Greville G. Corbett (Ed.

)
The Expression of Gender
The Expression
of Cognitive Categories

Editors
Wolfgang Klein
Stephen Levinson

Volume 6
The Expression
of Gender

Edited by
Greville G. Corbett
ISBN: 978-3-11-030660-6
e-ISBN: 978-3-11-030733-7

Library of Congress Cataloging-in-Publication Data

A CIP catalog record for this book has been applied for at the Library of Congress.

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available in the Internet at http://dnb.dnb.de.

© 2014 Walter de Gruyter GmbH, Berlin/Boston

Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Berlin
Printing: Hubert & Co. GmbH & Co. KG, Göttingen
♾ Printed on acid-free paper
Printed in Germany

www.degruyter.com
Contents
Greville G. Corbett
Introduction  1

Sally McConnell-Ginet
Gender and its relation to sex: The myth of ‘natural’ gender  3

Michael Dunn
Gender determined dialect variation  39

Peter Hegarty
Ladies and gentlemen: Word order and gender in English  69

Greville G. Corbett
Gender typology  87

Marianne Mithun
Gender and culture  131

Niels O. Schiller
Psycholinguistic approaches to the investigation of grammatical gender  161

Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

Plural as a value of Cushitic gender: Evidence from gender congruency effect
experiments in Konso (Cushitic)  191

Author index  215
Language index  219
Subject index  221
Greville G. Corbett
Introduction
Gender is an endlessly fascinating category. It has obvious links to the real world,
first in the connection between many grammatical gender systems and biologi-
cal sex, and second in other types of categorization such as size, which underpin
particular gender systems and also have external correlates. While in some lan-
guages the way in which gender is assigned to nouns is semantically transparent,
for example, in Dravidian languages such as Tamil, others are rather opaque:
though their systems still have a semantic core, there is much more to be said
about gender in familiar languages like French or German. Of course, there are
other grammatical categories with links to the real world, but compared with
these gender is surprising in that it appears to be an “optional extra”. That is,
many of the world’s languages have gender, but many (probably somewhat over
half) do not. The differences continue: in some languages gender is a relatively
superficial matter, while in others it is central, being found through the noun
phrase and on the verb by agreement, and interacting in morphology with other
features, typically number, case and person. Thus the description of some lan-
guages requires constant detailed reference to gender, and for others it is absent.
Given its links to the real world, gender is a feature that speakers are partly
aware of. There is discussion of the appropriate use of gendered pronouns; learn-
ers comment on the difficulty of acquiring gender in some languages; the gender
of borrowings may arouse the curiosity of first language speakers. But gender is
like an iceberg since most of its interest is not apparent to normal speakers.
The authors of the volume take seven complementary perspectives on gender.
Sally McConnell-Ginet tackles one of the most approachable and yet most diffi-
cult issues: ‘Gender and its relation to sex: The myth of “natural” gender’, and
asks what meaning can be attached to gender. It also seems evident that in some
respects women and men speak differently. In some languages, such as Chukchi,
the differences are dramatic, as documented and analysed by Michael Dunn in
his chapter ‘Gender determined dialect variation’. In interesting contrast, Peter
Hegarty homes in on the way in which gender affects an apparently small lin-
guistic choice, the order of conjoined noun phrases, in his chapter ‘Ladies and
gentlemen: word order and gender in English’. The core linguistic types of varia-
tion between gender systems are laid out in ‘Typology of gender’ by Greville G.
Corbett. One way of understanding complex systems is to examine their behav-
iour over time. This is the approach of Marianne Mithun in ‘Gender and culture’,
where she examines subtle gender distinctions and the ways they came about
within the Iroquoian family, particularly in Mohawk. The variety found in gender
2       Greville G. Corbett

has naturally attracted psycholinguists, and Niels Schiller reviews this field and
presents recent research findings in ‘Psycholinguistic approaches to the inves-
tigation of grammatical gender’. Finally these psycholinguistic techniques are
applied to Konso, a language with a challenging gender/number system in ‘Plural
as value of Cushitic gender: Evidence from gender congruency effect experiments
in Konso (Cushitic)’ by Mulugeta Tarekegne Tsegaye, Maarten Mous and Niels
Schiller.
Some of the most exciting gender systems are found in languages whose sur-
vival is uncertain; the proceeds from this volume are therefore going to the Foun-
dation for Endangered Languages.
We thank the Max Planck Institute for Psycholinguistics (Nijmegen) for
funding and hosting a workshop on gender, which made it possible for authors
to discuss each other’s presentations and so to shape the different contributions
into a single volume, as well as to gain from the helpful comments of the audi-
ence. We are also grateful to Lisa Mack and Penny Everson for their help in pre-
paring the typescript and to Wolfgang Konwitschny for seeing the volume through
to publication.
Sally McConnell-Ginet
Gender and its relation to sex:
The myth of ‘natural’ gender
1 ‘ Grammatical’ vs ‘natural’ gender systems:
First pass
In McConnell-Ginet ([1988], 2011), I included the following:

The word gender in the title of this chapter refers to the complex of social, cultural, and
psychological phenomena attached to sex, a usage common in the behavioral and social
sciences. The word gender also, however, has a well-established technical sense in linguis-
tic discussions. Gender in this technical sense is a grammatically significant classification
of nouns that has implications for various agreement phenomena.

The title of the present chapter also includes the word gender: my thesis is that
something like sociocultural gender as delineated in the first sentence of the
extract above mediates connections between sex and pronoun choice in English
(often said to be based on ‘natural’ gender, discussed below) and also, though
less straightforwardly, between sex and the technical notion of ‘grammatical’
gender in linguistics.
The word sex includes the division of humans and many other animals
into female and male classes, based on reproductive potential; it also includes
matters of sexuality, not just sexual identity but also sexual desire and activity.
Sociocultural gender is not a matter of the sexual division of people into female
and male as such, what people typically mean by ‘natural’ gender, but of the
significance attached to that division, the institutions and ideologies, the pre-
scribed and claimed identities, and the array of social practices that sustain those
institutions, ideologies, and identities. What we have in English, I argue, is not a
‘natural’ gender system but what Terttu Nevalainen and Helena Raumolin-Brun-
berg (1994) call a ‘notional’ gender system: concepts and ideas about biological
sex matter at least as much as sex itself to the choice of English third-person pro-
nouns. (In principle there could be ‘notional’ gender systems involving first- or
second-person pronouns – Japanese might illustrate – or even agreement beyond
pronouns, but this chapter just contrasts the English notional gender system
with grammatical gender systems in earlier stages of English and in various Indo-
European languages.)
As a technical linguistic notion, gender is about agreement. Nouns are
assigned to classes such that which class a noun belongs to determines or ‘con-
4       Sally McConnell-Ginet

trols’ formal properties of other expressions linked to the phrase it heads, what
Corbett 1991 calls ‘agreement targets’. Agreement targets include articles and
attributive adjectives, often also adjectival and verbal expressions predicated of
that phrase, numerals, and relative pronouns and, importantly for my purposes,
anaphoric pronouns for which a phrase headed by the noun in question serves as
antecedent. The technical issue of most direct relevance to this chapter involves
what Corbett (this volume) dubs the assignment problem.
How nouns get assigned a gender is, as Corbett’s chapter documents in detail,
a very complex matter, and it varies considerably cross-linguistically. For some
languages, the main principles for assignment are formal; for others, they are
semantic in the sense of involving features of a noun’s referents; for many, both
semantic and formal features play a role in assignment. Not surprisingly, assign-
ment principles may shift over time, and at any given time there may be some
lexical items whose gender cannot be determined by the general principles then
operative in the language more broadly (and there may be variation in gender
assignment). Mithun’s illuminating discussion in this volume of gender in Iro-
quoian languages, especially Mohawk, shows this very clearly. Nonlinguists do
not always realize that gender semantic assignment principles need not connect
to sex at all: animacy, shape, and many other features figure in grammatical
gender systems crosslinguistically. Still, as Corbett’s chapter points out, sex is far
and away the most common feature to figure semantically in gender assignment.
What is of interest in this chapter are languages where sex and, crucially,
ideas about it play some sort of role in either a full-blown grammatical gender
system such as is found in many Indo-European languages (e.g., German, French,
Russian, Hindi) or in a limited system like that of English where only pronouns
show gender agreement. In languages with grammatical gender linked to sex, it
is often the case that inanimates for which sex is irrelevant can be assigned to the
same gender classes as sexed humans (or, for that matter, that nouns designating
sexed humans can on occasion be assigned to the ‘wrong’ gender class). For non-
native speakers acquiring such languages, such mismatches can seem very odd.
Mark Twain’s 1880 essay “The Awful German Language” seizes on such clashes
to underscore the difficulties many speakers of languages like English have in
learning a grammatical gender language like German, where they are tempted
to assimilate the German feminine gendered pronoun sie (and other marks of
feminine gender agreement) to something like what English she conveys, mascu-
line er to he, and inanimate es to it. Native speakers of German, in spite of what
Twain suggests, are not thinking of the referents of ‘wife’ (das Weib) or ‘girl’ (das
Mädchen), both neuter gender nouns with female referents, as unsexed beings.
Seeming to make the connection of grammatical gender to sex even looser,
there are also many cases where, e.g., one language assigns feminine gender to
 Gender and its relation to sex: The myth of ‘natural’ gender       5

a word designating referents that are designated by masculine gender words in

another language. Why, asks many a naive language learner, do nouns designat-
ing ‘table’ and ‘moon’ get assigned feminine gender in French (la table, la lune)
and masculine in German (der Tisch, der Mond)? We can even find words desig-
nating the same referent that are assigned different genders in a single language.
In German ‘head’ (in the sense of the thing on the top of the neck where we find
eyes, nose, ears, and mouth) these days is usually der Kopf (masculine) but was
earlier and poetically das Haupt (neuter), still used in some contexts.
Such cases establish clearly a strong conventional, ‘non-natural’, component
to gender class assignment. (In some cases, as Corbett suggests in his chapter in
this volume, there are formal principles at work.) In contrast, English is said to
be a language with ‘natural’ gender. What this is supposed to mean is that sex
of a noun’s referent is what determines gender agreement – in English, amount-
ing just to the form of a referentially linked pronoun. A little reflection shows
some of the problems here. What about epicene nouns, which refer to humans of
either sex? When, for example, we speak of a specific child (say, Lee’s youngest
child), the pronoun chosen will depend on the gender ascribed to its referent. But
if we talk about any child and have a linked pronoun, the gender of that pronoun
cannot be determined on the ‘natural’ basis of sex, given that children are some-
times female, sometimes male. In this case, the ‘rules’ of prescriptive grammar
say to use he, but other choices, discussed later, are common. Even if prescriptive
rules always held sway, thus fully conventionalizing the decision in cases where
sex is for some reason not determinate, more is going on than ‘natural’ agreement
with biological sex. English has gendered pronouns used in other contexts where
the antecedent either does not provide a definite referent and allows for either
sex or where referents provided are of mixed sex or not sex-differentiable. As we
will see later, these ‘indeterminate’ cases are just immediately obvious instances
of the non-naturalness of gender in English. English pronoun selection is not a
matter of ‘natural’ but, as I have already indicated, ‘notional’ gender.
What I will also argue is that even conventionalized grammatical gender
systems often connect to sex in non-arbitrary but non-natural ways, drawing on
notional gender though doing so in different ways than languages like English.
Research by psychologist Lera Boroditsky and her colleagues (e.g., Boroditsky,
Schmidt, and Phillips 2003) strongly suggests that speakers do treat grammati-
cal gender classifications even of nouns outside the conceptual core (e.g., those
designating inanimates in Indo-European languages) as notionally relevant.
Even though French speakers don’t think of moons or tables as female beings and
German speakers don’t take them to be male beings, associating feminine and
masculine characteristics with feminine and masculine gendered nouns does
occur more than chance would predict. The fascinating phenomenon of feminine
6       Sally McConnell-Ginet

(and, to a lesser extent, masculine) pronouns for inanimates in English gives

considerable support for treating gender in contemporary English as ‘notional’
rather than ‘natural’, but I can only point to it in this chapter; Curzan (2003) has
an excellent discussion of historical data, and Curzan (1999) is also relevant.
My focus will be on how gender systems notionally connected to sex classifi-
cations deal with human referents, the ‘home turf’ of gender for Indo-European
languages and the primary arena in which contemporary English pronoun usage
is said to depend on ‘natural’ gender. Sex is, of course, an important classifica-
tory principle in many languages. My thesis, however, is that it seldom stands
alone. So-called ‘natural’ gender almost always gives way to ‘notional’ gender,
both in languages like English where it dominates the show, and in grammatical
gender languages, where it is also plays a role, more limited (and varying across
languages) but important.
As I will argue below, grammatical gender systems involve matters of sex and
sociocultural gender more than is sometimes thought. My evidence comes from
work others have done involving a variety of languages with grammatical gender.
From surveying some of the relevant recent research, I conclude that even for lan-
guages with conventionalized grammatical gender, ideas about sex and sexuality
can interact in somewhat surprising ways with the gender system. Indeed, I suspect
that most grammatical gender systems in which sex plays some role have at least an
attenuated ‘notional’ (or ‘natural’) gender system as a part. This point was essen-
tially made for English in Moore 1921, quoted in Baron 1971: “natural [notional]
gender did not replace grammatical gender in Middle English but survived it.”
My own research, especially McConnell-Ginet ([1979], 2011), shows that
gender in English, while not ‘grammatical’ in the fullest sense because pronouns
are the only agreement targets, is not really ‘natural’ either. English-like lan-
guages have what I now call notional gender systems: pronominal usage cannot
be understood without considering sociocultural gender and the ideas about sex
and sexuality current at a given time. And it is such gender ‘notions’ that can
be embedded in and affect agreement phenomena, especially but not only pro-
nouns, even in languages where grammatical gender predominates.
Notional gender systems like that in English have long been called ‘natural’
because linguistic form seems to link to sex quite directly: there’s no need to learn
what can look from the outside like arbitrary nominal classes in order to choose
the correct pronoun. The choice of an English pronoun is apparently determined
by the sex that the speaker attributes to the referent of its nominal antecedent or
of the individual designated deictically (cases where something like a demonstra-
tion establishes reference rather than an antecedent noun phrase). If someone is
referred to using words like woman, wife, girl, or daughter, words that semanti-
cally indicate femaleness of those so denoted, then a pronoun for subsequent
 Gender and its relation to sex: The myth of ‘natural’ gender       7

reference to the same individual is standardly the feminine she. Similarly, man,
husband, boy, or son are semantically male and can head antecedent nominals
for the masculine he whereas turnip, fish scale, shoe, and stapler, none of which
designate sexed or even animate referents, all select the neuter pronoun it, which
designates nonsexed referents, including all inanimates.
Of course, what I have just said ‘usually’ holds. Even my introductory linguis-
tics texts admitted that there were some exceptions in English to this ‘natural’
arrangement in which sex and gender supposedly coincide. For example, inani-
mates like boats and cars, which have no sex, are (still) often referred to with
feminine pronouns, and there are also cases in informal speech where inani-
mates are referred to using masculine pronouns. And the choice of pronouns
with nonhuman animals is quite a mixed bag: some, e.g., always use masculine
forms for dogs and feminine for cats, a practice that pet-owners invested in their
female dogs or male cats often deplore. Although such practices do indeed shed
some light on gender ideologies, they lie outside the scope of this chapter. Actual
English pronominal usage even just for human referents is both more interesting
and complex than is evident from the usual stories told of its supposedly ‘natural’
gender system.
We have often felt comfortable calling gender in English ‘natural’ (and some-
times using that term also for certain phenomena in languages whose gender
systems are primarily grammatical) because choice of English gendered pro-
nouns is seldom surprising in light of familiar cultural ideas about sex and sexu-
ality. Gender ideologies have overwhelmingly viewed gender beliefs and arrange-
ments as ‘natural’, as dictated by biological imperatives, generally not realizing
that sociocultural gender is far from static and unchanging but varies historically
and crossculturally. Curzan (2003, 30) agrees that notional gender

may better capture the psychological and social aspects of gender assignment in the lan-
guage. It is possible, however, and I think pragmatically preferable to retain the description
natural gender with the understanding that its definition rests not purely on biological sex
but instead on social concepts of sex and gender [i.e., on sociocultural gender]. … [G]iven
the clear correlation between linguistic and social gender, and the growing understand-
ing of what the latter involves, the description natural gender could naturally [italic added]
come to encompass and appropriately refer to both biological sex and the social construc-
tions engendered by it.

Corbett (1991: 32) … notes that semantic gender categories reflect the world view of speak-
ers … it is possible to predict variation to some extent given knowledge of extralinguistic
factors. Instances of gendered anaphoric pronouns that cross biological lines are not excep-
tions to an underlying “real” or “unmarked” system of natural gender; they are part of a
natural gender system which is natural because [italic in original] it corresponds to speak-
ers’ ideas about and constructions of gender in the world about which they speak.
8       Sally McConnell-Ginet

What Curzan has done here is provide an excellent account of why calling
such gender systems natural has persisted even among those who know well that
biological sex is by no means the only factor in pronominal choice. But I fear
that there is still far less understanding of sociocultural gender than is needed,
particularly of its contingent and historical character. Perhaps most important,
natural is still too readily understood as encoded in our genetic make-up, the
inevitable outcome of human biology; linguists may better understand sociocul-
tural gender than they once did, but even among that group the term natural can
mislead. For linguists, a shift to ‘notional’ gender positions us better not only for
understanding languages like English in the early 21st century but also for explor-
ing changes not only in English-like languages but also in those with grammati-
cal gender systems.

2 Grammatical gender

2.1 R
 eferential pressures and (dis)agreement

Many (perhaps most) words apparently referring semantically to female humans

are assigned to a single gender class (called feminine in Indo-European lan-
guages) and many (perhaps most) referring semantically to male humans are
assigned to a different single gender class (called masculine in Indo-European
languages). Furthermore, languages generally use the same personal pronouns
for anaphora (i.e., where there is an antecedent for the pronoun) and for desig-
nating an individual indicated deictically or through a proper name (which might
or might not be gender-marked).
Specifically demonstrative pronouns do occur in a number of languages but
in English their use to refer to human beings is limited to identificational con-
texts such as who is that? or this is my sister but not *this is married to Lee or
*I haven’t been introduced to that. Even in languages where demonstrative pro-
nouns can readily be used for human referents, anaphoric pronouns can, so far as
I know, also be used deictically, and they are often a primary resource for deictic
reference. Not surprisingly, then, in grammatical gender languages deictic refer-
ences to human beings to whom the speaker attributes female sex typically use
feminine pronouns, whereas masculine pronouns are standardly used for deictic
references to human beings to whom the speaker attributes male sex. I qualify
this claim because of the expressive uses discussed below, but what the claim
amounts to is that so far as deictic pronouns go, grammatical gender languages
often look very English-like, which is to say that sociocultural gender is what is
 Gender and its relation to sex: The myth of ‘natural’ gender       9

relevant for pronominal form. Ivan Sag (pc) points out that this fits well with the
analysis of agreement that he and Carl Pollard have developed (see Pollard and
Sag 1994, chapter  2), which takes agreement in gender (as well as person and
number) to arise from co-indexing. English, they propose, has a pragmatic con-
straint that she can only be anchored by an entity presented as female, he by an
entity presented as male, it by an entity presented as inanimate (or at least as not
sex-differentiable).
Notice that deictic pronominal references to inanimates in languages like
French that have only masculine and feminine gender must choose a gendered
pronoun without any direct linguistic controller. How this happens might seem
mysterious, but Corbett 1991 plausibly hypothesizes that what Rosch 1978 calls
a basic level expression comes into play and its gender is what determines the
pronoun chosen. Pollard and Sag 1994 agree that an expression is involved but
note that there may not always be a single ‘basic level’ expression to do the job.
They cite (p.  78) Mark Johnson’s 1984 unpublished discussion of grammatical
gender and pronoun reference, which notes cases where alternative forms might
be available. Johnson observes that certain small structures could be designated
in German by either das Haus ‘the house’, neuter, or die Hütte ‘the hut’, femi-
nine, the latter suggesting something small and rundown. Interestingly, the same
suggestion could be conveyed by using the feminine deictic pronoun sie to refer
to the structure rather than the neuter es. The pragmatic constraint operative in
grammatical gender languages, according to Pollard and Sag, is that an entity
can ‘anchor’ a pronoun’s index only if the features of the index, which include
gender, match those of some common noun “that effectively classifies that entity
at a level of granularity appropriate to the context” (78, italics added), which
allows for competing expressions in some settings.
Deictic usage is important because when there is some kind of conflict in
grammatical gender languages between gender of a controlling noun and attrib-
uted sex of the individual designated by a target pronoun, the pronominal form
appropriate for target sex is often chosen, and this is more frequent the greater the
distance of the pronoun from its controller. Corbett 1991 speaks of ‘hybrid nouns’,
those that permit variable gender agreement where that variability is sensitive to
the kind of target involved. He posits an agreement hierarchy in which attributive
modifiers (in a very ‘close’ syntactic relation to the controller) outrank predicates,
which in turn outrank relative pronouns, with personal pronouns least likely to
agree with the gender of the controller. So if someone speaks of the masculine le
professeur and is referring to Julia Kristeva, a distinguished woman, then subse-
quent pronominal references are highly likely to use elle rather than the il that my
French teacher in high school insisted upon. There seems to be considerable room
for referential leakage from deictic practice, leading to what looks like gender
10       Sally McConnell-Ginet

disagreement between antecedent and anaphoric pronoun when attributed sex

of the referent does not match grammatical gender of the supposedly control-
ling antecedent. Such gender disagreement under deictic pressures is reported
for a wide range of grammatical gender languages (including, as documented in
Curzan 2003, Old English). Details vary significantly, however. Formal agreement
pressures appear much stronger in some languages than others (in, e.g., Czech,
Polish, and Greek as compared to German and French, according to descriptions
of these languages in Hellinger and Bussmann 2003). (Such differences may be
linked to the same kind of grammatical differences that affect the gender con-
gruency effects in language processing that Schiller reports in this volume.) Yet
disagreement, although probably more common in informal language usage than
in ‘careful’ writing, is always lurking as potentiality, presumably because deictic
practices strengthen the salience of attributed referent sex in pronominal selec-
tion.
To bring home my point that it is sociocultural or notional gender that is at
stake in gender (dis)agreement, let’s consider again the German Das Mädchen
‘girl’, another ‘hybrid noun’ in Corbett’s terminology, one that requires (at least
in formal Standard German) neuter agreement for attributives and relative pro-
nouns but allows feminine personal pronouns (providing evidence for the agree-
ment hierarchy). The -chen suffix is a diminutive, and Corbett notes that diminu-
tives are often assigned to neuter rather than feminine or masculine genders,
a phenomenon related to the frequent assignment in IndoEuropean languages
of the young of ‘sex-differentiable’ categories to neuter gender. What struck me
particularly in Corbett’s discussion of Mädchen is his mention of a suggestion
from semanticist Manfred Krifka that the older a girl, the more likely the feminine
personal pronoun sie will be used (conversely, the younger the more likely she
will be referred to using the neuter es). Now age does not make someone who
is straightforwardly a biological female on all counts (genetic or chromosomal,
hormonal, genital) any more a female. As a girl matures and moves towards men-
arche and potential fertility, however, the sociocultural significance of her female
sex certainly does increase. Sex as such is not what matters here but sociocul-
tural gender considerations are coming into play in (variably) conditioning the
form of the personal pronoun. Grammatical gender is not, of course, irrelevant.
It might well be that it is more likely that someone will be referred to using es if
Mädchen antecedes the personal pronoun than if that same individual is referred
to deictically (by the same speaker in a similar context) – i.e., with no anteceding
nominal. Nonetheless, sociocultural gender often puts pressures on grammatical
gender systems, with the precise details dependent on many specific linguistic
and other factors but with pronominal reference a particularly vulnerable point,
probably in large measure because of deictic usage. It is tempting to speculate
 Gender and its relation to sex: The myth of ‘natural’ gender       11

that a notional pronominal gender system lies embedded in any grammatical

gender system, its impact and pervasiveness depending at any given time both
on language structure and history (long established conventions take on lives of
their own) and on language and gender ideologies.

2.2 F eminist–inspired change in languages with grammatical

gender

In languages with feminine and masculine grammatical genders, nouns denot-

ing male humans exclusively are overwhelmingly masculine gender, and nouns
denoting female humans exclusively are overwhelmingly feminine gender.
Epicene nouns, those which can readily denote humans of either sex, can often be
assigned to either grammatical gender class (assignment may depend on phono-
logical factors in such cases). So long as there is a sharp sexual division of labor,
however, the range of genuinely epicene nouns is limited, and occupational ter-
minology and social titles will be gendered. In such circumstances, women and
men occupy very different positions and do very different things so that speakers
are likely to find the gender assignment easy, ‘natural’ in some sense; of course,
it’s le maçon ‘mason’ and le médecin ‘doctor’ in the masculine and la dentellière
‘lace maker’ and la sage-femme ‘midwife’ since it is men who are masons and
doctors, women who make lace and practice midwifery. Yet when gender arrange-
ments are changing, especially when women are moving into positions and jobs
once reserved for men (movement in the other direction, though it does happen
occasionally, is less common and typically less prestigious), grammatical gender
often strains at the seams, perhaps in part because of deictically driven disagree-
ment.
Burr 2003 offers a very interesting account of linguistic politics in France. In
France the major thrust in attempted language reform has been the feminization
strategy that tries to make women visible as they enter into public life in positions
previously reserved exclusively for men. It was in 1997 that six of the eight woman
nominated by the new prime minister, Lionel Jospin, put forth a circular demand-
ing that they be addressed as Madame la Ministre and had stationery drawn up
with feminine job titles and signs changed on their office doors to feminine forms.
And, importantly, Jospin made sure that indeed feminized designators would be
used for top women in his government, issuing a circular to that effect in early
1998 and drawing strong condemnation from the Académie Française and others
rushing to defend the French language from this latest assault.
The start of the Jospin government was by no means the first time that
the topic of making women linguistically more visible was raised in France but
12       Sally McConnell-Ginet

this particular moment had far more effect than earlier efforts. Burr explains
(120)

[C]hange could only come from the very top of the hierarchy. … [R]eluctance to use feminine
denominations has always been strongest with respect to high-level professions, functions,
grades and titles traditionally reserved for men, and women who had themselves reached
the top of the hierarchy were among the fiercest opponents of this use. … [Nothing could]
effect a real change so long as this hostile attitude towards feminisation prevailed among
almost all high-level women.

Indeed a terminological commission had been formed in 1984 by Yvette Roudy,

who was charged with developing policies to promote women’s rights. That
group, which included a number of linguists, proposed femininizing via allowing
hitherto masculine forms optionally to take feminine agreement  – i.e., making
previously masculine forms double-gender (allowing, e.g., la professeur along-
side le professeur) – and also in some cases through creating new feminine nouns
(la maçonne alongside le maçon). With new feminine forms there were also new
feminine plurals (e.g., les maçonnes in addition to les maçons) whereas the double
gender strategy resulted in a single plural form (e.g., les professeurs). The commis-
sion also proposed masculine forms to partner with feminines as in sage-homme
for any male midwives who might present themselves. The rules published in
1986, however, dealt only with feminization, prescribing its use in official docu-
ments and providing rules for producing feminine forms, specifying the double
gender strategy with forms ending in e muet ‘silent e’ and some others. There are
potential hazards even with specific rules. In some cases existing feminine forms
paired with masculine agentives are, as Connors 1971 had noted earlier, already in
service for other purposes. Adriaen and King 1991 (32) offer examples like these:
un trompette ‘a male trumpet player’ vs une trompette ‘a female trumpet player’
or ‘a trumpet’ and un manoeuvre ‘male laborer’ vs une manoeuvre ‘female laborer’
or ‘a maneuver’.
These rules were slightly modified in 1999 in guidelines from the Insti-
tut National de la Langue Française. At the same time, the 1999 document took
account of developments in Canada (see King 1991), Switzerland, and Belgium,
offering more lexical options than the 1986 document and also showing more
appreciation of why such issues might matter, something that did not happen
during the discussion of the 1980s. The first female prime minister of France,
Edith Cresson, was virtually always referred to using masculine gender forms,
le premier minister ‘prime minister’ or le chef du gouvernement ‘head of govern-
ment’. It’s probably no accident that focus on this issue required some male lead-
ership as well as female: the first women in high-level positions have special pres-
sures to show that they are doing exactly the same job (or perhaps even better) as
 Gender and its relation to sex: The myth of ‘natural’ gender       13

their male predecessors and having a different job title might seem to suggest oth-
erwise. We’ll return to this potential pitfall of feminization in our later discussion
of English, but it is important to acknowledge that the strategy has apparently
enjoyed some success not only in France but also in a number of other countries
where grammatical gender languages are used.
Of course, actual usage continues to be very mixed and generally is guided by
social and political rather than linguistic gender ideologies. Burr concludes (132):

That the way women are addressed and talked about does matter in French society can
be seen by the fact that every time the question of feminine personal nouns arises there is
loud protest from [some/many] men, from patriarchal institutions like the French Academy,
and from women who either accept the structures and values men have created or who are
afraid of losing their face, being ridiculed or attacked. … [T]he specific linguistic means …
are … secondary in nature in comparison with the fundamental question of the social func-
tion of change.

