have been
posited with a view to shedding light on the text-compositional and rhetorical
functions of such imperatives in research articles, including language-contrastive
issues as well as comparisons of disciplines. How did we proceed to arrive at these
eight types, how are they related to text composition and the rhetorical aims of
academic authors, and what has been gained by our positing them?
We started by searching for occurrences of the imperatives in the corpus using
the search program (see section 1.2.6). The contents of the resulting concordances
then had to be sorted manually in order to exclude irrelevant hits (notably, the
automated search in the French part of the corpus had to include all words ending
in -ons, but only a small fraction of these are ‘let us’-imperatives). Having com-
pleted the sorting, we were in possession of our material. (We could now state the
number of occurrences in each article and in each subcorpus and combinations of
them. We also performed statistical analyses on these data, but we will leave the
quantitative methods aside here.)
The next step was a verb-based classification of the occurrences. We began by
marking each occurrence with a label for the verb (or verbal construction) em-
ployed (in English and Norwegian: the infinitive following the pronoun meaning
‘us’). The occurrences in each language could then be sorted alphabetically based
on the verb, and we were able to ascertain which verbs occurred more frequently
and which ones more seldom, or even only once (as was the case for the majority of
verbs). So far, the procedure had been straightforward (albeit time-consuming).
We were now in a position to try to discern patterns of use in the material as a
whole. In this task, we could profit from the conceptual apparatus used elsewhere
in the book (notably, the system of author roles in 4.2) and in previous KIAP work
(e.g. Fløttum 2003g, 2004d; Kinn 2005c), which in turn was inspired by the works
of others (e.g. Swales et al. 1998; Hyland 2002b; see 5.3 for further references). We
tried to sort the employed verbs according to their meaning, separately but in the
same fashion for each language. The result was a kind of semantic map, networks
of verbs clustering and clusters overlapping and interconnecting in intricate ways.
There turned out to be obvious similarities across the languages, but also some
interesting differences; e.g. some clusters found in French were largely absent in
English and Norwegian.
On the basis of these maps, we attempted to establish usage types, with classes
of verbs (and, indirectly, the imperative occurrences containing them) constitut-
48 Academic Voices

ing what we perceived to be natural clusters, keeping the amount of overlap to a

minimum. The exact number of types and the labels that we chose to attach to them
were determined gradually, as we attempted to provide coherent and distinctive de-
scriptions of the types. The system of types became more complex than the system
of author roles employed in 4.2 and the system for Norwegian ‘let us/me’-impera-
tives in Kinn (2005c), but this complexity was considered to be motivated in the us-
age patterns that we discovered. (Some occurrences were left out because they did
not obviously pattern with others, i.e. the classification was not exhaustive.)
The analysis of the corpus-based material had so far taken a paradigmatic per-
spective, focusing on the range of verbs employed and their meanings. But we also
considered the data from a syntagmatic perspective, trying to generalise about
the text-compositional role of the imperatives, to find a common function in the
cotexts. In addition, we moved our attention to the contextual level and related
these analyses to the rhetorical aims of research articles, making explicit how the
imperatives are used in order to convince the reader.
The interpretation of the observed (cross-linguistic, cross-disciplinary and
individual) variation in the corpus is without doubt the most difficult part of the
analysis. Explanations can be sought in terms of the language systems, national
or linguistic writing cultures, disciplinary writing traditions, text types, objects of
study etc. In the case of ‘let us’-imperatives, we have been able to point to a few
factors that we think have contributed to the (near-) absence of them in medical
articles, and a factor related to the language systems is pointed out in a later case
study (5.5.2). But here, and elsewhere, we do not contend to be in possession of full
explanations of the data (or anything close to such a thing). We do, however, con-
sider the large number of new empirical findings to contribute in valuable ways to
the larger picture of academic discourse and varieties thereof.
Already on the basis of the raw figures for the material, it could be observed
that there were no occurrences of ‘let us’-imperatives in English and Norwegian
medicine. Thus, there were differences between, first, medicine and the other dis-
ciplines and second, between French and the other languages. During the subse-
quent analysis, we discovered several other differences and similarities between
languages and disciplines. Thus, the types that we established allowed us to gen-
eralise about relations between language- and discipline-specific tendencies, and
this strengthened our belief that our approach was on a promising track.
In this way, we feel that we have provided the research field with a deeper in-
sight into the use of ‘let us’-imperatives in research articles, in general and in many
specifics. This relates both to the presence of the author in text (the self-dimension),
which is the focus of Chapter 4, and to the relation between author and reader (in-
cluding metatext; the self- & other-dimension), which is the focus of Chapter 5. In-
 Chapter 2. Theoretical and methodological frameworks 49

corporated in this is a broadened knowledge of similarities and differences between

disciplines and between languages – and between individual texts.

2.3.2 Our use of case studies

In the main theoretical sections of the book, we concentrate on the textual phe-
nomena as found in the KIAP Corpus and its various subcorpora. With such an
approach, there is a danger that the individual texts are forgotten – or, at least, that
they appear to be forgotten. In order to make sure that neither is the case, and to
provide further illustrations of the various phenomena at work, we complement
the more theoretical sections with a number of case studies. In the case-study sec-
tions, we look in detail at one or a few articles. For instance, 4.6.2 is a study of
the explicit and implicit presence of ‘we’ in a Norwegian medical article. The case
studies also serve to draw lines between the phenomenon investigated and other
textual features that it works in tandem with. An example is the co-occurrence of
‘let us’-imperatives with various types of metatext.
The choice of themes and texts for the case studies is variously motivated. The
primary reason for including them is as just described, and we have tried to let
the different disciplines and languages have their reasonable share. The individual
articles have been chosen for various reasons, which we provide in the individual
studies. We should mention in this connection that the case studies on medical
texts have been included specifically to remedy the admittedly scarce presence of
medical examples in several of the theoretical sections. One reason for this scarcity
is that the majority of the linguistic features that we focus on are less frequent in the
medical texts than in texts from the other two disciplines. We hope that the medical
case studies not only compensate for this, but help to throw light on the reasons for
the relatively low frequency in medical articles of the features in question.
 chapter 3

Quantitative results

3.0 Introduction

This chapter is primarily a presentation of the outcome of our main quantitative

studies, i.e. statistical analyses of differences in the relative frequencies of a selec-
tion of six linguistic features. Documentation of the results of these statistical tests
is provided in Appendix B. Detailed data for the prevalence of the individual fea-
tures for the various subcorpora as well as some derived statistics are presented in
the three following chapters: 4.1 contains the data for first person subjects (with
the meanings ‘I’ and ‘we’) and for indefinite pronoun subjects (‘one’); 5.1 presents
data for metatext (e.g. ‘article’, ‘section’, ‘below’); and 6.1 deals with the details for
bibliographical references, adversative conjunctions (‘but’) and negation (‘not’).
Since the present chapter is a presentation of quantitative results, the discussions
are not found here, but in Chapters 4–7.
The remainder of this chapter is organised as follows. Section 3.1 serves as a
bridge between the discussion of cultural identities in section 1.3 and a quantita-
tive understanding of such a concept made available by statistical analyses. Sec-
tion 3.2 answers the question of whether it is discipline or language differences
that have the greatest effect on the relative frequency of the linguistic features that
we study. Sections 3.3 and 3.4 present comparisons of the disciplines and the lan-
guages, respectively, for each of the six features investigated.

3.1 Cultural identities as tendencies in linguistic practices

The main question of the KIAP project has been whether there are cultural identi-
ties in academic discourse, and, if that is the case, whether these are more strongly
tied to the discipline or to the language of the authors. Since our project is a linguis-
tic one, what we investigate is linguistic practices. A cultural identity as reflected in
linguistic practices amounts to linguistic usage that members of a culture tend to
adhere to and that members of certain other cultures (but not necessarily all other
cultures) adhere to only to a lower degree; i.e. we are looking for similarities within
52 Academic Voices

the group and differences between it and (some) other groups. All groups exhibit
internal variation, and groups overlap considerably in most cases.
As an example, take the relative frequency of negation with ‘not’ in French and
Norwegian medical articles. Norwegian articles have more negation than French
ones. This does not mean, of course, that every Norwegian article has more nega-
tion than every French article. What it means is that the median relative frequency
is higher in Norwegian than in French medicine – 0.57 % and 0.35 %, respectively.
The Mann–Whitney–Wilcoxon test shows the difference between the corpora to
be significant (at a significance level of 0.05). The two samples are far from homo-
geneous. In the French medical articles, the relative frequency of negation ranges
from 0.06 % to 0.85 %, while for Norwegian, it ranges from 0.12 % to 1.22 %. That
is, the range in French is about 0.79 %, and in Norwegian, about 1.10 %. The range
of the overlap between the two samples is about 0.73 %, i.e. most of the range for
both samples. In spite of the great variation in both samples and the overlapping
ranges, the samples are different in that the observations tend to cluster around
different medians.
The tendency to gather around a certain median is what we may call a similar-
ity within the group. This ”similarity-in-variation” is easier to see when the group is
compared with other groups, some of which differ from the former. We have inves-
tigated a limited number of linguistic features. They have been selected as presumed
indicators of differences that may be related to the discipline- or language-based
textual practices of the authors. They all have to do with the authors themselves
and their relation to their readers or other researchers. Needless to say, a broader
selection of features would have provided a more detailed picture of the textual
practices. Such investigations will hopefully be undertaken in future studies.
We discuss our findings for the individual features in later chapters, and those
chapters also contain a number of qualitative discussions. In the remainder of this
chapter, we present our main results in such a way that the various features can be
considered together. We choose not to present large amounts of statistical detail
here. Instead, we describe our findings informally and represent similarities and
differences among disciplines and among languages graphically. This will hope-
fully help the reader to get an overall impression that will make the rest of the book
easier to read and to use. Some descriptive statistical information is provided in
the following three chapters, while the results of statistical tests are presented in
Appendix B.
Recall that we use non-parametric tests that are based on ranking (see 2.2). In
these rankings, the article with the lowest relative frequency for a certain linguistic
feature is assigned the rank 1, and the higher the relative frequency, the higher the
rank of the article. The highest rank is 450 when the whole corpus is included, while
 Chapter 3. Quantitative results 53

in less comprehensive investigations it is 300, 150 or 100. As a measure that indi-

cates the general tendency of a subcorpus, we provide mean ranks for each of the
subcorpora involved. These can be compared to the expected mean ranks, which
are the numbers that are expected if there is no difference between the text popula-
tions that the subcorpora have been sampled from. If the total number of articles
involved is N, the expected mean rank is always N/2 + 0.5. That is, it is 225.5 when
all of the articles are included, and otherwise 150.5, 75.5 or 50.5. When the actual
mean rank is smaller, it means that articles in the subcorpus in question tend to
have lower relative frequencies of the relevant feature than what is the case for the
compared subcorpora considered together.
Before we present the results, we would like to emphasise one thing that is
obvious to those who are well-versed in statistics, but may not be equally obvious
to those who are not: The fact that there is a significant difference between two
subcorpora does not necessarily mean that the difference is great, nor that it is
important. What it means is that there is a risk smaller than 5 % that the observed
difference is incidental (since our α = 0.05), i.e. that the text samples misrepresent
the text populations that they are culled from (and are assumed to be representa-
tive of). The larger the samples, the smaller differences suffice in order for the dif-
ferences to be found statistically significant. We have found more frequent use of
adversative conjunctions in Norwegian economics than in French economics, and
the difference is significant. The mean relative frequency of Norwegian men (‘but’)
is 0.65 %, while in French it is 0.37 % for mais (‘but’) (see 6.1.3). From these figures
we see that the relative frequency is, on average, almost twice as high in Norwegian
as in French, and this can surely be said to be a great difference. How important
it is depends, of course, on our interpretation, which also needs to take into ac-
count reservations of the kind that we pointed to in 2.2.4. Above all, we need to
ask whether adversativity in French and adversativity in Norwegian are appropri-
ately measured by the relative frequencies of mais and men, respectively – there
are other words, phrases and constructions in both languages that serve similar
functions as these two adversative conjunctions. We will not discuss this further
here, but merely indicate that what kind of importance we attribute to findings of
significant differences is very much a matter of interpretation.

3.2 Is discipline or language the most important factor?

The question posed in the title of this section should be understood as follows:
Is it differences among disciplines or among languages that have the greatest ef-
fect on the relative frequency of the linguistic features that we study? (The final
restrictive relative clause of the question is crucial: We do not pretend to have the
54 Academic Voices

answer to the much more general question of whether discipline or language is the
most important factor determining features of language use.) In order to answer
the question, we have performed a two-way analysis of variance (see 2.2.5). The
statistical results are presented in full in Appendix B, but a summary is provided
in Table 3.2-1. The residue is the amount of variance that is not accounted for by
the two factors or their interaction.

Table 3.2-1. Effects of the discipline and language factors and their interaction on variance
in the KIAP Corpus
Feature Discipline Language Interaction Residual
Metatext 52 % 12 % 2% 34 %
Bibliographical references 44 % 5% 2% 49 %
Adversative conjuctions 25 % 9% 1% 65 %
Negation 18 % 18 % 0% 64 %
First person subjects 15 % 11 % 1% 73 %
Indefinite pronoun subjects 9% 49 % 3% 39 %

Both the discipline factor and the language factor have a significant effect for all
the features. But for the majority of features, discipline turns out to be more im-
portant than language. That is, authors of research articles tend to write more like
their disciplinary colleagues writing in other languages than like their language-
community co-members writing in other disciplines – with respect to most of the
features that we have chosen to study here.
The answer is different for each of the features. The feature where discipline
has the strongest influence is metatext, where 52 % of the variance in the popula-
tion of texts can be attributed to differences between disciplines. As much as 12 %
is due to the language factor, and another 2 % to discipline–language interaction.
The strong contribution of the discipline factor has above all got to do with a lower
relative frequency of metatext in medicine. For bibliographical references, 44 % of
the variance can be accounted for by reference to disciplinary differences, whereas
only 5 % is due to differences among the languages, and 2 % to discipline–language
interaction. The main contribution to the importance of disciplinary differences
comes from the fact that medical articles have many more references than do arti-
cles from the other disciplines. For adversatives, too, the discipline factor accounts
for a larger amount of variance than does the language factor; their respective
contributions are 25 % and 9 %, whereas their interaction is not significant. In this
case, the main contribution comes from linguists’ using adversatives more than
authors from the other disciplines. For first person subjects, disciplinary differ-
ences contribute to 15 % of the variance and language differences to 11 %, with no
 Chapter 3. Quantitative results 55

significant interaction. Medical articles have fewer such subjects than there are in
economics and linguistics. For negation, the factors are equally important, both
accounting for 18 % of the variance, with no significant interaction. Negation is
used considerably more in linguistics and in Norwegian articles than in the other
disciplines and languages. Finally, for indefinite pronoun subjects, the language
factor contributes to a vastly greater share of the variance than discipline; their
respective effects are 49 % and 9 %, whereas the interaction between them ac-
counts for 3 %. Not unexpectedly, this result is due to the fact that English uses
the indefinite pronoun one much more seldom than is the case for French on and
Norwegian man, en/ein. Many of these issues will be discussed in greater detail in
later chapters.
It is important to realise that for all the features we have studied, much of the
variation cannot be accounted for by reference to the factors of discipline or lan-
guage or their interaction. For metatext, the share of variance that is unaccounted
for is about 34 %, whereas it is as much as about 73 % for first person subjects. The
numbers for the remaining features lie between these two. This implies that much
of the variance is due to other factors, such as age, gender and number of authors,
subdisciplines, editorial practices, etc. No doubt, individual variation that cannot
be subsumed under any of the more easily accessible factors also plays an impor-
tant role in most of these cases.

3.3 Differences and similarities between disciplines

In this section, we sketch our main results of discipline comparisons. Investiga-

tions at the superlevel are here those that compare disciplines for all three lan-
guages taken together, while investigations at the sublevel are those that compare
disciplines separately for each language. (See 2.2.5 for explanations of the terms
superlevel and sublevel, and that section and Appendix B for the statistical tests).
We first present the results in terms of the mean ranks of the articles in the
disciplinary subcorpora under comparison. The individual differences and simi-
larities between disciplines are commented on in sections 3.3.1–2. Measures such
as absolute frequencies and median and mean relative frequencies are provided in
tables in later sections, viz. 4.1, 5.1 and 6.1.
Figures 3.3-1–4 employ the concept of ‘mean rank’, and an understanding of
this concept is necessary in order to understand the figures. We explain the con-
cept below. Figure 3.3-1 shows the mean ranks of the disciplinary subcorpora at
the superlevel (expected mean rank: 225.5). All the 450 corpus articles are includ-
ed, and we compare three subcorpora of 150 articles each. Figures 3.3-2–4 show
the mean ranks at the sublevel, for English, French and Norwegian, respectively
56 Academic Voices

(expected mean rank: 75.5). In each case, 150 corpus articles are included, and we
compare three subcorpora of 50 articles each. (In these figures, the mean ranks
have been rounded to integers.)

Figure 3.3-1. Differences between disciplines for the whole KIAP Corpus

Figure 3.3-2. Differences between disciplines in English

 Chapter 3. Quantitative results 57

Figure 3.3-3. Differences between disciplines in French

Figure 3.3-4. Differences between disciplines in Norwegian

The concept of ‘ mean rank’

As stated above, an understanding of the concept of ‘mean rank’ is needed to un-
derstand the figures. We therefore provide an explanation of this concept here.
We will use the sublevel comparison of English economics (engecon), linguistics
(engling) and medicine (engmed) with respect to indefinite pronoun subjects as
an example (cf. Figure 3.3-2).
There are 50 articles in each of these subcorpora, i.e. 150 altogether. The 150
articles are first combined into one group and ranked in order of increasing rela-
58 Academic Voices

tive frequency of indefinite pronouns. We assign the rank 1 to the article with the
lowest relative frequency, the rank 2 to the article with the second lowest relative
frequency, and so on up to the article with the highest relative frequency, to which
we assign the rank 150.1 Next, we calculate sums of ranks for each discipline sub-
corpus separately. These are 4,180 for engecon, 5,024 for engling and 2,121 for
engmed. Finally, we divide the sums by the number of articles (50) in order to
obtain the mean ranks. The mean ranks are approximately 84 for engecon, 100 for
engling and 42 for engmed; they can be found in Figure 3.3-2.
If the discipline subcorpora had been more or less the same with regard to
the relative frequency of indefinites, they would have had approximately the same
ranks, viz. the expected mean rank (expected under the null hypothesis of no dif-
ference between the populations that the subcorpora have been sampled from).
The expected mean rank is between the middle ranks 75 and 76, i.e. 75.5. We
see that engecon has a slightly higher mean rank, and the one for engling is even
higher, while the one for engmed is much lower. The mean ranks tell us that eng-
ling articles tend to have a higher relative frequency than engecon articles (100 >
84), and engecon articles a higher one than engmed articles (84 > 42).
Note that the size of a mean rank does not tell us how frequent the feature
in question is. We use the mean ranks to compare subcorpora, and they tell us
whether articles in one subcorpus tend to have higher or lower relative frequen-
cies than articles in the other subcorpora that the former subcorpus is compared
to. To make this point clearer, let us look at the comparison of engling with French
and Norwegian linguistics (frling and noling), again with respect to indefinites.
The mean ranks obtained in this comparison can be found in Figure 3.4-3 in sec-
tion 3.4 on language differences and similarities. The three mean ranks are 31 for
engling, 113 for frling and 83 for noling. Note that engling, which has a mean rank
of 100 when compared to engecon and engmed, in this case has a mean rank of
31 (see Figure 3.3-2). Of course, the relative frequency of indefinites in engling is
the same in both cases; the median is 0.036 % (cf. Table 4.1-4 in Chapter 4). When
engling is compared to the other English disciplines, it is the subcorpus with the
highest relative frequencies of indefinites, but when it is compared to linguistics in
the other languages, it is by far the subcorpus with the lowest relative frequencies.
Thus, the mean rank of one subcorpus for a specific feature is properly com-
pared only to the mean ranks of the other subcorpora for that same feature in that
very comparison. Comparisons across features and across different sets of com-
pared corpora should only be made in relative terms. For instance, it can be seen

1. There is a simple formula to calculate ranks when two or more articles have exactly the
same relative frequency of the feature in question. This almost exclusively occurs when the rela-
tive frequency is 0, i.e. when there are no occurrences of the feature in question.
 Chapter 3. Quantitative results 59

from Figures 3.3-3 and 3.3-4 that with respect to indefinites, the differences between
the disciplines are greater in French (with mean ranks of 80, 106 and 40) than in
Norwegian (with mean ranks of 71, 84 and 72). However, the apparent similarity of
frecon (mean rank when compared to frling and frmed: 80) and noling (mean rank
when compared to noecon and nomed: 84) is only apparent; it makes no sense to
compare these two mean ranks since they stem from different rankings.

3.3.1 Which differences between disciplines are significant?

The focus of this section is on differences between disciplines rather than simi-
larities. This is a natural consequence of the use of statistical methods that test
whether the differences between subcorpora are significant. The similarities be-
tween disciplines can be extracted from the description in the present section, but
are highlighted in 3.3.2.
In general, the differences between the disciplines are a little bigger in English
than in French, and a little bigger in French than in Norwegian. This observation
is based on the sum of differences in mean ranks between disciplines (three pairs)
for all six features – in each of the languages.
First person subjects
Medical articles have far fewer first person subjects than articles from economics
and linguistics, and the differences between medicine and economics and between
medicine and linguistics are significant at the superlevel as well as at the sublevel (in
each language), whereas the difference between economics and linguistics is not.
Indefinite pronoun subjects
In general, linguistics articles have more indefinite pronoun subjects than articles
from the other disciplines, and economics more than medicine. These differences
are significant at the superlevel and at the sublevel in English and French. In Nor-
wegian, however, there are no significant differences between the disciplines.
Metatext
There is more metatext in economics articles than in linguistics articles, and much
more in linguistics than in medicine. These differences are significant at the su-
perlevel, and at the sublevel in French and Norwegian. In English, the difference
between economics and linguistics is not significant, but the differences between
economics and medicine and between linguistics and medicine are.
Negation and adversatives
There is much more frequent use of negation and adversatives in linguistics than
in economics and medicine, and the differences between linguistics and econom-
ics and between linguistics and medicine are significant at the superlevel as well as
60 Academic Voices

at the sublevel (in each language), whereas the difference between economics and
medicine is not.
Bibliographical references
There are many more bibliographical references in medical articles than in articles
from the other disciplines, and more in linguistics than in economics. These differ-
ences are significant at the superlevel and at the sublevel (in each language), except
that the difference between economics and linguistics is not significant in French.

3.3.2 Similarities between disciplines

In this section we first specify for each feature which two disciplines are most
similar and then summarise those of our findings that indicate that disciplines are
similar, in the sense that the difference between them is not significant.
With respect to similarities across disciplines, it is easy to generalise: Only with
regard to one feature in one language, viz. bibliographical references in English,
are linguistics and medicine more similar to one another than either of them is to
economics. (The distances between the disciplines are calculated as differences be-
tween mean ranks.) In all other cases, it is either economics and linguistics that are
most similar, or economics and medicine. The distribution of most similar pairs of
disciplines over feature–language pairs is shown in Figure 3.3-5.

Figure 3.3-5. Most similar disciplines (smallest differences between mean ranks)

As became clear in 3.3.1, most discipline pairs are significantly different with re-
spect to the relative frequencies of the various features, at the superlevel and at the
individual language sublevels. We summarise below, for each discipline pair at the
various levels, which features do not exhibit any significant difference, i.e. where
the disciplines resemble one another closely. (Note that this is a different issue
from the one illustrated in Figure 3.3-5, where the question is which discipline
pair is the most similar, rather than whether the differences are significant or not.)
We begin at the superlevel:
• Economics and linguistics are similar for first person.
• Economics and medicine are similar for negation and adversatives.
 Chapter 3. Quantitative results 61

At the sublevels for the three languages, we find the following resemblances:
• English economics and linguistics are similar for first person and metatext.
• English economics and medicine are similar for negation and adversatives.
• French economics and linguistics are similar for first person and biblio-
graphical references.
• French economics and medicine are similar for negation and adversatives.
• Norwegian economics and linguistics are similar for first person and indefinites.
• Norwegian economics and medicine are similar for indefinites, negation
and adversatives.
• Norwegian linguistics and medicine are similar for indefinites.

3.4 Differences and similarities between languages

In this section, we sketch our main results of language comparisons. Investigations

at the superlevel are here those that compare languages for all three disciplines tak-
en together, while investigations at the sublevel are those that compare languages
separately for each discipline. (See 2.2.5 for explanations of the terms superlevel
and sublevel, and that section and Appendix B for the statistical tests.)
We first present the results in terms of the mean ranks of the articles in the language
subcorpora under comparison. The concept of ‘mean rank’ was explained in section
3.3. The individual differences and similarities between languages are commented on in
sections 3.4.1–2. Measures such as absolute frequencies and median and mean relative
frequencies are provided in tables in later sections, viz. 4.1, 5.1 and 6.1.

Figure 3.4-1. Differences between languages for the whole KIAP Corpus
62 Academic Voices

Figure 3.4-1 shows the mean ranks of the language subcorpora at the superlev-
el (expected mean rank: 225.5). All the 450 corpus articles are included, and we
compare subcorpora of 150 articles each. Figures 3.4-2–4 show the mean ranks of
the language subcorpora at the sublevel, for economics, linguistics and medicine,
respectively (expected mean rank: 75.5). In each case, 150 corpus articles are in-
cluded, and we compare three subcorpora of 50 articles each.

Figure 3.4-2. Differences between languages in economics

Figure 3.4-3. Differences between languages in linguistics

 Chapter 3. Quantitative results 63

Figure 3.4-4. Differences between languages in medicine

3.4.1 Which differences between languages are significant?

The focus of this section is on differences between languages. The similarities can be
extracted from the description in the present section, but are highlighted in 3.4.2.
The differences between the languages are clearly bigger in linguistics than in
economics, and bigger in economics than in medicine. This observation is based
on the sum of differences in mean ranks between languages (three pairs) for all six
features – in each of the disciplines.
First person subjects
French has fewest first person subjects and Norwegian most. The differences be-
tween English and French and between French and Norwegian are significant at
the superlevel and at the sublevel (in each discipline). The difference between Eng-
lish and Norwegian is significant at the superlevel and, at the sublevel, in linguis-
tics, but not in economics and medicine.
Indefinite pronoun subjects
English has far fewer indefinite pronoun subjects than the other languages. French
has more than Norwegian, except in medicine, where the situation is reversed. All
the differences, on the superlevel as well as the sublevel (in each discipline), are
significant.
Metatext
There is clearly more metatext in English articles than in Norwegian ones, and
more in Norwegian than in French. The differences between the languages are
significant at the superlevel and, at the sublevel, in economics and linguistics. In
64 Academic Voices

medicine, the differences between English and the other languages are significant,
but there is virtually no difference between French and Norwegian.
Negation
Norwegian has more negation than English and French, and the differences be-
tween English and Norwegian and between French and Norwegian are significant
at the superlevel as well as the sublevel (in each discpline), whereas the difference
between English and French is not.
Adversatives
Norwegian has more use of adversatives than English and French, and the differ-
ences between English and Norwegian and between French and Norwegian are
significant at the superlevel and at the sublevel (in each discipline), except for eco-
nomics, where there is no significant difference between English and Norwegian.
There are no significant differences between English and French, except at the sub-
level in linguistics, where there are more adversatives in French than in English.
Bibliographical references
On the whole, English and Norwegian have similar frequencies of bibliographical refer-
ences, and these languages have more references than French. At the superlevel and at
the medical sublevel, the differences between English and French and between French
and Norwegian are significant, but not the difference between English and Norwegian.
At the economics sublevel, there are no significant differences between the languages.
At the linguistics sublevel, English has more bibliographical references than Norwegian,
and Norwegian more than French. These differences are all significant.

3.4.2 Similarities between languages

In this section we first specify for each feature which two languages are most simi-
lar and then summarise those of our findings that indicate that languages are simi-
lar, in the sense that the difference between them is not significant.
It is more difficult to generalise here than for disciplines, because the pairs
of most similar languages are fairly evenly distributed over the feature–discipline
combinations. The distribution is shown in Figure 3.4-5.

Figure 3.4-5. Most similar languages (smallest differences between mean ranks)
 Chapter 3. Quantitative results 65

As became clear in 3.4.1, most language pairs are significantly different with re-
spect to the relative frequencies of the various features, at the superlevel and at the
individual discipline sublevels. We summarise below, for each language pair at the
various levels, which features do not exhibit any significant difference, i.e. where
the languages resemble one another closely. (Note that this is a different issue from
the one illustrated in Figure 3.4-5, where the question is which language pair is the
most similar, rather than whether the differences are significant or not.) We begin
at the superlevel:
• English and French are similar for negation and adversatives.
• English and Norwegian are similar for bibliographical references.

At the sublevels for the three disciplines, we find the following resemblances:
• English and French economics are similar for negation, adversatives and
bibliographical references.
• English and Norwegian economics are similar for first person, adversa-
tives and bibliographical references.
• French and Norwegian economics are similar for bibliographical refer-
ences.
• English and French linguistics are similar for negation.
• English and French medicine are similar for negation and adversatives.
• English and Norwegian medicine are similar for first person and biblio-
graphical references.
• French and Norwegian medicine are similar for metatext.

The quantitative results presented in this chapter will serve as a backdrop for our
thematically oriented analyses and discussions in the next three chapters.
 chapter 4

Presence of the author

4.0 Introduction

Every text has a responsible author, or a responsible group of authors. This also
holds for the research article, although this and related genres have often been
considered as relatively objective and impersonal, exhibiting few traces of the re-
searcher. In the last few decades, expressions pointing to a personal presence (es-
pecially personal pronouns) in the research article and other academic genres have
been at the centre of attention in a number of studies. These studies have often
focused on the rhetorical and strategic functions of personal and indefinite pro-
noun use. As Hyland (2001b: 223) observes, “first person pronouns […] are not
just stylistic optional extras but significant ingredients for promoting a competent
scholarly identity and gaining accreditation for research claims”. Previous research
has documented that there are considerable differences between disciplines in this
respect (cf. e.g. Kuo 1999; Hyland 2001a, b; Harwood 2005), as well as differences
between languages (cf. Vassileva 2000). The present study is, to the best of our
knowledge, the first doubly contrastive study (except for other, earlier KIAP stud-
ies like Fløttum (2003b)), taking both the discipline and the language factor into
systematic consideration.
This chapter is primarily an investigation of the use of first person singular
and plural subjects and indefinite pronoun subjects like English one in the bodies
of research articles. In addition, it addresses aspects of author manifestation in
research article abstracts. Thus, we focus primarily on the self-dimension rather
than the other-dimension. The use of first person pronouns is the most important
way for authors to make themselves visible in their texts. Hence, such pronouns
constitute our primary source when we address our first main question (see 1.1):
How do article authors manifest themselves in the texts?
Our study of first person singular subjects (for which we will often write ‘I’,
abstracting away from the actual forms in the three languages) in 4.2 is largely an
attempt to systematise the various roles that an author assigns to him- or herself
in the text, viz. the author as writer, researcher, arguer and evaluator. These roles
are employed to varying extents in the different disciplines and languages. Espe-
cially interesting in connection with ‘I’ is French, where there is a long tradition
68 Academic Voices

of avoiding the use of this pronoun in scholarly writing (cf. Loffler-Laurian 1980:
135): To what extent is this avoidance still observed, and does it affect the use of
the other pronouns (‘we’ and ‘one’)?
When we turn to first person plural subjects (‘we’) in 4.3, our main focus is on
the kinds of reference that this pronoun has in research articles: ordinary exclusive
and inclusive use as well as different types of metonymic use; we discuss the rhe-
torical functions of these referential types of ‘we’ and the extent of their presence
in the disciplines and languages under study. One interesting use is inclusive ‘we’,
where the author creates a common ground for him- or herself and the reader, a
central rhetorical means of creating agreement. Another point of interest is so-
called authorial ‘we’, where one author refers to him- or herself alone by means of
the plural. Such a usage may be an expected result of ‘I’-avoidance in French, but
in English there is, at least in some disciplines, a tendency to avoid such usage,
since it is considered stylistically bad – but as is evident from previous research
(Hyland 2001b; Harwood 2005), the authorial ‘we’ is far from absent in English
academic writing (see below).
It might perhaps surprise some readers that indefinite pronouns like one (for
which we will write ‘one’ when referring to more than one language) are included
in a chapter on authorial presence. English one is rather infrequent. It is gener-
ally considered as formal, even pretentious and affected, and it appears that many
people avoid using it altogether (Wales 1996: 82–83). The closest corresponding
pronouns in French and Norwegian are more frequently used and are stylistically
different from English one. Especially French on may function as a substitute for
personal expressions, including first person pronouns. The functional affinity be-
tween ‘one’ and ‘we’ is evident, something that has also been observed for English
(Mühlhäusler & Harré 1990: 178; Wales 1980: 37). This fact is one of our motiva-
tions for including ‘one’ in the present chapter. As in the case of ‘we’, questions of
reference are central to our discussion of ‘one’.
In the following paragraphs, we will review several relevant previous studies.
Loffler-Laurian (1980) looks at French research articles from physics and chem-
istry and finds no use of the first person singular je (‘I’) at all in single-author
articles. This absence is explained with reference to the general discouragement
of je-use in French academic writing (see above). Loffler-Laurian finds, however,
that nous (‘we’), as opposed to je, is used. Her examples seem to exhibit a mixture
of exclusive and inclusive uses of nous. The use of nous, too, is rather modest,
however, and she thinks the restrictions on je-use have led to a certain reluctance
to use nous as well.
Kuo (1999) is a study of 36 articles from three disciplines: computing science,
electronic engineering and physics. Each of her corpora has one single-author and
nine multi-author articles. The articles are written in English, apparently by both
 Chapter 4. Presence of the author 69

native and non-native authors. None of the single-author articles employ I. Kuo
finds relative frequencies of we ranging from 0.413 % for physics to 0.749 % for
computing science. These figures can be compared to our results for multi-author
articles, although the disciplines are different. Kuo mentions finding examples of
exclusive we for the author alone in the three single-author articles. She classifies
the examples of we according to various types of inclusive and exclusive reference
(with or without the reader or third persons), including a residual group of am-
biguous cases (cf. our discussion of problems of classification in 4.3.1). About 30 %
of the examples represent inclusive we, and about 70 % are exclusive uses.
Vassileva (2000) investigates single-author linguistics articles in five languag-
es: English, French, Bulgarian, German and Russian. Her findings for English and
French are especially interesting to us. She does find use of the first person sin-
gular in both languages, but does not provide relative frequencies, which makes
comparisons difficult. She does, however, compare the frequencies of first person
singular and first person plurals and finds the distribution of first person singular
versus plural pronouns in single-author articles to be 69 % ‘I’ and 31 % ‘we’ in
English, and 40 % ‘I’ and 60 % ‘we’ in French.
Hyland (2001b) is an investigation of self-mention in research articles and is
based on a corpus of 240 articles in English from eight disciplines. Among these,
the closest to our three disciplines, economics, linguistics and medicine, are soci-
ology, applied linguistics and microbiology, respectively. The same corpus is used
in Hyland (2001a), which is a study of addressee features, including inclusive we.
The articles in his corpus are all written in English, but may not be limited to na-
tive speakers. As far as we can see, no systematic information about the number
of single-author versus multi-author articles is provided in Hyland (2001a, b) (al-
though there are some remarks to the effect that the number of single-author ar-
ticles is low in the hard disciplines and high in the soft ones (2001b: 217–218)).
Neither is such information available from Hyland (2000: 179–181), where the
article sources are listed, but without author information. The relative frequencies
of the first person singular I are 0.127 % for sociology and 0.361 % for applied
linguistics, while there are no examples in microbiology (Hyland 2001b: 212). We
can find no information, however, about what these frequencies are relative to;
hence, we do not know whether they are based on the total numbers of words in
single-author articles or on those of all articles, including the multi-author ones.
This makes a comparison with our results for I very difficult. As for the plural,
the relative frequencies provided in Hyland (2001a) appear to include subject and
non-subject forms as well as possessives, while in Hyland (2001b) the frequencies
of we, us and our are specified separately. Of the three disciplines most relevant in
our connection, applied linguistics has the most frequent use of first person plural
pronouns and possessives and microbiology the least frequent. More interesting
70 Academic Voices

than the relative frequencies is the fact that Hyland (2001b) finds quite large shares
of exclusive we even in the soft disciplines, which have many single-author arti-
cles. For first person singular and plural pronouns and possessives as a whole, it
can be seen from Hyland’s (2001a, b) figures that applied linguistics clearly has a
higher frequency than sociology, and sociology has a much higher frequency than
microbiology. Hyland (2001a) also includes an addressee feature called “Indefi-
nite”. It appears from his examples that this category covers both you and one with
indefinite reference. It is therefore hardly comparable to our study of English one,
French on and Norwegian man, en/ein.
An early KIAP study is that of Fløttum (2003b), which is based on 180 research
articles, viz. the first 20 from each of the subcorpora employed in the present study.
This is the only previous doubly contrastive study that we are aware of, and it will
be central to our formulation of hypotheses below. It involves the use of ‘I’, ‘we’ and
‘one’. The relative frequencies reported are summarised in Table 4.0-1. The relative
frequency for the singular is calculated on the basis of only single-author arti-
cles. The medical subcorpora are not included in that column because of their low
number of single-author articles, but the few examples of ‘I’ in these subcorpora
are included in the sum in the column for ‘I’ + ‘we’. The figures for ‘we’ include both
single- and multi-author articles.

Table 4.0-1. Pronoun use as found in Fløttum’s (2003b) KIAP study

1st p. sg. 1st p. pl. Sum 1st p. Indefinite

engecon 0.22 0.47 0.60 0.03

engling 0.26 0.40 0.55 0.03
engmed – 0.37 0.38 0.00
frecon 0.02 0.25 0.26 0.44
frling 0.12 0.32 0.44 0.68
frmed – 0.14 0.14 0.22
noecon 0.11 0.58 0.67 0.25
noling 0.35 0.83 1.17 0.30
nomed – 0.30 0.32 0.31
Relative frequencies in %.

As can be seen from the table, Fløttum (2003b) finds that there is more first person
singular use in linguistics than in economics, and more in English and Norwegian
than in French. English has more use of ‘I’ than Norwegian in economics, while
the situation in linguistics is the reverse. If we disregard a few exceptions, some
general tendencies seem to be discernible in the use of ‘we’ and of ‘I’ and ‘we’ to-
gether, viz. that there is most frequent use in Norwegian and least frequent use in
 Chapter 4. Presence of the author 71

French, and that the use is most frequent in linguistics and least frequent in medi-
cine. As regards indefinite ‘one’, English has fewer such pronouns than French and
Norwegian, while the relation between French and Norwegian differs among the
disciplines. In English, medical writers, as opposed to economists and linguists,
appear not to use one at all. In Norwegian, the disciplinary differences are rather
small. French linguists use on more than economists, and economists use it more
frequently than medical writers.
Harwood (2005) is a study of 80 articles from the disciplines of economics,
business and management, physics and computing science. The articles are all sin-
gle-author articles in English. From the corpus contents list (ibid.: 370–372) it
seems that some authors are native speakers of English and some are not. Most
non-body parts of the articles have been excluded, but not the acknowledgements.
For I, the relative frequencies range from 0.010 % in physics (all examples in the
acknowledgements, however) to 0.424 % in business and management; the figure
for economics is 0.324 %. For we, Harwood gives separate figures for exclusive
and inclusive uses. Adding these together, we find that the relative frequencies
range from 0.063 % in economics, 0.105 % in business and management, 0.597 %
in physics to 0.717 % in computing science. We have also calculated Harwood’s
(2005) distributions of inclusive and exclusive we, which are quite revealing. They
are: economics: 98 % inclusive, 2 % exclusive; business and management: 100 %
inclusive; physics: 9 % inclusive, 91 % exclusive; computing science: 33 % inclu-
sive, 67 % exclusive. That is, while exclusive we used by single authors is absent
or almost so in the first two disciplines, it is more common than inclusive we in
the last two. Finally, adding together the frequencies for ‘I’ and ‘we’ provided by
Harwood, we find that the frequency of first person ranges from 0.387 % in eco-
nomics, 0.529 % in business and management, 0.607 % in physics to 0.740 % in
computing science. It is noteworthy that Fløttum’s (2003b) and Harwood’s (2005)
results for we (and therefore for first person in general) in (English) economics are
vastly different, but it should be kept in mind that the former study includes both
single- and multi-author articles, while the latter only has single-author articles.
In the formulation of our hypotheses, we will refer to the KIAP study of Fløt-
tum (2003b), since the corpus used there is included in the corpus of the present
study and amounts to 40 % of it in terms of the number of articles. Of course, it is
to be expected that the present study will mostly corroborate the findings of the
2003 investigation. The results would have been stronger if they confirmed studies
on independent data, but as will be evident from the section on previous studies,
the possibilities of comparison with results from those are limited. In addition, it
is not easy to see any clear patterns in previous studies that would help us much in
formulating hypotheses.
72 Academic Voices

First, we posit the following hypothesis regarding disciplinary differences in

the use of first person subject pronouns. Hypothesis 1: There is more frequent use
of first person in linguistics than in economics, and more in economics than in
medicine. This hypothesis is meant to hold for both ‘I’ and ‘we’, as well as for the
two together. (Medicine will not, however, be included in the comparisons for ‘I’,
because of the low number of single-author articles.)
Second, we posit the following hypothesis regarding language differences in the
use of first person subject pronouns. Hypothesis 2: There is more frequent use of
first person in Norwegian than in English, and more in English than in French. This
hypothesis is meant to hold for both ‘I’ and ‘we’, as well as for the two together.
As regards the use of ‘one’, it is hardly worth mentioning that we expect it to
be less frequent in English than in the other languages. Concerning the relation
between French and Norwegian as well as between the disciplines, we expect the
complex findings of Fløttum (2003b) to be corroborated as our corpus has been
extended (see above).
In the study of authorial presence as realised by first person and indefinite pro-
nouns, the polyphonic perspective (see 2.1.2) has a clear relevance. The abstract
locutors, corresponding to our authors, organise polyphonic dramas where their
choice of role assignment may be signalled in different ways. They can choose to
make their own presence explicit in more or less direct ways. In single-author
articles, the presence of ‘I’ constitutes the most direct author manifestation. De-
pending on the verb combined with the pronoun, the author may present different
“self-performances” or assign different roles to him- or herself. This author-voice
perspective will be the subject matter of section 4.2. In multi-author articles, ‘we’
can be used in a parallel way. The polyphonic perspective is even more relevant
in connection with ‘we’ than with ‘I’, since the variable reference of this pronoun
complicates the polyphonic play considerably. This is especially clear in single-
author articles. Major parts of section 4.3 are devoted to the issue of reference.
Similar issues as for ‘we’ are central to the study of how ‘one’ is used. In section
4.4, we investigate such pronouns, with a special focus on the French pronoun
on, by which a spectrum of different voices or points of view may be included or
excluded: the author (alone or including others), the reader or third persons.
Section 4.1 presents the main quantitative data and discusses the results of sta-
tistical analyses, which are also compared to previous research, while section 4.2 is
primarily a discussion of the roles of the author in connection with self-mention in
the form of ‘I’ grammatical subjects in single-author articles. Section 4.3 looks at
the use of ‘we’ in multi- as well as single-author articles, with a focus on the refer-
ence and functions of such grammatical subjects. Section 4.4 deals with the use of
indefinite pronoun grammatical subjects, especially the various types of reference
that such pronouns have in actual use. In section 4.5, we make a brief detour into
 Chapter 4. Presence of the author 73

the research article abstracts in order to study author roles and evaluation in such
texts. Finally, section 4.6 contains case studies, of pronoun use, of the implicit and
explicit presence of ‘we’ and of author roles and promotion in abstracts.

4.1 Quantitative and comparative results

As we saw in Chapter 3, the use of first person and indefinite pronouns as gram-
matical subjects exhibits considerable variation both among the disciplines and
among the languages. Here we will deal with first person subjects and indefinite
pronoun subjects in turn.

4.1.1 First person subjects

Our investigation of first person manifestation is limited to the use of first person sin-
gular and plural pronouns used as grammatical subjects. This means that our searches
in the KIAP Corpus include English I and we, French je/j’ and nous, Norwegian Bok-
mål jeg and vi and Nynorsk eg and vi/me. Other uses of first person pronouns than
those with subject function were not included, i.e. we did not search for English me
and us, French me/m’ and moi (all ‘me’), Norwegian meg (‘me’) and oss (‘us’). Further,
non-subject uses of French nous were excluded. Adnominal and pronominal first
person possessives were not included, i.e. English my, mine, our and ours and corre-
sponding forms in French and Norwegian. Myself and ourselves were also excluded.
There are three main reasons for our not including personal pronouns in other
than subject functions (i.e. primarily objects and complements of prepositions).
The first and most important is that by using the pronouns in these other functions,
the authors typically take on a semantically less prominent role than when referring
to themselves by subject pronouns, since subject referents are figures and object
referents are grounds (cf. Langacker 1987 and the cognitive linguistics literature
more generally). We are more interested in the most prominent instances of author
presence. The second reason is that the non-subject instances are less frequent than
the subject ones.1 The third is that including objects would have created problems
in relation to reflexive constructions, where there are two coreferent pronouns, and
whose frequency is very much subject to cross-linguistic variation that is irrelevant
for present purposes. The inclusion of possessives could have added an interesting

1. To give an example: in the 35 single-author English linguistics articles, there are 872 exam-
ples of I subjects, but only 30 instances of me. 19 of these are used as complements of a prepo-
sition, while 11 are used as objects (of which 6 are objects of let in a let me-imperative). Similar
figures are found by Hyland (2001b).
74 Academic Voices

dimension to the investigation, but we needed to restrict our focus. Also, reference
to first person by means of possessives means an even less prominent author pres-
ence than does, e.g., reference by means of object pronouns.
We now turn to the results. Disciplinary differences in the relative frequency
of first person subjects (‘I’ and ‘we’) account for a larger share of the variance than
do differences among languages, but about 73 % of the variance cannot be attrib-
uted to any of these factors or their interaction (cf. 3.2 and Appendix B). There
is less use of first person subjects in medicine than in the other disciplines, while
economics and linguistics are very similar in this respect. Thus, hypothesis 1 (see
4.0) has only partly been confirmed. Our finding that medicine has less first per-
son use than the other disciplines corresponds with Hyland’s (2001a, b) results for
microbiology in relation to sociology and applied linguistics. His results of more
frequent use in applied linguistics than in sociology appear to conflict with ours
for linguistics and economics. As for the language comparison, French uses first
person subjects less than the other two languages. For English and Norwegian, the
difference is smaller. It is significant in linguistics (and at the superlevel), where
Norwegian has a higher relative frequency, but not in the other disciplines. Thus,
hypothesis 2 has also only partly been confirmed.
The feature of first person subjects is a composite one – the frequencies are the
sums of three separate frequency counts: first person singular (‘I’) and plural (‘we’)
subjects in single-author articles and first person plural (‘we’) subjects in multi-
author articles. These three feature components contribute in rather different ways
to the overall first person feature. It is obvious that the difference in numbers of
single- and multi-author articles in the three disciplines complicates disciplinary
comparisons, at the superlevel as well as at the sublevel. To a certain degree, it also
complicates language comparisons, since the distributions of single- and multi-
author articles differ in that respect, too, albeit to a smaller extent.
Because of these complications, we have also studied the three feature compo-
nents of first person separately. We have performed statistical analyses (Kruskal–
Wallis2 and Mann–Whitney–Wilcoxon tests, α = 0.05, cf. 2.2.5) of the differences
in relative frequency between disciplines and languages.3 Since the corpus con-
tains few medical articles written by one author, medicine has been excluded from
the study of ‘I’ and ‘we’ in single-author articles. Similarly, there are few linguistics
articles with several authors, and this discipline is not included in the study of

2. The results of the Kruskal–Wallis test (one-way analysis of variance) are reported here only
when no significant variance is found.
3. No details about these and other subsidiary statistical analyses have been included, only
the information about significant results or lack of such.
 Chapter 4. Presence of the author 75

‘we’ in multi-author articles. (However, the data for the excluded subcorpora are
included in Tables 4.1-1–3.)

First person singular subjects

Table 4.1-1 presents data for the use of first person singular subjects (I, je/j’ and
jeg/eg) in articles with one author. The most important figures are the median rela-
tive frequencies. As we can see, there is considerable variation both among the
disciplines and among the languages. In general, it can be said that ‘I’ is used more
by linguists than by economists (and possibly more by economists than by medical
authors). The frequency that we find for engecon is considerably lower than what
Harwood (2005) finds for economics (0.324 %); even engling has less use of ‘I’
than that. Further, we find that ‘I’ is used more in English and Norwegian articles
than in French ones.

Table 4.1-1. The use of first person singular subjects in single-author articles

Corpus No. of + ’I’ No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 23 11 242 0.184 0.000 0.261 0.152

engling 35 32 872 0.273 0.234 0.235 0.186
engmed 1 1 7 0.309 0.309 n.a. 0.000
frecon 31 1 15 0.010 0.000 0.055 0.000
frling 48 24 182 0.082 0.007 0.126 0.077
frmed 4 0 0 0.000 0.000 0.000 0.000
noecon 33 19 212 0.107 0.039 0.158 0.062
noling 48 47 991 0.373 0.346 0.272 0.203
nomed 10 2 9 0.047 0.000 0.136 0.000

econall 87 31 469 0.093 0.000 0.180 0.033

lingall 131 103 2045 0.240 0.169 0.251 0.196
medall 15 3 16 0.052 0.000 0.132 0.000

engall 59 44 1121 0.239 0.172 0.245 0.199

frall 83 25 197 0.051 0.000 0.107 0.019
noall 91 68 1212 0.241 0.130 0.264 0.210
KIAP 233 137 2530 0.173 0.054 0.233 0.141
Relative frequencies and deviations in %; + ‘I’ = number of articles using ‘I’;
n.a. = not applicable.
76 Academic Voices

The discipline factor. The difference between economics and linguistics is statisti-
cally significant both at the superlevel and at the sublevel (in each language).4 This
confirms hypothesis 1 (see 4.0) for ‘I’.
The language factor. The differences between English and French and between
Norwegian and French are statistically significant both at the superlevel5 and at the
sublevel. The smaller differences between English and Norwegian that can be seen
in the table are not statistically significant at either level. Hypothesis 2 is therefore
only partly confirmed for ‘we’.
The gender factor. We have also compared articles written by one male author with
those written by one female author. There are no significant differences in the use
of ‘I’ subjects between the genders in any of the six discipline–language subcor-
pora, nor when they are considered together (in various ways).6
First person plural subjects
Table 4.1-2 presents data for the use of first person plural subjects (we, nous, vi/me)
in articles with one author, i.e. the same subcorpora as for ‘I’ in Table 4.1-1. The
results are rather different from those found for ‘I’. In fact, ‘we’ is used more than
‘I’ in all the subcorpora (except in the single engmed article) and has a median
relative frequency that is more than five times as high for the whole KIAP Corpus
as that of ‘I’ – 0.28 % as compared to 0.05 %.
If we compare our figures to those of Harwood (2005), we see that the fre-
quency in both engecon and engling is considerably higher than in his economics
(0.063 %) and business and management (0.105 %) corpora, but lower than his
results for physics (0.597 %) and computing science (0.717 %).
Adding together our median frequencies for ‘I’ and ‘we’ in single-author arti-
cles, we can again compare with Harwood’s findings. Engecon has a frequency of
only 0.189 %, which is much lower than Harwood’s result for economics (0.387 %).
Our engling reaches a little higher than that (0.471 %), but that is also lower than
what Harwood finds for his other three disciplines, with computing science top-
ping the list at 0.740 %.
Our findings can also be compared to those of Vassileva (2000). We find rather
different distributions of ‘I’ and ‘we’ in single-author linguistics articles than she
does. Based on the number of tokens in Tables 4.1-1–2, they can be calculated
to: 48 % ‘I’ and 52 % ‘we’ in engling (Vassileva: 69 % vs. 31 %); 26 % ‘I’ and 74 %
‘we’ in frling (Vassileva: 40 % vs. 60 %); 37 % ‘I’ and 63 % ‘we’ in noling. Thus, we

4. For an explanation of our terms superlevel and sublevel, see 2.2.5.

5. Recall that only economics and linguistics are included, so that the superlevel involves two
rather than three disciplines here.
6. Mann–Whitney–Wilcoxon test, α = 0.05; see section 2.2.5.
 Chapter 4. Presence of the author 77

find considerably higher shares of ‘we’-use in both English and French than Vas-
sileva does. On the other hand, our results corroborate hers that the share of ‘we’ is
greater in French than in English (with Norwegian assuming a middle position).

Table 4.1-2. The use of first person plural subjects in single-author articles

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 23 426 0.300 0.189 0.374 0.197

engling 35 947 0.286 0.237 0.250 0.147
engmed 1 0 0.000 0.000 n.a. 0.000
frecon 31 674 0.409 0.312 0.379 0.261
frling 48 527 0.242 0.195 0.250 0.159
frmed 4 45 0.278 0.071 0.463 0.124
noecon 33 1118 0.571 0.382 0.544 0.302
noling 48 1705 0.670 0.564 0.502 0.299
nomed 10 52 0.219 0.133 0.217 0.153

econall 87 2218 0.441 0.312 0.456 0.280

lingall 131 3179 0.411 0.263 0.412 0.200
medall 15 97 0.221 0.107 0.284 0.162

engall 59 1373 0.287 0.219 0.302 0.167

frall 83 1246 0.306 0.210 0.320 0.178
noall 91 2875 0.584 0.475 0.510 0.343

KIAP 233 5494 0.410 0.281 0.424 0.236

Relative frequencies and deviations in %; n.a. = not applicable.

Table 4.1-3 presents the data for first person plural subjects (we, nous, vi/me) in
articles with more than one author, i.e. the complement subcorpora of those of
Tables 4.1-1–2. Recall that in our statistical studies, linguistics has been excluded
because of the low number of relevant articles, while medicine is included. The
median relative frequencies are higher than in single-author articles in most of the
subcorpora, including the whole corpus, with a median relative frequency of 0.36
%. But in other respects, the results resemble those for ‘I’ more than those for ‘we’
in single-author articles. ‘We’ is clearly used more frequently in economics than in
medicine, and this goes for each language as well as when the language subcorpora
are treated together (0.78 % in economics compared to 0.21 % in medicine). Just
as for ‘I’ in single-author articles, the frequencies are higher in English and Nor-
wegian than in French.
78 Academic Voices

Our results for multi-author articles can be compared to those of Kuo (1999):
engmed has a lower frequency of we than has Kuo’s physics corpus (0.413 %),
while engling and engecon have lower and higher frequencies, respectively, than
Kuo’s computing science corpus (0.749 %).

Table 4.1-3. The use of first person plural subjects in multi-author articles

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 27 1491 0.879 0.949 0.383 0.238

engling 15 1093 0.728 0.657 0.417 0.269
engmed 49 470 0.303 0.248 0.244 0.160
frecon 19 441 0.392 0.411 0.248 0.123
frling 2 21 0.273 0.273 0.336 0.119
frmed 46 230 0.203 0.166 0.209 0.117
noecon 17 1230 1.110 0.941 0.509 0.351
noling 2 40 0.512 0.512 0.592 0.209
nomed 40 272 0.344 0.293 0.281 0.191

econall 63 3162 0.795 0.776 0.476 0.306

lingall 19 1154 0.657 0.600 0.429 0.301
medall 135 972 0.281 0.209 0.250 0.147

engall 91 3054 0.544 0.451 0.415 0.341

frall 67 692 0.259 0.201 0.236 0.169
noall 59 1542 0.570 0.416 0.502 0.305

KIAP 217 5288 0.463 0.360 0.419 0.257

Relative frequencies and deviations in %.

The discipline factor. In single-author articles, there is no significant difference be-

tween economics and linguistics in the relative frequency of ‘we’, at either super-
level or sublevel (in any language). However, in articles with several authors, the
differences between economics and medicine are significant both at the superlevel
and at the sublevel (in each language). Thus, there is only partial confirmation of
hypothesis 1 (see 4.0).
The language factor. In single-author articles, there are differences between the
languages: Norwegian uses ‘we’ more than the other languages. At the sublevel,
there are significant differences in linguistics between Norwegian and English
and French, respectively, while the difference between English and French is not
significant. The same three results hold at the superlevel. At the economics sub-
level, the variance due to language differences is too small to warrant pairwise
 Chapter 4. Presence of the author 79

comparisons, but the mean rank for Norwegian is higher than those for the other
languages.7 To sum up for single-author articles, hypothesis 2 must be rejected for
several reasons. In articles with several authors, there is least use of ‘we’ in French.
The differences between French and English and Norwegian, respectively, are sig-
nificant both at the superlevel and at the sublevel (in both disciplines), while there
are no significant differences between English and Norwegian at any level. Again,
hypothesis 2 has only partly been confirmed.
The authorship factor. Since the economics subcorpora have a reasonable number
of both single-author and multi-author articles, this discipline allows us to com-
pare the use of ‘we’ in these two categories of articles (cf. especially the median
relative frequencies for engecon, frecon, noecon and econall in Tables 4.1-2–3).
For all of these subcorpora, there is more frequent use of ‘we’ in multi-author than
in single-author articles, but the difference is clearly greater for English and Nor-
wegian than for French. At the sublevel, the difference is significant for the first
two languages, and there is also a significant difference at the superlevel. However,
the difference at the French sublevel is not significant.
The gender factor. We have also compared articles written by one male author with
those written by one female author. There are no significant differences in the use
of ‘we’ subjects between the genders in any of the six discipline–language subcor-
pora, nor when they are treated together (in various ways).

4.1.2 Indefinite pronoun subjects

Our investigation of indefinite pronouns meaning ‘one’ is limited to grammatical

subjects. The corpus searches, then, included English one, French on/l’on and Nor-
wegian man and en/ein. For English, non-subject examples were excluded manu-
ally, as were examples of the numeral one, and the same goes for Norwegian exam-
ples with en/ein in non-subject functions and as a numeral or indefinite article (a
huge number of examples).
The data are presented in Table 4.1-4, and the picture looks rather different
from that of first person pronouns. Here, differences among the languages account
for almost half (49 %) of the total variance in the corpus (see Appendix B for
details). This is because English one is used much less frequently than the corre-
sponding French and Norwegian pronouns (an unsurprising confirmation of our
expectation from 4.0). French uses indefinite pronouns more than Norwegian in

7. Pairwise comparisons indicate that the differences between Norwegian and French and
between French and English are not significant, but that the difference between Norwegian and
English is.
80 Academic Voices

economics and linguistics (sublevels), and when the three disciplines are consid-
ered together (superlevel), whereas the opposite is the case in medicine (sublevel).
This is consonant with Fløttum (2003b). In addition to the massive effect of lan-
guage differences, disciplinary differences account for 9 % of the corpus variance.
In English and French (sublevels), linguistics uses indefinite pronouns more than
economics, and economics uses them more than medicine. This also holds when the
three language subcorpora are considered together. Such differences are not found
in Norwegian (at least not to a statistically significant degree; sublevel), although the
median is a little higher in linguistics than in the other disciplines here too.

Table 4.1-4. The use of indefinite subjects

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 50 108 0.036 0.019 0.052 0.023

engling 50 210 0.049 0.036 0.047 0.034
engmed 50 4 0.001 0.000 0.006 0.000
frecon 50 1213 0.431 0.386 0.300 0.200
frling 50 1564 0.680 0.625 0.399 0.245
frmed 50 238 0.169 0.117 0.163 0.106
noecon 50 834 0.262 0.266 0.177 0.151
noling 50 1019 0.366 0.311 0.285 0.196
nomed 50 296 0.277 0.263 0.199 0.122

econall 150 2155 0.243 0.170 0.259 0.182

lingall 150 2793 0.365 0.255 0.383 0.257
medall 150 538 0.149 0.069 0.186 0.131

engall 150 322 0.029 0.000 0.045 0.021

frall 150 3015 0.427 0.346 0.367 0.243
noall 150 2149 0.302 0.275 0.229 0.151

KIAP 450 5486 0.252 0.141 0.301 0.178

Relative frequencies and deviations in %.

One of the few examples of significant gender differences that we have found for
our features concerns indefinite pronoun subjects in English linguistics. Male Eng-
lish-writing linguists use one more frequently than their female colleagues. Their
respective median relative frequencies are 0.09 % and 0.03 %, and the difference is
statistically significant.8 The median relative frequency for male economists is also
higher than for female authors, but the difference is not significant. However, the

8. Mann–Whitney–Wilcoxon test, α = 0.05; see section 2.2.5.

 Chapter 4. Presence of the author 81

difference between the genders is significant when the disciplines are considered
together. For other subcorpora and combinations of them, there are no significant
gender differences.

4.2 Author roles

Behind every utterance in a text, there is a responsible locutor, corresponding to

the author(s) of the text, even in the research article, traditionally considered as
“neutral” with few traces of the author. However, the presence or the responsibility
of the author manifests itself in different ways in different contexts. In a single-
author text the presence of the first person singular pronoun constitutes the most
explicit manifestation of the author’s voice, as I in the following example:
(1) In fact, the availability of contracts makes marriage less understandable; with
perfect contracts, why marry? I argue that the social distinctions between
marriages and informal relationships are better understood by recognizing
that marriage is not a contract. (engecon18)

However, by choosing specific words constituting the immediate cotext of the

pronoun, the author assigns different roles to him- or herself. In the polyphonic
perspective adopted in this book, we may say that the author’s voice is presented
in different self-performances. The voice presented may be scientifically, strategi-
cally, interactionally or evaluatively biased (see below).
In the previous section we have seen that the frequency of the first person sin-
gular pronoun varies considerably according to discipline, language and individu-
al author. French in particular is known to avoid the use of first person pronouns,
especially the singular, which is reflected in the expression “le moi haïssable” (‘the
I to be hated’; see Loffler-Laurian 1980).9 However, when it is present, in French
and in the other languages, the following question arises: what kind of I or je or
jeg/eg is this? The presence of the pronoun tells us that this is the author speaking,
but considered alone it does not tell us anything about the type of voice manifested
or the type of presence assumed. Quantitative data need to be complemented by
a qualitative analysis concerning the pragmatic-rhetorical nature of this kind of
manifestation.
In this section we will try to answer the question raised above, reformulated as
follows: What are the rhetorical roles the author takes on when referring to him- or

9. We return to this in section 4.4.

82 Academic Voices

herself by means of a first person singular pronoun?10 We do so by looking at the

immediate cotext where the pronoun occurs, i.e. first and foremost the verb con-
struction which it is combined with, but also metatextual expressions surround-
ing it. The verb meaning contributes to the nature of the author’s presence in this
context. We will try to identify the activities – abstract or concrete – which the
author performs by means of a verb. The semantic-pragmatic meaning of the verb
will constitute the first criterion in the determination of the rhetorical role that the
author takes on. The second criterion will be expressions referring to the utterance
situation (deictic expressions other than the first person pronoun) or to text struc-
turing (metatextual expressions). We propose four roles, which will be illustrated
below: the author as writer, as researcher, as arguer and as evaluator.11
As far as we know, there are no directly comparable studies undertaken pre-
viously concerning the quantitative presence of these or comparable roles in re-
search articles (our categories are nevertheless inspired by various studies related
to “author presence”, in particular Vassileva 2000 and Hyland 2000; see below).
However, it is possible to formulate some hypotheses concerning the distribution
of author roles in different disciplines and languages. According to the traditional
conception of academic discourse, one might think that the researcher role will
be the dominant one. However, given the present knowledge about academic dis-
course and the competition taking place in every research community, there is
reason to believe that both writer and arguer will be important roles as well. The
evaluating author might be the less directly present. As regards language differenc-
es, previous studies may lead us to formulate a hypothesis that English authors are
more explicitly argumentative and interactional than Norwegian and especially
French authors.
As indicated above, there are studies related to authorial presence to take into
account in the existing literature. Here we would like to refer to Hyland’s studies
on authorial presence in English language articles taken from different disciplines.
Our work is inspired by his studies on self-mention in research articles (Hyland
2001b) and on authorial identity in academic writing (Hyland 2002a). These have
broader perspectives than our verb-oriented one; however, the classification of
verbs undertaken by Hyland (2000) in his study of citation processes contains cat-
egories similar to ours. His research and cognition acts correspond to some extent
to our researcher role, and his discourse act to our writer role. We will come back

10. We focus here on first person singular pronouns, but the roles are relevant both to first
person plural and to indefinite pronouns when these refer to the author(s), or even to author
and reader.
11. In our previous studies, we have only treated the first three roles (Fløttum 2003g and
2004d); the “evaluator” role is introduced here.
 Chapter 4. Presence of the author 83

to his quantitative results in sections 6.2 and 6.3, in connection with bibliographi-
cal references.
Vassileva’s (2000) study of authorial presence is also relevant, especially since
she investigates the frequency of first person pronouns in different languages (see
4.0). In order to answer her question “who is the author”, she proposes different
so-called “micro-speech acts” relevant as characterisations of first person singu-
lar presence (the ‘I’ perspective). The micro-speech acts are the following (ibid.:
61–63): analysis/argumentation, aims/advance organisers, personal view, personal
experience, focusing, self-reference, conclusions, terminology/procedures, exem-
plification, reference, back organisers and permission. Her classification is quite
fine-grained, and the various acts are very much related to specific parts of the
article structure and content. It may therefore be difficult to generalise on the basis
of them. We would like our role categories to be more general and thereby relevant
to every part of the article (even if some roles might be more frequent in some
parts than in others). In our view a restricted number of roles may contribute to
a more general rhetorical characterisation of author presence in research articles
and enable a comparison between different disciplines.

4.2.1 Verb groups

In order to define the proposed roles, we have chosen the verb meaning as the
first criterion to be considered. We are aware that lexical-semantic and semantic-
pragmatic classifications are problematic, and the four verb groups we propose
are open to discussion. However, they have been established in order to serve the
purpose of the study undertaken here; i.e. we want them to help us in characteris-
ing author manifestation (the author’s voice) and the author’s presentation of own
work. The proposed verb groups have already been used with good results in vari-
ous exploratory studies.12 We propose the four following groups:13
• discourse verbs (e.g. summarise, (re)turn to):14
typically manifested in the writer role
• research verbs (e.g. analyse):
typically manifested in the researcher role

12. See for example Fløttum (2003b, 2003g); Dahl (2004b, c); see also Kinn’s (2004) article on
cognitive research agents.
13. For practical reasons we only give examples of English verbs here.
14. In earlier versions of this classification, the verb group in question has been called rhetori-
cal verbs (RH verbs; see Fløttum 2003b: 41). That term has been replaced by discourse verb since
most of the verbs classified in this group are not argumentative.
84 Academic Voices

• position verbs (e.g. claim):15

typically manifested in the arguer role
• evaluation and emotion verbs and verb constructions (e.g. be sceptical
about; feel):
typically manifested in the evaluator role

The first group consists of what we have called discourse verbs or verb construc-
tions. These verbs, which to some extent correspond to Hyland’s discourse act
verbs (Hyland 2000: 27), denote either processes involving verbal or graphical re-
presentation, such as describe, discuss, illustrate, outline, present, repeat, show, sum-
marise, or processes directly related to the text structuring and the guiding of the
reader (Dahl 2004c), such as begin by, focus on, move on, (re)turn to, conclude by.
The research verbs, in the second group, refer to the action or the activities
directly related to the research process, such as analyse, assume, consider, choose,
compare, explore, find, follow, limit, study, test, use. Some of these verbs (like follow
and use) have a general meaning and are commonly used in non-specialist con-
texts as well. The reason why we put them in this group is that they appear with
complements related to the research context (I use stock return data to calculate …
(engecon06); I follow Whelpton (1995) in assuming that … (engling47)).
Hyland (2000: 27) proposes a separate cognition act concerning mental pro-
cesses as exemplified by verbs like conceptualise and view. We have chosen to in-
clude such verbs in the group of research verbs used in the classification of first
person pronouns. The reason is that it is difficult to make a clear distinction be-
tween the two groups. Another and maybe better justification is the evident fact
that there is a substantial part of cognition in all kinds of research processes. This
means that in the present study we include verbs and verb constructions like be
aware of, know of, see, understand in the research group. However, it might have
been interesting to study to what extent research verbs of a clearly cognitive nature
are used, in other words the proportion of cognitive to other research verbs.16
The position verbs, constituting the third group, denote processes related to
position and stance, explicit argumentation concerning approval, promotion or
rejection. The following are examples of position verbs: argue, claim, dispute, main-
tain, propose, reject.
Finally, we have established a fourth group which contains various emotional
and evaluating constructions. They are used to represent emotional and evaluating

15. In earlier versions, this verb group has been called opinion verbs (OP verbs; see Fløttum
2003b: 41). However, the term position verb seems more appropriate for the group in question.
16. In section 4.4, in the study of indefinite pronouns, we do establish a group of cognitive
verbs.
 Chapter 4. Presence of the author 85

voices of the author – reactions towards different observations made by the author
him- or herself or by somebody else. Examples of such verbs and constructions are
feel, be content to, be sceptical about, be struck by, find something + evaluative adjec-
tive. It may be difficult to maintain a clear distinction between these constructions
and the position verbs, for example between be sceptical about and dispute. How-
ever, the verb dispute is likely to be preferred to being sceptical about in a situation
where the author wants to take a firmer stance. One reason to establish a fourth
group, instead of making the position verb group more open, is that these verbs
are interesting in the study of the potentially individual style of the article.
As indicated above, semantic-pragmatic classification of verbs is quite prob-
lematic. Verbs are lexical items susceptible to different semantic-pragmatic inter-
pretations in different co(n)texts. For that reason some verbs belong to more than
one group. We have chosen not to include exhaustive verb lists, but hope that the
explanations given below will indicate how we have used the verbs as indicators of
specific author roles.
The four verb groups are different in terms of size. The research verb group is
by far the largest one; a wide range of verbs of this type are found in our corpus. As
regards the discourse and the position verbs, they are less numerous, and less fre-
quent. They also seem to be more “general” in that the same verbs tend to be used
across disciplines. Finally, the evaluation and emotion group is rather large, since
we have included open constructions such as find something + evaluative adjective.
However, this group is by far the least frequently manifested.
The four verb groups indicate the nature of the author’s presence in the article.
Through discourse verbs he or she performs explicitly the activity of verbal or
graphical representation or of structuring the text, assuming the writer role, inter-
actionally biased in the guiding of the reader through the text. Through research
verbs the author refers to some activity in the research process, assuming the re-
searcher role, scientifically biased in the reporting of the research activities under-
taken. Through position verbs, he or she takes position in relation to the relevant
research community, presenting own views on a given subject and thus assuming
the arguer role. This role is strategically biased in that the author’s own opinion is
explicitly presented and expressed. Finally, by combining the first person pronoun
with an evaluation or emotion verb construction, the author takes on the role of
evaluator (of own or others’ findings); evaluatively or subjectively biased in that
his or her own feeling related to a specific matter is expressed.
This presentation may suggest a perfect correspondence between verb groups
and author roles. But, as already indicated, there is no clear-cut division between
the verb groups. Different verbs may have different meanings in different (syntac-
tic) contexts, and they may be susceptible to the influence of metatextual elements
in such a way that they switch roles. We will return to this question below.
86 Academic Voices

Before illustrating how the roles are realised in the present corpus, we have
to define what we consider as the main verb, i.e. the verb which constitutes the
basis for the role classification. The notion of main verb should be understood in
a broad semantic-pragmatic sense. It represents the activity the author engages
in. In addition to temporal and aspectual auxiliaries (be/have and correspond-
ing verbs in French and Norwegian) used to form periphrastic constructions, we
exclude modal verbs (will/shall, can, may) as well as other modalising or hedging
elements placed between the pronoun and what we define as the main verb. The
following examples illustrate the semantic-pragmatic conception of main verb (in
bold) used in this classification of author roles:
(2) I will return to this problem below. (engecon06)
(3) I wish to suggest otherwise […]. (engecon10)
(4) I do not want to give the impression here that […]. (engecon08)
(5) I would like to contrast […]. (engling45)
(6) […] as I will try to show […]. (engling31)
(7) Secondarily I hope to have shown […]. (engling37)

The following are examples taken from the French subcorpus:

(8) […] je vais maintenant décrire […]. (frecon07)
‘I will now describe …’
(9) En conclusion, je tenterai de dégager quelques pistes de recherche. (frling08)
‘In conclusion, I will try to provide some research orientations.’

And the following from the Norwegian subcorpus:

(10) Jeg vil nå fokusere på […]. (noecon01)
‘I will now focus on …’
(11) Så langt har jeg forsøkt å forklare […]. (noling01)
‘So far I have tried to explain …’

The modalising elements of course constitute important aspects in the characteri-

sation of author presence, and consequently in the characterisation of the whole
article, especially the article’s “tone”. However, except for auxiliary verbs used to
form periphrastic tense constructions in all languages (have, be / avoir, être / ha,
være) and the future in English (will) and Norwegian (vil) (in French the simple
 Chapter 4. Presence of the author 87

future tense is made by adding an inflectional suffix to the infinitive form),17 the
number of such modalising elements is not very high. In English, excluding be,
have, will, shall, there are about 100 out of 1121 cases where some modalising ele-
ment precedes the “main verb”; in French, if we exclude avoir, être, aller, venir de,
the number is about 30 out of 197; and in Norwegan, if we exclude ha, være, skal,
vil, the number is about 100 out of 1212. Examples of such modalising expressions
are – in English: will like to (15 occ.) and want to (10 occ.), in French: voudrais
(‘would like’, 8 occ.) and aimerais (‘would like’, 2 occ.), and in Norwegian: kunne
(‘could’, 18 occ.) and the verb forsøke (‘try’) in different forms (13 occ.). Since the
number of such expressions is relatively low, we have not studied them systemati-
cally. Our main focus has been to describe the nature of explicit authorial voice
and activity within the article.
Finally, it should be noted that the various adverbial constructions found in
the immediate cotext of both pronoun and verb do not influence the verb clas-
sification as such, but they will be important in the interpretation of author roles,
especially the metatextual expressions.

4.2.2 Author roles and cotext

Let us now see to what extent there is correspondence between the verb groups
and the four proposed author roles. We first present some examples which are eas-
ily identified as one of the four roles.
• The writer role, manifested by the presence of a discourse verb:
(12) In Sections 3.1, 3.2 and 3.3, I describe each of the possible perspectives […].
(engling34)
(13) I shall return to this sequence later, […]. (engling20)

In (12) the verb describe assigns a writer role to the author by denoting a process
involving verbal representation (of some research, idea, etc.). In (13) the verb re-
turn (to) also assigns a writer role to the author, in this case concerning the struc-
turing of the text. We want to emphasise that the writer role is one that the author
carries throughout the article. In many cases, however, this role is not explicit (for
obvious reasons) or it is set aside in order for the author to assume another role
explicitly, expressed by a verb from one of the other groups.

17. There are 197 hits of I will in our corpus of 1121 occurrences of I, but only 74 of the corres-
ponding Norwegian expressions jeg/eg vil/skal of a total of 1212 jeg/eg. As regards French, there
are 40 future forms in the group of 197 je-occurrences.
88 Academic Voices

• The researcher role, manifested by the presence of research verbs:

(16) As before, I calculate welfare gains […]. (engecon06)
(17) […] I assume that this position is Spec, VP […]. (engling01)
(18) I conducted these interviews in Spanish. (engmed04)

In these examples the author is assigned the researcher role by typical research
verbs such as calculate, assume and conduct (interviews).
• The arguer role, manifested by the presence of position verbs:
(19) Yet, I argue that these on-line studies only examine the immediate activation
of individual word meaning […]. (engling22)

In the English articles, the most typical position verbs indicating the arguer role
are argue (as in (19)) and claim. The verb believe is another example:
(20) I believe that experimental economics has progressed […]. (engecon06)

When assigning the arguer role to the author, the verb believe is of course denoting
that the author has the opinion that something is true.
• The emotional or evaluating role, manifested by the presence of emotional
or evaluative verb constructions:
(21) […], I have been struck by the practical differences that separate present-day
econometrics and experimental economics, […]. (engecon30)
(22) But I am sceptical about extending iconicity to distance phenomena. (engling03)

In (21) the author is expressing an emotional reaction (be struck by), and in (22) he
is evaluating a phenomenon treated in the article (be sceptical about).
The above examples have presented no serious problems as regards the clas-
sification into the four different roles. This has to do with the relatively clear cor-
respondence between verb meaning and role activity. We shall now look at a few
examples that are less clear-cut.
First, one and the same verb may appear in different roles according to mean-
ing, to different syntactic constructions or to position in the article structure. The
verb conclude, for example, can be interpreted as a discourse or position verb. In
the construction I conclude by + -ing-form of the succeeding verb, we normally
classify it as representing the writer role (structuring the text), as in the following
example, taken from the introduction part:
(23) I conclude by outlining some further predictions […]. (engling45)
 Chapter 4. Presence of the author 89

In the construction I conclude that, which represents a way of arguing and taking
position, we may classify it as an example of the arguer role:
(24) I conclude that there is a clear correlation between […]. (engling03)

Other examples of clearly polyfunctional verbs are find and show. In (25), find +
the adjectival phrase slightly odd assigns the evaluator role to the author:
(25) I have always found this formulation slightly odd, […]. (engecon30)

When the verb is used in the construction find that, it may get the meaning of a
position verb and may thus assign the arguer role to the author, as in (26), taken
from the conclusion section:
(26) I found that when a model is formulated to allow for complex integration
strategies the issue of interdependencies across countries becomes important
to understanding the structure of FDI. (engecon44)

In this case one might also object that I found that … is a manifestation of the
researcher role; however, we think that the cotext of the utterance and its place in
the conclusion section constitute support for the interpretation of it as manifesting
the arguer role.
We will now look at two possibilities for the verb show. In (27) show assigns the
writer role to the author (an interpretation clearly supported by the metatextual
expressions).
(27) In Section 4 below, I show how force-dynamic relations constitute a sufficient
condition for dynamicity and in Section 5, I show how thematic relations
must be linked in order for the predicate to have a determinate aspectuality.
(engling39)

In (28), on the other hand, show in the construction show that may assign the ar-
guer role to the author:
(28) I show that two recent accounts of this are inadequate, both conceptually and
empirically. (engling49)

There is of course no clear-cut distinction between the two constructions show

how and show that, but there are two possible interpretations according to the
cotext in which they occur.
Discourse verbs typically occur in metatextual cotexts, especially in the out-
lining of the article structure presented in the introduction:
(29) In Section 2, I outline the model. (engecon44)
90 Academic Voices

When a verb normally understood as being of the research type occurs in a similar
metatextual cotext, we classify the whole sequence as a manifestation of a dual role,
the writer-researcher role. Example (30) is in fact the direct continuation of (29):
(30) In Section 3, I consider the model in a partial equilibrium setting […].
(engecon44)

The verb consider is a typical research verb corresponding to the researcher role.
However, through the influence of the metatextual expression (here: In section 3)
and the position in the introduction section presenting the structure of the article,
the author takes on a dual role. The roles as writer and researcher are to some ex-
tent blurred. When there is an explicit reference to a specific section (as in (30)) or
a deictic adverb like here or now also referring to a specific section in the article, we
classify the example as representing a dual writer-researcher role. The same holds
for position verbs combined with expressions referring to specific sections in the
article. Because of the metatextual phrase In this section, example (31) is classified
as being of the dual role type, in this case writer-arguer:
(31) In this section I will argue that there is no logophoricity condition on ration-
ale clause antecedence […]. (engling47)

Example (32), on the other hand, is interpreted as the arguer role only, despite the
presence of the metatextual expression In this paper:
(32) In this paper I argue that we should abandon a speech act theoretic approach
[…]. (engling05)

The reason is that the expression In this paper refers to the whole article; the pur-
pose here is not to outline text structure. It is more likely that the author wants to
position herself by making explicit (one of) the main claim(s). The role assumed
here is clearly the arguer role. The same holds for the combination of the same type
of metatextual expression with a research verb. If example (30) had taken the form
of (33), we would have classified the author role as researcher:
(33) In this article, I consider the model in a partial equilibrium setting …

The above considerations are presented in order to show that our four quite gener-
al roles need to be more fine-grained in some cases. We have, as just stated, added
two dual roles to our main classification. These are used when the verbs concerned
appear in cotexts containing metatextual expressions indicating functions which
are different from the function inherent in the meaning of the verb in question.
Finally, it should be noted that not all verbs combined with a first person sin-
gular pronoun fit easily into one of the four verb groups. For these we have estab-
lished a fifth, residual, group. However, the large majority of the verbs do in fact
 Chapter 4. Presence of the author 91

fall into one of the four groups. Of all the occurrences with a first person singular
pronoun, 2530 in all, only 63 are placed in the residual group.
There will always be a certain degree of subjective interpretation in this kind of
taxonomic work. However, in our view the results give a good indication of what
the authors are doing, i.e. through which voice they enter the polyphonic drama
they set up in their own text. The overall results are presented in Table 4.2-1, where
the percentages indicate the distribution of the roles within the total of 2530 first
person singular pronoun occurrences.

Table 4.2-1. Distribution of author roles with first person singular pronouns

Writer 26 %
Writer & Researcher 05 %
Researcher 46 %
Arguer 14 %
Writer & Arguer 02 %
Evaluator 04 %
Residual 03 %

Total occurrences 2,530

As can be seen from this table, the researcher role is the dominant one, constitut-
ing around half of all the first person singular pronoun occurrences (46 % plus 5 %
in the dual writer-researcher role). What may be more surprising is the relatively
prominent position of the writer role, with around one fourth of all the occur-
rences (about one third if the dual roles are included). As regards the arguer role,
the result may also be surprising, but the surprise goes in another direction than
for the writer role. Given the well-known competitive situation within research,
we might have expected a more explicit author manifestation by an arguer role
combined with the ‘I’ presence. In fact, the arguer role only represents about one
seventh of the total (14 % plus 2 % in the dual writer-arguer role).
The evaluator role is the least frequent one, and this result corresponds well to
what we expected. It represents only about 4 % of the occurrences, and these are
mostly found in a few articles. Thus, the role is relevant to the characterisation of the
tone in individual articles, but not to the general characterisation of research articles.
As regards the distribution of the three main roles, we conclude as follows:18
When academic authors use ‘I’, they use it first and foremost “scientifically”, i.e. to
make explicit their presence as researchers. However, they also use ‘I’ to a surpris-

18. This interpretation is based on the results from economics and linguistics; the use of ‘I’ in
medical articles is too modest to be considered here.
92 Academic Voices

ingly great extent interactionally, i.e. to indicate that they are present as writer,
or text organiser and readers’ guide. This is a sympathetic trait, if one may say
so talking about research writing. In a world full of stress and competition for
presenting original results, the authors take the time to tell their reader what to
find and where to find it in the articles. Finally, the authors use ‘I’ strategically to
a smaller extent than expected when taking into consideration the competitive re-
search world: the arguer role is relatively infrequent. This seems to imply that there
is still some reluctance to commit oneself to direct arguing, and that the tradi-
tional conception of scientific discourse as “neutral” is still quite strong. However,
the relatively small number of occurrences of the arguer role does not imply that
there is little argumentation in the articles. It just means that the author hesitates
to involve him- or herself in this form of direct commitment; a certain prudence
is favoured as regards the use of ‘I’. With a rhetorical conception of science, one
might in fact expect a certain balance between positioning oneself (strategic use of
‘I’) and inviting cooperation (interactional use of ‘I’).
We will now look at the results in more detail in order to identify similari-
ties or differences between languages and disciplines. In Table 4.2-2 we exclude
the three medicine subcorpora as well as French economics because of the low
number of articles using ‘I’. We only look at the three main roles, writer, researcher
and arguer, with the rest (evaluator, dual and residual) in an “other” category (see
Table 4.2-2).

Table 4.2-2. Distribution of author roles with first person singular pronouns in subcorpora

Subcorpus Writer Researcher Arguer Other Occurrences

engecon 21 % 56 % 7% 16 % 242
noecon 41 % 47 % 3% 8% 212
engling 24 % 41 % 21 % 14 % 872
frling 24 % 47 % 12 % 17 % 182
noling 28 % 47 % 12 % 13 % 991

For both discipline and language the researcher role is the most important; the
percentage varies from 41 in engling to 56 in engecon. The interesting differences
appear in the frequency of the writer and arguer roles. A first observation indicates
that Norwegians are more writers than arguers and that English authors argue more
explicitly than Norwegian ones. This is an interesting result, especially since we
knew very little about Norwegian scientific style at the outset of the KIAP project.
As regards English scientific style, earlier comparative studies (Mauranen 1993a;
Vassileva 2000) have found that it tends to be more explicit and direct than Nordic,
Romance and Slavic styles, and our results here point in the same direction.
 Chapter 4. Presence of the author 93

If we look at discipline tendencies, linguists definitely argue more than econ-

omists. The particularities to note here are that Norwegian economists use the
writer role (41 %) a good deal more than the others, and that English linguists use
the arguer role (21%) a lot more than the others. The French linguists are close to
the general percentage for all three roles; much the same is true for Norwegian
linguists (see Table 4.2-1).
We will now take a closer look at the verbs that are combined with the first per-
son singular pronoun in the different subcorpora. The verbs that are used explain
to a certain extent the frequency of the different roles attributed to the pronoun ‘I’.
We only list the top five verbs for each corpus, with the number of occurrences for
each, and with the total number of verbs (types) (see Table 4.2-3).

Table 4.2-3. Top five verbs used with first person singular pronouns

Corpus N Verb 1 Verb 2 Verb 3 Verb 4 Verb 5

engecon 98 consider 19 use 17 assume 16 focus 13 show 13

noecon 95 anta 23 diskutere 13 betegne 5 drøfte 5 presentere 5
‘assume’ ‘discuss’ ‘characterise’ ‘discuss’ ‘present’
engling 240 argue 82 show 48 suggest 37 assume 34 believe 31
frling 102 proposer 9 revenir 7 considérer 6 utiliser 5 constater 5
‘propose’ ‘return’ ‘consider’ ‘utilise’ ‘establish’
noling 249 me(i)ne 53 kalle 37 se/sjå 20 finne 20 bruke 20
‘think’ ‘call’ ‘see’ ‘find’ ‘use’
N = number of different verbs.

First, we note the large number of different verbs in all the subcorpora. Second,
there are not many verbs that appear “regularly” with the first person pronoun,
except perhaps for the top five in English linguistics. The table shows that argue is
used 82 times and show 48 times (out of 872 ‘I’ occurrences; see Table 4.2-2). We
also note that Norwegian mene is used 53 times in Norwegian linguistics (out of
991 ‘I’ occurrences). The table explains to a certain extent the high frequency of
the writer role in Norwegian economics (the verbs diskutere, drøfte and presentere
are typical discourse verbs), as well as the high frequency of the arguer role in Eng-
lish linguistics (the frequency of the verb argue is 82 and 31 for believe).
We will close this section by providing some typical examples of the three main
roles in the five subcorpora studied above:
• English economist as writer:
I summarize this result in the following proposition: […]. (engecon44)
94 Academic Voices

• English economist as researcher:

Therefore, I assume zero network costs in the foreign market […].
(engecon39)
• English economist as arguer:
In this paper, I show that this intuition is not true in general. (engecon06)
• Norwegian economist as writer:
Jeg skal nå diskutere noen mulige argumenter […]. (frecon20)
‘I will now discuss …’
• Norwegian economist as researcher:
Som utgangspunkt for analysen skal jeg anta at hjemlandet er kapitalim-
portør. (noecon45)
‘As a point of departure for the analysis I will assume that …’
• Norwegian economist as arguer:
Her skal jeg imidlertid argumentere for at […]. (noecon41)
‘However, here I will argue that …’
• English linguist as writer:
[…], first, I will present a non-quantificational grammar […]. (engling48)
• English linguist as researcher:
[…], I will assume Sells’ (1987) characterisation of logophoricity […].
(engling47)
• English linguist as arguer:
I argue that the strongest hypothesis is one which maintains […]. (engling45)
• French linguist as writer:
[…] pour simplifier je décris brièvement les procès conjugués […]. (frling30)
‘… in order to simplify I describe briefly …’
• French linguist as researcher:
J’envisage donc a priori la possibilité d’une relation entre […]. (frling13)
‘I envisage …’
• French linguist as arguer:
Par ailleurs, je propose des hypothèses explicatives concernant […]. (frling32)
‘Moreover I propose …’
• Norwegian linguist as writer:
Eg skal ikkje her gå utførleg inn på […]. (noling13)
‘I will not go into detail here …’
• Norwegian linguist as researcher:
Jeg opererer med en teoretisk definisjon […]. (noling14)
‘I use a theoretical definition …’
• Norwegian linguist as arguer:
Jeg vil hevde at et tredje synspunkt også er mulig, […]. (noling16)
‘I will claim that …’
 Chapter 4. Presence of the author 95

4.3 First person plural subjects

In the previous section, we discussed the various roles that the single author as-
sumes when referring to him- or herself with the first person singular subject
pronoun I – or je or jeg/eg. Most importantly, the author may be mentioned in
contexts where he or she can be seen as primarily filling one of the roles of writ-
er, researcher, arguer or evaluator. In these roles, the author engages to varying
degrees in interaction with the reader: telling the reader what research has been
done, making the reader aware of the structure of the text, trying to convince the
reader of his or her own views. In articles written by several authors, the pronouns
we, nous and vi/me are used for the same purposes.
However, first person plural pronouns are pragmatically much more complex
than the corresponding singulars. This complexity is reflected in the number of
epithets used to describe different kinds of ‘we’: inclusive we, exclusive we, autho-
rial we, royal we, generic we, coaxing we, etc. (cf. e.g. Loffler-Laurian 1980; Rounds
1987; Faarlund, Lie & Vannebo 1997: 323–326; Kuo 1999; Vassileva 2000; Rast-
all 2003; Fortanet 2004). The picture can be broken down into two main aspects:
questions of reference (who the pronoun refers to – it often borders on ‘one’ or ‘I’,
and sometimes even on ‘you’) and questions of role relations and rhetorical moti-
vations (how the choice of ‘we’ is influenced by the roles that the authors take on
and by what aims they attempt to achieve linguistically). We will start by sketching
the referential properties of first person plural pronouns (4.3.1), with some added
discussion of rhetorical functions. Next, we look at ‘we’ in combination with Eng-
lish find and French proposer (‘propose’) (4.3.2). We then go on to a discussion of
disciplinary and language differences and certain special uses of ‘we’ (4.3.3), before
ending with a note on ‘we’ and polyphony (4.3.4).

4.3.1 The reference and functions of first person plurals

Concerning questions of reference, we first need to acknowledge the fundamental

referential vagueness of ‘we’. Pronouns meaning ‘we’ typically refer to more than
one person, at least one of whom is the author(s) (or the speaker in spoken lan-
guage, but we will concentrate on written language here). That is, it includes the
author(s), but there is variation in terms of whether the reader is included and
whether others are, as well as who these potential others might be. The question of
reference is further complicated by the fact that ‘we’ is frequently used figuratively
(or, to be more specific, metonymically), especially referring to a single author
alone (‘we’ for ‘I’) or to the reader (‘we’ for ‘you’). This section presents a sketch of
the various referential possibilities, using selected examples from the KIAP Cor-
pus and drawing on existing literature, and we also discuss briefly some of the
96 Academic Voices

rhetorical effects of the various uses. As pointed out by Myers (1989), personal
pronouns are frequently used strategically, and this is particularly true for ‘we’.
Many authors have discussed the reference of first person plurals and em-
ployed classificatory systems to sort out the possibilities (cf. e.g. Loffler-Laurian
1980; Wales 1980, 1996; Mühlhäusler & Harré 1990; Fortanet 2004; Kinn 2005c)
One simple system is the one of Rounds (1987: 17–19), who distinguishes between
“traditional semantic mappings” and “semantic remappings”. One could alterna-
tively speak of literal versus metonymic uses of the pronoun. Such a distinction is
useful as a point of departure. As regards traditional semantic mappings, Rounds
recognises the distinction between (1) inclusive ‘we’, which includes the second
person(s), and (2) exclusive ‘we’, which does not, a dichotomy that is well known
from the literature.19 She further lists three semantic remappings, viz. ‘we’ used
about (3) a single speaker, i.e. ‘we’ for ‘I’, (4) ‘we’ used about the hearer(s), i.e. ‘we’
for ‘you’, and (5) anyone potentially involved in a certain kind of event, i.e. ‘we’ for
‘one’. In the following, we will first look briefly at literal examples of exclusive and
inclusive ‘we’ and variants of those, before turning to metonymic uses, where we
will concentrate on ‘we’ for ‘you’ and two kinds of ‘we’ for ‘I’.

Exclusive ‘we’
A typical example of ‘we’ referring to several authors and nobody else is given in
(1); it is clear that it is the authors that perform the emphasising and the arguing.
(1) Igjen vil vi understreke at vi ikke argumenterer imot reguleringer av dette
markedet. (noecon01; several authors)
‘We want to emphasise again that we are not arguing against regulations …’

A clear example of the inclusion of third (but not second) persons is provided in
(2). This is taken from an article with one author, and the referent of ‘we’ can rea-
sonably be taken to be the (medical) staff of the institution in question.
(2) Graden av reinnleggelse gir derfor ikke noe mål på kvaliteten av tilbudet vi
har gitt, […]. (nomed29; one author)
‘… the quality of the services that we have offered, …’

Somewhat less clear is (3), where the interviewers and examiners include the au-
thors, but where it is possible that more people have been involved in the research
process referred to. (The exact reference might be possible to determine from a
wider cotext.)

19. Many languages have separate expressions depending on clusivity, but the three languages
in this study do not distinguish between inclusive and exclusive subject pronouns. (However,
both English and French have specifically inclusive imperative constructions, and the ’s of En-
glish let’s-imperatives can be said to be an inclusive pronoun; cf. section 5.3.)
 Chapter 4. Presence of the author 97

(3) Vi benyttet oss av intervju og en fysisk undersøkelse av pasientene. (no-

med05; several authors)
‘We made use of interviews/an interview and a physical examination of the
patients.’

Inclusive ‘we’
The three examples above have in common that they clearly exclude the reader
from the referent. The opposite of this is the inclusive ‘we’, where the reader is
included. Several authors have discussed such pronoun use in research articles,
e.g. Kuo (1999) and Hyland (2001a); see also Breivega (2003: 146). Inclusive ‘we’ is
exemplified in (4), where what is had is data that the author and the reader share.
(4) I (7b) og (7c) derimot har vi et argument som uttrykker hvem som synger,
[…]. (noling49; one author)
‘In (7b) and (7c) [linguistic examples], on the other hand, we have an argu-
ment that expresses who is singing …’

As several authors have pointed out, such use of the inclusive ‘we’ is common in
many disciplines and is a way of involving the reader that has a strong rhetorical
potential (see e.g. Mühlhäusler & Harré 1990 (172–175) and Hyland 2001a). The
reader is (on the face of it) engaged in the research process and is assumed to
agree: What he or she has done, obtained and observed together with the author
will more likely be accepted as true, relevant and reasonable. We return to the rhe-
torical functions of reader inclusion in 5.3, in the study of ‘let us’-imperatives.
The ‘we’ in (4) does not include any third persons. In (5), however, the pro-
noun appears to refer to Norwegians in general (or, more properly, those Norwe-
gians who have a say in deciding the duty levels).
(5) Enkelte hevder at vi bør unngå et høyere nivå på norske avgifter enn det våre
konkurrentland har. (noecon43; several authors)
‘Some claim that we should avoid a higher level of Norwegian duties …’

This referent includes the authors and many others. Assuming that the reader is
Norwegian, he or she is also included, i.e. the ‘we’ is inclusive. But when the reader
is not Norwegian, the ‘we’ will be exclusive. This illustrates an important point: The
reference of ‘we’ is not fully determined by the author, but depends also on the read-
er’s identity and his or her text interpretation. The ‘we’ in (6) includes at least the
society of economists and approaches a generic reference where alternatively ‘one’
could have been used. More or less vague references to the discipline community is
an important function of inclusive ‘we’ (cf. also Myers 1989 and Kuo 1999).
98 Academic Voices

(6) Hovedpoenget mitt er […] å indikere hvordan vi kan benytte denne littera-
turen til å belyse hvorfor fattigdomsproblemet bør løses. (noecon20; one au-
thor)
‘My main point is … to indicate how we can use this literature …’
In Rounds’ (1987) system one or both of the last two examples could possibly be
classified as a remapping (‘we’ for ‘one’) rather than a traditional inclusive map-
ping; the distinction is quite hard to make.

Metonymic uses of ‘we’

Rounds (1987), in her system sketched above, is not very explicit about the rela-
tion between traditional (literal) mappings and (metonymic) remappings. But her
choice of the term remapping appears to indicate a kind of process where one map-
ping is changed into another one. Given such a remapping, an important question
is whether the original mapping remains. For instance, when an author uses a
pronoun meaning ‘we’ about the reader, is the referent only ‘you’, or is it ‘we’ and
‘you’ simultaneously?
In many discussions of the reference of first person plurals, relations between
speaker (author) and hearer (reader) are discussed, in terms of the participants’
roles in the context and of the rhetorical functions of pronoun use (cf. e.g. Rounds
1987; Mühlhäusler & Harré 1990; Hyland 2001a, b; Fortanet 2004). These issues
are especially central in connection with ordinary inclusive ‘we’ (see above) and
metonymic pronoun use. For example, Kuo (1999: 125) writes: “The fact that the
writer of a single-authored article uses we, instead of I, as he/she is referring to
himself/herself, may suggest an intention to reduce personal attributions.” The
crux here is not Kuo’s specific interpretation, but a more general point: For such an
interpretation to be at all reasonable (which we think it is), the pronoun must have
a double reference when used in such a way: a metonymic “real” one, viz. to ‘I’ (the
author), and one suggested by the literal meaning of the pronoun, viz. an exclusive
or inclusive ‘we’. And indeed, the term metonymy that we employ here presupposes
such a double reference.
The upshot of this is necessarily a referential system that is more complex than
most of those that have been proposed in the literature: When Wales (1996) discuss-
es metonymic uses of English we, she refers to them as “disguises” (p. 66), which is
another way of saying that there is one “real” referent hidden behind another. Thus,
the double referential classification that we assume is not strictly new, but is implicit
in much of the literature. The important point is that a classification based only on
the “real” referent, like the remappings of Rounds (1987), is not complete. When
‘we’ is used for ‘I’ or ‘you’, the question arises whether the source of the remapping
is exclusive or inclusive ‘we’. In the following, three metonymic uses are discussed:
inclusive ‘we’ for ‘you’, inclusive ‘we’ for ‘I’ and exclusive ‘we’ for ‘I’.
 Chapter 4. Presence of the author 99

When pronouns are used metonymically, then, the literal reference of the pro-
noun and the metonymic (“real”) one are at play simultaneously. The result is what
in cognitive science is called a blend (cf. Fauconnier & Turner 2002; Kinn 2005a). To
the extent that language users are at some (conscious or subconscious) level aware of
the blend and its interplay between different references, it has the potential of being
emotionally effective (cf. the discussion of the rhetoric of ‘let us’-imperatives in 5.3).

Inclusive ‘we’ for ‘you’

The ordinary inclusive ‘we’ that is used for the author(s)–reader dyad sometimes
blends into one used primarily to refer to the reader. A relatively clear example is
given in (7), where the authors are telling the reader to keep something in mind;
they themselves are obviously aware of it when writing. Of course, writing ‘you
must remember’ would have been highly unusual.
(7) Nivået på ledigheten […] var da […] likt nivået ved fullstendig koordinering
[…], men vi må huske på at det første tilfellet ikke bare er en situasjon uten
koordinering av lønnsdannelsen, […]. (noecon23; several authors)
‘…, but we must remember that the first case …’

The interpretation of ‘we’ in (8) is less clear. It can be argued that what the authors
are saying is ‘if you look at (≈ study) the table, you will see (≈ understand) that …’.
Whether the authors study the table or not, is rather irrelevant; they are telling the
reader that by doing so, he or she will acquire a certain knowledge. Again, it would
have been odd to write ‘as you see’, but that is the most likely interpretation.
(8) Som vi ser av tabellen avtar kostnaden signifikant ved økt barn/voksen-rate.
(noecon21; several authors)
‘As we see from the table, …’

As discussed by Rounds (1987: 21), this usage promotes a sense of cooperation

rather than making explicit the different role assignments. In (8), the authors point
to a piece of information and the reader observes it, but they are portrayed as fel-
low observers (cf. also Wales 1980, 1996; Mühlhäusler & Harré 1990).

Inclusive ‘we’ for ‘I’

An example that is in one way similar to those in the preceding section is given in
(9): There is little doubt that the subject pronoun here at one level, the literal one,
refers to the author–reader dyad, i.e. is an inclusive ‘we’. But at the metonymic, or
“real” level, the reference is to the author alone.
(9) Vi begynner med to NPer inneholdende substantivet čaëk ‘te (diminutiv)’
som attributt i genitiv: […]. (noling06; one author)
‘We begin with two NPs …’
100 Academic Voices

The verb ‘begin’ denotes an activity performed by the author, but the reader can
join the author in the process. In this case, then, it seems legitimate to assume that
we are dealing with an instance of inclusive ‘we’ for ‘I’.
Inclusive ‘we’ for ‘I’ has rhetorical functions similar to those of inclusive ‘we’
for ‘you’: It creates a sense of togetherness and blurs the author–reader divide, and
this community promotes agreement. As Mühlhäusler & Harré (1990: 175) point
out, the use of ‘we’ instead of ‘I’ also diminishes the responsibilities of the speaker,
since he or she is portrayed as collaborating with the hearer.
The example in (10) illustrates well the intricacies of reference. In addition to
the main verb få (‘obtain, get’) whose overt subject is ‘we’, there is an infinitival
construction with the verb sette (literally: ‘put, set’, but translated here by ‘letting’),
which is part of an adverbial.
(10) Ved å sette (5) = (6) får vi (10). (noecon41; one author)
‘By letting (5) = (6), we obtain (10) [an equation].’

The verb få in the present is typical of ordinary inclusive ‘we’; since the calcula-
tions take place in the text, author and reader are together in obtaining the results.
But this togetherness is less clear for sette: It is the author alone, in the role of re-
searcher, who equates the expressions, and the understood subject is arguably the
author, i.e. inclusive ‘we’ for ‘I’. But such an interpretation interferes with that of
the overt subject: For grammatical reasons, the understood subject of the infini-
tive is (or should be) coreferential with the subject of the main verb. Whether the
subject pronoun should then be regarded as an instance of ordinary inclusive ‘we’
or one of metonymic inclusive ‘we’ for ‘I’ remains a moot point. It is not at all clear
that it is possible to decide.

Exclusive ‘we’ for ‘I’

In the last usage to be discussed, ‘we’ is also employed for ‘I’, i.e. a single author
refers to him- or herself in the plural. An example is given in (11), where it would
have been possible to use ‘I’ instead of ‘we’.
(11) På bakgrunn av at […], argumenterer vi for at resultatene trekker i retning av
forsiktighetsmotivert sparing. (noecon37; one author)
‘On the background that …, we argue that …’

The verb ‘argue’ denotes an activity performed by the author where the reader is
given a different role, viz. as the one that the argumentation is directed at. It seems
reasonable, then, to assume that there is no reader inclusion here, i.e. it is an in-
stance of exclusive ‘we’ for ‘I’. This is what in English is called “authorial we” (see
Wales (1996); various other terms are also in use), and in French “nous de modes-
tie” (‘humble we’), cf. Loffler-Laurian (1980). We return to this usage in 4.3.3.
 Chapter 4. Presence of the author 101

Summing up
We have been at some pains to clarify the referential possibilities of first person plu-
ral pronouns and would not be surprised to learn that some readers disagree with
certain interpretations. One aim of the preceding sections has been to illustrate
the difficulties of hard and fast classification. It can be quite difficult to distinguish
between ordinary inclusive ‘we’, inclusive ‘we’ for ‘you’ and inclusive ‘we’ for ‘I’. The
distinction between inclusive ‘we’ for ‘I’ and exclusive ‘we’ for ‘I’ is also elusive.
Other authors have also pointed to difficulties in distinguishing between the vari-
ous uses (cf. e.g. Wales 1980: 36; Mühlhäusler & Harre 1990: 174; Kuo 1999: 134;
Fortanet 2004: 54). Notable exceptions are Hyland (2001a, 2001b) and Harwood
(2005), who have classified examples of we into inclusive and exclusive ones.
When we started investigating first person plural subjects, it was our aim to
classify the 10,782 examples in the KIAP Corpus on the basis of reference. We
wanted, among other things, to identify the instances where ‘we’ is used about
the author(s), in order to investigate the distribution of author roles in the same
way as for the first person singular in 4.2. As it turned out, the classification was
so fraught with problems that we decided to leave it for future studies. A possible
solution would have been to (try to) classify only on the basis of “real” reference.
Such a classification could have been aided especially by looking at the use of verbs
(cf. 4.2 and also Wales 1996: 66, who points to differences in typical verb use be-
tween inclusive and exclusive ‘we’). But as should be clear from the discussion
above, such a procedure would have meant excluding half of the reference for the
metonymic uses. Further, it will be clear that even the “real” reference can be quite
difficult to pin down. As Rounds (1987: 22–24) points out, the speaker (or, in our
case, the author) can use the referential vagueness of ‘we’ rhetorically. Exclusive
and inclusive uses blend together, and the inequality of roles of speaker and hearer
(or author and reader) can be as good as erased from language use.

4.3.2 ‘We’ in combination with two verbs

To illustrate in some more detail how ‘we’ is used in research articles, we will study
the co-occurrences of this subject pronoun with two verbs:
• English find in multi-author economics and medical articles
• French proposer in multi-author economics and single-author linguistics
articles

The verb find is normally used in connection with research findings, i.e. with
the researcher role (see 4.2) that we have established for ‘I’ in single-author arti-
cles, but it can also be used with an arguer role for the grammatical subject. We
102 Academic Voices

include all constructions with this verb, viz. find + noun phrase and find that +
clause, as well as some other semantically similar uses.
French proposer (‘propose, suggest, intend, aim’) is typically used in connec-
tion with suggestions for methodological or theoretical approaches, i.e. with an
arguer or researcher role for the subject. We include all constructions with this
verb. We find three patterns in the two subcorpora that we are looking at here:
proposer + noun phrase (‘propose, suggest’), proposer de + infinitival verb phrase
(‘propose to’), se proposer de + infinitival verb phrase (‘intend/aim to’). (There are
no occurrences of proposer que + clause (‘propose that’).)
There are 94 occurrences of find with a we-subject in the English multi-author
economics articles. Examples of non-modalised present tense (we find) account
for 83 of these, while there are only 5 occurrences of non-modalised past tense
(we found). (In addition, there are 6 occurrences with a modal verb.) Three of the
occurrences are given in (12)–(14).
(12) What happens as we move outwards from the centre, while staying within
zone I? We find the answer by totally differentiating ((10) and (11) [equations
already introduced]). (engecon36; several authors)
(13) We present a model of industry-level employment dynamics […]. We then
estimate econometric models […]. We find strong evidence that movements
in real exchange rates significantly affect gross job flows in U.S. manufactur-
ing. (engecon35; several authors)
(14) […] the model usually had difficulty closely matching the autocovariances of
output growth. In contrast to the results for output growth, we found no clear
pattern for inflation. (engecon22; several authors)

(12) is a typical example of inclusive we; the authors are demonstrating a methodo-
logical procedure and engage the reader in this activity. The use of the present tense
is typical of such examples. (13) is from an article introduction. The we in the first
sentence appears not to include the reader, since he or she is the receiver of the pres-
entation. The we in the second sentence could be of the referential type illustrated
in (12) and include the reader, but this is a less likely reading given the preceding
text. Since the finding referred to in the third sentence is something the reader has
yet to experience, it seems that this is an exclusive we. (14) is also from an article
introduction, i.e. the finding referred to is in the past for the authors, but not yet
demonstrated to the reader. The use of the past tense makes it clear that this is an
exclusive we. Thus, there is a connection between verb tense and pronominal refer-
ence: Inclusive ‘we’ tends to go with the present tense, exclusive ‘we’ with the past
tense. However, there is no straightforward relation between reference and tense.
 Chapter 4. Presence of the author 103

The picture with respect to the distribution of tenses is quite different in the mul-
ti-author medical articles. In this subcorpus, there are 54 occurrences of find with a
we-subject. 47 occurrences are in the past tense, 5 are in the present perfect (we have
found), and 2 are in the present tense. Three examples are provided in (15)–(17).
(15) Using SAS regression diagnostics, we found no evidence of collinearity.
(engmed24; several authors)
(16) Using this approach, we have identified a monosialylated oligosaccharide
(HPAgly1) […]. HPAgly1 was originally observed in a cell line that bound HPA
strongly (MCF-7), and we have found that it is over-expressed in excised breast-
cancer specimens that stain with HPA. (engmed08; several authors)
(17) A number of studies have also found increased mortality rates among peo-
ple who have undergone cataract extraction.3 5 6 9 10 13 Although we find an in-
creased mortality rate in univariate analyses for people who had undergone
cataract surgery, this finding was not statistically significant in our multivari-
ate analyses. (engmed37; several authors)

(15) is quite typical of this subcorpus. The research process (here the use of a statis-
tical method on data) is presented as a past activity. Hence, the verb is in the past
tense, and the pronominal reference is clearly exclusive. As can be seen from (16)
and (17), we can also be exclusive when the tense is present perfect or present.
Turning now to French proposer, there are 26 occurrences of this verb with a
nous-subject in the French multi-author economics articles. 25 of the occurrences
are in the present tense (nous (nous) proposons) and 1 in the present perfect or passé
composé (nous avons proposé). The situation is very similar in the single-author lin-
guistics articles, with 30 occurrences. 24 of these are in the present tense and 5 in the
passé composé. In addition, there is one occurrence of nous pourrions proposer (‘we
could propose’), i.e. with the modal verb pouvoir in the present conditional. Thus,
both economists and linguists mainly use proposer in the present tense, but the
passé composé is not infrequent, either. Four examples are given in (18)–(21).
(18) Nous avons proposé dans cette étude un modèle analytique d’équilibre général
[…]. Nous avons montré théoriquement que l’effet de la distorsion fiscale […]
était non intuitif, […]. (frecon41; several authors)
‘We have in this study proposed an analytic model …’
(19) En d’autres termes, est-ce qu’un comportement mimétique s’explique par la
connaissance détenue par les individus ? Pour répondre à cette question cent-
rale, nous proposons au préalable de préciser les principaux fondements con-
ceptuels de la notion de mimétisme rationnel, […]. (frecon20; several authors)
‘… we propose initially to make the principal conceptual foundations precise …’
104 Academic Voices

(20) Partant de la caractérisation du préfixe sous- comme un mot relateur de la

forme X R Y, nous proposons une hypothèse sur l’identité sémantique de sous-
que nous désignerons comme sa forme schématique8. (frling48; one author)
‘… we propose a hypothesis on the semantic identity …’
(21) L’analyse d’exemples classiques et d’un corpus oral nous a permis de dégager
quatre types d’emplois12 que nous nous proposons maintenant d’examiner.
(frling31; one author)
‘… four types of use which we now intend to examine.’

In these examples, nous is most naturally interpreted as exclusive, i.e. the reader
is not included. In the multi-author economics articles, this can be regarded as
an ordinary exclusive nous, referring to a group of authors. In the single-author
linguistics articles, however, we have examples of ‘we’ for ‘I’. This is the authorial or
humble ‘we’, in French called ”nous de modestie” (see 4.3.1 and 4.3.3).

4.3.3 Differences between languages and disciplines

Having illustrated the referential flexibility and vagueness of first person plural
pronouns, we will now address four issues in relation to the quantitative findings
presented in 4.1:
• the use of ‘we’ as such (irrespective of its reference)
• the use of inclusive ‘we’ when ‘I’ would have been more direct
• the use of inclusive ‘we’ when ‘you’ would have been more direct
• the use of exclusive ‘we’ when ‘I’ would have been more direct

The first issue concerns the explicit textual presence of the authors, which is the
overarching theme of this chapter. The three other issues concern metonymic uses
of first person plural pronouns.
We saw in section 4.1 that ‘we’ is used much more than ‘I’ in single-author
articles, except in English linguistics, where the difference between I and we is
negligible (cf. the figures for the economics and linguistics subcorpora in the col-
umns for median relative frequency in Tables 4.1-1–2). The plural was also seen to
be more frequent in multi-author than single-author articles, as is to be expected
(cf. the figures for engecon, engling, frecon, noecon and nomed in the columns for
median relative frequency in Tables 4.1-2–3; the remaining subcorpora either have
too few single-author or multi-author articles for a comparison to be tenable).
There is little doubt that the explicit presence of ‘we’ is a frequent type of author
manifestation in research articles, possibly the most frequent one.
As for disciplinary differences, we found that economics and linguistics have
about the same frequency of ‘we’ in single-author articles, while economics has
 Chapter 4. Presence of the author 105

much more use of ‘we’ in multi-author articles than medicine. The low frequency
in medicine corroborates the general picture emerging from our studies that this
discipline has fewer personal traces than the other disciplines. Although we have
not classified the occurrences of ‘we’ for their referential properties (for reasons
discussed in 4.3.1), it is our firm impression that a major difference between medi-
cine and the other two disciplines is that the former has very little use of inclusive
‘we’. This fits well with the low frequency of metatext (see 5.2) and ‘let us’-im-
peratives (5.3) in medicine; these are all reader-oriented features that are clearly
less commonly used in this discipline than in economics and linguistics. This also
implies, however, that when medical writers use ‘we’, they typically refer to them-
selves and not to the reader. The exclusive presence of the authors in medicine
is therefore not as weak compared to economics and linguistics as our statistics
might appear to indicate.
As for differences between the languages, Norwegian has more frequent use
of ‘we’ in single-author articles than English and French,20 while in multi-author
articles English and Norwegian are similar in having more occurrences of ‘we’
than French. In general, then, it is French that has the fewest occurrences of ‘we’.
This confirms the observation in Loffler-Laurian (1980) that the French avoidance
of ‘I’ (cf. sections 4.2 and 4.4) extends to ‘we’. It is interesting to note that Norwe-
gian linguistics has more use of ‘we’ (in single-author articles) than English, but at
present we have no satisfactory explanation for this. One possibility is that there is
more ‘we’ for ‘I’ in Norwegian; it is well known that the authorial ‘we’ is discour-
aged in English-speaking communities (see below), although as previous studies
have shown (see 4.0), it is definitely still used in various disciplines. There is less
focus on this in Norway, and we are not able to point to any general tendency to
encourage or discourage ‘we’ for ‘I’ (although there are certainly those who think
that ‘I’ should be avoided, as well as those who ridicule the use of ‘we’ for ‘I’). It
is tempting to speculate that a contributing factor could be that there is more in-
clusive ‘we’ in Norwegian than in English (in terms of the frequency of inclusive
use in itself, rather than relative to exclusive use), and that Norwegian authors
might be even more reader-oriented than their English-writing colleagues (cf. sec-
tion 5.2). Two related observations supporting such a view are, first, the fact that
the pronoun–verb combination ‘we see’ is 13 times more frequent in Norwegian
than in English single-author economics articles (Norwegian: 159 examples, or a
relative frequency of 78 per 100 000 words; English: 8 examples, or 6 per 100 000
words), and, second, that the combination is almost twice as frequent in Norwe-
gian as in English single-author linguistics articles (Norwegian: 262 examples, or

20. The difference between Norwegian and English is not significant in economics.
106 Academic Voices

101 per 100 000 words; English: 169 examples, or 56 per 100 000 words). Verbs of
seeing are typically used with inclusive ‘we’ (cf. Wales 1996: 66).
The second and third issue to be addressed, viz. the use of inclusive ‘we’ when,
in principle, ‘I’ or ‘you’ would have sufficed, need to be considered together. What
they have in common is that they treat author(s) and reader as collaborators with
the same roles in the activities referred to. The creation of common activities tends
to be centered around the text. For example, when the authors write ‘as we see
from the table’ in (8) above, they include themselves as participants in the visual-
cognitive activity that they want their reader to perform. This activity is directly
connected to the text. Further, when the single author writes ‘we begin with two
NPs’ in (9) above, he includes the reader as a co-researcher in his presentation
of linguistic data. Here, too, the text is present, as a space where research-related
activities begin, but the focus is on the data. Such textual orientation is much more
common in economics and linguistics than in medicine, and while medical re-
search is normally presented as an activity that belongs to the past, using the past
tense, the other two disciplines often have present tense (or related tenses, e.g.
the present perfect), connecting up with the textual here and now. Of course, the
creation of an activity with common roles for author(s) and reader is more natural
in relation to the textual present, and this is no doubt a major factor behind the
higher frequency of inclusive ‘we’ and (consequently) ‘we’ in general in economics
and linguistics than in medicine. As regards rhetorical aspects of inclusive ‘we’ for
‘you’ or for ‘I’, which we touched upon in 4.3.1, we will return to these in connec-
tion with ‘let us’-imperatives (section 5.3), since these phenomena have more to
do with author/reader interaction than with author manifestation as such. Suffice
it here to point to the textbook- and lecture-like character of such language use.
Finally, we need to consider the use of exclusive ‘we’ for ‘I’ (authorial or hum-
ble we, ”nous de modestie”). This usage is especially interesting in connection with
French, where the use of je (‘I’) has long been eschewed in academic discourse
(cf. sections 4.2 and 4.4) – although, as we have seen, the first person singular
is indeed used in French linguistics. When je cannot be used, the author has to
resort either to impersonal constructions (e.g. the passive) or use other pronouns,
above all nous (‘we’) (but also on (‘one’); see 4.4). Loffler-Laurian (1980: 138) writes
about this nous that “le locuteur-auteur fait mine de se fondre dans une pluralité,
parmi un nombre indéfini de personnes, de n’être qu’une sorte de porte-parole
anonyme – ou presque”.21 She regards this pronominal use, then, as a way for the
single author to make him- or herself less visible, seemingly speaking not on his
or her own behalf, but as a member of a (fictitious) larger group. Even though this

21. In English: “the speaker-author pretends to melt into a plurality, among an indefinite num-
ber of people, to be only a kind of anonymous mouthpiece, or almost so”.
 Chapter 4. Presence of the author 107

humble nous to some extent replaces the singular je in French, there is still less
frequent use of the first person plural in French than in English and Norwegian.
Clearly, then, French avoids the use of first person as such to a higher degree than
the other two languages.
The use of nous in combination with proposer in single-author articles that we
looked at above is an example of the humble nous. It should not be assumed, how-
ever, that all uses of nous in French research articles are of the exclusive type. In
(22), it is obvious that the single author does not only refer to his own seeing; the
reader is included. The same can be said about (23), which comes from a multi-
author article. Thus, the reader can be included in the reference of nous used in
single- as well as multi-author articles.
(22) Nous avons vu que les profits par tête et le taux d’intérêt s’expriment par les
équations : […]. (frecon43, one author)
‘We have seen that …’
(23) Nous allons voir qu’il peut toutefois être optimal pour le monopole de pro-
poser un rabais en t = 0, pourvu qu’il rationne la quantité mise en vente en
primeur. (frecon23; several authors)
‘We are going to see that …’

The inclusion of the reader in French first person plurals is further evidenced by
‘let us’-imperatives (see 5.3). Imperative expressions like montrons (‘let us show’)
indicate that even verbs that apparently exclude the reader by assigning a differ-
ent role to him or her than to the author – the author shows the reader something
– can be used with an inclusive ‘we’. This tells us that even the clearest cases of
exclusive nous above (e.g. (18)–(21)) may not be so clear after all: If the inclusive
montrons (‘let us show’) is used in a single-author article, how can we be certain
that nous montrons (‘we show’) in the same article is exclusive and not inclusive?
The various types of reference blend into each other, and it proves to be exceed-
ingly hard (if at all possible) to distinguish with certainty e.g. the ”nous de modes-
tie” from the inclusive nous.
Authorial or humble ‘we’ (exclusive ‘we’ for ‘I’) appears to be more commonly
used in French than in Norwegian and English. However, it is used to some extent
in these other languages too. Some examples are shown in (24)–(27).
(24) Vi vil deretter relatere de ulike teoretiske momentene til faktiske mobilitets-
kostnader i ulike OECD-land, og diskutere effektiviteten i landenes arbeids-
markeder i lys av dette. (noecon06; one author)
‘After that, we will relate the various theoretical points … and discuss the ef-
fectivity …’
108 Academic Voices

(25) Vi skal ikke avvise dette fullstendig, men hypotesen er heller ikke helt overbe-
visende. (noling24; one author)
‘We will not dismiss this completely, but the hypothesis is not quite convin-
cing either.’
(26) Although its effects are indirect […] we will argue in the conclusion that
they may be large. (engecon05; one author)
(27) Throughout the paper, when we refer to ”English”, we mean the spoken and/
or written language typically found in England or the United States. (eng-
ling30; one author)

It is our impression that the authorial ‘we’ is least commonly used in English. As
discussed by Wales (1996: 65), English authorial we has been subject to ridicule
and condemnation, and it has been called formal and old-fashioned. She writes
that some journals require authors to replace it with I or use the passive (but this
is evidently not the case for all journals, as the examples above demonstrate). The
usage is also sometimes discouraged in textbooks on academic writing, even vehe-
mently, e.g. in Day (1998: 210): “Do not use the ‘editorial we’ in place of ‘I.’ The use
of ‘we’ by a single author is outrageously pedantic.” Of course, the very observation
that there is a need to prohibit or discourage such usage shows that it is alive and
fairly well, and Wales (1980: 28) observes that in academic registers it is “extremely
common” and “far from being ‘old-fashioned’”, a fact made abundantly clear by
the findings of Hyland (2001b) and Harwood (2005). As noted there and in Wales
(1996: 65–66), the authorial we and the (literal and metonymic variants of) inclu-
sive we of instructional language use (lectures, workshops) tend to blend together,
possibly allowing the former to regain popularity. Of course, this blending under-
scores our observation that a classification of examples according to reference is
extremely problematic.

4.3.4 ‘We’ and polyphony

As noted in 4.0, the use of ‘we’ opens up for polyphonic plays of considerable com-
plexity. The complexities of the first person plural will no doubt have become clear
from the preceding sections. Of course, the main problem is to determine who is
meant by ‘we’. What reference is attributed to the pronoun has consequences for
the interpretation of the content of the points of view as well as the relation of
these points of view to the discoursal beings that are their sources. In 4.3.1, we did
not systematically discuss how it can be decided who ‘we’ refers to. In ScaPoLine it
is assumed that the polyphonic structure provides instructions for the interpreta-
tion of utterances. In the case of ‘we’, the default instruction is to assume a referent
 Chapter 4. Presence of the author 109

of more than one person, one of whom is the locutor. Who else is implied must
be decided by means of other instructions from the cotext or context. Let us look
at the passage in (28), with no less than nine occurrences of we – the text has two
authors.
(28) We1 turn now to the functional aspect of the protoscene in Fig. 4, namely the
claim that the TR and LM are within each other’s sphere of influence. A con-
sequence of being within potential reach of the LM, is that the TR can affect
the LM in some way and vice versa. For instance, […], we2 conventionally
understand power and control being associated with an entity who is higher
than the entity being controlled (we3 will discuss this in more detail when we4
deal with the control sense for over). In physical terms we5 can only control
someone or something […] if we6 are physically proximal to the entity we7
seek to control. If, then, […], control, […], is dependent upon being higher
than and physically close to the entity we8 seek to control, we9 would expect
that these notions can be designated by over but not above. (engling17; sub-
script numbers added)

The authors here employ several kinds of ‘we’, e.g. occurrences 8 and 9, used in the
same sentence, evidently have different referents. This text refers to events in three
different spheres: the text, the research process and human existence in general,
and a ‘we’ is involved in each of them. We will not discuss occurrences 2 and 5–7
more than to point out that we there refers to human beings in general (a very wide
inclusive ‘we’).
Occurrences 1, 3 and 4 are related to the textual sphere. Instructions to that
effect come from expressions referring to the utterance context and to the cotext:
now, turn to, Fig. 4, discuss etc. This means that the relevant place to look for refer-
ents is the utterance situation. Since this is a written text, it includes both the writ-
ing and the reading, and the relevant participants are the authors and the reader.
Since engling17 has two authors, we can refer literally to them, but it can also
include the reader. To decide between an exclusive and an inclusive reading, we
have to look for additional instructions, primarily those of verbs, viz. turn (to), dis-
cuss, deal (with). Who performs the turning, discussing, dealing? In literal terms,
it is the authors; i.e. we is exclusive. But as seen in 4.3.1, ‘we’ is frequently used
metonymically. The fact that the authors could very well have written “Let’s turn
now to …” shows that reader inclusion is possible at least for occurrence 1. This we
could, then, be an example of a type not discussed in 4.3.1, namely inclusive ‘we’
for exclusive ‘we’!
Occurrence 9 is situated in the research sphere. It is part of a long sentence
of the form “If p, then we would expect q”, p referring to human experience in
general (functioning as a premise) and q to a state of affairs in the primary field of
110 Academic Voices

study, viz. prepositional semantics. The sentence is a formulation of a prediction,

q. Instructions pointing to the research sphere (and away from the sphere of hu-
man experience in general) come primarily from the combination of the epistemic
verb expect and the proposition q, which pertains to the research field. The modal-
ised expression would expect locates the epistemic attitude in a generic time-space,
which includes the present as a central point in time. This is compatible with the
tendency of linguistic research articles to portray the research process as going on,
so to speak, as the text evolves, which implies that the research situation is (figu-
ratively) the same as the utterance situation, i.e. the participants are the authors-
cum-researchers and the reader. All participants in the situation have access to the
same information and may therefore be assumed to be co-responsible for the con-
clusions drawn on the basis of it. The authors have just argued, in the two previous
sentences, that p is the case; hence they are responsible for the point of view that
p is true. Further, they are responsible for the pov that ‘if p, then probably q’, and
therefore also the pov ‘probably q’. But since the reader is also made part of the re-
search situation, it would appear that he or she is also responsible for these points
of view. The modalisation of would expect further generalises the whole context,
so that the source responsible for the points of view might not only include the
authors and the reader, but any linguist. That is, the source is ‘authors and col-
leagues, including reader’. (This use of we resembles closely the use of French on
called ON3 in 4.4.) It is evident that the polyphonic drama set up by the author(s)
is central in the rhetoric of the research article, in that points of view are attributed
not only to the authors and other researchers mentioned in bibliographical refer-
ences, but to the reader and to the disciplinary collective of researchers.

4.4 “Indefinite” authors

4.4.1 Indefinite pronouns in English, French and Norwegian

Like many others we claim that first person pronouns represent the most explicit
author presence, whatever the discoursal genre is. However, in our studies of au-
thor presence in research articles, we realised already in a pilot study (see Breivega
et al. 2002) that the differences between English articles, on the one hand, and
French articles, on the other, are so great that we found it necessary to investigate
the extent to which other possible personal constructions are used in French. It
became obvious that, for the French corpus, we had to take the indefinite pronoun
on into consideration. For comparative reasons, we decided to study the frequency
of the corresponding pronouns in English and Norwegian as well, viz. one and
man, en/ein respectively.
 Chapter 4. Presence of the author 111

Another and equally important reason for including the indefinite pronoun
in the present study is that it reflects the polyphonic drama in a particular way, at
least in French. It can include (and exclude) the voices of the author(s), the read-
ers and third persons (corresponding to ScaPoLine’s locutor, allocutor and third
discoursal beings; see 2.1) as well as an indefinite community of persons (corre-
sponding to the collective third). Since the reference of these pronouns typically
has to be determined by the co(n)text, they represent a very specific means for hid-
ing as well as for bringing forth different voices. On the one hand, we can say that
the pronoun is very useful to authors who want to hide; on the other, it is useful to
authors who want to bring in the voice of a larger community, without having to
be specific about the extension of this community.
Even if these indefinite pronouns, one, on, man, en/ein, are not directly compara-
ble and exhibit quite different uses in the three languages, all of them can be used with
general as well as with specific and personal reference.22 For this reason, we think it is
interesting to study their actual frequency in the empirical material under investiga-
tion. As shown in section 4.1, French researchers are the dominant “indefinite” users,
but Norwegians are not far behind. This is an interesting result since, as far as we
know, no large empirical study of this phenomenon exists for Norwegian. As regards
the use of one as an indefinite pronoun in the English subcorpus, our study shows that
English authors are relatively modest “indefinite users”. Our findings correspond to
a great extent with Johansson’s (2002) results. He finds that German man is far more
frequent than the corresponding pronouns in English and Norwegian. As regards
frequency, Norwegian man seems to take a middle position between German man
and English one. Johansson also argues that “German man seems to be more compat-
ible with specific reference than the English and Norwegian generic pronouns” (ibid.:
527). This might be explained by a different semantic value of English one compared
to French and Norwegian, or perhaps by the relatively frequent use of I and we in
English. However, according to the material studied here, the latter explanation does
not hold, as the Norwegian subcorpus also contains many first person pronouns.
Before going on to what will be our focus here, the French pronoun on, we
will take a quick look at what English and Norwegian grammars say about these
indefinite pronouns.
In A Grammar of Contemporary English, Quirk, Greenbaum, Leech & Svartvik
(1972) have, in fact, little to say about one:
indefinite one means ‘people in general’, in particular with reference to the speak-
er. This use of one is chiefly formal and is often replaced by the more informal you:
One would/You’d think they would run a later bus than that! (Quirk et al. 1972: 222)

22. Johansson (2002: 516) also argues that the referential possibilities of English one, German
man and Norwegian man are comparable even if there are great differences in use.
112 Academic Voices

Biber, Johansson, Leech, Conrad & Finegan (1999), in their Grammar of spoken
and written English, argue that when referring to people in general, English one
has no “personal overtones” (ibid.: 331). However, they point to examples of one
with specific reference used in academic discourse, where one may convey “a veiled
reference to the author; this is characteristic of the impersonal style often adopted
in academic prose” (ibid.: 354). They conclude as follows:
The relatively high frequency of generic one in academic prose should be com-
pared with the unexpectedly high frequency of we […], and with the very high
passive frequency in this register […]. These are all connected with the preoccu-
pation in academic work with making generalizations and with the wish to adopt
an impersonal, objective style. (Biber et al. 1999: 355)

As regards the Norwegian indefinite pronouns man and en/ein, very few stud-
ies have been undertaken on these pronouns. An exception is Lundeby (1996),
who refers to a discussion from the beginning of the 1950s on the use of what in
Norway is called “departemental man/en” (i.e. being used in the ministries), with
reference to an institution or office, and where the advice is to use the plural vi
instead of the indefinite pronoun in such contexts. The most comprehensive Nor-
wegian grammar, Norsk referansegrammatikk, by Faarlund et al. (1997), states that
the indefinite pronouns can be used with reference to the sender. The descriptions
of Norwegian man and English one are thus more or less parallel. Both pronouns
are classified as generic or indefinite, but the grammars seem to accept the use of
these pronouns with specific reference.
The need for studies on indefinite pronouns in scientific discourse genres is
obvious.23 In section 4.1, we observed many interesting differences as regards the
frequency of these pronouns, and it seems reasonable to think that the differences
are related to the different referential values the pronouns can assume. For exam-
ple, the disciplinary differences in the Norwegian subcorpus are not very great
(not even significant) when compared to the disciplinary differences in the French
subcorpus. A possible explanation of this observation might be that the general
reference is predominant in Norwegian; thus, the use of Norwegian man, en/ein
does not reveal any differences of frequency between soft and hard disciplines. The
opposite might be the case in French. The pronoun on is quite frequently used with
a personal reference; thus we may expect it to be more frequent in linguistics than
in economics and medicine. This is also the case in our corpus (see 4.1). However,
we will see below that this cannot be the whole explanation. The personal values of
on are in fact quite frequent in French medicine.

23. For previous KIAP-studies on French on, see Fløttum (2003b, 2003i). For studies of on in other
genres, see Norén (2004a); Jonasson forthcoming; Fløttum, Jonasson & Norén forthcoming.
 Chapter 4. Presence of the author 113

4.4.2 Previous studies on the French indefinite pronoun on

Defined as referring to an indefinite group (‘ensemble’) of humans, the French

pronoun on is traditionally classified as indefinite. On is also considered to be a
rival to nous (‘we’), but can, in fact, replace any personal pronoun.
The reference of on is notoriously elusive. It is this elusiveness which seems to
enable it to function as a substitute for all personal subject pronouns (see Riegel,
Pellat & Rioul 1994: 197). In addition to Riegel et al. (ibid.), other recent grammars
tend to classify it as a personal pronoun, like Arrivé, Gadet & Galmiche (1985).
However, Wilmet (1997) continues to classify it as an indefinite pronoun. This
disagreement between grammars may be explained by the referential complexity
of the pronoun; it demonstrates personal as well as indefinite uses.
In more specific studies, different aspects are treated. However, in addition to
the fact that on always has the grammatical function of subject, most seem to agree
on the following characteristics of on:
• On refers to one or more persons.
• On can be substituted for all personal pronouns.
• On has an unclear enunciative status (i.e. relation to speaker or locutor
and receiver).

Muller (1979: 70) ends up pointing to three uses of on: indefinite use (on = on),
stylistic use (on = je, tu, nous, vous) and personal use (on = nous). We note that he
does not include third person pronouns, il/elle, ils/elles, in his classification.
More relevant to our work is Loffler-Laurian’s (1980) study of the presence of
the locutor represented by the pronouns je, nous and on in scientific discourse, in
this case texts taken from chemistry and physics. Taking as her point of departure
the well-known French cultural maxim “le moi haïssable” traditionally taught in
French schools, she argues that, in general, French researchers avoid the first per-
son singular, use the first person plural as little as possible, and therefore have to
resort to all kinds of impersonal, reflexive and passive constructions (ibid.: 135).
Among the five values of on that she proposes, it is the on for nous which seems
to be characteristic of Loffler-Laurian’s corpus. We also note her observation of on
changing reference according to the verbs it is combined with. This point will be
important in the classification proposed here (4.4.3).
The aim of François (1984: 34) is to propose a classification of the enunciative
values of on (taking into consideration referential, deictic, discoursal and stylis-
tic criteria). He proposes five different values, forming a scale, ranging from on
referring to some inanimate being (like in On était en juin, corresponding to an
impersonal construction in English: ‘It was June’) to on referring to some specific
human being, via uses with more or less vague reference.
114 Academic Voices

Atlani (1984), in her study of on in journalistic texts, does not want to reduce
the pronoun neither to an indefinite nor to a personal pronoun. We agree that it
is important to insist on the heterogeneity of the pronoun in order to capture its
different uses. We also agree with the importance Atlani attributes to the interpre-
tation process of on. Always subject to interpretation, on represents a particularly
complex discoursal phenomenon.
In his master’s thesis from 1984, Larsen focuses on the relation between the
pronouns on and nous (‘we’). Larsen’s basis for the semantic and distributional op-
position between these pronouns is a corpus constituted by a philosophical essay
(Genèse, by M. Serres). The 354 occurrences (231 nous and 123 on) are classified
according to the verb they combine with. The verbs avoir, devoir, être, penser and
savoir are preferred with nous, while on seems to co-occur with verbs like dire and
pouvoir. As regards expressions like dit-on (‘one says’) and comme on dit (‘as one
says’), Larsen believes that they are undergoing a lexicalisation process. (ibid.: 60).
One of Larsen’s conclusions is that on shows a clearer tendency to combine with
the abstract than nous does.
In a study based on a corpus of spoken language (14–15 year old Parisian ado-
lescents), Viollet (1988) proposes four values of on, which are clearly related to the
corpus studied (ibid.: 69): two personal (deictic and anaphoric), one non-personal
(doxa) and one undecidable.
Like many other researchers, Rey-Debove (2001) describes on in relation to
personal pronouns. However, she is particularly concerned with the case of the
pronoun nous. In two relatively strong hypotheses, she claims that 1) on tends to
replace nous, and that 2) on = nous progressively erases the real indefinite use of
on (ibid.: 280).
Further, Rey-Debove considers the verbal mode and tense as important for the
disambiguation of on; she argues that, in general, it assumes its greatest extension
(generic and often deictic) in combination with a verb in the present tense.
In her morpho-syntactic study of “the double play of the pronoun on”, Blanche-
Benveniste (2003) begins with the observation that even if on may have very dif-
ferent kinds of references, French-speaking people are very seldom confused by its
use. The interpretation of on is not problematic to native speakers. However, she
continues, this pronoun is particularly complex: It may both include and exclude
the author (or locutor). In her comparison between nous and on, she concludes
that the typical “composition” of nous is ‘moi + non-moi’ (‘I + non-I’) and that
the composition of on may, but does not have to, include ‘moi’ (the locutor) (ibid.:
49–51). We also note the following observation: “on s’oriente vers tous les humains
placés dans les circonstances que mentionne l’énoncé” (ibid.: 46). This is a general
but important instruction for the interpretation of on: On orients itself towards all
the humans placed in the circumstances mentioned by the utterance.
 Chapter 4. Presence of the author 115

In her master’s thesis, related to the KIAP project, Gjesdal (2003) studies on
and its use in linguistics articles. The results obtained in the comparison between
two corpora (32 articles from c. 1980 and 20 articles from c. 2000) do not confirm
Rey-Debove’s (2001) hypothesis that on is replacing nous (see above). In fact, Gjes-
dal observes that the relative frequency of on is lower in the 2000 corpus than in
the 1980 corpus. On the other hand, she finds that the use of je has increased (to
a statistically significant extent) between 1980 and 2000. These are observations
which are limited to the genre under study, i.e. the research article, and should of
course not be taken as a general development of the French language. They are,
however, interesting findings. In the qualitative part of her study, Gjesdal intro-
duces various rhetorical functions contributing to the disambiguation of on. She
also observes considerable differences as regards the articles’ individual styles in
their use of the pronouns je, nous and on.
To conclude, the previous studies of on show that it is necessary to undertake
more empirical analyses of the phenomenon in question. The need for studies that
go beyond the analysis of on versus nous is also obvious. In the next section we will
study the pronoun in its diversity, as manifested in research articles. Its potential
referential variation, also emphasised in the previous studies, will constitute the
point of departure for our studies. Our hypothesis is that the genre in which on
manifests itself will explain, to a large extent, the kind of heterogeneity it displays.

4.4.3 Six values of on in French research articles

Based on previous – more or less empirical – studies (see above), and our knowl-
edge of the research article, we formulate the hypothesis that the pronoun on will
take on a series of different values in this specific genre. Its fuzzy contours and
inherent indefiniteness seem to make this pronoun a perfect means of expres-
sion for a genre which would like to be objective and non-expressive. Whatever
its reference might be, it never refers directly to the author. Even when it can be
replaced by the pronoun je, it will always maintain a flavour of indefiniteness in
its meaning.24
The central challenge is perhaps not so much its interpretation in a specific
context as the identification of criteria which can explain these interpretations.25
In order to understand its functioning, we have to disambiguate on. Our main aim

24. The play between the personal and the indefinite allows us to consider the use of certain
values of on as a metonymic relation. When on refers to the speaker, for example, there is a me-
tonymy from ‘people in general’ to je (see our discussion of ‘we’ in 4.3).
25. We remind the reader of the observation made by Blanche-Benveniste (2003: 43). Accord-
ing to her, the use of on very rarely causes confusion for native speakers of French.
116 Academic Voices

in this section is to add qualitative explanations of the figures presented in 4.1; we

will identify and describe relevant criteria which may explain different values of
on, on a scale from indefinite to personal.
Even if we insist on the importance of the genre for the determination of rel-
evant criteria, there are certain criteria which seem valid for any text genre:
• verb tense; we consider mainly the present tense (présent, abbreviated as
PR), the present perfect (passé composé, PC) and the future (futur, FUT)
• the presence of modal verbs
• the presence of adverbial complements

Other criteria are more clearly related to the genre under investigation. Thus, for
the research article, the following seem relevant:
• the presence of metatextual and deictic elements (like dans cet article, au
tableau suivant, ici, maintenant, etc.)
• the meaning of the verb with which on is combined (like in on analysera,
with a research verb)
• the presence of bibliographical references (like on affirme (Duval 1998) que …)
• the presence of discipline-specific expressions

These criteria are mostly lexical-semantic, except for verb tense. As regards mor-
pho-syntactic criteria, like adjective and participle agreement (in feminine and/or
plural forms, such as in the example On est allées à l’école, where the participle
form (allées) shows that on refers to a group of females) as well as the presence of a
direct or indirect pronominal object, these are in general not relevant to our genre.
The reason is that such constructions are rare or do not occur at all in research
articles.
Our general hypothesis is that on is more personal and less indefinite in cotexts
where the following traits are present:
• metatextual elements
• verbs in future (FUT) or present perfect (passé compose, PC) tense
• verbs referring to the research process or to the text organising

The following constructed example contains all of these features, and on can be
interpreted as referring to the author (corresponding to je):
(1) Dans cet article, on examinera / a examiné …
‘In this article, one will examine / has examined …’

We consider the semantics of the verb as particularly important. As regards verb

classification, which is notoriously difficult, we refer to section 4.2, where we de-
fine discourse verbs (example: présenter), research verbs (example: analyser) and
 Chapter 4. Presence of the author 117

position verbs (example: défendre ‘defend’). In this section, we have found it use-
ful to refine the category of research verbs. This gives us three additional groups:
cognitive verbs, referring to a mental process (examples: comprendre ‘understand’,
concevoir ‘conceive’); reporting verbs, referring to the act of uttering (examples:
énoncer, dire ‘say’) and perception verbs, referring to different types of perception
(examples: entendre ‘hear’, voir ‘see’). Our final group is the ragbag group other
verbs, where we, inter alia, put emotional, evaluative and volitional verbs. These
verbs may signal different author roles, as is the case with first person pronouns
(see section 4.2). As in that study, we emphasise that it is the main verb (defined in
a broad semantic-pragmatic sense) that we consider. This means that in construc-
tions like on voudrait analyser (‘one would like to analyse’) or on tentera d’identifier
(‘one will try to identify’) it is the verbs analyser and identifier respectively which
are the basis for the classification. However, modal verbs are taken into account
when it comes to determining the value of on.
The considerations presented above allow us to propose six values that the
pronoun on can have in a French research article:

Table 4.4-1. Values of the French pronoun on in research articles

ON values referring to corresponding to

ON1 author(s) je (‘I’) / nous (‘we’)

ON2 author(s) + reader(s) je/nous + vous (‘you’, the reader(s))
ON3 author(s) + limited je/nous + vous (‘you’, my/our col-
discourse community leagues)
ON4 author(s) + non-limited je/nous + tout le monde (‘everybody’)
community
ON5 reader(s) vous (the reader(s))
ON6 other(s) il(s)/elle(s) (‘he/she/they’,
the other researcher(s))

In a polyphonic perspective, the pronoun on is similar to ‘we’ in many respects.