Not surprisingly, when reference might be to sexually mixed groups or to indi-

viduals of unknown sex or generic and not really to individuals at all, the various
groups put in charge of directing linguistic change in French in pursuit of social
gender equity have all opted for masculine gender. Burr comments that even
many feminists perceive plural masculine forms as generic, not really masculine.
In contrast in some other countries using languages with grammatical gender
(including other French-speaking countries), issues of masculine generics and
of principles dictating masculine plurals for groups of mixed or indefinite sex
have received considerable public discussion. I will discuss the issues of the sup-
posed ‘unmarkedness’ of masculine gender only when I turn to English, where
the push has generally been for neutralization rather than for the feminization
of occupational terminology and other personal nouns that has predominated
feminist-inspired interventions in European languages with grammatical gender.

2.3 E
 xpressive uses of grammatical gender: Beyond sex

Deixis is not relevant only for choice of third-person pronominals. A language

can have sex-invariant first- and second-person pronouns but show feminine
or masculine agreement on predicates depending on whether the speaker or
addressee is being ascribed feminine or masculine gender. In English and other
languages where gender is a matter of third-person singular pronominal form
(and of some semantically feminine or masculine nominals), saying something
about the speaker or the addressee need carry no gender messages. In contrast,
the French speaker uttering je suis heureuse ‘I’m happy’ is self-attributing femi-
14       Sally McConnell-Ginet

nine gender whereas one uttering je suis heureux ‘I’m happy’ self-attributes mas-
culine gender. And second-person forms that predicate adjectives of the subject
(whether expressed explicitly or, as in imperatives, evoked implicitly) take gen-
dered forms that depend in the most straightforward cases on the sex attributed
to the addressee. What matters for our present purposes is that languages with
sex-linked grammatical gender provide rich resources for challenging gender
binaries and for policing gender boundaries not only in speaking of third person
individuals but also in speaking of the self or of an addressee.

2.3.1 Gendering minority identities in French literature

Anna Livia (2001) focuses on literary uses of linguistic gender and includes a
chapter devoted to what she calls ‘liminal identity’. There she includes a detailed
account of the exploitation of the French gender system in Appelez-moi Gina, the
autobiography of Dr. Georgine Noël, born Georges in Belgium and assigned male
sex at birth and reared as a boy but from a young age identifying as female and
finally at age 39 undergoing sex reassignment surgery. Of particular interest are
the switches in gender Noël uses in speaking of the self, which proceed far less
straightforwardly than might be thought. As Livia puts it (168),

Noël uses the binary opposition of the French linguistic gender system throughout her auto-
biography to express or underscore many of her changes of mood, attitude, and identifica-
tion, an expressivity that goes far beyond the simple polar opposites of “binary thinking.”
Appelez-moi Gina demonstrates how a structural binary may be subverted into expressing
more than one simple opposition.

Importantly, her initial announcement of her birth evades gender by using the
passé simple (je naquis), which though not very colloquial does not mark gender,
rather than the more frequent passé composé (je suis né, m, or je suis née, f). She
apparently wants to resist, at least in this narrative, sex assignment of that baby
from whom she has become an adult. Before beginning a basically chronologi-
cal narrative of her life, Noël does speak, using conditional verb forms, of what
might have been had she undergone sex reassignment surgery at age 15 (Livia
2001, 169).

(1) Heureuse d’être femme et épouse, j’aurais aimé adopter les enfants.
Happy to be a woman and a wife, I would have liked to adopt children.

This is a hypothetical femininity, but in recounting her early childhood she uses
masculine forms (170–171, italic added).
 Gender and its relation to sex: The myth of ‘natural’ gender       15

(2) a. J’étais un en­fant souffreteux.

I was a sicklym child.

b. J’étais bon élève.

I was a goodm pupil.

c. Engoncé dans mon costume trois pièces

Stuffedm into my three-piece suit

Again and again she speaks of herself as seul ‘alonem’, noting how out-of-place
she felt with male classmates, but then she says

(3) La fille cherchait à naître isolée au milieu de tant de garçons

The girl was struggling to be born, isolatedf among so many boys

It is only after she decides to keep a diary in code, not trusting her parents, that
her account includes recurrent uses of first person feminine without the protec-
tion of the conditional form. Particularly telling is her account of following the
male friend Edgard for whom she saw her affection as a heterosexual attraction:

(4) Je suis revenue dans la cour avec Edgard.

I came backf into the playground with Edgard.

She tells of her parents’ discovery of her diary and their speaking of her in the
masculine while she listens from the next room, speaking of herself in the femi-
nine. But in an expression of disgust at her body she resumes the masculine

(5) Sûr que ce n’ était pas à moi ce machin-là

Surem that it wasn’t mine that thinggummy-jig there

Yet when she learns of the possible availability of surgery and hormone treatment
that would make her body conform to her self-perception she returns to feminine
self-description, punctuated with occasional masculine renderings of frustration
at her actual constrained situation, too young and too poor to embark on medical
sex reassignment. Eventually she leaves home, going first to medical school and
then to Africa. Although she wears women’s clothes only in the privacy of her own
quarters, her self-descriptions are consistently in feminine form. The one striking
exception, Livia notes, occurs when Noël is speaking disparagingly of the African
men being treated for AIDS after having in a careless drunken moment shown
a feminine side to European colleagues by appearing in women’s clothes. Such
slippage, Noël notes, must be avoided in the future.
16       Sally McConnell-Ginet

(6) Né homme, je devais continuer à manifester ma supériorité sociale et

sexuelle.
Bornm a man, I had to continue to show my social and sexual superior-
ity.

The masculine past participle né contrasts sharply with the nongendered inflected
verb form naquis used when the birth is first mentioned. A final example comes
from Noël’s discussion of her surgery. Although Noël has identified as a woman
for more than 25 years, masculine forms describe the self awaiting surgery (rev-
eillé tôt le matin ‘waked upm early in the morning’) and feminine for the one
recovering (couchée sur le dos ‘lyingf on my back’). Noël sums up that event (168):

(7) Avant j’étais un transsexuel; maintenant j’étais une femme.

Before I was a transsexualm; now I was a woman.

Yet the use of gender as she narrates her past does not show such a sharp bifurca-
tion; the preoperative self is presented sometimes as masculine but often as femi-
nine, and gender choice expresses Noël’s and the larger social views of sex class.
That is, deictically anchored gender agreement in French does not just fall out
from sex of an individual. For first-person discourse it may express the speaker’s
sometimes conflicted views of their own relation to sociocultural gender systems.
Noël’s autobiography is just one sample of the works Livia examined in this
general category of transsexuals. What is so striking is that transsexuals in speak-
ing of themselves do not just switch gender marking at some point but there is
considerable vacillation, especially in speaking of themselves prior to reassign-
ment of sex. And the same thing holds when others speak of them in the third
person but it is first-person discourse where grammatical gender languages offer
gender-related possibilities and also constraints not present in languages like
English, where gender-marking is a third-person phenomenon.
Livia also examined writings of and about people who challenge gender
binaries in other ways, e.g., people who fall in the small group of those for whom
assignment to either of the standard two sex categories is problematic, often
called intersexual or intersex. Anne Fausto-Sterling (2000) estimates that for
some 1.5 to 2% of babies, standard criteria give mixed results for sex assignment
due to chromosomal configurations other than the standard XX or XY, hormonal
conditions of various kinds, or non-canonical genitalia. In spite of the absence of
definitive critieria, however, the practice has long been to assign female or male
sex to the child, often also prescribing various medical procedures to try to make
recalcitrant bodies conform to the binary imperative. As Fausto-Sterling (2000,
3) explains:
 Gender and its relation to sex: The myth of ‘natural’ gender       17

labeling someone a man or a woman is a social decision. We may use scientific knowledge
to help us make the decision, but only our beliefs about gender – not science – can define
our sex.

Not surprisingly, like transsexuals, some people forced in infancy into a sex class
in order not to rock the binary boat do not fully identify with the sex to which
they were assigned and may eventually embrace the other sex identification.
Such was the case of Herculine Barbin, reared female but eventually embracing a
male identity. Michel Foucault edited Barbin’s memoirs and it is very interesting
to read them with a focus on Barbin’s use of gender to express shifting and con-
flicting gender identifications through childhood and adolescence.
Livia also considers writings of and about people who put themselves in the
same sex category as that to which they were assigned at birth – as she puts it
(185–186), their sexual anatomy and gender identification are not conflicted – yet
for whom same-sex desire leads them to reject the ‘compulsory heterosexual-
ity’ so firmly embedded in gender ideologies in many sociocultural and histori-
cal settings (though not all – see, e.g. Cameron and Kulick 2003, 22, of ancient
Roman sexualities). She speaks of “cross-expressing” of gay men and lesbians,
noting that “[a]t a time when … gender distinctions in nouns and adjectives in
French are tending to disappear, many gay men [sometimes] use feminine terms
to address or refer to others in the community [sic], including their lovers and
themselves” (188) and citing a 1994 sociolinguistic survey reported in Pastre 1997.
Livia observes that “this linguistic strategy is intended not to reflect a feminine
persona so much as to dissociate the speaker, addressee, and any third party
described in the feminine from heterosexual alliance and to create a homosexual
alliance … [These men] are proudly designating themselves as well as the refer-
ents as traitors to heterosexual masculinity.” (188)
Although Livia’s analysis of the significance of particular gender choices may
sometimes be open to debate, her larger point clearly holds. “Grammaticalized
gender … provides linguistic devices for expressing gender fluidity.” (192) It is not
that there is some particular gender identity that gender choices index but that
gender choices offer a way to position oneself outside socioculturally enforced
gender binarisms.

2.3.2 Gendering the hijra minority identity in Hindi

Another kind of liminal identity is involved in cases where alternatives to female

and male identities are institutionalized. Kira Hall with various colleagues has
written extensively on the hijras, an Indian group who early in life are treated as
18       Sally McConnell-Ginet

members of a third ‘sex’ – or at least as not either (fully) female or male. Some
(by no means all) were born intersex, most of them are raised male, many but not
all undergo voluntary penectomies on entry into the hijra community, and they
adopt female attire and many other aspects of female appearance and behavior,
including speech  – at least to some extent and in some respects (e.g., there is
frequent commentary on their ‘masculine-sounding’ voices and on the ways they
use obscenities). Most hijras, however, describe themselves as na mard na aurat
‘neither man nor woman’. Hijras typically live communally, often with a guru
heading a family-like unit, and there are also much larger networks and a rela-
tively recent political awareness and organization to assert rights such as being
able to vote ‘as women’ or even hold political offices reserved for women. Hijras
often go in groups to weddings or to households where a son has been born and
dance and sing, sometimes on invitation but more often not. Families pay them
to get hijra blessings bringing the promise of many sons and general prosper-
ity (or, depending on perspective, to avoid the hijra curses, believed by many to
bring infertility and other misfortunes). The word hijra literally means ‘impotent’,
but the belief in hijra powers, though still widely held, is waning in many Indian
communities. Many hijras do sex work (though some activists claim ascetic spiri-
tuality), some threaten curses or obscene behavior like showing their genitals to
claim alms from passers-by. Most hijras lead economically precarious lives and
experience considerable social ostracism from outsiders: they find support and
affection within the hijra community.
There are hijras across India from many different language groups, but Hall
and O’Donovan 1996 and Hall 2003 focus on the speech of several Hindi-speak-
ing hijras in Banaras. In particular, Hall 2003, which I only read well after I had
begun writing this chapter (having mistakenly assumed that it mostly repeated
Hall and O’Donovan 1996), makes quite explicit a major point of this chapter: that
‘natural’ gender is a myth.
She sets out to

challenge the very assumption implicit in the term “natural gender”, i.e., that gender is
a fixed phenomenon, rooted in biology and therefore free of ideological influences. What
happens to a language’s classification system in instances when the referent’s gender
can no longer be assumed as either male or female? And what might these instances of
“unnatural gender” tell us about the relationship between gender in language and gender
in society? (137–138)

I’m challenging the term ‘natural gender’ on other grounds as well, but Hall’s
formulations and her linguistic data are very relevant. Hall 2003 includes con-
siderable background on the linguistic varieties grouped under the label ‘Hindi’
and some discussion of prior sociolinguistic work on Hindi so it is a rich resource
 Gender and its relation to sex: The myth of ‘natural’ gender       19

for present purposes and helps expand conceptions of the varied ways ‘notional
gender’ can affect usage in a language with grammatical gender.
Except for some pidginized or ‘contact’ forms of Hindi, all varieties of Hindi
have feminine and masculine grammatical gender. Details of the gender system
vary across varieties, however. In some dialects feminine gender is now reserved
for nouns designating female referents and has been lost for inanimate nouns.
And many varieties show far less extensive gender concord than might be found
in ‘standard’ Hindi, where the system of concord extends even to postpositions
(like the word translating English ‘of’, which is ka when modifying a singular
masculine, ke with plural masculine, and ki for both singular and plural femi-
nine). Furthermore standard Hindi requires not only gender concord between
subject and predicate adjectives but also gender-marked endings for verb forms
in all three persons and both singular and plural, requiring first-person utter-
ances to identify the speaker as masculine or feminine and second-person to
assign gender to the addressee. In Banaras, where Hall worked, however, the
local Banarsi Bhojpuri variety, used alongside standard Hindi, makes much less
use of gender concord, and extensive use of gender-marking is often heard as
‘foreign’ or ‘over-educated’. And Hall reports a study with evidence that such
attitudes toward the strict gender-marking of standard Hindi are most prevalent
among poorly educated women.
Hijras in Banaras, however, generally come from elsewhere and are said
to speak Hijra Boli, which is in many ways is like a lingua franca or pidginized
variety. Where Hijra Boli differs strikingly from pidginized forms of Hindi reported
for the streets of Bombay or Calcutta, where grammatical gender is essentially
absent, and also from Banarsi Bhojpuri, where grammatical gender is on the
decline, is in its strong emphasis on grammatical gender. Hijras adopt, e.g., the
verbal morphology that marks their first-person statements (usually) feminine or
(less often and for special effect) masculine. As Hall puts it

[T]he variety of Hindi adopted by the hijras tends to overemphasize gender, using mascu-
line and feminine gender in places where it normally would not appear … or treating nouns
that are masculine in standard Hindi as feminine and vice versa. To give one illuminating
example: The word hijra is grammatically masculine in standard Hindi, but the hijras fre-
quently treat the noun as feminine through verbal agreement when it acts as subject of the
sentence. As I argue here, this usage reflects a kind of gender overcompensation or even
hypercorrection. Upon entering the community, Benaras hijras work to distance themselves
from masculine representations. … The fact that the word hijra is grammatically masculine
sometimes gets in the way of this communal distancing, so hijras will mark the noun as
feminine as part of “doing gender.” … [G]rammatical gender is most often overemphasized,
not underemphasized, in the hijras’ construction of a “more feminine” self. (139–140)
20       Sally McConnell-Ginet

Non-hijras in Benaras were often shocked in helping Hall transcribe tapes

to hear a hijra use morphological resources to mark herself or another hijra as
feminine for they thought of the hijras as male, even if imperfectly male, and
these usages seemed highly ‘unnatural’. Not surprisingly, gender marking in first
person is part of what hijras count as zanana boli ‘feminine language’, something
they adopt along with saris and female names. Of course eschewing mardana boli
‘masculine language’ is significantly harder than putting on a sari rather than a
lungi (cloth wrap worn by men in northern India). The hijras themselves spoke of
zanana boli as less direct and more intimate than mardana boli, which of course
would involve much more than gender agreement in first-person utterances.
Those who have been within hijra communities for at least six months gen-
erally do indeed adopt the feminine gender markings for themselves and other
hijras that are viewed as shocking by outsiders: more informative in many ways
are instances of hijras marking themselves or other hijras as masculine. Least sur-
prising, perhaps, are uses of masculine gender for reference to the pre-hijra self,
which are not infrequent. And hijras occasionally live in non-hijra households
where they find it more comfortable to present themselves ‘as men’, wearing male
clothes and masculinizing their first-person utterances as they did in boyhood.
There are also frequent masculine references to new initiates, those who are still
acclimatizing themselves to feminine ways of doing things. But, perhaps more
surprising, there are also sometimes masculine references to other hijras of
whom the speaker disapproves, a distancing maneuver. Yet masculine reference
can also be used positively – e.g., of a hijra guru much respected by the speaker or
in self-reference in order to strengthen a statement. First-person masculine refer-
ences also occur in angry speech. Hall suggests that hijras draw on cultural ide-
ologies of femininity and masculinity at the same time they also challenge them,
using grammatical gender as a powerful resource.

3 English and ‘natural’ gender

Anne Curzan (2003) opens her very interesting book on the history of gender in
English with an account of the American Dialect Society’s choice in 2000 of she
as the word of the millennium then closing. She, first cited in 1154, represents, as
Curzan puts it, “change at the very core of the English vocabulary.” Phonologi-
cal shifts had by then been conspiring to render inaudible the distinction once
carried by the contrast between heo ‘she’ and he ‘he’ (inanimate hit ‘it’ stayed
quite distinct). Somehow she, whose origins are still debated but probably lie in
one of the Scandinavian languages that Anglo-Saxons heard around them, was
 Gender and its relation to sex: The myth of ‘natural’ gender       21

pressed into service, and English was spared the gender-impoverished fate of
those languages with a single pronoun for all animate beings.
It was none too soon, as the system of grammatical gender found in the Ger-
manic ancestors of English and in Old English was under serious threat by the
time she entered the language. Grammatical gender persisted for several centu-
ries, however, in some regional varieties though Chaucer’s English did not show
evidence of the grammatical gender system of its ancestors. No matter where it is
spoken, contemporary English is a notional rather than grammatical gender lan-
guage, with pronominal selection depending not simply on sex but on sociocul-
tural gender, ideologies and ideas about sex. Gendered pronouns for inanimate
referents and gendered pronouns that don’t take account of sex for non-human
animals might seem the most striking ‘non-natural’ aspects of contemporary
English gender, but, as I noted earlier, even restricting attention to human refer-
ents shows us that what matters is less sex as such than attitudes linked to sex
and sexuality.
What I was told in English classes many years ago was that we use he for indi-
vidual male humans, she for individual female humans, and they for groups of
whatever sexual composition. Well, that was basically the story given for deictic
uses. Where there was a linguistic antecedent, choice of pronoun was said to
depend on the semantics of that NP. A singular NP semantically denoting female
humans (girl, sister, daughter, wife, woman, compounds with –woman, and a few
other forms) takes she. Singular NPs semantically denoting male humans (boy,
brother, son, husband, and some others) select he. But, as noted earlier, that
leaves unanswered what happens in other cases: an epicene antecedent where
the referent’s sex is unknown to the speaker (e.g., Lee’s new tenant), a disjunctive
(“split”) antecedent with a feminine and a masculine disjunct (e.g., my mother
or my father), a ‘notionally’ plural but grammatically singular antecedent (e.g.,
everyone), or, of course, a singular generic (the careful shopper or an intelligent
child, where these are used to say something about careful shoppers or intelli-
gent children generally). Some said that, although she is marked feminine, he is
simply unmarked for gender and this is why we get so-called masculine generics.
But that view cannot be maintained, as we will see.

3.1 R
 eferential practices: Enforced gendering

By and large, pronominal reference in English to a single specific human being,

whether involving a linguistic antecedent and its linked pronominal or a deicti-
cally anchored pronoun, involves a forced choice between the feminine pronoun
she or the masculine he, which in such uses is most definitely not neutral.
22       Sally McConnell-Ginet

(8) a. (pointing) What is she/he doing?

b. Someone called, but she/he didn’t leave her/his name.

The choice of she is marked feminine and attributes female sex (or at least notional
feminine status) to the person at whom I am pointing. And, contrary to the view
that he is simply unmarked, in such contexts he is marked masculine and attri-
butes male sex (or at least notional masculine status). It simply cannot be used as
gender-neutral or unmarked in references like these to a specific person, whether
deictic as in (9a) or controlled by a pronominal antecedent as in (9b). Now it is
possible to avoid gender attribution in such references but not easy.

(9) a. What is that person/are they doing?

b. Someone called but they didn’t leave their name.

To my ear the notionally singular they sounds better in the second case, but it is
certainly possible in the first – especially if the person in question is not readily
assigned masculine or feminine gender (perhaps too far away to see clearly or,
rarer, someone whose external appearance is sex-neutral). Still, they seems virtu-
ally impossible in certain contexts – e.g., when the antecedent is a proper name
or a notionally gendered noun.

(10) a. ?#Chris could tie their own shoes when they were three.

b. ?#Joan’s dad has Alzheimer’s and apparently forgets their own

birthday.

Notionally singular they has certainly extended its reach since I first began think-
ing about the meaning of English pronouns, but it is still not available for every
occasion when an epicene pronoun might be wanted.
Those who advocate gender-neutral policies for reference as essential for
gender equity (see, e.g., Beardsley 1973–74) almost universally fail to consider
how difficult it is to avoid gender-marking in third-person singular references
to specific people. The difficulty is especially acute in extended discourse about
the same person when repetition of a name (these too are very often gendered)
or description sounds very odd, sometimes seeming to suggest that a different
person is being discussed. In the current state of English, gendered pronouns
seem more or less inevitable. Of course, to speak of some other human, I choose
between she or he. How could it be otherwise? That’s ‘natural’. Or so it seemed to
me for a long time. But many languages do not in fact require such a choice, and
 Gender and its relation to sex: The myth of ‘natural’ gender       23

the potential pitfalls presented are beginning to be felt by substantial numbers

of English speakers. This is not to say that speakers of languages that make gen-
der-neutral reference easy do not consider the division of people into men and
women of great importance. It is just that their languages do not require them to
attend constantly when talking of others to the distinction between women and
girls, on the one hand, and men and boys, on the other.
English speakers have never had a straightforward gender-neutral pronomi-
nal choice for referring to another person. The neuter OE pronoun hit, still heard
in the North Carolina mountains where I lived during my early adolescence,
was pretty much reserved for inanimates by the time grammatical gender was
on the wane. There are some instances of hit in earlier OE sources controlled by
a nominal with human reference that was assigned neuter gender; e.g., Curzan
(2003, 62) reports the neutral wifman ‘woman’ anteceding the neuter pronoun in
two instances as compared to 16 feminine agreement patterns in one OE corpus.
But these most emphatically were not notionally gender-neutral uses  – they
simply represented grammatical gender agreement, which in the particular case
happened to conflict with notional gender.
Nowadays it is occasionally heard in reference to very young humans, and it
does sometimes serve generically with antecedents like baby or child. Here’s an
example:

(11) [T]he child comes to believe that its power to command … reflects its
superior value, and this is what awakens and flatters its amour-propre.
(Philosophical Review 119.2, April 2010, 184)

But it tends to sound dehumanizing if used in speaking about a particular

human, even a very young one. When told a baby has been born people often
respond with what is it?, to which the answer is supposed to be it’s a girl or it’s a
boy, and from then on they’re supposed to speak of she or he. Doctors are advised
when talking to parents of intersex babies to use something like your baby rather
than it.
Using it of a human being past infancy is almost always an act of derision, a
refusal to attribute fully human status and often an expression of disapproval of
that individual’s apparent attitudes towards sex and sexuality.

(12) Where did it come from?

With accented it, I have heard this and similar utterances used by people who
disapproved of the long-haired man or flannel-shirt wearing woman whom they
indicated. The pronoun it is simply not available as a neutral way to refer to adult
24       Sally McConnell-Ginet

humans. Its imputation of non-humanity, however, is certainly consistent with

the widespread view that true personhood cannot be gender-neutral, which
leaves intersex people in something of a limbo (and makes speaking of them chal-
lenging).
Being either she or he is not only considered essential to one’s humanness.
Sex is also considered a permanent personal characteristic, which indeed it is
for most people. But not for all and in particular not for transsexuals. Though
a first-person transsexual memoir written in English would not yield data like
that from Appelez moi Gina, trans men and women do often encounter problems
in getting long-term friends and family to refer to them as he or she, to speak
of them as they would of other men and women. Some professionals who work
with people undergoing sex reassignment use they to speak of them during the
period of transition. Even speakers who are trans-friendly encounter difficulties
when speaking of the person at times prior to sex reassignment: does one use the
pronoun that was then appropriate or the one that now fits? Nonetheless, first-
person reminiscences in English would avoid this difficulty.
It is no accident that the wonderfully comic crime fiction novels by Sarah
Caudwell featuring Oxford don of jurisprudence, Hilary Tamar, are first-person
narratives. In English, in contrast to many languages with grammatical gender,
one readily speaks of oneself in gender-neutral terms. Others speak to Hilary or use
the name – one never knows for sure whether Hilary is woman or man. Caudwell
manages this so skillfully that people reading the novels do not always notice the
absence of definitive gendering of Hilary: they sometimes mentally provide she or
he on the basis of whichever familiar gender assumptions happen to attract their
attention. But the novels are full of little hints that Hilary and friends do not lead
gender-conventional lives. Sarah Caudwell, sadly, died in her (his? – the author
used a nom de plume) early 60s, leaving open the question of Hilary’s sex.
Many possible epicene or gender-neutral third-person pronoun choices for
referring to humans have been devised for English over the years, not with much
success. Baron 1986 surveys neologisms proposed for this purpose before the
1980s, and Livia 2001 devotes a chapter to neologistic epicene forms in English-
language fiction. Such innovations have not had a very good track record, as
linguists often point out, but they do sometimes find considerable support from
various groups of nonlinguists. Because enforced gendering of standard singular
pronouns is particularly problematic for transgender people, especially in others’
accounts of their lives (as we noted above, there is an issue for others of which
pronoun to use for the person in their pre-transition life), they have been particu-
larly interested in finding new alternatives.
Gender-neutral options are attractive for other reasons, especially for those
who would like to challenge strict gender binaries. One fairly popular set of pro-
 Gender and its relation to sex: The myth of ‘natural’ gender       25

nouns in recent years is ze (ze is here), zir (we see zir), zir (is that zir coat?), zirs
(is that zirs?), and zirself (Chris has learned to get dressed by zirself); another set
includes E/e or Ey/ey for the nominative, em, eir, eirs, and eirself for the other cases.
These are called Spivak pronouns because mathematician Michael Spivak intro-
duced them; they have achieved some currency online and are easy to remember
because of their strong similarity to they, one reason they might become more
widely embraced. Whether some one of these gender-neutral neologisms actually
becomes widespread depends on factors that are difficult to foresee. The history
of English shows that personal pronouns can change: speakers added she when
the inherited female and male forms were losing their distinctiveness and aban-
doned singular thee/thou in favor of plural you for singular addressee reference.
If gender-neutral reference to specific individuals does become widespread, my
guess is that it will come through widening uses of they to include usages like
those in (10) but some neologism might indeed beat out they.
English speakers have been interested in what is happening in Swedish,
where the gender-neutral singular third-person pronoun hen introduced in the
1960s got a boost in the spring of 2012. Not only was it added to the Swedish
online National Encyclopedia as an alternative to han ‘he’ and hon ‘she’,
but it was also used in a children’s book about Kivi, “who wanted a dog for
hen’s birthday” (http://www.slate.com/articles/double_x/doublex/2012/04/
hen_sweden_s_new_gender_neutral_pronoun_causes_controversy_.html; site
accessed 11 June 2012). What distinguishes the Swedish case from most attempts
to add a gender-neutral option in English is that influential figures in govern-
ment and education seem to be urging that gendered pronouns be used no
longer (or at least this is how matters are interpreted by many observers). Such
apparent gender-neutral prescriptivism has enraged some Swedes and perhaps
even more non-Swedes. Many websites include comments both for and against
hen, sometimes but by no means always well-informed on linguistics or social
gender. This buzz in the blogosphere vividly illustrates the ‘passion’ associated
with gender in language of which Corbett, this volume, reminds us, with a quote
from Ranko Matasović.
It is often difficult, especially for those accustomed to think of enforced
gendering as ‘natural’, virtually inevitable, to imagine alternatives between the
constraints of enforced gendering and those of coerced gender-neutralization.
In principle, English discourses might eventually have some/many occurrences
of gendered pronouns alongside some/many occurrences of gender neutral pro-
nouns for individual human referents. Such changes could not happen overnight,
however.
26       Sally McConnell-Ginet

3.1.1 Expressive uses

There are uses of she to refer to people who are attributed and claim male sex.
Rudes and Healy 1979 give many examples collected in their ethnolinguistic inves-
tigation among gay males in Buffalo, NY. Rudes and Healy found what they saw as
both positive (youth, glamour) and negative uses (excess, disorganization, lack of
‘naturalness’, nastiness or a more global rejection). We are of course dealing with
notional gender, where that includes ideologies operative among these gay men
about self-presentation and other characteristics as well as their playing with
gender stereotypes others might embrace. Such uses do not show anything like
gender-inversion among gay men  – they are not ascribing ‘feminine’ identities
to one another but disavowing prominent styles of presumptively heterosexual
masculinity. Nor can research like that of Rudes and Healy be taken as evidence of
broader practices in some supposed ‘gay community’. Indeed, most discussions
of this phenomenon are decades old, and it is not clear that the practice contin-
ues much nowadays. As Barrett 1997 makes clear, there are multiple communities
within which gay men and others rejecting heterosexual norms engage in social,
including linguistic, practice. Pronoun switching is one available resource and
certainly has been used in a variety of ways among some groups of gay men.
Reports of analogous practices among English-speaking lesbians don’t seem
to exist, and linguist Robin Queen (personal communication) speculates that the
rather different histories of gay men and lesbians may account for this gap. Of
course this does not mean that he might not be used by some lesbians for refer-
ring to one another on occasion. What it does mean is that the use of pronoun-
inversion to signal ingroup status and to comment on (and perhaps thereby to
‘police’) self-presentation by others in the group seems confined to gay men and
perhaps mainly to past generations of gay men, far more of whom were closeted
and not generally ‘out’ than is now the case. (Pastre 1997 surveyed French-speak-
ing gay men and lesbians on the topic of using masculine forms to refer to one
another; she not only got a much lower rate of response from the lesbians than
from the gay men but some expressed hostility and only a couple reported using
masculine forms, which they reserved not for reference but for address in inti-
mate, romantic settings.)
In general there are few reports of uses of masculine pronouns to refer to
those who claim and are attributed female identities. Mathiot and Roberts 1979
report what seemed to be instances of ‘elevating’ the referent, highlighting com-
petence or other attributes of someone claiming and attributed female sex by
referring to that person using he. I have not directly observed such uses nor are
they reported elsewhere in the literature so it is not clear how widespread this
phenomenon might be.
 Gender and its relation to sex: The myth of ‘natural’ gender       27

There are also dismissive uses of feminine pronouns heard in presumptively

heterosexual communities of practice, often accompanied by other feminine
forms like girl. Coaches (male) of male athletic teams sometimes engage in this
kind of disparagement. Like pronoun-inversion among gay men, this practice is
probably on the decline.