As with ‘we’, the polyphonic play is obvious. By the use of on, points of view are
attributed to the voices of a wide range of persons, or groups of persons, partici-
pating, or invited to participate, in the polyphonic drama set up. More specifically,
we may say that ON1, ON4 and ON6 represent single dimension voices in that
they refer only to the author(s), only to the reader(s) and only to third persons
respectively. ON2, ON3 and ON4 are multiple dimension voices: They refer to the
author(s) + reader(s) in ON2, the author(s) + colleagues in ON3 and the author(s)
+ ‘everybody’ in ON4.
118 Academic Voices

As regards the personal versus general distinction, ON1, ON2, ON5 and ON6
are more personal (and specific in their reference) than ON3 and ON4; ON3 and
ON4 are more complex, referring to indefinite groups or groupings of human be-
ings. In ON1–ON4 the author is included; in ON5 and ON6 the author is not
included. These values correspond in the following way to the self- and other-
dimensions employed in different contexts in this book: self: ON1; self & other:
ON2, ON3, ON4; other: ON5 and ON6.
In the following presentation of the different values of ON, we indicate which
criteria we consider as the most relevant to each type. We emphasise that the genre
itself is considered as a general criterion, relevant to all values.
ON1  referring to the author(s)
Criteria which are particularly relevant to the interpretation of on as referring to
the author(s) are the presence of metatextual or deictic elements, the verb tense
being FUT or PC, and the verb being of the discourse or research type, as in the
following, constructed, examples:
(2) On illustrera ici … / Dans cet article, on a comparé …
‘One will here illustrate … / In this article one has compared …’

Here are some authentic examples:

(3) On partira ici de deux formulations de cette hypothèse, […]. (frling26)
‘One takes as a point of departure two formulations of this hypothesis …’
(4) Dans un premier temps (par. 1), on présentera une liste de caractéristiques,
[…]. (frling06)
‘First (paragraph 1), one will present …’
(5) Dans ce papier, on se propose précisément de developper un test simple […].
(frecon25)
‘In this paper, one proposes, precisely, to develop …’

All these examples contain deictic or metatextual expressions (ici, Dans un pre-
mier temps (par. 1), Dans ce papier); the verbs are of the discourse or research type
and in the future tense, except for propose in (5), which is the present (PR) tense.
However, the presence of the metatextual Dans ce papier leaves no doubt about the
interpretation of on as referring to the author in this example.
In articles reporting on experiments or on the research process itself, we also
find on with this value when combined with verbs in the present perfect (PC); note
the adverbials dans un premier temps (‘firstly’) and ensuite (‘then’):
(6) Dans l’étude, on a utilisé ces données pour déterminer dans quelles propor-
tions l’analyse longitudinale remettait en question les estimation habituelles.
 Chapter 4. Presence of the author 119

[…]. Dans un premier temps, on a comparé les estimations en coupe trans-

versale avec celles issues de l’enquête sur les budgets familiaux. On s’est ensuite
intéressé à deux variantes […]. (frecon01)
‘In the study, one has used … Firstly, one has compared … then one has taken
an interest in …’
ON2 – referring to the author(s) and the reader(s)
A criterion which is particularly relevant to the interpretation of this value is the
semantic nature of the verb. For the pronoun on to be interpreted as referring to
sender (author) and receiver (reader), i.e. je/nous and vous, the successive verb
must refer to an action that at least two people can do together. Cognitive and
perception verbs are typical in this context. The verb tense may vary, but it is often
FUT or PC. In addition, there is often a metatextual element in the immediate
cotext. The following are constructed examples:
(7) On notera / remarquera …
‘One will note / will remark …’
(8) (Comme) on l’a vu (dans le tableau 1) … / Comme on va le voir …
‘As one has seen (in Table 1) … / As one will see …’

Authentic examples include:

(9) Comme le montre le tableau III, […] on remarquera qu’elle [l’évolution de la
valeur prismatique préscrite] n’a jamais dû être augmentée […]. (frmed47)
‘As table III shows, … one will remark that …’
(10) Comme on le verra par la suite, l’estimation du modèle […]. (frecon25)
‘As one will see in the following …’
(11) Les exemples [27] […] et [32] […] seront soumis aux six critères introduits
dans la section 1. On constate en premier lieu que […]. (frling06)
‘The examples … will be submitted to the six criteria introduced in section 1.
One observes …’

Example (12) contains the quite frequent expression comme on l’a vu (‘as one has
seen’) where the perception verb is in the PC tense. The author includes the reader
by referring to something already presented.
(12) Le facteur individuel incorpore en effet les orientations du couple en mat-
ière de descendance et, comme on l’a vu, celles-ci se traduisent probablement
dans les choix du ménage en matière d’habitation ou de biens durables, […].
(frecon01)

The message is as follows: “as we – you and I together – have seen here”.
120 Academic Voices

Especially in economics articles, many equations are introduced. The author

often invites the reader to participate in the reasoning or calculating process by
expressions like on obtient (‘one obtains/gets’) or on a (‘one has’):26
(13) Nous prenons dorénavant le taux de rémunération du travail comme numérai-
re, ainsi #. En substituant les quantités produites à l’équilibre dans l’expression
(2), et en dérivant cette expression par rapport au temps, on obtient le taux de
croissance […]. En remplaçant # par sa valeur et en reprenant l’équation (5), si
la recherche est active dans le secteur l, on a : (10) #. (frecon05).
‘By substituting … one obtains … By replacing # … one has …’

ON3 – referring to the author(s) and a more or less defined limited community
Criteria which are particularly relevant to the interpretation of on as referring to
the author(s) and a limited discourse community constituted by a more or less de-
fined group of colleagues include the following: verb tense being PR or main verb
being preceded by a modal verb (especially a form of pouvoir (‘can/may’)) and the
presence of specialised vocabulary in the immediate cotext. Since the verb may be
of very different types, but often of the research, cognitive or perception type, a
model example could be one with the verb analyser:
(14) On analyse (généralement) / On peut analyser … + cotext of specialised vo-
cabulary.
‘One analyses (generally) / One can/may analyse …’

In all three disciplines, the pronoun on is used to refer to the more or less vaguely
defined community consisting of author(s) and colleagues. In contrast to ON1
and ON2, ON3 is a value clearly situated on the indefinite side of the scale indefi-
nite–personal. The present tense (PR) in its general meaning is easily combined
with this on. The semantics of the verb does not seem important in this context,
but these verbs are naturally often of the research or cognitive type. However, the
criterion related to the presence of specialised expressions is essential. It contrib-
utes to the distinction between ON3 and ON4, the two indefinite values, which
otherwise are easily confounded. Let us look at some examples:
(15) […] dans le cas de verbes intransitifs à sujet inanimé et auxiliés en avoir, on
n’observe pas, généralement, de formation transitive à sens causatif corre-
spondent […]. (frling30)
‘… one does not observe, generally, transitive formations with a causative
sense …’

26. The symbol # replaces equations and symbols in the examples.

 Chapter 4. Presence of the author 121

(16) D’un point de vue catégoriel, on peut isoler 2 groupes pathologiques distincts.
(frmed13)
‘… one can isolate 2 distinct pathological groups …’
(17) On pourrait améliorer cet estimateur en changeant la fonction objectif […].
(frecon40)
‘One could improve this estimator by changing the objective function …’

All of these examples contain discipline-specific expressions; the verb in (15) takes
the present tense (the general sense is supported by the adverb généralement) and the
verbs in (16) and (17) are modified by the modal pouvoir. Below is a final example:
(18) De toutes ces prépositions, seules près de et hors de n’ont pas de correspondant
nominal (leur origine est comme on le sait adverbiale). (frling14)
‘… (their origin is as one knows adverbial).’

The inserted comme on le sait resembles the comme on l’a vu typical of ON2; how-
ever, the verbs and the tense are different: voir in PC tense in ON2 and savoir in PR
tense in ON3. The “translation” of the value in ON2 could be “as I/we and you the
readers of this article have seen” while the translation for ON3 could be “as I/we
and you my/our colleagues know”.
ON4 – referring to the author(s) and a non-limited community
The ON4 value is not easily distinguishable from ON3. The following traits can
often be used to distinguish between them: The verb in ON4 is not of the research
type, and there is no specialised vocabulary in the immediate cotext (no vocabu-
lary that cannot be understood by non-experts of the field). The other criteria are
the same as for ON3, i.e. PR for verb tense, or main verb preceded by a form of
pouvoir. A model example could be as simple as the following:
(19) On fait / On peut faire …
‘One does / One can/may do …’

The occurrences which we have interpreted as belonging to ON4 are normally

found in contexts referring to activities not related to research (and therefore there
is no specialised vocabulary). On the contrary, the activities referred to are in most
cases activities that “everybody” can undertake, such as in the following example:
(20) […] ; sans cette confiance qui caractérisait autrefois les relations qu’on en-
tretenait avec son médecin de famille (et qu’on entretient peut-être encore)
et que la promotion du médecin référent cherche à institutionnaliser, il est à
craindre que […]. (frecon04)
‘… the relations that one had with one’s family doctor (and that one still has
perhaps) …’
122 Academic Voices

In linguistics articles, there are often discussions about what one can say and what
one cannot say in a language. This is an activity in which every speaking subject of
the relevant language can participate, everybody is “qualified” for the discussion
(in these contexts, the “non-limited’ community is, of course, limited to those who
speak the language in question):
(21) On peut dire tout contre mais pas *tout sur, *tout jusqu’à, […]. (frling16)
‘One can say tout contre but not *tout sur, *tout jusqu’à …’
We will now turn to the last two values we have identified in our analysis of on
occurrences: ON5 and ON6. Both are more personal than indefinite, in the sense
we use these terms here, and as we will see below, they are not very frequent in
our corpus.
ON5 – referring to the reader(s)
The criteria we have identified in this case are the following: presence of a meta-
textual element, FUT as verb tense and verbs of the cognitive or perception type.
A model example could be:
(22) On trouvera dans le tableau 5 …
‘One will find in table 5 …’

The ON5 value manifests itself in contexts containing a metatextual reference. The
FUT tense takes on an instructive value by which the author indicates to the read-
ers what they should do in order to follow the presentation in the article. This is a
trait which distinguishes ON5 from ON2. Let us look at some examples:
(23) On trouvera dans le Tableau 1 ci-après un échantillon représentatif des prépo-
sitions spatiales du français. (frling14)
‘One finds in Table 1 below …’
In example (24), on is part of the more or less lexicalised construction si l’on veut
(‘if one likes’):
(24) Implicitement, je considère des entreprises ayant déjà réalisé ces investisse-
ments initiaux (ou si l’on veut : ayant déjà payé les coûts fixes). (frecon07)
‘(… or if one likes …)’

This seems to be a special case of ON5 which can be translated into si vous voulez
(‘if you like’). The author proposes new expressions to the readers in order to re-
place what he or she has already said – new expressions which might be consid-
ered as better (by others, and therefore also by the readers) than the first ones.
 Chapter 4. Presence of the author 123

ON6 – referring to the other(s)

For the interpretation of ON6, i.e. the pronoun on referring to other individuals,
more or less specific, and different from both author and reader, we search for a
bibliographical reference (name or name + year of publication) in the immediate
cotext. A model example could be the following:
(25) Dans Duval 1980, on dit …
‘In Duval 1980, one says …’
Instead of referring directly by il(s)/elle(s) (‘he, she/they’) to the persons or studies
that are introduced, the author uses the fuzzy pronoun on:
(26) Ce qui se dégage de cette sorte de brassage ”pluridisciplinaire” (Kleiber 1982:
3b), c’est que l’attrait pour la sémiotique de Peirce et celle de Morris (où on
employait le mot pragmatique sans nécessairement amener la chose jusqu’à
ses conséquences extrêmes) et pour la philosophie analytique d’Austin et de
Searle (où on appliquait la chose pragmatique sans nécessairement la nom-
mer) semble imposer son sort actuel à la problématique énonciative: elle est
désignée ”pragmatique” mais, surtout, elle est analysée comme un pragma-
tisme. (frling40)
‘… where one used the word pragmatics … where one applied the thing prag-
matics …’

In the following example there is in fact a double reference:

(27) En outre, dans la littérature, on ne sait prendre en compte l’absence d’exogénéité
stricte [6] que lorsque la variable est dichotomique, ce qui n’est pas le cas ici.
(frecon01)
‘… in the literature one does not know how to take into account …’
First there is the vague expression dans la littérature and then note 6, referring to a
specific publication. At the end of this passage, by the proposition ce qui n’est pas le
cas ici (‘which is not the case here’), the author indicates explicitly that this on does
not include himself; he is referring to other researchers. Here is a final example
taken from a medical article:
(28) On a récemment observé une telle baisse en Suède [12]. (frmed02).
‘One has recently observed this kind of decline in Sweden [12].’

The great majority of the references in medical articles are only represented by a
number in the body text of the article, as here by [12] (see section 6.3). The note
number refers to a bibliographical list. Thus the pronoun on in (28) refers to the
authors of the publication indicated by [12].
124 Academic Voices

Summing up
In the presentation above we have considered examples which are more or less
easy to interpret. There are, of course, numerous occurrences of the pronoun on
that have been difficult to classify as one of the six proposed values. However, the
criteria that we have identified have proven to be useful. Many of these criteria are
genre-specific. This confirms our hypothesis that the genre in which on manifests
itself may explain the kind of heterogeneity the pronoun displays. We have seen
that the use of on may bring along a range of different values, involving different
voices. We argue that the genre of the research article to a large extent is “respon-
sible” for this voice complexity.

4.4.4 Distribution of on-values

Table 4.4-2 represents our classification of the 1079 on-occurrences found in 180
of the 450 articles constituting the KIAP Corpus; we have used articles 01–20 in
each subcorpus.

Table 4.4-2. Distribution of values of on in different disciplines

Discipline Total ON1 ON2 ON3 ON4 ON5 ON6

frecon 456 29.6 12.7 52.2 3.1 0.7 1.8

frling 486 8.4 9.7 70.6 9.9 1.0 0.4
frmed 137 37.2 4.4 46.7 6.6 0.0 5.1
Total 1079 21.0 10.3 59.8 6.6 0.7 1.6
ON1 = je/nous; ON2 = je/nous + vous;
ON3 and ON4 = je/nous + community;
ON5 = vous; ON6 = il(s)/elle(s).

ON3, referring to the author(s) + a limited discourse community (self- & other-
dimension), is the dominating value (almost 60 % of the total). Thus we can con-
clude that the indefinite perspective wins over the personal. However, the other
indefinite value, ON4, corresponding to ‘everybody’, is only represented to a lim-
ited extent (6.6 %). This finding should not come as a surprise: the topic of a re-
search article is related to the world of specialisation and does not commonly offer
activities or facts concerning the non-scientific world.
Even if the indefinite values are in the majority, the frequency of the personal
values should not be underestimated. All in all they constitute about a third of all
the occurrences (33.6 %). This supports the conception of academic discourse as
including personal traces. It is particularly interesting to observe that ON1, refer-
ring to the author(s), is the most frequent; in fact, it represents 21 % of all the oc-
 Chapter 4. Presence of the author 125

currences. Thus, the pronoun on seems an important tool to avoid the use of the
pronouns je/nous.
We also note that the value ON2, corresponding to the author and the reader,
is quite frequent (10.3 %). This result indicates that the author, without directly
addressing the reader, employs on in order to create or encourage common par-
ticipation in the activities and reasoning presented.
Finally, the last two personal values, ON5 (corresponding to vous) and ON6
(corresponding to il(s)/elle(s)), are very rare. This shows that on is not commonly
used to refer to the reader alone (moreover, direct address of the reader made ex-
plicit by the pronoun vous, for example, is almost absent; see, however, section 5.3)
and that on is not used to refer to other specific researchers.
To conclude the general discussion of the personal values attributed to on,
we may say that, in the materials studied here, the on corresponding to je/nous or
to je/nous+vous wins over the on corresponding to vous or to il(s)/elle(s). In other
words, the self-perspective wins over the other-perspective.
We will now take a quick look at the disciplinary differences. It is particularly
interesting to observe the strong presence of ON1 in economics and medicine
compared to linguistics. This finding is surprising if we take into account the tra-
ditional conception of academic discourse, according to which especially the hard
sciences are characterised by an absence of personal constructions. However, these
results should be seen in relation to the results presented in sections 4.1 and 4.2.
The relatively weak presence of ON1 in linguistics is clearly modified by the rela-
tively strong presence of the pronouns je and nous. It seems that linguist authors
do not have to use ON1 since they use je or nous. On the other hand, ON3 is by far
the most frequent in linguistics articles. This value is also frequent in the other two
disciplines. In general terms, this finding indicates that the researchers prefer a
presentation form potentially including the relevant discourse community. How-
ever, with the pronoun on, this inclusion remains indefinite and rather fuzzy. An
interesting question in this context is to what extent the first person pronoun nous
carries a less indefinite and more specific and personal reference.

4.4.5 Final remarks

The results we have presented above indicate that the pronoun on has a clear per-
sonal potential, but also that its indefinite dimension is strong. We emphasise
that the criteria we have found may be further developed and refined and that the
six proposed values by no means are absolute. The criteria we have used are only
126 Academic Voices

“helpers” in the interpretation process related to the disambiguation of on.27 There

are no clear dividing lines between the different values, and whatever seems to be
the strongest dimension on the indefinite–personal scale, the pronoun on remains
personal (or human) in its reference but, at the same time, retains its inherent
semantic trait of indefiniteness (see Rabatel 2001 for a similar view). Thus, in our
view, there is no clear distinction between personal and non-personal values.
As regards the characterisation of the French research articles, the strong pres-
ence of on is important in itself. Within a conception of academic discourse as be-
ing non-personal, the pronoun on seems very suitable to the genre we are studying.
Its remarkably flexible nature makes it an efficient tool for the authors, enabling
them to balance between a personal presence coloured by the indefinite and an
indefinite presence coloured by the personal. There may be different explanations
for the use of each value (Fløttum et al. forthcoming); however, for each of these
the pronoun on represents a possibility for the authors to refer both to themselves
and to others in an indirect and non-explicit way. The academic authors often try
to hide behind various formulations, and the on-construction presents a series of
possibilities which they can choose from.28 In this sense our hypothesis is con-
firmed: The genre is an explanatory factor of the kind of heterogeneity displayed
by this pronominal practice.

4.5 Author roles and evaluation in abstracts

4.5.1 Introduction

In the present study our main focus is on the actual article, but as we have also
stored the abstracts as part of our electronic corpus, this allows us to focus briefly
on how the presence of the author is manifested in these short but complex texts.
Abstracts are complex in the sense that they may be read as independent texts
representing a genre of their own (e.g. when they appear in abstracts databases),
but still they are never completely independent since their relationship with the
text they ”belong” to is always present in some sense, whether we read them as

27. Atlani (1984: 21) argues that in the end, it is the receiver who has to choose which interpre-
tation is most relevant.
28. There are a number of questions which deserve further discussion concerning the
role of on, particularly its interaction with other personal pronouns in a text and its role in
(co-)referential chains (cf. 4.6.1). Another interesting question is how the transition between the
indefinite and the personal perspectives is realised, and whether this can be explained, to some
extent, by formulating hypotheses related to the presence of semantic isotopies (see Fløttum et
al. forthcoming).
 Chapter 4. Presence of the author 127

discipline professionals or study them as text researchers. They are also complex
in the sense that they may serve several communicative purposes. Their overall
function is that of a decision tool, helping the reader to decide whether it is worth
reading the whole article. If the reader does not go on to read the full text, the
abstract becomes a replacement for the whole text; if the reader is led on to the
article, the abstract has served the purpose of relevance indicator. A somewhat dif-
ferent aspect is that of promotional value. In today’s competitive research world,
the researcher must make great efforts to stand out in the crowd. The abstract as a
promotional tool becomes an important device in that context, an issue taken up
in studies by e.g. Berkenkotter & Huckin (1995), Melander, Swales & Fredrickson
(1997) and Lindeberg (1998, 2004).
There is by now quite a comprehensive literature on abstracts (cf. e.g. Dahl
2000; Hyland 2000). In the present context we are concerned with the issue of au-
thor roles (cf. section 4.5.2), since this has been discussed for the main text (4.2).
As abstracts are quite short and therefore more manageable texts than whole ar-
ticles when it comes to qualitative text analysis, we have here included both first
person singular and plural pronouns in our discussion (cf. 4.2 and 4.3). The same
applies to another aspect of author manifestation, viz. the use of linguistic signals
of a clearly evaluative nature. In order to address the promotional function of ab-
stracts, we briefly discuss the use of such signals in the abstracts (4.5.3). Finally, we
attempt to sum up by pointing to characteristic features of our abstracts from three
disciplines and three languages that have emerged from the discussion (4.5.4).
Our investigation is based on 40 abstracts for all language/discipline combina-
tions, apart from French linguistics. The reason why not 50 abstracts were included
in each subcorpus is that some articles lacked abstracts, and 40 turned out to be the
highest common number available. As for French linguistics, one of the journals
in the corpus (Revue de Sémantique et Pragmatique) did not have abstracts; in ad-
dition, one other journal (Langue Française) only had a few articles with abstracts
in French (but quite a few articles with abstracts in English). The total number of
French linguistics abstracts included in this study is therefore only 27.

4.5.2 Author roles

The three main roles are, as stated in section 4.2, the writer, the researcher and the
arguer.29 In contrast to the discussion of author roles in the main text, we have, as
already stated, for the abstracts recorded examples with both singular and plural
first person pronouns. In addition, we have for the French texts included instances

29. The evaluator role turned out to be of little significance in the abstracts and has not been
included in the discussion.
128 Academic Voices

of on when the reference is clearly personal (cf. section 4.4). The following three
examples are cases in point:
(1) On utilise l’enquête sur la Formation et la Qualification Professionnelle (FQP)
réalisée en 1993. (abstract frecon21)
‘We use the survey …’
(2) On étudie, notamment, un équilibre de court terme (le cas d’une ville fermée)
et un autre de moyen terme (le cas d’une ville ouverte où l’offre de logements
est fixe et les travailleurs libres d’entrer, ou non, sur le marché). (abstract
frecon37)
‘We study especially a short-term equilibrium …’
(3) On opposera à ceux-ci l’approche interne, d’inspiration plus résolument lin-
guistique, dont les principaux représentants sont Jakubinskij et Jakobson. (ab-
stract frling24)
‘We contrast to these the internal approach …’

As stated in section 4.2, the academic author switches back and forth between
these three roles. The researcher role may perhaps be said to be underpinning aca-
demic work. For rhetorical purposes, all three roles move in and out of focus. As
is the case for the main text, in the abstracts, too, there is no perfect correspond-
ence of verbs and rhetorical role. In many cases we need to take elements in the
cotext into consideration before a role may be assigned. The cotext may swing the
interpretation in one or the other direction, or even result in two roles being si-
multaneously present. Here is an example that illustrates such a dual role, in this
case researcher-arguer:
(4) I critically evaluate this new work and suggest that it does not contribute suf-
ficient evidence against the direct access view. (abstract engling22)

However, in the majority of the cases one rhetorical role is clearly assumed by
the author. In the relatively few cases in the abstract corpus where a dual role is
present, we have chosen to assign the example to the category that was seen as the
most dominant one in each case. The example above was classified as an instance
of the researcher role. The following abstract illustrates how the author switches
back and forth between the three rhetorical roles:30
(5) Quantification and the nature of crosslinguistic variation

The standard analysis of quantification says that determiner quantifiers (such

as every) take an NP predicate and create a generalized quantifier. The goal

30. For convenience, when longer stretches of text are discussed, English examples are most
often used.
 Chapter 4. Presence of the author 129

of this paper is to subject these beliefs to crosslinguistic scrutiny. I begin by

[writer] showing that in St’at’imcets (Lillooet Salish), quantifiers always re-
quire sisters of argumental type, and the creation of a generalized quantifier
from an NP predicate always proceeds in two steps rather than one. I then
explicitly adopt [researcher] the strong null hypothesis that the denotations of
quantifiers should be crosslinguistically uniform. Since the Salish data cannot
be captured by the usual analysis of English, I pursue [researcher] the idea that
English is reducible to the Salish pattern. Reanalysis of many English con-
structions is required. I argue that [arguer] the reanalysis has advantages over
the standard analysis for partitives, as well as for non-partitive all- and most-
phrases, which I analyze as containing bare plurals of argumental type. Even
where the new analysis faces some challenges (for example, with every), the
attempt still leads to fruitful results. It forces us to view familiar constructions
in a new light, and to redefine, I believe [arguer] correctly, which quantifica-
tional constructions are ‘basic’ and which stand in need of further explana-
tion. (abstract engling45)

Our investigation revealed that quite a few abstracts within each subcorpus did
not contain instances of first person pronouns at all. The figures for how many
abstracts did have such pronouns are provided in Tables 4.5.1–3. With one excep-
tion, abstract length is the only restriction the journals represented in our corpus
have in their author instructions. (For the economics abstracts the requirement is
typically 100 words, while within linguistics and medicine up to 200 words may be
allowed.) The exception is one of the two Norwegian linguistics journals, Maal og
Minne, which specifically prohibits the use of first person pronouns in the abstract.
The six abstracts from that journal have still been included in order to study their
use of evaluative elements.31 What we very often see in abstracts (both with and
without first person pronouns), however, is that the inanimate noun paper/article
is used with active verbs, as in the following examples:
(6) This paper explores social norms as a mechanism of how neighborhood char-
acteristics can affect individual behavior. (abstract engecon19)
(7) The paper argues that, on the other hand, Praguian functional syntax has a
great deal in common with more ‘formal’ functionalist approaches and with
much work in formal semantics. (abstract engling04)
(8) Cet article examine comment une baisse des cotisations sociales employeurs sur
le travail non qualifié peut être financée par des hausses d’autres taxes ou para-
taxes, et quelles sont les conséquences pour l’économie. (abstract frecon13)

31. Of the 44 articles in the noling subcorpus from the other journal, Norsk Lingvistisk Tids-
skrift, eight lacked an abstract.
130 Academic Voices

‘This paper examines how …’

(9) Cet article souhaite contribuer à une théorie du marché du logement loca-
tif où les frictions de recherche sont rendues endogènes par une fonction
d’appariement entre propriétaires et locataires potentiels. (abstract frecon37)
‘This paper wants to contribute to …’
(10) Artikkelen studerer virkninger av ansettelses- og oppsigelseskostnader for ar-
beidsmarkedets funksjonsmåte og effektivitet. (abstract noecon06)
‘The paper studies …’
(11) Artikkelen diskuterer en type dobbelt bestemte nominalfraser som hyppig
forekommer i muntlig norsk. (abstract noling09)
‘The paper discusses …’

Such usage has been shown to be a typical feature of scientific language (Prelli
1989; Master 2001). The academic author hence has a choice between stronger tex-
tual presence through a pronoun (‘I/We examine and present’) or weaker presence
through personification of inanimate nouns (‘The paper examines and presents’).
In the present corpus, nearly all the abstracts in English and Norwegian econom-
ics and linguistics, and to some extent the French economics abstracts, with no
first person pronoun used ‘paper/article’ in this way.
Tables 4.5.1–3 provide figures for the distribution of the examples over the
three posited author roles for the three disciplines and the three languages. The
number of abstracts with occurrences of first person pronouns is also given for
each subcorpus.

Table 4.5-1. Author roles in economics abstracts

English French Norwegian

(20 abstracts) (29 abstracts) (19 abstracts)

Writer 3 15 10
Researcher 36 23 19
Arguer 7 17 7

Total no. of occurrences 46 55 36

Average per abstract 2.3 1.9 1.9
 Chapter 4. Presence of the author 131

Table 4.5-2. Author roles in linguistics abstracts

English French Norwegian

(17 abstracts) (9 abstracts)* (10 abstracts)

Writer 6 4 6
Researcher 19 6 5
Arguer 25 6 4
Total no. of occurrences 50 16 15
Average per abstract 2.9 1.8 1.5
* Out of a total of 27 abstracts.

Table 4.5-3. Author roles in medical abstracts

English French Norwegian

(12 abstracts) (16 abstracts) (17 abstracts)

Writer – 3 7
Researcher 19 18 10
Arguer 2 1 2

Total no. of occurrences 21 22 19

Average per abstract 1.8 1.4 1.1

We will now comment on the results for the three disciplines one by one, before
comparing disciplines and languages.

Economics
As can be seen from Table 4.5-1, half of the 40 English and 40 Norwegian abstracts
in this discipline contained overt traces of the author in the form of a personal
pronoun, while for French, close to three quarters did. In terms of the average
number of occurrences in the abstracts with pronouns, we see that the English
abstracts have more occurrences (2.3) than the Norwegian and the French, which
are identical in this respect (1.9)
As regards the difference in distribution over author roles, we see that the
researcher role is the dominant one in all three languages, and for English it is
overwhelmingly so. However, in Norwegian and French, the author also assumes
the writer role, and in French, the difference between the arguer role and the re-
searcher role is not as prominent as in English.
132 Academic Voices

Linguistics
Table 4.5-2 shows that for English, roughly 40 % of the linguistics abstracts con-
tain at least one personal pronoun, while the figure for Norwegian is only 25 % (of
course keeping in mind here that six of the abstracts come from the journal that
prohibits the use of such pronouns). For French, the subcorpus comprises only
27 abstracts; hence, the percentage here is 33. In terms of average per abstract,
English has the highest number of occurrences with 2.9, while French has 1.8 and
Norwegian 1.5.
As for roles, author behaviour varies greatly between the three languages. In
English, the arguer role is very much in focus, with the researcher in second place
and the writer as a modest number three. For French and Norwegian, the numbers
are so small that little can be deduced from them.

Medicine
In Table 4.5-3, we see that 30 % of the English abstracts display author presence
through a first person pronoun, while the percentage for both Norwegian and
French is about 40 %. As regards the average number of occurrences per abstract,
English has 1.8, while French has 1.4 and Norwegian 1.1.
Concerning author roles, the researcher role is the dominant one in all three
languages, and very clearly so in English and French. In Norwegian the writer role
is also important, while the arguer role is negligible in all three languages.

Comparisons
As regards the percentage of abstracts that contain a personal pronoun indicating
author presence, the results show that economics is the discipline which has the
highest overall figure, with medicine in second place and linguistics in third place.
No clear language pattern can be discerned across the disciplines. We see that for
economics, French has the highest percentage with 75 %, while for linguistics it is
English (40 %) and for medicine Norwegian and French (40 %). In terms of the
”strength” of the presence, we see that English across all three disciplines has the
highest average number of occurrences per abstract, with French in second place
and Norwegian in third (tied with French in economics).
When it comes to the issue of what role(s) the author assumes in the abstract,
we have seen that there is some variation both in terms of language and in terms
of discipline. The only general statement we may make is that the researcher role is
the dominant one across disciplines and languages (the only exception being Nor-
wegian linguistics where in fact the writer role has one more recorded occurrence).
As for English, each discipline has a clearly favoured role, viz. the researcher in
economics and medicine and the arguer in linguistics. For the other two languages
the picture is more blurred. In Norwegian, the economist is mainly present in the
 Chapter 4. Presence of the author 133

abstract as researcher, but the other two roles are also in evidence; in linguistics
there is little presence overall, but the few occurrences divide fairly evenly over the
three roles; in medicine the researcher and the writer roles are both exploited. In
French, the situation for economics and linguistics is a distribution across the three
roles, while for medicine the researcher role is again clearly the favoured one.
We stated in the introduction to this section that we would compare our find-
ings regarding author roles for the abstracts with what we found for the articles
(section 4.2). The comparison can only provide us with an impression of this re-
lationship, since for the article our investigation is limited to occurrences of first
person singular pronouns, while for the abstracts the plural forms are included as
well. In addition, since the number of single-author articles is so small in some
language/discipline combinations (all medical texts, as well as French economics),
these were excluded as well. These limitations apart, what is clearly seen when we
compare author roles in the two text genres is that in both genres the researcher
role is the dominant one. For the medical texts in the abstract corpus, this is the
only role that matters (Table 4.5-3). However, there are also differences in behav-
iour between the two kinds of text. In the articles, the writer role is the second
most important one, while we for the abstracts see that the arguer role holds this
position in several of the subcorpora, notably French economics and English lin-
guistics (Tables 4.5-2 and 4.5-3). The fact that the arguer role seems to be more
important in the abstracts than in the articles may perhaps be explained by the
different functions of the two genres. The promotional function of the abstracts
is emphasised by the author through the presence of a persona who more openly
promotes research ideas and findings to attract the reader.
Let us now take a closer look at some of the figures given above. In Norwegian
linguistics, only one quarter of the abstracts display author presence through per-
sonal pronoun use at all. Of the 30 abstracts that do not contain such pronouns, 6 are
from the journal Maal og Minne which, as stated above, explicitly prohibits them. This
still leaves us with 24 abstracts from the other Norwegian linguistics journal (Norsk
Lingvistisk Tidsskrift), which does not explicitly prohibit their use. Hence, we must
conclude that Norwegian linguists prefer the impersonal abstract to a greater extent
than their English and French language colleagues. The Norwegian linguists also dif-
fer from their fellow countrymen the economists in this respect. In order to check
whether this feature is linked to certain authors or to the abstract as genre within
Norwegian linguistics, we compared the use of first person pronouns in the Norwe-
gian linguistics articles (i.e. the main text) and the use in the abstracts. We found that
in the cases where the abstracts are impersonal, the accompanying main text does use
first person pronouns roughly to the same extent as in the articles where the abstracts
do have such personal pronouns. Hence, at this point we cannot offer any explana-
tion for this observation other than stating that it seems to be a genre practice that is
134 Academic Voices

specific to this language/discipline combination. The Norwegian linguists are perhaps

closer to the prevailing norm for abstracts as portrayed in textbook discussions of this
genre than their colleagues writing in the other two languages.

4.5.3 Evaluative elements

The investigation of evaluative elements in the abstracts relates to our third research
question, viz. how the author’s attitudes are expressed through the presentation
and promotion of his or her own research (see section 1.1). We have limited our
search for evaluative elements to what Lindeberg (2004), in her interesting cross-
disciplinary study of promotion and politeness in research articles, calls boosts,
defined as ”positive assessment of contribution” (pp. 40−41) and other evaluative
elements linked to the results obtained. Even the handling of such a seemingly
simple category is not always straightforward, and the topic deserves a much more
thorough treatment than what we have had the opportunity to undertake here.
Lindeberg discusses the difficulties text researchers may experience when trying
to assess how rhetoric works within a discipline where he or she is not an expert.
Her solution to this problem has been to focus on linguistic items that are easily
identifiable, while acknowledging the fact that
[…] for the subject expert there are large numbers of signals that only they under-
stand the significance of. Thus it can be assumed that linguistic signals of evalua-
tion in fact form a continuum: at one end are items that are undisputably evalua-
tive (e.g. good/bad, important/unimportant) and at the other end are items that are
only contextually evaluative. (Lindeberg 2004: 60)

Let us now present some impressions from our corpus, illustrated by a few exam-
ples. In general, we find that for this category there is very little use of linguistic
signals in the abstracts which may be described as ”undisputably evaluative” in
Lindeberg’s words. On the other hand, we do find relevant examples in all nine
subcorpora. What is also clear, however, is that in six of the nine, only a couple
of abstracts have such elements. The three subcorpora with most overt evalua-
tion of results are English linguistics (8 out of 50 abstracts), and Norwegian and
French medicine (with 9 and 10 out of 50 respectively). The subcorpus with the
lowest occurrence is Norwegian linguistics (1 abstract). As shown in Table 4.5-2,
the Norwegian linguistics abstracts also have the lowest number of occurrences of
first person pronouns.
 Chapter 4. Presence of the author 135

Here is a selection of examples from all nine subcorpora:

Linguistics
(12) […] thus providing a more comprehensive framework for analysing […]. (ab-
stract engling23)
(13) [ …] the attempt still leads to fruitful results. (abstract engling45)
(14) We conclude that computational corpus methods provide an important check
on […]. (abstract engling21)
(15) We also demonstrate that modelling this interaction allows us to achieve a
more refined interpretation than […]. (abstract engling09)
(16) Des statistiques complètent cet article sur des emplois de plus de dix mille
verbes […]. (abstract frling03)
‘… more than ten thousand verbs…’
(17) La deuxième partie de l’article développe la théorie B, plus solide, à savoir la
théorie de la pertinence de Sperber et Wilson. (abstract frling11)
‘… more solid …’
(18) Dans la deuxième partie de notre travail, nous présenterons deux nouveaux
traits pour la préposition […]. (abstract frling16)
‘… two new features …’
(19) Artikkelen kan likevel gi et bidrag til ytterligere forståelse og kunnskap innen
disse feltene […]. (abstract noling48)
‘… may still contribute to increased understanding and knowledge …’

Economics
(20) Also, contrary to conventional wisdom, the differences in gains from the two
approaches do not arise from treating stock returns as exogenous rather than
endogenous. (abstract engecon06)
(21) […] ceci permet de fournir une base analytique intéressante dans la tentative
de définir une consommation soutenable. (abstract frecon02)
‘… to provide an interesting analytical base …’
(22) […] nous permet d’obtenir des résultats pertinents. (abstract frecon36)
‘… allows us to obtain relevant results …’
(23) En viktig konklusjon fra den empiriske delen er at bruttonasjonalprodukt i
faste priser går noe ned ved en overgang fra dagens avgiftssystem. (abstract
noecon43)
‘An important conclusion …’
136 Academic Voices

Medicine
(24) This is the first time that histochemical lectin staining has been correlated
with biochemical mapping of oligosaccharides. (abstract engmed08)
(25) Because BP1 is expressed abnormally in breast tumors, it could provide a use-
ful target for therapy, particularly in patients with ER-negative tumors. (ab-
stract engmed28)
(26) suggesting NQO1 may be important in chemosensitivity as well as the
pathogenesis of lung cancer and NQO1 genotyping may be a useful com-
ponent of pharmacogenetic strategies for the treatment of NSCLC. (abstract
engmed41)
(27) Comme les données obtenues à partir de MED-ÉCHO couvrent pratique-
ment l’ensemble des naissances québécoises, […] elles permettent de fournir
un portrait plus précis des anomalies congénitales au Québec. (abstract
frmed01)
‘… to give a more accurate picture …’
(28) Notre étude est un état des lieux utile pour mieux définir les conditions
d’accueil et d’accompagnement des toxicomanes aux Urgences. (abstract
frmed30)
‘Our study is a useful point of departure to better define …’
(29) Nos 3 observations illustrent bien les particularités de cette association mor-
bide. (abstract frmed34)
‘Our 3 observations provide a good illustration of …’
(30) Il nous semble donc utile de les proposer à tout patient porteur d’une DMLA
et se plaignant de difficultés à la lecture. (abstract frmed40)
‘It seems useful to …’
(31) La lentille non pliable en PMMA donne de relativement bons résultats […].
(abstract frmed44)
‘… gives relatively good results …’
(32) Le laser excimer InPro-Gauss donne des résultats réfractifs très satisfaisants.
(abstract frmed47)
‘… gives very good results …’
(33) Våre resultater er meget gode […]. (abstract nomed01)
‘Our results are very good …’
(34) Risperidon virker derfor lovende som et hjelpemiddel ved behandling av van-
skelig kontrollbar aggresjon […]. (abstract nomed04)
‘Risperidon therefore seems promising …’
 Chapter 4. Presence of the author 137

(35) Korttidsprognosen var god, og i vår serie langt bedre enn det norske data for
trombolysebehandlede pasienter har vist tidligere. (abstract nomed21)
‘… in our series far better than …’
(36) BFM-regimet gir gode resultater som primærbehandling ved Burkitts lym-
fom både hos barn og voksne […]. (abstract nomed35)
‘… gives good results …’
(37) Ved vanskelig pankreatikobiliær diagnostikk synes derfor S-MRCP å være
et nyttig og komplikasjonsfritt supplement til eksisterende diagnostiske
hjelpemidler. (abstract nomed41)
‘… seems to be a useful supplement without complications …’
As for the grammatical nature of the evaluative items, we see from the examples
that the majority are evaluative adjectives like new, important and useful. There is
also an instance of a quantifier expressing positive amount (example (16): plus de),
a prepositional phrase implying novelty (example (20): contrary to conventional
wisdom), a temporal noun phrase implying novelty (example (24): the first time)
and an adjective with discipline-specific positive meaning (example (37): kompli-
kasjonsfritt).

4.5.4 Summing up

Our primary concern in this section has been author presence in the form of a
first person pronoun. We have seen that, with the exception of subcorpus frecon,
the majority of the abstracts do not have any occurrences of such pronouns. The
economics abstracts in general are in fact the ones with most author presence in
this form. In terms of what we have called the ”strength” of the presence, i.e. the
number of pronouns per abstract, the English abstracts have the strongest pres-
ence in all three disciplines and the Norwegian ones have the weakest. As regards
author roles, manifested in pronoun + verb-combinations, we see that for eco-
nomics and medicine the researcher role is the dominant one, while for linguistics
the arguer role is even more important.
We have also looked briefly into another clear manifestation of author pres-
ence, viz. the use of evaluative elements in the abstracts, and find that there is little
overt evaluation in terms of lexical signals to promote own results (see section
5.4). However, English linguists as well as French and Norwegian medical authors
are somewhat more willing to promote their findings in this way than authors in
the other language/discipline combinations.
Our investigation has shown that within all three languages and disciplines
studied here, the author has a choice whether to be visible or invisible in the ab-
stract. What we have also seen in our corpus is that if an author chooses a persona
138 Academic Voices

that is overtly present in the abstract, he or she is also more likely to make use of
overt evaluation of own results than authors that take on a persona that is less vis-
ible. Whether there is a link between author presence in the form of first person
pronouns and such promotional elements in the research articles as well perhaps
deserves further treatment.

4.6 Case studies

4.6.1 Pronoun use in individual articles

This subsection contains three studies related to pronominal use. In the first one,
we look at two French linguistics articles in order to see to what extent different
pronouns (and to a certain extent different voices) may complement each other,
as regards reference and author roles. In the second case we look at two French
linguistics articles produced in similar contexts but displaying different pronomi-
nal uses. In this case we are particularly interested in author roles in relation to
rhetorical strategies. The third case focuses on medical writing practices. We have
selected one article from each of the three languages in order to study personal
versus indefinite perspectives related to the use of personal and indefinite pro-
nouns. In all three studies, we use the classification of author roles established in
section 4.2 (i.e. the three main roles: writer, researcher and arguer). We will also
refer to the classification of the French indefinite pronoun on from section 4.4,
where we listed six reference values which this pronoun can assume in research
articles. Even if the nature and frequency of indefinite pronouns are different in
the three languages, we argue that these six values can be posited as general poten-
tial values of the indefinite pronoun, including English one and Norwegian man,
en/ein. We repeat the six values here for convenience:
• ON1 referring to the author(s)
• ON2 referring to the author(s) + reader
• ON3 referring to the author(s) + the relevant discourse community
• ON4 referring to people in general
• ON5 referring to the reader only
• ON6 referring to other researchers

First case
Different authors mark or colour their texts by their individual style by means of
exploiting the semantic-referential potential of pronouns and by combining them
with specific verb types. We will now look at two French linguistics articles in or-
 Chapter 4. Presence of the author 139

der to see to what extent different pronouns may complement each other. The two
articles display the distribution of pronouns shown in Table 4.6-1:32

Table 4.6-1. Pronoun use in two French linguistics articles

Body je nous on
words N R.f. N R.f. N R.f.

frling13 4,951 27 5.5 3 0.6 58 11.7

frling14 3,364 1 0.3 7 2.1 38 11.3
R.f. = relative frequency per thousand.

We see that frling13 contains relatively many occurrences of je. They realise in fact
all the three main author roles; writer in (1), researcher in (2) and arguer in (3):
(1) Je commencerai par noter que […]. (frling13)
‘I start by noting that …’
(2) […] j’utiliserai le corpus de […]. (frling13)
‘I will use the corpus …’
(3) Cette solution fera appel à des notions que j’ai utilisées dans la théorie des
stéréotypes que je défends depuis 1990. (frling13)
‘… which I have defended since 1990.’

The use of the first person plural pronoun nous seems to be limited to the inclusive
type, referring to author + reader (see section 4.3):
(4) Mais nous avons également vu que les essentielles sont représentées par des
phrases génériques admettant des exceptions. (frling13)
‘But we have also seen that …’

Finally, the use of the pronoun on in this text seems to be primarily indefinite; this
pronoun appears with the values of ON3 referring to the author + the relevant
discourse community of linguists, as in example (5), and ON4 referring to people
in general (in this context to people speaking French) as in example (6):
(5) Dès lors qu’on ne parle plus de propriétés intrinsèques, la combinaison re-
devient possible. […]. (frling13)
‘From the moment that one does not talk about intrinsic characteristics any
more …’
(6) En effet, on peut très bien dire […]. (frling13)

32. As mentioned in Chapter 3, there are no clear differences in the KIAP material related to
gender differences. Note that frling13 has a male author and frling14 a female.
140 Academic Voices

‘In fact, one may very well say …’

To sum up, this article contains a large number of first person je, manifesting itself
in three different roles. Nous is not used very much and manifests mostly only
one value. Finally, the pronoun on is frequent, but manifests itself mainly in two
indefinite and not personal values.
Let us now compare frling13 with frling14, containing one je, 7 nous and 38
on. The most striking difference is that frling14 contains only one je, which mani-
fests a dual writer-researcher role (the writer or text organiser role is marked in
the metatextual sequence Pour en donner un exemple concret and the researcher in
the verb isoler):
(7) Pour en donner un exemple concret, je voudrais isoler une catégorie par-
ticulière de prépositions, […]. (frling14)
‘In order to give a concrete example, I would like to isolate …’

This very limited use of je explains why the plural pronoun nous is used, in this
article, with reference to the author (“nous de modestie”; cf. Loffler-Laurian 1980),
either as researcher or as writer, like in dans le Tableau 1 nous ne faisons figurer que
(‘… in Table 1, we only represent …’).
So far, this gives a homogeneous and simple picture of an author only referring
to herself. However, when looking at the numerous occurrences of on, we discover
a more heterogeneous picture. This pronoun is in fact used with at least four values
– ON1 referring to the author in (8), ON2 referring to the author + reader in (9),
ON3 referring to the author + the relevant discourse community of linguists in
(10) and ON5 referring to the reader only in (11):
(8) On peut signaler en passant que dedans, dessus peuvent fonctionner lorsque
[…]. (frling14)
‘One can indicate …’
(9) Comme on va le voir, le remplacement du syntagme prépositionnel […]. (fr-
ling14)
‘As one will see, …’
(10) On sait que des Prep simples comme chez, parmi, jusque, etc. sont incapables
de fonctionner sans nom régime. (frling14)
‘One knows that …’
(11) On trouvera dans le Tableau 1 ci-après un échantillon représentatif […]. (fr-
ling14)
‘One will find in Table 1 …’
 Chapter 4. Presence of the author 141

To sum up, frling14 contains only one occurrence of je (and thus only one role). Nous
is more frequent, but displays mostly one value, viz. referring to the author alone. Fi-
nally, on is frequent and appears with at least four different types of reference.
A comparison of the two articles offers an interesting picture of how the use
of pronouns may form different patterns in a text. Frling13 contains many occur-
rences of je, also assuming various roles, and may thus limit itself to the inclusive
value of nous. In order to open up the perspective from the personal (je and je +
vous (the reader)) to the more indefinite ”we, the linguists” or ”all of us” (meaning
here people who speak French), on is used. Through different reference but with
relatively clear values, the three pronouns complement each other.
As regards frling14, the author uses je only once and thus has to fall back on
nous in order to refer to herself. This makes it necessary to use on when widening
the perspective. The author attributes, in fact, several different values to on, prob-
ably because je and nous are used in few cases and with few values. The pronouns
complement each other here too, but the inherent vagueness of on has the effect of
a less clear author manifestation than in frling13.
For these two articles we can conclude that extensive and varied use of je leads
to few and more easily definable values of the poly-referential on. On the other
hand, modest use of je leads to many values of on, which are not so easily defin-
able. These phenomena deserve further study. For example, it would be interesting
to investigate to what extent it is possible to predict the distribution of referential
values, relative to the distribution of different pronouns. It would also be inter-
esting to study whether the pronominal distribution may have an effect on text
structuring.33

Second case
In this case, we also study two French linguistics articles, frling20 (2,376 words in
the body part) and frling32 (2,629 words), but in a somewhat different and more
rhetorically oriented analysis. These single-author articles are produced in similar
contexts. The two authors – we call them “author 20” and “author 32”, respectively
– are both fairly young female linguists, with university positions. Thus, some of
the factors which could have explained differences between the articles are elimi-
nated (same discipline, same language, same sex and similar age).
Rhetorically, the articles are quite similar too. Even if their issues and strat-
egies are different,34 they are both clearly marked by the prominent wish of all
rhetorical texts, viz. to be persuasive. Both authors seem to want to induce coop-

33. For an elaboration of these questions, see Fløttum (2004d).

34. For further elaboration of the analysis of these two articles, see Fløttum (2005a). For the
rhetorical perspective, see Prelli (1989) and section 2.1.
142 Academic Voices

erative acts and attitudes at the same time as they want to position themselves in
relation to the discourse community they belong (or want to belong) to. In other
words, the rhetorical factors do not indicate that we should anticipate great differ-
ences between the two articles. However, there are clear differences. Let us take a
look at how their linguistic choices are different and to what extent these support
the rhetorical balance between inducing cooperation and claiming positions (a
balance which in our view is not sufficiently discussed in Prelli 1989). Our hy-
pothesis is that the choice of pronoun may indicate different balancing. (However,
as shown above, the verb combined with the pronoun (i.e. the immediate cotext
of the pronoun) constitutes an important corrective to possible unidimensional
interpretations of the use of pronouns.)
Article frling32 contains 12 je, 4 nous and 14 on. Since all the three pronouns
are used, we might think that they complement each other, more or less as in
frling13, analysed in the first case above. However, in frling13, we found clear
cases of the arguer role in the use of je, while in frling32, je is typically used in the
researcher or the writer role, as in (12), taken from the Introduction section:
(12) Mais avant de passer à l’analyse proprement dite de ces constructions, je vais
décrire brièvement les conditions matérielles dans lesquelles cet échange s’est
produit […]. (frling32)
‘… I will describe …’

The arguer role is only modestly present in one sequence: je propose des hypothèses
… (‘I propose some hypotheses …’).
As regards the use of the other two pronouns, the few occurrences of nous
are used with inclusive reference (referring to the author–reader dyad, see 4.3 and
Kinn 2005a), in constructions like nous verrons que (‘we will see that’). The 14 on
occurrences are also used mainly in this inclusive manner, in constructions like on
observe (‘one observes’) and on a vu que (‘one has seen that’), corresponding to the
value ON2, typically with cognitive verbs, and typically occurring in the middle
sections of the articles. We conclude that the pronoun/verb use in frling32 is bal-
anced between a moderate je-positioning and a rather strong cooperative line as
manifested by the inclusive use of nous and on.
Article frling20 contains 27 nous and 15 on. Since author 20 does not use je,
it is reasonable to expect at least some of the 27 occurrences of nous to represent
the authorial nous. This nous is exclusive in the sense that it does not include the
reader.35 In fact most of the nous occurrences in this article are of the authorial
type, but the author assumes different roles – the writer role as in Nous ne revenons

35. However, frequently in the writer and researcher roles, it can be difficult to distinguish
between exclusive and inclusive use. (See sections 4.3 and 5.3.)
 Chapter 4. Presence of the author 143

pas sur … (‘We will not come back to …’) and the researcher role as in nous n’avons
relevé aucun emploi … (‘… we have not found …’).
However, what is perhaps most noteworthy is the clear presence of the author
as arguer, in combination with nous, as in the following examples:
(13) […] nous postulons que, dans ce cas, il s’agit de préposition. (frling20)
‘… we postulate …’
(14) […] nous avons soutenu que des constructions attestées à l’écrit illustrent un
emploi prépositionnel de genre. (frling20)
‘… we have supported/defended …’

Because of the meaning of the verb, it is more natural to interpret the pronoun
nous as authorial in examples (13)–(14). However, the multifunctional nous is also
used with a reference clearly including the reader, as in nous allons voir maintenant
(‘… we are now going to see …’) with the deictic adverb maintenant ‘now’ and
typically with the verb voir ‘see’. Further, the pronoun on is also used with differ-
ent reference potentials in this article, mostly as ON3, referring to the author +
the relevant linguistic discourse community. We conclude that the pronoun/verb
use in frling20 is balanced between the absence of the first person je and a full
exploitation of the reference potential of the pronouns nous and on – using both
positioning and cooperating constructions. The cooperation strategy, related to a
clear problem–solution argumentation line, is in addition clearly manifested by
numerous questions addressed directly to the reader throughout the text, as in
Comment analyser … (‘How to analyse …’).
The analysis of these two articles has shown that individual authors may as-
sume particular discourse identities by choosing different linguistic expressions.
However, it might seem that the two authors arrive at more or less the same effects
by exploiting pronominal constructions in quite different ways. It is also worth-
while to note that the first impression or tone given by the actual pronominal use
– isolated from other linguistic choices – can be deceptive. The use of first person
pronouns, like je in French, is not necessarily self-promoting, and the use of first
person plural, like nous in French, is not necessarily only inclusive and coopera-
tive. (Another point of interest, discussed in Kinn (2005a), is the imposing effect
the first person plural form may have; see also 5.3.) By taking a closer look at the
immediate cotext where the pronouns appear, we have seen that the authors can
construct their own subtle rhetorical strategy in the balancing between self and
others, between cooperation with the reader (and the relevant research commu-
nity) and self-positioning in relation to one or more research communities. In an
e-mail correspondence (October 31, 2004), author 20 points to tradition as an
important factor for using nous. She also points to the use of first person plural
144 Academic Voices

pronouns as a means of assuming the personality of the researcher as a member of

a community. Author 32 (e-mail of November 10, 2004) points to the possibility of
taking a personal position that the use of je offers. As regards the use of nous, she
explains it as a way of including the virtual interlocutor.

Third case
In section 4.1, we saw that our results correspond well to the general concep-
tion of medical researchers as “non-personal” authors in the sense that they use
first person and indefinite pronouns much less frequently than their linguist and
economist colleagues. However, as regards which pronouns they use when they
choose to express themselves in personal terms, our corpus also shows that there
are clear differences between the three languages: Norwegian medical authors use
pronouns the most, both ‘we’ and ‘one’; French medical authors are positioned in
the middle, also using both ‘we’ and ‘one’; their English colleagues use only ‘we’,36
more than the French, but less than the Norwegians.
We will now take a closer look at how these pronouns are used in three medi-
cal multi-author articles, focusing on reference and author role: engmed31 (6,054
words in the body part), frmed22 (3,997 words) and nomed39 (1,491 words).
Article engmed31 contains 15 we and 2 one. The main question we ask here is
to what extent the 15 we are used with a personal reference. Given the traditional
conception of medical discourse as “non-personal”, one might expect we to be
used with a more or less indefinite reference. However, the 15 occurrences of we
in engmed31 are in fact clearly personal and specific, referring to the authors of
the article. This interpretation is justified by the immediate cotext in which they
occur: We is regularly combined with a research verb referring to the activities
undertaken throughout the project which is reported, as in the sequences we com-
puted frequencies, we did not conduct proxy interviews taken from the “Data analy-
sis methods” section. The majority of the 15 we pronouns occur in the Discussion
section, and three of these are combined with the verb find. Thus, the authors
of this article do not hesitate to manifest themselves as researchers carrying out
their research as well as obtaining results. The homogeneity is striking in that the
instances of the pronoun we refer to the authors alone and are combined with
verbs indicating that the authors assume the researcher role in the text. As re-
gards the verbs, one exception should be mentioned. There is one occurrence of
we combined with the discourse verb focus on which manifests the writer role: In
this article we have focused on … As shown in section 5.1, metatextual expressions
are generally almost non-existent in medical articles. Finally, we note that there is
no explicit and overt argumentation manifested by rhetorical verbs like argue and

36. There are only 4 occurrences of the pronoun ‘one’ in the whole engmed subcorpus.
 Chapter 4. Presence of the author 145

claim, verbs that are very frequent in English linguistics articles. The argumenta-
tion in this article is realised by the fact-oriented presentation of the research proc-
ess and findings. As regards the two occurrences of one, both modified by modal
auxiliaries, they open up for a wider reference group (corresponding to ON3; see
the introduction to this section) including an indefinite community of medical
researchers. Here is one of the occurrences:
(15) It is also possible that inability to lactate or suppressing lactation after giving
birth might have a deleterious effect on breast physiology [20], but we did not
request this information from subjects. Given the relatively late ages of first
birth in this population, one might not expect to see a protective effect of hav-
ing ever given birth, […]. (engmed31)

Even if one in the sequence one might not expect to see … may refer to a wider
group than the authors alone, it is also possible to interpret this pronoun as re-
ferring to the authors only. The reason is that the sequence is closely related to
a discussion linked to their own specific project in the preceding sentence. The
same uncertainty with regard to interpretation may be attributed to the other oc-
currence of one:
(16) Although it was not the purpose of this study to explain the high rates of
breast cancer in Marin County, one hypothesis of the study was that if expo-
sures specific to Marin County were to some degree responsible for the high
rates there, one would expect women with breast cancer to have lived longer
in Marin County than control women […]. (engmed31)

This kind of interpretation uncertainty shows us that the author exploits the vague
reference inherent to the pronoun.
Since the French and Norwegian medical articles of the KIAP Corpus display
frequent use of indefinite pronouns, we have selected two articles, frmed22 and
nomed39, containing only indefinite pronouns (4 on and 11 ein respectively) in
order to study to what extent they may take on personal reference values. We have
already seen that French medical authors use on to refer to themselves to a rela-
tively large extent (see section 4.4.4). Frmed22 contains 4 occurrences of on and
no nous (‘we’) in subject position. Only one of these seems to refer clearly to the
authors only:
(17) Chaque médecin généraliste devait recruter les 10 premiers patients auxquels
il prescrivait un traitement de substitution par buprénorphine haut dosage.
On définit 3 sous-groupes […]. (frmed22)
‘One defines 3 subgroups …’
146 Academic Voices

However, even here we may ask whether this on includes a larger group than the
authors of the article. In the present case, a relevant question is whether the gener-
alists referred to in the previous sentence are included in on. We often encounter
this kind of interpretation uncertainty in medical articles. Indefinite pronouns (as
well as ‘we’) may refer to the authors + an unspecified number of other researchers
or hospital personnel included in some way or another in the research project in
question. We propose to call this value ON1+ (see also example (2) in 4.3.1 and
section 4.6.2).
The other occurrences of on in this article seem to be of the ON3 type, includ-
ing the relevant but indefinite discipline community, i.e. medical researchers who
take an interest in the reported project, as in the following example:
(18) Ces recommandations ont d’autant plus d’intérêt si on s’attache aux propos de
Greenfield [14]. Ce dernier précise que, […]. (frmed22)
‘These recommendations are even more interesting if one accepts what Green-
field [14] says …’

We may conclude that the authors of frmed22 manage well without using personal
expressions like personal pronouns or indefinite pronouns with a clear personal
reference. However, as emphasised many times throughout this book, the individ-
ual variation is considerable. By taking a quick look at another French medical ar-
ticle, frmed19, we see a quite different kind of manifestation. This article contains
5 nous (‘we’), all clearly referring to the authors of the article, mostly combined
with research verbs, displaying the researcher role. Like engmed31 (see above),
frmed19 also contains one example where the author assumes the writer role: nous
allons passer en revue quelques-unes des hypothèses … (‘we will list some of the
hypotheses’). As regards the occurrences of the indefinite on in this article, they
display a more heterogeneous picture. The 12 on pronouns seem to be used in the
following way: 2 ON1 referring to the author(s), 2 ON2 referring to the authors
+ the reader(s), 7 ON3 referring to the authors + the relevant indefinite discourse
community and 1 ON4 referring to people (here: the French) in general. Even
if the indefinite perspective is important, we see that the personal perspective is
admitted too. The ON2 occurrences are of special interest since they generally are
very rare in medical articles. Here is an example where the reader seems to be in-
cluded in the reference of on and directly invited to “conclude” preliminarily:
(19) On peut donc provisoirement énoncer de ce qui précède que si la peur du
SIDA est un déterminant de la réduction des SD et des autres phénomènes
étudiés dans notre pays, elle a eu un ”catalyseur” chez les injecteurs de drogues,
[…]. (frmed19)
‘One may thus preliminarily state from what precedes that if ...’
 Chapter 4. Presence of the author 147

Let us now look at the Norwegian article, nomed39 (written in Nynorsk), contain-
ing no vi/me (‘we’) but 11 ein (‘one’). The question we ask is the same as the one
asked in relation to frmed22: To what extent are these pronouns used with per-
sonal reference values? From another perspective we may reformulate the ques-
tion as follows: To what degree do medical authors manage without using personal
expressions? In nomed39, the authors allow the indefinite pronouns to assume
personal values. In fact 7 of the 11 occurrences are of the ON1 type. However, in
many of these cases it may be that it is not the pure ON1 value that is used, but
rather the ON1+, including an unspecified number of other researchers, hospital
personnel or generalists working outside the hospital:
(20) Spontan perforasjon av oesophagus (Boerhaaves syndrom), var utvilsamt
ikkje det første ein tenkte på, men samstundes er sjukehistoria nokså typisk
for denne tilstanden […]. (nomed39)
‘… was not the first thing one thought about …’

Interestingly, in addition to a couple of ON3 values, we also find two ON values, refer-
ring to other researchers and excluding the authors. Here is one of the occurrences:
(21) I ein studie med seks pasientar med oesophagusperforasjon (fem med Boer-
haaves syndrom), der fem av lesjonane var eldre enn 24 timar, behandla ein
med debridement og skylling og deretter primær sutur, som vart forsterka
med ein velvaskularisert stilka omentlapp. Alle desse pasientane vart utskrivne
innan 15 dagar (9). (nomed39)
‘ In a study … one treated …’

The reference to the other researchers is clear since a specific study is mentioned
and referred to by the number (9).
We see that nomed39 displays a less homogeneous picture as regards the op-
position indefinite–personal. The personal perspective (both self and other) is
clearly present even if there are no personal pronouns.37 The article contains other
expressions manifesting personal presence. These will not be commented on here,
but in the next section we provide another analysis of a medical article.

4.6.2 Implicit and explicit references to ‘we’ in a Norwegian medical article

In sections 4.3–4, we looked at how first person plural pronouns (‘we’) and in-
definite pronouns (‘one’) are used as grammatical subjects in research articles, and
4.2 introduced a set of roles that single authors assign to themselves in connec-

37. Further comparisons of French and Norwegian usage with regard to indefinite pronouns
are needed.
148 Academic Voices

tion with the use of first person singular pronouns (‘I’) – writer, researcher, arguer
and evaluator. This system can to some extent also be applied to the use of ‘we’
in articles written by several authors, although the reference of first person plu-
rals is much more varied than that of the first person singular (cf. 4.3). In 4.4, we
discussed a number of possible referential values for indefinite pronouns. Both
author roles and the reference of first person plural and indefinite pronouns will
be at issue in the present case study.
As shown in 4.1, multi-author medical articles have less frequent use of ‘we’
than articles from economics. In English and French, there is also less use of the
indefinite ‘one’ in (single- plus multi-author) medical articles than in econom-
ics and linguistics, but the disciplinary differences in Norwegian are not signifi-
cant. It is common knowledge that academic discourse tends to use more passives,
nominalisations and various other constructions that allow the agent of reported
actions to be left implicit, sometimes giving the texts an impersonal character.
Kinn (2004, 2005b), in a study of the presence of cognitive research agents in con-
nection with a selection of verbs in 60 Norwegian research articles from the KIAP
Corpus, shows that the ratio of passive to active clauses with such verbs is clearly
greater in medicine than in economics and (in particular) linguistics, and that, for
the active, there is also more use of ‘one’ in medicine.
In the present section, we examine one multi-author Norwegian medical
article, viz. nomed43. This article has 11 occurrences of vi (‘we’) and 6 of man
(‘one’). We will focus, above all, on how the authors describe processes where they
themselves (or a group of people that they belong to) are participants, i.e. partici-
pants that can be referred to by an exclusive vi (‘we’). The processes in question
are mainly parts of the research process, but some have to do with their textual
and medical-advisory relation to the readership. We will look at the sections of
the article in the sequence that they appear in: (an untitled) Introduction, Mate-
rial/Method, Results, Discussion and (a very short) Conclusion.
The Introduction can be analysed as having three parts: 1) definition and sub-
division of the disorder in question (hydrocephalus) as well as current treatment
options (two paragraphs), 2) treatment history (two paragraphs) and 3) introduc-
tion of the present study/article (one paragraph). The agents of the first part are
the medical expert community; these are never explicitly referred to. The example
in (22) is typical of this text part.
(22) Hydrocephalus behandles enten ved å eliminere tilgrunnliggende årsak slik
at hjernevæskesirkulasjon gjenopprettes, eller ved kirurgisk etablering av
permanent alternativ drenasjevei for cerebrospinalvæsken.
‘Hydrocephalus is treated either by eliminating … so that circulation … is re-
established, or by the surgical establishing of …’
 Chapter 4. Presence of the author 149

Three agent-suppressing construction types are employed: the passive (twice: be-
handles, gjenopprettes), an infinitival phrase (å eliminere …) and a nominalisa-
tion (etablering). The agents in the historical part are mostly medical writers and
practitioners of the past, several of whom are mentioned by name and with biblio-
graphical references. The presentation deals with two alternative procedure types
and reaches the present time twice. In both cases, the agent is generalised to man
(‘one’), referring to the international community of surgical experts. The examples
are given in (23) and (24).
(23) Fra slutten av 1970-årene har man foretrukket å shunte cerebrospinalvæsken
til bukhulen pga. faren for alvorlige komplikasjoner ved å ha drenet liggende
i blodbanen (6, 7).
‘Since the end of the 1970s, one has preferred to shunt … (6, 7).’
(24) I de etterfølgende år skjedde det en revolusjon innen endoskopisk teknikk,
og etter hvert skiftet man fra ventrikkelstomi med stereotaktisk teknikk til
endoskopisk teknikk som nå er blitt et etablert behandlingsalternativ ved hy-
drocephalus […].
‘In the following years …, and gradually one changed techniques …’

Due to the difference in grammatical tense in the immediate cotext (present per-
fect in (23), past in (24)), the first man more clearly includes the authors them-
selves than the second. That is, the reference is to ON3, but bordering on ON6 in
(24) (cf. 4.4).
When, in the last paragraph of the introduction, the authors address their own
study and article, they do refer to themselves by vi (‘we’) in an active sentence; (25)
is the entire paragraph. This is an example of vi with a writer role. Note that there
is a cohesive link between ‘the Neurosurgical unit’ in the first sentence and ‘we’ in
the second; they are approximately the same referent.
(25) Nevrokirurgisk avdeling, Rikshospitalet anskaffet for noen år siden utstyr for
endoskopisk tredjeventrikkelstomi. Vi presenterer i denne artikkelen våre re-
sultater ved behandling av hydrocephalus med denne metoden i perioden 1.1.
1999–31.12. 2000.
‘The Neurosurgical unit … some years ago acquired the equipment … We
present in this article our results of the treatment …’

The Material and Method section describes 1) the patients and the treatment
procedures and 2) the registration of data and statistical methods, i.e. two fairly
separate research stages. The agents here are the medical staff of the surgical unit
(a group that presumably includes the authors and some others) and the authors,
respectively. There is reference to a fairly large number of treatment, registration
150 Academic Voices

and calculation acts, and in all instances but one, passives, nominalisations and
other agent-suppressing devices are used. A typical example is given in (26).
(26) Det ble i denne perioden foretatt 136 prosedyrer på 120 pasienter, […]. 14
pasienter gjennomgikk to prosedyrer og en pasient tre. […] Figur 1 viser al-
dersfordelingen, og tabell 1 viser årsaken til hydrocephalus.
‘In this period, 136 procedures were performed on 120 patients … 14 patients
underwent two procedures, and one patient, three. … Figure 1 shows the age
distribution, and table 1 shows the cause of hydrocephalus.’
Here we have a passive (ble … foretatt), a morphologically active verb with a pa-
tient-role38 subject (gjennomgikk), two nominalisations (prosedyrer in both cases)
and two personifications (of a figure and a table). Most of the implicit references to
‘we’ concern the researcher role, but the personifications serve as substitutes for ‘we’
with a writer role. The single appearance of an explicit ‘we’ in this part of the article
is shown in its cotext in (27). This is an instance of vi with a researcher role.
(27) Ikke vellykket behandling ble definert som persisterende symptomer på
forhøyet intrakranialt trykk og/eller ytterligere utvikling av patologisk hode-
omkrets. Hos pasientene med ikke-vellykket behandling la vi inn shunt eller
gjentok endoskopiprosedyren. Alle pasientdata er oppdatert t.o.m. 30.6. 2001,
slik at oppfølgingstiden er 6–30 måneder etter ventrikkelstomien.
‘Not successful treatment was defined as … For patients with a non-success-
ful treatment, we inserted a shunt or repeated the endoscopy procedure. All
patient data are updated up to June 30, 2001, so that the follow-up time is 6 to
30 months after the ventriculostomy.’

This is from the part dealing with registration procedures, formulated mostly in
the passive, but the middle sentence interrupts this with additional treatment in-
formation. The deviating use of the active and the personal pronoun in la vi inn …
eller gjentok (‘we inserted … or repeated’) functions well as a formal marking of a
thematic sidetrack.
The Results section is almost complectely devoid of agents, being similarly
characterised by many passives and nominalisations. The beginning of (28) is typi-
cal, but the subordinate clause at the end presents an interesting departure from
the rest of the text.
(28) I fire tilfeller ble prosedyren avbrutt fordi det ikke var mulig å oppnå fe-
nestrasjon av gulvet i tredje hjerneventrikkel. […] I det fjerde tilfellet kollab-

38. The pun is hard to avoid.

 Chapter 4. Presence of the author 151

erte ventrikkelsystemet hos et prematurt barn under prosedyren slik at man

mistet oversikten over de anatomiske landemerker.
‘In four cases the procedure was interrupted because it was not possible to
achieve fenestration … In the fourth case, the ventricular system in a pre-
mature child collapsed during the procedure so that one lost view of the ana-
tomical landmarks.’
It is difficult to tell to what extent the referents of man (‘one’) here overlap with
the author group, i.e. whether it would have been an alternative for the authors to
use vi (‘we’) in connection with this reported treatment failure. It appears, though,
to be a special case of ON1. Of course, the use of ‘it was not possible’ without any
direct reference to those who failed to achieve fenestration is equally inexplicit. We
hasten to add that their reports of treatment successes, too, are agentless.
The authors are far more visible in the Discussion, although there are agent-
less expressions here, too, especially nominalisations. There are a few references to
the community of medical experts, where the indefinite man (ON3) is used a few
times, e.g. as in (29).
(29) Endoskopisk tredjeventrikkelstomi har i de senere år i økende grad blitt tatt
i bruk som et alternativ til shunting. […] Man håper at endoskopibehandlin-
gen hos utvalgte pasienter skal gi et like godt eller bedre funksjonelt resultat,
[…].
‘In the last years, endoscopic third ventriculostomy has to an increasing extent
been put to use as an alternative to shunting. … One hopes that endoscopy
treatment for selected patients will give an equally good or better functional
result …’
In (30), man presumably refers to the community of experts (ON3), although the
authors of the publications referred to are prominent and do not include the au-
thors of nomed43; i.e. the reference has a flavour of ON6.
(30) Men kun hos et fåtall av pasientene som er vellykket behandlet på denne måten,
finner man normalisering av ventrikkelstørrelsen, og hos noen pasienter for-
blir ventriklene uendret forstørret selv om behandlingen åpenbart har effekt
(15–18).
‘But only for a minority of the patients successfully treated in this way does
one find normalisation … (15–18).’

We note that the authors use vi (‘we’) when referring to decisions about treatment
policies at the surgical unit (special cases of the researcher role). An example is given
in (31), where their policy of the period in question is motivated with reference to a
152 Academic Voices

previous study, and a revised policy is introduced for the future because of reported
low shares of treatment success for a specific patient group in their own study.
(31) På bakgrunn av den positive rapporten fra Cinalli og medarbeidere valgte vi å
tilby endoskopisk behandling til alle aldersgrupper. Ved gjennomgang av vårt
materiale fant vi imidlertid en meget lav andel vellykket behandlede […]. For
fremtiden vil vi derfor fortrinnsvis tilby endoskopisk tredjeventrikkelstomi til
pasienter over seks måneder.
‘On the basis of the positive report from Cinalli and collaborators, we chose
to offer … But in the examination of our material, we found a very low share
… In the future, we will therefore primarily offer …’
Towards the end of the Discussion, the authors use vi in connection with expres-
sions of belief and opinion as well as recommendations directed at the presumed
primary readership, i.e. the community of neurosurgeons. Such cases of vi with an
evaluator and arguer role are illustrated in (32).
(32) Vi har fått stor tro på endoskopisk tredjeventrikkelstomi og anbefaler at me-
toden også prøves som første behandlingsalternativ hos pasienter over seks
måneder med kommuniserende hydrocephalus. Unntaket er etter vår men-
ing pasienter med tydelig ekstern supratentorial hydrocephalus. Vi mener at
det per i dag ikke foreligger sikre hydromekaniske tester som kan predikere
hvem som vil ha nytte av metoden.
‘We have developed great belief in endoscopic third ventriculostomy and
recommend that the method should be attempted … The exception, in our
view, is patients with … We believe that there are currently no …’

The entire Conclusion is quoted in (33). Here the authors do not refer explic-
itly to themselves, but are implicitly present as evaluators in both sentences, e.g.
in the epistemic verb synes (‘appears’; cf. Vold 2005) and in the positive adjectives
godt (‘good’) and oppmuntrende (‘encouraging’).
(33) Endoskopisk tredjeventrikkelstomi synes å være et godt behandlingsalternativ
til pasienter eldre enn seks måneder med behandlingstrengende ikke-kom-
muniserende hydrocephalus. Resultatene av prosedyren hos pasienter over
seks måneder med ikke-kommuniserende hydrocephalus er oppmuntrende.
‘Endoscopic third ventriculostomy appears to be a good treatment alternative
… The results of the procedure … are encouraging.’

To sum up, the authors of nomed43 mainly report treatment and research
procedures without making explicit who the agents are. This is achieved by, inter
alia, passives, nominalisations and non-finite verbal constructions. In most cases,
there is little doubt as to who the agents are, viz. the authors themselves or a collec-
 Chapter 4. Presence of the author 153

tive that they are central members of, but the pronoun vi (‘we’) is not often used. In
some cases, e.g. the event reported in (28) above, it appears likely that only some of
the authors (if any) are involved, but exactly who this is is admittedly not impor-
tant for the readers, and it would have been strange to use a more specifically refer-
ring expression than man (‘one’). It is interesting to observe in what contexts the
authors choose to use vi, viz. primarily in the Discussion and notably in reports on
decisions and in recommendations. They do not hesitate to enter the stage, so to
speak, when defending their views, but otherwise they do so only rarely.

4.6.3 A study of two English abstracts

In this case study of abstracts we have selected two texts from the English subcor-
pus. The contrasting variable is discipline, with one abstract from economics and
one from linguistics. Both texts are rendered in full below. These two particular
texts have been chosen because they enable us to illustrate the issues that are dealt
with in our discussion of abstracts (section 4.5), and because each is a fairly typi-
cal example of an abstract within the discipline they belong to in our corpus. We
start with the economics abstract, engecon04, which is written by a single female
author. Author roles have been added and will be referred to when we comment
on these issues below.
Economics
(34) How do foreign patent rights affect U.S. exports, affiliate sales, and licenses?
This paper analyzes how foreign patent rights (FPRs) affect US exports, affili-
ate sales, and licenses. Our approach is distinctive in three ways. We apply [re-
searcher] ownership, location, and internalization concepts to link FPRs with
servicing decisions. We account for [researcher] the simultaneity of servicing
decisions. We estimate [researcher] the relative effects of FPRs on exports, af-
filiate sales, and licenses. Empirical findings show strong FPRs increase US
affiliate sales and licenses, particularly across countries with strong imitative
abilities. Further, FPRs have a larger effect on US knowledge transferred out-
side the country and firm, relative to knowledge located inside the country
and internalized inside the firm. (abstract engecon04)

As stated in section 4.5.2, one of the very few comments regarding language-
related issues which journals tend to mention in their instructions to authors is
abstract length. In economics journals, 100 words is in most cases the required
154 Academic Voices

number.39 The text we are discussing here matches this almost perfectly: the ab-
stract is 101 words, and hence must in this respect be said to be a typical instance
of an economics abstract.
As regards author roles, the first thing to note is that the author is overtly
present in the abstract through the use of first person pronouns. As this is a single-
author text, the author has a choice whether to use the singular or the plural form.
The plural has been chosen, which must be considered the default choice in this
particular subcorpus40 (see the discussion of authorial ‘we’ in 4.3). As can be seen
from our labelling in the abstract, the author is visible as a researcher in all three
instances. (This is also the dominant role in the economics abstracts corpus as a
whole; cf. Table 4.5-1 in section 4.5.2.) In addition, she is present through the use
of the possessive our in sentence 2.41 We also see that the inanimate noun paper
is used with an animate verb, analyze, making the noun take on a personalised
meaning which has been shown to be typical for scientific language (Prelli 1989;
Master 2001). With three instances of a first person pronoun, author presence in
this abstract is stronger than the average for this subcorpus, which is 2.3 occur-
rences (Table 4.5-1).
Let us then take a look at how the author makes use of evaluative elements in
the abstract to persuade potential readers that it is worth taking the trouble to read
the accompanying full article. In our discussion of the abstract corpus in section
4.5, we focused only on evaluative elements linked to findings. In this section we
comment on promotion in a slightly wider sense. This particular abstract has one
clearly promotional element, viz. the whole of sentence 2: Our approach is dis-
tinctive in three ways. This sentence has an interpersonal metadiscourse function,
intended to persuade the reader of the original nature this research contribution
represents. The use of the ”undisputably evaluative” (Lindeberg 2004) adjective
distinctive makes this clear not only to the discipline professional, but also to the
uninitiated reader of this text.
Next, we turn to the linguistics abstract, engling17, which is written by two
authors.

Linguistics
(35) Reconsidering prepositional polysemy networks: the case of over

39. A well-known Norwegian economics professor once said in a discussion with us about
academic discourse that when writing an abstract, he always makes a point of making sure that
it ends up at exactly 100 words (Agnar Sandmo, personal communication.)
40. Three instances of the singular form I from three different abstracts are found in the abs-
tract subcorpus of 40 texts. Two of these are written by a female author.
41. Possessives are not included in our discussion in section 4.5, and do not affect the figures (cf. 4.1.1).
 Chapter 4. Presence of the author 155

This article explores lexical polysemy through an in-depth examination of the

English preposition over. Working within a cognitive linguistic framework, the
present study illustrates the nonarbitrary quality of the mental lexicon and the
highly creative nature of the human conceptual system. The analysis takes the
following as basic: (1) human conceptualization is the product of embodied
experience, that is, the kinds of bodies and neural architecture humans have,
in conjunction with the nature of the spatio-physical world humans inhabit,
determine human conceptual structure, and (2) semantic structure derives
from and reflects conceptual structure. As humans interact with the world,
they perceive recurring spatial configurations that become represented in
memory as abstract, imagistic conceptualizations. We posit that [arguer] each
preposition is represented by a primary meaning, which we term [researcher]
a PROTOSCENE. The protoscene, in turn, interacts with a highly constrained
set of cognitive principles to derive a set of additional distinct senses, forming
a motivated semantic network. Previous accounts have failed to develop ad-
equate criteria to distinguish between coding in formal linguistic expression
and the nature of conceptualization, which integrates linguistic prompts in a
way that is maximally coherent with and contingent upon sentential context
and real-world knowledge. To this end, we put forward [researcher] a meth-
odology for identifying the protoscene and for distinguishing among distinct
senses. (abstract engling17)
This abstract has 214 words distributed over 8 sentences. Instructions to authors in
English linguistics journals often indicate that abstracts ought to be between 100
and 200 words, and the abstracts in our linguistics subcorpus generally adhere to
this norm. Hence, this text is somewhat longer than what is typically the case. It
is, however, not the longest one in the English linguistics subcorpus; the engling24
abstract contains 294 words, while the engling23 abstract has as many as 336.42
The authors are present through a first person pronoun in three instances,
which in terms of the strength of the presence is practically identical with the aver-
age of 2.9 for this subcorpus (Table 4.5-2). As regards author roles, our labels show
that in the first occurrence the author takes on the arguer role (We posit that),
while the other two are instances of the researcher role (we term, we put forward).
That both the arguer and the researcher roles are present also corresponds well
with our expectations for such texts (Table 4.5-2). As was the case in the econom-
ics abstract above, here, too, we find a personalised use of article.43 Another thing

42. These three longest ones are from three different journals.
43. As regards the variation between article and paper in English, our studies of metatext in the
articles show that the economists almost exclusively use paper, while the linguists occasionally
use article, even though they, too, show a very clear preference for paper.
156 Academic Voices

to note in this abstract is the overt distinction between the research project and the
textual realisation of it, visible through the references to [t]his article (sentence 1)
and the present study (sentence 2).44
As the present abstract is quite long, the potential for finding traces of promo-
tional elements is higher than for shorter texts. In fact, an evaluative element is
found in the first sentence, viz. ”an in-depth examination”. Since a thorough analy-
sis is positive in all kinds of research, no discipline-specific knowledge is required
to recognise the value of this modifier. A clearly promotional effect is also created
in the following text passage from the second half of the abstract:
(36) Previous accounts have failed to develop adequate criteria to distinguish be-
tween [….]. To this end, we put forward a methodology for identifying the
protoscene and for distinguishing among distinct senses.

Here the authors are performing the rhetorical moves of establishing a niche and
then occupying the niche (Swales 1990). The positive evaluative effect is created
through the contrast between the knowledge level of the field before and after the
research reported on in the abstract is taken into account.

4.7 Concluding remarks

In this chapter, we have been concerned primarily with the textual presence of
authors as manifested in the use of first person pronouns as well as indefinite
pronouns meaning ‘one’. Taking into consideration the great referential flexibility
of both ‘we’ and ‘one’, we have also addressed the inclusion of reader and third
persons and various rhetorical functions of such pronoun uses. Thus, although
reader/writer interaction and the presence of others are the foci of attention in
the next two chapters, it will be evident that the various aspects of person mani-
festation cannot be kept entirely separate. As indicated already in Chapter 3 (see
Table 3.2-1), both the discipline and the language factor are co-determinants of
the frequency of use of both first person (‘I’ and ‘we’) and indefinite (‘one’) subject
pronouns. For the former, discipline is somewhat more important than language,
while for the latter, language has a much stronger effect. First person is further the
feature where individual variation is greatest, 73 % of the variance being unac-
counted for by reference to effects of discipline and language differences.

44. As will be seen in our discussion of metatext in Chapter 5, medical articles give first and
foremost the impression of being accounts of a research project, rather than texts in their own
right (cf. section 5.2).
 Chapter 4. Presence of the author 157

As expected, medicine has less use of the first person than the other two dis-
ciplines, which are not very different in this respect, although linguists do use
‘I’ more than economists. Overall, linguists also use indefinite pronouns more
than economists, and economists more than medical writers, but here there are
clear differences between the different language sublevels. It seems to be clear that
there is a weaker pronominal manifestation of the first person in medicine than
in economics and linguistics, while the difference between those two disciplines is
smaller than we had expected.
As regards the language factor, French has the least frequent use of the first
person, whereas English and Norwegian are fairly similar, with Norwegian tend-
ing to range the highest. As expected, English has far less use of ‘one ‘than the other
languages, while the relation between French and Norwegian varies depending on
the discipline. The complex relation between the usages of these two languages
indicates possible interaction with the use and non-use of the first person.
Our discussion in 4.2 of author roles posited four central types, viz. the author
as writer, researcher, arguer and evaluator, and it was shown that there are clear
differences between subcorpora in the distribution of these roles for ‘I’, and espe-
cially between the disciplines. Turning to ‘we’ and ‘one’, in 4.3 and 4.4 we looked
at a number of (in part very similar) referential possibilities that both of these
pronouns open up for, and we discussed a number of functions that they have
in research articles. In section 4.3, some aspects of the polyphonic properties of
‘we’ were also briefly discussed. Section 4.5 addressed author manifestation in
abstracts, an important genre outside of the research article body, which is our
primary focus of attention in this book, while several case studies of person mani-
festation were performed in 4.6.
Pronoun use has a central place in the rhetoric of the research article, as in
many – probably most – other genres. Our study of pronouns has mostly had its
focus on low-level phenomena: reference, author roles, verbs and metatextual ex-
pressions. Many other researchers, although also providing quantitative data (cf.
references in 4.0), have been more overtly concerned with rhetorical functions
than we have here. However, combining detailed studies of particular expression
types with large-scale quantitative analyses, we think that our perspective comple-
ments those previous studies in important ways and contributes significantly to
our collective understanding of pronoun use in academic discourse.
 chapter 5

Reader/writer interaction

5.0 Introduction

Having dealt primarily with the presence of the author(s) in the previous chapter,
we now turn to aspects of texts that are more closely connected to the interaction
between authors and readers, the interplay between self and other. We address the
third main question: How do the authors present and promote their own research?
Specifically, we look at metatextual expressions (such as ‘article’ and ‘below’; de-
tails about the items we have included in this investigation are given in section 5.2)
and expressions of the type ‘let us’ (and ‘let me’), as well as the overt textual presen-
tation of ‘results’ and ‘conclusions’. Hence, we discuss linguistic phenomena which
in different ways address the issue of reader/writer interaction. One of the most
classical instances of such interaction is metadiscourse. For reasons which will be
given below, we have in this study chosen to focus not on metadiscourse as such,
but (primarily) on the use of metatext. Metadiscourse is a wider concept which is
often discussed with a point of departure in Vande Kopple’s rather vague defini-
tion of it as “discourse about discourse or communication about communication”
(1985: 83). His somewhat tentative classification system developed in the same
work, as well as further developments in studies involving Crismore (Crismore
1989; Crismore & Farnsworth 1990; Crismore, Markkanen & Steffensen 1993)
are early well-known contributions to the field. The concept of metadiscourse is
clearly a relevant one in discussions of academic text,1 a fact reflected by the many
studies devoted to it (see below). Such studies very often take a contrastive angle,
focusing on either language or discipline. English is commonly used as a point of
reference, bearing witness to the importance of English as a lingua franca in the
global education and research community. Many of these studies have a pedagogi-
cal purpose as the most prominent one, often stressing that the findings may be
used to help non-native speakers of English in their publishing efforts. Interesting
studies of this kind are e.g. Mauranen (1993b), Valero-Garcés (1996) and Bunton
(1999), all on written discourse, and Mauranen (2002) and Thompson (2003), on
spoken discourse.

1. However, other genres and settings have been studied as well (e.g. Fuertes-Olivera et al. 2001).
160 Academic Voices

As witnessed by Vande Kopple’s definition rendered above, metadiscourse is a

concept which is difficult to delimit. It may be realised by various linguistic forms
and may fulfil a number of pragmatic functions in the text. Many metadiscourse
studies make use of the Hallidayan distinction between the textual and interper-
sonal macrofunctions of language (Halliday 1973), most of them with their own
modifications of Vande Kopple’s classification system. Textual metadiscourse, ac-
cording to Vande Kopple, refers to devices which primarily play the role of organ-
ising the text for the reader, while interpersonal metadiscourse is mainly used to
interact with the reader about the propositional content. This division of metadis-
course into two categories has given rise to the term metatext, used in studies
of textual metadiscourse (Mauranen 1993b; Valero-Garcés 1996; Moreno 1997;
Bunton 1999), while the term metadiscourse tends to be used in studies discussing
textual as well as interpersonal functions (Crismore 1989; Crismore & Farnsworth
1990; Hyland 1998b, 1999).
We may in a somewhat loose way say that metadiscourse represents overt
expressions of the writer’s acknowledgement of the reader. As indicated above,
the concept is a multifunctional one and may be realised through many different
linguistic forms. In this book, several phenomena are investigated which could
be classified as metadiscourse. Cases in point are first person pronouns and evalu-
ative expressions (both instances of interpersonal metadiscourse in the Crismore
framework, which is used in many studies on metadiscourse, e.g. Hyland 1998b,
1999) as well as references (textual metadiscourse subclassified as evidentials in
Hyland 1998b, 1999). We have decided to keep these phenomena apart, since we
find it important to study the different kinds of presence they manifest. We do,
however, here deal with ‘let us’-constructions, which are typically included in
studies of metadiscourse (e.g. Swales et al. 1998 and Hyland 2002b; references to
other studies of this construction will be given in section 5.3). As stated above, we
in this chapter also discuss lexical realisations of ‘results’ and ‘conclusions’, a topic
which to our knowledge has not been dealt with in a cross-linguistic or cross-dis-
ciplinary context before (an exception in the latter category is Grinde 2003; other
relevant studies will be referred to in the discussion in 5.4).
Based on previous studies of metadiscourse, our main hypothesis related to
language in this chapter was that reader/writer interaction would be more visible
in English (cf. e.g. Mauranen 1993b) than in French (cf. Valero-Garcés 1996, com-
paring English and Spanish, another Romance language; Moreno 1997, however, a
study of causal metatext, finds more similarities than differences between English
and Spanish from a quantitative point of view). For Norwegian, no previous stud-
ies exist, but our tentative feelings as native speakers were that Norwegian authors
are somewhat less visible in the text than their English-writing colleagues. As for
the discipline variable, Hyland (1998b) includes (applied) linguistics among the
 Chapter 5. Reader/writer interaction 161

four disciplines investigated (the other three being microbiology, astrophysics and
marketing). In that study, linguistics is ranked last in terms of textual metadiscourse
markers (1998b: 447), the category that comes closest to our metatext study. Our
pilot study (Breivega et al. 2002) did not yield clear indications of disciplinary differ-
ences. However, we assumed metatext to be more prevalent in linguistics and eco-
nomics than in medicine, based on the fact that medical writers in general seemed to
be less visible in their texts than writers from the other two disciplines (ibid.).
In the study of reader/writer interaction, a lot of implicit polyphony may take
place. However, what we focus on here is the explicit signalling undertaken by
the author(s) in order to bring the reader into the text. This, as already indicated,
can be done in many ways, for example by guiding the reader through the text by
means of metatextual expressions (section 5.2) or by calling directly on the reader
by ‘let us’-constructions (5.3). In a Bakhtinian dialogical perspective, this has to
do with the interlocutive influence the meaning of an utterance may undergo (see
Bres 1999). The locutor sets up the play in a way that anticipates the interaction of
a potential interlocutor or reader.
The chapter is organised as follows: after the presentation of quantitative and
qualitative results in section 5.1, we discuss metatext in section 5.2, ‘let us’-con-
structions in 5.3, while our discussion of ‘results’ and ‘conclusions’ takes place in
section 5.4. In section 5.5 we present three case studies, the first on metatext in
three English economics articles (5.5.1), the second on ‘let’-imperatives in one
Norwegian economics article (5.5.2), while the third deals with metatext and au-
thor manifestation in a French medical article (5.5.3). The chapter ends with a
concluding section summing up the main findings relating to reader/writer inter-
action (5.6).

5.1 Quantitative and comparative results

This section presents the results of our third main quantitative investigation, viz.
that of a selection of metatextual expression types. Statistical details can be found
in Appendix B. We emphasise that what we have studied is a selection, and this
is discussed in 5.2. The data are presented in Table 5.1-1. As we saw in 3.2, a very
large share of the variance in the relative frequency of metatext, 52 %, is due to dif-
ferences between disciplines. Another 12 % is contributed by language differences,
and 2 % to the interaction of the discipline and language factors. Hence, about
34 % of the variance is left unaccounted for.
162 Academic Voices

Table 5.1-1. The use of metatext

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 50 976 0.327 0.314 0.140 0.088

engling 50 1230 0.279 0.256 0.143 0.125
engmed 50 47 0.027 0.008 0.041 0.022
frecon 50 474 0.149 0.126 0.123 0.065
frling 50 187 0.076 0.054 0.082 0.056
frmed 50 8 0.007 0.000 0.017 0.000
noecon 50 697 0.223 0.209 0.107 0.059
noling 50 432 0.161 0.139 0.129 0.084
nomed 50 9 0.009 0.000 0.024 0.000

econall 150 2147 0.233 0.213 0.143 0.086

lingall 150 1849 0.172 0.138 0.146 0.096
medall 150 64 0.014 0.000 0.030 0.010

engall 150 2253 0.211 0.199 0.177 0.151

frall 150 669 0.077 0.039 0.103 0.060
noall 150 1138 0.131 0.108 0.133 0.113

KIAP 450 4060 0.140 0.095 0.151 0.115

Relative frequencies and deviations in %.

The discipline factor. In general, metatext is used more in economics than in lin-
guistics, and more in linguistics than in medicine. The differences between eco-
nomics and linguistics, respectively, as well as the differences between linguistics
and medicine, are significant at the superlevel and at the sublevel (in each lan-
guage) – except that the difference between economics and linguistics at the Eng-
lish sublevel is not significant.2
The language factor. Metatext is more frequent in English articles than in Nor-
wegian ones, and more frequent in Norwegian than in French. These differences
– between English and French, between English and Norwegian, and between
French and Norwegian – are significant both at the superlevel and at the sublevel
(in each discipline), the only exception being the difference between French and
Norwegian at the medical sublevel, which is not significant.
The gender factor. In some of the subcorpora, male authors use metatext more
than do female authors. The differences are significant in French economics and
Norwegian economics, and there is also a significant difference in French when

2. The terms superlevel and sublevel are explained in 2.2.5.

 Chapter 5. Reader/writer interaction 163

economics and linguistics are considered together.3 (Recall that the few medical
articles with one author are not included in our gender studies.)

5.2 Metatext

We will in this section discuss meta-elements which have a primarily textual func-
tion. An earlier study of metatext in roughly half the KIAP Corpus discussed a
much wider selection of items (Dahl 2004c). Many of these occurred only spo-
radically. Based on those findings we in this study have narrowed our list of search
items to those that turned out to be the most common ones in our texts.4 Table 5.2
presents the items that have been included in the present analysis.

Table 5.2-1. Metatextual search items in all three languages

English French Norwegian

article, paper article, papier artikkel

(sub)section (sous-)section avsnitt
above ci-dessus, plus haut ovenfor/ova(n)for, over
now maintenant, à présent nå
below ci-dessous, plus bas nedenfor/neda(n)for

These metatextual items are of two kinds. Items in the first group refer to the text
and specific parts of it by the use of ‘article’/’paper’ and ‘(sub)section’, while those
in the second express basic spatial relations within the text, pointing backwards
(so-called reviews), forwards (so-called previews) and to current point, through
the use of ‘above’, ‘below’ and ‘now’ (the latter of course also has a temporal aspect).
The main function of such meta-elements (apart from ‘article’/’paper’) is to help
the reader navigate the text. The category corresponds quite closely to what Hyland
(1998b, 1999), in his broader metadiscourse studies requiring a more fine-grained
classification, calls endophoric markers. Hyland’s studies do not explicitly mention
elements which refer to the text itself. In addition, we have, as already noted, in-

3. Mann–Whitney–Wilcoxon test, α = 0.05; see section 2.2.5.

4. In this earlier study (Dahl 2004c) we also looked at so-called discourse verbs (e.g. ‘describe’,
‘emphasise’, ‘summarise’ and ‘conclude’), which were classified as rhetorical metatext, while the
kind of metatext discussed here was termed locational metatext. These verbs are in the present
study discussed in connection with author roles in section 4.2.
164 Academic Voices

cluded ‘now’, which Hyland (1998b) classifies as a topic shift indicator, a subclass of
his frame marker category. Here are some typical examples from our material:
(1) The present paper studies a quantitative dynamic-optimizing business cycle
model of a small open economy in which nominal prices and wages are sticky.
(engecon01)
(2) Cet article propose une analyse théorique de ces politiques en considérant
que chaque hôpital peut agir sur son coût de production par un effort de ré-
duction du coût et par un effort d’accroissement de la qualité. (frecon35)
‘This article proposes a theoretical analysis …’
(3) I denne artikkelen vil vi presentere eksperimentelle resultater fra preteri-
tumsbøyning i norsk, fra barn i alderen 4-8 år og fra voksne. (noling07)
‘In this article, we will present experimental results …’
(4) Although this complication has not been observed in most of the clinical
reports cited above (also reviewed by Holland and Schwartz24), a recent re-
port of three cases did show that the donor eye developed pseudopterygium.
(engmed32)
(5) Une borne inférieure est obtenue avec une stratégie de tarif binôme comme
décrite ci-dessus. (frecon24)
‘… as described above.’
(6) Effekten av markedsstørrelse som ble drøftet ovenfor kombineres ofte med
diskusjoner av andre typer koblinger mellom bedriftene. (noecon03)
‘… that was discussed above …’
(7) The specimen was routinely fixed in formalin, processed to paraffin-wax
blocks and tested for its HPA-binding pattern according to the method shown
below. (engmed08)
(8) Le raisonnement ”par défaut” ci-dessous ne peut en conséquence avoir valeur
de preuve définitive. (frmed19)
‘The reasoning ”by default” below …’
(9) I setningen nedenfor regner vi med at jenta har blitt flyttet fra objektsposis-
jonen etter verbet ‘kysse’ og frem på subjektsplass, og et spor (t) står i objekts-
posisjonen: […]. (noling11)
‘In the sentence below, …’
(10) The analysis now turns to the optimal production policy of a regulated in-
cumbent threatened with entry. (engecon13)
(11) Je vais maintenant appliquer ce type d’explication à quelques uns des cas évo-
qués par D.L. (frling13)
‘I will now apply this type of explication …’
 Chapter 5. Reader/writer interaction 165

(12) Nå vil vi på mer allment grunnlag undersøke om den er riktig. (noling06)

‘Now we will … investigate whether …’
In addition to leaving out metatextual elements that our preliminary study showed
to be relatively rare in our material, such as e.g. ‘chapter’, ‘part’, ‘previously’, ‘so far’
and ‘already’ (Dahl 2004c), we also decided to exclude the potentially metatextual
‘here’ from the analysis. This word is − at least in English and Norwegian5 − cha-
meleon-like in nature, displaying, inter alia, locational, temporal and anaphoric
characteristics. Here are a few English examples illustrating this:
(13) The data presented here support Prince’s (1984, 1997) postulation of a distinc-
tion between LD1 and LD2. (engling21)
(interpretation: ‘in this article’)
(14) Here is a speculation about what might actually be going on. (engling46)
(interpretation: ‘now follows’)
(15) The important example here is (93), where the infinitive is controlled by the
subject of the sentence. (engling47)
(interpretation: ‘in this case’?)
(16) But he offers only one really strong argument in favor of not leaving copies with
A- movement, and that is the one which will concern us here. (engling49)
(interpretation: ‘in this article’?; ‘at this point’?)

The difficulties experienced in classifying each instance led to the decision of

leaving out ‘here’ altogether. We also decided to exclude the items ‘figure’, ‘table’,
‘equation’ and ‘example’ from our investigation. These words refer to phenomena
which are an integral part of academic presentation and the discussion of claims
and results. The first two can be found in all our three disciplines, while the third
belongs to economics and the fourth (primarily) to linguistics.6 In our view, these
words are not typical indicators of writer/reader interaction, since it is the subject
material which typically decides whether they are used or not.
We will now take a closer look behind the figures for metatext given in Table
5.1-1. As regards the use of ‘article’, the English texts have more occurrences than
the Norwegian (second in all three disciplines) and the French ones (third in all

5. The Norwegian texts contain 963 occurrences of her, the English texts 670 here, while the
French texts have 432 occurrences of ici.
6. In Dahl (2004c), involving roughly half the KIAP Corpus, only one occurrence of example
was recorded in the English medical texts. In the economics texts, 27 occurrences were found,
while in the linguistics texts as many as 261 occurrences of example were recorded (two English
linguistics articles alone contained 61 and 41 occurrences of example respectively). The same
trend was seen for French and Norwegian.
166 Academic Voices

three disciplines). As regards the discipline variable, the economists refer to the
whole text more often than the linguists, while in medicine there are hardly any
such references at all. As for ‘section’, this is used only in economics and linguistics.
In the medical texts (all three languages) there are no relevant hits.7 In economics
(all three languages) and in English linguistics, ‘section’ is much more frequent
than ‘article’. The economists in all three languages make use of the almost uniform
pattern of article structure presentation towards the end of the Introduction sec-
tion of the text. Swales (2004: 232) describes this as providing a roadmap of the
structure of the paper, and states that ”the occurrence of this step seems to be in-
versely related to whether the disciplinary field has an established IMRD-like sec-
tional arrangement” (ibid.). This pattern greatly contributes to the high number of
occurrences in the economics subcorpus:
(17) The rest of this paper is organized as follows. Section 2 briefly describes […].
We present our model of information sharing in Section 3, where we show
that […]. Section 4 describes some of the central bank’s activities that are
[…]. Section 5 concludes and suggests avenues for further investigation.
(engecon03)
(18) La section 2 est consacrée à la présentation du modèle […]. Dans la troisième
section, nous étudions la transition vers un système par capitalisation en […].
La quatrième section conclut l’article. (frecon09)
‘Section 2 is devoted to … In the third section, we study … The fourth section
concludes the paper.’
(19) Artikkelen er delt inn som følger. Avsnitt 2 presenterer modellen og gir […].
I avsnitt 3 benytter vi modellen til å analysere […]. Avsnitt 4 drøfter mulige
tiltak for å […]. Avsnitt 5 konkluderer. (noecon05)
‘… Section 2 presents the model … In section 3, we employ the model … Sec-
tion 4 discusses … Section 5 concludes.’

This feature is, however, much less universally present in the linguistics articles,
even though these to an even lesser degree than the economics articles are IM-
RAD-structured.

7. The search in the English medical corpus returned many hits for section, but they are all
references to Caesarean section!
 Chapter 5. Reader/writer interaction 167

As for the category of items specifying spatial relations within the text, it turns
out that reviews (i.e. ‘above’) are far more common than previews (‘below’). This
trend is valid for all the subcorpora, independently of language and discipline.8
The last three examples given above illustrate another phenomenon of
metatext/metadiscourse, viz. clustering (cf. e.g. Mauranen 1993b). Sentences con-
taining a meta-element are in fact very likely also to comprise other elements in-
dicating author presence. The following example from one of the Norwegian eco-
nomics articles may serve as an additional illustration of this point:
(20) Vi vil komme tilbake til dette punktet senere i kapittel 3 i denne artikkelen,
der vi ser på modellen til Greenwald og Stiglitz (1990). (noecon09)
‘We will get back to this point later in section 3 of this article, where we look at …’

As indicated in section 5.1, metatext is a feature that is highly sensitive to the

discipline factor. Economists use significantly more metatext (at least of the kind
we focus on here) than the linguists, and authors in both these two disciplines use
much more of it than the medical authors. That the medical texts turn out to be
very different from the texts in the other two disciplines is not very surprising.
Our pilot study showed the same trend (Breivega et al. 2002). In fact, the impres-
sion one gets from reading the medical texts is that the authors do not seem to
focus on them as texts in the same way as the economists and linguists do. This
is for instance reflected in practically a complete absence of references to the arti-
cles themselves, as already mentioned. In the medical articles, ‘study’ is frequently
used, a word which we have not included in our selection of metatextual search
items, as it in our view primarily refers to the (physical) research project rather
than the textual account of it.
As stated in section 4.5, medical reporting tends to be formally organised
through the IMRAD format. Hence, the content presented is forced into informa-
tion categories in a given sequence, and no extra processing effort is needed by
the expert reader in orienting him- or herself within the text. This explains the
complete lack of references to parts of the article (by means of the word ‘section’),
a word which we saw above occurs frequently in the other two disciplines. The
same applies to the use of reviews and previews. They are not needed, because the
reader already knows where in the text the information is to be found. It is in fact
a point for the medical writer to make sure that information belonging to the vari-
ous article sections is kept apart. As has been shown in a study of argumentation

8. Pisanski Peterlin (2005), in a study of previews and reviews in 32 English and Slovene
research articles from mathematics and archaeology, finds that in her corpus previews are twice
as common as reviews. One explanation given for this finding is that the articles are quite short
and hence reviews may seem unnecessary. The articles in the present corpus tend to be longer,
which may explain why our data yield the opposite result.
168 Academic Voices

strategies in Norwegian research articles (Breivega 2003), the strategy in medical

articles is not to create a high degree of functional connection in the text, but
rather to maintain a sharp distinction between research data and the interpreta-
tion of these.
In contrast to the medical writers, neither economists nor linguists have a
standardised text format to rely on (cf. section 4.5). The consequence is that ar-
ticles within these two disciplines have a more heterogeneous organisation. This
in turn has implications for how information can be tracked in the text. Metatext
thus becomes a very useful tool in offering assistance to the reader in the text
navigation process.
The finding that economists use significantly more metatext than linguists is
valid for French and Norwegian, but not for English (cf. sections 3.3 and 5.1).
However, there are statistically significant differences in frequency between the
languages, with most frequent use in English in all three disciplines. The Norwe-
gian authors in all three disciplines in turn use more metatext than the French.9
How can this be explained? Within the Anglo-American world, particularly the
US, much focus is placed on teaching students at various levels to write effectively.
This is witnessed by compulsory college composition classes. Emphasis is put on
communication with a reader, hence making this an explicit factor in the writing
process. The fact that this is taught explicitly in the educational system most likely
reflects a similar ideal prevalent in this society in general. In our discussion of na-
tional identity (section 1.3), we quoted from Mauranen’s (1993b) study of metatext
in English economics texts written by native speakers and Finns respectively. As
this quote is very relevant for our discussion here, it is repeated:
The careful and explicit guidance practised by Anglo-American writers, together
with frequent signalling of the personal presence of the author […] conveys the
impression that the reader is invited to take a tour of the text together with the
author, who acts as a guide. (Mauranen 1993b: 16)

Norway has less of a tradition in this matter. However, throughout the school sys-
tem today there is in general increased focus on writing as a process and on argu-
mentation, both orally and in writing. At university level there is also an increase
in courses and seminars on academic writing. It seems likely that this trend will
favour the role of the writer as someone who is clearly present in his or her texts
(Stålhammar 2002 reports a similar trend for Swedish). And as the results from
our investigation show, Norwegian authors within the disciplines of linguistics
and economics, too, clearly exploit metatext as a tool in helping their readers pro-
cess the text. On the basis of these results, we may claim that both English and

9. However, in medicine the difference between French and Norwegian is negligible.

 Chapter 5. Reader/writer interaction 169

Norwegian are representatives of so-called reader-oriented, or from another an-

gle, writer responsible, cultures. Hinds, in a paper discussing reader versus writer
responsibility, states that English clearly belongs in the second category: “English
speakers, by and large, charge the writer, or speaker, with the responsibility to
make clear and well-organized statements” (Hinds 1987: 143). Norwegian has here
been shown to belong to the same category. The following example from our Nor-
wegian economics subcorpus may serve as a good illustration of this:
(21) La meg derfor bare gi et lite hint om sammenhengen nå, og invitere leser-
en til å lese seg videre ned gjennom artikkelen for å få med seg detaljene.
(noecon07)
‘Let me therefore just hint at the connection now, and invite the reader to read
on through the article to pick up the details.’

It is of course quite possible that academic writing traditions in Norway over time
have been influenced by the Anglo-American tradition. To us as native speakers of
Norwegian it is, however, somewhat unexpected that Norwegian academic writers
are so similar to their English-speaking colleagues in this respect. French writers,
on the other hand, must according to our findings for metatext be classified as
belonging to the reader responsible cultures, in the company of e.g. Finnish and
Polish (Mauranen 1993b; Duszak 1994).
Hall and Hall (1990), discussing cultural differences between various national
cultures including, inter alia, the French, Americans and Scandinavians, intro-
duce the concepts of high context (HC) and low context (LC) communication.
They describe the French as representatives of a HC culture, stating that “most of
the information is already in the person, while very little is in the coded, explicit,
transmitted message” (ibid.: 6), while Americans and Scandinavians are said to be
representatives of LC cultures, implying that “the mass of information is vested in
the explicit code” (ibid.). Our findings here fit well with their description of these
cultures. We do, however, realise that as our analysis is based on form rather than
function, the inclusion of other metatextual elements specific to French might
have yielded a somewhat different picture (cf. Moreno 1998). As will become clear
in 5.3, the French are more active users of ‘let us’-imperatives than their English-
and Norwegian-writing colleagues, a fact that serves to modify our findings for
metatext.

5.3 ‘Let us’-imperatives

Text features related to the interaction between writer and reader and their inter-
action with the text have received markedly more attention in research on aca-
170 Academic Voices

demic discourse since the 1990s. The author wishes to create a convincing text and
to gain acceptance for his or her research and for him- or herself as a researcher.
Hence, there is a need to develop a suitable relation with the readers and let the
text address their expectations (see Bazerman 1988; Swales 1990; Hyland 2000,
2001a, 2002b). In this picture, features such as author presence and metatext be-
come important. Among the most explicitly interpersonal and textual features
used in this connection are ‘let us’-imperatives and other imperatives (including
others with ‘let’). Overt references to the reader by use of second person pronouns
or expressions like the reader are rather uncommon (although they do occur, pace
Swales (1990: 137); see example (21) in section 5.2). Hence, ‘let us’-imperatives are
in many cases the closest we get to overt addressivity.10

5.3.1 Types of ‘let’-imperatives

In English, there are several formal types of let-imperative (see Collins 2004 for
various classifications). We will simply distinguish between three types based on
the grammatical object of let (which is also the implicit subject of the following
infinitive). The first type is let us-imperatives, of which there are two subtypes, viz.
exclusive and inclusive let us-imperative, corresponding to exclusive and inclusive
we (see 4.3). The difference is whether the addressee is included or not, as illustrat-
ed in (1) and (2) below. The contracted form let’s is only used for the inclusive type,
as a slightly informal alternative to let us. The second type is let me-imperatives,
as illustrated in (3). The third type is let-imperatives with a third person object of
let, as illustrated in (4).
(1) Kim and I want to go to the park alone. Please, let us go!
(2) Why don’t you and I go to the park? Come on, let us (let’s) go!
(3) I want to go to the park. Please, let me go!
(4) Kim and Sam want to go to the park. Just let them go!

Let us-imperatives, let me-imperatives and let-imperatives with third person objects are
all employed in English academic discourse. Typical examples are given in (5)–(7).
(5) Having accounted for basic reconstruction effects with A-movement, let us
now go back to the data that motivated Chomsky’s and Lasnik’s conclusion
that there are no reconstruction effects with A-movement. (engling49)

10. We have not systematically investigated the frequency of ordinary imperatives. The main
reason is that they are not formally recognisable by any inflectional suffix or the like in English
or Norwegian and therefore not possible to search for automatically. Hence, a systematic inves-
tigation would have meant reading all the corpus articles in order to excerpt the imperatives.
 Chapter 5. Reader/writer interaction 171

(6) Let me summarize this section with a succinct statement by Joseph and Janda
on why we must abandon any appeal to ‘panchrony’ in linguistic explanation:
[…]. (engling33)
(7) To formalize this, let Tp, Te, and Ts denote the finite sets of distinct prefer-
ences, expectations, and socially relevant attributes found in the population.
(engecon30)

Our main focus will be on let us-imperatives and their closest correspondences in
Norwegian and French. Norwegian ‘let us’-imperatives are formally quite simi-
lar to English ones, using the imperative la (‘let’) (or, less commonly in Nynorsk,
lat), the pronoun oss (‘us’) and an infinitival verb phrase whose implicit subject is
coreferential with oss. Having no distinction like English let us vs. let’s, Norwegian
la oss-imperatives are basically ambiguous between an exclusive and an inclusive
reading. An example is given in (8).
(8) La oss så sjå på dømet lysbrytar, eit tilfelle av type c. i Kristoffersens typologi:
[…]. (noling32)
‘Let us then look at the example …’

No attempt will be made to distinguish between exclusive and inclusive ‘let us’-
imperatives in English or Norwegian (but it would seem that there are few, if any,
clearly exclusive examples).
French ‘let us’-imperatives are formally different and are only inclusive (at
least when used literally). French employs a special first person plural imperative
verb form in -ons, which is in most, but not all, cases identical with the first person
plural present indicative. The forms are nevertheless easily distinguished, since
the imperative does not have an overt subject, whereas nous (‘we’) is used with the
present indicative. An example is given in (9).
(9) Considérons à présent le problème du monopole à la date t = 0. (frecon23)
‘Let us now consider the problem of monopoly …’

We will study how ‘let us’-imperatives are used in research articles. In particular,
we ask what ”actions” the writer invites the reader to participate in, and we discuss
the functions of ‘let us’-imperatives from a pragmatic perspective as well as the
perspectives of text composition and rhetoric.
From a Searlean speech act perspective, we will regard most of them as being
typically combined directives (that they “are attempts by the speaker to get the
addressee to do something” (Levinson 1983: 240)) and commissives (that they
“commit the speaker to some future course of action” (ibid.)). This is in line with
Schiffrin (1994: 73). Collins (2004: 304) observes that they are “associated with
a range of illocutionary forces”. According to Kuo (1999: 134), the most impor-
172 Academic Voices

tant function of let us-imperatives is to seek agreement or cooperation. Hyland

(2002b) classifies them as directives. Swales et al. (1998: 107) state that in the aca-
demic texts that they have looked at, ordinary imperatives function as commands
or requests. Let us-imperatives, on the other hand, issue invitations. Some cases
that might be classified as declarations rather than directives or commissives are
discussed in 5.3.5.
In our corpus, there are 562 examples of ‘let us’-imperatives (see Table 5.3-1).
They are clearly most frequent in French, with 389 examples. There are 67 ex-
amples in English and 106 in Norwegian (counted as instances of let and la/lat,
respectively; the number of infinitives is slightly higher because of a few examples
of coordination). They are primarily used in economics and linguistics, with 284
and 267 examples, respectively. There are no examples in English or Norwegian
medicine, but 11 in French medicine.

Table 5.3-1. ’Let us’-imperatives

Economics Linguistics Medicine All disciplines

English 6 61 0 67
French 238 140 11 389
Norwegian 40 66 0 106

All languages 284 267 11 562

English: let us / let’s V; French: V-ons; Norwegian: la(t) oss V.

We have analysed this material statistically.11 We present the results of the analyses
here, but do not include the numerical details, since ‘let us’-imperatives are not one
of the main linguistic features investigated in our study (see 1.1 and Chapter 3).
The differences in relative frequency between economics and medicine and
between linguistics and medicine are significant both at the superlevel and at the
sublevel (in each of the languages).12 In addition, English linguistics has more fre-
quent use of this feature than English economics. The differences between eco-
nomics and linguistics are not significant at the superlevel or at the French and
Norwegian sublevel. (This is especially noteworthy in the case of the great differ-
ence in absolute frequency between French economics and linguistics.)
The differences in relative frequency between French and English and between
French and Norwegian are significant both at the superlevel and at the sublevel (in
each of the disciplines). In addition, the difference at the sublevel between Norwe-
gian and English economics is significant (the former having more frequent use).

11. Mann–Whitney–Wilcoxon test, α = 0.05; cf. 2.2.5.

12. For the terms superlevel and sublevel, see 2.2.5.
 Chapter 5. Reader/writer interaction 173

The rather large difference in absolute frequency that can be observed between
English and Norwegian at the superlevel is not significant. Neither is there any
significant difference at the sublevel between English and Norwegian linguistics,
nor (of course!) between English and Norwegian medicine.

5.3.2 A classificatory overview of different uses

We will now try to draw a picture of how these imperatives are used. The examples
in the corpus involve at least 125 different verbs in the three languages. They do
form some rather evident clusters, but – needless to say – a clear and exhaustive
classification of examples is hardly feasible. It might be more useful to think of the
groups as clusters of varying coherence, with outliers and connections to other
clusters. In spite of the classificatory difficulties, we posit eight functional types
for which we use the terms structuring, awareness, scope, analysis, manipulation,
definition, imagination and demonstration. This taxonomy builds on, extends and
modifies the one in Kinn (2005c), which is a study of all three kinds of ‘let’-im-
peratives in Norwegian, based on 60 KIAP Corpus articles. Below we will deal
with the eight functional types in turn.
Hyland (2002b) distinguishes between three main types of directive speech
acts in written texts based on the types of actions that they are intended to induce:
textual ones, which guide the reader through the text or direct him or her to other
texts, physical ones, which demand bodily action, and cognitive ones, which guide
the thinking of the reader. As will become evident from the examples below, the first
and the third type dominate among the ‘let us’-imperatives of research articles.

Structuring
This type directs attention to the formal structure of the text itself as well as the
relation of formal structure to thematic structure. Above all, the expressions refer
to where writer and reader are in the text and where they are going. This kind of
metatext makes the structure of the text clearer to the reader and involves expres-
sions of beginning, going on and concluding, as well as stopping, returning etc.
Examples (10)–(12) are of this type.
(10) Commençons par illustrer la question en parlant d’exemples non préposition-
nels. (frling45)
‘Let us begin by illustrating the issue …’
(11) Men la oss gå tilbake til MÅ igjen. (noling36)
‘But let us return to MÅ (‘must’) again.’
174 Academic Voices

(12) Let us conclude this section with a general observation about the methodo-
logical consequences of the fact that unidirectionality is not true across the
board. (engling33)
This type normally involves verbs that Bunton (1999) classifies as text act markers
(for further references to similar concepts, see his article).

Awareness
This type serves to direct the attention of the reader to information that the author
wants him or her to have in mind. It includes expressions of mentioning, noticing
and keeping in mind. Thus, some of the verbs in this cluster refer to enunciative
acts, while others refer to sensory or mental observations. This kind of imperative
is especially common in French, where notons (‘let us note’) is the most frequent
verb; rappelons (‘let us remind’), remarquons (‘let us remark’) and citons (‘let us
mention (quote)’) are also among the more commonly employed French verbs.
Some examples are provided in (13)–(15).
(13) Notons enfin que les distorsions entre prix et coût marginal sont d’autant plus
importantes que la demande est inélastique [...]. (frecon24)
‘Let us finally note …’
(14) Signalons enfin que, dans les deux méthodes, les paramètres λij sont estimés
en premier, […]. (frecon40)
‘Let us finally point out that, …’
(15) Enfin, la dernière explication considère la concurrence spatiale avec inter-
actions stratégiques entre les entreprises (n’oublions pas qu’en concurrence
monopolistique, les interactions stratégiques sont inexistantes). (frecon36)
‘… (let us not forget that …).’

Scope
This type serves to fix, narrow or broaden the focus of the ongoing discussion. It
is related to the preceding type, but whereas awareness is primarily about calling
something to the attention of the reader, the scope type has to do with delimita-
tion of the focus of attention. The affinities between the types pertain especially to
visual verbs, where the distinction between awareness and scope is hard to draw.
This type is common in all three languages, with verbal expressions like consider,
take, see, look at, concentrate on and focus on (and the corresponding French and
Norwegian ones). Some illustrations are given in (16)–(18).
(16) Let us focus on how universal quantifiers can scope under negation when
they are ‘clause-mates’ […]. (engling49)
 Chapter 5. Reader/writer interaction 175

(17) À titre d’exemple, intéressons-nous plus particulièrement au cas des modèles

incorporant l’hypothèse d’un processus d’appariement entre des chômeurs et
des emplois vacants […]. (frecon26)
‘As an example, let us take specific interest in the case …’
(18) La oss se på ett eksempel til. (noling36)
‘Let us look at one more example.’

Analysis
This type comprises various kinds of expressions that refer to cognitive operations
and their verbal expression, especially the analysis of data. While the preceding scope
type has to do with what should be attended to, the present type includes expressions
that refer to attempts at gaining an understanding of the matter. The analysis type
includes verbs like study (bordering on the scope type), analyse, compare and discuss,
as well as more specific ones. Some examples are provided in (19)–(21).
(19) Let us examine (108). (engling46)
(20) Étudions donc séparément les sous-cas de type a […] et ceux de type b […].
(frecon10)
‘Let us therefore study separately the subcases …’
(21) La oss sammenlikne verbene dette og rette: […]. (noling05)
‘Let us compare the verbs …’
Manipulation
This type is related to the preceding category, but comprises expressions that refer
to more specific actions dealing with the thematic matter. The type includes ref-
erences to methodological acts, which tend to be rather discipline-specific. (The
term manipulation should not be understood in its negative sense, but rather in
the direction of ‘handling, changing’.) Some illustrations are given in (22)–(24).
(22) Calculons tout d’abord la stratégie de prix optimale en l’absence de toute coa-
lition. (frecon24)
‘Let us first calculate the strategy …’
(23) La oss for klarhets skyld sette opp paradigmene for de personlige pro-
nomenene i entall: […]. (noling01)
‘Let us, for the sake of clarity, establish the paradigms …’
(24) Choisissons A = n(1 – ε)μ et utilisons l’inégalité de Bienayme-Chebyshev
pour obtenir la majoration : […]. (frecon24)
‘Let us choose A = n(1 – ε)μ and use the inequation …’
176 Academic Voices

Definition
This type has to do with decisions on how the subject matter should be categorised
and talked about. In particular, it includes references to terminological acts, theo-
retical assumptions and the creation of analytic models; the latter subtype is, of
course, closely related to the methodological acts of the previous type (manipula-
tion). Examples (25)–(27) illustrate this.
(25) Let us call such a type of phenomenon a ‘distinct process’. (engling33)
(26) Posons de plus que c est nul. (frecon23)
‘Let us further assume that c is zero.’ (‘Let c = 0.’)13
(27) La oss anta at den enkelte konsument betrakter funksjonen for miljøkvalitet i
likning (2) som gitt. (noecon34)
‘Let us assume that …’

Imagination
This type, illustrated in (28) and (29), has to do with the mental creation of imagi-
nary states of affairs. It is related to those varieties of the manipulation type that
have to do with establishing models.
(28) La oss som et eksempel tenke oss at vi har modellert etterspørsel etter en vare
betinget med hensyn på prisen. (noecon40)
‘Let us, as an example, imagine that …’
(29) Imaginons, pour commencer, que la gravité de la catastrophe augmente.
(frecon31)
‘Let us, to begin with, imagine that …’

Demonstration
This type is about showing how things are and about illustrating certain phenom-
ena. It is related to both awareness (noticing how things are) and analysis (finding
out how things are). The type is found primarily in French; two examples are given
in (30) and (31).
(30) Montrons, maintenant, que la situation où tous les individus choisissent la
stratégie π = 1 correspond également à un équilibre de Nash. (frecon34)
‘Let us now show that …’
(31) Illustrons cette analyse sur l’exemple: […]. (frling13)
‘Let us illustrate this analysis with the example: …’

13. See the discussion of declarations in 5.3.5.

 Chapter 5. Reader/writer interaction 177

5.3.3 Functions of text composition

The classification of ‘let us’-imperatives into eight usage types above is based on
the verbs and verbal constructions, that is, on the authors’ explicit invitations. But
at a different level, these directives-cum-commissives all appear to have one com-
mon function: By using them, the authors seek to make the reader aware of how
the text will be developing. This function is inherent to the imperatives themselves
(and hence cannot serve as the basis for subclasses): An utterance that presents
the receiver with an offer or a demand claims his or her attention, since he or she
needs to accept the offer or comply with the demand. This can only take place after
the actual utterance, and in academic texts, this future-directedness orients the
reader towards the development of the text.
In Kinn (2005c) it is shown that there is a striking degree of co-occurrence be-
tween ‘let us/me’-imperatives and metatextual expressions; 53.9 % of the impera-
tive sentences contain metatextual adverbials. (Only constituents with an adver-
bial function were counted.) These metatextual elements were found to be of two
types: locational metatext, which refers to a part of the text or to text sequencing
(e.g. ‘in this section’, ‘first’; cf. 5.2) and motivational metatext, which provides a
reason for the act proposed by means of the imperative (e.g. ‘for the sake of clar-
ity’). (These types overlap to a considerable extent with Hyland’s (1998b, 1999)
categories of logical connectives, frame markers and endophoric markers, but his
categories are more diverse in terms of grammatical status.) Several of the exam-
ples in 5.3.1–2 contain locational metatext (e.g. now in (5), tout d’abord (‘first’)
in (22) and så (‘then, further’) in (8)). Examples of motivational metatext are for
klarhets skyld (‘for the sake of clarity’) in (23) and perhaps à titre d’exemple (‘as (to
give) an example’) in (17).
Kinn (2005c) also showed there to be a high degree of co-occurrence between
‘let us/me’-imperatives and (in particular) beginnings and ends of paragraphs
(first and last sentences), as well as a certain tendency for the imperatives to ap-
pear in the first or last paragraph of a section. Thus, there is an evident connection
between the use of such imperatives and textual transitions, something which is
also indicated by the functions noted by Swales et al. (1998: 107). This, of course,
ties up with the fact that the imperatives direct the reader to what follows, either
internally in the paragraphs or sections where they occur or in the immediately
following ones. Such textual means are useful above all when the text assumes a
new direction, and to signal thematic changes is also the function of textual tran-
sitions between paragraphs or sections (although, naturally, there is not always a
nice and tidy fit between the thematic and the formal structure of texts). We return
to these matters in the case study on ‘let’-imperatives in 5.5.2.
178 Academic Voices

5.3.4 Rhetorical functions

The author and the reader of a research article perform different actions in relation
to the text. The author has done research that he or she tries to write a compre-
hensible and convincing article about. Presumably, the article is typically read be-
cause the reader wants to gain insight into the research and the author’s mediation
of it, out of interest and expected gain. Normally, the author and the reader will
be members of a common disciplinary community, but not necessarily any more
closely connected than that – socially, temporally or spatially. However, the author
can make use of linguistic means that establish a dialogue across time and space,
for instance by addressing the reader directly. ‘Let us’-imperatives (as well as ‘let
me’-imperatives) are excellent examples of such means.
The use of ‘let us’-imperatives not only implies directly addressing the reader,
thus establishing a dialogue: The author constructs, at least in the case of the inclu-
sive type, an author–reader togetherness and creates a kind of solidarity between
the two (see Myers 1989). They become, so to speak, collaborators in the structur-
ing of text and in mental activities, and when we are talking about research arti-
cles, these mental activities may be part of the research itself. By using the inclu-
sive ‘we’ and inclusive ‘let us’-imperative, the author draws the reader closer and
partly erases the author–reader asymmetry. From the perspective of politeness,
this can be seen as an attempt to give the reader a feeling that the author shares
his or her knowledge, opinions, aims etc. This is positive politeness (see Brown &
Levinson 1987).
The argumentative function is evident, because, given that they view things
in similar ways, the author can more easily convince the reader. In addition, the
author brings the reader along in the construction of an understanding of the data
and in the production of research results: When the reader has contributed in the
intersubjective construction of a piece of knowledge, he or she will more easily
accept it as true knowledge. ‘Let us’-imperatives both direct the reader and invites
him or her into a common space, and this combination potentially has a strong
rhetorical effect (see Mühlhäusler & Harré 1990: 174), directing text reading as
well as understanding. Such rhetorical functions are important, not least in edu-
cation, and some of the examples of ‘let us’-imperative are reminiscent of usage
associated with pedagogical genres such as the textbook and the lecture (see Wales
1996: 65–66).
These imperatives aid the reader. But the principal rhetorical aim of the author
in using them is, no doubt, to convince. The reader is to become interested in the
subject, accept the author’s treatment of it and acknowledge the author and his
or her research and texts. Whether these aims are always achieved is, of course, a
different matter. Some readers probably experience inclusive ‘we’ and imperatives
 Chapter 5. Reader/writer interaction 179

as impositions, especially if they do not agree with the author. Whether such reac-
tions are common is harder to estimate.
We will take a closer look at the use made of ‘let’-imperatives in the case study
in 5.5.2.

5.3.5 Variation between disciplines, languages and individual authors

The quantitative data in Table 5.3-1 above reveal clear differences between eco-
nomics and linguistics on the one hand and medicine on the other. There are very
few examples of ‘let us’-imperatives in medicine, and they are all from the French
subcorpus. This finding is in line with the general tendency for medical authors
to use overtly personal expressions less frequently than economists and linguists.
It also fits well with our impression from 4.3 (which we have not tested quantita-
tively) that the ratio of inclusive to exclusive ^[we^] is smaller in medicine than in
the other disciplines. The (near-) absence of ‘let us’-imperatives in medical articles
is discussed in section 7.1.
As noted, there are 11 instances of ‘let us’-imperatives in the French medical
articles. It is interesting to note that all of these actually belong to the awareness
type posited above, a type that is virtually absent in English and Norwegian. The
verbs that are used are notons (‘let us note’; 3 examples), rappelons (‘let us remind
(ourselves)’; 3), soulignons (‘let us underline/emphasise’; 2), citons (‘let us men-
tion’), remarquons (‘let us remark’) and signalons (‘let us point out’). These are all
used to call attention to or stress some piece of information which the authors
treat as a fact. The imperatives are not limited to this type in French economics
and linguistics.
In addition to differences between the disciplines, there are also considerable
differences between the languages: French uses ‘let us’-imperatives more frequently
than English and Norwegian. This is also true for medicine, where only the French
subcorpus has any examples. As already noted, some of the eight types above are
used primarily in French, or more in French than in the other languages. The dif-
ferences between the languages are greatest in economics. In addition to ‘let us’-
imperatives, Norwegian economists use inclusive ‘we see’ more than their English-
writing colleagues; both of these features contribute to a relatively reader-oriented
style in the Norwegian economics texts. The high frequency of ‘let us’-imperatives
in French economics (when compared to the other languages) is partly due to the
fact that they are sometimes used in French where ‘let’ plus a third person object is
used in English and Norwegian, as in (7) above. A French example is given in (32),
where English could have used let plus the equation.

(32) Posons w( x ) = x /(1 − x ) − (1 − Fz A (g + ( x ))) /(1 − Fz B (g − ( x ))) (frecon45)

180 Academic Voices

As noted by Kinn (2005c), in terms of speech act theory, such utterances are often
best classified as declarations rather than directives or commissives, since they “ef-
fect immediate changes in the institutional state of affairs” (Levinson 1983: 240).14
For instance, by uttering (32) the author establishes a state of affairs in his math-
ematical model. Such expressions are characteristic of academic texts from fields
that employ mathematical or logical modelling (see also Swales et al. 1998).
For ‘let us’-imperatives, as for the other linguistic features that we look at,
there is much variation from article to article within each of the language–disci-
pline subcorpora (except for English medicine and Norwegian medicine, where
the articles uniformly lack such imperatives). A total of 313 out of 450 articles have
no examples of this feature, i.e. 69.6 %. 43 articles have one example, 26 have two,
and so on. The frequency curve falls fairly evenly down to one article that has 20
examples – and then there is one article (frecon24) with the rather extreme fre-
quency of 38 examples (a relative frequency of 0.32 %). That is, most articles lack
this feature, some have a few scattered tokens, and a limited number of articles
have many examples. The use of the feature thus appears to be a personal style
trait that many authors do without, but which some find useful. It is possible that
frequency covaries with text type – e.g. it may be that ‘let us’-imperatives are more
common in argumentative text. Also, the history of the individual text may be im-
portant. For instance, it could be the case that the feature is more frequent in texts
that have been developed from an oral genre, e.g. a conference paper. These are
questions that would deserve looking into, but which we will not go deeper into
here. We do, however, return to the subject of ‘let’-imperatives in 5.5.2, where we
study their use in one particular Norwegian economics article.

5.4 Presentation of “results”

5.4.1 Introduction

The study undertaken in this section relates to our third research question, con-
cerning the author’s presentation and promotion of own research in research ar-
ticles (see section 1.1). The entire research process on which an article is based
is normally of interest to colleagues working on related phenomena. However, it
is the results which are often asked for first. In fact, the presentation of results

14. Other ways of expressing such declarations are also found, like ordinary indicatives (e.g.
English we let) and, in French, the subjunctive soit (‘(it) be’). It could also be argued that utte-
rances with English let us assume and the (roughly) equivalent Norwegian la oss anta belong in
the category of declarations.
 Chapter 5. Reader/writer interaction 181

constitutes a central element in the interaction with the readers, expert as well as
non-expert.15
In today’s competitive research communities, there is a constant demand for
results, preferably results which are original, important and far-reaching. The de-
mands come from different contexts: the researchers’ academic institutions, their
“home field”, media and journals.
Generally speaking, research constitutes an activity which, by definition, in-
volves formulating questions and finding answers to these questions, i.e. search-
ing for results. However, the nature of the results is different given different fields
and different research designs (see section 3.1). This self-evident fact allows us to
present the hypothesis that there will also be differences in the verbal presentation
of research findings. From the perspective of the rhetoric of science, we propose
that different styles are chosen “because of their presumed persuasiveness” (Prelli
1989: 104). The differences may relate to many factors; in this section we will focus
on the degree of explicitness and on lexical choice. Let us first look at the follow-
ing, constructed, utterances:
(1) The results indicate that Norway is a rich country.
(2) It seems possible to conclude that Norway is a rich country.
(3) I argue that Norway is a rich country.
(4) Norway is a rich country.

The claim which is presented is the same (‘Norway is a rich country’), but the
presentation forms are quite different. We may find all four types in any research
article, but our knowledge about different research and writing traditions would
probably influence us in the interpretation of their origin. However, one might
ask which utterance will have the best effect, i.e. which one will contribute most
directly in the process of persuading the reader?16 Which one will win, (1), (2), (3)
or (4)? This is an interesting question, but one which goes beyond the purpose of
this section. Our primary aim is not to evaluate the effects of scientific discourse,
but to find out more about the creative or inventive side of this specific kind of
rhetorical discourse (see Prelli 1989).

15. To a non-expert journalist, for example, a clear and explicit communication of research
results is essential. At a seminar on scientific writing traditions held at the University of Bergen
in April 2003, the journalist Normann Kirkeeide (from the local newspaper Bergens Tidende)
said that even though research texts are admittedly heterogeneous and differ according to the
field they belong to, journalists expect to find a conclusion where one has been announced. See
Kirkeeide (2003).
16. See Flyum (2004).
182 Academic Voices

In the present section, we want to investigate how researchers cope linguisti-

cally with this pressure. This particular study contains a quantitative part based on
the whole corpus consisting of 450 research articles, a lexical-semantic part based
on the English subcorpus (150 articles), and a qualitative part, based on a selection
of three linguistics articles, one English, one French and one Norwegian. The in-
vestigation itself will adopt two approaches: quantitative and qualitative. In section
5.4.2 we will first show that there are quantitative differences between languages, as
well as between disciplines in the use of the word family RESULT (including verb,
noun and their derivatives and corresponding expressions in the other languages;
in the following, RESULT implies the word families in the three languages). Next,
we will focus on the ways of expressing “final results” in English-language articles.
Some comments on evaluative expressions will also be given. Section 5.4.3 is a
discussion based on the quantitative investigation, and we suggest there that it is
necessary to undertake qualitative analyses in order to get a better picture of how
researchers express their results. This will be illustrated by an analysis of three lin-
guistics research articles written in English, French and Norwegian, respectively
(section 5.4.4). The texts are analysed with a view to find out more about the rhe-
torical and argumentative strategies and styles that the authors have chosen.
To our knowledge, there are no previous studies focusing particularly on the
use of RESULT in research articles. Swales (1990: 173) said in his presentation of the
Discussion section of the research article (where results are usually discussed) that
“we know little about disciplinary variation and little about the linguistic exponents
of the moves [of the discussion]”. Since then, the number of studies in this field has
increased (see Ruiying and Allison 2003 for a useful review of relevant literature).17
Most of the studies, however, are related to organisational patterns of the research
article structure, and most often to the moves within relevant sections like Results,
Discussion and Conclusion, i.e. sections of the IMRAD structure.18 Questions re-
lated more specifically to lexical choice or to the linguistic realisation of presenting
results are less studied than the organisational patterns. In this context, however,
we would like to mention the semantic study undertaken by Cavalla & Grossmann
(2005), even if it does not concern RESULT. They focus on “scientific nouns” used in
French research articles (their corpus also includes the KIAP articles), like hypothèse,
postulat, thèse and théorie; some of their findings indicate sizable differences between
disciplines in the use of these nouns. Grinde (2003), in her master’s thesis related to

17. See also Brett (1994) and Lindeberg (1994, 2004); for evaluative aspects, see Hunston
(1993), Stotesbury (2003) and Koutsantoni (2004).
18. It is surprising that so little has been done on articles that do not follow the IMRAD struc-
ture, but instead display the classic structure of Introduction, thematic middle sections and
Conclusion (in one form or another), typically used in non-“hard” disciplines.
 Chapter 5. Reader/writer interaction 183

the KIAP project, studied some of these and similar “keywords” in 10 articles from
each of the three fields studied in KIAP: hypothèse, modèle, méthode, résultat and
théorie. She found that economists often used modèle, the linguists hypothèse (but
not résultat), and the medical researchers often used both méthode and résultat. The
noun théorie is very little used, except for one single linguistics article which traces
the history of a specific theory. Grinde’s observations are interesting and reflect some
of the particularities of the research processes characterising these fields.

5.4.2 Quantitative analysis of RESULT

Introduction
The first step of our study is related to the degree of explicitness, aiming at answer-
ing the question of how results are expressed in scientific discourse. It is a quanti-
tative analysis of the frequency of the group of lexemes result (noun), result (verb)
and various adjectives like resulting, resultant, resultative and their corresponding
forms in French (noun: résultat) and Norwegian (noun: resultat).19 As mentioned
above, all of these lexemes will be referred to as RESULT.
Although limited in scope, our investigation will show that the study of the
presence of RESULT can tell us quite a lot about explicitness. Moreover, the fact
that RESULT appears in a number of different contexts also provides relevant in-
formation on the nature of the research article. Even more interesting is perhaps
the fact that there are considerable differences between languages, disciplines and
individual articles.
The use of RESULT depends, of course, on the type of research presented in the
article at hand. In one perspective, research can vary from nomothetic (relating to
or involving the search for abstract universal principles) to idiographic (relating to
or involving the study of individual phenomena). Another perspective is the issue
of theoretical versus empirical research. Further, there are important distinctions
to be made between, for example, quantitative and qualitative and between ex-
perimental and non-experimental research. These distinctions, and others, which
can be combined in different ways, are relevant to our investigation. Generally,
one can say that research based on quantitative methods will bring about findings
which are naturally expressed by RESULT. This will apply to most of the articles
in the medical subcorpus. As regards the economics subcorpus, there are many
articles presenting research involving equations, calculations and data subject to
quantitative methods. In linguistics, there is great variation. There is, of course, not
only qualitative research there, but also research based on quantitative and experi-
mental methods. Most of the articles selected for our linguistics subcorpus do not

19. We have also looked at the frequency of the word finding; see below.
184 Academic Voices

deal with general linguistics, but with questions related to the national language
in which they are written. This may give the articles a more idiographic character
than articles in general linguistics. Given our knowledge of the corpus, we would
expect the following ranking: RESULT will be more frequent in the medical and
the economics articles than in the linguistics articles.
As regards possible language differences, it is more difficult to posit a hypothesis.
In general, there is little research on this issue; however, if we look at previous studies
based on the KIAP Corpus, some indications may be found. Our studies on the use
of personal pronouns and metatext, for example,20 suggest that English researchers
are the most explicit; close behind are the Norwegians, while the French are the least
explicit. To the extent that RESULT can be considered as a mark of explicitness, we
would expect English to use it most frequently and French to use it least frequently.