3.1.2 Presumptive leaps

Back in the early 1970s some feminists told a story along these lines.

A young boy was rushed to the hospital from the scene of an accident in which his father
was killed and prepped for emergency surgery. The surgeon walked in, took one look, and
said “I can’t operate on him – that’s my son.”

Many people hearing the story would say things like

(13) How could the boy be his son? The boy’s father was killed.

They were assuming that the surgeon was male and overlooking the possibility
that the surgeon was the boy’s mother. (They were also assuming only one father,
though that would not have affected pronoun choice.) Though recognizing the
existence of a stereotype that surgeons are male (perhaps less robust now than
when the story was first told but by no means completely moribund), seman-
ticists tended to cringe when people took the story to illustrate that surgeon is
semantically ‘male’. After all, it is not contradictory to utter sentences like

(14) She/my mother/Jennifer is a surgeon

What McConnell-Ginet (2008) proposed is that words have cognitively associated

with them not only information about their linguistic properties, including their
conventional meaning, but also what I call conceptual baggage. Although concep-
tual baggage is not part of what the words conventionally mean, it can be essential
for understanding how what is said using those words contributes to the inferences
interlocutors draw, often with neither speaker nor hearer consciously accessing the
conceptual baggage on which those inferences rely. There are many cases:

(15) A: I’ve got to drive the babysitter home.

B: How far away does she live?

28       Sally McConnell-Ginet

(16) A: The police have identified the burglar who took Kim’s silver.

B: Was he local?

Babysitters and burglars can be male and female respectively, but the pronouns
used by the questioners reveal their feminine and masculine notional gender.
People make such presumptive leaps often and not only on the basis of single
words like surgeon, babysitter, or burglar. An example like one I’ve used from my
own experience illustrates.

(17) Sally: I’ve got a student who does very well on homework and in
class but freaks out taking exams.
Sally’s Give her some old exams so she can practice working prob-
­colleague: lems under time pressure.

In so far as referent sex matches pronoun choice, such gender assumptions are
generally not noticed by either party. It’s the mismatches we catch. The student
whom exams tripped up happened to be male; my correction embarrassed the
colleague who would never have said that exam anxiety was a distinctively femi-
nine phenomenon. And when my friend told me she’d taken her sick cat to the
vet and I inquired “What does he think is wrong?”, I was the one corrected, to my
chagrin. In fact I know that these days more women than men enter veterinary
practice, I even know some women who are vets, and I have given considerable
thought to gender-stereotyping. Nonetheless, my tongue betrayed me. These are
not cases where he is gender-neutral: any reference to a specific individual by he
attributes masculinity, perhaps only figurative as with the inversion uses but not
neutral.
And of course pronoun selection can evince presumptions of heterosexuality.

(18) A: I heard from my former student Carolyn that she’s getting married
this summer.

B: Is he a linguist too?

A: I think she is an anthropologist.

We have seen that uses of he and of she that are either deictically anchored or ana-
phorically dependent on a noun phrase referring to a specific individual person
are equally gendered. There is no neutral ground for he in this arena: notionally
singular they is used in some contexts (e.g., if sex is unknown or the speaker
would prefer not mentioning it) but still avoided even by non-purists in others.
 Gender and its relation to sex: The myth of ‘natural’ gender       29

A special case of ‘unknown’ sex occurs when the apparent antecedent for a
pronoun is an overt or implied disjunction that makes the possibility of either
male or female referent very salient.

(19) a. If Mary or John calls, tell *him/?her or him/them I’ll be back in

an hour.

b. If my mother or father/one of my parents calls, tell *him/?him or

her/them I’ll be back in an hour.

Of course him would be fine if I’m someone whose parents are both men. There is,
however, no possibility of using masculine pronouns in cases like these without
implying a masculine caller. Notionally singular they is a common choice in such
cases of sex-indeterminate reference to a single individual (who is, of course, the
one who calls).

3.2 F rom specific to general

3.2.1 Shakespeare and company: Generalizing quantifiers

The grammatically plural they has a long history of being used with notionally
plural but grammatically singular antecedents, a usage proscribed from the 18th
century on but nonetheless encountered widely from William Shakespeare to
Jane Austen to Virginia Woolf. The so-called pronominal forms anyone and every-
one or NPs where any or every are the determiners are a primary arena where
English speakers of my generation were drilled on using he generically. But even
here he is sometimes problematic.

(20) a. Everyone should be sure that ?he understands the assignment.

b. Everyone got Chris’s e-mail, didn’t *he? (cp. didn’t they?)

The (b) sentence is especially odd and this oddness is only slightly ameliorated
if the addressees are all male. With anyone, the masculine form may work a bit
better but not much.

(21) a. Anyone turning in work that is not ?his own will be subject to
disciplinary action.
30       Sally McConnell-Ginet

b. Anyone who solves the bonus problems gets an A, doesn’t ?*he?

(cp. don’t they?)

Without the valiant efforts of generations of teachers and editors, they would
prevail in such contexts, its plurality fitting well with the notional plurality of the
antecedents.
In the prose of my youth, however, it was not just everyone and anyone con-
trolling masculine anaphoric pronouns (avoiding cases where the target pronoun
was in one of those pesky tag questions). We also frequently encountered every
and any as universal quantifying determiners with what would seem to be epicene
content nouns. Sentences like those in (22) were common, even in contexts where
it was clear that there might be female humans to whom the generalization was
supposed to apply.

(22) a. Every student must bring his own writing materials to class.

b. Any politician who talks regularly to his constituents knows that

he will sometimes disappoint some of them, perhaps most even-
tually.

Yet such usages were mainly encountered in written material, with they very fre-
quent in speech. Matossian 1997 and Newman 1997, two interesting studies done
well after my youth, found they with grammatically singular antecedents very
common indeed. Strikingly, Matossian found women in her Philadelphia and
Minneapolis groups less likely to use the putatively generic he and more likely
to use they with singular ‘general’ antecedents than the men; among these same
populations, women’s use of other ‘incorrect forms’ (e.g., double negatives) is
usually somewhat smaller than that of men. (Notice that such frequential differ-
ences in usage among women and men in the same communities are not evidence
for “gender dialects” as that label is used in Dunn’s contribution to this volume.)
Other research (e.g., Khosroshashi 1989) shows clearly that both interpretation
and production of so-called masculine generics is tied to gender ideologies as
well as to language ideologies. Women who are committed to pushing for greater
gender equity are less likely to produce sentences with masculines serving gener-
ically but more likely to interpret such sentences as applicable to females as well
as males.
Grammatically plural they with a notionally plural antecedent is increasingly
seen in print and heard on the airwaves as well as being produced often online
(not surprising, given the relative informality of much online usage). An example
from the February 28, 2011 issue of Newsweek, not exactly a ‘radical’ publication:
 Gender and its relation to sex: The myth of ‘natural’ gender       31

(23) No one’s just a “peace activist” anymore – they have a specialty. (p 18)

Notice that he would be quite bizarre here. This might be because the pronomi-
nal reference (what semanticists sometimes call E-type) in this case is actually
linked to the unexpressed but evoked everyone (in the contextually given domain
of young actors), but that too would be grammatically singular though notionally
plural.
In the semantics literature sentences of the sort we have looked at in
this section are analyzed as universally quantified. The term generic is often
reserved for sentences that generalize but (1) lack overt quantifiers in the plural
and (2) seem to permit exceptions. Bare plural subject generics (e.g., women
love babies, men can’t cook, etc.) have received considerable attention (see
McConnell-Ginet 2012 for some discussion), but of course gender is neutralized
in the plural so I will focus my discussion on singular generics, beginning with
feminine generics.

3.2.2 S
 ingular feminine generics

I will begin with some examples I used in McConnell-Ginet ([1979], 2011). Some
of them aren’t strictly speaking generic but they are all cases where there could
be male referents, and where the choice of he as an anaphoric pronoun is taken
as reference to a particular male whereas she is compatible with the referent’s
being male.

(24) a. When the nurse comes, he’ll take your blood pressure.

b. When the nurse comes, she’ll take your blood pressure.

Sentence (24a) makes reference only to a male nurse whereas (24b) could be used
quite correctly in a context where the nurse in question might be of either sex but
the speaker was simply making a general statement applicable to whatever nurse
happened to come. In professions like nursing, still predominantly populated by
women, feminine pronouns are the norm. A sentence like (25) is distinctly odd.

(25) ?*The careful nurse will be sure he takes the right medicines to the
right patients.

And teacher also often still takes a generic or indefinite she. I was struck when
one of my kids brought home something written by the male head of a junior
32       Sally McConnell-Ginet

high school department with roughly equal number of women and men teachers
involved in the self-instruction program, including himself, reported:

(26) Students have to check with the teacher regularly so she knows how
they’re progressing.

Generic feminines of this kind have been in use for some time, and they show
clearly that English gender is notional.
But feminine generics do not require predominance of women as possible
referents. Sometimes the choice seems clearly to reflect anticipation of a largely
female audience as in (27), which occurred in a mass-circulation women’s maga-
zine.

(27) Behaviorists believe that what a person does is determined by the situ-
ation in which she finds herself.

And, as with the presumptive leaps in choosing pronouns for specific reference,
a speaker’s gender stereotype can lead to a feminine generic. Two more examples
from my 1970s collection:

(28) Physicians fear – with some reason – that a patient who discovers she
can lose weight quickly, without hunger pangs, may fast unwisely on
her own.

(29) When I walk down to them and point my fingers and say “you,” why
that person even forgets her own name in the excitement. [interview
with emcee of quiz show that did use contestants of both sexes]

I’ve often extended the benefit of the doubt to speakers using what seemed to be
generic he. It is noteworthy when such a person switches pronouns. I listened to
a lecture on teaching with professor, student, and so on controlling he. So the fol-
lowing offered strong evidence of gender stereotyping, almost certainly operating
below the level of conscious attention.

(30) When a student finally says something after sitting silently half the
semester, don’t intimidate her.

Even a noun like parent, wearing its sex-neutrality on its sleeve in its primary
use to provide a sex-indefinite alternative to mother and father, can control an
indefinite or generic she.
 Gender and its relation to sex: The myth of ‘natural’ gender       33

(31) When a parent hears her baby cry, she rouses quickly.

A lecture by a distinguished psychologist for whom speaker, child, and subject all
were anaphorically linked to he contained such a generic feminine occurrence of
parent. Three decades later such a switch would be far less likely, as more and
more speakers move away from masculine generics.
Generic she was once more or less restricted to cases where women were
especially salient. Sometimes there was their predominance in some occupation,
other times their role as audience, and other times the operation of strong stereo-
types. Nowadays, however, we also find generic she, sometimes alternating with
generic he, in writing or planned talk when there is no special contextual promi-
nence of women. These uses are self-conscious expressions of certain gender ide-
ology: their use signals the user’s disavowal of default masculine generics and
usually recognition that many such generics were not interpreted as truly inclu-
sive or even fully meant as inclusive. Here’s an example of this kind of generic
feminine from a recent article in one of the leading philosophy journals:

(32) [A] policy is legitimate only if it is generally acceptable – only if there’s

an adequate case for enforcing the policy coercively such that, for every
member of the society, and for every qualified belief that she holds, she
can accept that case consistently with her holding that belief.
(Philosophical Review 119.2, April 2010, 256)

At least in certain circles, the generic she has emerged as a more forceful way to
reject generic he than rephrasing with the plural to eliminate gendering or opting
for the disjoined he or she (or she or he). Some of my students have reported
English teachers allowing generic she but rejecting he or she as much too clumsy
(which it can be if there is a long string of linked forms, perhaps possessives or
reflexives). Generic she may seem downright ‘unnatural’ and it still carries a little
‘shock’ value, which is part of the point, but its increasing usage is changing the
generic landscape and, with it, the relation of sex and gender in English.

3.2.3 Other generics

Treichler and Frank 1989 remains the most thorough discussion of generic usage
and possible alternatives to the then still very common singular he with intended
general force. It was the existence of generic he that led me first to question the
view of gender in English as ‘natural’. Leonard Bloomfield described gender in
English this way:
34       Sally McConnell-Ginet

“The distinction, then, between the pronoun-forms he and she, creates a clas-
sification of our personal nouns into male (defined as those for which the definite
substitute is he) and female (similarly defined by the use of the substitute she).
Semantically, this classification agrees fairly well with the zoological division
into sexes.” (1933, 253) This formulation seems to overlook generic he in such
instances as (33a,b) from the same book.

(33) a. Suppose, for instance, that day after day the child is given his doll
(and says da, da, da) immediately after his bath (p.30).

b. Even if we know a great deal about a speaker and about the

immediate stimuli which are acting upon him, we usually cannot
predict whether he will speak or what he will say (p.32).

How could a system that treats words like child and speaker as male be thought
to “correspond fairly well” with zoological sex? Presumably, because Bloomfield
and many others were accustomed to thinking of maleness as the ‘natural’ con-
dition for humans, with femaleness only a distinctive condition. It is certainly
possible to have a default rule that has singular epicene controllers serve as ante-
cedents for singular masculine pronouns. What is not possible – at least not for
many contemporary English users – is to view such an arrangement as one where
sex as such is the principle for pronoun assignment, a ‘natural’ gender system.
It does continue to be the case that male human beings are far more visible
in most contexts than female. And it also continues to be easy for many people,
especially but probably not only men, to think and also speak of prototypical
humans as male. It is conflating male humanity with humanity generally that
permits what Black and Coward 1981 dub “false generics,” where there is no
pronoun but the clear implication is that what would seem to be a clear case of
an epicene noun is being used as if its referents were all (adult) males. Here’s an
example, adapted from an anthropological essay.

(34) The villagers all left, leaving us behind with the women and children.

Variants of this example have been widely discussed, and there are many similar
cases where even though the language used looks gender-neutral, the message
conveyed is not.
And even when the context begins with reference to a particular person of a
specified sex, if one party is focused on that individual not as such but as exem-
plary of a category there can be gender confusion. An advice column in the real
estate section of my local paper on the 16th of February, 2011 began with a reader’s
 Gender and its relation to sex: The myth of ‘natural’ gender       35

question about selling their house to a tenant living in the house. The tenant is
referred to several times in the course of the rather involved question as she as
in she gave me a down payment. The ‘expert’ begins the response by speaking
of “your tenant”, “ your buyer”, “the buyer,” “your borrower,” etc., with no pro-
nouns. And then after several paragraphs I was astonished to read

[I]f you have not sold the house to this person and he is merely renting the house from you
until he can finance the $50,000 to complete the purchase, then the payments he or she
is making to you … [A]n accountant can help you determine whether the buyer/tenant is
paying you rent or interest on the loan you gave him to purchase the house.

The advice-giver seems to simply forget the sex of the tenant/buyer and speak of
a presumptively male generic tenant/buyer, though recalling at one point that the
actual individual in question could be female.
The point is simple: it is not only linguistic practices like the so-called generic
masculine that tend to render women less visible than men even though these
practices certainly are consequential.

4 ‘ Grammatical’ vs ‘natural/notional’ gender

systems: Reform strategies
I will close by briefly considering the rather different linguistic strategies that
have been taken by English-speaking feminist activists and most of those in lan-
guages with grammatical gender. In languages with grammatical gender, gender
marking is far more pervasive than it is in English-like languages, where it is
primarily limited to personal pronouns. There are, of course, some feminizing
affixes (-ess, -woman, -ette), and at certain points in history they were relatively
widely used. In the 19th century, for example, it was proposed that the women
faculty members at Vassar College be called professoresses, and for a long time
women chairing organizations were called chairwomen. In the case of, say, chair-
man vs chairwoman, the two forms are morphologically parallel, but in many
other cases the feminine forms have been viewed as suggesting the women and
men are doing different things in their positions. Speaking of a poetess, many felt,
implied less significance that the plain term poet. In the case of acting, one might
want to say that what males and females do is indeed different as casting in the
modern era typically matches the sex of character and the person acting the role.
Typically but not always, and even in acting many women now call themselves
actors. A woman I met in the 1970s had recently managed to land a good job in
36       Sally McConnell-Ginet

the traditionally masculine field of drafting: “I am a draftsman and that’s what

I should be called,” she said, “because I do just as good a job as any of the men
holding that title.”
There’s been debate over how masculine the -man suffix ‘really’ is – too bad,
I’ve often thought, the spelling didn’t become -mun to reflect that the suffix is pro-
nounced differently from the independent word man, which did actually begin
life as a neutral form, only later specializing to contrast with woman. Sometimes
the -man suffix is replaced by -person, but there is a tendency for chairpersons
and the like to be overwhelmingly female, with males continuing as chairmen.
Nonetheless, in spite of Bloomfield, there are many words in English that strike
users as gender-neutral, and there has been a concerted and moderately suc-
cessful push in many locales to replace gender-marked occupational labels like
fireman and stewardess by gender-neutral labels like fire fighter and flight atten-
dant. The occupations too have changed; many more women are now rushing off
in fire engines and many more men are urging passengers to buckle their seat
belts.
The push in the US at least has generally been to remove gendered refer-
ences, to promote gender-neutral ways of speaking of people, including the use
of they, he or she, and alternating pronouns for generic antecedents. But the push
in many countries with grammatical gender has been towards providing feminine
alternatives to designate women in positions previously open only to men and
designated by a masculine noun. Sometimes there is a new word, sometimes the
old word becomes double-gendered, able to control feminine as well as mascu-
line targets. When every noun is gendered (and in a number of languages, either
masculine or feminine), then feminization probably seems the more ‘natural’
strategy and it does have the advantage of increasing women’s visibility, espe-
cially if it is embraced by high-ranking women.
In a talk at the September 2010 IGALA meetings, Malka Muchnik detailed the
problems she sees for Israeli women who are trying to change Hebrew so that it
will better serve their interests. Feminization, she argued, did not work because
it was not embraced by powerful women; neutralization, she claimed, is really
impossible, and thus her argument was that the language was becoming more
rather than less masculinized. How widespread that view might be I do not know,
but what struck me is Muchnik’s experience of grammatical gender in modern
Israeli Hebrew as deeply problematic for gender equity.
To what extent the kind of gender system in a language constrains or pro-
motes gender equity is not clear. What is clear, however, is that the relation of
sex to linguistic gender, whether ‘grammatical’ or ‘notional’, is a moving target:
languages and those speaking them continue to change in ways that we still do
 Gender and its relation to sex: The myth of ‘natural’ gender       37

not fully understand. Neither linguistic nor sociocultural gender is ‘natural’ if by

that we mean impervious to change, isolated from human ideas and actions.

5 References
Adriaen, Monique and Ruth King. 1991. Job titles, ch. 5. In R. King, (ed.). 27–40.
Baron, Dennis. 1986. Grammar and Gender. New Haven: Yale University Press.
Baron, Naomi. 1971. A reanalysis of English grammatical gender. Lingua 27. 113–140.
Barrett, Rusty. 1997. The “homo-genius” speech community. In A. Livia & K. Hall (eds.), Queerly
Phrased: Language, Gender, and Sexuality. 181–201. New York & Oxford: Oxford University
Press.
Beardsley, Elizabeth Lane. 1973–74. Referential genderization. Philosophical Forum 5.
285–293.
Black, Maria and Rosalind Coward. 1981. Linguistic, social and sexual relations. Screen
Education 39. 111–133.
Bloomfield, Leonard. 1933. Language. New York: Holt, Rinehart and Winston.
Boroditsky, Lera, Lauren A. Schmidt, and Webb Phillips. 2003. Sex, syntax and semantics. In
Dedre Gentner & Susan Goldin-Meadow (eds.), Language in Mind: Advances in the Study
of Language and Cognition, 61–80. Cambridge, MA: MIT. Press
Burr, Elisabeth. 2003. Gender and language politics in France. In Hellinger & Bussman, (eds.),
vol. 3. 119–139.
Cameron, Deborah and Don Kulick. 2003. Language and Sexuality. Cambridge: Cambridge
University Press.
Connors, Kathleen. 1971. Studies in feminine agentives in selected European languages.
Romance Philology 24. 573–598.
Corbett, Greville. 1991. Gender. Cambridge: Cambridge University Press.
Curzan, Anne. 1999. Gender categories in early English grammars: Their message to the
modern grammarian. In Barbara Unterbeck and Matti Rissanen, (eds.), Gender in Grammar
and Cognition. The Hague: Mouton. 561–576.
Curzan, Anne. 2003. Gender Shifts in the History of English. Cambridge: Cambridge University
Press.
Fausto-Sterling, Anne. 2000. Sexing the Body: Gender Politics and the Construction of
Sexuality. New York: Basic Books.
Frank, Francine W. and Paula A. Treichler (eds.). 1989. Language, Gender, and Professional
Writing: Theoretical Approaches and Guidelines for Nonsexist usage. New York: Modern
Language Association.
Hall, Kira. 2003. ‘Unnatural’ gender in Hindi. In Hellinger & Bussmann (eds.), vol. 2, 133–166.
Hall, Kira and Veronica O’Donovan. 1996. Shifting gender positions among Hindi-speaking
hijras. In Victoria L. Bergvall, Janet M. Bing, and Alice F. Freed (eds.), Rethinking Language
and Gender Research: Theory and Practice. London and New York: Longman. 228–266.
Hellinger, Marlis and Hadumod Bussmann (eds.). 2003. Gender across Languages, Volumes
1–3. Amsterdam and Philadelphia: John Benjamins.
Khosroshahi, Fatemeh. 1989. Penguins don’t care, but women do: a social identity analysis of a
Whorfian problem. Language in Society 18. 505–525.
38       Sally McConnell-Ginet

King, Ruth (ed.). 1991. Talking Gender: A Guide to Nonsexist Communication. Toronto: Copp
Clark Pittman Ltd.
Livia, Anna. 2001. Pronoun Envy: Literary Uses of Linguistic Gender. New York and Oxford:
Oxford University Press.
Livia, Anna and Kira Hall (eds.). 1997. Queerly Phrased: Language, Gender, and Sexuality. New
York and Oxford: Oxford University Press.
Mathiot, Madeline (ed.). 1979. Ethnolinguistics: Boas, Sapir and Whorf Revisited. The Hague:
Mouton.
Mathiot, Madeline, assisted by Marjorie Roberts. 1979. Sex roles as revealed through
referential gender in American English. In M. Mathiot (ed.). 1–47.
Matossian, Lou Ann. 1997. Burglars, Babysitters, and Persons: A Sociolinguistic Study of
Generic Pronoun Usage in Philadelphia and Minneapolis. Philadelphia: University of
Pennsylvania PhD dissertation.
McConnell-Ginet, Sally. 2008. Words in the world: How and why meanings can matter.
Language 84(3). 497–527.
McConnell-Ginet, Sally. [1979], 2011. Pronouns, prototypes, and persons. In S. McConnell-
Ginet. chapter 9. 185–206.
McConnell-Ginet, Sally. [1988], 2011. Linguistics and gender. In S. McConnell-Ginet. chapter 2.
37–62.
McConnell-Ginet, Sally. 2011. Gender, Sexuality, and Meaning: Linguistic Practice and Politics.
New York and Oxford: Oxford University Press.
McConnell-Ginet, Sally. 2012. Generic predicates and interest-relativity. Canadian Journal of
Linguistics 57(2). 261–287.
Moore, Samuel. 1921. Grammatical and natural gender in Middle English. Proceedings of the
Modern Language Association 36. 79–103.
Nevalainen, Terttu and Helena Raumolin-Brunberg. 1994. Its strength and the beauty of it: the
standardization of the third person neuter possessive in Early Modern English. In Dieter
Stein and Ingrid Tieken-Boon van Ostade (eds.), Towards a Standard English, 1600–1800.
Topics in English Linguistics 12. Berlin: Mouton de Gruyter. 171–216.
Newman, Michael. 1997. Epicene Pronouns: The Linguistics of a Prescriptive Problem.
Outstanding dissertations in linguistics series. New York: Garland Press.
Pastre, Genviève. 1997. Linguistic gender play among French gay men and lesbians. In A. Livia
& K. Hall (eds.). 369–379.
Pollard, Carl and Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar. Chicago: University
of Chicago Press and Stanford: CSLI Publications.
Rosch, Elinor. 1978. Principles of categorization. In Elinor Rosch and Barbara B. Lloyd (eds.),
Cognition and Categorization. 27–48. Hillsdale, NJ: Erlbaum.
Rudes, Blair A. and Bernard Healy. 1979. Is she for real?: The concepts of femaleness and
maleness in the gay world. In Madeline Mathiot (ed.). 49–61.
Treichler, Paula A. and Francine W. Frank. 1989. Guidelines for nonsexist usage. Part 2. In Frank
and Treichler (eds.), 137–278.
Wagner, Suzanne. 2003. Gender in English Pronouns: Myth and Reality. Accessed at http://
www.freidok.uni-freiburg.de/volltexte/1412/pdf/Diss_Freidok.pdf. 16 February 2011.
Michael Dunn
Gender determined dialect variation
1 Introduction
On the Telqep tundra of the Russian arctic in the early 20th century, a young man
drives his reindeer sled towards a snow-covered jaraŋə, the traditional dome-
shaped reindeer hide tent of the Chukchis. The elderly Chukchi herdsman whose
home it is cries out okkaj! in surprise, and asks meŋin pəkirɣʔi? ‘Who has arrived?’.
His wife comes out to look and echoes his question, ii, meŋin pəkiccʔi? ‘Yes, who
has arrived?’. The young man, seeing their confusion, shouts out ətcajqaj, waj
raɣtəɬʔeɣəm! ‘Aunty, it’s me, Raɣtəɬʔən’. They both call out to him in delight,
raɣtəɬʔeɣət?! ‘It’s you, Raɣtəɬʔən?!’. The aunt welcomes him, qəcecqikwi! jara-
cyko! qəcajoccən! ‘Come inside, into the house, have tea!’; ee, agrees the uncle,
qərecqikwi, qəcajorkən! This cozy scene of homecoming and tea-drinking is a
pastiche of welcoming scenes from folktales, but illustrates some of the typical
features of real Chukchi usage. There are small differences in the speech of men
and women, such as how the uncle says ee for ‘yes’ while the aunt says ii. Myste-
riously, many instances of r or consonant clusters with -r- in the uncle’s speech
correspond to the affricate -c- or -cc- in the aunt’s speech. But the aunt’s speech
does include -r- too. Longer acquaintance with the old couple would convince you
that these correspondences are completely regular: both of them say jaraŋə for
‘house’; aunty never says *jacaŋə. But whereas the old man says mren ‘mosquito’
the old woman would only ever say mcen.
In 1658 the author1 of the Histoire naturelle et morale des Iles Antilles
described a peculiar linguistic situation of the Garifuna people of the Caribbean
islands. Men and women had, at least in part, different vocabularies, with the
same object named in one way by men, and another by women so that “in much
of their conversation, one could say that women speak a different language to
men” (Rochefort 1658: 392; Section 3.9).