Quantitative results
The suggestions above are to a large extent confirmed by the findings provided by
the automatic search for RESULT (result.* in English and Norwegian and résult.*
in French), as shown in Tables 5.4.1–3.21
A general summary of Tables 5.4-1–3 can be formulated as follows: For the
language dimension, English has the highest frequency (0.20 %), Norwegian is
second and French third. For the discipline dimension, economics has the high-
est frequency (when all the articles are taken together, this gives a total of 0.25
%), medicine is second (0.14 %) and linguistics is third (0.09 %). When language
and discipline are considered together, English economics ranks as number one
(0.32 %) and French linguistics is ranked the lowest, as number nine (0.04 %).

Table 5.4-1. Frequency of RESULT in English subcorpus

Subcorpus N %

engecon 957 0.32

engling 494 0.11
engmed 313 0.19

Total 1,764 0.20

Search string: result.*

20. See Fløttum (2003b); Kinn (2005a); Dahl (2004c); Chapter 4 and sections 5.1–2. above.
21. The percentages have been calculated on the basis of whole subcorpora, rather than as the
means for individual articles. The two articles engling18 and frling13 are not included in the
tables. They contain 233 and 19 RESULT occurrences respectively, which are mostly used as
metalinguistic terms.
 Chapter 5. Reader/writer interaction 185

Table 5.4-2. Frequency of RESULT in French subcorpus

Subcorpus N %

frecon 612 0.21

frling 92 0.04
frmed 89 0.06

Total 793 0.12

Search string: résult.*

Table 5.4-3. Frequency of RESULT in Norwegian subcorpus

Subcorpus N %

noecon 657 0.21

noling 224 0.08
nomed 161 0.16

Total 1,042 0.15

Search string: result.*

For English, a search for the commonly used noun finding (including plural forms)
was also undertaken to ensure a more correct picture of the situation.22 The total
number of hits was 362. Medicine was the discipline which used finding most fre-
quently (viz. 0.08 %; in economics: 0.04 %; linguistics: 0.03 %).
Since we are especially interested in what may be termed final (or major and
general) results, i.e. the main point(s) or result(s) of the research presented in
the article, a search for the word conclusion (and derivatives and corresponding
forms)23 was also undertaken. The search revealed that conclusion is much less
used; while the total number of occurrences of RESULT is 3,599, the correspond-
ing number for conclusion is only 888. What is more interesting is that the average
use is the same within the three languages taken as a whole (around 0.04 %); the
differences between disciplines are great, however. Economics has the highest fre-
quency (0.05 %) here too, but linguistics is close behind (0.04 %), while medicine,
a discipline which frequently uses finding, has the lowest frequency (0.02 %).
There are also interesting observations as regards the distribution of RESULT
in the text. Concerning the preferences for placing RESULT either at the begin-

22. The verb find has not been taken into consideration here.
23. The search was made by the search string concl.* for English and French and by konkl.* for
Norwegian.
186 Academic Voices

ning or at the end of the article, there are no clear language distinctions.24 As for
discipline differences, one can say that economists use RESULT more in the In-
troduction section than in the Conclusion section; the medical researchers do it
the other way round: RESULT is generally used more in the Discussion and the
Conclusion part than in the Introduction. The linguists do not use RESULT very
much in either Introductions or Conclusions; when they use it, it is in the middle
of the text, reporting the research process.
In addition to the findings reported above, a last point, perhaps the most impor-
tant one, has to be noted: The individual differences are clearly visible. Cases in point
are articles engecon11 with 45 occurrences of RESULT and engecon14 with none.

Observed categories of RESULT

In the second step of our investigation, we focus on the ways of using RESULT, and
particularly for expressing “final results” in the English articles. Some comments
on evaluative expressions will also be given. The observations will form the basis
for the claim that it is necessary to undertake a qualitative analysis in order to get
a good picture of how researchers express their results.
The 3,599 occurrences of RESULT represent a very heterogeneous sample,25
and all the examples are not equally interesting for our purposes. The main catego-
ries are as follows: RESULT used in a general meaning, RESULT used as a techni-
cal linguistic term, RESULT related to research. The last category is further divided
into three subtypes: metatextual reference, process results and final results.
1) RESULT used in a general meaning
There are many examples of RESULT used with a more or less general meaning, i.e.
they have nothing to do with results coming from a specific research project. They
are simply used to denote something that happens or exists because of something
else that happened or existed (often situated outside the research world presented
in the article). Here are some examples:
(1) Matches that would have resulted in illegitimate births still form; now they
are marriages, but marriages that immediately dissolve. (engecon18)
(2) Patients incur extra costs, however, as a result of surgery, and hospitals can be
expected to benefit from these, as well as from the higher bed occupancy rates
that result from programming. (engmed04)

24. See Mauranen’s study on cultural differences in academic rhetoric (Mauranen 1993a). She
indicates that Anglo-Americans prefer to start with their main point(s), while the Finns seem to
prefer the end for their main point(s).
25. For practical reasons we give only examples of noun and verb variants here – result(s) and
to result; only English examples are given.
 Chapter 5. Reader/writer interaction 187

2) RESULT used as a technical linguistic term

This is a use which typically occurs in linguistics, where RESULT can denote a
semantic and pragmatic or a text-structural category in different contexts:
(3) However, Grimshaw (1990: section 3.5.1) argues that nouns such as an-
nouncement, statement, observation, etc. with propositional CP comple-
ments are never argument-taking, but show all the characteristics of ‘result’
or ‘simple event’ nominals with the single referent argument (c)R(tm), […].
(engling10)
(4) The topic-based format is, however, of a very different kind with only one of
the theses in the collection having what might be considered a materials and
methods type section and none of them having separate results and discus-
sion sections. (engling11)

3) RESULT related to research

For present purposes, we have established the following three subcategories:
3A) Metatextual reference
In this category, we find occurrences of RESULT in a metatextual context. The word
is used only to indicate in which part of the article the results are to be found:
(5) This paper’s primary results are in Section 4, which takes the model to the
data and confirms its main predictions. (engecon40)
(6) The results of the experiment are presented in Table 1. (engling19)
(7) Figure 2 shows the result of dot-blot analysis, identifying the presence of
HPA-binding glycoproteins after HPA affinity chromatography of cell line
preparations. (engmed08)

3B) Process results

These are part results which it is important for the researchers to report before ar-
riving at their final results and conclusion. The context is typically the description
of some step in the analysis, in an experiment, etc.:
(8) Throughout, estimation results were obtained by maintaining the adding up,
homogeneity, and symmetry restrictions, and by deleting the chicken equa-
tion. (engecon33)
(9) The intonational devices that give rise to the structures in (1) cannot be used
to induce the intonational structures indicated in (2) without absurd results
[…]. (engling16)
(10) We obtained results on 398 specimens distributed 105, 130, 85, and 78 by the
4 time periods, of which 74.6% were HIV-uninfected. (engmed17)
188 Academic Voices

3C) Final results

By final results we mean the major outcome of the total research process reported
in the article: what the author of the article ends up with when summing up, con-
cluding and evaluating his or her research, and what can be important for the
promotion of the work carried out.26 The following is an example taken from the
Conclusion of an economics article:
(11) Our results imply that discriminatory trade barriers may be the preferred
instrument of enforcement in those agreements that are susceptible to the
problems of renegotiation. (engecon38)

These are the RESULT occurrences which we will look at in the following.

Final results
Since final results, for obvious reasons, are typically expressed in the Introduc-
tion and/or in the Discussion and Conclusion sections,27 we restrict our analysis
to these sections.28 Two main categories are proposed: accounting for results and
qualifying results. This dichotomy represents an attempt to generalise the observed
occurrences; for a more detailed classification, see Ruiying & Allison (2003).

Final results 1: accounting for results

RESULT is used to report in one way or another the final outcome of the article;29
a model example could be: My/our results show x. The verb show is used here in
order to generalise over the meanings communicated by various verbs. The actual
verbs of this type may be more or less specific, but the aim of the author is in all
cases to account for the outcome of his or her research. Here are some examples:
(12) We have developed a two-country two-sector model of the business cycle.
Our results show that such a model is capable of generating a positive inter-
national transmission of the business cycle, […]. (engecon31)

26. The term final result corresponds more or less to the term bottom-line result used, for exam-
ple, by Berkenkotter and Huckin (1995: 35).
27. See Berkenkotter and Huckin (1995: 30–39) who point to the tendency to place the most
important points in the abstract, the title and the Introduction (in a study based on articles in
biology and physics).
28. One reason why we do not include the Result section is that one finds most typically “pro-
cess results” (see above) in this section. Moreover, it is mostly the medical articles which contain
a Result section. The other disciplines typically use other headings, but most of them have a
section called conclusion, concluding remarks, summing up, and the like.
29. RESULT can, of course, be used to refer to own findings as well as findings reported by
others; we only look at own results here.
 Chapter 5. Reader/writer interaction 189

(13) Our results indicate that hematospermia occurs in 15 % of all patients and may
persist in approximately 6 %. (engmed06)
(14) The three main theoretical results of the paper are that increasing the average
return to ”mainstream” behavior in the community non-linearly decreases the
percentage of ”deviants” in the community; decreasing the variance of returns
to ”mainstream” behavior in the community also non-linearly reduces devi-
ance. Finally, if the upper bound on the distribution of these returns for the
community is below some threshold value then there is stable equilibrium
where all individuals will want to be ”deviant”. (engecon19)
(15) Our results suggest that the down-regulatory effect of CRH and ACTH is
exerted not at a transcriptional level, as evidenced by a lack of change in 11ß-
HSD-1 mRNA, but rather by […]. (engmed47)
(16) Our results may suggest avenues for future research. First, our results estab-
lish that particular directions of quota reform are strict Pareto improving
when […]. (engecon07)
(17) These results inform us significantly about various natural typological ar-
rangements cross-linguistically and about certain kinds of recurrent changes.
(engling31)

Final results 2: qualifying results

In this category, the outcome reported by RESULT is qualified or evaluated in one
way or another; a model example could be: My/our results are useful. While the
first category of RESULT aims at accounting for results in an apparently objective
way, the second type constitutes a sort of self-evaluation.30 In today’s competitive
research society, it is reasonable to expect such qualifications to be added to the
results. Generally, however, their presence seems to depend on individual factors
and style. Since we only consider the qualifications of the authors’ own results
here, they are generally positive. (There are, however, more negative or “neutral”
qualifications too, but they will not be considered here.) The results can be quali-
fied as original, solid or useful, as illustrated in (18)–(19), (20)–(22) and (23)–(26),
respectively.
(18) The preliminary results discussed above are an initial step in investigating the
determinants of countries’ external wealth. (engecon02)
(19) To my knowledge, the results are the first to link FPRs with simultaneous
servicing decisions via the ownership, location, and internalization concepts
of the multinational literature. (engecon04)

30. For finer categories, see Hunston (1993) and Stotesbury (2003).
190 Academic Voices

(20) The results are robust with respect to the grouping of countries by imitative
ability and the delineation of these groups. (engecon04)
(21) Finally, we note that our results are consistent with, and tend to confirm,
those reported in the recent work of Engel and Hakkio (1996). (engecon08)
(22) The experimental results reported here provide positive support for the nom-
inalization analysis of children’s errors in interpreting adjunct PRO construc-
tions. (engling19)
(23) Therefore, these results should be useful for future research on measuring
international risk-sharing gains. (engecon06)
(24) Our results may suggest avenues for future research. (engecon07)
(25) These results have some interesting implications for fiscal policy in the Euro-
pean Union, particularly as the EU continues to evolve. (engecon11)
(26) The results of our study provide a means for distinguishing between distinct
senses and the process of on-line meaning construction, which is primarily
conceptual in nature. (engling17)

Having given examples of two main categories of final result presentation, let us
now focus on the frequency of these in all three languages, within the maze of RE-
SULT use in the research articles. For this part, we limit the analysis to 180 articles
of the corpus (the 01–20 articles of each subcorpus). There are 1220 occurrences of
RESULT in total in this sample; 306 of them are found in the Introduction and Dis-
cussion/Conclusion sections (175 English, 68 French and 63 Norwegian). Among
these there are only 102 which qualify as “final result” examples (21 in the Introduc-
tion sections and 81 in the Discussion and Conclusion); there is a fairly even dis-
tribution between the accounting type and the qualifying type. In short, RESULT
is not much used in presenting final results. The relation between languages and
disciplines is the same as for RESULT in general: English is ranked first, then Nor-
wegian and French; economics is ranked first, then medicine and linguistics.
To sum up the quantitative investigation of the use of RESULT in general
terms:31
• English researchers use RESULT more frequently than Norwegian and
French researchers.
• Economists use it more often than medical researchers and linguists.

31. We have not studied systematically the distribution of RESULT between presentations of
own results and of others’ results. However, the tendency seems clear: RESULT is more often
used to present own results than others’ results.
 Chapter 5. Reader/writer interaction 191

• The most frequent use of RESULT turned out to be in articles present-

ing some statistical/numerical or experimental research (economics and
medicine).
• When RESULT is used to present final results, it seems that the type “ac-
counting for results” is as important as “qualifying results”.

5.4.3 Discussion

The semantic content ‘result’ can, of course, be expressed in many ways. Thus, the
quantitative results given above can only be indicative. They should be extended
through the analysis of other lexical expressions and more qualitative analyses of
entire articles. This is particularly important for the expression of final results.
However, some interesting tendencies have been observed. As we have shown,
our corpus indicates that as regards RESULT, English researchers, especially Eng-
lish economists, are the most explicit, while French researchers, and particularly
French linguists, are the least explicit. Norwegian is situated between the other
two languages.
The explanation for the differences between disciplines seems to be differences
in type of research and research design. The economists and to a larger extent the
medical researchers obtain results which can be referred to by result (or finding),
while linguists (at least in our corpus) often obtain results which are more appro-
priately referred to by nouns like solution, interpretation, etc.
As regards language differences, our observations point to cultural differences
between English and French as explicit and non-explicit, respectively (also ob-
served by for example Vassileva 2000). However, as regards Norwegian, the indica-
tions differ somewhat from the trends seen in some of our other investigations.
Generally, in the KIAP studies, the English articles are more explicit than the
others; the Norwegian articles are quite close to English. We tend to explain this by
English influence on Norwegian. Geographically, historically and politically, Nor-
way and Great Britain are close. Norwegian researchers see the contact with the
Anglo-American world as important; this is not necessarily so in all French sci-
entific communities. In addition, Norwegians use English a lot in various profes-
sional settings and may therefore be influenced by the English language. However,
in the present study, the difference between English and Norwegian is notable, and
we have to look for other explanations. It does not seem that the tradition of Eng-
lish-speaking researchers to be direct and explicit in their article writing has influ-
enced Norwegian researchers in all matters. The study undertaken here suggests
that Norwegian researchers are more modest and careful than English-speaking
researchers are, possibly because of the smaller discourse community.
192 Academic Voices

As for the discipline dimension, there is also an important difference to note

when compared to other KIAP investigations, i.e. the ranking of the three disciplines
studied. Our initial working hypothesis suggested that economics is situated between
medicine and linguistics as regards explicitness and author manifestation (Breivega
et al. 2002). This had to be revised because several of our studies tend to situate eco-
nomics close to linguistics, i.e. closer to so-called soft disciplines than to hard ones.
Thus, rather than pointing to three distinct author profiles, we could instead study
two (in very general terms). The reason is that the profiles of the economist and the
linguist appear to be quite similar, while the profile of the medical author seems to
take a clearly different shape. In the present analysis, however, economics seems to
be closer to medicine. As already indicated, this probably has to do with the research
processes and methods and the object of investigation, factors which should be inte-
grated in all studies of scientific discourse (see section 7.1).
However, all in all, the most important observation is perhaps the one indicat-
ing clear and considerable individual variations, which to a certain extent override
both language and discipline variables.
The quantitative outcome of the present study might lead to the conclusion that
Norwegian and French researchers are less explicit and direct, and perhaps more
careful, than their English-speaking colleagues when presenting results. But the low
number of occurrences of RESULT used to present final results (especially in linguis-
tics) as well as our knowledge about scientific writing in general point to the necessity
of applying other methods in order to get a more complete picture. To complement
our quantitative investigation, we have analysed three linguistics articles with respect
to lexical choices pertaining to results. These analyses are presented in 5.4.4.

5.4.4 Analysis of three linguistics articles

In the analysis of the three selected texts, we will change the perspective to some
extent and look for other lexical means of expressing RESULT or notions that
are close to it. Since we are mainly interested in final results, we will focus on the
Introduction and Conclusion sections of engling01, frling01, and noling01.32 The
three articles all investigate a particular aspect of the three languages: the gerun-
dive relative clause in English, so-called eclipse prepositions in French and the
demonstrative pronoun denne (‘this’) in Norwegian, but take different theoretical
perspectives. They are single-author articles written by female researchers.
Prelli’s theory of the rhetoric of science (Prelli 1989) provides us with the fol-
lowing points as a basis for formulating a hypothesis regarding the expressing of

32. A possible approach for the analysis of whole texts could be to study semantic isotopies
(see Rastier 1987), i.e. chains of recurrent semes or semantic traits realising RESULT chains.
 Chapter 5. Reader/writer interaction 193

result in these articles. Scientific discourse is rhetorical in the following sense: “it
involves selective use of symbols to induce cooperative actions and attitudes re-
garding particular orientations for attaching meanings to situations” (Prelli 1989:
87). This means that a scientific rhetor has to select one or several suitable ends,
to locate a relevant problem or issue and to choose an appropriate argumentative
strategy. There is no reason that our three articles should not be considered rhe-
torical in this sense. In fact they have all chosen some kind of argumentative prob-
lem–solution strategy, and of course they try their best to find a solution to the
problem presented, i.e. they search for a valid result of their research. We will now
see to what extent they fulfil the rhetorical ideal by selecting appropriate linguistic
expressions, to induce cooperation and to persuade the readers that their research
should be given an appropriate place in the discipline’s literature.

Engling01
There are five occurrences of RESULT in this article, but they are found in the
middle part of the text and are of the types “process result” and “general result”, as
in the following example:
(27) Given that the Event and Speech time readings are the result of the linking of
different times in tense structure, the syntactic issue is how these readings are
represented in the sentence structure.

Thus, these occurrences are not interesting to us from the perspective taken here.
So, how are final results or findings expressed?
If we go to the abstract first, it clearly formulates a main claim as well as a
claim related to the empirical phenomenon studied (expressions which seem rel-
evant for the present discussion appear in bold):
(28) [abstract] This article explores the interface between the syntactic and seman-
tic representation of natural language with respect to the interpretation of
time. The main claim of the paper is that the semantic relationship of tem-
poral dependency requires syntactic locality at LF. Based on this claim, I ex-
plore the syntax and semantics of gerundive relative clauses. I argue that since
gerundive relatives are temporally dependent on the tense of the main clause,
they need to be local with a temporal element of the main clause at LF. I show
that gerundive relatives receive different temporal interpretations depending
on their syntactic position at LF. This analysis sheds light on the behavior of
gerundive relatives in constructions involving coordination, existential there,
scope of quantifiational and cardinality adverbials, extraposition, presupposi-
tionality effects and binding-theoretic reconstruction effects.
194 Academic Voices

These claims are repeated in the Introduction section, introduced by formulations

including the verb argue. Other manifestations of the results of this study are indi-
cated by expressions like I claim and I show:33
(29) 1. Introduction [455 words]
Recent work on the syntax of tense shows that there is a principled relation-
ship between the meaning and the phrase structure representation of tem-
poral information […]. [first claim:] I contribute to this discussion here by
arguing that syntactic locality constrains the interpretation of temporal rela-
tions; […].
I adopt the semantic analysis of Reichenbach (1947) […], and I claim
that […]. Based on this syntax and the claim […], I present an analysis of the
syntax and semantics of gerundive relative clauses. [second claim:] I argue
that gerundive relatives are temporally dependent on the main clause tense
and thus are required to be local with a matrix time at LF.
This analysis of gerundive relatives has implications for […]. […].
This paper is organized as follows: in section 2, I outline […]. In section
3, I show that […]. […], I claim that […], in section 4.
Section 5 argues that […]. Evidence for the proposal is discussed […] in
section 6.
Section 7 discusses […] and section 8 presents a discussion […]. I argue
that […].
The two main claims are repeated in the conclusion:
(30) 9. Conclusion [247 words]
[first claim:] In this paper, I have shown that the LF syntax of subjects provides
evidence for the structure of tense proposed here. [second claim:] Gerundive
relatives in subject position are temporally dependent on the matrix Event or
Speech time, and the Event time is represented in VP. […]

The content of the first claim is said to have been shown, while the second claim
is presented without any introductory phrase. This “objective” presentation, not
related to any person, is the most categorical way of expressing a result. It presents
something as a self-evident fact with no reservation or limitation (hedging) what-
soever. The author implicitly expresses that this “fact” has been proven and should
be taken as true in all contexts; there is no need to use an expression like for in-
stance I argue.

33. Another interesting point, also mentioned in our discussion of author roles in abstracts
(4.5.2), is the personification of the inanimate subject that we find in utterances like Section 5
argues that … (see also Master 2001).
 Chapter 5. Reader/writer interaction 195

In the seven middle sections, there is continuous argumentation related to

specific linguistic examples, as well as to the theoretical discussion. There are ex-
pressions of the type This explanation provides evidence and the present analysis
correctly predicts that, not referring directly to the author, as well as personalised
expressions like I assume, I propose, I present evidence and I claim. Sometimes the
reader is included as a witness of the results: In this section, we have seen …
To sum up, the major or final results are in fact reported three times, in the
Abstract, in the Introduction and in the Conclusion. The Introduction seems to be
the section in which they are presented in the most complete way. Formulations
like I contribute … by arguing, I argue and I claim make the presentation explicit
and direct. The presence of an author who wants to position herself is clear. Except
for the example from the Conclusion mentioned above, the results are not pre-
sented as self-evident but as part of an explicit argumentative strategy for which
the author is responsible. This is supported by frequent use of the negation not,
often polemic (see 6.4 and Fløttum 2004f, 2005c), accompanied by the contrastive
and concessive connective but, as here in the last part of the Conclusion:
(31) Full relatives avoid reconstruction effects by adjoining directly to the wh-ele-
ment in Spec, CP, but this adjunction is not permitted with gerundive rela-
tives, since in Spec, CP, they are not in a local relation with either the Speech
or Event time, and hence cannot receive a temporal interpretation.

In conclusion, this is an author with a manifest wish to “win”, on her own, in the
struggle for a position within her preferred community, a wish that is manifested
by an explicit, direct and argumentative presentation of final results.

Frling01
Frling01 contains no occurrences of RESULT. As stated above, the article presents
a study of so-called eclipse prepositions in French. In the English abstract, the
subject topic is introduced as follows:
(32) [abstract] In French as in English, ”weak prepositions” à, de, en, cannot pre-
cede a QUE, whether it be a conjunction or a relative pronoun. [...] Two gram-
matical solutions can be used, a ”poor one”, zeroing the preposition, […], and
a ”rich one”, adding a demonstrative CE between the preposition and QUE,
[…]. […]
Both solutions, in both cases, cannot be predicted by regular rules. The
choice depends on variable usages and no diachronic ordering can be stated.
196 Academic Voices

From both Abstract and Introduction (see below), it is clear what the author is
going to study, but there is no explicit presentation of what will be her main claim
or final outcome:
(33) [Introduction. 615 words]
Les prépositions spécifiques des valences verbales sont généralement consid-
érées comme indispensables [...]. […]
Cette particularité est mentionnée depuis longtemps par les grammair-
iens français […]
Il me semble que cette interprétation est fondée sur des bases peu solides
et qu’on doit accepter l’idée que les trois tournures coexistent actuellement, et
qu’elles ont sans doute coexisté à d’autres époques. Dans cette perspective, il
faudrait admettre que ce phénomène de ”préposition à éclipses” n’est pas un
accident de l’évolution mais qu’il fait partie de la grammaire. Comme d’autres
langues semblent présenter des faits analogues (Dixon 1991: 14, 48, 66, 68,
281–286), on peut être tenté d’y voir un phénomène plus général, qui ne se-
rait pas spécifiquement français.
Je propose de rappeler les principales circonstances dans lesquelles
s’observe ce phénomène de la “préposition à éclipses”, et d’en envisager
quelques interprétations1.
‘The specific prepositions of verbal valence are generally considered as indis-
pensable ...
This particularity has been mentioned for a long time by French gram-
marians …
It seems to me that this interpretation is founded on not very solid bases
and that one must accept the idea that three expression types coexist at present,
and that they have no doubt coexisted in other periods. In this perspective, one
must admit that this phenomenon of ”eclipse preposition” is not an accident of
evolution, but that it is part of grammar. Since other languages seem to exhibit
analogous data (Dixon 1991: 14, 48, 66, 68, 281–286), one may be tempted to
see in this a more general phenomenon that is not specific to French.
I propose (intend) to report the main circumstances in which this phe-
nomenon of the “eclipse preposition” may be observed and to suggest some
interpretations of it.’
Thus, after having offered a thorough presentation of the issue, she ends the In-
troduction with an explicitly formulated intention. In the preceding presentation,
however, the author is very modestly manifesting herself, by modalised or hedged
expressions like ‘it seems to me’, ‘one must accept’, ‘one must admit’, ‘one may be
tempted to see’. This indirect style, characterised by the use of the indefinite pro-
noun on (see 4.4) or impersonal constructions accompanied by different hedging
 Chapter 5. Reader/writer interaction 197

devices, is also quite typical of the middle sections. There are, for example, only
two occurrences of je, both in the first sentence after a subheading, combined with
the verb s’intéresser à (‘be interested in’ in the sense ‘study’, ‘look at’) with a more or
less metatextual function pointing to what is going to be studied in the section in
question. There is also the expression On se donnera donc le droit de poser … (‘One
will give oneself the right to postulate …’), where the verb clearly indicates that a
claim is presented, but where it is difficult to relate the claim directly to the author
because she uses the pronoun on. Let us now look at the Conclusion:
(34) 4. Conclusion [559 words]
La conclusion est provisoire. Elle porte à la fois sur les données et sur l’analyse.
Pour les données, d’abord. La ”préposition à éclipse” n’a généralement pas
reçu une bonne légitimation grammaticale, que ce soit chez les grammairiens
classiques ou chez les linguistes contemporains. Ce n’est pas par hasard. Elle
oblige à considérer deux situations désagréables. […] Pour l’analyse, la clef des
embarras vient du statut grammatical à accorder à qu-. [...].
Berrendonner (1997), convaincu que ce sont des constructions distinctes,
envisageait même que [...].
Je ferais plutôt l’hypothèse qu’il s’agit, dans les trois cas cités, d’une
seule et même structure syntaxique clivée, quelle que soit la réalisation mor-
phologique qui en est donnée, soit par des pronoms qui supportent les prépo-
sitions, soit par une conjonction qui les refuse. [...]
[…]. Ces linguistes [just referred to] fondaient leurs analyses essentielle-
ment sur [...]. Je proposerai d’étendre l’analyse aux emplois prépositionnels
manifestés dans les tournures syntaxiques focalisantes que je viens de citer.
[...]. Cela reviendrait à dissocier, dans les deux cas, le niveau des structures
syntaxiques et le niveau des réalisations morphologiques.
‘4. Conclusion
The conclusion is provisional. It concerns at the same time the data and the
analysis. First, as regards the data: The ”eclipse preposition” has generally not
received a good grammatical justification, neither by classic grammarians nor
by contemporary linguists. This is not accidental. It necessitates the considera-
tion of two disagreeable situations. … As regards the analysis, the key to the
embarrassment stems from the grammatical status of qu-. ...
Berrendonner (1997), convinced that they are distinctive constructions,
even imagined that ... I would rather propose the hypothesis that, in the three
mentioned cases, it is a question of one and the same syntactic cleft structure,
whatever the morphological realisation of it is, either by pronouns supporting
the prepositions, or by a conjunction that prohibits them. ...
198 Academic Voices

… These linguists founded their analyses mainly on ... I would propose to ex-
tend the analysis to prepositional uses manifested in the focalising syntactic con-
structions which I just mentioned. ... This implies a dissociation, in the two cases,
of the level of syntactic structures and the level of morphological realisations.’
After having presented views that differ from her own (and with the manifesta-
tion of a polemic negation by the use of ne … pas (‘not’)), the author rejects their
solutions, and then states that she will go further: ‘I would rather propose the
hypothesis …’, ‘I would propose to extend the analysis ...’. Thus, at last, the reader
is informed of what is really her final result or conclusion. The argumentation
throughout the entire text is thorough and quite explicit, but the author chooses to
wait until the very end before presenting her final result and main claim. However,
the polemic of the article is clear from the beginning. After having referred to
several grammarians and linguists, she says in the Introduction, cf. example (33)
above: ‘It seems to me that this interpretation is founded on not very solid bases ...’
This rejection of others’ interpretations is also manifested in the Conclusion.
Frling01 differs from engling01 in two ways. First, in engling01 the final results
are presented both in the Introduction and in the Conclusion, but in a more direct
and explicit form in the Introduction. Frling01 only indicates the final results in
the Conclusion. Second, engling01 uses forms of expression which are more direct
and more clearly position-taking. The author of frling01 does not position herself
as clearly as the author of engling01 until the end of the article. The results are
presented more as suggestions and propositions than as claims. In conclusion, this
is an author who also wishes to “win” and who maintains a line of argumentation
throughout the article. But this wish to “win” is manifested in a less explicit and
less direct way than in engling01, through the presentation of final results.

Noling01
Like frling01, noling01 contains no occurrence of RESULT. Noling01 presents a study
of the Norwegian pronoun denne (‘this’). The Norwegian abstract is as follows:
(35) [abstract] Artikkelen belyser de selvstendige, ikke-attributive pronomenene
denne, dette og disse. Først og fremst undersøkes det deskriptivt hvilke bruks-
betingelser pronomenene har i forhold til mulig antesedent. Det vises at de er
gjenstand for en generell subjektsvegring. Både semantiske og pragmatiske
betingelser undersøkes og forkastes. Deretter foreslås en syntaktisk analyse
basert på bindingsteorien i Chomsky (1986) og på Hestviks (1992) analyse
av anti-subjektsorientering. Den gjør rede for de setningsinterne begrensnin-
gene på koreferanse.
‘(abstract) The article studies the independent, non-attributive pronouns
denne, dette og disse (‘this’, ‘these’). First and foremost usage restrictions related
 Chapter 5. Reader/writer interaction 199

to possible antecedents are investigated. It is shown that they are subject to a

general anti-subject orientation. Both semantic and pragmatic conditions are
investigated and rejected. Then a syntactic analysis based on the binding theo-
ry in Chomsky (1986) and on Hestvik’s (1992) analysis of anti-subject orienta-
tion is proposed. It explains the clause-internal restrictions on coreference.’
The main claim is that the use of the pronoun denne is syntactically rather than
semantically or pragmatically constrained: ‘It is shown that they are subject …’ The
proposal of a syntactic analysis of the phenomenon is also announced.
The Introduction section, provided in (36), is a long and complex section con-
sisting of three parts (limitation of the study, proposition of analysis and previous
research); however, we find some indications of what the author’s claim is. After
having made explicit what the subject matter of the article is, the author uses the
introductory formulation ‘We shall see that …’ to announce her claim of anti-sub-
ject orientation. She also states clearly what kind of analysis she will propose for
the explanation: ‘I will propose that…’.
(36) 1. Innledning1 [1102 words]
1.1. Avgrensning
I denne artikkelen skal jeg belyse flere sider ved pronomenet denne, med de
alternative formene dette og disse. […]
Det artikkelen skal handle om, er den anaforiske, selvstendige bruken av
pronomenet. […]
1.2. Forslag til analyse
[…]. I denne artikkelen skal vi se på hvilke bruksbetingelser som finnes. Vi
skal se at […].[…]
Pronomenet er altså offer for subjektsvegring. […] Vi vil derfor ende opp
med et større syntaktisk apparat, […]. Jeg skal foreslå at bruksbetingelsene til
DENNE kan forklares ved […]. […]
1.3. Tidligere omtale i litteraturen
[…]
‘1. Introduction1
1.1. Delimitation
In this article, I will study several aspects of the pronoun denne (‘this’) and its
alternative forms forms dette and disse. …
What the article will deal with is the anaphoric, independent use of the
pronoun. …
1.2. Analysis proposal
[…]. In this article we will look at what the conditions of use are. We will see that …
200 Academic Voices

The pronoun is thus subject to anti-subject orientation. … We will there-

fore end up with a larger syntactic apparatus, … I will propose that the condi-
tions of use for DENNE can be explained by […].
1.3. Treatment in previous literature
…’
There is an interesting alternation of perspective in this introduction between jeg
(‘I’) and vi (‘we’). In the rest of the article, the first person plural dominates (in all:
62 occurrences of vi and 3 of jeg). The use of vi, especially combined with a verb like
se (‘see’), indicates a wish to include the reader. This invitation to participate in the
interaction is continued throughout the article. Such author manifestations, with
various uses of inclusive and exclusive we (see 4.3 and Kinn 2005a), constitute an
important part of the argumentation; they switch from cooperating initiatives like
Vi ser at (‘We see that’) to relatively firm and conclusive claims like Vi kan derfor
konkludere at (‘We can therefore conclude that’) and Vi kan altså fastholde at (‘Thus
we can maintain that’). These refer to what we have called process results above.
As regards final results, we have shown that they are stated in the introduction.
Let us now look at the conclusion.
(37) 5. Sammenfatning [111 words]
Vi har sett på ulike pragmatiske, semantiske og syntaktiske bruksbetingelser
for pronomenet DENNE. Vi har sett at betingelsene er såpass komplekse at
en syntaktisk forklaring har vært nødvendig, med bindingsteori og subjekts-
vegring som følge av LF-flytting: DENNE må være fritt for en subjektsante-
sedent, og dette gjelder for alle subjekter i helsetningen, uansett hvor mange
trinn ned pronomenet befinner seg.
Analysen tar for seg alle forekomster (så nær som én) av setningsinternt
DENNE, men sier ikke noe om antesedens over helsetningsgrense. Det ser
uansett ut til at subjektsvegringen er langt svakere setningseksternt, noe som
tyder på at det der finnes andre betingelser – kanskje mer i tråd med logofor-
iske pronomener.
‘5. Summary
We have looked at different pragmatic, semantic and syntactic conditions of
use for the pronoun DENNE. We have seen that the conditions are so com-
plex that a syntactic explanation has been necessary, with binding theory and
anti-subject orientation as a result of LF-movement: DENNE must be free of
a subject antecedent, and this holds for all subjects in the matrix clause, what-
ever the number of steps down the pronoun is.
The analysis considers all occurrences (except one) of sentence-inter-
nal DENNE, but does not say anything about antecedence across sentence
boundaries. In any case it appears that the anti-subject orientation is much
 Chapter 5. Reader/writer interaction 201

weaker sentence-externally, which indicates that there are other conditions

– perhaps more in line with logophoric pronouns.’
Compared to the Conclusion in frling01, this one is very brief. However, the final
result or main claim is stated, but in a complex and long formulation, in the sen-
tence introduced by ‘we have seen that …’.
To sum up, the final results are reported both in the Abstract, Introduction
and Conclusion section. The we-author of noling01 is not as direct and explicit as
the I-author in engling01, but more direct than the author of frling01 (who could
almost be called an indefinite on-author because of the use of the pronoun on).
As in the English article, the final results or claims of noling01 are reported in the
most complete way in the introduction (different from frling01, which states and
develops this in the Conclusion). The presence of an author who wants to position
herself is not as clear as in engling01, because of the use of vi (‘we’), but clearer
than in frling01. However, the argumentative strategy is clear throughout the arti-
cle. As in the other two articles, this strategy is supported by a considerable use of
the negation particle not.
In conclusion, this is an author who first and foremost seems to want the par-
ticipation of the reader and thereby of the scientific community she is, or wants to
be, a part of. In addition, there is the wish to “win” in the position struggle. This
wish is manifested by a frequent use of vi in expressions like vi kan fastslå (‘we
can conclude’). The author presents her results in an explicit, but not always very
direct way.
The qualitative analysis has shown that these authors have chosen quite differ-
ent styles to convey their message: From the English and most direct, I argue, via
the Norwegian not so daring and direct but quite manifest, Vi fastslår (‘We con-
clude’), to the least direct and most modalised French, Je ferais plutôt l’hypothèse
que (‘I would rather make the hypothesis that’).

Summary
With a point of departure in the theory of the rhetoric of science, we can character-
ise all three articles as exhibiting problem–solution strategies. From this it follows
that one could expect explicit presentation of solutions of the problems functioning
as the object of study. And the solutions were there. However, even if the authors
have chosen similar argumentative strategies, they have opted for quite different
styles to realise it. Three individual author profiles may hence be sketched:
English. The author of engling01 expresses her results and solutions in a clear, ex-
plicit and direct way. Through the use of expressions like I argue, she seems to cre-
ate a research space of her own, which marks a kind of independence with respect
to the larger scientific community.
202 Academic Voices

Norwegian. The author of noling01 expresses her results and solutions in a clear
and relatively explicit way, but in a less direct manner than her English colleague.
Through the extensive use of the pronoun vi in expressions like vi kan derfor kon-
kludere at (‘we can therefore conclude that’), she seems to create a research space
common to her and a larger scientific non-specified community. The linguistic
expressions used in this article point to a careful author as well as an author who
wants to be an integrated part of the research community that is established.
French. The author of frling01 expresses her results and solutions in a less explicit and
direct way than the other two. By the use of different kinds of hedges, impersonal con-
structions and the indefinite pronoun on (‘one’), she, in contrast to the English author,
does not seem to create her own personal research space. On the contrary, she seems to
create a distance between herself and the research community. By using expressions like
On se donnera donc le droit de poser… (‘One will give oneself the right to postulate…’),
she gives the impression of being careful and distant. However, this “impersonal” man-
ner of expression may camouflage subtle argumentation which even rejects other re-
searchers and their points, like in the following utterance taken from the Conclusion: Je
ferais plutôt l’hypothèse qu[e] ... (‘I would rather propose the hypothesis that …’).
Even if, on the basis of these three articles, we cannot point to profiles which
clearly indicate different cultural identities, the variation that we have observed is
considerable. However, there is one characteristic trait common to all three: They
manifest a wish to “win”. But there is no straightforward answer to the question of
who is most likely to be a winner. That depends to a large extent on the reader and
the discourse community he or she belongs to.

5.5 Case studies

5.5.1 Metatext in three economics articles

This case study will take a closer look at how metatext is used in three economics
articles from our English subcorpus. The first article, engecon03, has been selected
as a “typical” text, based on the fact that the relative frequency of metatext in this
article is the same as the mean for the whole engecon subcorpus, viz. 0.33 %. As
for the other two texts, engecon06 and engecon38, they are among the articles
with the highest relative frequency for this feature, 0.61 %, as well as the highest
number of occurrences (47 and 44 respectively, as against 15 in engecon03).34 As
will be seen from the discussion below, the two texts which make frequent use

34. Article engecon25, a very short text of nine pages only, has an even higher frequency, 0.67 %.
 Chapter 5. Reader/writer interaction 203

of metatext do this in different ways. Engecon03, our so-called typical text, has
two authors, as has engecon38, while engecon06 is written by a single, female,
author. When it comes to the structure of the articles, all three texts follow a very
typical pattern for economics texts with an Introduction, followed by several sec-
tions with subject-related headings (e.g. ”The channels of intervention”, ”The gain
function”), constituting a so-called Middle part, and ending with a Conclusion (cf.
section 1.2.4). Tables 5.5-1–3 below show the pattern of metatext in each of these
three texts (see section 5.2. for a general discussion of metatext and the items in-
cluded in our study).

Table 5.5-1. The pattern of metatext in engecon03

article/paper section above now below Sum

Introduction 2 4 – – – 6
Middle 2 3 1 – – 6
Conclusion 3 – – – – 3

Total 7 7 1 0 0 15

Table 5.5-2. The pattern of metatext in engecon06

article/paper section above now below Sum

Introduction 4 5 – – 1 10
Middle – 10 5 5 13 33
Conclusion 4 – – – – 4

Total 8 15 5 5 14 47

Table 5.5-3. The pattern of metatext in engecon38

article/paper section above now below Sum

Introduction 5 5 – – – 10
Middle 2 16 11 2 – 31
Conclusion 3 – – – – 3

Total 10 21 11 2 0 44

When comparing this with the other two texts with more frequent use of meta-
text (Tables 5.5-2 and 5.5-3), we see that the pattern of usage differs mainly in the
use of previews (below) and reviews (above). In addition, there is more frequent
reference to specific text parts through the use of section in the Middle part. In
204 Academic Voices

both engecon06 and engecon38 the five instances of section in the Introduction
are found in the road-mapping paragraph. As for the use of previews and reviews,
we see from Table 5.5-2 that engecon06 mainly makes use of the former, with 13
instances of below, while Table 5.5-3 shows that engecon38 has no previews at all,
but 11 instances of the review above. Engecon38, then, displays the most common
pattern of previews/reviews found in the entire KIAP Corpus (section 5.2). We
notice that most of the metatext is found in the Middle part of the two articles.
Let us now take a closer look at the context for the various instances of above
and below in engecon06 and engecon38. As already stated, engecon06 primarily
makes use of the preview below, but there are also some instances of above.
(3) Using utility functions similar to those used in the general equilibrium lit-
erature, I show below that this simple partial equilibrium framework implies
welfare gains of at least 20% of permanent consumption and often-times near
100%. (engecon06)
(4) This relationship will be determined endogenously below. (engecon06)
(5) Below, I examine the equity-based and consumption-based literature on risk-
sharing gains using the gain function Eq. (5). (engecon06)
(6) Below, I begin by calculating the gain function for two benchmark cases as-
suming plausible preference parameters. (engecon06)
(7) I follow this approach below although clearly this approach ignores the poten-
tial for estimation 12 risk to affect portfolio decisions. (engecon06)
(8) I will return to this problem below. (engecon06)

Examples (3)–(8) are all typical instances of preview use, preparing the reader for
what will follow. They all seem good illustrations of the text navigation assistance
we in section 5.2 claimed is the most prominent function of such meta-elements.
We notice how, apart from example (4), all the cotexts contain the personal pro-
noun I. The author in question is hence clearly visible in these cases. When it
comes to the use of the review above in this article, the cotexts tend to be less per-
sonal than those for below. Examples (9) and (10) are typical cases:
(9) As noted above, these measures represent the lower bounds for the true gains
since the feasible set of portfolios is restricted to linear combinations of the
US stock market and a fixed mutual fund of foreign stocks. (engecon06)
(10) As described above, low intertemporal substitution mitigates the gains from
risk-sharing. (engecon06)
 Chapter 5. Reader/writer interaction 205

If we then turn to text engecon38, with no previews, but with 11 instances of above,
it turns out that 10 of these occur as prenominal modifiers, as shown in examples
(11)–(13):
(11) The above two first-order conditions specify unilaterally optimal tariffs by
country Y on imports from each country […] and reflect familiar terms-of-
trade incentives in which countries impose a positive tariff on imports so as
to lower the world price of the good. (engecon38)
(12) In the following Lemma we establish that, as a result of the above incentives,
tariffs will be complementary. (engecon38)
(13) While the above qualification should be kept in mind and the limitations of
the partial equilibrium model are acknowledged, nevertheless similar results
are also present in general equilibrium models. (engecon38)

The only instance of an adverbial use of above is the following example:

(14) It should be apparent that the non-discriminatory punishment path specified
above satisfies the first three properties. (engecon38)

In all 11 instances, above points back to information found in the near vicinity. It is
thus quite local in scope. In addition, as it is part of a noun phrase with exact refer-
ence, what it refers back to tends to be very precise and limited stretches of text. This
is somewhat different from the use of below in text engecon06, which in most cases
points to information found at some distance and with less specific reference.
This case study has once again demonstrated individual variation in linguistic
behaviour within the corpus. In this case engecon06 and engecon38 are numeri-
cally almost identical in their use of metatext. However, the ways the authors of
the two texts exploit this pragmatic feature are somewhat less similar. When com-
paring metatext (and to some extent pronoun) use in the three articles, the author
of engecon06 gives the impression of being the most personal and visible author,
and can hence perhaps be described as the one closest to the personification of the
typical English-speaking author who acts as a guide when the reader ”tours” the
text (Mauranen 1993b: 16).

5.5.2 ‘Let’-imperatives in a Norwegian economics article

In this section, we take a closer look at how ‘let’-imperatives are used in one eco-
nomics article in Norwegian, viz. noecon12. This article is special in that it is based
on the author’s so-called trial lecture for the doctoral degree in economics and
hence may not be a typical research article. It is interesting, nevertheless, because
it contains 12 examples of ‘let’-imperatives: five sentences with la oss (‘let us’) and
206 Academic Voices

seven with la meg (‘let me’). The examples of ‘let us’ will be related to the usage
types posited in 5.3, and we will see that there is a rather clear functional distribu-
tion of ‘let us’ and ‘let me’ in this article (see also Kinn 2005c).
The article has a number of reader-oriented features, as well as expressions of
author manifestation. In addition to ‘let us’ and ‘let me’, it contains several occur-
rences of the ordinary imperative and some occurrences of ‘let’-imperative with
a third person object. There are also direct questions, singular and plural (mostly
inclusive) first person pronouns and a large amount of metatext. (This case study
takes into account all types of metatext, including those of our investigation in
5.2.) The article even has a couple of direct references to the reader, viz. by the
use of leseren (‘the reader’). All this gives the text a very reader-oriented and fairly
lecture-like character. It is quite possible that the oral history of the article has at-
tributed to this.
The five examples of ‘let us’ are provided in (15)–(19).
(15) For å illustrere enda en Nash-likevekt, la oss gå tilbake til eksempelet med de
to bilene som møter hverandre uventet på en smal vei.
‘…, let us go back to the example …’
(16) La oss konkludere diskusjonen av angrepskoordinerings-problemet.
‘Let us conclude the discussion …’
(17) La oss betrakte den følgende vrien på historien om Radar og Blind før jeg
forsøker å definere «nesten åpen kunnskap» mer presist.
‘Let us consider the following twist of the story …’
(18) La oss først se nøyaktig hvordan sjåførenes nytte er gjensidig avhengig av
hverandre i den følgende spillmatrise, før vi finner Nash-likevektene.
‘Let us first see exactly how …’
(19) La oss definere åpen oppfatning mer presist.
‘Let us define common belief more precisely.’

Examples (15) and (16) with ‘go back’ and ‘conclude’ illustrate the structuring type
posited in 5.3. While the former primarily orients the reader towards the thematic
structure of the text, the latter both orients the reader and involves the reaching of
an opinion about the matter in question. Example (17) with ‘consider’ belongs in
the scope type, defining the focus of attention, while example (18) with ‘see’ should
probably be classified in the analysis type (bordering on the scope type). The last
example, (19), with ‘define’, illustrates the definition type, referring to an act of
terminologisation. As argued in Kinn (2005c), the Norwegian la oss-imperative is
typically used in cases where it is easy to create the impression of the author and
reader performing the act together and having the same role.
 Chapter 5. Reader/writer interaction 207

On the other hand, according to Kinn (2005c), when the creation of a dyad
where author and reader have the same role is less easy, there is a tendency to use a
la meg-imperative instead. This is illustrated by the first six of examples (20)–(26),
which are the ones found in noecon12. The enunciative acts of exemplifying, men-
tioning, saying, explaining and telling relate a sender and a receiver. In ”reality”,
the sender here is the author, and the receiver is the reader of the article. Since the
two have different roles in the act, it is more felicitous to use la meg than la oss.
(20) La meg gi et eksempel på et spill i følge denne definisjonen.
‘Let me give an example …’
(21) La meg nevne tre av disse [antakelsene].
‘Let me mention three of these (assumptions).’
(22) Før jeg forsøker å besvare dette spørsmålet, la meg si noen ord om ambi-
sjonene til artikkelen.
‘…, let me say a few words about the ambitions of the article.’
(23) La meg forklare intuisjonen bak dette resultatet.
‘Let me explain the intuition behind this result.’
(24) I stedet for å gjøre det, la meg fortelle om en viktig re[ak]sjon på angreps-
koordineringsproblemet i litteraturen.35
‘…, let me tell (you) about an important reaction …’
(25) Før jeg forteller mer utførlig hvordan Monderer og Samet definerte nesten
åpen kunnskap, la meg gi intuisjonen bak deres tilnærming.
‘…, let me give the intuition behind their approach.’
(26) La meg skissere et argument – som jeg tror gir en plausibel forklaring på
hvordan folk klarer å koordinere handlingene sine uten at spillet er åpent
kjent.
‘Let me sketch an argument – …’

The last example of la meg-imperative, (26), involves sketching, which less clearly
differentiates between a sender and a receiver role. Some of the examples with
la meg can be classified according to the types of usage for ‘let me’ in 5.3. For in-
stance, (21) with nevne (‘mention’) is of the awareness type.36 Other examples are
more difficult to place, e.g. (24) with fortelle (‘tell’); it is probably safe to claim that
this is an unusual locution for a research article.

35. The article has relasjon (‘relation’), but it is clear that reaksjon (‘reaction’) is meant.
36. It is probably significant that this function is performed with a ‘let me’- rather than a ‘let
us’-imperative in Norwegian – recall that ‘let us’ in connection with awareness is more or less
confined to French examples. There appears to be a functional distribution of ‘let us’ and ‘let me’
corresponding to one formal type in French.
208 Academic Voices

As observed in 5.3, there is a tendency for ‘let us/me’-imperatives to co-occur

with paragraph transitions and (to a lesser degree) with section transitions. This is
also the case in this article, and more clearly for paragraphs than for sections. Out
of 12 imperatives, four are in the first sentence of a paragraph (viz. examples (15),
(16), (23) and (26) above), two in the second (examples (17) and (24)), and one or
two in the last (examples (25) and, possibly, (18)).37 As regards co-occurrence with
section transitions, this is not a strong tendency here: One imperative is in the first
paragraph of a section (example (21)), two in the second (examples (20) and (25)),
one in the second to last (example (16)) and one in the last (example (24)).
We will look more closely at two passages involving four of the examples. The
first passage, given as (27) below, is in the introductory section; it begins on the
second page and ends on the third and includes examples (15) and (18) above.
(The section goes on for about two thirds of a page.)
(27) [example of a Nash equilibrium] En Nash-likevekt er altså et par av handlin-
ger […]. […]
For å illustrere enda en Nash-likevekt, la oss gå tilbake til eksempelet
med de to bilene som møter hverandre uventet på en smal vei. Sjåførene kaller
vi for henholdsvis Adam og Eva. Anta at både Adam og Eva verdsetter […].
[…] La oss først se nøyaktig hvordan sjåførenes nytte er gjensidig avhengig av
hverandre i den følgende spillmatrise, før vi finner Nash-likevektene.
[figure: game matrix]
‘(example of a Nash equilibrium) Thus, a Nash equilibrium is a pair of actions …
To illustrate yet another Nash equilibrium, let us go back to the example
with the two cars that meet unexpectedly on a narrow road. We will call the
drivers Adam and Eve, respectively. Assume that both Adam and Eve value …
Let us first see exactly how the drivers’ gains depend on one another mutually
in the following game matrix, before we find the Nash equilibria.
(figure: game matrix)’
The paragraph with the two ‘let us’-imperatives is a very clear example of their ten-
dency to co-occur with (other bits of) metatext. The first imperative is preceded
by an adverbial of motivational metatext (For å illustrere …). This expression mo-
tivates the textual move of returning thematically to a previous scenario, which is
announced with the imperative. The second imperative co-occurs with not one,
but three metatextual adverbials, viz. først, i den følgende spillmatrise and før vi
finner … These are all instances of locational metatext. The first and last of these

37. The classification depends on whether a specific figure is considered to be in the middle
of or after the paragraph in question (i.e. whether the paragraph continues after the figure). The
problematic imperative is in the last sentence before the figure (see example (27)).
 Chapter 5. Reader/writer interaction 209

relate two text parts to one another sequentially; there is first an examination of a
matrix and then a calculation of equilibria. The second metatextual expression ex-
plicitly identifies the first element of the sequence as (primarily) consisting of the
matrix referred to. This example is an unusually clear illustration of the frequent
co-occurrence of ‘let’-imperatives and metatext.
This passage also illustrates the tendency for ‘let’-imperatives to co-occur with
paragraph transitions. The first imperative is in the first sentence of the paragraph,
the second in the last sentence before the matrix figure. The paragraph is not close
to a section transition, however.
The second passage, in (28), amounts to the last two paragraphs in the
third main section of the article and includes examples (16) and (24) above. There
is less metatext here than in the previous passage, but the adverbial I stedet for …
preceding the second imperative takes on a metatextual character in this context.
(28) [paragraph] La oss konkludere diskusjonen av angrepskoordineringsprob-
lemet. Angrepskoordineringsproblemet forteller oss at […]. […]
Hva jeg ikke har sagt, og vil heller ikke si noe om, er hvor generelt det
gjelder […]. I stedet for å gjøre det, la meg fortelle om en viktig re[ak]sjon på
angrepskoordineringsproblemet i litteraturen.
[new section]
‘(paragraph) Let us conclude the discussion of the coordinated attack prob-
lem. The coordinated attack problem tells us that …
What I have not said, neither will say anything about, is how generally it is
the case … Instead of doing that, let me tell (you) about an important reaction
to the coordinated attack problem in the literature.
(new section)’

The first imperative begins a new paragraph, and the second ends a paragraph
that is also the end of the section. In effect, the last two paragraphs of the section
conclude the section thematically and introduce the theme of the next, and the
imperatives make this explicit.
In the passages above, we see how ‘let’-imperatives direct the attention of the
reader and help to guide him or her through the article. There is a fairly strong
presupposition built into these locutions that the reader agrees with the author
and is willing to follow. More than is typically the case, this author (in this specific
article) takes on a pedagogical role, assigning a student role to the reader. This style
is more typical of lectures and text books (see Wales 1996: 65–66), and, as men-
tioned in the beginning of this section, it is possible that the history of the article
as a trial lecture has contributed to its tone.
210 Academic Voices

5.5.3 Metatext and author manifestation in a French medical article

In this section, we take a closer look at a French medical article that has no ex-
amples of the kinds of metatext that we investigated in section 5.2, nor of ‘let
us’-imperatives (5.3), viz. frmed32. The investigation in section 5.2 is limited to
a selection of metatextual expression types. The present case study also looks for
metatextual elements that were not included there.
We also look at author manifestation. The article has a lower relative frequency
of nous (‘we’; 4.3) and on (‘one’; 4.4) than the median for French medicine (with
absolute frequencies of four and three examples, respectively). That is, explicit
mention of the authors themselves is rare.38
We begin with author manifestation. In addition to the four examples of nous
(‘we’), there are 17 examples of the possessive notre/nos (‘our’). The explicit pres-
ence of the authors (and possible other non-author colleagues included in the ‘we’)
is therefore stronger than the figure for nous might lead us to believe. However, all
the examples of notre/nos and three of the examples of nous are in the discussion
section. The remaining example of nous is the first one in the text and is found in
the section ”Patients et méthode”; it is used in a description of the procedure for
registering information for each patient case. Except for this one example, there
is in fact no mention of the researchers or authors in the introduction, material/
method or (rather lengthy) results sections, which take the shape of a report and
are formulated in past tenses. Needless to say, the researchers are implicitly present
as the agents of research processes referred to by various impersonal expression
types, e.g. passives, nominalisations and non-finite verbal constructions (see the
case study in 4.6.2).
Most of the types of author manifestation found in the article are illustrated in
the passages in (29) and (30), which are both taken from the discussion section.
(29) Notre étude retrouve le caractère nosocomial de l’infection à CD. Cependant,
près de la moitié des cas sont considérés comme des infections acquises en de-
hors de CHU […]. Cette place importante des infections « extra-hospitalières »
dans notre série nécessite d’être confirmée par une étude prospective. Mais ce
premier constat mérite que l’on s’attache à informer davantage nos confrères
extra-hospitaliers. […] La mention « recherche de CD » doit également fig-
urer dans les prescriptions hospitalières.
‘Our study identifies the nosocomial character of CD infection. But almost
half the cases are considered to be infections acquired outside of the CHU

38. The relative frequency of negation is slightly higher than the median for French medicine,
while the ones for adversative conjunctions and bibliographical references are well above the
respective medians; see Chapter 6 for these features.
 Chapter 5. Reader/writer interaction 211

(hospital) … This important place of ”extra-hospital” infections in our series

needs to be confirmed by a prospective study. But this first report indicates
that one should seek to inform our extra-hospital colleagues. … Mention
should also be made of ”test for CD” in hospital prescriptions.’
(30) Le motif de prescription de l’antibiothérapie n’est pas mentionné dans les
séries de la littérature. Cependant, au vu de nos résultats, il paraît intéressant
de noter que dans 90 % des cas l’infection est banale […]. C’est probablement
pour cette raison que les épidémies hospitalières constatées dans notre étude
étaient surtout hivernales.
‘The reason for the prescription of antibiotic therapy is not mentioned for the
series in the literature. But in the light of our results, it seems interesting to
note that in 90 % of the cases the infection is banal … It is probably for this
reason that the hospital epidemics observed in our study occurred mostly in
the winter.’

As mentioned above, the research is mostly presented as an activity of the past.

However, in a few cases the research process is extended into the textual here-and-
now, e.g. when they note with interest the high share of ”banal” infections in (30).
Other examples of this include nous constatons (‘we observe’) and nous enregis-
trons (‘we register’), also in the discussion.
There is some use of persuasive adjectives in the text. Two instances are im-
portante in (29) and intéressant (‘interesting’) in (30). The clearest example is a use
of surprenant (‘surprising’). Other types of expressions that mark the presence of
thinking subjects include the epistemic modality markers sembler (‘seem, appear’)
and probablement (‘probably’) (see Vold 2005).
The passage in (29) contains two other interesting features, viz. the combined
motivation and ”self-obligation” to inform colleagues and the piece of advice that
testing for CD infection should become part of hospital prescriptions. This illus-
trates an aspect of applied science that is characteristic of many medical articles:
The research is conducted in order to gain knowledge that may help us fight dis-
ease and cure disorders, and this knowledge can serve as the basis for advising
colleagues in the medical profession. We see, then, that the presence of the authors
is a little stronger than the pronoun figures seem to indicate.
The features that we have mentioned primarily have to do with authorial pres-
ence. There is, however, also a certain amount of metatext in the article. One ex-
ample can be seen in (29) above: ce premier constat (‘this first report’). Some of the
metatext is so conventionally engrained as to be hardly noticeable, e.g. quant à (‘as
regards’) and il s’agit de (‘concerning’), à savoir (‘namely’). More noteworthy is the
way that reference is made to figures and tables, illustrated in (31) and (32). These
are taken from the section ”Patients et méthode”.
212 Academic Voices

(31) La Figure 1 fait apparaître deux périodes enregistrées au sein de l’unité post-
urgence.
‘Figure 1 shows two periods …’
(32) Les principaux motifs de prescription sont regroupés dans le Tableau 1.
‘The principal reasons … are grouped in Table 1.’

Significantly, the authors avoid mentioning themselves. In (31) this is achieved by me-
tonymy, in the personification of the figure. In (32) it is done by means of the passive.
Apart from these examples, there is virtually no metatext until the end of the
article. The last two paragraphs are quoted in (33).
(33) Enfin, il est important d’insister sur le retentissement économique qu’engendre
une telle complication. D’abord, elle occasionne un surcroît de travail […].
Ensuite, elle isole le patient […]. […]
En conclusion, une infection à CD doit toujours être évoquée […].
De plus, cette complication est classiquement connue comme infection
nosocomiale, mais notre expérience montre que l’infection à CD complique
aussi l’antibiothérapie prescrite en ville.
‘Finally, it is important … To begin with, it causes … Further, it isolates …
To conclude, a CD infection … In addition, this complication is classi-
cally known as …, but our experience shows …’

The enfin of the second to last paragraph signals the arrival of the last new point
to be mentioned, and the two sentence-initial adverbials d’abord and ensuite struc-
ture this part of the text. The last, informationally important, paragraph is simi-
larly flagged and structured by means of two metatextual adverbials (en conclusion
and de plus).
To conclude this case study, then, the amount of metatext is fairly sparse in
frmed32, but it is not completely absent. It is probably significant that it is most
clearly present in the concluding paragraphs, where the essence of the message to
be conveyed is given in brief.

5.6 Concluding remarks

In this chapter we have seen that discipline is a very important factor to take into
account when discussing reader/writer interaction, particularly regarding the
phenomenon of metatext. In addition, the language factor comes into play and
contributes to the observed variation in the corpus (cf. Table 3.2-1). In section 5.1
we stated that about 34 % of the variance is left unaccounted for. This is in fact the
lowest percentage observed for any of our features, i.e. it is the feature with least
 Chapter 5. Reader/writer interaction 213

individual variation. In terms of discipline, economics makes most frequent use

of metatext, while linguistics represents a clear second in this respect. Medicine −
most likely due to its fixed rhetorical text structure − has less need for providing
the reader with overt text structural signposts, and uses metatext only to a negli-
gible degree. In terms of language, English uses most metatext, illustrating its so-
called reader-oriented approach; Norwegian, in second place, also belongs in the
reader-oriented group, while French, ranked third in all three disciplines, may on
this basis be characterised as a writer-oriented language. French, however, turned
out to be the most frequent user of ‘let us’-constructions, which in our material
turned out to typically co-occur with metatextual expressions (5.3.3). Hence, the
inclusion of this construction to some extent blurs the picture drawn above of the
three languages as belonging in two different groups in terms of reader/writer in-
teraction. As for the discipline variable, the ‘let us’-constructions basically display
the same pattern as our metatext elements, implying that they are more frequent in
economics and linguistics than in medicine. Thus, our findings show that medicine
is clearly the discipline with least overt interaction between reader and writer.
In terms of the overt textual presentation of ‘results’ and ‘conclusions ‘, when
the author is signalling his or her knowledge claims to the readers, our study re-
veals that English-writing authors are again more explicit than their French col-
leagues, with Norwegian authors occupying a middle position between the two.
As regards the discipline dimension, economics in this case, in contrast to several
of the other features studied, turns out to be more similar to medicine than to lin-
guistics; hence in this respect economics shows an affiliation to the hard sciences.
Thus, our hypotheses regarding reader/writer interaction (section 5.0) have
on the whole been confirmed. English does display overt interaction to a higher
degree than French; for Norwegian, our expectation that author behaviour would
be similar to English (albeit a little less explicit) also turns out to hold true. With
respect to the discipline variable, economics and linguistics authors on the whole
are more visible in their texts than their colleagues within the medical profession.
 chapter 6

Presence of the others

6.0 Introduction

We have summarised our main object of study in the notion of person manifesta-
tion (see section 1.1). In a general textual perspective, this notion may easily be
associated with author- or self-manifestation (first person) only. However, in our
study of the research article, the notion of person manifestation covers reference
to all three persons (first, second and third). In our distinction between self, self &
other and other, the self-dimension corresponds to first person as manifested by
the presence of the author(s), self & other to first and second person as manifested
through the interaction between author and reader, and other mainly to third per-
son as manifested by the presence of researchers other than the author(s), but also
to second person manifested through references to the reader(s). The two previous
chapters have dealt primarily with the presence of the author (self-dimension in
Chapter 4) and with the interaction between writer and reader (self- & other-di-
mension in Chapter 5). We now direct our attention towards linguistic features
that the authors use in order to bring other researchers (third persons), their re-
search and viewpoints into the text. Thus, the focus is on the other-dimension,
and we mostly address our second main question: How are the voices of other
researchers reflected?
When studying person manifestation in the research article, bringing in the
third person or other researchers is obviously relevant. Given the fast develop-
ing and competitive nature of research today, scientific authors need to refer to
previous publications and to compare their results with previous findings in order
to present their claims and discuss scientific progress. With a rhetorical view of
scientific discourse as something which is created in a particular and multivoiced
communicative situation, it becomes natural to pay special attention to the mixing
of voices of self and others, i.e. polyphony.
In recent years, Bakhtinian ideas on dialogism and polyphony in all kinds of
discourse studies have pervaded studies of academic discourse as well (see Bondi
2005; Bondi & Silver 2004). Placing ourselves to some extent in this tradition, we
have chosen as our theoretical framework the ScaPoLine version of linguistic po-
lyphony (see section 2.1.2). This framework is particularly relevant for the present
216 Academic Voices

chapter. It gives us the opportunity to go further than the study of reported speech
as realised through bibliographical references and citations alone. As already stat-
ed in section 2.1.2, the ScaPoLine theory provides a framework where a series of
linguistic means function as signals of voices in text. We follow Thompson (1996:
502) who states that “many reports are expressed by means of structures other
than quotes or reported clauses”. In an approach which is more functional than
structural, his purpose is “to suggest an overall framework for the description of
‘language reports’” (ibid.: 502), defined as “signalled voices in text”. He takes as
his point of departure the four dimensions of choice for the reporter: voice, mes-
sage, signal and attitude. We find the voice dimension particularly relevant, which
specifies the five options of self, specified other(s), unspecified other(s), commu-
nity and unspecifiable other(s) (ibid.: 524). Even though structured and labelled
somewhat differently, these options are covered by the discourse beings assumed
by ScaPoLine (see 2.1.2). In our view, the advantage of ScaPoLine is that it covers
a larger spectrum of signals, within one homogeneous framework. Thompson lists
reporting clause, adjunct, noun, adjective and verb as structural options for lan-
guage reports (ibid.: 524), while ScaPoLine considers in addition various adverbs
and connectives as important signals of voices or points of view.
In this chapter we focus on the manifestation of others’ voices in two perspec-
tives: explicit presence, mainly in the form of bibliographical references (section
6.2), and implicit presence, mainly through the manifestation of others in negation
and in concession constructions signalled by the adversative connective ‘but’ (sec-
tion 6.4). Our decision to study the use of general features such as negation and
adversative connectives in research articles may call for an explanation. The main
reason is that both negation and the adversative ‘but’ signal polyphonic patterns.
In addition there is an interesting rhetorical aspect related to these features: Both
negation and adversatives may convey polemic attitudes. Salager-Meyer, Alcaraz
Ariza and Zambrano (2003) study the socio-pragmatic phenomenon of academic
conflict, realised by various linguistic features, in Spanish, French and English
medical articles. We see our analysis of negation and adversative ‘but’ as comple-
menting their study.
The distinction between explicit and implicit presence is a methodological
one. Implicit and explicit manifestations of voices are, in fact, often interwoven, as
can be observed in the following examples (emphasis added):
(1) Blanchard and Watson (1986) and Blanchard and Quah (1989) found evi-
dence that demand shocks were the main source of US fluctuations, but Sha-
piro and Watson (1988) and Gali (1992) found that supply shocks predomi-
nated. (engecon22)
 Chapter 6. Presence of the others 217

(2) There is an interrelation between the absence or presence of the TO-phrase

experiencer and the agency of the subject of these verbs. It is not enough to
claim, as Brinton (1988, p. 31, following Mourelatos, 1978, p. 419) does, that
there is aspectual multivalence. The different aspectual behaviour we see with
these verbs is associated with the presence or absence of an argument – the
experiencer – and with the polysemy of the verbs of appearance. (engling39)
In (1) there is explicit other-presence through references combined with a con-
cessive but-construction manifesting implicit polyphony; in (2) there is explicit
presence through references, introduced by a negation construction indicating
implicit polyphony (It is not enough to claim, …). In addition, in (2) the plural pro-
noun we indicates that the author is bringing the reader and/or a larger discourse
community into the argumentation (engling39 is a single-author article).
These examples illustrate that in academic language explicit other-manifesta-
tion realised through bibliographical references and implicit polyphonic construc-
tions are often interrelated, and they are, of course, both polyphonic. The notion of
linguistic polyphony can handle both the explicit and implicit perspectives, as well
as the combination of the self- and other-dimensions within a coherent theory
and by means of a homogeneous methodological apparatus. It can handle both
explicit citations in some form of reported speech, as well as implicit presence as
manifested, for example, by an adversative and concessive connective like but, or
by the negation not (see examples above).
As regards bibliographical references, what we call explicit presence of others,
these constitute a domain within academic discourse which has been extensively
studied. In our qualitative analysis of referencing behaviour in research articles,
we can therefore lean on previous research (e.g. Swales 1986, 1990; Salager-Meyer
1998, 1999; Valle 1999; Hyland 2000; we will come back to some of these in section
6.2). However, as far as we know, no studies take into account the doubly con-
trastive language–discipline perspective adopted in this book. This implies that
there is little prior research on which to base well-founded hypotheses concerning
quantitative characteristics of the referential behaviour in the three languages and
the three disciplines studied. (We will however return to a possible comparison
with Hyland’s 2000 study of eight different disciplines.) Hence the hypotheses we
set forth are somewhat speculative in nature. If we take into consideration the size
and the age of our three disciplines, we might nevertheless hypothesise, tentatively,
that references are more frequent in medicine than in linguistics, and more fre-
quent in linguistics than in economics. Likewise, if we take into consideration the
extension of the three languages and the amount of research reported in English,
French and Norwegian language journals, we expect English to have the highest
frequency of references, French to rank second highest and Norwegian third. In
218 Academic Voices

order to contribute to answering our key research issue related to potential cul-
tural identities within academic discourse (see section 1.1), we need to undertake
quantitative analyses which can provide new insight into the frequency of bib-
liographical references in research articles in different disciplines and written in
different languages (see section 6.1).
Polyphonic constructions as presented within ScaPoLine, in particular con-
structions indicating implicit presence of others, are a much less studied domain,
at least in the Anglo-American tradition. In 6.4 we will return to previous studies
relevant to the qualitative analysis of these constructions. However, as far as we
know, the quantitative aspect of negation and adversative connectives in academic
discourse has only been studied within the KIAP project. If we focus on the po-
lemic aspect of these constructions, we might, as mentioned above, relate our is-
sue to the studies of criticism in Spanish, French and English medical discourse
undertaken by Salager-Meyer et al. (2003). We can on that basis hypothesise that
French medical authors use more negations and adversatives than their English
counterparts (for further comments see 6.1).
Our quantitative data are presented and discussed in section 6.1. Section 6.2
provides a qualitative study of different types of bibliographical references as well
as their distribution in different disciplines and languages. We focus particularly
on references with grammatical subject function. In section 6.3, we relate the oth-
er-dimension to the self-dimension. This section focuses on a comparison of what
roles the authors assign to themselves and to the others in the polyphonic drama.
By comparing verbs combined with the reference source in subject position with
verbs combined with first person pronominal subject referring to the authors, we
attempt to answer the question of “who can do what?”. Finally, negation (‘not’) and
concessive constructions signalled by the adversative connective ‘but’ are issues
dealt with in section 6.4. Section 6.5 presents a case study of the use of negation in
three English articles (one from each discipline), while we in section 6.6 sum up
and discuss our main results in some concluding remarks.

6.1 Quantitative and comparative results

This section presents results of our last three main quantitative investigations. Bib-
liographical references are the subject of 6.1.1, negation is treated in 6.1.2, and
6.1.3 deals with adversatives.
 Chapter 6. Presence of the others 219

6.1.1 Bibliographical references

The bibliographical references we have included in our quantitative study com-

prise publications of the years 1900–1999 and 2000–2003. We have not counted
instances where the reference takes the Latin forms of ibid., op.cit., loc.cit., etc.
Further, instances with a reference of the form Hansen 1999a, b, c have only been
counted as one occurrence. In the medical articles we have counted the note num-
bers referring to the list of references at the end of the article. References integrated
in notes have not been counted, the reason being that we have chosen to limit the
study to the main article text (the body part; see 1.2). Central data concerning the
frequency of bibliographical references are presented in Table 6.1-1.
We saw in 3.2 (Table 3.2-1) that as much as 44 % of the variance in the relative
frequency of bibliographical references in the KIAP Corpus is due to differences
between disciplines, whereas language differences contribute only 5 % (cf. also
Appendix B). The contribution of factor interaction is also significant, viz. 2 %.
About 49 % of the variance is unaccounted for.

Table 6.1-1. The use of bibliographical references

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 50 1279 0.437 0.396 0.258 0.173

engling 50 4184 0.930 0.814 0.484 0.234
engmed 50 2663 1.630 1.604 0.604 0.371
frecon 50 1366 0.445 0.350 0.347 0.148
frling 50 1040 0.485 0.443 0.401 0.174
frmed 50 2019 1.411 1.040 1.673 0.410
noecon 50 1583 0.493 0.472 0.321 0.191
noling 50 2088 0.868 0.674 0.744 0.238
nomed 50 1819 1.837 1.500 1.202 0.450

econall 150 4228 0.458 0.401 0.310 0.168

lingall 150 7312 0.761 0.637 0.592 0.238
medall 150 6501 1.626 1.427 1.243 0.447

engall 150 8126 0.999 0.821 0.678 0.486

frall 150 4425 0.780 0.510 1.102 0.298
noall 150 5490 1.066 0.744 1.007 0.470

KIAP 450 18041 0.948 0.688 0.952 0.421

Relative frequencies and deviations in %.
220 Academic Voices

The discipline factor

The relative frequency of bibliographical references is clearly highest in medicine,
and it is higher in linguistics than in economics. The differences between medicine
and economics and between medicine and linguistics are significant both at the
superlevel and at the sublevel (in each language).1 The same holds for the differ-
ences between linguistics and economics, except at the French sublevel, where the
difference is not significant.
The overall findings confirm our hypothesis tentatively formulated in 6.0,
namely that references are more frequent in medicine than in linguistics, and more
frequent in linguistics than in economics. Our selection of three disciplines is ob-
viously too narrow, however, for us to posit a claim of general validity that there
is a relation between the size and age of a particular discipline and the number of
bibliographical references used in the same discipline.
As regards our English subcorpus, it is interesting to compare our findings
with those presented by Hyland (2000) in his study of 80 articles from eight disci-
plines. His general conclusion related to frequency is formulated as follows:
The figures broadly support the informal characterisation that softer disciplines
tend to employ more citations, with engineering and physics well below the aver-
age, although the frequencies for molecular biology seem to differ considerably
from this picture. (Hyland 2000: 24)

Since the disciplines he studies are not the same as those included in the KIAP
Corpus, a direct and general comparison is not possible, but we can look at the fig-
ures related to some of the disciplines studied by Hyland. If we consider molecular
biology as a discipline relatively close to medicine, we see a certain correspond-
ence as regards relative frequency: In Hyland’s study, the relative frequency of bib-
liographical references in molecular biology is 1.55 %; in our study, it is 1.63 % in
English medicine (1.83 % for Norwegian medicine). Other comparisons which
could be made, still keeping in mind that the disciplines are not the same in our
corpus, are the following: In his study of applied linguistics, Hyland observes a
relative frequency of 1.08 %, while we have observed a relative frequency of 0.93
% in English linguistics. In Hyland’s study, the relative frequency in sociology is
1.25 % while it is only 0.44 % for English economics in our material. In conclusion,
we see that our findings related to the English subcorpus correspond fairly well to
those of Hyland (2000), except for the discipline of economics to which there is no
directly comparable counterpart.

1. The terms superlevel and sublevel are explained in 2.2.5.

 Chapter 6. Presence of the others 221

The language factor

English and Norwegian have similar frequencies of bibliographical references, and
there are more in these two languages than in French. At the superlevel, the dif-
ferences between English and French and between Norwegian and French are sig-
nificant, but not that between English and Norwegian. At the sublevel, all the three
language differences are significant in linguistics, with English first, Norwegian
second and French third. In medicine, the differences between English and French
and between Norwegian and French are significant, but the difference between
English and Norwegian is not. In economics, there are no significant differences
(and there are slightly more references in Norwegian than in English).
The overall findings correspond only to a modest degree to our hypothesis for-
mulated in 6.0 suggesting more references in English than in French and Norwe-
gian, and more in French than in Norwegian. Norwegian articles are in fact quite
close to English articles while French articles have a substantially lower frequency
of bibliographical references, even if this language has a considerably larger ex-
tension than Norwegian (see also 6.6). One tentative explanation might be that
Norwegian culture is traditionally more oriented towards Great Britain and the
US, and that Norwegian researchers commonly orient themselves towards inter-
national research. As regards French academics, they may traditionally have been
less oriented towards research undertaken and published outside the francophone
world.
However, the differences are more modest in the language perspective than in
the discipline perspective. This might be taken as support for our general hypoth-
esis of greater differences between disciplines than between languages.2

The gender factor

In single-author articles, there are no significant differences in the relative fre-
quency of bibliographical references between articles written by a man and articles
written by a woman in any of the subcorpora, nor in combinations thereof.3

2. The results of an exploratory study comprising 180 articles, reported in Fløttum (2003e),
are very similar to the results obtained for the total corpus. The difference is that the frequency
of bibliographical references in medical articles is higher when the whole corpus is taken into
account (1.41 % in the 180 articles and 1.56 % for the whole corpus). The difference holds for all
the language corpora with the following figures: engall 0.82 % versus 1.00 %; noall 0.85 % versus
1.07 %; frall: 0.65 % versus 0.71 %.
3. Mann–Whitney–Wilcoxon test, α = 0.05; see section 2.2.5.
222 Academic Voices

6.1.2 Negation

The search for expressions in this main quantitative investigation includes English
not, French ne … pas and Norwegian Bokmål ikke, Nynorsk ikkje. (The English
contractions -n’t and cannot were not included in the quantitative analysis. Nor
were any other kinds of negating expressions.) As we saw in 3.2, differences be-
tween disciplines and differences between languages contribute to almost exactly
the same share of the variance in the relative frequency of negation in the KIAP
Corpus, viz. 18 % (cf. also Appendix B). There is no significant interaction be-
tween the discipline and language factors for negation. Most of the variance, about
64 %, cannot be attributed to either factor or their interaction.

Table 6.1-2. The use of negation

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation

engecon 50 1194 0.388 0.323 0.193 0.127

engling 50 2822 0.623 0.595 0.264 0.198
engmed 50 717 0.442 0.410 0.252 0.142
frecon 50 1084 0.371 0.359 0.128 0.067
frling 50 1541 0.654 0.612 0.282 0.212
frmed 50 511 0.354 0.345 0.177 0.131
noecon 50 2059 0.646 0.644 0.192 0.106
noling 50 2604 0.967 0.893 0.336 0.237
nomed 50 628 0.583 0.566 0.246 0.171

econall 150 4337 0.468 0.430 0.214 0.156

lingall 150 6967 0.748 0.716 0.332 0.219
medall 150 1856 0.460 0.423 0.245 0.151

engall 150 4733 0.484 0.440 0.258 0.176

frall 150 3136 0.460 0.394 0.247 0.134
noall 150 5291 0.732 0.682 0.312 0.172

KIAP 450 13160 0.559 0.507 0.300 0.192

Relative frequencies and deviations in %.

The discipline factor

Negation is clearly used more in linguistics than in economics and medicine. The dif-
ferences between linguistics and economics and between linguistics and medicine are
significant both at the superlevel and at the sublevel (in each language). There is no sig-
nificant difference, however, between economics and medicine, at either level. The same
picture of disciplinary differences and similarities is found for adversatives (see 6.1.3).
 Chapter 6. Presence of the others 223

The language factor

Negation is used more in Norwegian than in English and French. The differences
between Norwegian and English and between Norwegian and French are signifi-
cant both at the superlevel and at the sublevel (in each discipline). There is no
significant difference, however, between English and French, at either level. It is
probable that one factor behind the higher frequency of negation in Norwegian is
a grammatical one, viz. the fact that the negative quantifier ingen (‘no’) has a more
restricted use than corresponding English and French expressions; constructions
involving ikke/ikkje (‘not’) are frequently used instead. But Norwegian also has
more adversatives than the other languages (see 6.1.3), and this, together with the
high frequency of negation, might indicate a stronger argumentative character of
Norwegian research articles (see discussion in section 7.1.) (For further comments
on language differences, see 6.1.3.)

The gender factor

In single-author articles, no significant differences in the relative frequency of ne-
gation between articles written by a man and those written by a woman have been
found in any of the subcorpora or combinations thereof.4

6.1.3 Adversatives

Our quantitative investigation of adversatives is limited to adversative conjunc-

tions, viz. English but, French mais and Norwegian men. Thus, words like however,
nevertheless etc. and corresponding French and Norwegian expressions were not
included in the corpus searches.
The reasons that we limit this part of our study to only three conjunctions,
or connectives as we often call them, are both qualitative and quantitative. First,
these connectives are interesting in a polyphonic perspective and thus fit well with
the overall approach of this book. Second, they are the most frequent and rel-
evant connectives (already confirmed in a pilot study; see Breivega et al. 2002).
And third, the three connectives but, mais and men are semantically compara-
ble. A comparison of connectives in different languages requires in fact careful
semantic-pragmatic considerations. Even though some connectives correspond
more or less to each other, they often have particular semantic-pragmatic values
in different languages. Adversative ‘but’ in different languages may also convey
different meanings; however, they have in common a concessive meaning. This is
the meaning which is particularly interesting to the polyphonic perspective and
which will be further developed in section 6.4 (when concessive, these connectives

4. Mann–Whitney–Wilcoxon test, α = 0.05; see section 2.2.5.

224 Academic Voices

function as markers of polyphony). There are of course other connectives which

it could have been interesting to study. If we look at connectives that may convey
a concessive meaning, the following clearly deserve further study in the context
of academic discourse (number of tokens in our corpus in parentheses): English
however (1,282), French cependant (378), Norwegian imidlertid (846), English al-
though (521), French bien que (100) and Norwegian selv om (350). However, they
are less frequent than the three we have chosen to study (the figures for but, mais
and men are 2,296, 1,632 and 2,445 respectively; see Table 6.1-3) and it would have
been difficult to pick out clear counterparts in each language for a systematic con-
trastive analysis. Our investigation hence has limited value for making generalisa-
tions, but it seems unique in its kind up to now.
As seen in 3.2, 25 % of the variance in the relative frequency of adversative
conjunctions in the KIAP Corpus can be attributed to disciplinary differences,
while language differences contribute 9 % (cf. also Appendix B). The interaction
contributes 1 %, which is not significant. This means that about 65 % of the vari-
ance is unaccounted for.

Table 6.1-3. The use of adversative conjunctions

Corpus No. of No. of Mean Median Standard Quartile

articles tokens rel. fr. rel. fr. deviation deviation
engecon 50 592 0.197 0.188 0.105 0.063
engling 50 1428 0.319 0.270 0.195 0.059
engmed 50 276 0.171 0.137 0.120 0.072
frecon 50 513 0.163 0.148 0.093 0.070
frling 50 870 0.382 0.379 0.141 0.096
frmed 50 249 0.181 0.161 0.099 0.068
noecon 50 815 0.254 0.235 0.147 0.098
noling 50 1293 0.463 0.454 0.152 0.065
nomed 50 337 0.325 0.304 0.223 0.125

econall 150 1920 0.205 0.194 0.122 0.070

lingall 150 3591 0.388 0.377 0.173 0.115
medall 150 862 0.226 0.175 0.171 0.103

engall 150 2296 0.229 0.207 0.158 0.085

frall 150 1632 0.242 0.210 0.150 0.097
noall 150 2445 0.347 0.340 0.197 0.127

KIAP 450 6373 0.273 0.239 0.177 0.116

Relative frequencies and deviations in %.
 Chapter 6. Presence of the others 225

The discipline factor

Adversative conjunctions are clearly used more in linguistics than in economics
and medicine. The differences between linguistics and economics and between
linguistics and medicine are significant both at the superlevel and at the sublevel
(in each language). There is no significant difference, however, between economics
and medicine, at either level. It is noteworthy that this picture is exactly the same
as for negation (see 6.1.2).

The language factor

Adversatives are more frequently used in Norwegian than in English and French.
The differences between Norwegian and French are significant both at the super-
level and at the sublevel (in each discipline). The same holds for the differences
between Norwegian and English, except at the economics sublevel, where the dif-
ference is not significant. At the linguistics sublevel, there is more frequent use
of adversatives in French than in English, and the difference is significant. At the
economics and medical sublevel, there are no significant differences between these
two languages, nor is there a significant difference at the superlevel.
In section 6.0 we indicated that if focusing on the polemic aspect of both ne-
gations and adversatives (for further details on the polemic aspect, see section
6.4), we may compare our findings related to the studies of academic conflict un-
dertaken by Salager-Meyer et al. (2003). They investigated academic conflict in
Spanish, French and English medical articles, published between 1930 and 1995.
Their overall findings indicate that French (and Spanish) medical authors tend
to be more critical (and authoritarian) than their English-speaking counterparts
(and for French this discoursal pattern has not changed substantially over time).
These findings correspond to ours, but only to a very modest degree, and only to
the extent that negation and the adversatives but and mais are used with a polemic
meaning. In our corpus French medical authors use slightly more negation and
adversatives than their counterparts writing in English, but the difference is not
significant.5
If we transfer this line of argument from the discipline of medicine to a general
discipline perspective, our findings indicate that linguists are more polemic than
economists and medical researchers. Further, Norwegian authors are more so than
their English counterparts.

5. In their study of negative appraisals in academic book reviews (BR), Salager-Meyer and Al-
caraz Ariza (2004: 168) conclude that “[t]he frequency, […] would tend to be language/culture
dependent, English- and French-written BRs being less critical than their Spanish counterparts.”
However, as regards the quantitative differences between English and French, the percentage of BRs
with negative appraisals is slightly higher in French (76 %) than in English (70 %), (ibid.: 156).
226 Academic Voices

The gender factor

For single-author articles, one subcorpus displays a significant gender difference,
viz. Norwegian linguistics, where male authors use adversatives more frequently
than female authors. There are no other significant gender differences for adversa-
tives in any of the other subcorpora, nor combinations thereof.6

6.2 Explicit presence: bibliographical references and citations

The topic of this section is the explicit presence of others as manifested in biblio-
graphical references. With the quantitative data presented in 6.1.1 as a background,
we first present the four subtypes studied (6.2.1) and then look into the distribu-
tion of these subtypes (6.2.2). Further, we focus on the type of bibliographical
references in which the reference is used as a grammatical subject (6.2.3). In 6.2.4
we discuss functions that bibliographical references can assume.

6.2.1 Four subtypes

Looking behind the general figures presented in 6.1.1, we propose to differenti-

ate four subtypes of bibliographical references. These subtypes can tell us a lot
more about how and to what extent other researchers are given the opportunity
to manifest themselves in a research article. We have studied different realisations
of explicit references through four relatively broad categories proposed below
(R1–R4).7
Like Salager-Meyer (1998, 1999), Koskela (1999), Hyland (1999, 2000), Bondi
& Silver (2004) and many more, we were inspired by Swales’ distinction between
integral and non-integral references in the setting up of these categories:
An integral citation is one in which the name of the researcher occurs in the actual
citing sentence as some sentence-element; in a non-integral citation, the research-
er occurs either in parenthesis or is referred to elsewhere by a superscript number
or via some other device. (Swales 1990: 148)

However, since to us there seems to be an important qualitative distinction between

those references where the name of the researcher does not occur in the main text
and those where it does (be it within parentheses or not), we have defined our
categories somewhat differently. There is a second important distinction between
references where the actual researcher is only mentioned in one way or another

6. Mann–Whitney–Wilcoxon test, α = 0.05; see section 2.2.5.

7. The same categories were used in Fløttum (2003e).
 Chapter 6. Presence of the others 227

and those where he or she is actually given the floor, i.e. by a verbatim quote. These
reflections have resulted in a division into four categories in the present study:
R1 – Non-integral reference: Little Lake is polluted [1].
R2 – Partly integral reference: Little Lake is polluted (Clark 1999).
R3 – Semi-integral reference: Clark (1999) has observed that Little Lake is polluted.
R4 – Fully integral reference: Clark (1999) claims: “Little Lake is polluted.” / Clark
(1999) claims that “Little Lake is polluted”.

As regards R1, numbers inserted in brackets or parentheses or superscript num-

bers refer to a bibliographical list (typically an end-list). As regards R2 and R3,
they are characterised by the presence of a cited author’s name and the year of
publication, within parentheses (R2) or as an integrated element of the sentence
in which the reference occurs (R3). Finally, R4 represents the case where the cited
author is quoted; quotes are here defined as 3 or more words included by direct or
indirect reported speech.
In the following, some typical authentic examples are provided (for practical
reasons most of the examples are taken from the English subcorpus). First the R1
– Non-integral type:
(1) One pupillary manifestation of DAN is small pupil size but the mechanism for
this is unclear despite a number of previous studies that have assessed pupil
function in diabetic patients using both static and kinetic methods. [2-12]
(engmed34)

It is clear that the authors refer to eleven specific studies, indicated by a number of
previous studies and the numbers [2-12], but the name and year of each publica-
tion is placed in a reference list at the end of the article, outside the main text. The
reader gets to know the results of the research referred to, but when it comes to
who carried them out or how old the studies are, this information is only available
through the bibliographical list. Let us look at another example:
(2) A large number and wide variety of medications approved for use by the US
Food and Drug Administration (FDA) are available to the US population,
and expenditures on drugs have increased dramatically in recent years. [1-
3] New prescription drugs are continually introduced, and older drugs are
increasingly available over the counter (OTC), making self-medication com-
monplace. Adverse reactions to drugs are among the leading causes of hospi-
talization and death in this country. [4,5] (engmed20)

This example simply refers to a set of facts, without any explicit mention of a study.
We must, however, believe that the presented facts are accounted for in the publica-
228 Academic Voices

tions referred to by the numbers [1-3] and [4,5]. This type of more or less objective-
ly presented fact is also quite commonly found in the partly integral reference, R2:
(3) Another difficulty is the sheer size of theses and dissertations as texts for anal-
ysis (Swales, 1984; Swales and Thompson, 1999). (engling11)

This example is not very different from (2). But the mention of name and year
makes this reference more informative. There might, however, be some confusion
as to who did or found exactly what when there is a cluster of references in the
same sentence:
(4) Hence, we follow recent work in cognitive linguistics (Fauconnier 1994, 1997,
Fauconnier & Turner 1998, Turner 1991, 1996), which posits that formal lin-
guistic expression underspecifies for meaning. (engling17)

When faced with a reference of this type, if one does not know the publications
referred to, it becomes legitimate to ask the following question: Do all the publica-
tions referred to say “that formal linguistic expression underspecifies for meaning”
and in the same way?
In a study concerning the specific phenomenon of cluster, defined as the ac-
cumulation of three or more bibliographical references represented within paren-
theses as in the R2 type, Fløttum (2004i) claims that even if clustering represents
an efficient way of referring to and positioning oneself in relation to other relevant
research, it may lead to considerable ambiguity and even go against the academic
ideal of clarity.
There are, of course, situations where clustering is practical (for example when
referring to a group of publications treating the same subject or belonging to the
same tradition), but in many cases it may be pertinent to ask why such accumula-
tions occur. Here is an example from engling16:
(5) However, the relevant dimensions of discourse meaning have proved quite
hard to characterize. The tunes have variously been associated with social
attitude (O’Connor and Arnold 1961 and in a different sense Merin 1983),
illocutionary acts (Liberman and Sag 1974, Sag and Liberman 1975, Liber-
man 1975), propositional attitudes and maintenance of mutual belief (Ward
and Hirschberg 1985, Pierrehumbert and Hirschberg 1990, Bartels 1997), and
various notions of information structure or packaging (Halliday 1967a, Jack-
endoff 1972, Schmerling 1976, Ladd 1980, 1996, Gussenhoven 1983, Selkirk
1984, Terken 1984, Terken and Hirschberg 1994, Morel 1995, Rochemont
1986, Steedman 1991a,b, Zubizarreta 1998). (engling16)

The first two clusters are in fact introduced and specified in a quite informative
way; but the last one, containing 14 references, is introduced rather vaguely by
 Chapter 6. Presence of the others 229

the complex noun phrase various notions of information structure or packaging.

With many clusters like these, it is not surprising that the total number of occur-
rences in this article is as high as 185. However, this example must be considered
an extreme one.
The semi-integral reference, R3, is more informative than R2 due to the pres-
ence of the name of the cited researcher as an integrated member of the citing
utterance. The reader is told who did or found what:
(6) Deci and Ryan (1985) and Lepper et al. (1973) find that if people are given
external rewards for pursuing an activity, their existing internal motivation is
undermined. (engecon20)

The most prominent syntactic role a cited researcher can get seems to be the syn-
tactic subject (more about these constructions below), as in example (6). But there
is considerable variation within category R3, and often researchers are only re-
ferred to through a passive construction, as in (7):
(7) As shown by Ludema (2001), the above properties can be satisfied by a punish-
ment path in which countries play symmetric Nash tariffs […]. (engecon38).

The final category, R4, fully integral reference, is the one in which the voice of
the other(s) is the most explicit. This type comprises the cases where an author is
directly quoted. The author(s) might be introduced formally by a reporting verb
like ‘say’ or ‘write’ (or by other, non-reporting verbs) followed by a colon and with
or without quotation marks, and often marked off graphically, as in the following
reference to Poustma:
(8) Jespersen (1942: 117) coined the expression ‘light verb’, but the phenomenon
in English was noted earlier by Poutsma (1926: 394–400) and Kruisinga (1932:
198–200). Poutsma writes (p. 394):
There is a marked tendency in Modern English to express a verbal idea by
means of a combination consisting of a verb with a vague meaning and a noun
of action. The latter is then the real significant part of the predicate, while the
former mainly serves the purpose of the connective. (engling10)
In this example, the citation ‘light verb’ taken from Jespersen is not counted, while
the quote taken from Poutsma is (There is a marked tendency … the connective).
We also find many complex mixed forms where the citing author mixes his
own words with one or several cited authors’ words in quotation marks:
(9) Discussing SEE, Croft (1991, p. 220) says that ”since experiencer and stimulus
are both simultaneously initiator and endpoint, they are identical in causal
230 Academic Voices

structure”, but this observation is at odds with Langacker’s (1991, p. 304) claim
that (for SEE) ”the object’s semantic role is zero”. (engling39)
The explicit presence of other(s) as realised through R3 and R4 corresponds to
what is commonly called reported or represented speech. If we understand po-
lyphony as the presence of several voices in one and the same text, or even in
one utterance, reported speech is the polyphonic phenomenon par excellence (see
Nølke et al. 2004: 57). The locutor (corresponding to the author) gives the floor
to another person, the represented locutor, either directly by transferring com-
plete responsibility for the reported sequence to the represented locutor, as in R4,
or indirectly by transferring responsibility of content only (and not form) of the
reported sequence to the other, as in R3. In the R1 and R2 types, the other is also
introduced explicitly, but without being given the floor. However, the reference
convention indicates that the other’s voice or point of view is represented in some
way or another.

6.2.2 Quantitative distribution of the four subtypes

In a disciplinary perspective, our findings related to the distribution of the four

subtypes of bibliographical references more or less confirm our working hypoth-
esis that medical authors rarely give the floor to others while economists and es-
pecially linguists report or cite other researchers more frequently. In the KIAP
Corpus, R1 is the type typically used by medical authors (relative frequency 1.4
%8 for the total medical corpus of 150 articles). This is not surprising given “the
highly professionalized, structured and conventional character of the late 20th
century medical research and medical research writing”, as Salager-Meyer (1999:
300) states in her well-known diachronic study of referential behaviour in Eng-
lish medical discourse. The numerical end-list pattern has become “the almost
unique referential pattern in today’s medical discourse” (ibid.: 295). This is a pat-
tern which is also imposed on medical authors by the large majority of medical
journals. However, even if medical authors typically relate to colleagues that are
nameless in the body text, these colleagues are present by the number referring to
the end-list, and their voices are mixed in some way or other with the voice of the
authors. It may also be argued that this type of hiding the represented voice makes
it easier for the authors to present the reported text “in a way that most effectively
supports their own argument” (Hyland 2000: 26).

8. For R1 the relative frequency is calculated as the percentage of reference numbers in relation
to the total number of words; for R2–R4 the relative frequency is calculated as the percentage of
publication year references (counted as 1 word) in relation to the total number of words.
 Chapter 6. Presence of the others 231

R2 and R3 are the types preferred by linguists and economists. For R3, the
relative frequency is the same for the two disciplines, viz. around 0.3 %. As regards
R2, linguists (and especially English linguists) use it much more than economists
(the relative frequency is around 0.4 % and 0.15 % respectively). R4, the fully inte-
gral reference, i.e. with a direct quote, is used only to a very limited extent (relative
frequency of 0.03 % for the total corpus), and mainly by linguists (very seldom by
economists, and hardly ever by medical researchers; see also Salager-Meyer 1999).
The R3 and especially the R4 types represent the most faithful way of bringing the
voices of others into the text; however, the preference seems to be for the other
types.
Even though Hyland’s study (2000) is not directly comparable to ours, it might
be interesting to report some of his findings in this context. He investigates the use
of references (or “citations”, which is the term he uses) in eight different disciplines
(only English). He adopts the binary distinction of Swales (1990) between integral
(corresponding to our R3 and R4) and non-integral (corresponding to our R1 and
R2) references. In contrast to the variation observed in the frequency of refer-
ences, Hyland finds that “there was far less variation in the ways disciplinary com-
munities refer to sources, with all but philosophy displaying a distinct preference
for non-integral structures” (Hyland 2000: 24). To some extent this corresponds to
our general findings (for example that R4, the fully integral reference with a direct
quote, is very infrequent and is used mainly by linguists). However, our more fine-
grained categories have made it possible to reveal that, for example, the linguists
use much more of the R2 type than economists.
While there are notable differences between disciplines as regards the use of
the four subtypes R1–R4, the language differences are smaller and will not be fur-
ther commented on here.
There are several particular features which could be further studied in rela-
tion to the analysed reference types. However, our presentation here is limited to
examples where a bibliographical reference assumes the grammatical function of
subject, as in example (6) above. Our purpose is to study which roles are assigned
to the others in research articles.

6.2.3 Bibliographical references as grammatical subjects

We will here look at the cases where a bibliographical reference assumes the gram-
matical function of subject. The examples studied correspond, on one hand, to
the semi-integral (R3) and the fully integral (R4) types, and, on the other, to the
non-integral reference type indicated by (superscript) number (R1). The reason
for including the R1 type, typical of the medical articles, is that in these articles, the
studies referred to may be given the prominent function of grammatical subject
232 Academic Voices

even if they are not referred to by exact name and year of publication. The typical
noun used in these contexts is study (or corresponding nouns in French and Nor-
wegian); other nouns used are report and survey. In some cases, the names of the
persons responsible for the study referred to are indicated, as in example (10):9
(10) In another study, Helzlsouer et al. [32] did not find lower measured levels of
vitamin A, […]. (engmed10)

So what are the other researchers allowed to do? What roles do the authors as-
sign to the other researchers when relating to them? A very general answer to this
question, based on the analysis of our material, is as follows: The others are present
as researchers. They find facts and solutions and show observations and proofs
related to their scientific issues, as in the following examples taken from the three
languages under study:
(11) Kormendi and Meguire (1985) and Aizenman and Marion (1993) find that
output growth is adversely affected by the volatility of monetary policy, […].
(engecon03)
(12) Pinchon (1972: 122-128) a montré qu’il ne s’agit pas d’une impossibilité
stricte, [...]. (frling06)
‘Pinchon (1972: 122-128) has showed that it is not a question of strict impos-
sibility, ...’
(13) Flere studier viser at MRCP er et godt alternativ til diagnostisk ERCP (1–9).
(nomed14)
‘Several studies show that MRCP is a good alternative to diagnostic ERCP
(1–9).’

This general conclusion is based on studies of the verbs combined with the ref-
erence assuming the grammatical function of subject and typically the seman-
tic-pragmatic function of agent. Table 6.2-1 presents the five most frequent verbs
– of 10 or more occurrences – in combination with bibliographical references in
subject position.

9. Since the year of publication is not indicated and the number referring to the bibliographi-
cal end-list is given, this type is counted as being of the R1 category in the quantitative classifi-
cation.
 Chapter 6. Presence of the others 233

Table 6.2-1. Verbs used with bibliographical references as subjects

Corpus RefS V Most frequent verbs

engecon 364 93 find 60 show 40 provide 18 report 16 argue 15
engling 713 167 argue 51 note 33 point to/ show 31 propose 26
out 18
engmed 233 51 report 47 find 36 show 28 demon- suggest 14
strate 23
frecon 255 97 montrer 35 soulig- proposer suggérer
démontrer ner 18 15 10
11 (‘empha- (‘propose’) (‘suggest’)
(‘show’) sise’)
frling 152 84 montrer 18 proposer
démontrer 16
2
frmed 153 69 montrer 29
démontrer
2
noecon 496 118 finne 63 vise 43 påpeke/ benytte få 13
(‘find’) påvise 1 peke på 17 16 (‘get’)
(‘show’) (‘point to’) (‘use’)
noling 573 150 si 44 hevde 35 skrive 29 påpeke/ mene 23
(‘say’) (‘claim’) (‘write’) peke på 28 (‘think’)
(‘point to’)
nomed 80 40 vise 22 finne 20
påvise 4 (‘find’)
(‘show’)
Total 3,019 869
RefS = number of references in subject position; V = number of different verbs.

The first notable finding is the high number of different verbs used in the con-
structions studied: 869 for the three languages. Most verbs are used only once or
twice. A general conclusion is thus that the lexical variation in this type of con-
struction is considerable. This corresponds to Hyland’s (2000) study. He also finds
great variation in the choice of verb used in referencing (over 400 were found in
his corpus of 80 articles written in English) and notes “substantial differences be-
tween disciplines” (ibid.: 27).
Another interesting point in our corpus is the modest presence of typical re-
porting verbs, except in English medicine, which employs report, and Norwegian
linguistics, which employs both si/seie (‘say’) and skrive (‘write’) quite frequently.
The table shows interesting similarities between both disciplines and lan-
guages: In all the nine subcorpora (except for Norwegian linguistics), the oth-
ers are presented as researchers who show or find something. But there are also
234 Academic Voices

interesting distribution and frequency differences concerning the use of different

verb types (see 4.2). Reporting verbs (like report, say, write) are, as stated above,
used in English medicine and Norwegian linguistics; research and discourse verbs
(like find, show) are more or less used by all but typically in economics and medi-
cine; position verbs (like argue, claim) are typically used in linguistics, especially
in English linguistics, but also in English economics. From a different perspective,
these verbs can be considered as factive (show, find) and non-factive (say, argue).
(A third group would be contra-factive verbs (pretend). The last group is for obvi-
ous reasons more or less irrelevant to our corpus.)10
The verb variation is also interesting from a polyphonic perspective. One of the
challenges for all discourse analyses of citations is the interpretation of the relation
between the speaker and the citation (or the represented/reported speaker), i.e. be-
tween the different voices integrated in the text. The ScaPoLine theory (see Nølke et
al. 2004: 49–50 and section 2.1.2) proposes only one general relation, that of repre-
sentation, for the cases of reported speech. This relation is of the non-responsibility
type, which implies that the locutor, or speaker, is not taking responsibility for the
represented speech, regardless of whether it is signaled by quotation marks and
regardless of the verb. However, it seems possible to imagine different subtypes of
relations according to the verb that is used in the inquit (i.e. the introduction to the
citation, as in Henry says …). We argue that there is a clear non-responsibility rela-
tion between author (self) and represented researcher (other) in the case of report-
ing (non-factive) verbs. However, in the case of factive verbs, this non-responsibil-
ity seems weakened. The author indicates that the proposition reported is one he or
she is clearly taking into account. The relation between self-voice and other-voice
is still one of non-responsibility, but at the same time there is an indication that the
author is adopting the other’s observation. The verbs provide important informa-
tion as regards how the author uses others’ research in his or her own presentation
and positioning, as well as the author’s attitude to the content of what is reported.
A further interesting elaboration related to this can be found in Perrin’s stud-
ies of reported speech.11 He proposes a distinction between referential and modal
reported speech. The referential reported speech is considered as a simple object of

10. Lyons (1977: 599) describes factive verbs as follows: “Anyone who says (21) John realizes
that it is raining (in order to make a statement) is committed by his use of the verb ‘realize’ to the
truth of the proposition expressed by the complement-clause: he presupposes that it is raining.”
Contra-factive verbs, on the other hand, commit the speaker to the falsity of the expressed pro-
position. Finally, a non-factive verb “commits the speaker to neither the truth nor the falsity of
the expression expressed by its complement clause” (ibid. 795).
11. Inspired by Clark & Gerrig (1990), but also by Bakhtinian ideas, Perrin has studied reported
speech and polyphony in a number of contexts; (cf. e.g. Perrin 2003, 2004a, 2004b, 2005, forthco-
ming). For a relevant study of “argumentation par autorité”, see also Norén 2004b.
 Chapter 6. Presence of the others 235

reference and is not activated in the author’s own discourse; the author just men-
tions that something has been said (more or less as an historical event):
[…] l’élément rapporté n’est […] qu’un simple objet de référence et n’est donc nul-
lement activé dans le discours effectif du locuteur, pris en charge par le locuteur ef-
fectif si l’on préfère, qui se contente alors de communiquer quelque chose à propos
(plutôt qu’au moyen) de ce qui est exprimé dans la reprise […]. (Perrin 2004b: 64)
‘the reported element is only a simple object of reference and not at all activated
in the effective discourse of the locutor, taken charge of, if you like, by the effective
locutor, who is then satisfied by communicating something concerning (rather
than by means of) that which is expressed in the report’

The modal reported speech, on the other hand, is taken into account by the author
and activated in his own discourse:
Dans le cadre d’un discours rapporté modal, […] l’élément rapporté est simultané-
ment activé dans le discours effectif du locuteur et ne peut donc être assimilé à
un simple objet de référence de ce qui est communiqué. L’élément en question est
bel et bien rapporté dans la mesure où il relève d’un discours objet auquel le locu-
teur réfère, mais d’un autre côté cet élément est simultanément pris en charge par
le locuteur, au même titre que n’importe quel élément constitutif de son propre
discours. (Perrin 2004b: 65)
‘In the context of modal reported speech, […] the reported element is simultane-
ously activated in the effective discourse of the locutor and thus cannot be as-
similated to a simple object of reference of what is communicated. The element
in question is clearly reported to the extent that it depends on a discourse object
to which the locutor refers, but on the other hand, this element is simultaneously
taken charge of by the locutor, in the same way as whatever element constitutive
of his own discourse.’

It seems reasonable to propose that reported speech with a factive verb should
be a good candidate for the modal type. In fact, only the context can provide an
indication of whether the reported speech is activated in the author’s discourse in
the sense that it is developed in his or her argumentation. Let us look at a French
example where the reported speech is introduced by the verb montrer (‘show’):
(14) Hudson et al. [25] montrent que 6 % des boulimiques (n = 49) sont égale-
ment dépendantes d’amphétamines, 13 % pour les anorexiques (n = 16), et 28
% pour les anorexiques-boulimiques (n = 25). (frmed13)
‘Hudson et al. [25] show that …’
236 Academic Voices

If one considers the sequence directly following this utterance, one might think
that it is an example of the referential type – a simple object of reference. In fact,
the author introduces another reference directly after the one in (14):
(15) Plus récemment, ont été effectuées des études contre témoins. Stern et al. [26]
retrouvent ainsi 15 % de ”substance abuse disorder” (abus de toxiques) chez
34 femmes souffrant d’anorexie ou d’anorexie-boulimie, contre 3 % chez les
témoins. (frmed13)
‘Stern et al. [26] find …’

However, if one takes a closer look at the article, Hudson et al.’s arguments are ac-
tually reported several times, one after the other; their results are even presented
in a separate table. There is no doubt that Hudson et al. constitute an argument of
authority and that the introduction by the verb montrer indicates modal reported
speech. The modal reported speech, in our view, is still not the author’s responsi-
bility; nevertheless it is, to a certain extent, blended with his or her discourse.12

6.2.4 Discoursal and rhetorical functions of bibliographical references

If one considers scientific writing as a tension between progress and continuity

(Fløttum 2003e) – progress in the sense of being original and presenting new find-
ings, and continuity in the sense of situating oneself as a researcher in a scien-
tific tradition – the use of bibliographical references assumes an important role.
However, in general, reference practices may seem to be more important for the
continuity dimension than for progress. In the KIAP Corpus, the most frequently
used verbs combined with references in subject position, i.e. factive verbs (see Ta-
ble 6.2-2), indicate that other researchers are cited first and foremost to present
what has already been found or shown in the research territory in question.13 As
mentioned in 6.2.3, English and Norwegian linguists constitute, to some extent,
an exception in that they often use non-factive verbs (like argue and hevde ‘claim’)
which emphasise argumentation. Norwegian linguists also use non-factive verbs

12. A construction which may indicate even more clearly modal reported speech is the com-
parison introduced by comme (‘as’), as in this example: “[…] comme le montrent Zaleski et Zech
(1996)” (frecon02). In this context, constructions like according to X or for X (in French: selon X,
pour X; in Norwegian: ifølge X, for X) are also interesting (see Coltier & Dendale 2004; Fløttum
2004b; Perrin 2005).
13. Some of the references we have studied may of course contain modal elements and hedges
which modify the meaning of the verbs, but in the majority of the around 3000 examples with
a reference in subject position, the function of this reference can clearly be interpreted on the
basis of the semantic-pragmatic meaning of the main verb.
 Chapter 6. Presence of the others 237

like si/seie (‘say’) and skrive (‘write’), which may play an argumentative part in the
play between self and other, but which do not indicate argumentation directly.
When researchers integrate bibliographical references clearly in their argu-
mentation, these references may of course also assume an important role within
the progress dimension. They may claim progress by refuting what has been said,
found or done in previous research; however, they also present directly what they
have done without comparing with or referring to others.
It is clear that our rhetorical conception of academic discourse as both inviting
cooperation and claiming position may also help to explain the citation practices
demonstrated in research articles.14 To some extent these two aspects of rhetoricity
parallel the dimensions of continuity and progress.
A third distinction which may contribute to the understanding of referencing
practice is the distinction proposed by Bondi and Silver (2004) between discipli-
nary participants and discourse actors in their study of textual voices in research
articles taken from economics and history. When the reported voices are inter-
preted as disciplinary participants, they seem to be part of a positioning process
or ongoing debate within the relevant disciplinary community; on the other hand,
when the reported voices are considered as discourse actors, they are only staging
voices of disciplinary agents situated in a context where the author emphasises the
object of study or a particular event, rather than an ongoing debate.
If it is relatively easy to point to general disciplinary tendencies as regards the
type of bibliographical references used (see 6.2.2), it is on the other hand rather
difficult to point to general tendencies as regards the discoursal and rhetorical func-
tions the references may assume.15 However, the general dimensions of continuity/
progress and cooperation/position constitute relevant guidelines for the interpreta-
tion of the function of references that in most cases seem to manifest some kind of
“collective intelligence” (see Valle 1999), while each article may demonstrate indi-
vidual patterns of reference use according to subject matter and the author’s posi-
tion in the discourse community. When studying individual articles, we believe
that the introduction of factivity combined with the modal/referential distinction

14. Tuomarla (1999), contrasting the citation practices in news articles characterised by dra-
matisation with academic practices, points to positioning (or “actualisation” in her French ter-
minology) as an important function of bibliographical references in research articles.
15. Bibliographical references used to indicate continuity, for example, can have different
subfunctions, like giving historical background, giving credit, assigning priority or gaining sup-
port (see Valle 1999: 401). But it is difficult to distinguish them clearly; they often overlap. We
find it more useful to establish a few, relatively general, functions or dimensions which can ope-
rate as guidelines for the interpretation of the individual references (and clusters of references).
For other relevant studies, see Rosier (1999) and Boch & Grossmann (2002).
238 Academic Voices

contributes to a better understanding of similarities and differences in knowledge

construction and argumentation in different disciplines and languages.

6.3 Who can do what? Self, other and verb use

In a polyphonic perspective, the locutor, corresponding to the author, constitutes

the dominating voice in every text, regardless of whether he or she is “visible” or
not. However, in addition to the locutor’s voice, others are allowed to take the floor.
But, as we have seen in section 6.2, the author lets other voices manifest them-
selves in many different ways, assigning different roles to them (often signalled by
the verb combined with the reference). In order to get a better picture of who does
what and to find out to what extent there are similarities and differences between
the representations of self and other, we will here compare the verbs combined
with self and the verbs combined with other (see Table 6.3). For the self-dimen-
sion, we present the three most frequent main verbs used in combination with ‘I’
in economics and linguistics and with ‘we’ in medicine.16 For the other-dimension,
we present the three most frequent main verbs used in combination with refer-
ences, represented by third person plural pronouns in the list.17
Throughout the book we have observed heterogeneity, in many respects, as a
characteristic feature of the research articles in our corpus. However, as already
stated in section 6.2, we note some homogeneity in the other-dimension, as can
be seen here through the use of verbs combined with third person nominal phras-
es referring to cited authors. The others, in all disciplines apart from Norwegian
linguistics, are allowed to show or find something. However, the discipline factor
seems stronger than the language factor in that the most homogeneous patterns
are found within economics and medicine. Linguistics is more heterogeneous; in
this discipline we observe that cited authors are allowed to argue more explicitly
(through verbs such as argue, proposer and hevde). Norwegian linguistics is spe-
cial; cited authors in this discipline are literally reported through verbs such as
si/seie and skrive.

16. The reason for this, again, is that medical articles are generally multi-author articles.
17. In some cases, four verbs are listed. The reason is that number three and four of the verbs
are (nearly) equally frequent. As regards the frecon subcorpus, we have only listed one verb. The
reason is that there is only one French economist author who uses the first person pronoun je in
our corpus. The presence of the verb considérer in this context should therefore be interpreted
only as a feature of an individual article.
 Chapter 6. Presence of the others 239

Table 6.3-1. The three most frequent verbs used in a selection of self- and other-contexts

Subcorpus SELF OTHER

engecon I consider they find
use show
assume provide
frecon je considérer (‘consider’) ils/ montrer (‘show’)
elles souligner (‘emphasise’)
proposer (‘propose’)
noecon jeg/ anta (‘assume’) de/ finne (‘find’)
eg diskutere (‘discuss’) dei vise (‘show’)
betegne (‘call’) påpeke (‘point to’)
engling I argue they argue
show note
suggest point out/to, show
frling je proposer (‘propose’) ils/ montrer (‘show’)
revenir à (‘return to’) elles proposer (‘propose’)
considérer (‘consider’) signaler (‘point to’)
noling jeg/ mene (‘think’) de/ si (‘say’)
eg kalle (‘call’) dei hevde (‘claim’)
se (på) (‘see’/’look at’) skrive (‘write’)
engmed we find they report
use find
examine show
frmed nous observer (‘observe’) ils/ montrer (‘show’)
constater (‘see’) elles trouver (‘find’)
analyser (‘analyse’) proposer (‘propose’)
rapporter (‘report’)
nomed vi/ finne (‘find’) de/ vise (‘show’)
me ha (’have’) dei finne (‘find’)
velge (‘choose’) rapportere (‘report’)

The discipline homogeneity observed within medicine, in several cases, extends in

this context to the self-dimension. When entering the scene, the medical authors
allow themselves to be represented as researchers who find something (in French
a similar semantic content is represented by the verb observer).
However, the general picture of the self-dimension is much less homogeneous.
The author roles manifested vary to a considerable extent, as shown in Chapter 4.
Apart from arguing linguists, we see all kinds of manifestations. The authors, when
manifested through first person pronouns, allow themselves to consider, to discuss,
to argue, to call, to return to, etc. Nevertheless, the cognitive and perceptive self
seems quite important through verbs like ‘consider’, ‘assume’, ‘think’ and ‘see’. We
also note that the result orientation marked by verbs such as ‘find’ and ‘show’ in
240 Academic Voices

the other-dimension is more or less absent in the self-dimension of economics and

linguistics. It should also be noted that both in linguistics and economics, the verbs
‘find’ and ‘show’ are frequently combined with ‘we’ (not shown in the table).
The clear presence of verbs taking as their complement expressions referring
to some result or finding in medicine may be related to the nature of the research
undertaken in this discipline. The research presented in medical articles is often
of a kind which provides observable and quantifiable findings; furthermore, the
medical research is clearly presented as completed in the articles. (For interesting
and very relevant studies of tense choice in verbs used in referencing, see Shaw
1992 and Hawes & Thomas 1997.) In linguistics, on the other hand, the research
is often non-experimental, having objects of study which are ontologically differ-
ent from the objects of study in medicine, providing other kinds of results (see
section 7.1). In many of the linguistics articles in the KIAP Corpus, the research
is presented as undertaken in the article itself. This may explain the greater pres-
ence of cognitive (or research) verbs and argumentative (or position) verbs in the
linguistics articles.
As regards economics, this discipline is similar to linguistics in the self-per-
spective but to medicine in the other-perspective (except for the low frequency of
bibliographical references). This observation corresponds to other observations
made in previous chapters: Economics is situated in a non-stable middle position,
sometimes biased towards linguistics and sometimes towards medicine.
To give a better picture of the French self-perspective, we have also studied the
verbs combined with the indefinite pronoun on. The quantitative data presented
in 4.1 show that on is by far the most frequent of the French pronouns which are
studied here. As shown in section 4.4, this pronoun, due to its referential poten-
tial, demonstrates personal as well as indefinite uses. Further, we saw that the use
of on in the KIAP Corpus is more self- than other-oriented in the sense that the
most frequent values of on refer to the author(s), sometimes alone but more often
included in a more or less indefinite reference. Table 6.3-2 presents the ten most
frequent main verbs combined with on in each subcorpus.18

18. We emphasise that many of these verbs are modalised by the auxiliary pouvoir (‘may/can’).
The sequence on peut (‘one may’) + infinitive form of the main verb is typical. If the pronoun on
is an indirect means of self-manifestation, the pouvoir hedging adds considerably to this indirect-
ness. It appears with about 1/5 of the on-occurrences in the economics subcorpus, with about
1/3 in the linguistics subcorpus and with about 1/6 in the medical subcorpus.
 Chapter 6. Presence of the others 241

Table 6.3-2. The ten most frequent verbs combined with the French pronoun on
frecon frling frmed
1,213 on 1,564 on 236 on
253 361 111
verb types verb types verb types
noter 69 avoir 88 observer 29
(‘note’) (‘have’) (‘observe’)
supposer 68 voir 88 constater 18
(‘suppose’) (‘see’) (‘see’)
avoir 66 dire 67 retrouver 17
(‘have’) (‘say’) (‘find’)
obtenir 64 trouver 62 noter 11
(‘obtain’) (‘find’) (‘note’)
considérer 53 considérer 37 considérer 8
(‘consider’) (‘consider’) (‘consider’)
montrer 43 noter 37 estimer 6
(‘show’) (‘note’) (‘estimate’)
observer 30 retrouver 37 savoir 6
(‘observe’) (‘find’) (‘know’)
déduire 29 constater 36 remarquer 5
(‘deduct’) (‘see’) (‘remark’)
remarquer 29 parler 36 assister 3
(‘remark’) (‘speak’) (‘assist’)
voir 29 observer 33 avoir 3
(‘see’) (‘observe’) (‘have’)

Even if the lists are made without taking into account the different on-values pos-
ited in 4.4, they give an interesting picture of what the authors allow themselves
to do when “hiding” behind on, alone or together with a more or less limited mul-
tivoiced community. Once again the heterogeneity is striking. Four of the listed
verbs are common to all three disciplines, but are ranked differently: avoir, con-
sidérer, noter and observer. A general conclusion which may be drawn from this
table is that the pronoun on does not appear with clear argumentative or position
verbs. In other words, French researchers do not represent themselves through the
pronoun on when assuming the arguer role. Thus, we do not find for example on
défend (‘one defends’) with on referring to the author(s).
To conclude, Tables 6.3-1 and 6.3-2 give a complex but interesting picture of
what the authors typically allow themselves and their cited colleagues to do. We
have pointed to some general tendencies, and in addition, the verb lists provide
valuable insight into the complexity of research and research processes in the three
different disciplines.
242 Academic Voices

6.4 Implicit presence: negation and concessive constructions

6.4.1 Introduction

The source of an assertion can be more or less overtly or explicitly marked. In this
context, it is quite common, particularly in Anglo-American discourse analysis, to
refer to the distinction described by Sinclair (1987) in terms of averral and attribu-
tion. Bondi and Silver explain the distinction in the following way:
Averral is the default condition of a text, where the reader can assume that the
responsibility for each proposition lies with the speaker or writer. Attribution,
on the other hand, is the case where a proposition is indicated as deriving from a
source. (Bondi & Silver 2004)

However, within the ScaPoLine theory (see section 2.1.2), the distinctions are
somewhat different. Our polyphonic framework argues that there is a responsi-
ble locutor (or speaker) behind every utterance. This locutor may give the floor
to other voices, explicitly or implicitly. In the previous section, we discussed the
explicit type as manifested by bibliographical references, particularly in the form
of reported speech (often considered as the prototypical form of discoursal po-
lyphony). However, polyphonic interactions very often take place in more subtle
ways, without explicit sources (see also Fløttum 2001a). In the present section, we
look at two kinds of implicit polyphony, as manifested by negation with ‘not’ and
by the concessive construction marked by the adversative connective ‘but’.
As mentioned above, implicit and explicit polyphony are often interwoven. For
methodological reasons, we distinguish between them here. Polyphony is actually ex-
plicit in most cases, through some distinctive marker. But there are different degrees of
explicitness. In the following example, the connective however indicates polyphony:
(1) Phenomenon X is difficult to describe in linguistic terms; however, it is es-
sential to the explanation of text structure.

In example (2) also, however indicates polyphony, but in this example there is an-
other kind of polyphony as well, viz. reported speech:
(2) Smith (2000) claims that phenomenon X is difficult to describe in linguistic
terms; however, in my studies (Johnson 2001) it is essential to the explanation
of text structure.

The methodological distinction corresponds to a distinction between two types of

polyphony: the explicit one, for example reported speech with an explicit source,
and the implicit one, for example concessive constructions indicated by however
or but, without explicit sources of the represented voice.
 Chapter 6. Presence of the others 243

What is particularly interesting in this context is that negation and concession,

by virtue of their inherent semantic content, indicate what the function of the
voices integrated in these constructions is – regardless of the source responsible
for the content presented. For the interpretation process, however, it is crucial to
identify the sources. These are issues that we will develop in the following.

6.4.2 Quantitative comments

Our quantitative analyses give reason to consider polyphony as realised by adver-

sative ‘but’ and negation ‘not’ as a general or typical trait of the research article
genre. As shown in section 6.1, there are some differences as regards frequency in
different languages and disciplines; however, negation and concession are clearly
present in all articles. To compare with another frequent feature, first person plural
pronouns: the mean relative frequency for adversatives is 0.27 %,19 for negation
0.56 % and for first person plural pronouns in multi-author articles 0.46 %.
There are, of course, variations related to the different variables we study. As
we saw in section 6.1, there is more frequent use of negation and of adversatives
in linguistics than in economics and medicine, and the differences between lin-
guistics and economics and between linguistics and medicine are significant at the
superlevel, as well as in each language, whereas the difference between economics
and medicine is not. Furthermore we have observed that Norwegian linguists use
both most frequently.

6.4.3 The case of negation

The ScaPoLine perspective allows us to propose various uses of the negation con-
struction. We find different nuances of meaning by taking as our point of depar-
ture the polyphonic structure of this construction (see section 2.1.2), for conven-
ience repeated here:
(3) This wall is not white.
pov1: [X] TRUE (‘this wall is white’)
pov2: [l0] UNJUSTIFIED (pov1)
Two points of view (pov) are indicated by the negation not. The polyphonic struc-
ture also consists of an instruction saying that the locutor (l0) is responsible for

19. It is of course not satisfactory to measure concession by the adversative markers but, mais,
men. First, these may have other values; second, there are other connectives which mark conces-
sive constructions. Cases in point are English however, French cependant and Norwegian imid-
lertid. For further explanation, see 6.1.
244 Academic Voices

pov2, but not for pov1, which has an unknown source (X). Pov1 is considered as
true by X, but as unjustified by l0.
Without going into the discussion of the old distinction between descriptive
and polemic negation,20 let us just mention that the challenge here is to see to
what extent there are two points of view, and if there are two, to try to identify the
individual (or group of individuals) responsible for the negated pov in order to
determine the function that the negation assumes in discourse. To a certain extent,
we follow Nølke (1993b: 241–242; 1993a), who characterises the primary value of
the negation as polemic. The descriptive value is only the result of a derivation. In
our studies, however, we question the traditional distinction between polemic and
descriptive negation and postulate the existence of a scale of uses manifesting dif-
ferent degrees of polemicity, depending on who is responsible for the negated pov
(self or other, internal or external voice).
There are certain linguistic contexts which are clearly more interesting than
others in this context: Different kinds of contrastive contexts, which are important
to the interpretation of the examples in the present study, often signal polemicity.
In order to determine the nature of polyphony, it is necessary to take into consid-
eration a larger context than the isolated utterance. We have to consider the situ-
ational context, in which the text genre itself plays an important role.
In the following, we propose four different uses of negation. Since our qualita-
tive studies in this domain have mainly been carried out on the French subcor-
pora,21 we refer to French examples containing the negation ne … pas (but the se-
mantic-pragmatic analysis is also valid for English not and Norwegian ikke/ikkje).
We use the term self when referring to the article or study from which the example
is taken, or to its author (self-author); we use the term other when referring to
what the self-author reports from (another author, study or article). We will dis-
cuss the four types of negation use in turn: refutative, delimitative, corrective and
descriptive.

Refutative use of negation

This use typically has a high degree of polemicity. It often concerns observations,
findings and choices attributed to others and refuted by the self-author. The pres-
ence of a positive pov1 seems evident, and justified by different contexts of contrast.
This use often manifests itself in an interactional situation where the self-author
wants to position him- or herself within the relevant scientific community and is
trying to occupy a niche (Swales 1990) that others have not (yet) occupied. There are

20. For this discussion see for example Nølke (1993a, b); Fløttum (2004f, 2005c); and additio-
nal references in these publications.
21. For a study of the Norwegian negation ikke/ikkje, see Fløttum (2004f).
 Chapter 6. Presence of the others 245

several subtypes within this use, according to the more or less explicit nature of the
implicated discoursal beings and according to the degree of accessibility of the pov1
source. Here is an example (emphasis added), taken from a multi-author article:
(4) Une idée jugée prometteuse par BLUNDELL et LEWBEL [1991] est celle
préconisée par l’école de Leyden (VAN PRAAG [1979]). Celle-ci consiste à
utiliser des variables d’opinion pour approcher les niveaux d’utilité atteints.
Cette approche ne résout cependant pas tous les problèmes d’identification
sur lesquels on reviendra. (frecon01).
‘An idea considered as promising by BLUNDELL and LEWBEL [1991] is the
one put forward by the Leyden school (VAN PRAAG [1979]). It consists in …
However, this approach does not solve the identification problems which we
will come back to.’

In this use, the negation is clearly related to self-research, as in (4). It is situated

enunciatively, often by a metatextual expression (like revenir sur, ‘come back to’)
or by a personal pronoun referring to the author(s) in one way or another (like the
pronoun on (‘one’), in (4)). Emphasising that their study will show the inadmis-
sibility of the reported pov, the authors, corresponding to the responsible locutor,
refute the idea put forward by the other. All in all, the construction contributes to
the positioning of the self-authors.
The source of the refuted pov can be explicit (as in (4) through l’école de Ley-
den) or not, as in the following example:
(5) Cet article montre que la prise en compte des inégalités n’est pas sans con-
séquence sur l’évolution de la pollution. (frecon10)
‘This article shows that taking the inequalities into consideration is not with-
out consequence for the evolution of pollution.’

This negation is also refutative, situated in relation to the self-research (cet article
‘this article’), but the source of the refuted pov is not explicit. In fact, in order
to determine the polyphony, it is not important to know if pov1 really has been
expressed or thought, or not. The interesting aspect is that the locutor attributes
this pov to a source. However, for the interpretation of the whole text, it can be
essential to identify the real person or community corresponding to the discoursal
being responsible for pov1.
246 Academic Voices

Delimitative use of negation

The delimitative use of negation, manifesting a lower degree of polemicity, is typi-
cally related to the self-author’s choices and objectives:
(6) Dans le présent modèle, nous ne proposons pas une analyse de la formation
endogène du chômage ; nous nous intéressons à l’interaction entre la localisa-
tion endogène des chômeurs, des employés et des entreprises. (frecon36)
‘In the present model, we do not propose an analysis of the endogeneous for-
mation of unemployment; we take an interest in the interaction between …’

At first sight, the presence of an underlying positive pov may not be obvious. How-
ever, by contextual inference, a pov1 appears. One can easily imagine that the self-
author, faced with his study, could reason as follows: “Given the fact that I am
studying A, a phenomenon consisting of the subphenomena a1, a2, a3 … an, the
community to which I belong might think that my objective is to study a1, a2, a3
… an. I have to specify my objective: I will not study a1.” By anticipating potential
questions from his or her colleagues, the author utilises a negation that is more
informative than polemic. The negation assumes a delimitative function.
The contrastive context here is the article itself, seen in relation to what may
exist – now or in the future – in other relevant studies or articles.
In this interaction, what is negated is normally a constructed pov, which it may
be difficult to access and attribute to a collective third, corresponding to the com-
munity that the author is addressing (including the reader).
Even if it is not very strong, the polemicity is present and is often supported by
a contrastive perspective in the text. In (6) above, the metatextual expression dans
le présent modèle implicitly indicates the contrast with other models.

Corrective use of negation

This subtype usually comprises negation concerning the utilised data in the self-
study or findings that the author’s analyses provide. Thus this negation type bears
on the analysis or interpretation of the object of study itself. The self-author cor-
rects a pov considered as wrong:
(7) Comme le montre la discussion de ces deux exemples, les éléments de la FS
n’ont pas nécessairement de réalisation lexicale. (frling19)
‘As shown by the discussion of these two examples, the FS elements do not
necessarily have a lexical realisation.’

While contributing to the positioning of the author, this negation use assumes
a corrective function in the sense that it rejects conceptions (real or potential) of
precise data considered as wrong or unjustified.
 Chapter 6. Presence of the others 247

As regards the polyphonic dimension, the presence of pov1 might not be im-
mediate. In the subtype illustrated above, it is the situational context which guides
us towards a polemic and polyphonic interpretation. The polyphonic reasoning
might go as follows: “Given the already acquired knowledge, the academic com-
munity that I address might think that my data indicate X. If research is to progress,
I have to correct this supposition: My data do not indicate X.”
The source of pov1 can be of various types. It might be a more or less undefined
collective third party, or an individual third party corresponding to a particular
researcher (who the self-author thinks of without necessarily referring to him or
her), or even the textual locutor, L, i.e. an image of LOC (constructor) at a previous
stage or in another study. In the latter case there is a “mise en scène” of different
images of the locutor.
Descriptive use of negation?
The last negation example we will present represents a subtype that might not be
polyphonic. At least this is the use of negation with the lowest degree of polemicity
(if any at all) among the examples treated here. It is close to the negation tradi-
tionally referred to as descriptive; in polyphonic terms a construction without an
underlying positive pov. Here is an example:
(8) Danielle Leeman m’a fait observer que l’emploi actif de planter s’étend à des
objets qui ne sont pas des plantes, comme des clous, […]. (frling30)
‘… the active use of planter extends to objects which are not plants …’

If this negation is interpreted as descriptive, it indicates that there are no other

(obvious) linguistic means by which the author could express what she wants to
express. We leave the discussion here – even if it is tempting to try to apply the
polyphonic perspective to such examples as well.
Before going on to concessive constructions, we sum up by emphasising that
the analysis undertaken above has as its point of departure the linguistic poly-
phonic structure of negation (see (3)). This basis has made us ask what we see as
very relevant questions in the interpretation process, especially so in the search for
contextual explanatory factors of the various negation uses proposed.

6.4.4 The case of concession

Let us now look at an example of the polyphonic marker but in its adversative and
concessive capacity. In simple terms, a construction with but can be formalised in
the following way:
(9) p but q
248 Academic Voices

where p and q constitute two propositions in contrast and where p represents the
concession and q an argument that the locutor identifies with, here and now. The
locutor does not refute p; the q argument is however the one that counts in the
specific context where the utterance is produced. In the polyphonic analysis these
arguments are treated as points of view (pov). Let us look at some authentic exam-
ples of French, Norwegian and English concessive constructions with mais, men
and but respectively:
(10) En situation imparfaite, le régulateur ne connaît pas les caractéristiques de
l’exploitation, mais il peut tenir compte de la connaissance statistique des ren-
dements des terres pour proposer un contrat d’exploitation. (frecon27)
‘In an imperfect situation, the regulator does not know the characteristics of the
exploitation, but it may take into consideration the statistical knowledge …’
(11) Her er ikkje data like uproblematiske. Rett nok skjer det genusskifte i norsk,
som Beito (1954) har dokumentert, men det er likevel svært små endringar i
den relative fordelinga av dei tre generaa frå gammalnorsk til nynorsk anno
1900-1950. (noling47)
‘… admittedly gender changes do occur in Norwegian, as documented by
Beito (1954), but there are still very small changes in the distribution of the
three genders …’
(12) The most general characterization of actually is as a marker of contrast and
revision. But such a broad characterization cannot capture the subtleties of its
use displayed so far. (engling20)

For practical reasons we use the English example (12) in our illustration. The anal-
ysis of the concessive but-construction, containing four points of view, goes as
follows in a somewhat simplified form:22
(13) pov1: [X] (TRUE (p))
pov2: [norm] (generally (if p then r))
pov3: [locutor] (TRUE (q))
pov4: [norm] (generally (if q then non-r))

22. The original analysis as developed in Nølke et al. (2004), for French, is presented as follows
(where TOP signifies topos; i.e. a notion inspired by the theory of argumentation in langue (see
Ducrot 1984) signifying a principle that ensures the passage from an argument to a conclusion):
pov1: [X] (VRAI (p))
pov2 : [ON] (TOP (si p alors r))
pov3 : [l0] (VRAI (q))
pov4 : [ON] (TOP (si q alors non-r))
 Chapter 6. Presence of the others 249

This can be read and interpreted as follows: Pov1 says that an unknown non-ex-
plicit person or group of persons (X) thinks that p is true. Related to example (12),
p corresponds to ‘the most general characterization of actually is as a marker of
contrast and revision’.
The connective but gives the instruction that the locutor might agree with this
– the claim is not rejected; however, what counts to the locutor in this context is q,
corresponding to ‘such a broad characterization cannot capture the subtleties of its
use displayed so far’ – a relation expressed by pov3.
Further, according to pov2, some norm or more or less indefinite scientific
community (here: community of linguists) implies that the argument ‘if p then r’ is
generally true. The symbol r represents a conclusion to be found in the interpreta-
tion process. In this context, we may imagine the conclusion to be something like
“go for this characterisation of actually”.
However, another community or norm is implicated in this discussion, the
one expressed in pov4. This one opposes the one in pov2 and believes that if q then
non-r is generally true. We might interpret the conclusion non-r as “do not go for
the first characterisation of actually”.
The linguistic polyphony structure cannot answer all the questions that are
relevant for the interpretation of a concessive structure in discourse. However, it
indicates relevant questions to ask. For example, who is the voice hiding behind
X in pov1? Who is responsible for this pov? The utterance alone cannot tell us.
Given the context of scientific discourse, we may of course imagine that another
researcher or researchers are responsible for this pov1 and some scientific commu-
nity, including X, for pov2. However, X could even correspond to the self ’s pov at
another, maybe earlier, stage in his work. It may be important to identify X for the
interpretation, i.e. to decide whether it is an external or an internal polyphony. If
it is internal, X may correspond to an image of the locutor at another point in the
research process. In this case, the internal polyphony indicates that the author is
conscious of potential objections, that he or she might have raised him- or herself,
leading to a given conclusion. If the polyphony is external, the author offers a con-
cession to an external person or group of persons. This concession indicates that
the author is ready to admit the existence of claims oriented towards a conclusion
that is different from the one indicated in the utterance in question.
In both cases, whether the polyphony is external or internal, it is a matter
of manifesting the intention to promote cooperation between different points of
view. However, according to the nature of the concession, this intention plays dif-
ferent roles as regards the author’s position. In the case of external concession, the
invitation contributes at the same time to a clear positioning vis-à-vis the external
community in question.
250 Academic Voices

Determining whether the concession is internal or external is also important

for the determination of the norm or community responsible for pov2. If the con-
cession is internal, we may ask whether this is a norm created by the author or a
more general norm. The same question is relevant to pov4. However, in that case,
we already know that the argumentation presented is one that the author is sup-
porting. The q is in fact considered as true by the locutor in pov3, and this very q is
part of a norm represented in pov4. But to what extent does this norm or topos cor-
respond to a more or less general, and real, opinion, in a more or less defined com-
munity? In other words, does this norm really exist in the community to which
the author belongs or would like to belong? Or is the norm created by the author
for this specific occasion? If so, in that context, the author imposes on the reader
something which is not necessarily generally accepted, i.e. the conclusion non-r.
Regardless of the level of detail one aims at in the analysis, the questions raised
would contribute to a better understanding of the complex interaction of voices in
research articles.

6.4.5 Summing up

Our findings show that the presence of the other(s) is an important aspect of aca-
demic discourse as realised in research articles. The use of bibliographical refer-
ences is the most obvious manifestation of this presence. However, when it comes
to the frequency of other-presence, polyphonic constructions are far more impor-
tant. In this study, we have mainly studied negation and concessive constructions
as manifested by adversative connectives, but there are many other polyphonic
constructions which could have been taken into account in this book, for example
other connectives (like the French donc ‘thus’, see Didriksen 2004) or epistemic
expressions like it seems, may/might, perhaps, probably (see Vold 2005).
We have tried to demonstrate how the polyphonic perspective may be an in-
teresting one. In order to understand and interpret academic discourse, it is im-
portant to examine not only explicit but also implicit manifestations of both self
and other, especially in order to modify the traditional view of scientific discourse
as objective and fact-oriented. Both self and other are present through the instruc-
tions given by different markers, like not and but, contributing to the argumenta-
tion implicitly incorporated in the text. Here is a final example where explicit and
implicit polyphony are interwoven:
(14) After this quick survey, then, it looks like NP-deletion is possible after every
determiner except no, every, a, and the (cf. Lobeck 1995:42-45). But it has
been argued for some time that, under certain conditions at least, one and a
 Chapter 6. Presence of the others 251

are phonological variants of the same lexical item (Perlmutter 1970; Stockwell
et al. 1973). (engling46)
It seems useful to have a frame like the linguistic polyphonic one when “attacking”
complex examples of self- and other-presence of this kind. Linguistic polyphony
is a subtle way of bringing both self and other into a text, which might otherwise
be considered as “objective” and deprived of traces left by the writing subject or
by other voices. It is clear that multivoiced or polyphonic presence of this kind is
manifested in different ways and to different degrees in different languages and
disciplines. However, our findings allow us to argue that there are some polyphon-
ic phenomena which, independent of language or discipline, constitute a common
and characteristic feature of the research article. Among these are negation and
concessive constructions.

6.5 Case study: Negation in individual articles

In this case study we take a closer look at the use of negation. Following ScaPoLine
(see 2.2.1), we argue that there is a responsible locutor (or speaker) behind every
utterance. This locutor may give the floor to other voices, explicitly or implicitly.
The negation construction with not is one type of implicit interaction between the
locutor and other voices. The source of the latter is frequently not explicit. The
negation not indicates two points of view, where the locutor is responsible for the
negated one (see 6.4.3).
We now want to focus on the disciplinary variable, and have therefore selected
three articles written in the same language (English) but taken from our three
different disciplines. In section 6.1, we saw that negation is clearly used more in
linguistics than in economics and medicine and that there is no significant differ-
ence in the use of negation between economics and medicine. The question we
ask in this case study is the following: Do the three disciplines display differences
in the nature of the negation constructions used? We will try to answer this ques-
tion through an analysis of the articles engecon28, engling06 and engmed44 (all
multi-author articles, displaying frequencies which are fairly close to the discipline
median for negation). The analysis will focus on the distribution of the four types
of use proposed in section 6.4.3, which we repeat here:
1) Refutative use of negation, typically polemic, often concerning observations,
findings and choices attributed to others and refuted by the author. The presence
of a positive underlying pov1 may usually be justified by different contexts of con-
trast. This use often manifests itself in an interactional situation where the author
wants to position him- or herself within the relevant scientific community.
252 Academic Voices

2) Delimitative use of negation, manifesting a lower degree of polemicity, typically

related to the author’s own choices and objectives. By anticipating potential
questions from his or her colleagues, concerning the delimitation of the actual
study, the author utilises negation of a more informative than polemic nature.
3) Corrective use of negation, typically concerning data utilised by the author or
referred to (used by others) in the study, or findings that the author’s analyses
provide. The author corrects a pov, considered as wrong, related to the analysis
or interpretation of the object of study itself.
4) Descriptive use of negation, with a low, if any, degree of polemicity. When
interpreted as descriptive (and not polyphonic), this negation indicates that
there are no other (obvious) linguistic means by which the author could ex-
press what he or she wants to express.