1 The identity of this author is somewhat mysterious: the preface of the first edition of the book
is signed LDP, presumed to be the initials of Philippe de Longvilliers de Poincy, the then Gover-
nor General of the French West Indies and by all accounts a very unlikely ethnographer. Later
editions are signed C. Rochefort, probably for Charles de Rochefort, a Huguenot pastor travel-
ling and writing in the West Indies at that time, but often confused with César de Rochefort, a
prominent French lexicographer. The Dominican missionary and ethnographer/naturalist Jean
Baptiste du Tertre further claimed that a considerable portion of the book is plagiarised from his
own work.
40       Michael Dunn

Since at least the 17th century there have been sporadic reports in the eth-
nographic record of language communities where men and women speak sig-
nificantly different forms of their language, so different, in fact, that it would be
impossible to speak without signalling gender identity. While gender variation
in language is common, perhaps ubiquitous, such obligatory, categorical dialect
differences determined by gender are rarer, and tend to be poorly described.
The complex correspondences between men’s and women’s Chukchi (above and
Section 3.6) are described as a simple substitution by women of c for men’s r and
č (Skorik 1961: 33). Gender dialects are often poorly documented with respect to
usage too. Many gender dialect systems occur in small, endangered languages,
and the gender dialect systems themselves tend to drop out of use faster than the
language itself does. This means that the rules for using gender dialects are often
inferred from the recollections of elderly (e.g. Section 3.1, 3.4, 3.8) or extracted
from older written sources (e.g. Sections 3.9, 3.12) rather than taken from direct
observations.
The term ‘dialect’ is used in the variationalist tradition to refer to system-
atic linguistic variation statistically associated with a sociolinguistic parameter,
and as such can be difficult to delimit (Labov 1972: 192). This paper approaches
gender-determined language usage from the ethnographic extreme, examining
a small set of instances of categorical gender dialects for which we have records.
These gender dialects are easily recognized as being the same language: they
are spoken by people who form a single speech community and their differences
only affect parts of the language: grammar, phonology and perhaps lexicon. They
differ from a frequentially characterised dialect in that they form complete lin-
guistic systems whose use is determined by the gender affiliation of the speech
participants (Section 2.1), and which is characterized by obligatory grammatical
differences rather than statistical tendencies (Section 2.2; Sherzer 1987: 96). In
languages which have such categorical gender dialects, to use language means to
use language like a woman or to use language like a man. This means that each
gender dialect must be learned separately (since presumably most individuals
in a gender dialect language community will have at least passive command of
the other gender dialect). This is interesting from an acquisitional perspective,
because the mutual autonomy of gender dialects makes for a considerable cogni-
tive load in acquisition and use. In turn, it is also indicative of the social/cultural
importance of signalling gender through language.
In this paper I take a pragmatic, cross-cultural approach to the notion
of gender. All known societies classify people at birth as “male” or “female”
according to the anatomical distinctions indicating their potential reproductive
role. But this is in practice a social classification, relating biological sex to a
wider set of social practices, norms, and relations (Eckert and McConnell-Ginet
 Gender determined dialect variation       41

1992: 463, McConnell-Ginet this volume). The specifics of the social construction

of gender vary from culture to culture, differing both in the manner of expres-
sion, and the magnitude of effect. But as a fundamental social category it is
highly likely to have significant consequences for ways of being (e.g. acting,
speaking) in any society.2 They provide a significant example of language-cul-
ture interaction, and have an important role in the documentation of human
cultural diversity.
Following from the strict categorical definition of gender dialect, I specifi-
cally exclude some phenomena called gender dialect or genderlect in the linguis-
tic literature from this survey. For example, Yokoyama (1999: 402) uses the term
“genderlect” to describe a gendered speech register in Russian which is a part of
colloquial speech, and which comprises “merely ‘typical’, rather than absolute
traits”. Likewise Sen ([1928] 1979) describes “women’s dialect in Bengali”, but
the description here shows a cluster of linguistic choices from within the range of
variation of a single language, rather than a pair of related, but structurally dif-
ferent, systems. I will not consider gender-determined variation such as that in
Yuchi, which shows morphological differences in possessive prefixes determined
by speaker gender, but which only occur with certain kinship terms (Wagner
1934: 339–340). Forty years after Wagner’s study, Ballard (1978: 107) could elicit
few of these terms. An example is the pair of terms for ‘my brother’, dzodane
in men’s speech and dodaʔone in women’s speech. Marking speaker gender in
kinship terminology is not the same sort of social signal that it is in a gender
dialect system. Kinship systems have an intrinsically egocentric orientation (e.g.
Scheffler 1987: 217), and quite naturally form a dyad between ego and relative,
where the sex of both participants in the relationship is relevant (see also the
‘same sex sibling’/‘different sex sibling’ oppositions in many languages). The
clearest examples of gender dialect have different phonological rules describable
in historical linguistic terms as the product of different regular sound changes
(e.g. Chukchi, Tangoan), or which have different morphological terms and gram-
matical categories (e.g Kũṛux, Yanyuwa).
Along with the recognition that gender is a culturally constructed category,
whose relation to biological sex is not straightforward, there are some scholars

2 The importance of biological sex in communication systems extends beyond humans too. For
example, in many bird species the songs of males and females are distinct. Furthermore, it is not
uncommon for birdsong to be transmitted through social learning, leading to vocal repertoires
which are differentiated by geographical region – referred to as regional dialects. There is even
an analogy to gender dialect amongst the birds: the slate-coloured boubou Laniarius funebris is
highly unusual in that it combines these two characteristics: songs are learned through imita-
tion of same-sex models, resulting in distinct male and female variants embedded within the
geographical variation (Wickler and Lunau 1996).
42       Michael Dunn

arguing that genderlect is not a valid category (Glück 1979; Motschenbacher 2007,
2010). Since these authors appeal to evidence from the languages treated in this
paper I will discuss the linguistic and ethnographic basis for their arguments
(Section 3.6). It is notable that the societies I will discuss are largely non-western,
non-technological and, in some cases, only attested historically. This is not by
chance: it may be that the stable transmission of gender dialect distinction is only
possible in languages which are used primarily as in-group codes, and not as
languages of inter-cultural communication.
This paper will treat the range of attested categorical gender dialects from
three interrelated perspectives. In Section 2.1 I will discuss the functional typology
of gender dialects, addressing how gender dialects are actually used. In Section
2.2 I present a description of the typical structural characteristics of gender, and
discuss whether any of these characteristics distinguish them from other kinds of
dialects. Section 2.3 takes a diachronic perspective: where are the attested gender
dialects and how have they come about? After the general characterization of
gender dialects in Section 2, Section 3 is devoted to 14 case studies of languages
with gender dialect distinctions.

2 Characteristics

2.1 Usage

Haas (1944: 147) provides a typology of gender dialect systems, whereby the

speech-act participant’s gender determines which gender dialect is used. She
defines three types:
–– Type I: Speaker-based systems
–– Type II: Addressee-based systems
–– Type III: Speaker-and-addressee-based systems

According to the Haas typology, systems based on gender of addressee are quali-
tatively different than speaker-based systems. Speaker-based systems comprise,
in effect, two language communities cohabiting in space (with passive compre-
hension), while addressee-based systems constitute two systems coexisting in
each speaker (and so are more akin to diglossia/bidialectalism). The Haas typol-
ogy doesn’t, however, seem to pick out the most common or important parame-
ters from the distributional point of view, since almost all attested gender dialects
belong to Type I. Type II systems intersect with notions of politeness and taboo:
they are contextually sensitive social norms of speech and behaviour and may be
 Gender determined dialect variation       43

difficult to fit in to the notion of “dialect” rather than “register”. For this reason
I treat Type I and Type III in this paper, although I do give a brief description of
Island Carib, a Type II language, since this is by far the best known example of
gender variation described as a gender dialect (Section 3.9).
In most language communities where gender dialects are used speakers are
bidialectal, at least passively and maybe actively. There is a basic asymmetry
here: in all the non-institutional language communities for which we have infor-
mation, children learn to produce the women’s dialect first, then male children
have to learn to produce the men’s dialect later. In most cases, boys’ acquisition
of the men’s dialect accompanies social and ritual recognition of their entering
the men’s world. This probably contributes to the historical instability of gender
dialects, as the interruption of traditional social practices may also interrupt
men’s dialect acquisition. A number of descriptions of gender dialects explicitly
mention that in quoted speech the gender dialect of the person quoted may be
used, even where this is otherwise not the gender dialect used by the speaker. In
Chukchi at least this is not completely regular, and is presumably related to vivid-
ness of the direct speech. Irish Sign Language seems to offer an exception to this
rule of bidialectalism, having developed from two different dialects spoken in
gender segregated institutions. In post-institutional life the female speakers have
adapted to the male speakers, and not vice versa (Section 3.1).
Japanese has strong social norms about gender-appropriate linguistic
behaviour (Inoue 2011). Some form of gender distinct speech has existed in Jap-
anese since at least as early as the Heian period (794–1185AD), although there
is no strong evidence that the practice was followed outside of social elites
over this entire period (Abe 1995: 654). It is however doubtful that Japanese
gender dialect variation forms a categorical system, although amongst many
features used differentially by male and female speakers, some scholars do
identify some features which are present only in one of the gendered varieties.
Examples of supposedly categorical phenomena in gender varieties of Japanese
include distinctive sets of personal pronouns and the forms of sentence final
particles. However, Abe (1995: 663) demonstrates that “[…] this categorization
of sentence-final particles based on gender is nothing but a representation of
longstanding stereotypes and fails to accurately represent the current usage
by both women and men.” The gender differences in Thai mentioned by Haas
(1944: 147–148) may be similar.
44       Michael Dunn

2.2 Structural typology

Gender variation in language is often expressed phonetically. Biological differ-

ences between males and females in vocal tract length determine fundamental
frequency: biological males have an average f0 around 100–120hz, and biologi-
cal females have an average f0 around 200–220hz (Simpson 2009: 622). This is a
statistical difference only, but is clearly used for social classification (Gelfer and
Mikos 2005). Hillenbrand and Clark (2009) show that f0 and formant frequen-
cies are major determinants of speaker sex identification. Listeners would fairly
reliably change their speaker sex identification from recordings when f0 and
formants were artificially manipulated. This effect was most reliable when both
f0 and formants were changed together, less so when just f0 was changed, and
weaker again when just the formants were changed. Interestingly, sex classifica-
tion changes of manipulated recordings happened more often with recordings of
isolated syllables than with entire sentences, demonstrating that there are other
phonological factors contributing to sex identification beyond these two. Hillen-
brand and Clark cite sources suggesting that women’s speech is breathier, and
(controversially) that women’s speech has wider prosodic variation (2009: 1191).
These features would seem to be under more conscious control, and I would
expect them to be subject to considerable cross-cultural variation, and available
for ‘emicization’ as overt socio-cultural gender markers.
While phonetic differences are outside the scope of this survey, many of the
languages with gender dialects do have significant phonetic differences between
men’s and women’s speech on top of phonological, morphological and lexical
differences. Women’s Chukchi has an affricate [ts] where men’s Chukchi has
(depending on the region) [s] or [tʃ]; Pirahã women’s speech uses a smaller articu-
latory space than men’s speech, with characteristic pharyngeal constriction and
more retroflexed articulation (Everett 2004: 7).
In the Journal of a second voyage for the discovery of a north-west passage from
the Atlantic to the Pacific, Parry (1824: 553) describes differences in the speech of
Eskimo men and women,

It is common for the Esquimaux to vary the pronunciation of their words at different times
without altering the sense. The women, in particular, seem frequently to make such altera-
tions as conduce to the softness of the words, as, for instance, by dropping the harsh final
k which occurs so commonly, as Inniloo for Innialook; by changing it into a vowel, as Ne-a-
ko-a for Neakoke, or by altering Oo-ce-ga into Oo-inga-a or Oo-ee-ma, and Hee-u-teega into
Hee-u-ting-a.

This kind of non-systematic phonological difference in men’s and women’s vari-

eties is attested widely. But categorical distinctions in men’s and women’s pho-
 Gender determined dialect variation       45

nology are also attested. There are two different kinds of systematic phonological
differences shown in the languages of this sample: Pirahã, Tangoan and (margin-
ally) Gros Ventre show phonological collapses in the women’s/common gender
dialect, which male speakers have to ‘undo’ when they produce the adult male
speech style. The phonological differences in Chukchi follow a different pattern:
three ancestral phonemes of proto-Chukotian, *s, *r and *ð, are collapsed dif-
ferentially into two in men’s (*r, *ð > /r/) and women’s (*s, *ð > /ts/) Chukchi.
Unlike the Pirahã and Tangoan, neither Chukchi gender dialect can be structur-
ally derived from the other.
Many gender dialect languages differ morphologically, either through having
different morphological forms, as in Kokama-Kokamilla, Koasati and Awetí, or
through expressing different morphological categories, as in Yanyuwa. Haas
Type III systems, i.e. systems which are determined by gender of speaker and
addressee together, are often expressed morphologically. Yanyuwa speakers have
different morphological paradigms depending on the gender of both speaker and
addressee; Kũṛux has an interaction of speaker gender, addressee, and grammati-
cal subject (the latter applies in the third person, where speech act reference is
neither speaker or addressee-oriented).
Some gender dialects possess lexical differences. These may be cryptic vari-
ants of the same word, where men’s and women’s forms of the lexeme are clearly
related but have some distinctive mutation. For example, certain nouns in Awetí
which are vowel initial in the women’s dialect are pronounced with initial n- in
the men’s dialect. There are also cases where men’s and women’s lexemes have
no obvious etymological relationship. This is also found sporadically in Awetí
with the word for the parrot species a. Amazonica, which is takänyt in men’s
dialect and takárï in women’s dialect. Often lexical differences in gender varieties
resemble euphemism, avoidance language, or other forms of word substitution
which are unlikely to be categorical (as found in Kalmyk, and yet other Awetí
terms).
There are some cases where one gender dialect can be formally derived from
the other – women’s dialect can be derived from men’s in the case of e.g. Pirahã
(Section 3.11) and Tangoan (Section 3.5), and men’s dialect can be derived from
women’s in the case of e.g. Yanyuwa (Section 3.4). But other gender dialect dis-
tinctions exist where neither men’s nor women’s systems are structurally deriv-
able from the other, e.g. Irish Sign Language (Section 3.1), Chukchi (Section 3.6)
and Kũṛux (Section 3.2).
46       Michael Dunn

2.3 O
 rigins and distribution

In general, gender dialects are originally statistical phenomena of gender varia-

tion in language use which have at some historical point become categorical.
Where the origins of gender dialects can be inferred, there seem to be three dis-
tinct evolutionary pathways:
–– Dialect merger
–– Internal change and conservatism
–– Isolation and diversification

In some cases discussed below the historical source of the women’s and men’s
dialect distinction is the merger of two different geographical dialects, or the
fossilization of other types of language contact. Chukchi (Section 3.6), Island
Carib (Section 3.9) and Kokama (Section 3.10) all provide examples of men’s and
women’s dialects incorporating elements for different genealogical sources. In
other cases, gender dialects have come about through long-term institutionaliza-
tion of linguistic conservatism in the speech of one gender. The mapping of con-
servative or colloquial speech onto gender categories can go either way: men’s
dialect can be conservative and women’s innovative (from the historical linguis-
tic perspective), or vice versa. In Pirahã (Section 3.11) and Tangoan (Section 3.5)
there is a learned phonemic distinction used by men, especially when speaking
in an elevated register. Comparison with related languages shows that this pho-
nemic distinction was present in the ancestral language, but subsequently lost in
colloquial speech. In Yanyuwa (Section 3.4), women’s language is more archaic,
preserving morpho-syntactic categories which have been collapsed in the men’s
language. Irish Sign Language (Section 3.1) is the sole example I have of a gender
dialect that evolved through neutral drift in isolation. While isolation and drift is
probably the default mechanism for the diversification of geographically-based
dialects, gender dialect could only develop this way in the kind of gender-segre-
gated institutional context recorded for this language.
Gender dialect systems do not seem to be diachronically stable. If they were
we might expect to see, for instance, entire linguistic subgroups with inherited
gender dialect systems. Rather, the gender dialect systems we see seem to be spo-
radic. There are hints that high levels of gender variation in language may be an
areal feature e.g. in Amazonia, and so the relatively frequent instances of gender
dialect systems in the Americas may be significant (Map 1 on page 48; see also
Kroskrity 1983: 88; Fleming 2012).
 Gender determined dialect variation       47

2.4 L anguage and other correlates of gender

Ochs (1988: 137–139) describes the effect of gender and social rank in Samoan
society, showing that while gender has significant effects on language, these
effects are outweighed by the effect of social rank. This demonstrates that while
gender is always an important social category, it is not always the most impor-
tant one. A similar phenomenon is found in ancient literary traditions. In San-
skrit (Indo-European, 1200–300 BCE) drama, only educated upper-class males
speak Sanskrit, whereas women and lower-class males speak colloquial Prākrit
(Hoch and Pandharipande 1978: 14–15). Likewise in Sumerian (language isolate,
2nd millenium BCE), men are portrayed as speaking high register Emeĝul and
women are portrayed as speaking the colloquial Emesal variety (Whittaker
2002). It is probably no coincidence that the Sanskrit and Sumerian examples
pertain to the written language, since literacy in ancient societies is a correlate
of high social status, and so writing itself should be oriented towards a higher
register.
The debate around the description of Koasati men’s and women’s speech by
Haas (1944) illustrates the arbitrariness of ‘gender’ as the social category deter-
mining linguistic variation (Section 3.8; Kimball 1987; Saville-Troike 1988). The
distinction between the two varieties of Koasati had all but disappeared at the
time that these languages were being studied, and it seems impossible to decide
conclusively whether gender or social status was at the root of the system. Luthin
(1991; cited by Mithun 2001: 278) showed that the Yana gender dialect system
described by Sapir (1929: 212) [as a Haas Type III system] was similarly a register
marking formality.
Labov ([2001] 2010: 266) shows that for stable sociolinguistic variables,
women show a lower rate of stigmatized variants and a higher rate of prestige
variants than men. Evidence for this principle was drawn from a wide range of
studies of common variables and a wide range of speech communities, rural
and urban, western and non-western. The evidence from gender dialects contra-
dicts this tendency however. In all the ethnographically attested cases of pos-
sible gender dialects with overtones of formality, it is the male (or male-to-male)
variety which is the elevated one (cf. Koasati, and Yana, as well as Tangoan).
48       Michael Dunn

3 Case studies
Gender dialect is distributed sporadically around the world. I know of no gender
dialects recorded in Africa, but the distribution in other parts of the world is so
thin that there is no reason to think that this absence – even if true – is signifi-
cant. The following case studies are ordered geographically, from west to east
according to the Pacific-centred view of Map 1.

6
1
3 7

8
2
9

10 11
14
4 5 13
Dialect merger 12
Conservatism
Isolating
Unknown

Map 1.: Distribution of attested gender dialects with evolutionary type where it can be inferred
(from internal evidence or from comparison with related languages). Languages: 1. Irish Sign
Language, 2. Kũṛux, 3. Kalmyk, 4. Yanyuwa, 5. Tangoan, 6. Chukchi, 7. Gros Ventre, 8. Koasati,
9. Island Carib, 10. Kokama, 11. Pirahã, 12. Chiquitano, 13. Awetí, 14. Karajá

3.1 Irish Sign Language

Irish Sign Language originated in a girls’ school for the deaf founded in Dublin
by a Dominican order of nuns in 1846. The nuns running the school were all
hearing, adult learners of a sign language, trained at a school for the deaf run
by the order in France. After the foundation of the girls’ school, there was little
or no further contact with sign languages from outside, so the language used at
the school rapidly developed unique characteristics. A second link to this chain
was added when a boys’ school for the deaf was founded a decade later, in 1857.
The language used at this school was based on a dictionary of signs produced
by the nuns from the girls’ school, but despite the physical proximity of the two
schools there was very little interaction between them, and the language at the
boys school likewise developed its own unique characteristics. From these diver-
gent varieties a consensus variety was formed when these children grew up and
began to socialize together. The consensus variety was based on the male dialect,
 Gender determined dialect variation       49

but female speakers maintained female sign alternants for use as an in-group,
women-only code. The linguistic differences that are documented are lexical:
some meanings are indicated by entirely different signs, and some signs have dif-
ferent meanings in the two dialects (see LeMaster 1999: 69–70 for illustrations). In
terms of usage, Irish Sign Language was a Haas Type III system: a woman talking
to women would use the women’s sign language; talking to men or a group con-
taining men she would use the so-called ‘men’s language’, the latter always being
used by male speakers.
The two Irish Sign Language dialects were the sole medium of language
socialization for many deaf Irish children for a century. In 1946 the girls’ school
abandoned sign language in favour of oralism, with the boys’ school following
suit in 1957 (although sign language continued to be used outside the classroom
at the boys’ school at least until 1987). LeMaster (1997, 1999) gives a description
of the lexical characteristics of the Irish Sign Language gender dialects, based on
dissertation research in 1990. At the time this research was carried out, the native
speakers of Irish Sign Language were primarily women over 70  years and men
over 55 years.

3.2 Kũṛux

Ekka (1972) describes a complex morphological interaction between speech act

participants within the verbal paradigms in the Dravidian language Kũṛux. A
sample is given in Table 1. This is a good example of a Haas Type III “speaker-
addressee determined” gender dialect (Section 2.1). When this system is con-
sidered from the speaker’s perspective, you can see that men and women have
to make different morphological choices from each other. A male speaker must
select different forms of the 2sg or 3sg.m agreement according to gender of
the addressee (the 2sg form is coreferential with the addressee, but the 3sg.m
form–presumably–is not). A female speaker uses distinct person-number
agreement forms according to gender of addressee in all person-number com-
binations except for 1pl.in, and two of the 3rd person forms. Men’s and women’s
forms only differ from each other when the addressee is a women; for a male
addressee the agreement form is identical irrespective of the gender of the
speaker.
It is unclear how this system developed. According to Ekka (1972: 31), there
is no evidence of similar phenomena in any of the neighbouring Dravidian or
Munda languages.
50       Michael Dunn

Table 1: Present tense of ‘come’ in Kũṛux, showing interaction of speaker and addressee

gender with specification of gender, number and person of grammatical subject. MM = Man
speaking to man, WM = Woman speaking to man, etc. (adapted from Ekka 1972: 26).

MM, WM MW WW

1sg barc-k-an barc-ʔ-an

1pl.ex barc-k-am barc-ʔ-am
1pl.in barc-k-at
2sg barc-k-ay barc-k-i barc-k-in
2pl barc-k-ar barc-k-ay
3sg.m barc-Ø-a barc-Ø-as
3sg.nm barc-Ø-a(d)
3pl barc-Ø-ar

3.3 Kalmyk

The Kalmyk (Mongolic) women’s language is an example of an avoidance lan-

guage (Aalto 1959: 3–4; Birtalan 2003: 227). The precise circumstances of its use
are unclear, but it seems that is consciously and overtly derived from general
Kalmyk. Many common terms are taboo, and replaced by near synonyms: for
terms meaning ‘girl, daughter’ the word nojɔxɔn ‘princess, lady’ is used; köwün
‘boy, son’ is replaced by a word derived from the adjective ajtɛ meaning ‘good,
proper‘. If no suitable or sufficiently unambiguous synonym is available, a word
can be transformed phonologically, by replacing the initial consonant with /j/:
shaghä :: yaghä ‘ankle bone’, tend :: yend ‘there’ and shaar :: yaar ‘tea’. Certain
words are referred to euphemistically or with cryptic paraphrases: bū “taxation”
is called ägarɣɔdɔɢ “the roisterer, rabble-rouser”.

3.4 Yanyuwa

The Yanyuwa language (Pama-Nyungan) of northern Australia has a complex

and well-described gender dialect distinction (Kirton 1988; Bradley 1988; Kirton
and Charlie 1996). The main difference between the dialects is in syntactic cat-
egories and their morphological marking: The female dialect distinguishes two
noun classes, ‘male’ and ‘masculine’, where the male dialect only has one.3

3 See Mithun, this volume, for discussion of the culturally specific properties of gender as a
grammatical category.
 Gender determined dialect variation       51

There is one exception in the male dialect, an archaic form suggesting to Kirton
and Charlie (1996: 3) that the ancestor of the male dialect also distinguished two
classes. Yanyuwa makes use of noun class prefixes, which differ according to
case. In the female dialect “male” and “masculine” noun classes are indicated by
different prefixes (see Table 2).

Table 2: Noun class prefixes in the female dialect of Yanyuwa

noun class nominative non-nominative

male nya- nyu-

masculine o̸ ji-

In the male dialect these correspond to a single noun class, marked by different
prefixes in non-nominative cases and by zero in the nominative (see Table 3), like
the women’s masculine-class.

Table 3: Noun class prefixed in the male dialect of Yanyuwa

noun class nominative non-nominative

male/masculine o̸ ki-

The women’s dialect also makes more distinctions in third-person pronouns than
the men’s dialect. These distinctions are highlighted in Table 4.

Table 4: Third-person pronouns in male and female Yanyuwa dialects

women’s dialect men’s dialect

he yiwa yiwa
she anda anda
it alhi anda

Bradley (1988: 127) gives a good example of how far-reaching the differences

between the two dialects are, contrasting the women’s dialect (1) with that of the
men (2):
52       Michael Dunn

(1) Nja-ja nya-wukuthu nya-rduwarra niya-wini

this-M M-short M-initiated_man his-name
nya-Wungkurli kiwa-wingka wayka-liya ji-wamarra-lu
M-personal_name he-go down-wards MSC-sea-ALL
niwa-yirdi na-ridiridi ji-walya-wu
he-bring ARB-harpoon MSC-dugong/turtle-DAT

(2) Jinangu ø-wukuthu ø-rduwarra na-wini

this short initiated man his-name
ø-Wungkurli ka-wingka wayka-liya ki-wamarra-lu
personal name he-go down-wards MSC-sea-ALL
na-yirdi na-ridiridi ki-walya-wu
he-bring ARB-harpoon MSC-dugong/turtle-DAT

“The short initiated man whose name is Wungkurli went down to the
sea, taking a harpoon with him for dugong or sea turtle”

The Yanyuwa language was no longer being transmitted at the time that the
gender dialects were documented, so we only have speakers’ reminiscences of
how language acculturation happened rather than direct observations. But the
situation seems to have been similar to that reported for Pirahã and Tangoan,
where all children acquire the women’s dialect first from their caretakers. In
Yanyuwa society, boys underwent formal initiation at the age of ten, after which
they were expected to speak men’s dialect, and rebuked if they spoke the women’s
dialect by mistake. Older speakers could use the inappropriate gender dialect for
various kinds of humorous or rhetorical effect.

3.5 Tangoan

Many of the languages of Santo and Malekula islands in Vanuatu have, or have
had, distinctive apico-labial phonemes, a cross-linguistically rare type articulated
with the tip of the tongue against the middle of the upper lip (described for the
neighbouring language Araki in François 2002: 15). This areal feature is evidently
unstable, as a number of the other languages of the region show apico-alveolar
stops and nasals corresponding to proto-Oceanic bilabials, a change which can be
explained by an intermediate apico-labial stage: *p > *t̼ > t and *m > *n̼ > n (Tryon
1976: 52). There is also evidence from the Tutuba language that the apico-labial
 Gender determined dialect variation       53

c­ onsonants can revert back to bilabials, e.g. *p > *t̼ > p (Naito 2006: 224). This
sound change would normally leave no trace, and was only detected because
Naito (2006) was able to observe the phonological change in progress between
generations.
The Tangoan language had a gender dialect distinction acquired by males
during a protracted initiation period, during which boys lived in seclusion in all-
male company (Baker 1928: 289–290). This distinction held only for phonological
features: boys had to learn to produce apico-labials in the appropriate contexts,
in effect undoing the historical collapse of the reflexes of e.g. *m and *mw, as in
Table 5.

Table 5: Phonological distinctions in Proto-Oceanic and Tangoan

‘eye’ ‘snake’

Proto-Oceanic *mata *mw ata

Male Tangoan t̼ata mata
General Tangoan mata mata

To the phonologist, this is actually quite remarkable, since the irreversibility of

phonological mergers is the basic diagnostic of directionality in phonological
change. In the Tangoan case the knowledge of pre-merger phonological distinc-
tions may have been supported by awareness of other languages in the vicinity in
which the distinction is also preserved, as demonstrated in Table 6.

Table 6: Phonological correspondences in Proto-Oceanic and several Oceanic languages

‘eye’ ‘snake’
w
Proto-Oceanic *m *mata *m *mw ata
Tolomako n nata- m mata
Araki t̼ t̼əri-ku m mařa
Tangoa (male) t̼ t̼ata- m mata
Tangoa (general) m mata- m mata
North Malo m mata m mata

William Camden described the male dialect as a prestige variety, used regularly
for “oratory, serious discussion, traditional storytelling, etc., and with less con-
sistency in ordinary speech.” (1979: 113). He noted that women and children are
not expected to use the phonological distinction, but implied that they occasion-
ally did without sanction. Inconsistency in the use of the male dialect is attested
54       Michael Dunn

in early sources. In his 19th century description of Tangoan, Joseph Annand

(1891: 1–2) described the phonological characteristics of the apico-labials quite
well, at least for the stop consonant. He went on to note, however, that [m] and [n]
are often used interchangably, even by the same speaker. His examples – lima/
lina ‘hand’, mae/nae ‘come’ and magi/nagi ‘animal’ – are all reflexes of *m, so
provide the expected environment for the apico-labial nasals.

3.6 Chukchi

The Chukchi gender dialects (Dunn 2000) were introduced at the beginning of
this chapter. They provide interesting evidence of pre-contact social dynam-
ics. Chukchi is a Chukotko-Kamchatkan language, spoken in the tundras of the
extreme north east of the Eurasian continent. Until the mid twentieth century
most Chukchis lived as nomadic reindeer herders in family units of perhaps 20
people. On the northern coasts there was intermixture with sedentary Yupik
Eskimo hunter-gatherers, the source of considerable Chukchi influence in the
Siberian Yupik languages, and in the south a Chukchi expansion over recent cen-
turies put pressure on culturally similar nomadic pasturalists speaking closely
related Koryak languages. There are also contacts with other coastal and riverine
groups speaking Chukotka-Kamchatkan languages: Kerek, Alutor, and further
varieties of Koryak.
Male and female Chukchi are distinguished by differences in the phonologi-
cal system. Two phonemes (men’s /r/ and /s/, women’s /r/ and /ts/) can be sorted
in to three correspondence sets, as shown in Table 7.

Table 7: Male and female phonological correspondences in Chukchi

Male Chukchi Female Chukchi Correspondence

mosquito mren mtsen /r/ :: /ts/

polar fox rekokalɣən tsekokalɣən /r/ :: /ts/
reindeer qoraŋə qoraŋə /r/ :: /r/
s/he went home raɣtəɣʔe raɣtəɣʔe /r/ :: /r/
trap utkusʔən utkutsʔən /s/ :: /ts/
sister sakəɣet tsakəɣet /s/ :: /ts/

Historical linguistic reconstruction of this family shows that the correspondences

between male and female Chukchi phoneme variation are the product of regular
sound change, albeit different sound changes in each dialect. In female Chukchi
*r is realized as /r/, while *ð and *s are realized as /ts/. In male Chukchi the three-
 Gender determined dialect variation       55

to-two phonological collapse has gone another way: *r and *ð are realized as
/r/, while *s is realized as /s/. This means that there is a phonological contrast
between male and female Chukchi in any word containing a reflex of *ð, as well
as the phonetic contrast between the afficate /ts/ of the female dialect and the
sibilant /s/ of the male.