Let us start with engecon28, containing 20 occurrences of not. This article dis-
plays a varied use of negation. All the four types are represented, but the corrective
use seems to dominate. Here is an example:
(1) The present paper contributes to this debate by presenting a case study of
Ireland’s break with sterling following its decision to join the exchange rate
mechanism (ERM) of the European Monetary System in 1979. If forming a
currency union has a large positive effect on trade, the break-up of a union
would be expected to have a significant adverse effect, but we do not find this
to be so. (engecon28)

In this example, the authors correct what is presented as a reasonable expectation,

probably based on already acquired knowledge (the source of this point of view is
not explicit), by referring to what their own data show. In addition to the frequent
corrective use, this article also includes instances of negation which may be inter-
preted as being refutative, as in the following example:
(2) While it was generally believed that trade diversification towards the con-
tinental countries was in Ireland’s interest, this was not a prime motive for
breaking the sterling link. (engecon28)

The polemic tone is present already in the clause introduced by the contrastive
connective while. The construction it was generally believed may be interpreted as
a form of inquit (see 6.2) followed by represented speech. This impersonal con-
struction does not indicate whether the locutor was part of the “general belief ”
represented. What is clear, however, is that the locutor refutes that the content of
this belief was a prime motive for breaking the sterling link. It seems reasonable to
think that the point of view refuted is attributed to a source that may also share
 Chapter 6. Presence of the others 253

the general belief represented in the subordinate clause, a voice the locutor clearly
is not part of.
Let us now turn to the linguistics article, engling06, containing 50 occurrences
of not. It is dominated by the refutative use of negation (about 20 of the 50 occur-
rences). This interpretation of not is often supported in this article by clear author
presence through introductions consisting of a first person pronoun combined
with a position verb, i.e. the authors are assuming the arguer role (see section 4.2),
as in this example:
(3) That is, although New Zealand English is now typologically very close to the
English of the southeast of England, we assert that it is not simply the result
of the transplantation of, say, London English to the other side of the world.
(engling06)

The source of the point of view constituting the outset of the polemics is not made
explicit in the same utterance as the negation is expressed. However, we are in-
formed about the source in the immediately succeeding cotext: This was errone-
ously argued for by Wall 1938; … Finally, a further mark of contrast, often accom-
panying the refutative use of negation, is found in the concessive although.
In addition to some negations which are not easily classified into one of the
four proposed types of use, the article contains a couple of delimitative uses (in the
section called Aims), some corrective and quite a few descriptive uses. Not unex-
pectedly, the delimitative use is related to the delimitation of the reported study:
(4) Our work on this topic is mainly, though not exclusively, concerned with pho-
netics and phonology, since most distinctive characteristic of New Zealand
English today is its accent. We have not, however, dealt with suprasegmental
phonology, […]. (engling06)

The descriptive uses may be explained by the fact that a major part of the text is
dedicated to the description of data, as in the following example:
(5) […] those of our speakers who are most obviously and consistently Scottish in
their phonetics and phonology do not have this feature […]. (engling06)
We will now turn to engmed44, the medical article, containing 15 occurrences of
not. In addition to some delimitative-like expressions like data not shown, placed
within parentheses and referring to data which are not shown in the figures in-
cluded in the article, the dominating use of negation is of a descriptive or descrip-
254 Academic Voices

tive-corrective type. We have classified the following example as descriptive, i.e.

with no apparent polyphonic aspect (at least to a non-expert):
(6) Addition of TNF or IL-1 had no effect on IL-6 production and IL-4 similarly
did not alter baseline levels of IL-6 […]. (engmed44)

The next example may be interpreted as being of a more corrective use:

(7) Untreated PANC-1 cells did not express RNA for IL-8 but this appeared when
cells were pre-treated with TNF or IL-1. (engmed44)

The author corrects a point of view which the readers could anticipate, given the
description of the study up till now. The justification of an underlying point of
view of this kind is given in the succeeding concessive construction but this ap-
peared when cells were pre-treated with TNF or IL-1.
The general impression of the negations used in this article is that they are
primarily non-refutative, in the sense that they are not used to contest points
of view attributed to specific studies or researchers. However, we find negations
which may indicate an implicit polemicity, as the following example taken from
the Introduction section, containing presentation of and dialogue with existing
literature in the relevant domain:
(8) The obvious question, which arises, concerns the source and nature of the
stimulus for acute phase protein production. In vitro studies have suggested
that re-prioritisation of hepatic protein synthesis in favour of acute phase
protein production is mediated by […] (5). It has been suggested that these
cytokines might account in part for many of the metabolic abnormalities and
consequent tissue wasting observed in the cancer host. Despite compelling
data demonstrating the potency of IL-6, IL-1 and TNF in stimulating acute
phase protein production in vitro, in animal models and in humans, these fac-
tors are often not detectable in the sera of tumor-bearing patients who have
an elevated APPR. (engmed44)
The analysis undertaken above has given a positive answer to the question
raised in the introduction of this section. There are, in fact, differences between
disciplines as regards the nature of the negation constructions that are used. The
linguistics article, engling06, demonstrates a varied use of negation with the refuta-
tive use as the dominating one. This corresponds well with other observations indi-
cating that linguists are more polemic and expressive than economists and medical
authors (see for example the study of author roles, section 4.2; see also section
7.1). Let us also mention that the frequency of first person plural pronouns is par-
ticularly high in this linguistics article. The economics article, engecon28, also dis-
plays a varied use of negation constructions. However, the refutative use is not the
 Chapter 6. Presence of the others 255

dominating one; engecon28 contains more corrective (and thus not very polemic)
use of not. In parallel to the comment on pronominal use in engling06, it is worth
mentioning that engecon28 displays a relatively low frequency of first person pro-
nouns. Finally, in the medical article, engmed44, the use of not is less varied than
in the other two articles, and is mostly confined to a descriptive or corrective use.
The polemic aspect seems to be very modest here. The use of first person pronouns
is also relatively small; however, the presence of adversative but is relatively impor-
tant. This last point may indicate that there are contrastive (and perhaps polemic)
aspects related to other expressions than the negation particle not.
The analysis undertaken above has also shown that the general quantitative
differences do not correspond directly to the qualitative differences between the
three articles investigated. From our quantitative data we know that economics
and medicine are similar disciplines as regards the frequency of negation; how-
ever, in the present study of specific articles, we have seen that there are clear quali-
tative differences between engecon29 and engmed44. By its varied use of nega-
tion constructions, including some refutative use, engecon29 is more similar to
engling06 than to engmed44. The presence of negation constructions may, in fact,
imply many different uses and thus contribute to different tones in various arti-
cles. As stated earlier, economics seems sometimes to be biased towards linguistics
and sometimes towards medicine. This particular study of an individual econom-
ics article has in fact demonstrated this instability in an interesting way: While
the quantitative data indicated similarity between economics and medicine, the
qualitative analysis of three individual articles from the three disciplines indicates
greater similarity between economics and linguistics.

6.6 Concluding remarks

In this chapter, we have focused on the presence of others and their voices (the
other-dimension) through analyses of the use of bibliographical references and of
polyphonic constructions. We have addressed our second main question related to
the way voices of other researchers are reflected. We have seen that the ScaPoLine
framework can handle both explicit presence through referencing in some form
of reported speech, as well as implicit presence as manifested, for example, by an
adversative and concessive connective like but, or by the negative particle not.
As regards the quantitative results presented in 6.2, we saw that our somewhat
loosely formulated hypothesis concerning a possible relation between the frequen-
cy of references and the size and age of a discipline was confirmed. In our corpus
medical articles have in fact the highest frequency of bibliographical references,
linguistics ranks second and economics third.
256 Academic Voices

To the extent that Hyland’s (2000) study of references in English articles is

comparable to our findings, we see some similarities and some differences in rela-
tion to our English subcorpus (see 6.2).
We also formulated a working hypothesis concerning language differences:
Since English is the language having the largest extension, articles written in Eng-
lish are likely to contain the highest number of references. This hypothesis was not
confirmed. Norwegian articles, written in the language with the smallest extension
of the three, were in fact very close to the frequency observed in English articles.
French articles, written in a language with a large extension compared to Nor-
wegian, contained the smallest number of references in all three disciplines. The
explanation for these differences clearly has to do with other factors than language
extension only. Being a small country with relatively small research communities,
Norwegian researchers probably feel the necessity of acquiring knowledge from
and relating their research to larger, international communities. In order to justify
this claim, a systematic study of national versus international references in the
Norwegian articles is required. We have looked at a selection of Norwegian refer-
ences that seem to provide some support for it.
As regards our study of the distribution of different types of bibliographical
references, we observed much clearer differences between disciplines than be-
tween languages. The non-integral type (R1) was by far the most frequent in medi-
cal articles and the partly integral (R2) and the semi-integral (R3) references were
the most frequent in linguistics and economics articles (linguists use more R2 than
economists). The fully integral type (R4; verbatim quote) was used only to a very
modest extent and then mainly by linguists.
In our discussion of rhetorical functions of bibliographical references (6.2.4), we
pointed to the dimensions of continuity/progress and cooperation/position as rel-
evant guidelines for the interpretation of the function of references. However, each
article may demonstrate individual patterns of reference use according to the socio-
professional context in which the research reported in the article is produced.
In section 6.3, called “Who can do what?”, we compared verbs combined with
first person subject pronouns (self) and verbs used with references in subject posi-
tion (others). Our findings showed a relatively homogeneous picture of the others,
who mainly are given the researcher role, signalled by verbs such as find and show.
The general picture of the self-dimension turned out to be clearly more heteroge-
neous. The three main author roles presented in 4.2 (researcher, writer, arguer])
were all represented with a large spectrum of different verbs. As regards verbs
combined with the French pronoun on (‘one’), we observed that some of the ten
most frequent verbs were used in all disciplines. We also noticed that no argu-
mentative or position verb was among the ten most frequently used, in any of the
 Chapter 6. Presence of the others 257

three disciplines, thus implying that the on-subject does not commonly assume
the arguer role.
Let us now turn to the implicit presence of the others through negation and
the adversative connective ‘but’. In 6.1, we saw that both negation and adversative
‘but’ were clearly used more in linguistics than in economics and medicine, and
more in Norwegian than in English and French. To the extent that negation is used
in a polemic way, this may indicate that linguists are more argumentative and even
more “critical” than economists and medical researchers, and that Norwegian au-
thors are more so than their English and French colleagues.
Negation and adversative ‘but’ are also obvious markers of polyphony, i.e. they
give instructions pointing to the presence of voices other than the author’s. The
presence of others, be they specified or unspecified, individual or collective, con-
stitutes an important aspect of the enunciative and rhetorical characterisation of
the research article.
In 6.4, we questioned the traditional distinction between polemic and de-
scriptive negation and posited the existence of a scale of uses manifesting different
degrees of polemicity, depending on who is responsible for the negated point of
view (self or other, internal or external voice). We proposed four uses: refutative,
delimitative, corrective and descriptive. In the case study (6.5), we demonstrated
how this classification may contribute to quite revealing results when applied to
the analysis of whole articles. The study showed that qualitative analyses of po-
lyphony are necessary in order to account for different uses of negation in articles
taken from different disciplines.
The ScaPoLine perspective has proven to be fruitful in that it unveils not only
explicit but also implicit sources of voices. The search for the interaction between
implicit and explicit voices helps us to discover important rhetorical strategies
in discourse. And the polyphonic perspective has made us ask questions that are
important in the interpretation process and in the search for relevant contextual
explanatory factors.
To conclude, in general the discipline factor is more important than the lan-
guage factor as regards the cultural identities we have pointed to (see also section
3.2): It is the discipline factor, i.e. differences among disciplines, which has the
greatest effect on the relative frequency related to both bibliographical referenc-
es and adversative connectives. As regards negation, discipline and language are
equally important (see Table 3.2-1).
 chapter 7

Conclusions

7.0 Introduction

In this book, we have focused on person manifestation in research articles. We

have studied how this genre displays traces of academic voices: of the authors
themselves (the self-dimension), of the reader and of third persons, in particular
other members of the discipline community (the other-dimension), and in rela-
tion to the interaction with their readers and others (self & other). Our aim has
been to provide answers to the following questions, repeated from 1.1:
• How do article authors manifest themselves in the texts?
• How are the voices of other researchers reflected?
• How do authors present and promote their own research?

Taking as our point of departure the textual manifestations of different aca-

demic voices, we have based our investigations on the electronic KIAP Corpus,
which is a collection of 450 refereed research articles distributed evenly over three
languages viz. English, French and Norwegian, and over three disciplines, viz.
economics, linguistics and medicine. This has allowed us to study texts from the
research article genre in a doubly contrastive perspective and on a solid quantita-
tive foundation. We have performed large-scale statistical analyses of the relative
frequencies of six selected linguistic features (see 7.1). In addition, we have un-
dertaken a number of qualitative and combined quantitative/qualitative studies
of these and other features. Among the qualitative investigations are a number of
case studies.
Applying an enunciative and polyphonic approach to language and beginning
with specific linguistic features at the micro-level, we have taken a primarily bot-
tom–up perspective. This perspective, however, has been assumed with a view to
providing a solid empirical basis for generalisations not only at the micro-level,
but also at the textual and contextual macro-levels. The genre of the research ar-
ticle is socio-professionally situated, has a historically established conventional
form (or, not to beg the question, one or more conventional forms) and is, by its
nature, rhetorical. Throughout the book, we have attempted to connect the micro-
260 Academic Voices

and macro-levels. We believe that our study has produced new knowledge which
we hope may inspire further research in this field.
In the present concluding chapter, we first sum up and discuss the quantitative
results of our investigations (7.1) We then return to the settings that we estab-
lished in Chapter 1 for the concept of cultural identity, and attempt to relate these
to our findings – of differences and similarities between languages and between
disciplines – and to our recurrent observation of the great amount of individual
variability (7.2). Finally, in 7.3 we consider our findings in the light of the concept
of academic voices.

7.1 Quantitative results: summing up and discussion

7.1.1 Overview

Focusing on person manifestation in research articles as our main object of study,

we have undertaken a number of quantitative studies based on the KIAP Corpus.
We repeat the list here for convenience: (1) first person grammatical subjects (‘I’
and ‘we’), (2) indefinite pronoun subjects (‘one’), (3) metatext (e.g. ‘section’), (4)
negation (‘not’), (5) adversative conjunctions (‘but’) and (6) bibliographical refer-
ences (e.g. Johnson (1999)).
The number of features investigated is thus fairly small, and we have stated,
several times, that our findings for these features can only be indicative of a larger
picture. Nevertheless, our doubly comparative study based on extensive empiri-
cal materials has allowed us to produce novel kinds of findings: Very little previ-
ous research has had such a double perspective, and no other studies have, to our
knowledge, had an equally broad empirical foundation.
The data on our main quantitative results have been presented in Chapter 3
and sections 4.1, 5.1 and 6.1. Several other sections also present results of addi-
tional quantitative investigations of various types and scales. In the present sec-
tion, we attempt to sum up the tendencies that we have documented, and discuss
what conclusions can be drawn on the basis of them.
As regards the discipline factor, linguistics exhibits the strongest degree of per-
son manifestation and medicine the weakest, with economics assuming a variable
middle position, sometimes closest to linguistics and sometimes to medicine.
As regards the language factor, English exhibits the strongest degree of person
manifestation but is closely followed by Norwegian, whereas French has a much
weaker degree of person manifestation.
 Chapter 7. Conclusions 261

7.1.2 The disciplines

Starting with disciplinary differences and similarities, we have found that medical
articles have fewer first person subjects than texts from the two other disciplines.
This less manifest presence of medical authors is related to a similar relative scarci-
ty of indefinite pronoun subjects in medicine (except in Norwegian, where the dis-
ciplines differ very little); although referring vaguely, indefinite pronouns, like first
person pronouns, refer to human agents: researchers, writers etc. In French, the
indefinite on is often used with a more specific reference, and it is possibly indica-
tive that it is used more in French than the corresponding pronouns are in Norwe-
gian (see below) – except in French medicine, which has less of it than Norwegian
medicine: It may be that this is because medical writers avoid a large amount of on
just as first person pronouns are avoided. However, as we saw in 4.4.1, to the extent
that French medical writers use on, the pronoun quite often refers to the authors,
so the reasons behind its low frequency appear to be more complex. A very small
amount of metatext and ‘let us’-imperatives in medical articles also contributes to
a weaker personal presence in medicine than in economics and linguistics. On the
other hand, medical articles have more bibliographical references than economics
and linguistics texts, positioning the medical research in question more firmly in
its disciplinary context. When we compare economics and linguistics with respect
to metatext and references, we find that economics has more of the former (except
in English), while linguistics has more of the latter (except in French).
With respect to the weaker personal presence of authors and interaction of
author and reader in medical articles, it might be tempting to ascribe this to a
stronger presence of an objectivist ideal of science in medical research than in
economics and (perhaps especially) linguistics. An overt presence of author and
reader makes the interpretative aspect of research more evident, and from an ob-
jectivist perspective this may be regarded as undesirable. In our view, however,
such possible differences concerning what science is should not be the first place
to look for explanations. Rather, we want to point to other differences between
medicine and our other two disciplines. For present purposes, we contrast medi-
cine primarily with linguistics.
First, medical research is more clearly cumulative than linguistic research, in
the sense that new research contributions are added to the collective knowledge
capital. It may safely be stated that it would be highly unusual for a medical re-
search article to start with discussions of the very basic medical conceptual system.
In the normal case, this is quite unnecessary. Even though linguistics, too, accu-
mulates knowledge, the linguist more often needs to discuss fairly fundamental
concepts and argue for his or her understanding of the research field in order to
try to ascertain a similar understanding on the part of the reader. The need for an
262 Academic Voices

overt personal presence and for appeals to the reader is greater, then, in linguis-
tics than in medicine. (For a related discussion, on data-driven and conceptually
driven academic discourse, see MacDonald (1989).) In our view, this situation is
above all due to the central objects of study being ontologically different between
the disciplines. The more cumulative nature of medical research is, we believe,
also reflected in the higher frequency of bibliographical references in this disci-
pline than in economics and linguistics, adding to the factor of the higher age and
greater size of medicine (see 6.1): The reported research builds on and modifies
the previous research that is referred to and is in general highly integrated in the
community research activities. This indicates a different potential axis for gener-
alisation than in Hyland (2000), where the distinction between ”soft” and ”hard”
disciplines is referred to (see 6.3).
Second, writing occupies different positions in the research processes. In medi-
cal research, writing is typically a complex collective process that takes place after
the ”real” research is finished (see Gilhus 2003). It is natural for this situation to
result in fewer addressee features (like ‘let us’-imperatives) in certain parts of the ar-
ticle (viz. Material/Method and Results), because it is harder to engage the reader in
what has already been done. Medical texts also employ the past tense more than lin-
guistic texts. In linguistics, it is more often (but far from always) the case that there
is no temporal distance between ”real” research and writing. When the research is
performed, so to speak, in the text, during writing, the reader can more easily be
engaged. In linguistics, the present is the typical grammatical tense employed.
Third, there is the relation between research group and author group in
medicine: There is not necessarily full overlap between the two. Some of the re-
search may have been performed by only some of the authors, and other medical
staff than the authors may have been involved. Quite often it is of no importance
exactly who did what, and the choice of some impersonal mode of expression may
simply be more appropriate than using ‘we’.
Fourth, and most importantly in connection with the low frequency of meta-
text and ‘let us’-imperatives in medicine, medical research articles typically exhibit
a simpler thematic structure than linguistic ones (see Breivega 2003: 235–236).
This is closely related to the strict adherence of the typical medical article to the
IMRAD format. There is a weaker need for medical writers than for linguists to
exhibit text composition, which is perhaps the principal function of metatext and
‘let us’-imperatives.
As regards economics and linguistics, the higher frequency of metatext in the
former is at least partly due to the fact that economics articles typically have an
article structure presentation at the end of the Introduction section. Swales (2004)
assumes that such presentations are used more the less IMRAD-structured the
articles from the discipline tend to be (see 5.2). However, this leads us to expect an
 Chapter 7. Conclusions 263

even higher frequency in linguistics than in economics, and our study of metatext
does not corroborate this prediction.
We have already touched upon the more frequent use of bibliographical ref-
erences in medicine than in the other disciplines above. If we are on the right
track in assuming that this is related to the more cumulative nature of medicine
and the stronger integration of medical research in an established and global re-
search community, the higher frequency of references in (English and Norwegian)
linguistics than in economics would seem to indicate that economics is the less
cumulative and integrated of the two. There may be other explanations, however
(see below).
Linguistics has more frequent use of negation and adversative conjunctions
than economics and medicine. It seems reasonable to assume that an important
factor behind this clear tendency is that linguistics texts are often more overtly
argumentative than medical and economics texts. This may in turn be related to a
less cumulative state of affairs in linguistics than in medicine, but this can hardly
be the sole reason, and it is not clear how economics fits in here. However, a more
argumentative tendency in linguistics than in the other disciplines may also be
perceptible in the verb use in connection with bibliographical references – the
views of others appear to be referred to a little more frequently with non-factive
verbs in linguistics than in the other disciplines (see 6.3). It is possible that the
higher frequency of references in linguistics than in economics comes partly from
a more extensive argumentation with disciplinary “opponents” in addition to ref-
erences to work that is in line with the author’s own contribution, in order to sub-
stantiate the claims made. In addition, linguistics has a longer research tradition
than economics.

7.1.3 The languages

English and Norwegian use first person subjects more frequently than French.
This result is more interesting when broken down into three contributing factors:
1) the use of ‘I’ in single-author articles, 2) the use of ‘we’ in single-author articles
and 3) the use of ‘we’ in multi-author articles. The use of ‘I’ refers to the author and
excludes the reader, and the use of ‘we’ in multi-author articles can (but need not)
refer to the authors and exclude the reader. That is, these two factors have a lot in
common. We also find that they behave similarly in ranking the languages: French
has less frequent use of them than English and Norwegian, while the last two are
very similar to one another in this respect. ‘We’ in single-author articles, on the
other hand, often includes the reader. This use is more frequent in Norwegian
than in English and French; the last two are similar. Our findings first of all con-
firm that French je (‘I’) is used less than the corresponding English and Norwe-
264 Academic Voices

gian pronouns (although it is indeed used in French linguistics). The “detestable”

first person singular – le moi haïssable – is frequently avoided (see 4.2 and 4.4.2).
The French also use the first person plural less in multi-author articles than their
English- and Norwegian-writing colleagues. Thus, our study corroborates the ob-
servations in Loffler-Laurian (1980) that the French avoidance of direct reference
to the first person carries over from ‘I’ contexts to ‘we’ contexts. There appears to
be little difference between English and French in the use of ‘we’ in single-author
articles, but Norwegians use this more. We have not studied the phenomenon in
detail, but our impression is that it involves a more frequent use of expressions that
portray the author and the reader as collaborators in the same activities (see 4.3).
This indicates, then, a clearer Norwegian reader-orientation in this respect than
even for English.
We have further found, unsurprisingly, that English indefinite one is used much
less than its French and Norwegian correspondences. The difference between Eng-
lish and Norwegian is evidently a stylistic one, where the indefinite pronoun is
regarded as very marked in English, but much less so in Norwegian. As concerns
French on, this pronoun is quite often used as a referential substitute for personal
pronouns, and this extended use is probably the reason that it is more frequently
used than the Norwegian pronouns. It is noteworthy that the situation is reversed
in medicine, where Norwegian has more indefinite pronouns than French (see
above). Possibly, this is related to the higher degree of avoidance of personal refer-
ence in medicine than in the other disciplines. As already mentioned, however, the
matter appears to be more complex.
For metatext and bibliographical references, we have found the highest relative
frequencies in English and the lowest in French. For ‘let us’-imperatives, however,
which are functionally similar to metatext, it is French that has the highest relative
frequency. We have discussed the focus on a reader orientation in the teaching of
writing in the Anglo-American world and the position of our three languages on a
scale of reader/writer orientation: Considering metatext, English and Norwegian
come out as more reader-oriented than French. The higher frequency of ‘let us’-
imperatives in French modify this picture somewhat, however, reducing the dif-
ferences between the languages. As regards bibliographical references, any attempt
on our part to explain the language differences at this stage must be speculative.
We have tried to relate them to differences in the size of disciplinary communities:
The English ones are the biggest and require more references to colleagues. The
Norwegian ones would in light of this be expected to have the fewest references,
but the general orientation of Norwegian research towards the international re-
search communities, especially the Anglo-American ones, might be a contribut-
ing factor to the higher frequency of references here than in French. Norwegian
articles typically have many references to publications in English. The French, at
 Chapter 7. Conclusions 265

least in the humanities, seem to have focused more on the francophone research
world. These speculations would need to be corroborated by more detailed studies
of the references, however, specifically of the language and national origin of the
publications referred to.
Finally, Norwegian authors use negation (with ‘not’) and adversative conjunc-
tions more frequently than their English- and French-writing colleagues. As noted
in 6.1, it is to be expected that there is more negation using ‘not’ in Norwegian
than in English and French, because the Norwegian negative quantifier ingen
(‘no’) is subject to special grammatical restrictions, so that constructions involv-
ing ikke/ikkje (‘not’) are often used instead. However, we see no similar possibility
for explanation of the higher Norwegian frequency of adversatives. In our disci-
plinary comparisons, we found that linguistics uses both negation and adversa-
tives more than the other disciplines, and we pointed to the possible explanation
that linguistics articles are more argumentative than economics and medical texts
(to the extent that these features are used polemically). Our findings of language
differences – restricted to the use of negation and adversative conjunctions – ap-
pear to indicate that Norwegian academic discourse is more overtly polyphonic
than academic discourse in the other two languages. Whether this tendency is also
present with regard to other polyphonic markers remains to be corroborated by
more detailed investigations.

7.1.4 Gender

Finally, we mention our most important finding regarding gender: There are hard-
ly any significant differences between single female and single male authors for the
features that we have looked at. We have found no significant gender differences
for the use of ‘I’ or ‘we’, nor for negation or bibliographical references. In English
linguistics, male authors use ‘one’ more than female authors, but the difference is
not found in English economics or in the other languages. In French and Norwe-
gian economics, males use metatext more than females, but the difference is not
found in English economics or in linguistics in any of the languages. In Norwe-
gian linguistics, men use adversatives more than women, but this difference is not
found in linguistics in the other languages, nor in economics. These few gender
differences appear rather haphazard to us and do not seem to reflect general ten-
dencies that call for further explanation.
266 Academic Voices

7.2 Community identity versus individual heterogeneity

In Chapter 1 we presented some reflections on the concept of cultural identity.

We established four settings which we saw as relevant to a discussion of how the
academic author develops a textual persona (section 1.3). The time has now come
to return to that discussion and see if these settings have been useful in our inves-
tigation of academic discourse across languages and disciplines.
As all authors belong in a national culture, we established that as our first set-
ting. This setting is linked in an obvious manner to the language variable in the
present study. In the various discussions of how the authors are visible in texts
written in the three languages represented in our study, we have linked author
presence, in the self- and other-dimensions, for several of our features to values
prevalent in the national cultures. An example is the high frequency of metatext
in English and Norwegian economics and linguistics texts compared with the
French texts. This may be seen as a possible reflection of the contrast between the
Anglo-American and Scandinavian cultures relying on explicit transmission of
information, and the French culture, where more implicit information coded in
the context would appear to be the norm (section 5.2). We did, however, also point
out that the form-based approach adopted for the study of metatext might have
led us to exclude functionally equivalent expressions in French which could have
resulted in a different impression of French in this respect (cf. Moreno 1998).
In section 1.3 we also stated that we expected each language to show linguistic
preferences at the lexical level. Our investigation has proved this to be the case in
quite a few instances, as reflected for instance in the discussions of author roles,
displayed by various pronoun and verb constructions (see e.g. sections 4.2 and 4.5),
as well as for negation (section 6.1.2). These linguistic preferences are not always
easy to explain by reference to cultural aspects, at least not at the present level of
knowledge. However, we see clear trends towards increased focus on cultural values
in linguistic research (e.g. Koutsantoni 2005), and future research in this field may
benefit from closer integration of more traditional text linguistic research and ap-
proaches common within intercultural communication studies. As regards simi-
larities observed between English and Norwegian, they might to some extent be
due to the typological closeness between the two languages. Whether this in turn
partly may explain the closeness in culture is difficult to ascertain. Analyses un-
dertaken within the framework of the so-called Quaestio model (cf. e.g. Hendriks
2005a: 7–8) have shown that languages differ in how they organise information on
domains such as persons, spaces and events in order to make coherent texts (e.g.
Hendriks 2005b; Stutterheim & Lambert 2005). Whether linguistic preferences
found in our data may be explained within this framework remains to be seen.
 Chapter 7. Conclusions 267

The second proposed setting was the academic world. This is the only setting
not implying a contrast, unless we want to see it as opposed to everyday commu-
nication, which was not the point in our study. Rather, we focused on the common
features of scientific communication, pointing to the rhetorical, and sometimes
more specifically, the persuasive nature of such communication. Persuasion was
touched upon in our discussion of the “author as arguer”-role (sections 4.2 and
4.5) as well as in our investigation of overt evaluation in abstracts (sections 4.5
and 4.6.2), but, as our quotes from Melander et al. (1997) and Lindeberg (2004) in
section 4.5 so clearly indicate, the actual persuasion is carried out in a much more
complex and sophisticated way than merely by means of “sales talk” items such
as innovative, good and We argue that … An academic text in addition becomes
persuasive through the interplay of a range of elements, including the choice of
co-author(s) and publishing channel, methodology, cited (and left out) sources,
information ordering and hedges, to name but a few.
As for the features we had selected for investigation, they in our view turned
out to be relevant for a description of academic discourse, and they also turned
out to be sensitive with respect to the language and discipline factors. For one
of the features, first person plural pronouns, for example, we had on the basis of
previous research expected variation, between languages in particular, but also
between disciplines. For another, bibliographical references, we had only general
preconceptions of what patterns to expect. This feature turned out to be particu-
larly discipline-sensitive.
Our third setting was disciplinary identity, hence representing a setting re-
lating directly to our second independent variable in this study. As indicated in
Chapter 1, our point of departure for the present study was that we assumed dis-
ciplinary differences to be more important than language differences. Our data
can be said to have confirmed this assumption (see Chapter 3 and 7.1.2), so that
we may postulate disciplinary identity as a meaningful construct in this context.
At the outset of our study we envisaged the three disciplines to display linguis-
tic behaviour which could be plotted as being spread out evenly along a continuum.
What we have found, however, is slightly different. With a few exceptions (e.g. meta-
text and bibliographical references; cf. section 3.3) the general picture reveals that
while linguistics and medicine, as expected, in most cases represent the end points
of the continuum, economics cannot simply be described as being equidistant from
these two along the line. For some of the features we have studied, especially meta-
text, first person subjects and ‘let us’-constructions, it is rather closer to linguistics.
Hence, linguistics and economics to some extent share linguistic practices, which
in turn may be seen as a reflection of shared epistemological features. On the other
hand, our lexical-semantic study of how research findings are reported by means
of the word family ‘result’ (section 5.4) points to differences in their epistemolo-
268 Academic Voices

gies. In Chapter 1 we quoted Breivega (2003: 16), who states that research results
in the humanities often take the shape of the actual argumentation in the text (sec-
tion 1.3.4). Our lexical study in Chapter 5 supports this claim for our humanities
discipline. There is very little use of ‘result’ in the linguistics texts (section 5.4). The
economists, on the other hand, are frequent users of ‘result’, and in fact in our mate-
rial use it more than even the medical authors. Still, as pointed out in another quote
from section 1.3.4, very few economics articles have a section termed Results (Bloor
& Bloor 1993). On the basis of these reflections, we may conclude that economics is
similar to linguistics in some respects and to medicine in other respects and hence
occupies a somewhat unstable middle position of the continuum. Our analysis can
thus perhaps be seen as providing support for Becher & Trowler’s (2001) statement
that “[a] detailed analysis of disciplinary discourse […] can help not only to bring
out characteristic cultural features of disciplines but also to highlight various as-
pects of the knowledge domains to which they relate” (p. 46).
Our fourth suggested setting was genre and discourse community. In Chapter
2 we defined genre in rather simple terms as a recurrent social practice, linguis-
tically realised through text containing traits which make up more or less fixed
patterns (section 2.1). We also stated that we did not see these traits as constitut-
ing absolute patterns. On the basis of this definition, the heterogeneity observed
between the subcorpora as well as in single articles in our view still allows us to
claim that the research article of the kind investigated here represents one genre.
Whether it is fruitful to establish subtypes of the genre is another matter which
must be discussed in studies focussing on that.1 This has not been a key issue in
the present book.
As regards the concept of discourse community, we in Chapter 1 acknowl-
edged the special status of English as a research language (cf. Montgomery 2004),
a fact which has consequences for how other, lesser used languages, are able to
exist as mediators in scientific communication. In our study this is particularly
relevant for Norwegian. Does the fact that we have used only one Norwegian lan-
guage journal in two of the disciplines studied here, viz. economics and medicine,
and the fact that the large majority of Norwegian economists and medical authors
publish primarily in English, make it meaningless to compare research articles
in these journals with those of the more numerous English and French ones, ad-
dressing much larger and perhaps also more specialised discourse communities?
It is our opinion that it has been meaningful to compare, and that the articles can
be said to belong to the same genre. The majority of the authors of the Norwe-

1. Swales (2004) in the chapter named ”The research article revisited” does argue for dividing
his 1990 research article genre into three separate genres, viz. the theoretical, the experimental
and the review article.
 Chapter 7. Conclusions 269

gian texts also publish in English (or other major languages), and their readers
also read articles in the international journals. The results obtained in our study
support this claim, in the sense that the Norwegian texts in each discipline have
been shown to share linguistic behaviour with their English, and to a lesser extent,
French counterparts.
These four settings that we have chosen to use as a backdrop for our study of
linguistic practices across languages and disciplines have, in our view, served us
well. They have enabled us to some extent to get beyond the mere reporting of
numerical data. On the other hand, the fact that our text corpus is so large makes
us more confident when we make claims about general or typical behaviour. How-
ever, we once more want to draw attention to the voice of the individual author.
Even from the outset of our study we realised that this dimension should not be
ignored, no matter how large the corpus and how comprehensive our battery of
statistical tests. We have attempted to provide snapshots of the individuals con-
tributing to the general impressions through the case studies included in the three
chapters dealing with our data (Chapters 4–6). Some of these qualitative studies
have confirmed the broad, general picture, e.g. the case study of author presence in
abstracts in section 4.6.2. Other case studies, again, have been illustrations of the
wide divergence in the corpus at the level of individual texts, such as our study of
pronouns (section 4.6.1). Yet other individual studies have shown how numerical-
ly similar behaviour in fact on closer inspection hides differences at the linguistic
level. A case in point is the qualitative study of metatext in two economics articles
(section 5.5.1).
All in all, then, we may say that the interaction of our two key concepts of
culture and identity as reflected in the linguistic practices of academic authors has
proved to be even more complex and intriguing than envisaged at the outset of
this investigation.

7.3 Academic voices in harmony and contrast

We have investigated person manifestation in research articles through different

kinds of textual traces of academic voices. In the broad pragmatic and rhetorical
meaning given to this concept, it has represented a unifying instrument for the
analyses of the various phenomena studied. The rhetorical situation or context
in which a research article is produced is particularly multivoiced, and our enun-
ciative and polyphonic theoretical approach has led to new insight into person
manifestation and thereby a new awareness of how academic discourse works. It
has become clear that multivoiced or polyphonic presence is manifested in differ-
ent ways and to different degrees in different disciplines and in different languages.
270 Academic Voices

Some disciplinary or language voices are in harmony in the sense that they use
similar linguistic features; others are in contrast in using different linguistic fea-
tures or similar features but with different frequencies.
The author voice is always present in an utterance, but the type of presence var-
ies between the two extremes of implicit or hidden presence and explicit presence
by means of the first person pronouns ‘I’ or ‘we’. As regards the self-dimension, we
have focused on explicit author voice in more or less performative constructions
of the type ‘I show’, ‘I argue’ and ‘I analyse’, manifesting the roles of writer, arguer
and researcher respectively. In this dimension, we have also studied the collective
author voice, including reader and/or other researchers through the pronouns ‘we’
and ‘one’. In the analyses of author/reader interaction (the self- & other-dimen-
sion), we have studied explicit addresses to and inclusion of the reader in metatex-
tual expressions and ‘let us’-imperatives. It should be noted, however, that, through
these constructions, the authors do not give the floor to the reader; the latter is
only addressed or included in the authors’ own voice. Finally, in the other-dimen-
sion, we have studied third persons’ voices as manifested through bibliographical
references. In some (but rather few) cases, third person voices are given the floor
through quotes. Most often, however, they are not allowed to speak for themselves;
their voices are represented through the authors’ own wording.
In addition to these types of explicit manifestation of voices, we have stud-
ied implicit manifestation as realised through the negative particle ‘not’ and the
adversative-concessive connective ‘but’. The study of these markers of polyphony
has shown that there is more to the presence of self than constructions with first
person pronouns and more to the presence of others than a straightforward iden-
tification of bibliographical references in the form of reported speech. In order to
get a fuller picture of the complex interaction of voices characterising the socio-
professional practice found in research articles, it is necessary to take into account
a larger spectrum of signals (as is also indicated by Thompson 1996). The study
of the polyphony markers of negation and concession has shown that voices are
present even if the sources of the points of view conveyed are not always identifi-
able. However, through the use of these markers, the different voices are given
specific roles in the polyphonic drama presented by the author: the negation ‘not’
in its polemic meaning always points to an “opponent” (to whom the refuted point
of view is attributed), and the concessive ‘but’ always points to a source whose
point of view is accepted, but not judged valid in the here and now of the text.
When there is explicit presence of third person voices through reported speech
(as realised by bibliographical references), the author attributes different roles to
the cited voices by using different verbs in the inquit. Whereas the role attributed
through non-factive verbs (such as argue) is not obvious, the role attributed by fac-
 Chapter 7. Conclusions 271

tive verbs, such as show, is quite obviously a “supporter” role (e.g. Johnson (2000)
shows that …).
Academic voices in medical articles are in harmony in the sense that they use
similar features whatever language the articles are written in: the other-dimension
is frequently present while the self-dimension is infrequent compared to the other
disciplines.
Linguistic articles in English and Norwegian display academic voices in har-
mony in the sense that both the self-dimension and the self- & other-dimension
are frequent. But their different use of the collective voices manifested by ‘one’ and
‘we’ indicates contrast. In general, academic voices in Norwegian articles contrast
with academic voices in English articles by their more frequent use of the poly-
phonic markers ‘but’ and especially ‘not’.
Academic voices in economics articles are in harmony across languages in the
sense that the self- & other-dimension as manifested through metatextual expres-
sions is more frequent than in the other disciplines.
As regards language, we have already pointed to some tendencies of harmony
between academic voices in English and Norwegian linguistics articles. French
articles in all disciplines display harmony in the sense that they prefer the collec-
tive author voice by using the indefinite on (‘one’) and avoid the self-dimension
as manifested by first person pronouns. The French preference of the collective
author voice is in clear contrast to English articles, but to some extent in harmony
with Norwegian articles.
To conclude, academic voices are developed through interchange within vari-
ous kinds of communities which the authors belong to. However, there is also
room in academic discourse for each author to develop his or her own individual
voice, a voice which may be exploited in very personal ways to achieve the ulti-
mate goal of scientific persuasion.
 References*

Adam, J.-M. 1999. Linguistique textuelle. Des genres de discours aux textes. Paris: Nathan.
Adam, J.-M. 2005. La linguistique textuelle. Introduction à l’analyse textuelle des discours. Paris:
Armand Colin.
Arrivé, M., Gadet, F. and Galmiche, M. 1985. La Grammaire d’aujourd’hui: guide alphabétique de
linguistique française. Paris: Flammarion.
Atkinson, D. 2004. ”Contrasting rhetorics/contrasting cultures: why contrastive rhetoric needs a
better conceptualization of culture.” Journal of English for Academic Purposes 3: 277-289.
Atlani, F. 1984. ”ON L’illusionniste.” In La langue au ras du texte, F. Atlani, L. Danon-Boileau, A.
Grésillon, J.-L. Lebrave and J. Simonin (eds), 13-29. Lille: Presses Universitaire de Lille.
Bakhtin, M. 1986. Speech Genres & other late Essays. Austin: University of Texas Press.
Bakhtine, M. 1970 [1929]. La poétique de Dostoïevski. Paris: Seuil (First edition in Russian: Mos-
cow 1929).
Bakhtine, M. 1978. Esthétique et théorie du roman. Paris: Gallimard. (Translated from Russian
by Daria Olivier.)
Bazerman, C. 1988. Shaping written knowledge. The genre and activity of the experimental article
in science. Madison, WI: University of Wisconsin Press.
Becher, T. and Trowler, P.R. 2001. Academic tribes and territories. Intellectual enquiry and the
culture of disciplines. 2nd ed. Milton Keynes: The Society for research into Higher Educa-
tion & Open University Press.
Benveniste, E. 1966. Problèmes de linguistique générale. Paris: Gallimard.
Berge, K.L. 1990. Tekstnormers diakroni: noen ideer til en sosiotekstologisk teori om tekstnormen-
dring. Stockholm: MINS 33.
Berge, K.L. 2003. ”The scientific text genres as social actions: text theoretical reflections on the
relations between context and text in scientific writing.” In Academic discourse. Multidisci-
plinary approaches, K. Fløttum and F. Rastier (eds), 141-157. Oslo: Novus.
Berge, K.L. and Ledin, P. 2001. ”Perspektiv på genre.” Rhetorica Scandinavica 18: 4-16.
Berkenkotter, C. and Huckin, T.N. 1995. Genre knowledge in disciplinary communication: Cogni-
tion/Culture/Power. Hillsdale, NJ: Lawrence Erlbaum.
Bhatia, V. 2001. ”Analysing genre: some conceptual issues.” In Academic writing in context, M.
Hewings (ed.), 79-92. The University of Birmingham Press.
Biber, D., Conrad, S. and Reppen, R. 1998. Corpus linguistics. Investigating language structure
and use. Cambridge: Cambridge University Press.

* A separate list of KIAP publications appears below. Some of these, but not all, are found
in this general list as well, which explains why some entries may seem irregular. The referen-
ces may have, e.g., Fløttum (2003b, e, g, i), but not (2003a, c, d, f, h). The latter are then listed in
the separate KIAP publications list.
274 Academic Voices

Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. 1999. Grammar of spoken and
written English. London: Longman.
Blanche-Benveniste, C. 2003. ”Le double jeu du pronom ‘on’”. In La syntaxe raisonnée. Mélanges
de linguistique générale offerts à Annie Boone à l’occasion de son 60e anniversaire, P. Hader-
mann, A. Van Slijcke and M. Berré (eds), 43-56. Louvain-la-Neuve: de Boeck Duculot.
Blom, J.-P. and Gumperz, J.J. 1972. ”Social Meaning in Linguistic Structure: Code-Switching in
Norway.” In Directions in Sociolinguistics. The Ethnography of Communication, J.J. Gumperz
and D. Hymes (eds), 407-434. New York: Holt, Rinehart and Winston, Inc.
Bloor, M. and Bloor, T. 1993. ”How economists modify propositions.” In Economics and lan-
guage, W. Henderson, T. Dudley-Evans and R. Backhouse (eds), 153-169. London/New
York: Routledge.
Boch, F. and Grossmann, F. 2002. ”Se référer au discours d’autrui: comparaisons entre experts et
néophytes.” Enjeu 54: 41-51.
Bondi, M. 2005. ”Polyphony in academic discourse: a cross-cultural perspective.” Keynote lec-
ture given at the conference Cross-cultural and Cross linguistic Perspectives on Academic
Discourse, University of Turku, 21 May, 2005.
Bondi, M. and Silver, M. 2004. ”Textual voices: a cross-disciplinary study of attribution in aca-
demic discourse.” In Evaluation in spoken and written academic discourse, L. Anderson and
J. Bamford (eds), 121-141. Roma: Officina.
Breivega, K.R. 2003. Vitskaplege argumentasjonsstrategiar. Ein komparativ analyse av superstruk-
turelle konfigurasjonar i medisinske, historiske og språkvitskaplege artiklar. Sakprosa 8. Oslo:
Norsk sakprosa.
Breivega, K.R., Dahl, T. and Fløttum, K. 2002. ”Traces of self and others in research articles. A
comparative pilot study of English, French and Norwegian research articles in medicine,
economics and linguistics.” International Journal of Applied Linguistics 12 (2): 218-239.
Bres, J. 1999. ”Vous les entendez ? Analyse du discours et dialogisme.” Modèles linguistiques XX
(2): 71-86.
Brett, P. 1994. ”A genre analysis of the Result section of sociology articles.” English for Specific
Purposes 13 (1): 47-59.
Brown, P. and Levinson, S.C. 1987. Politeness. Some Universals in Language Usage. Cambridge:
Cambridge University Press.
Bunton, D. 1999. ”The use of higher level metatext in Ph.D theses.” English for Specific Purposes
18, Supplement 1: S41-S56.
Butler, C. 1985. Statistics in Linguistics. Oxford: Blackwell.
Candlin, C. and Hyland, K. (eds). 1999. Writing: Texts, Processes and Practices. London: Longman.
Carter-Thomas, S. and Rowley-Jolivet, E. 2001. ”Syntactic differences in oral and written scientific
discourse: The role of information structure.” Asp 31 (3): 19-37.
Cavalla, C. and Grossmann, F. 2005. ”Caractéristiques sémantiques de quelques ”Noms scienti-
fiques” dans l’article de recherche en français.” Akademisk Prosa 3: 47-59. Skrifter fra KIAP.
Department of Romance studies, University of Bergen.
Clark, H. and Gerrig, R.J. 1990. ”Quotation as demonstrations.” Language 66 (4): 764-805.
Collins, P. 2004. ”Let-imperatives in English.” International Journal of Corpus Linguistics 9: 299-319.
Coltier, D. and Dendale, P. 2004. ”Discours rapporté et évidentialité.” In Le discours rapporté
dans tous ses états, J.M. Lopez et al. (eds), 587-598. Paris: L’Harmattan.
Connor, U. 2004a. Introduction. Journal of English for Academic Purposes 3: 271-276.
Connor, U. 2004b. “Intercultural rhetoric research: beyond texts”. Journal of English for Academic
Purposes 3: 291-304.
 References 275

Crismore, A. 1989. Talking with readers: metadiscourse as rhetorical act. New York: Peter Lang.
Crismore, A. and Farnsworth, R. 1990. ”Metadiscourse in popular and professional science
discourse.” In The writing scholar: Studies in academic discourse, W. Nash (ed.), 118-136.
Newbury Park/London: Sage.
Crismore, A., Markkanen, R. and Steffensen, M. 1993. ”Metadiscourse in persuasive writing. A
study of texts written by American and Finnish university students.” Written Communica-
tion 10 (1): 39-71.
Dahl, T. 2000. Lexical cohesion-based text condensation. An evaluation of automatically produced
summaries of research articles by comparison with author-written abstracts. Dr. art. thesis,
University of Bergen.
Dahl, T. 2003. ”Metadiscourse in research articles.” In Academic discourse. Multidisciplinary ap-
proaches, K. Fløttum and F. Rastier (eds), 120-138. Oslo: Novus.
Dahl, T. 2004b. ”Some characteristics of argumentative abstracts.” Akademisk Prosa 2: 49-67.
Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Dahl, T. 2004c. ”Textual metadiscourse in research articles: a marker of national culture or of
academic discipline?” Journal of Pragmatics 36 (10): 1807-1825.
Day, R.A. 1998. How to Write and Publish a Scientific Paper, 5th ed. Westport, Conn.: Oryx.
Didriksen, A.A. 2004. DONC et autres connecteurs argumentatifs dans les articles de recherche
français. Master’s thesis. Department of Romance studies, University of Bergen.
Ducrot, O. 1984. Le dire et le dit. Paris: Minuit.
Duszak, A. 1994. ”Academic discourse and intellectual styles.” Journal of Pragmatics 21 (3): 291-313.
Dysthe, O. 2002. ”Writing at Norwegian universities in an international perspective: From indirect
strategies of strengthening writing to the Quality Reform.” In Studies in writing, Volume 12:
Teaching academic writing in European higher education, G. Rijlaarsdam, L. Björk, G. Bräuer,
L. Rienecker and P. Stray Jorgensen (eds). Amsterdam: Kluwer Academic Publishers.
Ekelund Jr., R.B. and Hébert, R.F. 1997. A history of economic theory and method. 4th ed. New
York: McGraw-Hill.
Engberg, J. 1997. Konventionen von Fachtextsorten: kontrastive Analysen zu deutschen und dä-
nischen Gerichtsuhrteilen. Tübingen: Gunter Narr.
Faarlund, J.T., Lie, S. and Vannebo, K.I. 1997. Norsk referansegrammatikk. Oslo: Universitets-
forlaget.
Fabricius-Hansen, C. 1998. ”Informational density and translation, with special reference to Ger-
man – Norwegian – English.” In Corpora and Cross-linguistic Research. Theory, Method, and
Case Studies, S. Johansson and S. Oksefjell (eds), 197-234. Amsterdam/Atlanta, GA: Rodopi.
Fabricius-Hansen, C. 2004. ”Wessen Redehintergrund? Indirektheitskontexte aus kontrastiver
Sicht (Deutsch – Norwegisch – Englisch).” In Tempus/Temporalität und Modus/Modalität
im Sprachenvergleich, O. Leirbukt (ed.), 119-156. Tübingen: Stauffenburg.
Fauconnier, G. and Turner, M. 2002. The Way We Think. Conceptual Blending and the Mind’s
Hidden Complexities. New York: Basic Books.
Fløttum, K. 2001a. ”Le résumé scientifique – texte monophonique ou polyphonique ?” Techno-
style CATTW/ACPRTS. Sherbrooke, Canada. 17 (1): 67-86.
Fløttum, K. 2001b. ”Les liens énonciatifs : tentative d’une nouvelle typologie.” Polyphonie – linguis-
tique et littéraire, III. Roskilde: Samfundslitteratur Roskilde, 67-86.
Fløttum, K. 2003b. ”Personal English, indefinite French and plural Norwegian scientific au-
thors? Pronominal author manifestation in research articles: A cross-linguistic disciplinary
study.” Norsk Lingvistisk Tidsskrift 21 (1): 21-55.
276 Academic Voices

Fløttum, K. 2003e. ”Bibliographical references and polyphony in research articles.” In K. Fløt-

tum and F. Rastier (eds), 97-119.
Fløttum, K. 2003g. ”« JE » et le verbe.” TRIBUNE 14: 7-14.
Fløttum, K. 2003i. ”The French pronoun ‘on’ in academic discourse – indefinite versus personal.”
In Proceedings of CIL17, E. Hajičová, A. Kotěšovcová. and J. Mírovský (eds.), CD-ROM.
(Paper marked: S10_KjerstiFlottum). Matfyzpress, MFF UK, ISBN: 80-86732-215.
Fløttum, K. 2004b. ”Thèmes, topiques et marqueurs de cadres discursifs dans les articles scienti-
fiques français : la présence de personne(s).” Scolia (Strasbourg) 18: 167-201.
Fløttum, K. 2004d. ”La présence de l’auteur dans les articles scientifiques : étude des pronoms je,
nous et on.” In Structures et discours, A. Auchlin et al. (eds), 401-416. Québec: Ed. Nota Bene.
Fløttum, K. 2004f. ”Polyfonisk interaksjon via IKKE i vitenskapelig diskurs.” Rhetorica Scandi-
navica 31: 23-40.
Fløttum, K. 2004i. ”Traces of others in research articles: the citation cluster.” In New Directions in
LSP studies, K. Ahmad and M. Rogers (eds), 153-159 (160-166 in pdf-document). Guildford:
University of Surrey (UniS). ISBN: 1-84469-007-5.
Fløttum, K. 2005a. ”Individual linguistic variation and possibilities for academic self-hood?”
Akademisk Prosa 3: 87-100. Skrifter fra KIAP. Department of Romance studies, University
of Bergen.
Fløttum, K. 2005b. ”The self and the others – polyphonic visibility in research articles.” Interna-
tional Journal of Applied Linguistics 15 (1): 29-44.
Fløttum, K. 2005c. ”MOI et AUTRUI dans le discours scientifique : l’exemple de la négation NE...
PAS.” In Dialogisme, polyphonie: approches linguistiques, J. Bres, P.P. Haillet, S. Mellet, H. Nølke
and L. Rosier (eds), 323-337. De Boeck-Duculot.
Fløttum, K. and Rastier, F. (eds). 2003. Academic discourse. Multidisciplinary approaches. Oslo:
Novus.
Fløttum, K., Jonasson, K. and Norén, C. (forthcoming). ON – pronom à facettes.
Flyum, K.H. 2004. ”Mellom datapresisjon og tolkningspretensjon.” Akademisk Prosa 2: 29-47.
Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Fortanet, I. 2004. ”The use of ‘we’ in university lectures: reference and function.” English for
Specific Purposes 23: 45-66.
François, J. 1984. ”Analyse énonciative des équivalents allemands du pronom indéfini on.” In
Recherches en pragma-sémantique, G. Kleiber (ed.), 37-73. Paris: Klincksieck.
Fredrickson, K.M. and Swales, J.M. 1994. ”Competition and Discourse Community: introduc-
tions from Nysvenska studier.” In Text and Talk in Professional Contexts. Selected Papers
from the International Conference ”Discourse and the Professions”, Uppsala, 26−29 August,
1992, B.-L. Gunnarsson, P. Linell and B. Nordberg (eds), 9-21. Stockholm: ASLA.
Freedman, A. and Medway, P. (eds). 1994. Genre and the New Rhetoric. London: Taylor & Francis.
Fuertes-Olivera, P., Velasco-Sacristán, M., Arribas-Baño, A. and Samaniego-Fernández, E. 2001.
”Persuasion and advertising English: Metadiscourse in slogans and headlines.” Journal of
Pragmatics 33: 1291-1307.
Geertz, C. 1973.” Thick description: Toward an interpretive theory of culture.” In The interpreta-
tion of cultures, Chapter 1, 3-30. New York: Basic Books, Inc.
Gilhus, N.E. 2003. ”Medisinernes publisering; eksakt overføring av ny og nyttig informasjon
eller selvhevdende drøvtygging av gamle resultater?” Akademisk Prosa 1: 15-25. Skrifter fra
KIAP. Department of Romance studies, University of Bergen.
Gjesdal, A.M. 2003. L’emploi du pronom “on” dans les articles de recherche. Une étude diachro-
nique et qualitative. Master’s thesis. Department of Romance studies, University of Bergen.
 References 277

Gjesdal, A.M. 2005. ”Scientific writing and gender: the case of personal pronouns.” Akademisk
Prosa 3: 101-112. Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Grinde, L. 2003. Les choix lexicaux dans des articles de recherche français. Master’s thesis. Depart-
ment of Romance studies, University of Bergen.
Grossmann, F. 2002. ”Les modes de réference à autrui chez les experts: l’exemple de la revue.”
Langages. Faits de langue − Le discours rapporté, 255-262.
Gunnarsson, B.-L. 1994. Texter i Europeiska skrivsamhällen. En kontrastiv studie av svenska, en-
gelska och tyska textmönster. Forskningsprogram. FUMS, Uppsala Universitet.
Hall, E.T. and Hall, M.R. 1990. Understanding cultural differences. Yarmouth, ME: Intercultural Press.
Halliday, M.A.K. 1973. Explorations in the functions of language. London: Edward Arnold.
Halliday, M.A.K. 1994 [1985]. An introduction to functional grammar. London/Baltimore, MD:
Edward Arnold.
Harwood, N. 2005. ”‘We Do Not Seem to Have a Theory … The Theory I Present Here Attempts
to Fill This Gap’: Inclusive and Exclusive Pronouns in Academic Writing.” Applied Linguis-
tics 26 (3): 343-375.
Hawes, T. and Thomas, S. 1997. ”Tense Choice in Citations.” Research in the Teaching of English
31 (3): 393-414.
Hendriks, H. 2005a. ”The Structure of Learner Varieties: Introduction to the volume.” In The Struc-
ture of Learner Varieties, H. Hendriks (ed.), 1-17. Berlin/New York: Mouton de Gruyter.
Hendriks, H. 2005b. ”Structuring space in discourse: A comparison of Chinese, English, French
and German L1 and English, French and German L2 acquisition.” In The Structure of Learn-
er Varieties, H. Hendriks (ed.), 111-156. Berlin/New York: Mouton de Gruyter.
Hertzberg, F. 1995. ”Uttalte og uuttalte normer for vitenskapelig skriving.” In Virkelighetens for-
valtere, E.B. Johnsen (ed.), 187-205. Oslo: Universitetsforlaget.
Hinds, J. 1987. ”Reader versus writer responsibility: A new typology.” In Writing across lan-
guages: Analysis of L2 text, U. Connor and R. Kaplan (eds), 141-152. Reading, MA: Ad-
dison-Wesley.
Hofstede, G. 2001. Culture’s Consequences. Comparing Values, Behaviors, Institutions, and Organi-
zations Across Nations. 2nd ed. Thousand Oaks/ London: Sage.
Hunston, S. 1993. ”Evaluation and ideology in scientific writing.” In Register analysis: theory and
practice, M. Ghadessy (ed.), 57-73. London: Pinter.
Hyland, K. 1998a. Hedging in scientific research articles. Amsterdam/Philadelphia: John Ben-
jamins.
Hyland, K. 1998b. ”Persuasion and context: The pragmatics of academic metadiscourse.” Journal
of Pragmatics 30: 437-455.
Hyland, K. 1999. ”Talking to students: Metadiscourse in introductory textbooks.” English for
Specific Purposes 18 (1): 3-26.
Hyland, K. 2000. Disciplinary discourses: Social interactions in academic writing. Harlow, UK:
Longman.
Hyland, K. 2001a. ”Bringing in the reader. Addressee features in academic articles.” Written
Communication 18, 549-574.
Hyland, K. 2001b. ”Humble servants of the discipline? Self-mention in research articles.” English
for Specific Purposes 20: 207-226.
Hyland, K. 2002a. ”Authority and invisibility: authorial identity in academic writing.” Journal of
Pragmatics 34: 1091-1112.
Hyland, K. 2002b. ”Directives: Argument and Engagement in Academic Writing.” Applied Lin-
guistics 23 (2): 215-239.
278 Academic Voices

Hyland, K. 2005. Metadiscourse. Exploring interaction in writing. London/New York: Continuum.

Hyland, K. and Hamp-Lyons. L. 2002. ”EAP: issues and directions.” Journal of English for Aca-
demic Purposes 1: 1-12.
Ivanič, R. 1998. Writing and identity: The discoursal construction of identity in academic writing.
Amsterdam/Philadelphia: John Benjamins.
Johansson, S. 1998. ”On the role of corpora in cross-linguistic research.” In Corpora and Cross-
linguistic Research. Theory, Method, and Case Studies, S. Johansson and S. Oksefjell (eds),
3-24. Amsterdam/Atlanta, GA: Rodopi.
Johansson, S. 2002. ”Viewing language through multilingual corpora, with special reference
to the generic person in English, German and Norwegian.” In Studies in Contrastive Lin-
guistics, L.I. Rábade and S.M. Doval Suarez (eds), 515-554. Universidade de Santiago de
Compostela.
Johansson, S. and Oksefjell, S. (eds). 1998. Corpora and Cross-linguistic Research. Theory, Meth-
od, and Case Studies. Amsterdam/Atlanta, GA: Rodopi.
Jonasson, K. (forthcoming). Le pronom clitique on – un outil de non-partition référentielle
(Strasbourg 2003).
Kaplan, R. 1966. ”Cultural thought patterns in intercultural education.” Language Learning 16
(1): 1-20.
Kinn, T. 2004. ”Cognitive research agents in academic prose.” Akademisk Prosa 2: 137-149.
Skrifter fra KIAP. Department of Romance Studies, University of Bergen.
Kinn, T. 2005a. ”Plays of we-hood: What do we mean by we?” Akademisk Prosa 3: 129-142.
Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Kinn, T. 2005b. ”Denne artikkelen analyserer … Den tenkjande forskaren i norske forskingsar-
tiklar.” In MONS 10. Utvalde artiklar frå det tiande Møte om norsk språk i Kristiansand
2003, S. Lie, G. Nedrelid and H. Omdal (eds), 183-194. Kristiansand: Høyskoleforlaget.
Kinn, T. 2005c. ”Tilbod og innbydingar: imperativ med la i forskingsartiklar.” Norsk Lingvistisk
Tidsskrift 23: 145-175.
Kirkeeide, N. 2003. ”Uten nyheter, ingen fremdrift.” Akademisk Prosa 1: 65-67. Skrifter fra KIAP.
Department of Romance studies, University of Bergen.
Kjeldsen, J.E. 2004. Retorikk i vår tid. En innføring i moderne retorisk teori. Oslo: Spartacus.
Kjørup, S. 1996. Menneskevidenskaberne. Problemer og traditioner i humanioras videnskabsteori.
Roskilde Universitetsforlag.
Koskela, M. 1999. ”Om referenser till vetenskaplig litteratur i svenska filosofiske texter.” Nordica
Bergensia 20: 143-160.
Koskela, M. 2000. Svenska som filosofins språk. Tutkimuksia 235. Språkvetenskap 38. Vaasa.
Koutsantoni, D. 2004. ”Attitude, certainty and allusions to common knowledge in scientific re-
search articles.” Journal of English for Academic Purposes 3: 163-182.
Koutsantoni, D. 2005. ”Certainty across cultures: a comparison of the degree of certainty ex-
pressed by Greek and English speaking scientific authors.” Intercultural Pragmatics 2 (2):
121-149.
Krzeszowski, T.P. 1990. Contrasting Languages. The Scope of Contrastive Linguistics. Berlin/New
York: Mouton de Gruyter.
Kuo, C.-H. 1999. ”The use of personal pronouns: Role relationships in scientific journal articles.”
English for Specific Purposes 18: 121-138.
Kyvik, S. and Larsen, I.M. 1997. ”The exchange of knowledge: A small country in the international
research community.” Science Communication 18 (3): 238-264.
Lakoff, R. 1975. Language and Woman’s Place. New York: Harper and Row.
 References 279

Langacker, R.W. 1987. Foundations of Cognitive Grammar, vol. 1, Theoretical Prerequisites. Stan-
ford, CA.: Stanford University Press.
Larsen, R.J. 1984. Quelques regards sur le pronom ‘on’. Master’s thesis. University of Oslo.
Laurén, C. and Nordman, M. 1998. ”Om en granskning av två sociologers vetenskapliga idi-
olekter.” In Nation och individ i fackspråk. Forskning i Norden, C. Laurén and M. Nordman
(eds), 105-116. Vaasan yliopiston julkaisuja. Selvityksiä ja raportteja 34. Vaasa.
Ledin, P. 1999. Texter och textslag – en teoretisk diskussion. Lund: Institutionen för nordiska
språk. Rapporter från projektet Svensk sakprosa 27.
Levinson, S.C. 1983. Pragmatics. Cambridge: Cambridge University Press.
Lindeberg, A.-C. 1994. ”Rhetorical conventions in the discussion/conclusion sections of re-
search articles in finance, management and marketing.” In Applications and implications of
current LSP research. (LSP 1993), M. Brekke et al. (eds), 761-779. Bergen: Fagbokforlaget.
Lindeberg, A.-C. 1998. ”Promotional rhetorical steps and linguistic signalling in research articles
in three disciplines.” In LSP identity and interface: Research, knowledge and society. Proceed-
ings of the 11th European Symposium on Language for Special Purposes, Copenhagen, August
1997, L. Lundquist, H. Picht and J. Qvistgaard (eds), 689−698. Copenhagen: Copenhagen
Business School.
Lindeberg, A.-C. 2004. Promotion and politeness. Conflicting scholarly rhetoric in three disci-
plines. Åbo: Åbo Akademi University Press.
Loffler-Laurian, A.-M. 1980. ”L’expression du locuteur dans les discours scientifiques. “JE”,
“NOUS” et “ON” dans quelques textes de chimie et de physique.” Revue de linguistique
romane 44: 135-157.
Lundeby, E. 1996. ”Om ‘man’ og ‘en’ og ‘du’.” In Språket lever! Festskrift till Margareta Westman
den 27 mars 1996. Skrifter utgivna av Svenska språknämnden 80, 195-201. Stockholm:
Norstedts.
Lundquist, L. 2003. ”Interpretation of culture and culture of interpretation in law and linguis-
tics.” In Language and Culture, I . Baron (ed.), 111-124. Copenhagen Studies in Language
29, Samfundslitteratur.
Lyons, J. 1977. Semantics. Cambridge: Cambridge University Press.
MacDonald, S.P. 1989. ”Data-Driven and Conceptually Driven Academic Discourse.” Written
Communication 6 (4): 411-435.
Maingueneau, D. 1993. 3rd ed. Éléments de linguistique pour le texte littéraire. Paris: Dunod.
Malrieu, D. and Rastier, F. 2001. ”Genres et variations morphosyntaxiques.” Traitements au-
tomatiques du langage 42 (2): 547-577.
Marnette, S. 2001. ”The French théorie de l’énonciation and the study of speech and thought
presentation.” Language and Literature 10 (3): 243-262.
Master, P. 2001. ”Active verbs with inanimate subjects in scientific research articles.” In Academic
writing in context. Implications and applications. Papers in honour of Tony Dudley-Evans, M.
Hewings (ed.), 169-181. Birmingham: University of Birmingham Press.
Mauranen, A. 1993a. Cultural Differences in Academic Rhetoric. Frankfurt am Main: Peter Lang.
Mauranen, A. 1993b. ”Contrastive ESP rhetoric: Metatext in Finnish-English economics texts.”
English for Specific Purposes 12: 3-22.
Mauranen, A. 2001a. ”Descriptions or explanations? Some methodological issues in contrastive
rhetoric.” In Academic writing in context. Implications and applications. Papers in honour of
Tony Dudley-Evans, M. Hewings (ed.), 43-54. Birmingham: University of Birmingham Press.
280 Academic Voices

Mauranen, A. 2001b. ”Reflexive academic talk: Observations from MICASE.” In Corpus linguis-
tics in North-America: Selections from the 1999 Symposium, R.C. Simpson and J.M. Swales
(eds), 165-178. Ann Arbor, MI: University of Michigan Press.
Mauranen, A. 2002. ”One thing I’d like to clarify…”. Observations of academic speaking. http://
www.eng.helsinki.fi/hes/Corpora/one_thing.htm (site visited 9 May 2003).
Mauranen, A. Forthcoming. Spoken rhetoric − how do natives and non-natives fare?
McEnery, T. and Wilson, A. 2001. Corpus Linguistics. 2nd ed. Edinburgh: Edinburgh University
Press.
Melander, B. 1998. ”Culture or genre? Issues in the interpretation of cross-cultural differen-
ces in scientific papers.” In Genre Studies in English for Academic Purposes, I. Fortanet, S.
Posteguillo, J.C. Palmer and J.F. Coll (eds), 211-226. (Collegió ”Summa”. Sèrie Filologia 9.)
Universitat Jaume I: Castello de la Plana.
Melander, B., Swales, J.M. and Fredrickson, K. 1997. ”Journal abstracts from three academic
fields in the United States and Sweden: national or disciplinary proclivities?” In Culture
and styles of academic discourse, A. Duszak (ed.), 251-272. Berlin/New York: Mouton de
Gruyter.
Miller, C.R. 1994 [1984]. ”Genre as Social Action.” In Genre and the New Rhetoric, A. Freedman
and P. Medway (eds), 23-42. London: Taylor & Francis. (Revised version in Quarterly Jour-
nal of Speech 1984, 70: 151-167.)
Montgomery, S. 2004. ”Of Towers, Walls, and Fields: Perspectives on Language in Science.” Sci-
ence 303: 1333-1335.
Moreno, A. 1997. ”Genre constraints across languages: Causal metatext in Spanish and English
RAs.” English for Specific Purposes 16 (3): 161-179.
Mühlhäusler, P. and Harré, R. 1990. Pronouns and People: The Linguistic Construction of Social
and Personal Identity. Oxford: Blackwell.
Muller, C. 1979. ”Sur les emplois personnels de l’indéfini on.” Langue française et linguistique
quantitative. Genève: Slatkine, 65-72.
Myers, G. 1989. ”The pragmatics of politeness in scientific articles.” Applied Linguistics 10: 1-35.
Nølke, H. 1991. ”Contrastive pragmatic linguistics.” In Contrastive linguistics: Papers from the
CL-symposion, 28–30 August 1989, the Aarhus School of Business, K.M. Lauridsen and O.
Lauridsen (eds), 199-231. Aarhus: the Aarhus School of Business.
Nølke, H. 1993a. ”Formes et emplois des énoncés négatifs : polyphonie et syntaxe de ”ne...pas”.”
Le regard du locuteur. Paris: Kimé, 215-232.
Nølke, H. 1993b,” « NE...PAS : négation descriptive ou polémique ? Contraintes formelles sur son
interprétation ».” Le regard du locuteur, Paris : Kimé, 233-258.
Nølke, H. 1994. Linguistique modulaire: de la forme au sens. Louvain/Paris: Peeters.
Nølke, H., Fløttum, K. and Norén, C. 2004. ScaPoLine. La théorie scandinave de la polyphonie
linguistique. Paris: Kimé.
Norén, C. 2004a. ”‘On dit qu’on est speed’. L’emploi du pronom ON dans le français parlé.” In
Sept approches à un corpus, H.-L. Andersen and Ch. Thomsen (eds), 87-105. Frankfurt/
Bern: Peter Lang.
Norén, C. 2004b. ”Argumentation par autorité et autorité polyphonique.” In H. Nølke et al.,
129-146.
Oakes, M.P. 1998. Statistics for Corpus Linguistics. Edinburgh: Edinburgh University Press.
O’Barr, W. and Atkins, B.K. 1980. ”‘Women’s language’ or ‘powerless language’?” In Women and
Language in Literature and Society, S. McConnell-Ginet, R. Borker and N. Furman (eds),
93-110. New York: Praeger.
 References 281

Perrin, L. 2003. ”Citation, opacité, point de vue.” Polyphonie – linguistique et littéraire, 63-93.
Perrin, L. 2004a. ”Polyphonie et autres formes d’hétérogénéité énonciative: Bakhtine, Bally,
Ducrot, etc.” Pratiques 123/124: 7-26.
Perrin, L. 2004b. ”Le discours rapporté modal.” In Le discours rapporté dans tous ses états, J.M.
Lopez-Munoz et al. (eds), 64-74. Paris: L’Harmattan.
Perrin, L. 2005 “Polyphonie et séquence écho.” In J. Bres et al. (eds). Dialogisme, polyphonie: ap-
proches linguistiques. De Boeck-Duculot, 173-186.
Perrin, L. (forthcoming). Le rôle de pour et de selon relativement aux fonctions des séquences
échos dans la presse écrite.
Pisanski Peterlin, A. 2005. ”Text-organising metatext in research articles: an English-Slovene
contrastive analysis.” English for Specific Purposes 24: 307-319.
Poudat, C. 2003. ”Characterisation of French linguistic research articles using morphosyntactic
variables.” In Academic discourse. Multidisciplinary approaches, K. Fløttum and F. Rastier
(eds), 77-96. Oslo: Novus.
Poudat, C. 2004. ”Evaluation des manifestation pronominales de l’auteur dans l’article de revue
linguistique en français et en anglais.” Akademisk Prosa 2: 69-86. Skrifter fra KIAP. Depart-
ment of Romance studies, University of Bergen.
Poudat, C. 2006. Etude contrastive de l’article scientifique dans une perspective d’analyse des gen-
res. PhD thesis. University of Orléans.
Prelli, L.J. 1989. A Rhetoric of Science: Inventing Scientific Discourse. University of South Carolina
Press.
Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1972. A Grammar of Contemporary English.
London: Longman.
Rabatel, A. 2001. ”La valeur de « on » pronom indéfini/pronom personnel dans les perceptions
représentées.” L’information grammaticale 88: 28-32.
Rastall, P. 2003. ”What do we mean by we?” English Today 19: 50-53.
Rastier, F. 1987. Sémantique interpretative. Paris: Presses Universitaires de France.
Rastier, F. 2001. Arts et sciences du texte. Paris: Presses Universitaires de France.
Redding, G. and Stening, B.W. (eds). 2003. Cross-cultural management. Vol. I. The Theory of
culture. An Elgar reference Collection, Cheltenham, UK/Northampton, MA.
Rey-Debove, J. 2001. ”De on à je vers le nom propre : des pronoms personnels en français.” In
Quitte ou double sens, P. Bogaards et al. (eds), 279-304. Amsterdam/New York: Rodopi.
Riegel, M., Pellat, J.-C. and Rioul, R. 1994. Grammaire méthodique. Paris: Presses Universitaires
de France.
Rinck, F. 2005. ”Images of scientific activity through the research article: a comparison between
linguistics and literary studies.” Akademisk Prosa 3: 75-86. Skrifter fra KIAP. Department of
Romance studies, University of Bergen.
Rinck, F. Forthcoming. L’article de recherche comme genre ? Aspects linguistiques et socio-dis-
cursifs de l’écriture d’articles de recherche: identité disciplinaire et singularité des scripteurs
doctorants. Ph.D. thesis.
Rosier, L. 1999. Le discours rapporté. Bruxelles: Duculot.
Rounds, P.L. 1987. ”Multifunctional personal pronoun use in an educational setting.” English for
Specific Purposes 6: 13-29.
Rowley-Jolivet, E. 2002. ”Science in the making: Scientific conference presentations and the
construction of facts.” In The language of conferencing, E. Ventola, C. Shalom and S. Thomp-
son (eds), 95-125. Frankfurt: Peter Lang.
282 Academic Voices

Ruiying, Y. and Allison, D. 2003. ”Research articles in applied linguistics: moving from results to
conclusions.” English for Specific Purposes 22 (4): 365-385.
Salager-Meyer, F. 1998. ”Reference patterns in medical English discourse.” In LSP identity and
interface. Research, knowledge and society. Proceedings of the 11th European symposium on
LSP, Copenhagen, August 1997, L. Lundquist, H. Picht and J. Qvistgaard (eds), 495-504.
Copenhagen: Copenhagen Business School.
Salager-Meyer, F. 1999. ”Referential behaviour in scientific writing: a diachronic study (1810-
1995).” English for Specific Purposes 18 (3): 279-305.
Salager-Meyer, F. and Defives, G. 1998. ”From the gentleman’s courtesy to the expert’s caution: A
diachronic analysis of hedges in academic writing (1810-1995).” In Genre Studies in English
for Academic Purposes, I. Fortanet, S. Posteguillo, J.C. Palmer and J.F. Coll (eds), 133-171.
(Collegió ”Summa”. Sèrie Filologia 9.) Universitat Jaume I: Castello de la Plana.
Salager-Meyer, F., Alcaraz Ariza, M.A. and Zambrano, N. 2003. ”The scimitar, the dagger and
the glove: intercultural differences in the rhetoric of criticism in Spanish, French and Eng-
lish medical discourse (1930−1995).” English for Specific Purposes 22: 223-247.
Salager-Meyer, F. and Alcaraz Ariza, M.A. 2004. ”Negative Appraisals in Academic Book Re-
views: A Cross-linguistic Approach.” In Intercultural Aspects of Specialized Communication,
C.N. Candlin and M. Gotti (eds), 150-172. Bern: Peter Lang.
Schiffrin, D. 1994. Approaches to Discourse. Oxford: Blackwell.
Shaw, P. 1992. ”Reasons for the correlation of voice, tense and sentence function in reporting
verbs.” Applied Linguistics 13: 302-319.
Shaw, P. 2003. ”Evaluation and promotion across languages.” Journal of English for Academic
Purposes 2: 343-357.
Shaw, P. and Vassileva, I. Forthcoming. Co-evolving academic rhetoric across culture; Britain,
Bulgaria, Denmark, Germany in the 20th century.
Siegel, S. and Castellan Jr., N.J. 1988. Nonparametric Statistics for the Behavioral Sciences. New
York: McGraw-Hill.
Sinclair, J. 1987. Mirror for a Text. Manuscript (published 1988, Journal of English and Foreign
Languages, Hyderabad, India).
Sokal, R.R. and Rohlf, F.J. 1995. Biometry. The principles and practice of statistics in biological
research. 3rd ed. New York: Freeman.
Stålhammar, M. 2002. ”Abstracts of doctoral dissertations in the humanities and social sciences.”
In Porta Scientiae I. Lingua specialis, M. Koskela, C. Laurén, M. Nordman and N. Pilke
(eds), 264-275. University of Vaasa.
Stotesbury, H. 2003. ”Evaluation in research article abstracts in the narrative and hard sciences.”
Journal of English for Academic Purposes 2 (4): 327-341.
Stutterheim, C. von and Lambert, M. 2005. ”Cross-linguistic analysis of temporal perspectives
in text production.” In The Structure of Learner Varieties, H. Hendriks (ed.), 203-230. Ber-
lin/New York: Mouton de Gruyter.
Swales, J.M. 1986. ”Citation Analysis and Discourse Analysis.” Applied Linguistics 7 (1): 39-56.
Swales, J.M. 1990. Genre analysis. English in academic and research settings. Cambridge: Cam-
bridge University Press.
Swales, J.M. 2001. ”Metatalk in American academic talk: The cases of ‘point’ and ‘thing’.” Journal
of English Linguistics 29: 34-54.
Swales, J.M. 2004. Research genres. Exploration and applications. Cambridge: Cambridge Uni-
versity Press.
 References 283

Swales, J.M., Ahmad, U.K., Chang, Y-Y., Chavez, D., Dressen, D.F. and Seymour, R. 1998. ”Con-
sider This: The Role of Imperatives in Scholarly Writing.” Applied Linguistics 19 (1): 97-121.
Taylor, G. and Tingguan, C. 1991. ”Linguistic, cultural and subcultural issues in contrastive
discourse analysis: Anglo-American and Chinese scientific texts.” Applied Linguistics 12:
319-226.
Thompson, G. 1996. ”Voices in the text: discourse perspectives on language reports.” Applied
Linguistics 17 (4): 501-530.
Thompson, S.E. 2003. ”Text-structuring metadiscourse, intonation and the signalling of organi-
sation in academic lectures.” Journal of English for Academic Purposes 2: 5-20.
Tønnesson, J.L. 2001 [1998]. Vitenskapens stemmer. Vitenskapsbilder, dialogisme og forsk-
ernærvær i fire historiefaglige tekster for allmennheten. Oslo: INL, University of Oslo.
Tønnesson, J.L. 2003a. ”Model readers and other textual strategies in Norwegian historian
Professor Finn Olstad’s paper ”New perspectives on the rise and fall of the working class”
(2000).” In Academic discourse. Multidisciplinary approaches, K. Fløttum and F. Rastier
(eds), 158-180. Oslo: Novus.
Tønnesson, J.L. 2003b. Tekst som partitur eller Historievitenskap som kommunikasjon. Oslo: Uni-
versity of Oslo (doctoral dissertation).
Tønnesson, J.L. 2004. Tekst som partitur, eller Historievitenskap som kommunikasjon: nærlesning
av fire historietekster skrevet for ulike lesergrupper. Acta humaniora (Oslo) – no. 207. Oslo:
University of Oslo.
Triandis, H.C. 1994. Culture and social behavior. New York: McGraw-Hill.
Tuomarla, U. 1999. La citation mode d’emploi. Sur le fonctionnement discursif du discours rapporté
direct. Academia Scientarium Fennica, Helsinki. Humaniora 308.
Valero-Garcés, C. 1996. ”Contrastive ESP rhetoric: Metatext in Spanish-English economics
texts.” English for Specific Purposes 15 (4): 279-294.
Valle, E. 1999. A collective intelligence: The life sciences in the Royal Society as a scientific dis-
course community, 1665-1965. Anglicana Turkuensia no 17. Turku: University of Turku.
Vande Kopple, W. 1985. ”Some exploratory discourse on metadiscourse.” College Composition
and Communication 36 (1): 82-93.
Vassileva, I. 1997. ”Hedging in English and Bulgarian academic writing.” In Culture and styles of
academic discourse, A. Duszak (ed.), 203-222. Berlin/New York: Mouton de Gruyter.
Vassileva, I. 2000. Who is the author? A contrastive analysis of authorial presence in English, Ger-
man, French, Russian and Bulgarian academic discourse. Sankt Augustin: Asgard Verlag.
Vassileva, I. 2001. ”Commitment and detachment in English and Bulgarian academic writing.”
English for Specific Purposes 20 (1): 83-102.
Ventola, E. 1994. ”Finnish writers’ academic English. Problems with reference and theme.” Func-
tions of Language 1 (2): 26-293.
Ventola, E. and Mauranen, A. (eds). 1996. Academic writing: Intercultural and textual issues.
Amsterdam/Philadelphia: John Benjamins.
Viollet, C. 1988. ”Mais qui est on? Etudé linguistique des valeurs de on dans un corpus oral.”
LINX (Paris X Nanterre) 18: 67-75.
Vold, E.T. 2005. ”Expressions of uncertainty in scientific research articles.” Akademisk Prosa 3,
113-127. Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Wales, K. 1980. ”Exophora re-examined: The uses of the personal pronoun ‘WE’ in present-day
English.” UEA Papers in Linguistics 12: 21-44.
Wales, K. 1996. Personal pronouns in present-day English. Cambridge: Cambridge University Press.
Wilmet, M. 1997. Grammaire critique du Français. Louvain-la-Neuve: Duculot.
 KIAP publications

As at March 2006

Andersen, P.L.W. 2003. Le titre de l’article de recherche. Master’s thesis. Department of Romance
studies, University of Bergen.
Breivega, K.R., Dahl, T. and Fløttum, K. 2002. ”Traces of self and others in research articles. A
comparative pilot study of English, French and Norwegian research articles in medicine,
economics and linguistics.” International Journal of Applied Linguistics 12 (2): 218-239.
Dahl, T. 2002. ”Author abstracts and computer-generated summaries: similarities and differ-
ences.” In Porta Scientiae II. Lingua specialis, M. Koskela et al. (eds), 546-556. Proceedings
of the University of Vaasa. Reports 96. Vaasa.
Dahl, T. 2003. ”Metadiscourse in research articles.” In Academic discourse. Multidisciplinary ap-
proaches, K. Fløttum and F. Rastier (eds), 120-138. Oslo: Novus.
Dahl, T. 2004a. ”Absent doctors, shy economists and polemic linguists? Writer manifestation in
academic texts.” Synaps 14: 1-4. Department of professional and intercultural communica-
tion, Norwegian School of Economics and Business Administration (NHH).
Dahl, T. 2004b. ”Some characteristics of argumentative abstracts.” Akademisk Prosa 2: 49-67.
Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Dahl, T. 2004c. ”Textual metadiscourse in research articles: a marker of national culture or of
academic discipline?” Journal of Pragmatics 36 (10): 1807-1825.
Dahl, T. 2004d. ”Accommodating the reader: metatext in research articles. A report from the
KIAP project.” In New Directions in LSP studies. Proceedings of the 14th European Symposium
on Language for Special Purposes. 18–22 August 2003, Communication, Culture, Knowledge,
K. Ahmad and M. Rogers (eds), 167-171 (171-178 in pdf-document). Guildford: University
of Surrey (UniS). ISBN: 1-84469-007-5.
Dahl, T. 2005a. ”Cultural identity in academic texts.” Akademisk Prosa 3: 35-45. Skrifter fra KIAP.
Department of Romance studies, University of Bergen.
Dahl, T. 2005b. ”The academic author: traces in text.” In Text i arbete/Text at work. Essays in
honour of Britt-Louise Gunnarsson 12 January 2005, I. Bäcklund, U. Börestam, U. Melander
Marttala and H. Näslund (eds), 289-297. Uppsala Universitet.
Dahl, T. Forthcoming. ”Author identity in economics and linguistics abstracts.” In Cross-cultural
and cross-linguistic perspectives on academic discourse, E. Suomela-Salmi (ed.). University
of Turku.
Didriksen, A.A. 2004a. ”Val av artiklar ved hjelp av kvantitative metodar.” Akademisk Prosa 2:
99-115. Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Didriksen, A.A. 2004b. ”Konnektorar i franske vitskaplege artiklar.” TRIBUNE 15: 31-38.
Didriksen, A.A. 2004c. DONC et autres connecteurs argumentatifs dans les articles de recherche
français. Master’s thesis. Department of Romance studies, University of Bergen.
286 Academic Voices

Didriksen, A.A. and Gjesdal, A.M. Forthcoming. ”Genre Constraints and Individual Linguis-
tic Variation.” In Cross-cultural and cross-linguistic perspectives on academic discourse, E.
Suomela-Salmi (ed.). University of Turku.
Fløttum, K. 2001. ”Kulturelle likheter og forskjeller i akademisk prosa.” TRIBUNE 12: 27-34.
Fløttum, K. 2002. Corpus description and methodological design. http://kiap.aksis.uib.no/om_
prosjektet.htm
Fløttum, K. 2003a. ”Y a-t-il des identités culturelles dans le discourse scientifique ?” Estudios de
lengua y literatura francesas 14: 83-93.
Fløttum, K. 2003b. ”Personal English, indefinite French and plural Norwegian scientific au-
thors? Pronominal author manifestation in research articles.” Norsk Lingvistisk Tidsskrift
21 (1): 21-55.
Fløttum, K. 2003c. ”Om forfatternærvær i vitenskapelige artikler.” Rapport fra Nasjonalt fagråd i
romanske språk og litteraturer 2002, 30-47. Tromsø: Universitetet i Tromsø.
Fløttum, K. 2003d. ”Polyphonie dans les textes scientifiques. Etude de deux cas français.” In
Polyphonie – linguistique et littéraire, VII, M. Olsen (ed.), 111-129. Roskilde: Samfundslit-
teratur Roskilde.
Fløttum, K. 2003e. ”Bibliographical references and polyphony in research articles.” In Academic
discourse. Multidisciplinary approaches, K. Fløttum and F. Rastier (eds), 97-119. Oslo: Novus.
Fløttum, K. 2003f. Forskerne sier ... Kronikk i Bergens Tidende 3 April, 2003. http://www.bt.no/
meninger/kronikk/article149215
Fløttum, K. 2003g. ”« JE » et le verbe.” TRIBUNE 14, 7–14.
Fløttum, K. 2003h. Forskning og formidling. Kronikk i Bergens Tidende 17 September 2003
(Forskningsdagene 2003). http://www.bt.no/meninger/article193379.
Fløttum, K. 2003i. ”The French pronoun ‘on’ in academic discourse – indefinite versus per-
sonal.” In Proceedings of CIL17, E. Hajičová, A. Kotěšovcová. and J. Mírovský (eds.), CD-
ROM. (Paper marked: S10_KjerstiFlottum). Matfyzpress, MFF UK, ISBN: 80-86732- 215.
Fløttum, K. 2004a. ”Enonsiativ teori – et perspektiv for studiet av vitenskapelige tekster.” Akadem-
isk Prosa 2: 7-28. Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Fløttum, K. 2004b. ”Thèmes, topiques et marqueurs de cadres discursifs dans les articles scienti-
fiques français : la présence de personne(s).” Scolia (Strasbourg) 18: 167-201.
Fløttum, K. 2004c. ”Om fagtradisjoner, posisjonering og identitet i vitenskapelig skriving.” In
Språk i kunnskapssamfunnet. Engelsk – elitenes nye latin eller uunnværlig redskap for fri
kunnskapsutvikling?, D.F. Simonsen (ed.), 185-191. Oslo: Gyldendal.
Fløttum, K. 2004d. ”La présence de l’auteur dans les articles scientifiques : étude des pronoms je, nous
et on.” In Structures et discours, A. Auchlin et al. (eds), 401-416. Québec: Ed. Nota Bene.
Fløttum, K. 2004e. The expression of “results” in scientific discourse – a cross-linguistic and
cross-disciplinary analysis. Paper given at Societas Linguistica Europaea 37th International
Meeting, Kristiansand 29 July–1 August, 2004.
Fløttum, K. 2004f. ”Polyfonisk interaksjon via IKKE i vitenskapelig diskurs.” Rhetorica Scandi-
navica 3: 23-40.
Fløttum, K. 2004g. ”Den franske vitenskapelige forfatter – je, nous eller on ?” TRIBUNE 15:
56-69.
Fløttum, K. 2004h. ”Om formulering av forskningsresultater i vitenskapelige artikler.” In Kommu-
nikasjon, E. Lambertsson Björk and S. Vesterhus (eds), 61-76. Halden: Høgskolen i Østfold.
Fløttum, K. 2004i. ”Traces of others in research articles: the citation cluster.” In New Directions
in LSP studies. Proceedings of the 14th European Symposium on Language for Special Pur-
poses. 18–22 August 2003, Communication, Culture, Knowledge, K. Ahmad and M. Rogers
 KIAP publications 287

(eds), 153-159 (160-166 in pdf-document). Guildford: University of Surrey (UniS). ISBN:

1-84469-007-5. http://portal.surrey.ac.uk/pls/portal/docs/PAGE/COMPUTING/NEWS/
LSP%202003/LSP%20PROCEEDINGS_1.PDF.
Fløttum, K. 2005a. ”Individual linguistic variation and possibilities for academic self-hood?”
Akademisk Prosa 3: 87-100. Skrifter fra KIAP. Department of Romance studies, University
of Bergen.
Fløttum, K. 2005b. ”The self and the others – polyphonic visibility in research articles.” Interna-
tional Journal of Applied Linguistics 15 (1): 29-44.
Fløttum, K. 2005c. ”MOI et AUTRUI dans le discours scientifique : l’exemple de la négation
NE...PAS.” In Dialogisme, polyphonie: approches linguistiques, J. Bres, P.P. Haillet, S. Mellet,
H. Nølke and L. Rosier (eds), De Boeck-Duculot, 323-337.
Fløttum, K. 2005d. ”Om nekting i et polyfonisk og vitenskapsretorisk perspektiv.” Sproglig poly-
foni, Arbeidspapirer 2: 45-64. Roskilde: Roskilde Universitetscenter.
Fløttum, K. 2005e. ”Lingvistiske og diskursive studier i KIAP-korpuset. Om utvikling og bruk
av et flerspråklig og flerdisiplinært korpus av vitenskapelige artikler.” In Korpus i forskning
og undervisning. Hvor står vi i Norge?, 25-51. Kristiansand: Høgskolen i Agder. ISBN: 82-
7117-572-6.
Fløttum, K. 2006. ”En språklig analyse av norsk medisinsk prosa.” Tidsskrift for Den norske læge-
forening 126 (1): 65-68.
Fløttum, K. Forthcoming a. ”Interrelation de voix internes et externes dans le discours.” In Di-
alogisme et polyphoni en langue et en discours, L. Perrin (ed.). Metz: Recherches Linguis-
tiques. 2006.
Fløttum, K. Forthcoming b. ”The typical research article – does it exist?” In Cross-cultural and
cross-linguistic perspectives on academic discourse, E. Suomela-Salmi (ed.). University of
Turku.
Fløttum, K. Forthcoming c.” Polyphonic constructions as epistemic and evaluative qualifica-
tions in research articles.” In Evaluation in academic discourse, G. Del Lungo and E. Tognini
Bonelli (eds).
Fløttum, K. and Breivega K.R. 2002. ”Cultural identity in academic prose: national versus disci-
pline-specific.” In Porta Scientiae II. Lingua specialis. Proceedings of the University of Vaasa,
M. Koskela et al. (eds), 533-545. Reports 96. Vaasa.
Fløttum, K. and Rastier, F. (eds). 2003. Academic discourse. Multidisciplinary approaches. Oslo:
Novus.
Fløttum, K., Kinn T. and Dahl, T. Forthcoming. ”“We now report on ...” versus ”Let us now see
how ...”: Author roles and interaction with readers in research articles.” K. Hyland and M.
Bondi (eds). Academic Discourse across Disciplines. Bern/Frankfurt: Peter Lang.
Gjesdal, A.M. 2003. L’emploi du pronom “on” dans les articles de recherche. Une etude diachronique et
qualitative. Master’s thesis. Department of Romance studies, University of Bergen.
Gjesdal, A.M. 2004a. ”Om bruken av on i vitenskapelige artikler.” Akademisk Prosa 2: 87-97. Skrift-
er fra KIAP. Department of Romance studies, University of Bergen.
Gjesdal, A.M. 2004b. ”Bruken av pronomenet ”on” i franske, lingvistiske artikler. En diakronisk
studie 1980–2000.” TRIBUNE 15: 121-128.
Gjesdal, A.M. 2005. ”Scientific writing and gender: the case of personal pronouns.” Akademisk
Prosa 3: 101-112. Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Grinde, L. 2003. Les choix lexicaux dans des articles de recherche français. Master’s thesis. Depart-
ment of Romance studies, University of Bergen.
288 Academic Voices

Kinn, T. 2004. ”Cognitive research agents in academic prose.” Akademisk Prosa 2: 137-149.
Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Kinn, T. 2005a. ”Plays of we-hood: What do we mean by we?” Akademisk Prosa 3: 129-142.
Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Kinn, T. 2005b. ”Denne artikkelen analyserer … Den tenkjande forskaren i norske forskings-
artiklar.” In MONS 10. Utvalde artiklar frå det tiande Møte om norsk språk i Kristiansand
2003, S. Lie, G. Nedrelid and H. Omdal (eds), 183-194. Kristiansand: Høyskoleforlaget.
Kinn, T. 2005c. ”Tilbod og innbydingar: imperativ med la i forskingsartiklar.” Norsk Lingvistisk
Tidsskrift 23: 145-175.
Skiple, J. 2003. L’emploi de la négation dans les articles scientifiques français. Master’s thesis. De-
partment of Romance studies, University of Bergen.
Vold, E.T. 2004a. ”Vers une conception philosophico-logique de la modalité en linguistique.”
Akademisk Prosa 2: 117-135. Skrifter fra KIAP. Department of Romance studies, University
of Bergen.
Vold, E.T. 2004b. Comment sélectionner ses observables dans l’analyse d’une catégorie séman-
tique ? Paper given at Young Researchers’ Conference, 29–30 April, 2004. Paris-X, Nanterre.
Vold, E.T. 2004c. ”Epistemisk pouvoir/kunne i vitenskapelige artikler.” TRIBUNE 15: 229-240.
Vold, E.T. 2005. ”Expressions of uncertainty in scientific research articles.” Akademisk Prosa 3:
113-127. Skrifter fra KIAP. Department of Romance studies, University of Bergen.
Vold, E.T. 2006. ”Epistemic modality markers in research articles: a cross-linguistic and cross-
disciplinary study.” International Journal of Applied Linguistics 16 (1): 61-87.
Vold, E.T. Forthcoming. ” The choice and use of epistemic modality markers in linguistics a n d
medical research articles.” K. Hyland and M. Bondi (eds). Academic Discourse across Disci-
plines. Bern/Frankfurt: Peter Lang.

KIAP working papers, Department of Romance studies, University of Bergen:

Akademisk prosa 1. Skrifter fra KIAP: June 2003

Akademisk prosa 2. Skrifter fra KIAP: April 2004
Akademisk prosa 3. Skrifter fra KIAP: February 2005
 Appendix A The KIAP Corpus

Table A-1. The KIAP Corpus

Journals No. of articles Publication years
English economics 50
European Economic Review 18 2001–2003
Journal of Economic Behavior & Organization 5 2000–2002
Journal of International Economics 27 2000–2003
English linguistics 50
Computers and the Humanities 2 2003
English for Specific Purposes 8 2001–2002
Journal of Linguistics 10 1998–2001
Journal of Pragmatics 4 2001–2002
Language 4 2001
Language Sciences 11 2001–2002
Lingua 3 2002
Linguistic Inquiry 1 2000
Linguistics 1 2002
Natural Language & Linguistic Theory 2 2001–2002
Natural Language Semantics 4 2001–2002
English medicine 50
Breast Cancer Research 6 2002–2003
British Journal of Ophthalmology 7 2001
British Medical Journal 5 2000–2001
International Journal of Cancer 5 1999, 2001
International Journal of Oncology 6 2002
Journal of Clinical Endocrinology & Metabolism 6 2003
Journal of the American Medical Association 15 2001–2002
French economics 50
Annales d’Économie et de Statistique 34 1998–2001
Économie Appliquée 7 2000–2001
Revue Économique 9 2000–2001
290 Academic Voices

Journals No. of articles Publication years

French linguistics 50
Langue Française 13 2001–2002
Marges Linguistiques 2 2001
Revue de Sémantique et Pragmatique 17 1999–2001
Travaux de Linguistique 18 2001–2002
French medicine 50
Annales de Médecine Interne 30 1999–2003
Bulletin de la Société Belge d’Ophtalmologie 15 1999–2003
Maladies Chroniques au Canada 5 1995, 1997–1999
Norwegian economics 50
Norsk Økonomisk Tidsskrift 50 1993–2003
Norwegian linguistics 50
Maal og Minne 6 1998–2001
Norsk Lingvistisk Tidsskrift 44 1992–2002
Norwegian medicine 50
Tidsskrift for Den norske Lægeforening 50 2000–2003
 Appendix B Statistical tests

In this appendix, we present the results of four sets of statistical analyses of the
relative frequency of six linguistic features in the KIAP corpus. Each feature is
subjected to the same battery of analyses. The features are (see section 1.1 for an
overview and sections 4.1, 5.1 and 6.1 for details):
1. first person subjects
2. indefinite pronoun subjects
3. metatext
4. negation
5. adversative conjunctions
6. bibliographical references

The tests are the following, all non-parametric ones based on ranks:
1. two-way analysis of variance (using the Scheirer–Ray–Hare extension of
the Kruskal–Wallis test) of the whole corpus (i.e. of 3 × 3 × 50 = 450 arti-
cles) with discipline and language as the two factors. These are tests at the
superlevel (see section 2.2.5). The purpose of this analysis is twofold. First,
we test the significance of the effects of discipline and language and of the
interaction of discipline and language. Specifically, we test
a. whether there are significant differences (in the relative frequency of
the feature) between at least one pair of disciplinary subcorpora (or,
to put it differently, whether the discipline factor has a significant ef-
fect),
b. whether there are significant differences between at least one pair of
language subcorpora (or, in other words, whether the language factor
has a significant effect), and
c. whether the effects of discipline and language interact significantly,
so that, in order to tell how much discipline matters, we have to take
language into account, and vice versa.
Second, we compare the effects of discipline and language in order to tell
which factor has the stronger effect on the use of the feature. If we find that
a factor has no significant effect, no further comparisons for that factor
are motivated, i.e. pairwise comparisons of disciplinary subcorpora and
292 Academic Voices

of language subcorpora. If a significant effect is found, we proceed to such

pairwise comparisons.
2. pairwise comparisons (using the Mann–Whitney–Wilcoxon test) of dis-
ciplinary subcorpora (all languages together) and of language subcorpora
(all disciplines together), i.e. comparisons of 2 × (3 × 50) = 300 articles.
These are tests at the superlevel (see section 2.2.5). Since there are 3 pairs of
disciplinary subcorpora and 3 pairs of language subcorpora, the number
of such tests is 6 for each feature (given that significant effects have been
found for both discipline and language in the two-way analysis of vari-
ance). The purpose of these tests is to find out
a. for any given pair of disciplinary subcorpora (e.g. lingall and medall),
whether there is a significant difference between them (in the relative
frequency of the feature), and
b. for any given pair of language subcorpora (e.g. engall and noall),
whether there is a significant difference between them.
3. one-way analyses of variance (using the Kruskal–Wallis test), for language
differences within each discipline and for disciplinary differences within
each language (i.e. analyses of 3 × 50 articles). These are tests at the sublevel
(see section 2.2.5). Since there are 3 disciplines and 3 languages, there will
be 6 such analyses for each feature. The purpose of these tests is to find out
a. for a given discipline (e.g. econall), whether there is a significant dif-
ference between at least one pair of language subcorpora (with respect
to the relative frequency of the feature), and
b. for a given language (e.g. noall), whether there is a significant differ-
ence between at least one pair of disciplinary subcorpora.
If we find that there are no significant differences between subcorpora, no
further investigations are motivated. If significant differences are found,
we proceed to pairwise comparisons of language subcorpora within each
discipline and of disciplinary subcorpora within each language.
4. pairwise comparisons (using the Mann–Whitney–Wilcoxon test) of lan-
guage subcorpora for each discipline and of disciplinary subcorpora for
each language (i.e. comparisons of 2 × 50 = 100 articles). These are tests at
the sublevel (see section 2.2.5). Since there are 3 disciplinary subcorpora
with 3 language pairs each, and 3 language subcorpora with 3 disciplinary
pairs each, there will be 18 such tests for each feature (given that signifi-
cant differences have been found for both discipline and language in the
one-way analyses of variance). The purpose of these tests is to find out
a. for any given pair of language subcorpora within one discipline (e.g.
frecon and noecon), whether there is a significant difference between
them (with respect to the relative frequency of the feature), and
 Appendices 293

b. for any given pair of disciplinary subcorpora within one language (e.g.
engecon and engmed), whether there is a significant difference be-
tween them.

We have used the statistics program SPSS to do the statistical testing, with one
exception, viz. the two-way analysis of variance, for which SPSS does not provide
a non-parametric test. To do the two-way ANOVA (Scheirer–Ray–Hare extension
of the Kruskal–Wallis test), we first used Microsoft Excel to rank articles. We then
used SPSS to perform a parametric two-way ANOVA on the ranks. Finally, we
used Excel to do the remaining calculations.

Two-way analysis of variance

Our first statistical study, the two-way analysis of variance, takes all the 450 articles
in the corpus into account.
The analysis is based on a ranking of all the articles for the relative frequency
of the six features that are investigated: first person subjects, indefinite pronoun
subjects, metatext, negation, adversatives, and bibliographical references. Table
B-1 provides the mean ranks (rounded off to whole numbers) for the nine lan-
guage–discipline subcorpora (with 50 articles each) and for each of the disciplines
and languages (with 3 × 50 = 150 articles each). Since there are 450 articles in the
corpus, the expected mean rank is in every case 225.5 ≈ 226. If the mean rank is
smaller, articles in that sample tend to have a lower relative frequency for the fea-
ture in question than the median.

Table B-1. Mean ranks for the relative frequency of the investigated features
1st p subj econ ling med all
eng 288 265 166 239
fr 203 176 120 166
no 301 338 175 271
all 264 260 153 226
Indef subj econ ling med all
eng 110 134 51 98
fr 319 378 208 302
no 267 296 268 277
all 232 269 176 226
294 Academic Voices

Metatext econ ling med all

eng 372 346 124 281
fr 253 183 84 173
no 320 263 85 223
all 315 264 98 226
Negation econ ling med all
eng 143 263 173 193
fr 137 273 130 180
no 287 375 249 303
all 189 304 184 226
Advers econ ling med all
eng 171 265 140 192
fr 136 321 154 204
no 217 366 259 281
all 175 317 184 226
Bibl ref econ ling med all
eng 127 259 363 249
fr 124 137 294 185
no 144 220 363 242
all 131 205 340 226

Table B-2 presents the results of two-way factorial analysis of variance, viz. the
Scheirer–Ray–Hare extension of the Kruskal–Wallis test, two-tailed with α = 0.05.
The table gives numbers for the total variance (all observations together, labelled
Total). The total variance is composed of two parts: variance within groups (i.e.
sum of variances in each of the 9 subcorpora) and variance between groups. The
numbers for variance within groups are given in the row labelled Error. (It is called
error because the variance within groups cannot be accounted for with reference
to the factors (discipline and language) or the interaction between them.) The var-
iance between groups has three components: variance depending on the two fac-
tors, viz. variance between disciplines and variance between languages, and vari-
ance depending on the interaction between the factors. Numbers for these three
are given in the rows labelled Discipline, Language, and Disc*Lang, respectively,
and it is these measures that are tested for significance.
Mean square (MS): This a measure of variance. (In ANOVA result tables, the
sum of squares (SS) is sometimes given. SS = MS × df.) The mean square shows,
for disciplines, how much variance there is between disciplinary subcorpora (in
other words, how much it matters whether an article is from one discipline or
the other), and analogously for languages and the interaction between discipline
and language. The higher the mean square for a factor or the interaction between
 Appendices 295

them, the stronger is the effect of the factor or interaction on the relative frequency
of the feature in question – and the more likely it is that the effect is significant.
χ2 is the value of H, the test statistic for the Kruskal–Wallis test. The test is
two-tailed, and we use a 5 % level of significance. For the factors (discipline and
language), the degree of freedom is 2, which means that the critical value for χ2
is 5.99. If χ2 ≥ 5.99, the effect of the factor in question is significant. For the in-
teraction, the degree of freedom is 4, which means that the critical value for χ2 is
9.49. If χ2 ≥ 9.49, the effect of the interaction is significant. Significant values are
marked with an asterisk. Significance means that we can claim, with less than 5
% probability of being wrong, that there are differences in the populations from
which the samples have been drawn (with respect to the relative frequency of the
feature in question) depending on what discipline the article comes from, or what
language it is written in, or the interaction of discipline and language. For lower χ2
values, although there are differences, there is a probability greater than 5 % that
the differences are coincidences of the sampling. When a significant effect for a
factor is found, we proceed to a pairwise testing of differences between disciplines
and between languages (Mann–Whitney–Wilcoxon test). The p column provides
further information about significance (whether there is significance also at the
0.01, 0.005 and 0.001 level).
ω2 is a measure of strength of relationship. It tells us how much of the variance
in the population is accounted for by the factor or interaction in question.

Table B-2. Results of the Scheirer–Ray–Hare extension of the Kruskal–Wallis test

1st p subj df MS χ2 p ω2
Discipline 2 586164.50 69.33* < 0.001 0.15
Language 2 435403.09 51.49* < 0.001 0.11
Disc*Lang 4 39846.17 9.43 > 0.050 0.01
Error 441 12222.92
Total 449 16910.53
Indef subj df MS χ2 p ω2
Discipline 2 330927.77 39.47* < 0.001 0.09
Language 2 1854844.97 221.24* < 0.001 0.49
Disc*Lang 4 71637.95 17.09* < 0.005 0.03
Error 441 6509.60
Total 449 16767.99
296 Academic Voices

Metatext df MS χ2 p ω2
Discipline 2 1936700.83 235.38* < 0.001 0.52
Language 2 432709.41 52.59* < 0.001 0.12
Disc*Lang 4 51194.29 12.44* < 0.050 0.02
Error 441 5544.43
Total 449 16455.89
Negation df MS χ2 p ω2
Discipline 2 690915.13 81.70* < 0.001 0.18
Language 2 686855.12 81.22* < 0.001 0.18
Disc*Lang 4 20284.20 4.80 > 0.050 0.00
Error 441 10786.93
Total 449 16912.50
Advers df MS χ2 p ω2
Discipline 2 955371.29 112.98* < 0.001 0.25
Language 2 347264.45 41.07* < 0.001 0.09
Disc*Lang 4 37251.45 8.81 > 0.050 0.01
Error 441 10973.77
Total 449 16912.49
Bibl ref df MS χ2 p ω2
Discipline 2 1671869.04 197.71* < 0.001 0.44
Language 2 187757.59 22.20* < 0.001 0.05
Disc*Lang 4 45997.38 10.88* < 0.050 0.02
Error 441 8368.41
Total 449 16912.50
Two-tailed test, α = 0.05, N = 3 × 3 × 50 = 450.

Pairwise comparisons of disciplines and of languages

Our second statistical study first compares discipline subcorpora two and two
(with all languages included), and then we similarly compare language subcorpora
(with all disciplines included). This investigation is a two-tailed Mann–Whitney–
Wilcoxon test, α = 0.05. The test is based on a ranking of the 2 × 150 = 300 articles
in each pair of discipline or language subcorpora for the relative frequency of the
six features that are investigated. The results are presented in tables B-3 and B-4 for
discipline and language pairs, respectively.
Mean ranks: Since the total number of articles in each comparison is 300, the
expected mean rank for each sample is 150.5. If the mean rank is smaller, articles
in that sample tend to have a lower relative frequency for the feature in question
than the median. The greater the difference between the samples, the more likely
it is that the difference is statistically significant.
 Appendices 297

z: This is the value of the test statistic. Since we are using a two-tailed test and
a 5 % level of significance, the critical value for z is 1.96. If the absolute value of z
≥ 1.96, the difference between the samples is significant. (Such values are marked
with an asterisk.) This means that we can claim, with less than 5 % probability of
being wrong, that there is a difference between the populations from which the
samples have been drawn with respect to the relative frequency of the feature in
question. For lower z values, although there is a difference between the samples,
there is a probability greater than 5 % that the difference is a coincidence of the
sampling. The p column provides further information about significance (whether
there is significance also at the 0.01, 0.005 and 0.001 level).

Table B-3. Results of the Mann–Whitney–Wilcoxon test of differences between discipline

subcorpora
Mean ranks z p
Econ vs ling econ ling
1st p subj 152.03 148.97 –0.305 > 0.050
Indef subj 137.49 163.51 –2.599* < 0.010
Metatext 171.13 129.87 –4.120* < 0.001
Negation 111.57 189.43 –7.772* < 0.001
Advers 101.65 199.35 –9.754* < 0.001
Bibl ref 121.65 179.35 –5.761* < 0.001
Econ vs med econ med
1st p subj 187.02 113.98 –7.293* < 0.001
Indef subj 169.86 131.14 –3.904* < 0.001
Metatext 219.54 81.46 –14.202* < 0.001
Negation 152.79 148.21 –0.458 > 0.050
Advers 148.52 152.48 –0.396 > 0.050
Bibl ref 85.31 215.69 –13.016* < 0.001
Ling vs med ling med
1st p subj 186.13 114.87 –7.114* < 0.001
Indef subj 180.96 120.04 –6.122* < 0.001
Metatext 209.25 91.75 –12.226* < 0.001
Negation 189.89 111.11 –7.864* < 0.001
Advers 193.63 107.37 –8.612* < 0.001
Bibl ref 101.49 199.51 –9.786* < 0.001
Two-tailed test, α = 0.05, N = 2 × 150 = 300.
298 Academic Voices

Table B-4. Results of the Mann–Whitney–Wilcoxon test of differences between language

subcorpora
Mean ranks z p
Eng vs fr eng fr
1st p subj 176.09 124.91 –5.111* < 0.001
Indef subj 87.07 213.93 –12.817* < 0.001
Metatext 185.33 115.67 –7.040* < 0.001
Negation 154.71 146.29 –0.841 > 0.050
Advers 147.23 153.77 –0.654 > 0.050
Bibl ref 172.17 128.83 –4.326* < 0.001
Eng vs no eng no
1st p subj 138.70 162.30 –2.355* < 0.050
Indef subj 86.32 214.68 –12.953* < 0.001
Metatext 170.75 130.25 –4.072* < 0.001
Negation 113.81 187.19 –7.326* < 0.001
Advers 120.40 180.60 –6.011* < 0.001
Bibl ref 152.79 148.21 –0.457 > 0.050
Fr vs no fr no
1st p subj 116.66 184.34 –6.758* < 0.001
Indef subj 163.22 137.78 –2.540* < 0.050
Metatext 133.09 167.91 –3.565* < 0.001
Negation 109.41 191.59 –8.204* < 0.001
Advers 125.45 175.55 –5.001* < 0.001
Bibl ref 131.53 169.47 –3.788* < 0.001
Two-tailed test, α = 0.05, N = 2 × 150 = 300.

Analyses of variance for languages within disciplines and for disciplines within
languages
Our third statistical study, the one-way analysis of variance, first takes into account
the 3 × 50 articles in each language and tests the amount of variance among the
three disciplines. Second, it takes into account the 3 × 50 articles in each discipline
and tests the amount of variance among the three languages. The investigation is
a two-tailed Kruskal–Wallis test with α = 0.05 and df = 2. The test is based on a
ranking of the 3 × 50 = 150 articles in each discipline or language for the relative
frequency of the six features that are investigated. The results are presented in ta-
bles B-5 and B-6 for discipline and language triples, respectively.
Mean ranks: Since the total number of articles in each analysis of variance is
150, the expected mean rank for each sample is 75.5. If the mean rank is smaller,
articles in that sample tend to have a lower relative frequency for the feature in
 Appendices 299

question than the median. The greater the differences between the samples, the
more likely it is that the variance is statistically significant.
χ2: This is the value of H, the test statistic. Since the test is two-tailed with a de-
gree of freedom of 2 and we use a 5 % level of significance, the critical value for χ2 is
5.99. If χ2 ≥ 5.99, the variance between the samples is significant. (Such values are
marked with an asterisk.) This means that we can claim, with less than 5 % prob-
ability of being wrong, that at least one pair of populations from which the samples
have been drawn is different with respect to the relative frequency of the feature in
question. For lower χ2 values, although there are differences between the samples,
there is a probability greater than 5 % that the differences are coincidences of the
sampling. When significant variance is found, we proceed to a pairwise testing of
differences between languages (within each discipline) and between disciplines
(within each language) (Mann–Whitney–Wilcoxon test). The p column provides
further information about significance (whether there is significance also at the
0.01, 0.005 and 0.001 level).

Table B-5. Results of the Kruskal–Wallis test of variance among disciplines in language
subcorpora
Variable Mean ranks χ2 p
English econ ling med
1st p subj 92.96 84.48 49.06 28.730* < 0.001
Indef subj 83.60 100.48 42.42 54.647* < 0.001
Metatext 104.70 95.16 26.64 96.511* < 0.001
Negation 58.48 99.64 68.38 24.453* < 0.001
Advers 67.84 102.88 55.78 31.714* < 0.001
Bibl ref 33.96 76.24 116.30 89.821* < 0.001
French econ ling med
1st p subj 88.28 81.36 56.86 14.452* < 0.001
Indef subj 80.08 106.34 40.08 58.997* < 0.001
Metatext 107.07 80.71 38.72 67.738* < 0.001
Negation 62.38 106.64 57.48 38.849* < 0.001
Advers 52.64 114.34 59.52 60.569* < 0.001
Bibl ref 54.82 59.74 111.94 53.084* < 0.001
Norwegian econ ling med
1st p subj 84.60 98.24 43.66 42.749* < 0.001
Indef subj 70.57 83.77 72.16 2.751 > 0.050
Metatext 108.75 88.08 29.67 92.136* < 0.001
Negation 65.56 106.30 54.64 39.274* < 0.001
Advers 52.77 105.94 67.79 39.806* < 0.001
Bibl ref 42.60 67.70 116.20 74.165* < 0.001
Two-tailed test, α = 0.05, N = 3 × 50 = 150.
300 Academic Voices

Table B-6. Results of the Kruskal–Wallis test of variance among languages in discipline
subcorpora
Mean ranks χ2 p
Economics eng fr no
1st p subj 83.38 54.36 88.76 18.142* < 0.001
Indef subj 31.88 106.63 87.99 80.459* < 0.001
Metatext 104.38 47.11 75.01 43.454* < 0.001
Negation 57.54 57.62 111.34 51.040* < 0.001
Advers 74.50 60.90 91.10 12.120* < 0.005
Bibl ref 74.20 72.04 80.26 0.962 > 0.050
Linguistics eng fr no
1st p subj 76.46 46.42 103.62 43.373* < 0.001
Indef subj 30.50 112.58 83.42 91.762* < 0.001
Metatext 107.78 44.60 74.12 53.054* < 0.001
Negation 59.70 64.02 102.78 29.818* < 0.001
Advers 50.85 77.09 98.56 30.249* < 0.001
Bibl ref 97.40 48.89 80.21 32.050* < 0.001
Medicine eng fr no
1st p subj 81.24 61.10 84.16 8.359* < 0.050
Indef subj 31.94 87.72 106.84 85.436* < 0.001
Metatext 92.67 66.90 66.93 19.342* < 0.001
Negation 71.65 56.85 98.00 23.017* < 0.001
Advers 60.09 66.73 99.68 23.816* < 0.001
Bibl ref 85.64 55.84 85.02 15.363* < 0.001
Two-tailed test, α = 0.05, N = 3 × 50 = 150.

Results of pairwise comparisons of language–discipline subcorpora

Our fourth statistical study first compares discipline subcorpora two and two for
each language, and then we similarly compare language subcorpora for each dis-
cipline. The investigation is a two-tailed Mann–Whitney–Wilcoxon test with α =
0.05. The test is based on a ranking of the 2 × 50 = 100 articles in each subcorpus
pair for the relative frequency of the six features that are investigated. The results
for discipline pairs are presented in tables B-7–9 for English, French, and Norwe-
gian, respectively. The results for language pairs are similarly presented in tables
B-10–12 for economics, linguistics, and medicine, respectively. (Results for indefi-
nite pronoun subjects in Norwegian and for bibliographical references in econom-
ics have been included for the sake of a full presentation, although the results of
the one-way analysis of variance have not yielded significant variance for these
feature–subcorpus combinations.)
 Appendices 301

Mean ranks: Since the total number of articles in each comparison is 100, the
expected mean rank for each sample is 50.5. If the mean rank is smaller, articles in
that sample tend to have a lower relative frequency for the feature in question than
the median. The greater the difference between the samples, the more likely it is
that the difference is statistically significant.
z: This is the value of the test statistic. Since we are using a two-tailed test and
a 5 % level of significance, the critical value for z is 1.96. If z ≥ 1.96, the difference
between the samples is significant. (Such values are marked with an asterisk.) This
means that we can claim, with less than 5 % probability of being wrong, that there
is a difference between the populations from which the samples have been drawn
with respect to the relative frequency of the feature in question. For lower z values,
although there is a difference between the samples, there is a probability greater
than 5 % that the difference is a coincidence of the sampling. The p column pro-
vides further information about significance (whether there is significance also at
the 0.01, 0.005 and 0.001 level).

Table B-7. Results of the Mann–Whitney–Wilcoxon test of differences between discipline

subcorpora in English
Mean ranks z p
Engecon vs engling econ ling
1st p subj 53.70 47.30 –1.103 > 0.050
Indef subj 44.66 56.34 –2.041* < 0.050
Metatext 55.08 45.92 –1.579 > 0.050
Negation 36.80 64.20 –4.722* < 0.001
Advers 38.14 62.86 –4.260* < 0.001
Bibl ref 32.82 68.18 –6.094* < 0.001
Engecon vs engmed econ med
1st p subj 64.76 36.24 –4.915* < 0.001
Indef subj 64.44 36.56 –5.746* < 0.001
Metatext 75.12 25.88 –8.553* < 0.001
Negation 47.18 53.82 –1.144 > 0.050
Advers 55.20 45.80 –1.620 > 0.050
Bibl ref 26.64 74.36 –8.224* < 0.001
Engling vs engmed ling med
1st p subj 62.68 38.32 –4.198* < 0.001
Indef subj 69.64 31.36 –7.309* < 0.001
Metatext 74.74 26.26 –8.421* < 0.001
Negation 60.94 40.06 –3.599* < 0.001
Advers 65.52 35.48 –5.177* < 0.001
Bibl ref 33.56 67.44 –5.839* < 0.001
Two-tailed test, α = 0.05, N = 2 × 50 = 100.
302 Academic Voices

Table B-8. Results of the Mann–Whitney–Wilcoxon test of differences between discipline

subcorpora in French
Mean ranks z p
Frecon vs frling econ ling
1st p subj 53.62 47.38 –1.076 > 0.050
Indef subj 40.46 60.54 –3.461* < 0.001
Metatext 60.71 40.29 –3.533* < 0.001
Negation 35.06 65.94 –5.322* < 0.001
Advers 30.42 70.58 –6.921* < 0.001
Bibl ref 48.60 52.40 –0.655 > 0.050
Frecon vs frmed econ med
1st p subj 60.16 40.84 –3.333* < 0.001
Indef subj 65.12 35.88 –5.041* < 0.001
Metatext 71.86 29.14 –7.777* < 0.001
Negation 52.82 48.18 –0.800 > 0.050
Advers 47.72 53.28 –0.958 > 0.050
Bibl ref 31.72 69.28 –6.473* < 0.001
Frling vs frmed ling med
1st p subj 59.48 41.52 –3.097* < 0.005
Indef subj 71.30 29.70 –7.171* < 0.001
Metatext 65.92 35.08 –5.888* < 0.001
Negation 66.20 34.80 –5.412* < 0.001
Advers 69.26 31.74 –6.466* < 0.001
Bibl ref 32.84 68.16 –6.087* < 0.001
Two-tailed test, α = 0.05, N = 2 × 50 = 100.

Table B-9. Results of the Mann–Whitney–Wilcoxon test of differences between discipline

subcorpora in Norwegian
Mean ranks z p
Noecon vs noling econ ling
1st p subj 45.46 55.54 –1.737 > 0.050
Indef subj 46.19 54.81 –1.486 > 0.050
Metatext 59.48 41.52 –3.096* < 0.005
Negation 36.06 64.94 –4.977* < 0.001
Advers 33.02 67.98 –6.025* < 0.001
Bibl ref 40.76 60.24 –3.357* < 0.001
 Appendices 303

Noecon vs nomed econ med

1st p subj 64.64 36.36 –4.874* < 0.001
Indef subj 49.88 51.12 –0.214 > 0.050
Metatext 74.77 26.23 –8.746* < 0.001
Negation 55.00 46.00 –1.551 > 0.050
Advers 45.25 55.75 –1.810 > 0.050
Bibl ref 27.34 73.66 –7.983* < 0.001
Noling vs nomed ling med
1st p subj 68.20 32.80 –6.101* < 0.001
Indef subj 54.46 46.54 –1.365 > 0.050
Metatext 72.06 28.94 –7.850* < 0.001
Negation 66.86 34.14 –5.639* < 0.001
Advers 63.46 37.54 –4.467* < 0.001
Bibl ref 32.96 68.04 –6.046* < 0.001
Two-tailed test, α = 0.05, N = 2 × 50 = 100.

Table B-10. Results of the Mann–Whitney–Wilcoxon test of differences between language

subcorpora in economics
Mean ranks z p
Engecon vs frecon eng fr
1st p subj 60.54 40.46 –3.461* < 0.001
Indef subj 27.40 73.60 –7.999* < 0.001
Metatext 68.04 32.96 –6.046* < 0.001
Negation 49.24 51.76 –0.434 > 0.050
Advers 55.04 45.96 –1.565 > 0.050
Bibl ref 51.12 49.88 –0.214 > 0.050
Engecon vs noecon eng no
1st p subj 48.34 52.66 –0.745 > 0.050
Indef subj 29.98 71.02 –7.106* < 0.001
Metatext 61.84 39.16 –3.909* < 0.001
Negation 33.80 67.20 –5.756* < 0.001
Advers 44.96 56.04 –1.910 > 0.050
Bibl ref 48.58 52.42 –0.662 > 0.050
Frecon vs noecon fr no
1st p subj 39.40 61.60 –3.826* < 0.001
Indef subj 58.53 42.47 –2.768* < 0.010
Metatext 39.65 61.35 –3.740* < 0.001
Negation 31.36 69.64 –6.597* < 0.001
Advers 40.44 60.56 –3.468* < 0.001
Bibl ref 47.66 53.34 –0.979 > 0.050
Two-tailed test, α = 0.05, N = 2 × 50 = 100.
304 Academic Voices

Table B-11. Results of the Mann–Whitney–Wilcoxon test of differences between language

subcorpora in linguistics
Mean ranks z p
Engling vs frling eng fr
1st p subj 61.62 39.38 –3.833* < 0.001
Indef subj 26.02 74.98 –8.442* < 0.001
Metatext 70.66 30.34 –6.961* < 0.001
Negation 49.06 51.94 –0.496 > 0.050
Advers 41.27 59.73 –3.182* < 0.005
Bibl ref 66.64 34.36 –5.563* < 0.001
Engling vs noling eng no
1st p subj 40.34 60.66 –3.502* < 0.001
Indef subj 29.98 71.02 –7.078* < 0.001
Metatext 62.62 38.38 –4.178* < 0.001
Negation 36.14 64.86 –4.950* < 0.001
Advers 35.08 65.92 –5.315* < 0.001
Bibl ref 56.26 44.74 –1.985* < 0.050
Frling vs noling fr no
1st p subj 32.54 68.46 –6.191* < 0.001
Indef subj 63.10 37.90 –4.343* < 0.001
Metatext 39.76 61.24 –3.715* < 0.001
Negation 37.58 63.42 –4.453* < 0.001
Advers 42.86 58.14 –2.633* < 0.010
Bibl ref 40.03 60.97 –3.609* < 0.001
Two-tailed test, α = 0.05, N = 2 × 50 = 100.
 Appendices 305

Table B-12. Results of the Mann–Whitney–Wilcoxon test of differences between language

subcorpora in medicine
Mean ranks z p
Engmed vs frmed eng fr
1st p subj 57.32 43.68 –2.352* < 0.050
Indef subj 29.82 71.18 –7.807* < 0.001
Metatext 59.28 41.72 –3.619* < 0.001
Negation 55.59 45.41 –1.754 > 0.050
Advers 47.49 53.51 –1.038 > 0.050
Bibl ref 60.46 40.54 –3.433* < 0.001
Engmed vs nomed eng no
1st p subj 49.42 51.58 –0.372 > 0.050
Indef subj 27.62 73.38 –8.468* < 0.001
Metatext 58.89 42.11 –3.493* < 0.001
Negation 41.56 59.44 –3.082* < 0.005
Advers 38.10 62.90 –4.274* < 0.001
Bibl ref 50.68 50.32 –0.062 > 0.050
Frmed vs nomed fr no
1st p subj 42.92 58.08 –2.615* < 0.010
Indef subj 42.04 58.96 –2.919* < 0.005
Metatext 50.68 50.32 –0.100 > 0.050
Negation 36.94 64.06 –4.674* < 0.001
Advers 38.72 62.28 –4.060* < 0.001
Bibl ref 40.80 60.20 –3.344* < 0.001
Two-tailed test, α = 0.05, N = 2 × 50 = 100.
 Index

A concession 216, 243, 247–250, first person plural 40, 68, 69,
absolute frequency 40, 41, 270 74, 76–78, 82, 95, 96, 98,
172, 173 concessive 37, 195, 217, 218, 101, 104, 107, 108, 113, 139,
abstract 6, 7, 12, 13, 15, 19, 223, 224, 242, 243, 247–251, 143, 147, 148, 171, 200, 243,
32, 33, 35, 37, 38, 67, 72, 73, 253–255, 270 254, 264, 267
82, 114, 126–137, 153–157, connective 34, 37, 177, 195, first person singular 40, 45,
183, 188, 193–196, 198, 201, 216–218, 223-224, 229, 242, 67–70, 73–75, 81–83, 90–93,
267, 269 243, 249, 250, 252, 255, 95, 101, 106, 113, 127, 133,
adversative 3, 4, 51, 53, 54, 257, 270 147, 264
59–61, 64, 65, 210, 216–218, contrastive linguistics 46 frequency (see absolute fre-
222, 223–226, 242, 243, 247, convince 30, 48, 95, 178 quency, relative frequency)
250, 255, 257, 260, 263, 265, culture 2, 5, 14, 16–18, 20, fully integral reference 227,
270, 291, 293 23, 28, 48, 51, 169, 221, 225, 229, 231, 256
analysis of variance (see 266, 269
Kruskal–Wallis test, G
Scheirer– Ray–Hare exten- D gender 23, 24, 40, 44, 55, 76,
sion) dependent variable 40, 44 79–81, 138, 162, 163, 221,
ANOVA (see Kruskal–Wallis directive 171-173, 177, 180 223, 226, 248, 265
test, Scheirer–Ray–Hare discipline factor 43, 54, 76, genre 1, 2, 5, 17, 22–25, 27,
extension) 78, 162, 167, 220, 222, 225, 29–32, 46, 67, 110, 112, 115,
arguer 67, 82, 84, 85, 88–95, 238, 257, 260, 267, 291 116, 118, 124, 126, 133, 157,
101, 102, 127–133, 137–139, discourse community 1, 2, 178, 180, 243, 244, 259, 268
142, 143, 147, 152, 155, 157, 17, 22, 24, 25, 27, 30, 37, grammatical subject 4, 40,
241, 253, 256, 257, 267, 270 117, 120, 124, 125, 138–141, 42, 72, 73, 79, 101, 147, 218,
author role (see arguer, 142, 143, 146, 191, 202, 217, 226, 231, 260
evaluator, researcher, 237, 268
H
writer) E high context 169
authorial ‘we’ 68, 105, 108, encoding 12–14 hypothesis/hypotheses 2, 5,
154 enunciation 31–34, 38 6, 41, 44, 45, 58, 70–72, 74,
B evaluator 67, 82, 84, 85, 89, 76, 78, 79, 82, 124, 126, 160,
bibliographical reference 91, 92, 95, 127, 147, 152, 157 181, 184, 192, 213, 217, 220,
(see non-integral, partly exclusive ‘we’ 96, 98, 221, 230, 255, 256
integral, semi-integral, 100–102, 104, 106, 107, I
fully integral) 109, 179 identity 1–3, 6, 11, 16–20, 23,
body (of article) 7, 12–15, 41, exploratory 46, 83, 221 24, 51, 266–267, 269
71, 123, 139, 141, 144, 157, IMRAD (IMRD) 12, 13, 22,
F 24, 31, 166, 167, 182, 262
219, 230 factor (see discipline factor,
Bokmål 4, 73, 222 inclusive ‘we’ 68, 96–102,
interaction, language 104–107, 109, 178
C factor)
citation 27, 36, 82, 216, 217,
220, 226, 229, 231, 234, 237
308 Academic Voices

indefinite pronoun 42, 51, multivoiced 3, 4, 32, 36, 215, Q

54, 55, 57, 59, 63, 67, 72, 73, 241, 251, 269 qualitative 4, 41, 46-49, 52,
79–80, 110–113, 138, 196, 81, 116, 127, 161, 182, 183,
202, 240, 260, 261, 264, 291, N 186, 191, 201, 217, 223, 226,
293, 300 negation 2, 3, 4, 6, 37, 38, 255, 257, 259, 269
independent variable 40, 42–44, 51, 52, 54, 55, 59–61, quantitative 11, 39-45, 46,
44, 267 64, 65, 174, 195, 198, 201, 51-65, 73-81, 82, 104, 161-
interaction (of factors) 43, 210, 216–218, 222–223, 225, 163, 182, 183–192, 217–218,
54, 55, 74, 161, 219, 222, 224, 242–247, 250–255, 257, 260, 218-226, 230-231, 240, 243,
291, 294, 295 263, 265, 266, 270, 291, 293, 255-256, 259, 260-265
intercultural 18, 23, 28, 266 294, 296–305 quote 7, 14, 37, 216, 227-238,
interpersonal 4, 29, 154, neutral 1, 3, 81, 92, 189 256, 270
160, 170 non-integral reference 226–
227, 231, 256 R
K non-parametric statistics 42, rank 41, 52-53, 55-64, 79
Kruskal–Wallis test 43, 44, 44–46, 291, 293 reader 1, 3, 18, 29, 37, 39, 48,
74, 291–296, 298–299, normal distribution 45 52, 68, 69, 72, 82, 84-85, 92,
300–301 Nynorsk 4, 7, 73, 146, 171, 222 95, 97–110, 111, 117,–127,
133, 138–143, 146, 148, 152,
L O 154, 156, 159–161, 163, 165,
language factor 43, 54–55, other-dimension 3, 6, 29, 48, 167–169, 169-171, 173-174,
67, 76, 78, 156, 161, 212, 67, 124, 215, 217, 218, 238, 177–179, 181, 193, 195, 198,
221, 222, 223, 225, 238, 257, 240, 255, 259, 266, 270, 271 200-202, 204–207, 209,
260, 291 213, 215, 217, 227, 229, 242,
‘let me’-imperative 73, 170, P
parametric statistics 44–45, 246, 250, 254, 259, 261–264,
178, 206–209 269-270
‘let us’-imperative 46–48, 293
partly integral reference 227, referencing 217, 233, 237,
106, 107, 169–180, 205–209, 240, 255
210, 261, 262, 264, 267, 270 228, 256
person (see first, second, relative frequency 42–45,
locutor 33, 34, 37–39, 81, 109, 51–55, 58, 60-61, 65, 69-71,
111, 113, 114, 161, 230, 234, third person)
personal pronoun 73, 96, 113, 74-80, 104–105, 115, 139,
238, 242, 243, 245, 250, 161-162, 172, 180, 202, 210,
251–253 114, 131–133, 147, 150, 184,
204, 245, 264 219-224, 230-231, 243, 257,
low context 23, 169 260, 264, 291-301
person manifestation 3–5,
M 17, 29, 156, 157, 215, 259, representativity 9–11, 45, 53
Mann–Whitney–Wilcoxon 260, 269 researcher 67, 82-83, 85, 88-
test 43, 44, 52, 74, 76, 80, persuasion 19, 141, 181, 193, 92, 94, 95, 100-102, 127-133,
163, 172, 221, 223, 226, 292, 267, 271 137, 138-140, 142-144, 146,
295, 296–305 pilot study 4, 110, 161, 167, 223 147, 150-151, 153-155, 157,
mean rank 53, 55–64, 79, 293, point of view (pov) 34–38, 232-233, 239, 256, 270
296–305 72, 108, 110, 117, 216, 217, S
metadiscourse 26, 154, 243, 244, 248–249, 251–254, Scheirer–Ray–Hare exten-
159–161, 163, 167 257, 270 sion 43, 293–295
metatext 2, 4, 6, 30, 34, polyphony 29, 32, 34–39, 72, search 12, 14-16, 40-41, 47, 73,
40–42, 48, 49, 51, 54, 55, 81, 91, 95, 108, 110, 111, 117, 79, 134, 163, 166-167, 170,
59, 61, 63, 65, 85, 87, 89, 90, 161, 215– 218, 223, 224, 230, 183-185, 222-223
105, 116, 118, 119, 122, 144, 234, 238, 242, 243–259, 265, second person 96, 170, 215
155, 157, 159–170, 173, 177, 269, 270–271 self-dimension 3-6, 29, 34,
184, 187, 202–206, 208–213, pov (see point of view) 48, 67, 118, 124-125, 143, 147,
260–267, 269, 271, 291, 293, 159, 215, 217-218, 234, 237-
294, 296–305 240, 250-251, 256-257, 259,
metonymy 98–100, 104, 266, 270-271
115, 212
mother tongue 1, 10
 Index 309

semi-integral reference 227, third person 3, 29, 34, 38, 69, W

229, 231, 256 72, 97, 111, 113, 117, 156, 170, writer 67, 82-83, 85, 87-94, 95,
significance 43-44, 52-55, 59- 179, 206, 215, 238, 259, 270 127, 129-133, 138-140, 142,
60, 63-65, 74, 76, 78-81, 105, 144, 146-147, 149-150, 157,
112, 115, 148, 162, 167-168, U 256, 270
172-173, 219-226, 243, 251, utterance 29-39, 82, 89, 108-
265, 291-297, 299-301 110, 114, 161, 177, 180-181,
statistics 4, 40-45, 53, 59, 74, 230, 242, 244, 248-249, 270
172, 259 V
subcorpus 7-9 variable 39-40, 44-45, 243
subjective 21, 85, 91 variance 43–45, 54-55, 74,
sublevel 43-44, 55, 61, 292- 78–80, 102, 156, 161, 212,
293 219, 222, 224, 291–295,
superlevel 43, 55, 61, 291-292 298–300
T verb group 83–87, 90
tags 12-14
Pragmatics & Beyond New Series
A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com

151 SUZUKI, Satoko (ed.): Emotive Communication in Japanese. v, 233 pp. + index. Expected November 2006
150 BUSSE, Beatrix: Vocative Constructions in the Language of Shakespeare. xii, 485 pp. + index. Expected October
2006
149 LOCHER, Miriam A.: Advice Online. Advice-giving in an American Internet health column. 2006. xvi, 278 pp.
148 FLØTTUM, Kjersti, Trine DAHL and Torodd KINN: Academic Voices. Across languages and disciplines. 2006.
x, 309 pp.
147 HINRICHS, Lars: Codeswitching on the Web. English and Jamaican Creole in e-mail communication.
xii, 295 pp. + index. Expected September 2006
146 TANSKANEN, Sanna-Kaisa: Collaborating towards Coherence. Lexical cohesion in English discourse. 2006.
ix, 192 pp.
145 KURHILA, Salla: Second Language Interaction. 2006. vii, 257 pp.
144 BÜHRIG, Kristin and Jan D. ten THIJE (eds.): Beyond Misunderstanding. Linguistic analyses of intercultural
communication. 2006. vi, 339 pp.
143 BAKER, Carolyn, Michael EMMISON and Alan FIRTH (eds.): Calling for Help. Language and social
interaction in telephone helplines. 2005. xviii, 352 pp.
142 SIDNELL, Jack: Talk and Practical Epistemology. The social life of knowledge in a Caribbean community. 2005.
xvi, 255 pp.
141 ZHU, Yunxia: Written Communication across Cultures. A sociocognitive perspective on business genres. 2005.
xviii, 216 pp.
140 BUTLER, Christopher S., María de los Ángeles GÓMEZ-GONZÁLEZ and Susana M. DOVAL-SUÁREZ
(eds.): The Dynamics of Language Use. Functional and contrastive perspectives. 2005. xvi, 413 pp.
139 LAKOFF, Robin T. and Sachiko IDE (eds.): Broadening the Horizon of Linguistic Politeness. 2005. xii, 342 pp.
138 MÜLLER, Simone: Discourse Markers in Native and Non-native English Discourse. 2005. xviii, 290 pp.
137 MORITA, Emi: Negotiation of Contingent Talk. The Japanese interactional particles ne and sa. 2005. xvi, 240 pp.
136 SASSEN, Claudia: Linguistic Dimensions of Crisis Talk. Formalising structures in a controlled language. 2005.
ix, 230 pp.
135 ARCHER, Dawn: Questions and Answers in the English Courtroom (1640–1760). A sociopragmatic analysis.
2005. xiv, 374 pp.
134 SKAFFARI, Janne, Matti PEIKOLA, Ruth CARROLL, Risto HILTUNEN and Brita WÅRVIK (eds.): Opening
Windows on Texts and Discourses of the Past. 2005. x, 418 pp.
133 MARNETTE, Sophie: Speech and Thought Presentation in French. Concepts and strategies. 2005. xiv, 379 pp.
132 ONODERA, Noriko O.: Japanese Discourse Markers. Synchronic and diachronic discourse analysis. 2004.
xiv, 253 pp.
131 JANOSCHKA, Anja: Web Advertising. New forms of communication on the Internet. 2004. xiv, 230 pp.
130 HALMARI, Helena and Tuija VIRTANEN (eds.): Persuasion Across Genres. A linguistic approach. 2005.
x, 257 pp.
129 TABOADA, María Teresa: Building Coherence and Cohesion. Task-oriented dialogue in English and Spanish.
2004. xvii, 264 pp.
128 CORDELLA, Marisa: The Dynamic Consultation. A discourse analytical study of doctor–patient
communication. 2004. xvi, 254 pp.
127 BRISARD, Frank, Michael MEEUWIS and Bart VANDENABEELE (eds.): Seduction, Community, Speech. A
Festschrift for Herman Parret. 2004. vi, 202 pp.
126 WU, Yi’an: Spatial Demonstratives in English and Chinese. Text and Cognition. 2004. xviii, 236 pp.
125 LERNER, Gene H. (ed.): Conversation Analysis. Studies from the first generation. 2004. x, 302 pp.
124 VINE, Bernadette: Getting Things Done at Work. The discourse of power in workplace interaction. 2004.
x, 278 pp.
123 MÁRQUEZ REITER, Rosina and María Elena PLACENCIA (eds.): Current Trends in the Pragmatics of
Spanish. 2004. xvi, 383 pp.
122 GONZÁLEZ, Montserrat: Pragmatic Markers in Oral Narrative. The case of English and Catalan. 2004.
xvi, 410 pp.
121 FETZER, Anita: Recontextualizing Context. Grammaticality meets appropriateness. 2004. x, 272 pp.