Table 8: Sound changes in Chukchi and related languages

Alutor Palana Female Male Chavchuv Kerek

Koryak Chukchi Chukchi Koryak

*t t/tʃ t/tʃ t t t t/tʃ

*ð t/tʃ t/tʃ ts r j/tʃ j
*r r r r r j/tʃ j
*s s ts ts s tʃ tʃ
*j j j j j j/tʃ o̸/j

There is also a form class of adverbs and particles which is distinctive in having
a final r ~ t alternation. This includes forms such as qənur/qənut, ewər/ewət,
qənwer/qənwet, iɣər/iɣət, luur/luut, janor/janot and weler/welet. These are not
diagnostic of male/female speech, but there are statistical preferences, with
men using the -r final forms and women using the -t final forms, in each case by
a ratio of about four to one. Historically, the final -r/-t in these forms is *ð, the
proto-phoneme which is otherwise realized as /ts/ in female Chukchi. However,
the word-final allophone of *ð is /t/ in Palana Koryak and Alutor, so the /t/ pro-
nunciation is expected. What is surprising is that these forms are not recognized
as categorically male or female dialect forms, despite being high enough fre-
quency for the variants and the statistical associations with gender to be pre-
served. It seems that the salient feature of the gender dialects of Chukchi is the
/r/ :: /ts/ contrast. The low-frequency /r/ :: /t/ contrast seems to be treated as a
stylistic choice.
Using gender dialect correctly is part of being manly or womanly according
to Chukchi construction of gender, but there is no taboo against using the other
gender dialect in appropriate occasions. Chukchis freely quote speech in either
gender dialect during vivid story-telling. In traditional society, and well into
living memory, individual Chukchis sometimes change their gender affiliation as
part of shamanic inspiration. The individual “doomed to being a shaman” adopts
the distinctive characteristics of the other gender: clothing, work, social behav-
iour, and not least language (Bogoras 1901: 98–99).
Glück (1979) argues against the existence of gender dialects with some
dubious data from Chukchi. He claims that Russian loanwords into Chukchi
56       Michael Dunn

all enter in the form of the men’s language, rather than the women’s, on the
basis that where these loanwords can contain r in Russian they always have r
in Chukchi too (1979: 191). While it’s true that loanwords from Russian never
have the /r/ :: /ts/ correspondence, the most salient characteristic of the gender
dialect distinction, Glück was unaware that Chukchi men’s and women’s lan-
guages also have a systematic /r/ :: /r/ correspondence. Naturally Russian
loanwords with /r/ enter into the /r/ :: /r/ correspondence set, since the only
words in the /r/ :: /ts/ correspondence set are words descending from proto-
Chukotko-Kamchatkan /*ð/. Likewise, Motschenbacher’s (2010: 45) observation
that women and men can both use the other gender dialect in certain contexts
(such as quotation) is true, but this is not a valid argument that a gender dialect
distinction doesn’t exist.

3.7 Gros Ventre

The description of men’s and women’s speech in Gros Ventre (Algonquian) pro-
vides an example of a marginal gender dialect from both phonological and func-
tional perspectives (Flannery 1946; Taylor 1982). Phonologically, where women
have k or ky men have č or (for elderly speakers who preserve the phoneme dis-
tinction) ty. The women’s form is completely predictable from the men’s, condi-
tioned by the following vowel, as demonstrated in Table 9.

Table 9: Phonological correspondences in men’s and women’s speech in Gros Ventre

women’s k / __e, i
mens č (ty) corresponds to
women’s ky / __a, æ, ʌ

This shift was perfectly reversable for most speakers during the 1960s to 1980s,
since the phonological distinction between č and ty was only preserved by a
few elderly men. While the č ~ ty distinction was preserved, this would seem to
provide the same kind of acquisitional puzzle as in, e.g., Tangoan, where boys
acculturated in the women’s language would have to learn to reverse a phonologi-
cal merger as they matured.
The contexts of use of the so-called “men’s” and “women’s” varieties of Gros
Ventre are also not clear. While Flannery (1946: 133) describes them unproblemat-
ically as speaker-determined gender varieties, Taylor’s (1982: 304) observations,
in an admittedly highly endangered language situation, suggest that men might
have used the so-called “women’s” variety in certain sociolinguistic contexts too,
including as “foreigner talk”.
 Gender determined dialect variation       57

3.8 Koasati

Haas (1944) describes a morphological difference in the men’s and women’s

speech of Koasati (Muskogean), comprising a set of correspondences between the
respective forms for certain endings in the indicative and imperative. These rules
produce pairs (women’s form ~ men’s form) like ka·hâl ~ ka·hás ‘I am saying’.
í·sk ~ í·sks ‘you are saying’, and ka· ~ ká·s ‘he is saying’. Haas (1944: 145) gives
internal comparative evidence from other verbal paradigms that the women’s
speech preserves more archaic features of the language than the men’s. Haas
notes that at the time of observation (up to 1939 according to Kimball 1987: 30)
the women’s system was only used by the middle-aged and elderly women.
This entire description was challenged by Kimball (1987). In fieldwork starting
in 1977 Kimball found that the ‘male’ speech described by Haas was very nearly
extinct, only used by people quoting the speech of deceased elders. He also con-
cluded that rather than being a phonological transformation of certain positions
in the verbal paradigm, the male speech forms were actually produced by a suffix
serving as a phrase terminal marker. Two other phrase terminal markers – dele-
tion or nasalization of the final verb – were in use at the time of Kimball’s field-
work, although neither was functionally connected to speaker gender. Kimball
also established that this so-called ‘male’ marker was (i) not used by all men, (ii)
also used by some women, and (iii) that usage was probably related to high social
status.
Saville-Troike (1988: 242) contributed to this debate with observations from
fieldwork in 1968–1969. At that time her consultants explicitly described the vari-
eties as male and female terms, regretting:

that Koasati boys were no longer learning to speak like males, and had not done so for
about twenty years. He said that before then boys had learned the male forms when they
accompanied their fathers and other men for hunting and daily activities. (The female forms
were acquired first by both boys and girls in early childhood while with female caretakers
in the home.)

Saville-Troike (1988: 242) suggests that Kimball’s social status marking function

might be a reanalysis of the markers by non-users as the markers themselves were
dropping out of common speech.

3.9 Island Carib

Perhaps the most well-known example of a gender dialect is Island Carib. The
documentation of the gender dialect difference comes entirely from the 17th
58       Michael Dunn

century (Rochefort 1658; Breton [1667] 1877). The early sources related that the
gender dialect was said to have come about as the result of Carib conquest of
the Arawakan Iñeri speaking communities, with eradication of the male Arawaks
and capture of the female. Later anthropological analysis has questioned this
(Whitehead 2002).
Taylor (1954) and Hoff (1994) have reanalyzed the 17th century sources with a
better understanding of comparative Carib and Arawakan linguistics, which gives
us a reasonable picture of how the gender varieties were structured and used.
They demonstrate that Island Carib is structurally an Arawakan language with
a lexical stratum of Carib (where the etymologically Carib items correspond to
the men’s language, as would be expected). There are only a few morphologi-
cal differences between men’s and women’s Island Carib in Breton’s data (Taylor
1954: 29), and many of these may have been unwittingly recorded examples of
code-switching to mainland Carib, rather than authentic men’s style Island Carib
(Hoff 1994: 163). There are only a few regularly incorporated Cariban morphologi-
cal elements in Island Carib, and these are used in the same way as they are used
in Carib pidgins, rather than as in Carib proper. Hoff concludes that Island Carib
was a mixed language with gender determined diglossia that grew out of society
wide bilingualism in Iñeri (Arawakan) and Pidgin Carib. Finally, Hoff (1994: 164)
also interprets the sources to say that the gender variety choice is determined by
addressee gender rather than speaker gender, making Island Carib a rare example
of a Haas Type II language.

3.10 Kokama-Kokamilla

The Kokama-Kokamilla language is spoken in the Peruvian Amazon. It has been

classified as a member of the Tupi-Guarani family (Campbell 1997: 200), although
Cabral (1995: 2–3) has argued for a mixed origin, a conclusion only partially sup-
ported by Vallejos on the basis of a comprehensive description of the language
carried out in a more vital language community (Vallejos 2010: 38). The men’s
and women’s dialects of Kokama-Kokamilla differ in number particles, personal
pronouns, demonstratives and connectors. Some of these forms seem to result
from a semi-regular process where women’s y corresponds to men’s r, but other
forms lack any obvious etymological connection, such as penu ~ tana ‘1 plural
exclusive’.
 Gender determined dialect variation       59

Table 10: Particles used in the men’s and women’s dialects (Vallejos 2010: 42)

Gloss F M Gloss F M

1sg. ta tsa, etse distal dem. yukun yukan

1pl.excl. penu tana indef. dem. yama/yamua rama/ramua
3sg. short form ya [ja ~ za] ra ‘like this’ ajaya ikiara
3sg. long form ay uri ‘like that’ ya [ja] ria
3sg. object (=)ay (=)ura ‘also’ yay riay
3pl. inu rana ‘but’ iyan urian
plural clitic =kana =nu ‘there, then’ yaepe raepe
proximal dem. ajan ikian ‘after that’ yaepetsui raepetsui

These grammatical function words are very common in speech, making for highly
salient differences between men’s and women’s speech:

(3) a. uri tsenu ikian yawara=kara=uy tana ku=kuara

(male speaker)

b. ay tsenu ajan yawara=nu=uy penu ku=kuara

(female speaker)

3SG.L hear DEM dogs=PL=PAS1 1PL farm=INE

‘She heard the dogs on our farm.’ (Vallejos 2010: 41–42)

There are also minor phonological differences. In the Kokamilla dialect, women
have a tendency to produce the phoneme /r/ as a lateral rather than as a tap.
Vallejos speculates that this might be a residue of an earlier difference between
the Kokamilla dialect and the Kokama dialect (2010: 102), which if true would add
support for the hypothetical role of dialect mixing in the origins of the Kokama-
Kokamilla gender dialects.

3.11 Pirahã

The Pirahã language has gender differences on phonetic and phonological levels
(Everett 1979, 1986, 2004). On the phonetic level, Everett (2004: 7) reports that
women use more retroflexed points of articulation in comparison to men, and
they have a characteristic ‘gutteral’ articulation caused by pharyngeal constric-
tion. Phonologically, where men have two phonemes /s/ and /h/ women have a
single phoneme /h/.
60       Michael Dunn

In a personal communication (22 April 2011), Everett filled in some more of

the background to this. He confirms that early in language acquisition, Pirahã
boys speak using the women’s variety. Their later shift to the men’s variety seems
to be motivated by the desire to affiliate more with adult men’s society; there are
no formal social structures reinforcing this, however, but Everett gives anecdotal
evidence of an individual male who, along with other unconventional (for Pirahã)
social preferences, uses the women’s dialect.
The simplest historical scenario to produce /h, s/ :: /h/ correspondence between
male and female dialects of Pirahã is a simple phonological collapse, which is con-
sciously resisted and reversed by adult men. The source of the phonological col-
lapse is likely to be language-internal change, rather than contact with another
language or dialect, since the pharyngeal constriction characteristic of womanly
speech would act to reduce the distinctiveness of s and h in female pronounciation,
from which the full collapse of /s/ and /h/ phonemes would be a small step.

3.12 Bésɨro (Chiquitano)

Bésɨro is an endangered language isolate spoken in south-eastern Bolivia (Adelaar

and Muysken 2004: 477; Sans 2009). Nineteenth century forms of the language
are documented as having morphologically and lexically distinguished gender
variants (Adam and Henry 1880). It is not however clear that the system ever
included obligatory variation rather than a just register choice, and the examples
by Adam and Henry (1880: 67) show variation between the notional men’s and
women’s varieties even in a single utterance.
Adelaar and Muysken interpret this as a men’s variety and a general variety,
rather than a women’s variety (Adelaar and Muysken 2004: 479). The men’s
variety had morphological class and number markers which were used only in
elevated speech, when talking of men or of divine entities. The general dialect is
used by men in other contexts, as well as by women in all contexts. The gender
varieties also had other differences. These include: nouns never used by men;
pronouns and adverbs only used by men; nouns that women use inflected and
men uninflected, and vice versa (Adelaar and Muysken 2004: 478–479, excerpted
from Adam and Henry 1880: 67–68). New data is coming available on the Bésɨro
gender dialects which casts doubt upon the earlier sources. While gender variet-
ies are not treated in a sociolinguistic survey of the contemporary language by
Sans (2009), a complete description of the gender dialects will be presented in a
grammatical description currently under preparation by the same author. Rose
(submitted) contains a summary of this updated account which shows that Bésɨro
does in fact have a categorical gender dialect distinction.
 Gender determined dialect variation       61

3.13 Awetí

Drude (2002) describes another Amazonian gender dialect in the Awetí language.
This dialect difference, determined by speaker gender, manifests itself both mor-
phologically, with differences in the pronoun, verbal prefix and, deictic para-
digms, and lexically. The differences in the personal pronouns are illustrated in
Table 11.

Table 11: Personal pronoun differences by gender in Awetí

M F

1sg atit ito

2sg en
3sg nã i
1pl.ex kajã
1pl.in ozoza
2pl e’ipe
3pl tsã ta’i

Unlike the morphological distinctions between men’s and women’s Awetí, the
lexical distinctions are variational tendencies rather than categorical differences.
There are two types of lexical alternatives. The first type comprises pairs consist-
ing of entirely different words in the male and female dialects. In some cases, the
alternatives have transparent morphological structure. The men’s dialect tends
to form these morphologically complex, metaphorical terms by reference to func-
tion, whereas the women’s dialect refers to source material/species, as exempli-
fied in Table 12.

Table 12: Lexical differences by gender in Awetí

Gloss Male dialect form Female dialect form

drinking gourd y’a’jýt = ‘little round thing for mopo’j ýt = ‘small gourd’
water’
thatch tawypepo’apy (< tawypé, ‘roof, tapaj’jypo’apy (< tapaj’jyp, plant
large mat’) species used for thatching)
curica (parrot species) takänyt takárï

The second type of lexical difference relates to form class: vowel-initial words in
the women’s dialect correspond to n- initial words in the men’s dialect. The words
62       Michael Dunn

entering this class are mostly species’ names and tools, three examples of which
can be found in Table 13.

Table 13: Phonological differences in lexical items by gender in Awetí

Gloss Male dialect form Female dialect form

duck nypék ypék

parrot napúryt apúryt
bow nyzapát yzapát

Drude (2002: 189) suggests that the vowel-initial forms are ancestral, and that the
n- initial forms are innovated by analogy to the third person singular inalienable
noun prefix.

3.14 Karajá

The Karajá language of central Brazil has a well documented gender dialect
system in a highly gender-segregated society (Fortune and Fortune 1975; Fortune
1998; Ribeiro 2012). The differences are highly salient, with the gender dialect
differences showing up in every second or third word of running speech (Fortune
and Fortune 1975: 112). The most frequently occurring difference is a simple cor-
respondence between k in the women’s dialect and o̸ in the men’s dialect. There
is one exception to this, a small set of interjections and grammatical words where
the men’s form has the same k as the women’s form. There is also an irregular cor-
respondence between women’s č and men’s o̸, č or ǰ; a conditioned rule for drop-
ping n in the men’s dialect; and a few words which are etymologically unrelated
in the two dialects (Table 14).
These phonological correspondences (at least those involving dropping
women’s k in the men’s dialect) are the product of synchronically active phono-
logical rules (Ribeiro 2012: 130–139), and act equally on Portuguese loanwords,
e.g. women’s kararu :: men’s araru (< cabalo ‘horse’); women’s nobiku :: men’s
nobiu (< domingo ‘Sunday’).
Ribeiro (2012: 149) describes how speakers may use gender-incongruent
dialect forms, for example, when quoting speech by a member of the opposite
sex. Interestingly, this also seems to be used as part of gender dialect socializa-
tion: men may use women’s speech when baby-talking to baby girls and women
may use men’s speech when baby-talking to baby boys (Ribeiro 2012:149).
 Gender determined dialect variation       63

Table 14: Phonological and lexical comparison of men’s and women’s Karajá (Fortune and
Fortune 1975: 116–118)

correspondence women’s Karajá men’s Karajá gloss

k :: o̸ kətora ətora ‘fish’

k :: k kohe kohe ‘yes’
kai kai ‘you’
č :: o̸ ričɔre riɔre ‘child’
č :: č ričoko ričoko ‘doll’
č :: ǰ ičorɔθa iǰorɔθa ‘dog’
n :: o̸ / a__o anobo aobo ‘what’
lexical -bu- -hi- ‘cry’
-sira- -bu- ‘to be angry’
bebe! mi! surprise
wu ku calling form

4 Conclusion
Women’s and men’s dialects are a poorly documented phenomenon in language,
which, while rare, is nevertheless important to an understanding of the range
of possible culture and language interactions in a broad ethnographic perspec-
tive. In many of the cases described above, language dialect distinction is clearly
a reflex of wider social gender segregation. Physical separation of the genders,
e.g. with the “men’s houses”, found in traditional Tangoan society as well as in
Karajá, contribute to mutual reinforcement of gendered practices in culture and
language. The Irish Sign Language situation was different: the gender varieties
developed under isolation from each other, so didn’t have the distance-creating
function present in their origins. I know of no other languages like this, but they
might be expected to show up in places with e.g. gender-segregated monastic
traditions, Cossack-style military societies, etc.
Gender dialects are only attested in relatively small communities. Most likely
gender dialects are only stable in small scale societies, what Trudgill refers to as
“societies of intimates” (2011: 185). Maintaining a gender dialect distinction in a
language is evidently costly: a community must preserve society-wide bidialectal-
ism, and growing children must relearn basic linguistic principles of their native
language. As a result of this gender dialect systems are diachronically unstable,
and rarely survive major social upheavals within a speech community. Neverthe-
less, it is clear that gender dialect is a possible outcome of gendered sociolin-
guistic variation, which demonstrates the relative importance of the social sig-
64       Michael Dunn

naling function of language compared to acquisitional ease and communicative

efficiency.

5 Acknowledgements
Thanks to Agata Blichewicz and Kate Bellamy-Dworak for assistance in assem-
bling materials, and to Kate Bellamy-Dworak and the participants in the ‘Expres-
sion of Gender’ workshop (4 March 2011) at the Max PIanck Institute for Psycho-
linguistics, Nijmegen for their questions and comments.

6 References
Aalto, Pentti. 1959. Über Die Kalmückische Frauensprache. Studia Mongolica 1(3). 1–8.
Abe, Hideko Nornes. 1995. From Stereotype to Context: The Study of Japanese Women’s
Speech. Feminist Studies 21(3). 647–671.
Adam, Lucien and Victor Henry. 1880. Arte y vocabulario de la lengua Chiquita, con algunos
textos traducidos y explicados compuestos sobre manuscritos inéditos del XVIIIo Siglo.
Bibliothèque Linguistique Américaine VI. Paris: Maisonneuve.
Adelaar, Willem and Pieter Muysken. 2004. The Languages of the Andes. Cambridge:
Cambridge University Press.
Annand, J. 1891. A Grammar of the Tangoan-Santo Language. In Donald MacDonald (ed.), South
Sea Languages: A Series Of Studies On The Languages Of The New Hebrides And Other
South Sea Islands. 1–14.
Baker, John R. 1928. Depopulation in Espiritu Santo, New Hebrides. The Journal of the Royal
Anthropological Institute of Great Britain and Ireland 58. 279–303.
Ballard, William L. 1978. More on Yuchi Pronouns. International Journal of American Linguistics
44(2). 103–112.
Birtalan, Ágnes. 2003. Oirat. In Juha Janhunen (ed.) The Mongolic Languages. 210–228.
London: Routledge.
Bogoras, Waldemar. 1901. The Chukchi of Northeastern Asia. American Anthropologist 3(1).
80–108.
Bradley, John. 1988. Yanyuwa: ‘Men Speak One Way, Women Speak Another. Aboriginal
Linguistics 1. 126–134.
Breton, Raymond. [1667]1877. Grammaire Caraibe. In Charles Leclerc and Lucien Adam (eds.),
Grammaire Caraibe. Paris: Maisonneuve et Cle.
Cabral, Ana Suelly. 1995. Contact-induced Language Change in the Western Amazon. The
Non-genetic Origin of the Kokama Language. PhD dissertation, University of Pittsburgh.
Camden, W. G. 1979. Parallels in Structure of Lexicon and Syntax Between New Hebrides
Bislama and the South Santo Language as Spoken at Tangoa. In Peter Mühlhäusler (ed.),
Papers in Pidgin and Creole Linguistics, 2. 51–117. Pacific Linguistics Series A. Occasional
Papers 57.
 Gender determined dialect variation       65

Campbell, Lyle. 1997. American Indian Languages: The Historical Linguistics of Native America.
Oxford: Oxford University Press.
Drude, Sebastian. 2002. Fala Masculina e Feminina Em Awetí [Male and female speech in
Awetí]. In Aryon Dall’Igna Rodrigues and Ana Suelly Arruda Camara Cabral (eds.), Línguas
Indígenas Brasileiras: Fonologia, Gramática e História, 177–190. Belém: EDUFPA.
Dunn, Michael. 2000. Chukchi Women’s Language: A Historical-Comparative Perspective.
Anthropological Linguistics 42(3). 305–328.
Eckert, P. and S. McConnell-Ginet. 1992. Think practically and look locally: Language and
gender as community-based practice. Annual Review of Anthropology 21. 461–490.
Ekka, Francis. 1972. Men’s and Women’s Speech in Kũṛuk. Linguistics 81. 25–31.
Everett, Daniel L. 1979. Aspectos Da Fonologia Do Piranhã [Aspects of Piranhã phonology]. PhD
dissertation, Universidade Estadual de Campinas. Instituto de Estudos da Linguagem.
Everett, Daniel L. 1986. Pirahã. In Desmond Derbyshire (ed.), Handbook of Amazonian
Languages 1. 200–325.
Everett, Daniel L. 2004. Coherent Fieldwork. In Piet van Sterkenberg (ed.) Linguistics Today –
Facing a Greater Challenge. 141–162. Amsterdam: John Benjamins.
Flannery, Regina. 1946. Men’s and Women’s Speech in Gros Ventre. International Journal of
American Linguistics 12(3). 133–135.
Fleming, Luke. 2012. Gender indexicality in the Native Americas: Contributions to the typology
of social indexicality. Language in Society 41. 295–320.
Fortune, David L. and Gretchen Fortune. 1975. Karajá Men’s-Women’s Speech Differences with
Social Correlates. Arquivos De Anatomia e Antropologia 1. 111–124.
Fortune, Gretchen. 1998. Sex-exclusive differentiation in the Karaja language of Bananal island,
Central Brazil. PhD dissertation. Lancaster University.
François, Alexandre. 2002. Araki: A Disappearing Language of Vanuatu. Pacific Linguistics,
Research School of Pacific and Asian Studies, Australian National University.
Gelfer, Marylou Pausewang and Victoria A. Mikos. 2005. The Relative Contributions of Speaking
Fundamental Frequency and Formant Frequencies to Gender Identification Based on
Isolated Vowels. Journal of Voice 19(4). 544–554.
Glück, Helmut. 1979. Der Mythos von den Frauensprachen. Osnabrücker Beitrage Zur Sprach-
theorie 9. 60–95.
Haas, Mary R. 1944. Men’s and Women’s Speech in Koasati. Language 20(3). 142–149.
Hillenbrand, James M., and Michael J. Clark. 2009. The Role of F0 and Formant Frequencies in
Distinguishing the Voices of Men and Women. Attention, Perception, and Psychophysics
71(5). 1150–1166.
Hock, Hans Henrich and Rajeshwari Pandharipande. 1978. Sanskrit in the Pre-Islamic Sociolin-
guistic Context of South Asia. International Journal of the Sociology of Language 16. 11–26
Hoff, Berend J. 1994. Island Carib, an Arawakan language which incorporated a lexical register
of Cariban origin, used to address men. In Peter Bakker and Maarten Mous (eds.), Mixed
Languages: 15 Studies in language intertwining. 161–168. Amsterdam: IFOTT.
Inoue, Miyako. 2011. What does language remember? Indexical inversion and the naturalized
history of Japanese women. In Bambi B. Schieffelin and Paul B. Garrett (eds.), Anthro-
pological Linguistics: Critical concepts in language studies. Volume III. Talking about
language. 212–235 London, New York: Routledge.
Kimball, Geoffrey. 1987. Men’s and Women’s Speech in Koasati: A Reappraisal. International
Journal of American Linguistics 53(1). 30–38.
Kirton, Jean F. 1988. Men’s and Women’s Dialects. Aboriginal Linguistics 1. 111–112.
66       Michael Dunn

Kirton, Jean F. and Bella Charlie. 1996. Further Aspects of the Grammar of Yanyuwa, Northern
Australia. Dept. of Linguistics, Research School of Pacific Studies, Australian National
University.
Kroskrity, Paul V. 1983. On Male and Female Speech in the Pueblo Southwest. International
Journal of American Linguistics 49(1). 88–91.
Labov, William. 1972. Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press.
Labov, William. [2001]2010. Principles of Linguistic Change, Volume II, Social Factors. Wiley-
Blackwell.
LeMaster, Barbara. 1999. Reappropriation of Gendered Irish Sign Language in One Family.
Visual Anthropology Review 15(2). 69–83.
LeMaster, Barbara. 1997 Sex Difference in Irish Sign Language. In Jane H. Hill, P. J. Mistry and
Lyle Campbell (eds.), The Life of Language: Papers in Linguistics in Honor of William
Bright. (Trends in Linguistics 108), 67–86. Berlin Mouton de Gruyter.
Luthin, Herbert. 1991. Restoring the Voice of Yanan Traditional Narrative: Prosody, Performance,
and Presentational Form. PhD dissertation, University of California, Berkeley.
Mithun, Marianne. 2001. The Languages of Native North America. (Cambridge Language
Surveys). Cambridge: Cambridge University Press.
Motschenbacher, Heiko. 2007. Can the Term ‘Genderlect’ be Saved? A Postmodernist
Re-definition. Gender and Language 1(2).
Motschenbacher, Heiko. 2010. Language, Gender and Sexual Identity: Poststructuralist
Perspectives. Amsterdam: John Benjamins.
Naito, Maho. 2006. Tutuba Apicolabials: Factors Influencing the Phonetic Transition from
Apicolabials to Labials. Oceanic Linguistics 45(1) (June 1). 217–228.
Ochs, Elinor. 1988. Culture and Language Development: Language Acquisition and Language
Socialization in a Samoan Village. (Studies in the Social and Cultural Foundations of
Language 6). Cambridge: Cambridge University Press.
Parry, Sir William Edward. 1824. Journal of a Second Voyage for the Discovery of a North-west
Passage from the Atlantic to the Pacific. John Murray, publisher to the Admiralty, and Board
of Longitude.
Ribeiro, Eduardo. 2012. A Grammar of Karajá, PhD dissertation, Chicago University.
Rochefort, Charles de. 1658. Histoire naturelle et morale des îles Antilles de l’Amerique.
Enrichie de plusieurs belle figures des raretez les plus considérebles qui sont d’écrites.
Avec un vocabulaire Caraïbe. Rotterdam: Arnout Leers.
Rose, Françoise. submitted. On male and female speech and more. A typology of categorical
gender indexicality in indigenous South American languages.
Sans, Pierric. 2009. Éléments de sociolinguistique du bésɨro (chiquitano). Approche
bibliographique et approche de terrain d’une langue en danger de Bolivie. MA thesis.
Lyon: Université Lumière Lyon II.
Sapir, Edward. 1929. Male and Female Forms of Speech in Yana. In Josef Schrijnen and St. W. J.
Teeuwen, Donum Natalicium Schrijnen. 79–85. N. V. Dekker and van de Vegt.
Saville-Troike, Muriel. 1988. A Note on Men’s and Women’s Speech in Koasati. International
Journal of American Linguistics 54(2). 241–242.
Scheffler, Harold W. 1987. Markedness in Systems of Kin Classification. Journal of Anthropo-
logical Research 43(3). 203–221.
Sen, Sukamara. [1928]1979. Women’s Dialect in Bengali. Calcutta: Jijnasa.
Sherzer, Joel. 1987. A Diversity of Voices: Men’s and Women’s Speech in Ethnographic
Perspective. Language, Gender, and Sex in Comparative Perspective. 95–120.
 Gender determined dialect variation       67

Simpson, Adrian P. 2009. Phonetic Differences Between Male and Female Speech. Language
and Linguistics Compass 3(2). 621–640.
Skorik, Pjotr Ja. 1961. Grammatika čukotskogo jazyka. Tom 1, Fonologija i morphologija
[Grammar of the Chukchi language. Volume 1. Phonology and Morphology]. Leningrad:
Nauka.
Taylor, D. 1954. Diachronic Note on the Carib Contribution to Island Carib. International Journal
of American Linguistics 20(1). 28–33.
Taylor, Allan R. 1982. ‘Male’ and ‘Female’ Speech in Gros Ventre. Anthropological Linguistics
24(3). 301–307.
Trudgill, Peter. 2011. Sociolinguistic Typology: Social Determinants of Linguistic Complexity.
Oxford, New York: Oxford University Press.
Tryon, Darrell T. 1976. New Hebrides Languages: An Internal Classification. Australian National
University, Department of Linguistics Research School of Pacific Studies.
Vallejos Yopán, Rosa. 2010. A Grammar of Kokama. PhD dissertation, University of Oregon.
Wagner, Günter. 1934. Yuchi. In Franz Boas (ed.), Handbook of American Indian Languages.
Part 3. 293–384. Columbia University Press.
Whitehead, Neil L. 2002. Arawak Linguistic and Cultural Identity Through Time: Contact,
Colonialism, and Creolization. In F. Santos-Granero and J. Hill (eds.), Comparative
Arawakan Histories: Rethinking Language Family and Culture Area in Amazonia. 51–73.
University of Illinois Press.
Whittaker, G. 2002. Linguistic Anthropology and the Study of Emesal as (a) Women’s Language.
In Simo Parpola and R. M. Whiting (eds.), Sex and Gender in the Ancient Near East.
633–634. Helsinki: Neo-Assyrian Text Corpus Project.
Wickler, W. and K. Lunau. 1996. How do East African Bush Shrikes Laniarius funebris recognize
Male and Female Tutors during Gender Dialect Development? Naturwissenschaften 83,
12(12). 579–580.
Yokoyama, Olga. 1999. Russian Genderlects and Referential Expressions. Language in Society
28(3). 401–429.
Peter Hegarty
Ladies and gentlemen: Word order and
gender in English
1 Introduction
Why do speakers of English like their homes to be spic and span rather than span
and spic? Drink gin and tonic in the pub rather than tonic and gin? Stagger home
from that same pub in a zig-zig rather than zag-zig fashion? Or stop for fish and
chips rather than chips and fish on the way home? Linguists have long been fas-
cinated by preferences to order words in binomial phrases; “sequence[s] of two
words pertaining to the same form-class, placed on an identical level of syntac-
tic hierarchy, and ordinarily connected by some kind of lexical link” (Malkiel,
1959: 113). Speakers reference couples – such as Jack and Jill, Romeo and Juliet,
and The Queen and Prince Philip – and gender categories – such as ladies and
gentlemen, men and women, and boys and girls, in binomial phrases. In his
classic paper on binomials Malkiel (1959: 145) described the preference to name
men before women in Indo-European and Semitic languages as paradigmatic of
the way that pairs of words get ordered “with a hierarchy of values inherent in the
structure of a given society.” In a recent study, Mollin (in press) found that men
and women was the most common binomial including the conjunction and in
modern written English. Making sense of order preferences in binomials requires
an analysis of gender. In this chapter, I review past studies and present some
new evidence that semantics determines the order in which women and men are
referenced in binomials.

2 O
 rder Preferences in Binomials Referring to
Gender Categories
Is the preference for men and women over women and men a “natural” feature of
the English language, or something conventional that is subject to historical and
situational change? Very different answers to this question are suggested by two
essays, both published in 1975; Cooper and Ross’ (1975) chapter on preferences for
order in “frozen” binomials in English, and Bodine’s (1975) article on the history
of gendered conventions in English. Cooper and Ross’ (1975) chapter has been, by
far, the more influential of these two texts and I review it first. These authors pro-
70       Peter Hegarty

posed hypotheses about both semantic and phonological constraints that would
tend to fix or “freeze” order in a particular direction. They considered seman-
tic constraints to be primary and phonological ones to be secondary. Among the
twenty-six semantic constraints that Cooper and Ross (1975: 67) hypothesized,
several could be glossed by a parsimonious “me first” principle in which “first
conjuncts refer to those factors which describe the prototypical speaker.” This
principle is evidenced by preferences to first name things that are closer in space
(here and there), closer in time (now and then), or more friendly to the self (pro
and con). The authors also listed seven phonological features – listed in order of
presumed strength – that would be likely to distinguish first placed from second
placed elements in frozen binomials. Compared to first named terms, second
named terms were predicted to have (1) more syllables, (2) more resonant nuclei,
(3) more initial consonants, (4) more obstruent initial segments (if both words
start with the same consonant), (5) a vowel containing lower second formant fre-
quency, (6) fewer final consonants and (7) a less obstruent final segment.
Gender was not a particular focus of Cooper and Ross’ (1975) analysis, but
their assumptions about gender are theoretically interesting. Like Malkiel (1959),
these authors described the hypothesis that semantically male things would be
named before semantically female ones. On the way to deriving the “me first”
principle, Cooper and Ross (1975: 4–5) explicitly reject the possibility of a more
general rule to position prototypical unmarked terms first, partially on the
grounds of the male first preference. By so doing, they seemingly overlooked
how a male-first hypothesis fits with the me-first principle for men better than
for women. In other words, Cooper and Ross’ (1975) analysis is characterized by
androcentrism, “that is, males and male experience are treated as a neutral stan-
dard or norm for the culture of the species as a whole, and females and female
experience are treated as a sex-specific deviation from that allegedly universal
standard” (Bem, 1993: 41).
In addition, Cooper and Ross (1975) described the male-first preference as
both general and natural. In the context of explaining deviations from this prefer-
ence, they note that the preference for mother and son over son and mother, runs
counter to “the general law of males first” (p. 94, my emphasis). (Elsewhere the
authors explain preference to position mothers first by noting that “We believe
that mothers are special” p. 105.) In a footnote on another exception, ladies and
gentlemen, they hypothesize that the phrase is likely a result of a politeness con-
vention, and that such conventions are “in general contrary to natural tenden-
cies” (p. 105, my emphasis). In other words, they described idiomatic language
that positions women first as conventional and idiomatic language that positions
men first as natural.
 Ladies and gentlemen: Word order and gender in English       71

Were Cooper and Ross (1975) aware of this androcentrism? In their conclud-
ing paragraph, they note:

Finally … we attempt to relate our findings to a general framework of man’s view of himself
in the world. The principle of Me First, which appears to account for a fairly wide range of
freezing constraints, coupled with the assumption that place 1 conjuncts reflect the traits
of the prototypical speaker, might give some indication about how we view this speaker.
Although we have up until now been tacit on this matter, we hereby forsake the guise of
linguistics proper and admit to being card-carrying Whorfers (p. 103).

Here, at last, Cooper and Ross distinguish the “me first” and “prototypical first”
rules, and note that both rules are required. This conclusion could be read as
recognizing earlier androcentric assumptions about the extent to which ‘me’ is
a male person. But regardless of the authors’ intentions, their analysis was tacit
about the possibilities of future analysis beyond “linguistics proper.”
The second essay from 1975 was more concerned with hierarchical gender
relations in those societies where English is spoken. Bodine (1975) argued against
the view that the use of generics such as he and man to refer to all were neutral
or “traditional” uses of English (see also McConnell-Ginet, this volume). In so
doing, she traced the prehistory of the prescription to use he as a generic in the
1850 Act of the British Parliament, which contracted binomial phrases such as
he and she, on the grounds that the male term covered all legal persons. Bodine
(1975) described grammarians’ arguments for the correctness, naturalness, and
propriety to name male entities before female ones in English in the 16th and 17th
centuries. Wilson (1553, cited in Bodine, 1975) is typical of the earlier grammar-
ians she cites. Like Cooper and Ross some four centuries later, Wilson (1553, cited
in Bodine, 1975) proscribed that “in speaking at the leaste, let us kepe a natural
order, and set the man before the woman for manners sake.” Thus Bodine’s
(1975) suggests that the use of English is historical, and that gendered preference
for referring to men first in binomials might be dictated by conventions, just as
much as the polite phrase ladies and gentlemen. The difference is not that one is
natural and the other conventional, but that preferences for male-first binomials
have, from 1553–1975 been ontologized as natural and general, whilst female-first
binomials have been ontologized as conventional and exceptional, most often by
authors who, unlike Malkiel (1959), recognized that gender is a hierarchical social
relationship in many English-speaking societies.
72       Peter Hegarty

3 Empirical Studies of Binomial Order Preferences

Since 1975, several useful empirical studies have added to our knowledge of
why English language speakers order terms in binomials the way that they do.
Within this literature, Bodine’s (1975) historical research is rarely mentioned,
whilst Cooper and Ross’ (1975) work is routinely positioned as foundational. Both
psychologists and linguists have pursued the goal of teasing apart which con-
ventions truly account for order preferences. As a consequence of this strategy,
scholars in both disciplines have said comparatively less about situational and
historical change in such preferences.
Three sets of psychology experiments are particularly worthy of attention
within this literature. First, Pinker and Birdsong (1979) explored the strength
of Cooper and Ross’ (1975) hypotheses about phonology by asking speakers of
English and speakers of French to indicate their preferences for sentences con-
taining nonsense syllables within French and English sentences. Participants
consistently preferred sentences with binomial phrases in which the word with
fewer syllables was positioned first in each language. Cooper and Ross’ (1975)
hypotheses about phrases with particular vowel sounds or consonant sounds in
the first and second term received far more qualified support or no support at all.
These results constitute confirming evidence for the hypothesis that word length
is primary over other phonological determinants of order preferences.
A second set of experiments provided support for the semantic rule to posi-
tion prototypical things first and atypical things second in binomial phrases.
Kelly, Bock and Keil (1986) argued for a model of lexical access in which more
prototypical category members are more accessible than atypical ones because
they are more easily called to mind (Rosch and Mervis, 1975). They presented par-
ticipants with sentences including binomial phrases in which either a prototypi-
cal or an atypical item was named first (e.g., apples and lemons or lemons and
apples). Participants systematically mis-remembered these sentences to position
more prototypical things first. In a second study, participants reported a prefer-
ence for sentences that positioned prototypical things first.
Finally, McDonald, Bock, and Kelly (1993) examined both semantic and
phonological features concurrently in their experiments. These authors exam-
ined preferences by testing speakers’ memory for binomial phrases and noting
when order was spontaneously reversed in memory. They found that partici-
pants were particularly likely to reverse word order to place animate people
and animals before inanimate things, consistent with Cooper and Ross’ (1975)
“animate first” rule. However, there was no preference to put shorter words
before longer ones. In toto, this body of evidence confirms Cooper and Ross’
(1975) hypothesis that phonological features can affect preferences for order
 Ladies and gentlemen: Word order and gender in English       73

(Pinker and Birdsong, 1979), but are outweighed when semantic factors are rel-
evant (McDonald et al., 1993).
Linguists have also examined order preferences by studying the frequency of
their occurrence within corpora. Fenk-Oczlon (1989) introduced the new hypoth-
esis that more frequently used words in the language are positioned first in bino-
mial phrases, a preference that explained 84% of the cases in her corpus, more
than any other single rule could explain. Benor and Levy (2006) examined a large
number of semantic, metrical and phonological features that might constrain
word order. Their findings confirmed Cooper and Ross’ broad generalization in
that semantic features appeared to be the primary determinants of order prefer-
ences. Similar conclusions were drawn from a recent study of the British National
Corpus, where again “[s]emantic features tend to trump metrical ones” and
“[t]he phonological principles, on the other hand, clearly operate on a subordi-
nate level and only have influence on binomial order in the absence of the other,
more important factors” (Mollin, 2012, p. 94). Among the semantic factors, Mollin
noted particularly strong tendencies to position less marked, more powerful, and
more iconic entities first. Her definition of markedness distinguished perceptual
markedness, such as the “me first” principle, from formal markedness, in the
sense of the term with the more specific rather than the more general meaning.
Her findings regarding both markedness and power are consistent with the psy-
chological studies mentioned above (Kelly et al., 1986, McDonald et al., 1993).
Recent studies of corpora have also oriented attention toward the question
of how flexible or ‘frozen’ order preferences can be. Some binomials – such as
spic and span – appear to be relatively fixed in a particular order, whilst others –
such as Bill and Mary – do not. Benor and Levy (2006) posited that order prefer-
ences might be more frozen when Cooper and Ross’ (1975) rules align, and less
frozen when those rules are misaligned. Mollin (2012) found that only 18% of the
binomial pairs in her corpus were truly irreversible in the sense of appearing in
the same order 100% of the time. The other 82% were “distributed on a cline of
reversibility.” Most recently, Mollin (in press) used the Google Books n-gram to
examine the reversibility of binomials in written modern English between 1800
and 2000. She examined each decade for the frequency of occurrence of each
order of the most commonly-occurring binomials within the 1800–2000 period.
Of the 206 common binomials for which trends could be analyzed, 101 showed
trends towards freezing order preferences, 52 toward unfreezing, 10 showed non-
linear trends and 43 showed no trend at all. Thus, historical change toward both
freezing and unfreezing are ordinary in English. In sum, the literature on binomi-
als in English increasingly converges on the view that order preferences are pri-
marily a result of semantic features, particularly beliefs about agency, animacy,
power, prototypicality, and closeness to the self.
74       Peter Hegarty

Unsurprisingly, such preferences can and do change over historical time. The
preference for ladies and gentlemen is a case in point. Mollin notes that there is
a strong preference for ladies and gentlemen over gentlemen and ladies in 1800–
2000, but also that Potter (1972, cited in Mollin, in press) had earlier observed a
preference for gentil men before ladies in Chaucer’s texts. My suspicion that this
binomial may have reversed in the late 1800s was sparked by the on-line OED’s
quotation from 1808, in the entry on gentlemen, to the effect that “All public
addresses to a mixed assembly of both sexes, till sixty years ago, commenced
Gentlemen and Ladies: at present it is Ladies and Gentlemen.” A combined
searches of the following British Library databases: 17th–18th Century Burney Col-
lection Newspapers, The Times Digital Archive, and the Times Literary Supple-
ment Digital Archive for articles using the phrases ladies and gentlemen and gen-
tlemen and ladies over the period 1700–1900 confirmed that the OED informant
was a reliable informant as to patterns of late 18th century English. Figure 1 shows
a clear change of what Mollin (in press) has called ‘freezing with changing prefer-
ence.’ Within this period, ladies and gentlemen becomes preferred. The number
of newspaper articles including the phrase gentlemen and ladies and ladies and
gentlemen are both strongly correlated with chronological year over the period
1705–1800, Pearson’s r (51) = – .79, +.78 respectively, both p <.001. Correlation sta-
tistics of this magnitude are only observed in the absence of genuine correlation
in less than 1 in a 1000 cases.

4 E
 xplaining Order for Names in Binomials, I:
­Phonological and Frequency

Thus, assuming there exist, in real life or in fiction, two playmates, Ván’a, and Mít’a the reasons
for any mention of them, in conversation, report, oral story, or fine literature as R[ussian]
Ván’a i Mít’a rather than Mít’a i Ván’a may be effectively explored in sociological, psychologi-
cal or esthetic terms (margin of age, order of appearance, closeness to narrator, importance
of rôle, etc.) If there emerges a schema of definite preference, linguistic conditions are likely to
have acted, as best, as a lubricant (Malkiel, 1959: 119).

Given that preferences for binomial phrases that name categories can change
over time, it would be remarkable if preferences to name individual men
before women – such as Bill and Mary – were not yet more situation-specific.
After all, Bill and Mary could be anybody. Malkiel (1959) suggested that psy-
chological factors such as roles and closeness to the speaker might be more
determinative of order preferences than any ‘lubricating’ effects of the linguis-
 Ladies and gentlemen: Word order and gender in English       75

tics of names. Similarly, Allan (1987: 52) suggested that proximity might be the
primary determinant of order in binomials and that pairs of people are named
with the closer partner first. And yet, the most sustained linguistic investiga-
tion of the preference to name men before women in English de-emphasized
semantic constraints. Wright et al. (2005) noted how Cooper and Ross’ phono-
logical hypotheses overlapped with statistical differences between women’s
and men’s names in English. In English, men’s names have fewer syllables,
and are more likely to begin and end with consonants. These statistical dif-
ferences may reflect different historical roots (Hough, 2000), but whatever
their source, the differences are sufficiently robust that connectionist models,
and children and adults who speak English can all consistently predict the
gender of real names and nonsense names (Cassidy, Kelly and Sharoni, 1999;
Lieberson and Mikelson, 1995; Whissell, 2001). Fenk-Oczlon’s (1989) hypoth-
esis about word frequency may also explain a preference to name men first.
In English, the most popular names for boys are given to more children than
are the most popular names for girls (Lieberson and Bell, 1992). Wright et al.
(2005) described how these differences between the phonology of conven-
tional women’s and men’s names lead to a “conspiracy” to prefer to name
pairs of people with the man’s name first.

Figure 1: Number of Articles Mentioning Ladies and Gentlemen and Gentlemen and Ladies in
18th Century English Newspapers.
76       Peter Hegarty

This hypothesis has received some support in empirical studies. Wright and
Hay (2002) used internet searches to document a preference to name opposite-sex
pairs with men’s names first (e.g., Bill and Mary) rather than women’s names first
(e.g., Mary and Bill). Wright et  al. (2005) reported two experiments that tested
their hypotheses that phonological features and name popularity explain why
people prefer binomial phrases in which men’s names precede women’s names.
In the first, forty-six students ordered two names (e.g, Tammy and Freddy) in a
binomial phrase to complete a sentence (e.g., _____ and _____ went to the yogurt
factory). Participants, particularly men, demonstrated a preference to position
men’s names first when a man’s and a woman’s name were presented. When
the names were of the same gender, participants, particularly women, showed
a preference to position the shorter name first. Post hoc analysis also suggested
a preference to position the name with longer vowel sounds second. In a second
experiment with twenty-eight student participants, the obstruency of consonants
was varied across the names used as stimuli. This experiment replicated the pref-
erence to name men first. When the names were of the same gender, participants
preferred to position the name with a final obstruent consonant sound first. Post
hoc analysis of both experiments showed a preference to position more common
names first.

5 E
 xplaining Order for Names in Binomials, II:
Semantics
Wright et  al.’s (2005) phonological features could not completely account for
the strength of the preference to name men before women, nor for the particular
strength of this preference among men. Yet they concluded that “a number of
phonological constraints condition the optimal ordering of binomial pairs” and
that “male names tend to be characterized by first position phonology, whilst
female names tend to be characterized by second position phonology” (Wright
et  al., 2005: 558). However, the conclusion that features of names rather than
semantic beliefs about the people being named affect order preferences seems
unlikely for several reasons. First, it is inconsistent with the primacy of seman-
tics in the literature (Benor and Levy, 2006; Cooper and Ross, 1975; McDonald
et al., 1993; Mollin, 2012). Second, Wright et al.’s (2005) experiments may have
included “demand characteristics” (Orne, 1962) to focus on features of the names
to a greater extent than is typical in everyday use of English. As Hilton (1995)
notes, Gricean pragmatics shape the communication between experimenters and
researchers. Wright et al.’s (1995) demand that participants express a preference
 Ladies and gentlemen: Word order and gender in English       77

for one of two sentences that vary only in the order of names implicitly communi-
cates that the experimenter is interested in order preferences, directing attention
toward features of the names. This demand may not be relevant to situations in
which people discuss trips to yogurt factories and their attention is focused on
the actors named in the account, rather than the names of those actors per se.
Cooper and Ross’ (1975) work also predicts why men might be named before
women for semantic reasons. Several of their rules are relevant to the common
stereotype that men are more agentic kinds of people than are women (Diekman
and Eagly, 2000; Eagly, 1987; Hoffman and Hurst, 1990), including their “agentic
first rule” and “power source first” rule. Subsequent research has confirmed
these hypotheses about preferences to put animate things first (McDonald et al.,
1993) and powerful things first (Mollin, 2012). Can gender stereotypes explain
how the social gender hierarchy is translated into binomial order preferences?
Such a hypothesis would suggest that preferences for name order will be rela-
tively unfrozen, because people are not always viewed through the lens of gender
stereotypes. Rather, such stereotypes colour judgment about strangers about
whom we know little individuating information and not very much about people
that we know well (Fiske and Neuberg, 1990; Deaux and Major, 1987). Two empir-
ical studies demonstrated how the constraints on name order preferences vary
in predictable ways depending on whether speaker are addressing strangers in
public or referring to people that they know well (Hegarty, Watson, Fletcher, and
McQueen, 2011, Studies 1 & 2).
The first study was a replication of Wright and Hay’s (2002) internet search
using 200 popular name pairs (Hegarty et  al., 2011, Study 1). Approximately
three times as many hits were retrieved for men-first name combinations (e.g.,
John and Mary) than for equivalent women-first combinations (e.g,. Mary and
John). On the internet, there is a clear preference to name men before women.
However, preferences for order look different when we address the people we
know and love the most, as theories of gender stereotyping would predict. In a
second study, seventeen informants who inhabited opposite-sex couples shared
the greeting cards received over one Christmas period, and told us whether
each card was sent by someone closer to them, to their partner, or equally close
to both (Hegarty et al., 2011, Study 5). Among the 492 cards that were sent by
someone closer to one of the partners, 433 cards, or 88% of the total, addressed
the closer partner first. These findings are harder to square with a phonologi-
cal theory of name order based on statistical difference between women’s and
men’s names. However, they accord perfectly with gender stereotyping theory;
people stereotype men as agentic people and position their names first in rel-
atively anonymous communication only (Deaux and Major, 1987; Fiske and
Neuberg, 1990). Other semantic rules, such as the “me first” rule affect private
78       Peter Hegarty

communication with people we know well as individuals and are unlikley to

stereotype.
Earlier, I noted that Cooper and Ross’ (1975) theory had failed to fully
explore the consequences of the fact that the “me first” rule and the “proto-
typical first” rule were mismatched for women and matched for men speak-
ers. Another study examined consequences of the ‘me first’ rule for ordering
names of couples of varying levels of familiarity (Hegarty et al., 2011, Study 6).
Thirty nine women and thirty-eight men each listed the names of five couples
in their families, five couples among their friends, and the names of five imagi-
nary couples. Men positioned men’s names first when naming family and
friend couples in 62% and 73% of cases respectively. However, women posi-
tioned women’s names first in these two contexts, 66%, 68% respectively. Typi-
cally, people have more close friendships with members of the same gender
(Caldwell and Peplau, 1982). A me first preference might lead to a male-first
preference for most men and a female-first preference for most women when
naming well-known people.
In this same study, men more often positioned men’s names first when
naming imaginary couples than women did (74%, 55% respectively). This gender
difference is consistent with Wright et al.’s (2005) finding. Possibly, people call
imaginary couples to mind in ways that are influenced both by their habits of
naming friends and family, and by their gender stereotypes. Both factors lead
men to name men first, but the two factors have opposing effects on the order
preferences of women. In support of this interpretation, I present three more
experiments that evidence a direct influence of gender stereotypes on the order
in which members of imaginary couples are named.

6 D
 irect Evidence of the Effects of Stereotyping on
Order in Binomials
If gender stereotypes affect order preferences in binomials, then men should be
positioned first in those binomials when the couple being described is believed
to conform to gender stereotypes, but not when the couple are perceived to live
in a way that disconfirms those same stereotypes. In the first experiment to test
this hypothesis, eighty-six women and thirty-five men British students were
randomly assigned to read instructions to think about a couple who either ‘are
quite traditional, and who conform strictly to gender scripts about how the two
genders should behave’ or who ‘are quite non-traditional and who deviate radi-
cally from gender scripts about how the two genders should behave’ (Hegarty
 Ladies and gentlemen: Word order and gender in English       79

et al., 2011, Study 2). In both conditions, participants answered a series of ques-
tions about the division of labour and personal interests of the imagined couple
that confirmed that they had called to mind a relatively traditional or non-tradi-
tional heterosexual couple. Next, participants listed five name combinations for
the imaginary couple. Confirming predictions, combinations of names for tradi-
tional couples included a disproportionate number of male first combinations
whilst those for non-traditional couples did not (69%, 49% respectively). As in
the study described above, men positioned men’s names first more often than
women did (73%, 53% respectively), independent of the stereotyping manipula-
tion, F < 1 (see Table 1).
The second study used historical time as a proxy for gender stereotypical-
ity, and drew on the finding that students typically consider gender norms to be
eroding over historical time (Diekman and Eagly, 2000). Consequently, we pre-
dicted that students would be more likely to list men’s names first when con-
sidering couples who lived in earlier periods of time. Eighty women and eighty
men listed the names of ten couples living in either the 1920s, 1950s, 1980s or
the 21st century (Hegarty et al., 2011, Study 3). Again we calculated the percent-
age of combinations that positioned the man’s name first. As in the first study,
men produced more male-first name combinations than women did. Moreover,
the historical time manipulation affected the way women, but not men ordered
names for these couples (see Table 1).
The third study addressed a more subtle prediction from gender stereo-
typing theory. Gender stereotypes are not simply implicit beliefs about the
attributes of women and men, or ‘conceptual baggage’ associated with those
terms (McConnell-Ginet, this volume). Rather, gender stereotypes are ‘notional
gender systems’ that are not directly tied to the physical attributes that conven-
tionally define sex (see also McConnell-Ginet, this volume). Gender stereotypes
are networks of association that lead people to believe that a wide range of fea-
tures; physical attributes, interests, occupational choices, and sexual orienta-
tions will all co-vary as if they made up a single dimension of masculinity-fem-
ininity (Deaux and Lewis, 1984). Gender categories (e.g., women, men) are not
the most persistently active nodes in gender stereotypes. When placed under
cognitive load, semantic associations between other elements of gender stereo-
types remain active even after associations between such traits and gender cat-
egories have become inactive (Pratto and Bargh, 1991). Consequently, gender
stereotyping research predicts that people will tend to name the more mascu-
line partner in a romantic couple first, even when naming partners in a same-
sex couple.
80       Peter Hegarty

Table 1: Proportion of Couples Named with Male Name First by Participant Gender and Imag-
ined Decade (Historical Study).

Decade 1920s 1950s 1980s 21st Century

Participant Gender
Female .69* .68* .39 .42
Male .63 .70** .76*** .84***

* p <.05, **p <.01, ***p <.001. Significance tests are two-tailed t-tests testing deviation from a
theoretical mean of .50 within each cell of the experiment’s design.

To test this hypothesis, forty-seven women and thirty-nine men were randomly
assigned to conditions in which they called to mind either an imaginary lesbian
couple or an imaginary gay couple (Hegarty et  al., 2011, Study 4). In all cases,
partners were given names using the following prompt: ‘My imaginary couple are
called _____ and _____.’ Next, six items required participants to describe differ-
ences between the partners that were relevant to gender stereotypes by writing
the partner’s names into six comparative statements, endowing each attribute to
one partner more than to the other. For example, the final two items pertaining to
physical attributes were presented as follows:
_____ is physically smaller than _____
_____ is physically stronger than _____

Listing a partner’s name first in the first item suggests that the partner was con-
sidered stereotypically feminine. Listing a partner’s name first in the second
item suggests that this partner was considered stereotypically masculine. Notice
that these items were carefully written so that a participant who positioned one
partner’s name first in responding to all items would endow that partner with
an equal number of stereotypically feminine and stereotypically masculine attri-
butes.
Consistent with predictions, first named partners were endowed with attri-
butes that were more likely to be masculine than feminine (68% vs. 32%), irre-
spective of whether participants were women and men, or were asked to imagine
a lesbian or a gay couple. Some participants were recruited on one British campus
and others participated in a psychology of gender class at a second British
campus. Recruitment method did not affect this pattern or results. This experi-
ment provides strong evidence of an association between notional gender and
the order of names in binomials. Thus, even when same-sex couples are named,
the partner deemed to be masculine is named first.
 Ladies and gentlemen: Word order and gender in English       81

7 The Phonological Conspiracy Re-Examined

Jointly, these three studies provide compelling evidence of an influence of gender
stereotypes on the ways that we name couples that we do not know well. Next,
I report a new analysis of these data that tests elements of Wright et al.’s (2005)
theory. First, recall that in English, men’s names typically have fewer syllables than
women’s names (Cutler et  al., 1990), and that Wright et  al. (2005) found a pref-
erence for sentences in which shorter names appeared first. In the three studies
described above, the names that were generated for men partners consistently had
fewer syllables than the names produced for women partners (see Table 2). There
were weak trends to position shorter women’s names first in all three studies, but
to put shorter men’s names first in only one of three studies. Overall, these data do
not support the preference to position shorter words first in binomials.
Wright et  al. also found preferences to position names with longer vowels
in the second name and final obstruent consonants in the first name. Table  3
shows how more of the men’s names than women’s names began and ended
with obstruent consonants, whilst more of the women’s names began and ended
with vowels. Some supporting evidence for the phonological conspiracy emerged
only in the historical study. Chi-square tests confirmed that among the women’s
names in this study, a significantly greater number of those positioned first ended
in obstruent consonants and a significantly greater number of those positioned
second ended in vowels, χ2 (1, 1600) = 4.59, 6.45 respectively, both p<.05. Among
male names, a significantly greater number of those positioned first ended in
obstruent consonants and a significantly greater number of those positioned
second ended in sonorant consonants, χ2 (1, 1600) = 6.33, 7.57 respectively p<.05,
p<.01 respectively. These findings are consistent with Wright et al.’s (2005) predic-
tions. However, it is notable that they emerge only in this study in which partici-
pants were briefed that the study concerned knowledge about naming traditions.
A demand to focus on features of names may have been evident in this study as in
Wright et al. (2005), but not in the other experiments.
Finally, I tested Fenk-Oczlon’s (1989) hypothesis about word popularity. The
UK census office recorded the one hundred most popular names given to girls and
to boys in England and Wales across the 20th century, every ten years, in years
ending with the number 4. As study participants were largely born between 1984
and 1994, I used both the 1984 and 1994 data as proxy measures of name popular-
ity. Table 3 shows both the proportion of names mentioned by participants among
the top one hundred names, and the mean rank of those names according to both
the 1984 and 1994 lists. As participants in Study 2 were briefed that the study tapped
knowledge of naming trends, only data from the 21st century condition of that study
were analyzed here. Table 2 shows that men’s names produced in the studies were
Table 2: Phonological Characteristics and Popularity of First and Second Named Female and Male Names (Studies 4–6).
82   

Study Study 4 Study 5 Study 6

Gender of Names Female Male Female Male Female Male

Position 1 2 1 2 1 2 1 2 1 2 1 2

Count of Names 230 353 353 230 575 1025 1025 575 42 42 44 44
   Peter Hegarty

Syllables 1.99 2.20 1.51 1.50 2.18 2.22 1.65 1.63 1.79 1.93 1.38 1.45
Name Beginning
PVowel .157 .136 .105 .087 .151 .179 .126 .134 .149 .190 .159 .091
PSonorant .242 .252 .181 .213 .252 .247 .232 .236 .238 .333 .182 .159
PObstruent .622 .615 .714 .700 .597 .574 .642 .630 .619 .476 .659 .750
Name Ending
PVowel .609 .632 .091 .078 .508 .573 .127 .125 .619 .595 .114 .114
PSonorant .270 .240 .456 .496 .252 .233 .350 .417 .262 .262 .455 .386
PObstruent .110 .127 .453 .426 .240 .193 .523 .457 .119 .143 .432 .500
Name Popularity (1984)
PMention .477 .491 .537 .543 .810 .638 .433 .622 .452 .428 .409 .636
Rank 41.7 40.2 31.5 29.5 31.7 33.4 38.0 32.0 43.2 53.0 34.1 34.2
Name Popularity (1994)
PMention .365 .337 .567 .568 .608 .575 .488 .500 .286 .333 .409 .568
Rank 36.0 32.0 38.7 37.8 30.7 30.7 40.0 41.5 32.3 46.7 46.6 31.0
are positioned first because they are more popular than women’s.

Note: PVowel = Proportion of names beginning or ending in a vowel, PSonorant = Proportion of names beginning or ending in
sonorant consonants (i.e.,. /l/, /m/, /n/, /ŋ/, /ɹ/, /w/, /j/), PObstruent = Proportion of names beginning of ending in obstruent
consonants (i.e., all other consonants). PMention = Proportion of names mentioned among 100 most popular baby names for
girls and boys in 1984 or 1994, Rank = Mean rank of mentioned names among lists of most popular baby names.
popularity measure. This analysis lends little support to the idea that men’s names
more popular than women’s names produced in the studies. There are considerable
differences in trends across studies, and across the year used to operationalize the
 Ladies and gentlemen: Word order and gender in English       83

8 Conclusions
I have argued that people order names and category terms that reference gen-
dered people in situation specific ways that are influenced by historical and situ-
ation-specific beliefs about groups and individuals. Whilst the literature on bino-
mials has repeatedly commented on the reasons why it is idiomatic to name men
before women in many contexts, that literature has tended to emphasize features
of the words (including names), rather than speaker’s beliefs about the people
whom they reference. In particular, studies with nonsense terms (e.g., Pinker
and Birdsong, 1979) and studies using names without any semantic content (e.g.,
Wright et al., 2005) can yield results which isolate phonological features affecting
order preferences. However, the consensus emerging from analyses of corpora
(Benor and Levy, 2006; Mollin, 2012), and from experiments (Hegarty et al., 2011;
McDonald et al., 1993) is that semantic beliefs about the people and things that
we talk about affects the order in which we talk about them. I want to conclude
by positioning these findings in three broad contexts.
First, rules for naming people in binomials in English are variable and
subject to psychological and sociological explanation as Malkiel (1959) pre-
dicted. Speakers of English are guided by their beliefs about whom they are
closest to and by gender stereotypes when they chose one of two orders to use
when talking about a romantic couple. These rules may not be conscious choices
by English speakers, just as speakers of Mohawk often fail to notice that they
have two different ways of referring to women, or that they use these two systems
in different social contexts depending on familiarity, politeness, status and other
factors (see Mithun this volume). English speakers who may choose to address
their Christmas cards to Mary and Bill or to Bill and Mary are similar to speakers
of Mohawk in this regard.
Second the influence of semantics on word order in binomials that reference
gender is consistent with a growing body of psychological evidence about the
ways that semantic beliefs influence the way that information is ordered in both
linguistic and pictorial representations. People who read languages written left-
to-right or right-to-left learn spatial schemas that associate action with the direc-
tion of their written language, and agency with the starting position in their lan-
guage (Chatterjee, Southwood, and Basilico, 1999; Maass and Russo, 2003). When
couples are imagined to inhabit gender stereotypes they are not only named with
men first, but depicted as such (Maass, Suitner, Favaretto, and Cignacchi, 2009).
Even when graphing gender differences, stereotypes about power lead men to be
graphed first and women second (Hegarty, Lemieux, and McQueen, 2010). Con-
sistent with the stereotyping approach adopted here, conventions for represent-
ing women and men in portraits can change with time and vary with gender. For
84       Peter Hegarty

example, in recent centuries European women artists have increasingly broken

with the convention to represent men facing to the right and women facing to the
left (Suitner and Maass, 2007). Within the context of this body of evidence, the
effects of gender stereotypes on a form of linguistic order that is also affected by
semantic beliefs about agency are less surprising.
Finally, the gender stereotyping orientation presented here suggests the merit
of McConnell-Ginet’s (this volume) argument to de-couple assumptions about “sex”
and “gender” in linguistics. It is important to recognize that the sexist proscriptions
for male-first binomials described by Bodine (1975) continue to resonate in English,
because male-first binomial terms are often taken to be a non-sexist alternative to
masculine generics. Malkiel (1959, p. 144) noted that such expansions from man
power to man and woman power in his own time were rare. However, the increased
attention to masculine generics as sexist language in the late 20th century has not
always been coupled with a recognition of the sexist conventions that congeal in
the male-first binomials that masculine generics truncated.
Finally, I wish to return to Cooper and Ross’ vague conclusion about the
status of normative women and men speakers of English in their analysis. Pre-
scriptive gender stereotypes exclude or marginalize some people systematically;
gay men, lesbians, transgender people, and hjiras, for example. By analyzing the
language use of all groups on an equal footing, then the notional gender system
that is meted out in noun and pronoun use comes into view (McConnell-Ginet,
this volume). Similarly, asking people to name partners of imaginary same-sex
couples helped to decouple the effects of the semantics, phonology, and popular-
ity that are confounded with conventionally gendered names in English. As such,
I hope that this work spurs the impulse within English language scholarship to
queer the disciplines by taking the experiences of lesbian, gay and queer sub-
cultures as reference points in their own right rather than assuming them to be
deviations or particulars that cannot serve as paradigms or general frameworks
(Butler, 1990; De Lauretis, 1991). I wish to end this chapter being explicit on this
point, much as Cooper and Ross (1975) were implicit in the end.

9 References
Allan, Keith. 1987. Hierarchies and the choice of left conjucts (with particular attention to
English). Journal of Linguistics 23. 51–77.
Bem, Sandra L. 1993. The lenses of gender: Transforming the debate on sexual inequality, New
Haven, CT: Yale University Press.
Benor, Sarah B. and Roger Levy. 2006. The chicken or the egg? A probabilistic analysis of
English binomials. Language 82. 233–278.
 Ladies and gentlemen: Word order and gender in English       85

Bodine, Anne. 1975. Androcentrism in prescriptive grammar. Language in Society 4. 129–146.

Butler, Judith. 1990. Gender trouble. New York: Routledge.
Caldwell, Mayta A. and Letitia Anne Peplau. 1982. Sex differences in same-sex friendship. Sex
Roles 8. 721–732.
Cassidy, Kimberley W., Michael H. Kelly, and Lee’at J. Sharoni. 1999. Inferring gender from name
phonology. Journal of Experimental Psychology: General 128. 362–381.
Chatterjee, Anjan, M. Helen Southwood, and David Basilico. 1999. Verbs, events, and spatial
representations. Neuropsychologia 37. 395–402.
Cooper, William E. and John Robert Ross. 1975. World order. In: Robin E. Grossman, L.
James San and Timothy J. Vance (Editors). Chicago Linguistic Society: Papers from the
Parasession on Functionalism. 63–111. Chicago, IL: Chicago Linguistic Society.
Cutler, Anne, James McQueen, and Ken Robinson. 1990. Elizabeth and John: Sound Patterns of
men’s and women’s names. Journal of Linguistics 26. 471–482.
de Lauretis, Teresa. 1991. Queer theory: Lesbian and gay sexualities. differences: A Journal of
Feminist Cultural Studies 3. iii–xviii.
Deaux, Kay and Laurie L. Lewis. 1984. Structure of gender stereotypes: Interrelationships
among components and gender label. Journal of Personality and Social Psychology 46.
991–1000.
Deaux, Kay and Brenda Major. 1987. Putting gender into context: An interactive model of
gender-related behaviour. Psychological Review 94. 369–389.
Diekman, Amanda B. and Alice H. Eagly. 2000. Stereotypes as dynamic constructs: Women
and men of the past, present, and future. Personality and Social Psychology Bulletin 26.
1171–1188.
Eagly, Alice H. 1987. Sex differences in social behavior: A social-role interpretation. Hillsdale,
NJ: Erlbaum.
Fenk-Oczlon, Gertraud. 1989. Word frequency and word order in freezes. Linguistics 27.
517–556.
Fiske, Susan T. and Steven L. Neuberg. 1990. A continuum of impression formation from
category-based to individuating processes: Influences of information and motivation on
attention and interpretation. In: Mark P. Zanna (ed.), Advances in Experimental Social
Psychology 1–74. San Diego, CA: Academic Press.
Hegarty, Peter, Anthony F. Lemieux, and Grant McQueen. 2010. Graphing the order of the sexes:
Constructing, recalling, interpreting, and putting the self in gender difference graphs.
Journal of Personality and Social Psychology 98. 375–391.
Hegarty, Peter, Nila Watson, Katie Fletcher, and Grant McQueen. 2011. When gentlemen are
first and ladies are last. Effects of gender stereotypes on the order of romantic partners’
names. British Journal of Social Psychology 50. 21–35.
Hilton, Denis J. 1995. The social context of reasoning; Conversational inference and rational
judgment. Psychological Bulletin 118. 248–271.
Hoffman, Curt and Nancy Hurst. 1990. Gender stereotypes: Perception or rationalization?
Journal of Personality and Social Psychology 58. 197–208.
Hough, Carole. 2000. Towards an explanation of phonetic differentiation in masculine and
feminine personal names. Journal of Linguistics 36. 1–11.
Kelly, Michael H., J. Kathryn Bock, and Frank C. Keil. 1986. Prototypicality in a linguistic context:
Effects on sentence structure. Journal of Memory and Language 25. 59–74.
Lieberson, Stanley and Eleanor O. Bell. 1992. Children’s first names: An empirical study of
social taste. American Journal of Sociology 98. 511–554.
86       Peter Hegarty

Lieberson, Stanley and Mikelson, Kelly S. 1995. Distinctive African-American names: An

experimental, historical, and linguistic analysis of innovation. American Sociological
Review 60. 928–946.
Maass, Anne and Aurore Russo. 2003. Directional bias in the mental representation of spatial
events: Nature or culture? Psychological Science 14. 296–301.
Maass, Anne, Catarina Suitner, Xenia Favaretto, and Marina Cignacchi. 2009. Groups in space:
Stereotypes and the spatial agency bias. Journal of Experimental Social Psychology 45.
496–504.
Malkiel, Yakov. 1959. Studies in irreversible binomials. Lingua 8. 113–160.
McDonald, Janet L., Kathryn Bock, and Michael H. Kelly. 1993. Word and world order: Semantic,
phonological, and metrical determinants of serial position. Cognitive Psychology 25.
188–230.
Mollin, Sandra. 2012. Revisiting binomial order in English: Ordering constraints and revers-
ibility. English Language and Linguistics 16. 81–103.
Mollin, Sandra in press. Pathways of change in the diachronic development of binomial revers-
ibility in late modern American English. Journal of English Linguisitics.
Orne, Martin. 1962. On the social psychology of the psychological experiment; With particular
reference to demand characteristics and their implications American Psychologist 17.
776–783.
Pinker, Steven and David Birdsong. 1979. Speakers’ sensitivity to rules of frozen word order.
Journal of Verbal Learning and Verbal Behavior 18. 497–508.
Potter, Simeon. 1972. Chaucer’s untransposable binomials. Neuphilogoische Mitteilungen 73.
309–314.
Pratto, Felicia and John A. Bargh. 1991. Stereotyping based on apparently individuating
information: Trait and global components of sex stereotypes under attention overload.
Journal of Experimental Social Psychology 27. 26–74.
Rosch, Eleanor and Carolyn B. Mervis. 1975. Family resemblances: Studies in the internal
structure of categories. Cognitive Psychology 7. 573–605.
Suitner, Catarina and Anne Maass. 2007. Positioning bias in portraits and self-portraits: Do
female artists make different choices? Empirical Studies of the Arts 25. 71–95.
Whissell, Cynthia. 2001. Cues to referent gender in randomly constructed names. Perceptual
and Motor Skills 93. 856–858.
Wright, S. and Hay, J. 2002. Fred and Wilma: A phonological conspiracy. In: Sarah Benor, Mary
Rose, Devyani Sharma, Julie Sweetland, and Quing Zhang (eds.), Gender and Linguistic
Practice. 175–191. Stanford, CA: CSLI Publications.
Wright, Saundra K., Jennifer Hay, and Tessa Bent. 2005. Ladies first? Phonology, frequency, and
the naming conspiracy. Linguistics, 43. 531–535.
Greville G. Corbett
Gender typology
1 Introduction: Is gender special?1
If we compare gender with the other morphosyntactic features, it seems evident
that gender stands out. As Ranko Matasović puts it: ‘… gender is perhaps the only
grammatical category that ever evoked passion – and not only among linguists.’
(2004: 13). Non-linguists as well as linguists argue about what is “right” or appro-
priate in the use of gender, whether they are concerned with the appropriate rec-
ognition of the status of women, or with the agreement found with recent loan-
words. In the same source, however, besides suggesting that gender is ‘unlike all
other grammatical categories of nouns’, Matasović (2004: 18) also writes: ‘… there
is a sense in which gender is just one grammatical category among others.’ This is
clearly right too, as I shall show. I shall take a canonical approach (Corbett 2011),
examining what the canonical morphosyntactic feature would look like, and the
possible deviations from this canonical ideal. From this typological perspective,
we can contrast gender with the other morphosyntactic features. Current termi-
nology would suggest that the different types of deviation give grounds to sepa-
rate off gender as different from other morphosyntactic features. However, at this
level of abstraction, we see that the variations in gender systems are in fact inter-
estingly similar to those of other features. Gender is indeed special, but it is not
as different as is sometimes believed.2
I first outline the main ideas of Canonical Typology (§2); then look at the two
main issues in the typological analysis of gender systems. The first of these is the
analysis problem, which is concerned with whether a particular language has a
gender feature, and if so, how many values it has; we tackle this in §3. As will already

1 The support of the European Research Council (grant ERC-2008-AdG-230268 MORPHOLOGY)

and of the AHRC (grant: From competing theories to fieldwork: the challenge of an extreme
agreement system, and grant: Combining Gender and Classifiers in Natural Language) is grate-
fully acknowledged. I thank Marina Chumakina for help with Russian and Archi, Anna Thorn-
ton for suggestions for improving a draft, and Sebastian Fedden for advice on Mian, for helpful
comments on a draft and especially for discussion of recategorization. Versions were read at the
Workshop “The Expression of Gender”, Max Planck Institute for Psycholinguistics, Nijmegen,
4 March 2011, the Workshop “Exploring Grammatical Gender” within the 15th International Mor-
phology Meeting, Vienna, 9–12 February 2012, and the Third Summer School of the Marie Curie
Initial Training Network “Language, Gender and Cognition”, Potsdam, 14 June 2012. I am grate-
ful to all three audiences for their comments.
2 For a detailed recent bibliography on gender see Audring (2011).
88       Greville G. Corbett

be clear, I use ‘feature’ for constructs like gender, number, case and person (some,
like Matasović above, use the term ‘(grammatical) category’). Within each feature
there is a set of ‘values’: masculine, feminine, neuter and so on for gender, singular,
plural, dual, paucal and so on for number, and similarly for other features. Then
to the second main problem, the assignment problem, which covers the general
system according to which gender values are assigned to nouns, and the particular
rules for each value in a given language (§4). There is a brief conclusion (§5).

2 Canonical Typology
Let us think of a hypothetical language where every single verb, adjective and
adposition showed clear agreement in gender. We would propose a morpho-
syntactic feature gender, with certainty. If we did not, it would be hard to give
a convincing account of the syntax. On the other hand, if our hypothetical lan-
guage showed evidence of gender only in the personal pronoun, we would think
harder about proposing a gender feature. There are many real languages which
fall between these two extremes. With these we may be too ready to treat them
as though they were instances of the first type. The morphosyntactic features,
including gender, often have a ‘penumbra’ where the evidence is not straightfor-
ward, and needs careful analysis.
Typologists are naturally attracted to clusterings of properties. Certainly
where there is a problem with a particular feature it is often problematic in more
than one way. We want to establish whether these are significant clusterings of
properties or are no more than coincidences. One way in which we can address
this issue is to extend the theoretical space: then the clusters can be pulled apart.
In order to anchor this space, we start from the type of instance we mentioned
earlier  – the clearest instance of a feature. We use it to set up the properties
of a canonical feature and its values; we then have a point from which we can
measure the real examples we find. Naturally, the closer our real example is to
being canonical, the easier it is to argue for the use of a morphosyntactic feature.
As has been hinted at, to adopt a canonical approach is to take definitions to
their logical end point, which is the way we can build theoretical spaces of possibil-
ities. Only then do we investigate how this space is populated with real instances.
Canonical instances are those that match the canon: they are the best, clearest,
the indisputable ones. Given that they have to match up to a logically determined
standard, they are unlikely to be frequent. They are more likely to be rare, and may
even be non-existent. This is not a difficulty. The convergence of criteria fixes a
canonical point from which the phenomena actually found can be calibrated. This
 Gender typology       89

approach has been worked out particularly for inflectional morphology, as well
as for syntax. Inflectional morphology has been treated by Baerman, Brown and
Corbett (2005: 27–35), Spencer (2005), Stump (2005, 2006), Corbett (2007a), Niko-
laeva and Spencer (2008), Stump and Finkel (2008) and Thornton (2011). In syntax,
agreement has occupied centre stage, for instance in Corbett (2003, 2006), Comrie
(2003), Evans (2003), Polinsky (2003), Seifart (2005: 156–74) and Suthar (2006: 178–
198). There has been interesting work in other areas of linguistics too, from pho-
nology (Hyman 2009, 2011) to formal semantics (Fortin 2011) and computational
modeling (Sagot and Walther 2011). A working bibliography of this growing body
of research can be found at http://www.surrey.ac.uk/LIS/SMG/CanonicalTypology/
index.htm, and a volume of relevant work has appeared (Brown, Chumakina and
Corbett 2013). Particularly relevant to the current issue is the discussion of morpho-
syntactic features in Corbett (2011).
I should stress that canonical is not identical to prototypical (as the term is
normally used) since we have no requirement to produce a canonical exemplar;
rather we need to be able to define and so identify the canonical point. We should
also not confuse canonicity with being easy to find: the example which is fre-
quently cited may not be a fully canonical instance of a phenomenon.

3 The analysis problem

We need to be able to determine first, whether a particular language has a given
feature, and of course our main focus will be on gender, and second, how many
values that feature has. ‘Gender’ derives from Latin ‘genus’ via Old French ‘gendre’,
and it originally meant ‘sort’ or ‘kind’. Nouns come in many different kinds: those
with initial stress, those ending in a consonant, those with an irregular plural,
those denoting instruments, and so on. These are not gender values. ‘Gender’ is
normally used for kinds of noun which are ‘reflected in the behavior of associated
words’ (Hockett 1958: 231). This point is important, accepted by many, and often
forgotten. In order to compare across languages in a sensible way, we need this
means of classification which is motived from outside the noun itself. What then
does Hockett’s definition mean? The relevant ‘reflection’ in the associated words
is agreement (which for some linguists includes antecedent-anaphor relations).
That is, we divide the inventory of nouns into different kinds according to the
different agreements they control. This demonstrates the existence of a gender
system, and we can then ask about which gender values it contains. If we apply
this type of analysis, we find that some familiar languages, the different gender
values have a semantic core based on sex (thus Russian nouns divide into three
90       Greville G. Corbett

kinds, and nouns denoting males, though not only these, group together, and
those denoting females also group in another gender value). In other languages
the structures may be very similar but the semantic core is based not on sex, but
for instance on human versus non-human or animate versus inanimate. Thus a
language has a gender system only if noun phrases headed by nouns of different
types control different agreements. No amount of marking on a noun can prove
that the language has a gender system; the evidence that nouns have gender
values in a given language lies in the agreement targets which show gender.

3.1 A clear instance of evidence for a gender system

Let us take first a clear example from Russian:

Russian

(1) Žurnal by-l zdes’.

magazine be-pst[m] here.
‘The magazine was here.’ (And now it’s gone.)

(2) Kniga by-l-a zdes’.

book be-pst-f here.
‘The book was here.’

(3) Pis’mo by-l-o zdes’.

letter be-pst-n here.
‘The letter was here.’

In these three examples we see different forms of the verb: the bare stem in (1)
and different inflections in (2) and (3). The number of the head nouns has been
kept constant, as have all other potentially interfering factors. It follows that the
language has a gender system, and the three nouns have different gender values.
The approach to gender which I have outlined rests on the notion of ‘agree-
ment class’ (Zaliznjak 1964). For nouns to be in the same agreement class they
must take the same agreements under all conditions, that is, if we hold constant
the values of other features such as case and number. If two nouns differ in their
agreements, when all other relevant factors are held constant, they belong to
two different agreement classes. Normally this will mean that they have differ-
ent gender values. In our Russian examples, it is clear that žurnal ‘magazine’,
kniga ‘book’ and pis’mo ‘letter’ have different gender values. Each of these nouns
 Gender typology       91

r­epresents several thousand more. In (1) we could replace the noun with otec
‘father’ and many other nouns denoting males; this gender value is convention-
ally called ‘masculine’, even though most nouns it includes, like žurnal ‘mag-
azine’, do not denote males. Similarly in (2), in place of kniga ‘book’ we could
have a noun like mat’ ‘mother’, and these nouns are members of the feminine
gender. Again, the majority of the nouns involved do not denote females. Finally
the nouns like pis’mo ‘letter’ comprise the neuter gender value.
There are languages where we can see quite easily that they have a gender
system, and the number of gender values is evident. However, for some languages
the number of gender values has given rise to debate, even the passionate debate
of which Matasović speaks. A good example is Romanian, about which there
has been a long and sometimes quite heated discussion. The analytical problem
posed by Romanian is genuinely interesting, as we shall see shortly. It matters
that we are consistent in our analyses, since otherwise when we move to typology
we are not comparing like with like. For instance, it is regularly stated that the
Nakh language Batsbi (also known as Tsova Tush) has eight gender values. In a
way it does. However, if we were to analyse French in the same way, we would say
that French has three gender values. Alternatively, if we were to analyse Batsbi
as French is normally analysed we would find that Batsbi has five gender values.
Thus the analytical decisions are not always straightforward. We think of fea-
tures and values as being clean and neat. However they can have a less clear area,
a penumbra around the clear core. To investigate this tricky area it proves helpful
to start from the clearest examples, the canonical ones.

3.2 P
 rinciples for canonical features and their values

It will be helpful to have a yardstick against which to measure the examples we

find. We therefore consider the ideal morphosyntactic feature and how it would
behave. I suggest three overarching principles (Corbett 2011: 450, 458):
I: Canonical features and their values are clearly distinguished by formal
means.
II: The use of canonical morphosyntactic features and their values is deter-
mined by simple syntactic rules.
III: Canonical morphosyntactic features and their values are expressed by canon-
ical inflectional morphology.

Non-canonical instances may deviate in terms of the feature as a whole (thus

gender may be non-canonical in particular respects in a given language) and par-
ticular values may also be non-canonical (neuter might be non-canonical in a
92       Greville G. Corbett

way that did not affect other gender values). We shall not consider all the possi-
bilities, but rather home in on some which are particularly interesting for gender,
in comparison with the other morphosyntactic features. We shall concentrate on
Principle I and on some of the more specific criteria which it covers.
According to Principle I, canonical features and their values are clearly dis-
tinguished by formal means; given the general philosophy of the approach, it
follows that if we compare the formal means by which a feature or value is distin-
guished, then the clearer they are the more canonical that feature or value. But
what does ‘clear’ mean here? It means that there is a straightforward and regular
mapping from form to function. As a result, in the canonical situation there is
clear evidence for a given feature, gender for instance, and its values. We shall
consider four of the criteria which make this general principle more specific.

Criterion 1
Canonical features and their values have a dedicated form

We might very reasonably assume that in order to postulate a feature, and its
values, we should be able in each instance to point to an inflected form as justifi-
cation; ideally, it should be possible to explain this form only in terms of the par-
ticular feature and value. Values that can be justified in this way may be termed
‘autonomous’ (Zaliznjak 1973: 69–74; Mel'čuk 1986: 66–70). Naturally we can look
for non-canonical situations where there is no unique form to make the value
autonomous. Consider this situation, first at the abstract level, applicable to any
morphosyntactic feature:

a d
a e
b e

Figure 1: A non-autonomous case value

In Figure 1, ‘a’, ‘b’ and so on represent fully inflected forms. The paradigm repre-
sents two orthogonal features. For the feature represented by the rows, the issue
is whether there are three values or only two. Looking just at the left column, we
see there two values, and similarly in the right column. And yet, many linguists
would accept three values, based on the combinations a-d, a-e, and b-e; of these,
the combination a-e has no autonomous form. Let us check first with regard to
case values: Zaliznjak (1973: 69–74) discusses several instances. Here we will look
at Classical Armenian:
 Gender typology       93

SINGULAR PLURAL
am amk‘ NOMINATIVE
am ams ACCUSATIVE
ami ams LOCATIVE
ami amac‘ DATIVE

Figure 2: Classical Armenian am ‘year’ (partial paradigm)

In this example (from Klein 2007: 1053, see also Baerman, Brown and Corbett
(2005: 42–44) and references there) there is no unique form for the accusative; its
forms are always syncretic, with the nominative or with the locative, depending
on number. Yet we accept an accusative case value. The alternative, to avoid all
this, would be to say that transitive verbs take a nominative object in the singu-
lar and a locative object in the plural, but that would not allow simple rules of
syntax (Principle II). In terms of canonicity, we can say that the accusative is a
less canonical case value than the nominative or dative in this system.3
How does this relate to gender? A very clear comparable instance is the
gender system of Romanian. Key examples follow:

Romanian (Adina Dragomirescu and Alexandru Nicolae, personal communication)

(4) bărbat bun

man(m)[sg] good[m.sg]
‘a good man’

(5) film bun

film(n)[sg] good[m.sg]
‘a good film’

(6) femei-e bun-ă

woman(f)-sg good-f.sg
‘a good woman’

3 For nouns of this type the locative similarly does not have a unique form; however, elsewhere
in the system, there are nouns which do have a distinct locative. Daniel Kölligan points out (per-
sonal communication) that Classical Armenian has a preposition z-, used to distinguish objects
which are specific (for examples, see Jensen 1959: 146–150). This is an instance of differential
object marking but it does not make the accusative autonomous: it does not have a unique inflec-
tional form. For pronouns the preposition is obligatory.
94       Greville G. Corbett

(7) bărbaţ-i bun-i

man(m)-pl good-m.pl
‘good men’

(8) film-e bun-e

film(n)-pl good-f.pl
‘good films’

(9) femei bun-e

good woman(f)[pl] good-f.pl
‘good women’

If we had only the evidence from singular noun phrases we would conclude that
Romanian had two gender values. Similarly, if we had only plural noun phrases
we would propose two gender values. When we put the two together, we see the
need for three gender values, even though the third gender value has no unique
form to justify it. This third gender value is non-autonomous, just like the accusa-
tive of Classical Armenian. This can be seen in Figure 3:

SINGULAR PLURAL

bun bun-i MASCULINE

bun bun-e NEUTER (AMBIGENERIC)

bun-ă bun-e FEMININE

Figure 3: A non-autonomous gender value: Romanian bun ‘good’

Further discussion and sources can be found in (Corbett 1991: 150–154).4 This type
of gender value, what we are calling here a non-autonomous gender value, is also
known as genus alternans (as in Igartua 2006); once we recognize that gender
is not different in having such values, but is just like case (and as we shall see
shortly, like person too), we may prefer the term ‘non-autonomous’, since it is
common to the different morphosyntactic features. Given the nature of this third
gender value in Romanian, some prefer the term ‘ambigeneric’ to neuter.
It is important that from the point of view of the noun inventory Romanian
clearly has three gender values. There are substantial numbers of nouns in each

4 And more recently see Maiden (2011: endnote 36, pp. 701–702), and Nedelcu (2013) on Roma-
nian, and Loporcaro and Paciaroni (2011), for discussion of the development of such instances
elsewhere in Indo-European.
 Gender typology       95

of the three genders, and the neuter gender is gaining new members through
borrowings into the language. What is noteworthy about the Romanian gender
system is the means of agreement for one gender value: the distribution across
the lexicon is straightforward. Compare this with French, which arguably also has
nouns which are masculine in the singular and feminine in the plural: these are:
délice ‘delight’, orgue ‘organ’ and less clearly amour ‘love’. There is some ques-
tion about whether we are dealing with different lexical items, since they are not
straightforward singular-plural pairs and there is some variability. It seems clear
that these nouns should be treated as lexical exceptions and that the traditional
account which has two gender values for French is correct. But now consider the
Nakh language Batsbi, mentioned earlier. In various sources it is said that Batsbi
has eight genders. This can be found, for example, in the grammar by Dešeriev
(1953: 138–145), but the view goes back to the nineteenth century. Batsbi has just
four gender-number markers, and yet these occur in remarkable combinations
(Corbett 1991: 171). These combinations give eight agreement classes, all but one
of which are non-autonomous. The tradition has been to treat each of these agree-
ment classes as a gender value.
This traditional analysis may be represented as in Figure  4 (note that this
cannot be presented as in Figure  3, because of the ‘crossed’ relations between
some forms):

SINGULAR PLURAL

v
I

IV
b b

VII
VI
j III j
II
VIII
d V d

Figure 4: Agreement classes in Batsbi

96       Greville G. Corbett

However, the vast majority of nouns are found in just five gender values. The
remaining three agreement classes (all of which are non-autonomous) account
for around twenty nouns between them. None of these three classes should be
recognized as a gender value, since they have insufficient members: they are
“inquorate”. Each of the twenty or so nouns can be labeled as a lexical exception;
this is the analysis in Holisky and Gagua (1994: 162–163) and in Corbett (1991: 170–
172). This analysis can be represented as in Figure 5:

SINGULAR PLURAL

v I
b

III
j j

II

VI
b d

Figure 5: Gender in Batsbi (excluding inquorate genders)

Recall the earlier point about Batsbi and French. It is reasonable to say that Batsbi
has eight gender values (including the inquorate ones) provided we say also that
French has three gender values (also to include the inquorate gender for the very
few nouns which are masculine in the singular and feminine in the plural, noted
earlier). I believe it is preferable to treat French as having two gender values, in
which case Batsbi has five.
Having considered inquorate gender values, let us return to the main point,
that of non-autonomous values; we have seen a non-autonomous case value in
Classical Armenian and a non-autonomous gender value in Romanian. To dem-
onstrate the similarity across the features here let us look briefly at a non-auton-
omous person value, as found in Old Nubian (Figure  6); the source is Browne
(2002: 50), as cited in Baerman, Brown and Corbett (2005: 75):
 Gender typology       97

SINGULAR PLURAL
dollire dolliro 1
dollina dolliro 2
dollina dollirana 3

Figure 6: A non-autonomous person value: Old Nubian present indicative (doll- ‘wish’)

Old Nubian was a Nilo-Saharan language, with texts dating from the eighth to
the fifteenth centuries AD (according to Browne 2002: 1). The picture seen in
Figure 6 is fully comparable to that seen in Classical Armenian and in Romanian;
just the features have changed. Given only the singular, we would say that Old
Nubian makes a two-way distinction of morphosyntactic person, and similarly in
the plural. When we put the two together we recognize three person values, the
second person being non-autonomous. Thus non-autonomous gender values are
certainly interesting, but they are not unique to gender.

Criterion 2
Canonical features and their values are uniquely distinguished across other logi-
cally compatible features and their values.

This criterion makes intuitive sense. We know that in some instances, demon-
strating the existence of a particular feature or values means looking at a limited
set of environments, determined by combinations of other features and values.
In other languages, a particular feature and its values is evident almost wher-
ever one looks. The latter is the more canonical situation, since it is then easy
to argue that the particular feature and its values are required. This will be clear
from these Italian examples:

Italian (Marco Bertinetto, personal communication)

(10) il terren-o bass-o

def.m.sg ground(m)-sg low-m.sg
‘the low ground’

(11) i terren-i bass-i

def.m.pl ground(m)-pl low-m.pl
‘the low grounds’
98       Greville G. Corbett

(12) la port-a bass-a

def.f.sg door(f)-sg low-f.sg
‘the low door’

(13) le port-e bass-e

def.f.pl door(f)-pl low-f.pl
‘the low doors’

Let us ask whether we need a number feature. Just comparing (10) and (11) sug-
gests that we do, and that we need the values singular and plural. Equally, if we
compare (12) and (13) we reach the same result. Now consider gender. A com-
parison of (10) and (12) suggests that we need gender, and further work would
confirm masculine and feminine as appropriate values. Equally, comparing (11)
and (13) would give the same result. In other words, we find evidence for number
across the different gender values, and we find gender across the different
number values. We do not have to look, say, within the singular to find evidence
for gender. The situation is as in (14):

(14) Gender and number in Italian adjectival forms (both uniquely distin-
guished)
gender number
singular plural
masculine basso bassi
feminine bassa basse

Note that not all Italian adjectives are like basso ‘low’. Basso ‘low’ is canonical in
respect of Criterion 2. We return to other non-canonical Italian adjectives in our
discussion of Criterion 4 below.
Continuing with Criterion 2, according to which features and values are
uniquely distinguished across other logically compatible features and their
values, we now look for contrast as a clearly non-canonical instance. (15) shows
selected forms of the Archi verb aχas ‘lie down’ (data from Chumakina, Brown,
Quilliam and Corbett 2007: ix-xi and Marina Chumakina, p.c., following Kibrik
(1998: 457–458)):
 Gender typology       99

(15) Gender marking in Archi (aχas ‘lie down’)

number gender imperfective perfective
i w-a<r>χa-r a<w>χu8
ii d-a<r>χa-r a<r>χu
sg
iii b-a<r>χa-r a<b>χu
iv a<r>χa-r aχu
i
b-a<r>χa-r a<b>χu
ii
pl
iii
a<r>χa-r aχu
iv

We can extract the verbal gender/number markers, which may be prefixal (indi-
cated X-) or infixal (<X>), as in Figure 7:

GENDER (and assignment) NUMBER

SINGULAR PLURAL
I (male human) w-/‹w›
b-/‹b›
II (female human) d-/‹r›
III (some animates, all insects, some
b-/‹b›
inanimates)
Ø-/‹Ø›
IV (some animates, some inanimates,
Ø-/‹Ø›
abstracts)

Figure 7: Gender and number in Archi (evidence from verbs)

We should still recognize four gender values in Archi; however, it is clear that
the way in which they are realized in the morphology is less canonical than in
Italian. To see the four-way distinction, we must look at the singular, and even
here some markers are syncretic with other gender/number markers. Another
example is Lavukaleve (discussed in more detail in §4.2.2 below), which dis-
tinguishes three gender values in the singular and dual but not in the plural
(Terrill 2003: 142).
Again we find comparable non-canonical behaviour with other morphosyn-
tactic features: that is, instances where their values are distinguished fully only
in an environment defined by other morphosyntactic feature values. Examples

5 Realized as /uwχu /.
100       Greville G. Corbett

for number (from Biak, based on van den Heuvel 2006: 66), from person (from
Belhare, following Bickel and Nichols 2005: 51) and for case (from Russian) are
given in Corbett (2011: 462–464).

Criterion 3
Canonical features and their values are distinguished consistently across relevant
parts of speech (word classes).

In the canonical situation we find the same distinctions whichever part of speech
we look at. For instance, Russian nouns, adjectives, verbs and pronouns all show
two number values, singular and plural. From the perspective of systems like
the Russian number system, it may seem hard to imagine an alternative. And yet
there are various systems, gender systems and other morphosyntactic systems,
where different parts of speech behave differently.
Let us start with a problem which is relatively well known, concerning case.
In Guugu Yimidhirr (a language of North Queensland described in Haviland
1979), we find pronouns which, as part of a larger case system, distinguish sub-
jects and objects according to a nominative-accusative system. The forms are
given in (16):

(16) Guugu Yimidhirr pronouns (partial: Haviland 1979: 66–67)

nominative accusative
1 singular ngayu nganhi
2 singular nyundu nhina(an(in))
3 singular nyulu nhinhaan(in)

The nouns, however, are different, distinguishing ergative and absolutive:

(17) Guugu Yimidhirr nouns (partial: Haviland 1979: 47–51)

ergative absolutive
‘girl’ gabirr-inh gabiirr
‘head’ ngaabay-il ngaabaay

Given just the pronoun, we would imagine that Guugu Yimidhirr had a nomi-
native-accusative system. With just the evidence of nouns we would say that
it was ergative-absolutive. And indeed, such systems are sometimes treated
as ‘split ergative’. This makes sense if we are discussing the lexicon; it is true
that the lexicon is split between items with different morphological patternings
 Gender typology       101

of case. However, if we are concerned with syntax (especially the simple syntax
of our Principle II), it is unlikely that we would allow different case values for dif-
ferent controlees: we do not expect verbs to ‘peek’ at the type of a noun phrase
in order to establish which case they govern. Rather, we can consider pronouns
and nouns together: the data above give evidence for three distinct case values
(Guugu Yimidhirr has several additional simpler case values). This makes good
sense, and might be called a ‘combined case system’.

nom acc
pronoun 1sg ngayu nganhi
erg abs
noun ‘girl’ gabirr-inh gabiirr
combined case values erg nom-abs acc

Figure 8: Guugu Yimidhirr (as a combined case system): set-theoretical analysis

As we see in Figure 8, Guugu Yimidhiir has case values distinguishing the fol-
lowing:
1. the subject of a transitive verb stands in the ergative
2. the object of a transitive verb stands in the accusative
3. the single argument of an intransitive verb stands in the nominative-abso-
lutive; this name is only a helpful mnemonic: it is a single case value, which
we could have labeled simply absolutive.

This type of analysis is argued for in Goddard (1982). It is established by combin-

ing the evidence gained from pronouns with that gained from nouns.
Can we find similar situations with the other morphosyntactic features? Few
examples have been identified to date. For number there is Mele-Fila (data from
Ross Clark) see Corbett (2011: 467) and for person consider the Maybrat data (from
Dol 2007: 65) presented in Corbett (2011: 468). Of course, we are most interested
in instances of combined gender systems. One that has been discussed in some
detail is Mba (see Corbett 2011: 465–466 for sources and analysis), and a second
is Michif (Bakker 1997: 106–107 discussed in Corbett 2006: 269–270). Here we
examine an equally interesting example, which has been less discussed, namely
Burmeso, a language of the Mamberamo River area of Western New Guinea.
The material on Burmeso is entirely from Donohue (2001). Consider first these
instances of verb agreement (according to an ergative system):
102       Greville G. Corbett

Gender marking on verbs in Burmeso (Donohue 2001: 99–101):

(18) da nawak g-ihi-maru

1sg woman.sg ii.sg-see-today’s.past
‘I saw a woman.’

(19) da mibo j-ihi-maru

1sg banana.sg v.sg-see-today’s.past
‘I saw a banana.’

(20) jamo nawak n-akwa-ru

dog.sg woman.sg ii.sg-bite-today’s.past
‘The dog bit a woman.’

In these few examples (see Donohue 2001 for the full set) there is evidence that
Burmeso has a gender system, and also that gender agreement is marked dif-
ferently according to verb class. There are two inflection classes, as shown in
Figure 9:

assignment inflection class 1 inflection class 2

e.g. -ihi- ‘see’ e.g. -akwa- ‘bite’
sg pl sg pl
i male j- s- b- t-
ii female animate g- s- n- t-
iii miscellaneous g- j- n- b-
iv mass nouns j- j- b- b-
v banana, sago tree j- g- b- n-
vi arrows, coconuts g- g- n- n-

Figure 9: Full set of gender markers on Burmeso verbs (Donohue 2001: 100, 102)

These inflection classes are of great typological interest (see Corbett 2009), but
they are not our main concern here. The point which matters for us is that the
agreements of each inflection class show that there are six agreement classes of
nouns, labeled i-vi in (Figure 9). One of these, agreement class v can be consid-
ered inquorate, since it contains only two nouns, and the agreements are simply
an irregular combination: iv in the singular and vi in the plural. This system
 Gender typology       103

of gender agreement, is rather different from that shown on adjectives: (Donohue

2001: 105):

(21) Da de koya bek-abo

1sg 1sg.poss grandfather.sg good-m.sg
‘My grandfather is well.’

(22) Da d-asia bek-an.

1sg 1sg.poss-grandmother.sg good-f.sg
‘My grandmother is well.’

(23) Da de koysorad bek-odo

1sg 1sg.poss grandson.pl good-anim.pl
‘My grandsons are well.’

Again, I will not repeat all the examples to show the full system; the possible suf-
fixes are as follows:

(24) Gender agreement suffixes on adjectives (Donohue 2001: 106)

Gender singular plural
masculine -ab -od(o)
feminine -an -od(o)
neuter -ora -or(o)
masculine inanimate -ab -or
feminine inanimate -an -or
neuter animate -ora -od

It is evident that the distinctions made by Burmeso adjectives do not match

those of the nouns.6 However, they are not fully orthogonal either, as we see if
we plot the two against each other (Figure 10). I give the distinctions made on
the basis of verbs as i-vi on the left, and the adjectivally-induced distinctions
along the top:
In each cell I give the number of nouns found in Donohue’s representa-
tive word-list. This means that the figures are indicative, but not more.7 From

6 The adjective ‘white’ marks both gender systems.

7 For instance, the number of nouns in gender vi is actually higher; all terms for arrows belong
in gender vi (Donohue 2001: 102) but only the generic term kasarar appears in the representa-
tive word list.
104       Greville G. Corbett

Figure 10 we see clearly that it is not a matter of two systems combining freely.
Of the 36 possibilities, instances have been found of only 16. (It is possible, of
course, that with a larger noun inventory a few more cells might be filled, but the
distribution within Figure 10 shows considerable skewing.) Thus Burmeso shows
a combined gender system, comparable to the combined case system of Guugu
Yimidhirr; to see the full system we have to look at the differing evidence of verbs
and adjectives. Both systems are far from canonical: in a canonical system, the
evidence gained from each controllee or target would be the same.

m f n m inan f inan n anim

I 44 plus all 5 (4 birds) 1 (‘neck’) 2 (‘sea’,
male kin ‘wound’)
terms
II 7 plus all 4 1 (‘small 2 (‘sago
female kin goanna’) rinser
terms (lower)’,
‘string.
shapes’)
III 3 28, mainly 10, inani- 1 (‘goanna’)
inanimate mate
IV 9, inanimate
V 2 (‘banana’,
‘sago tree’)
VI 1 (‘arrow’) 1 (‘coconut’)

Figure 10: The combined gender system of Burmeso

It is not just larger systems that can be non-canonical in this way. Consider
the apparently simpler system of Dutch. Different targets distinguish different
numbers of gender values, as shown in Figure 11.
Dutch is non-canonical in that the targets are sensitive to different numbers
of gender values; the system appears relatively straightforward in that the smaller
system of two values is nested within the larger. There are additional very inter-
esting complications, however, in that several different combinations of gender
agreements are possible, for which see Audring (2009). See also De Vos and De
Vogelaar (2011) on the special interest of Dutch, and Schiller (this volume) for
some unexpected psycholinguistic results from its gender system.
 Gender typology       105

agreement target attributive relative personal

(type) pronoun pronoun
agreement target indefinite definite adjective relative personal
(category) article article pronoun pronoun
de [c] mooi-e [c] die [c] hij [m]
example forms een het [n] mooi [n] dat [n] zij [f]
het [n]
number of
0 2 2 2 3
gender values

Figure 11: Gender values of different agreement targets in Dutch (Audring 2009: 52)

Since we are discussing situations where different targets behave differently,

we should pause briefly to consider classifiers. In the clearest instances clas-
sifiers are rather different from gender systems, since they are free forms; see
Dixon (1982: 212–218) and Corbett (1991: 136–137) for ways of distinguishing
gender from classifiers. And the best known classifier systems and gender
systems are found in languages of different types (isolating and fusional
respectively), which was taken to suggest rather different functions. Since
those accounts the typology has become much more interesting. Classifiers
are of various types, and interesting examples have been found of each. For
comparison with gender, the most relevant are noun classifiers (or generics).
There have been two particularly significant steps forward. First, more and
more instances have been found of languages with both gender and clas-
sifiers. These include Tariana (North Arawakan; Brazil; Aikhenvald 1994,
2000), which has a gender system and three sub-types of classifier, Retuarã
(Tucanoan; Columbia; Strom 1992: 10–11, 34–36, 45–47), Tidore (West Papuan;
Halmahera, Indonesia; van Staden 2000: 77–81) and Mian (Trans New Guinea;
Papua New Guinea, Fedden 2011: 185–201). Ngan’gityemerri (Daly; northern
Australia) shows the development from generic classifiers into genders (Reid
1997).8 Such a development points to the second step forward, namely that
systems have been found which fall between the earlier ideas of gender and
classifier. A good example is Miraña (North West Amazon), as analysed by
Seifart (2005). Consider these examples:

8 See Sansò (2009) for discussion of an embryonic classifier construction in Ancient Greek,
whose development was cut short.
106       Greville G. Corbett

Miraña (Seifart 2005: 78–79)

(25) í-:baj ɯ́βi-:baj

prx-scm_cont basket-scm_cont
‘this one (container), a basket’

(26) ɯ́βi-:baj o:kε ɯ áhkɯ-:baj

basket-scm_cont 1sg.acc 2sg.sub give-scm_cont
‘the basket that you gave to me’

(27) kátɯ́:βε-:baj ɯ́βi-:baj

fall-scm_cont basket-scm_cont
‘The basket fell down.’

Note: Cont: container, PRX: proximal demonstrative, SCM: specific class marker;

Miraña has 66 specific class markers, mainly concerned with shape, which makes
one think of a classifier system; in addition, it has six general markers, based on
animacy, natural gender and number (Seifart 2005: 77); in terms of assignment,
then, this part of the system is like a gender system. Examples (25)-(27) are par-
ticularly interesting in terms of the targets involved: the markers are found on a
range of targets: on demonstrative pronouns (25), on the predicate of a relative
clause (26) and that of a main clause (27). (They also occur on the controller noun,
as all three examples show.) Their occurrence on different controllers makes the
system like a gender system. As Seifart (2005) shows, the canonical approach
is particularly fruitful here, since it allows to separate out the different parts of
the system, as more or less like a canonical gender system, rather than trying to
determine the typology of this complex system as a whole.
The final criterion we consider concerns lexemes. When we say that a lan-
guage has a gender system, we should consider what this means for individual
lexemes. In a canonical system, the following holds:

Criterion 4
Canonical features and their values are distinguished consistently across lexemes
within relevant parts of speech.

In the canonical system, then, if a particular part of speech – say the verb – marks
gender, then every verb does, and it can mark all the values. This implies two
ways to be non-canonical, in terms of (a) the feature or (b) a value of a feature. We
discuss each in turn.
 Gender typology       107

In Archi, almost every part of speech can agree in gender. I will illustrate
some of the less familiar targets:

Archi: agreeing adverb: Kibrik (1977: 186); Chumakina and Corbett (2008: 187) where original
bala is amended to balah

(28) o‹b›qˤa-tːu-b balah ditːa‹b›u

‹iii.sg›leave.pfv-atr-iii.sg trouble(iii)[sg.abs] soon‹iii.sg›

b-erχin
iii.sg-forget.ipfv
‘Past trouble gets forgotten quickly.’

Archi: agreeing postposition (Kibrik, Kodzasov, Olovjannikova and. Samedov 1977: 227)

(29) to-w-mi-s sin-t’u ɬːʷak-du-t

that.one-i.sg-obl.sg-dat know-neg near-atr-iv.sg

duχriqˤa-k e‹b›q’en
village(iv).sg.inter-lat ‹iii.sg›up.to

b-i-tːu-b deq’ˤ
iii.sg-be.prs-atr-iii.sg road(iii)[sg.abs]
‘He does not know the way to the next village.’ (how far it is)

Archi: agreeing emphatic particle (Kibrik 1977: 326)

(30) arša horoːk=ej<b>u iškul

Archi.in long.ago=emph.ptcl<iii.sg> school(iii)[sg.abs]

dabɬu
open.pfv
‘A school was opened in Archi already long ago.’

Even the personal pronoun, in some of its forms, agrees with absolutive argu-
ment:

Archi: agreeing pronoun (Chumakina and Corbett 2008: 188)

(31) d-ez un malgan

ii.sg-1sg.dat 2sg.abs be.dear
‘You (female) are dear to me.’ (uttered by a male)
108       Greville G. Corbett

Thus all these parts of speech, together with verbs (the agreement forms for
which we saw earlier in (15)) and adjectives (too familiar cross-linguistically to
deserve illustration) can in principle agree in gender. However, that is true at the
level of the part of speech. Individual lexemes may or may not agree; this appears
to be partly predictable, and partly lexically specified. Figures for the parts of
speech which are easier to count are given in Table 1; I omit the pronouns, because
it is not evident how to count the fact that some cells in the paradigm agree while
most do not.9 The data are derived from the Archi dictionary (Chumakina, Brown,
Quilliam and Corbett 2007), reported in Chumakina and Corbett (2008: 188):

Table 1: Non-canonical at the feature level: agreement targets in Archi

total agreeing % agreeing

adjectives 446 313 70.2

verbs 1248 399 32.0
adverbs 397 28 7.1
enclitic particles 4 1 (25.0)

Thus while these different parts of speech can agree in Archi, the system is less
canonical in respect of lexemes: many do not show gender (or number) agree-
ment. Earlier we noted the highly canonical behavior (in several respects) of
Italian adjectives. However, they too are less canonical when we look at the level
of lexemes. There is the type we saw in (10)-(13), which is canonical in respect of
Criterion 2. That is the main pattern, but there are others too, which give a picture
like Macedonian for Criterion 4. Thornton, Iacobini and Burani (1997: 74) give the
following statistics for a total of 1129 adjectives:

Table 2: Types of Italian adjectives (Thornton, Iacobini and Burani 1997: 74)

Four distinct forms, basso/ bassa/ bassi/ basse ‘low’ as in (10)–(13) 65.3%
Two forms, singular versus plural: verde/verdi ‘green’ 31.7%
One form, invariable: blu ‘blue’ 1.9%
Others 1.1%

9 In addition, there are 34 postpositions, of which one, namely eq’en ‘up to’, which shows infixal
agreement (as in (29)). However, its part of speech status is not fully clear, and so I have not in-
cluded postpositions in the table.
 Gender typology       109

We see that many Italian adjectives do not agree in gender (while agreeing in
number in most instances), which shows again that items which are fully canoni-
cal in one respect can be non-canonical in respect of others.
Let us turn to the other type of non-canonical behaviour, that which concerns
values. Consider these Latin adjectives:

(32) Latin adjectives of inflection class 3 (nominative singular forms)

masculine feminine neuter gloss
acer acris acre sharp
facilis facilis facile easy
vigil vigil vigil alert

Latin has three gender values, as shown by many adjectives (though not in all of
the case values). Other adjectives, those traditionally grouped in the third inflec-
tion class because of similarities elsewhere in the paradigm, show various possi-
bilities. Acer ‘sharp’, and others like it distinguish three gender values; adjectives
like facilis ‘easy’, on the other hand, show no distinction between masculine
and feminine. Vigil ‘alert’ and similar adjectives show no evidence of gender
agreement in the ­nominative singular (though certain other forms distinguish
neuter from the other gender values). Thus we have some adjectives which are
non-canonical in that they do agree in gender, but do not differentiate all the
available gender values.
Let us sum up the types of non-canonicity we have surveyed. Figure  12
includes for completeness other instances which were not illustrated above, since
we concentrate on gender. For full data see Corbett (2011).

Criterion 1: Criterion 2: Criterion 3: Criterion 4:

autonomous distinguished distinguished distinguished
across features/ consistently consistently
values across PoS across lexemes
number Biak Mele Fila Bezhta
gender Romanian Archi Burmeso Archi
person Old Nubian Belhare Maybrat English
case Classical Armenian Russian Guugu Yimidhirr Russian

Figure 12: Non-canonical feature values and the four criteria

110       Greville G. Corbett

The picture in Figure 12 shows that typologists have identified an interesting set
of phenomena. In terms of gender, we see that there are various ways in which
we may say that a language has a gender system, but where accuracy requires
us to specify this further. At another level of abstraction, gender shows the same
characteristics as the other morphosyntactic features. Each of them can have a
‘penumbra’ of unclear behaviour around the more canonical core.

4 The assignment problem

For any feature, we have to give the rules for the use of each of its values, the
assignment system. This is particularly interesting for gender, because typi-
cally values are assigned to nouns as a lexical matter. Thornton (2009: 14–15)
points out that assignment, particularly gender assignment, can be thought
of in two ways: nouns are assigned to gender values, conceptualizing gender
as containers; or gender values are assigned to nouns, where a feature value
is part of the specification needed for the noun to function properly in syntax.
Both views are important: the first deals with cognitive classification, the
second is more to do with the function of the feature in the grammar. It is the
second perspective which concerns us here: given a noun, which gender value
can be assigned to it.
When we analyse assignment systems of languages from different families
we find that genders always have a semantic core. This may involve biological
sex, or animacy, with other features also having a role (see McConnell-Ginet,
this volume, for discussion of the sociocultural aspect of this relation). In some
languages genders are assigned to nouns solely on the basis of semantics, but
in others this semantic information is supplemented by formal information, and
this may be phonological or morphological. Given this typology, we can see that
gender is distributed in interesting ways across the world’s languages.

4.1 S
 emantic assignment systems

Consider these examples from the Daghestanian language Bagwalal (from Kibrik
2001: 64–66):

(33) w-eš:a-w waša

m.sg-plump-m.sg boy
‘a plump boy’
 Gender typology       111

(34) j-eš:a-j jaš

f.sg-plump-f.sg girl
‘a plump girl’

(35) b-eš:a-b ʕama

n.sg-plump-n.sg donkey
‘a plump donkey’

Evidence from the attributive modifier gives us three agreement classes, each of
which should be recognized as a gender. If we look at the verb, we find evidence
for the same three classes. Bagwalal is a particularly clear instance of a consistent
type of assignment system; the examples given are representative of the whole
system. Nouns denoting male humans (and only those) are masculine, those
denoting female humans are feminine. All remaining nouns are neuter. Thus
waša ‘boy’ is masculine, jaš ‘girl’ is feminine, and ʕama ‘donkey’ is neuter (its
sex is not relevant because it is not human). The meaning is sufficient, and no
information about the form of a noun is needed to determine its gender. This is
what we term a ‘semantic assignment system’.
Bagwalal has a strict semantic system, since no other information is needed,
and such systems are found in various parts of the world. Kala Lagaw Ya, spoken
on the Western Torres Straits Islands, has a two-gender system, also with strict
semantic assignment: nouns denoting males (and the moon) are masculine and
all others belong in the feminine gender (Bani 1987). In Diyari, a language of South
Australia, we find the converse: there is a gender for nouns with female referents
(such as women, girls, doe kangaroos), and the other is for all remaining nouns
(Austin 1981). Strict semantic systems are particularly prevalent in the Dravidian
family, where there are three-gender systems (as in Kannada and Tamil) and two-
gender systems (as in Parji). The semantic distinction at the heart of the assign-
ment system of the languages discussed so far is biological sex (applying just to
humans or to animals too), and this is true both for strict semantic systems and
for those where other factors also have a role. Sex is the majority choice – three
quarters are of this type.10 There is another possibility, namely human vs non-
human or animate vs non-animate. These systems are found in the Niger Congo
and Algonquian families, but also in Khoisan and Austro-Asiatic, for instance.

10 There is variation as to what counts as being sex-differentiable: in Russian sex-differentiabil-

ity operates where it matters (as with humans, domesticated animals and those where there is
a striking difference), but shark is feminine and dolphin is masculine. In Tidore (Papuan, north
Moluccas), on the other hand, anything non-human is grammatically neuter.
112       Greville G. Corbett

Besides these strictly semantic systems, we find others that we call ‘pre-
dominantly semantic’ assignment systems. These are again clearly based on
semantics, but the rules are more complex and typically do not cover the full
noun inventory so successfully. A complex and interesting system is that of Bininj
Gun-wok (previously Mayali), a non-Pama-Nyungan language of northern Aus-
tralia as described in detail by Evans, Brown and Corbett (2002).11 To give a less
well known example, we turn to Mian. Mian is a member of the Ok family (part
of Trans-New-Guinea), spoken in Sandaun Province Papua New Guinea. Fedden
(2011) describes the eastern dialect, which has around 1400 speakers. Fedden
(2011: 169–184) presents evidence for four genders:

(36) The four gender values of Mian (Fedden 2011: 170)

gender example gloss
masculine naka ‘man’
feminine unáng ‘woman’
neuter 1 imen ‘taro’
neuter 2 am ‘house’

When we examine the assignments for the first two genders, the picture is
straightforward, as we see in Figure 13:

criteria gender
masculine
human
e.g. naka ‘man’
biological sex
animal (sex obvious or feminine
relevant) e.g. unáng ‘woman’
animate
masculine
animal (sex not obvious e.g. tolim ‘eagle’
conventionalized gender
or irrelevant) feminine
e.g. koból ‘cassowary’

Figure 13: The masculine and feminine genders in Mian (Fedden 2011: 172)

11 Bininj Gun-wok also shows a complex interaction of assignment to gender and to morphologi-
cal class: it has four genders and five morphological classes for nouns. This has been success-
fully implemented; for details of a formal model of this system see Evans, Brown and Corbett
(2002). Modelling has tended to be concentrated on synchronic assignment systems. However,
there is also interesting work on modeling the development of assignment systems over time, as
in Polinsky and van Everbroeck (2003).
 Gender typology       113

These follow biological sex in the main. Compare this with the remaining
two genders, which Fedden calls Neuter 1 and Neuter 2, as presented in Figure 14:

criteria gender
count nouns (e.g. mén ‘string bag’, imen ‘taro’)
liquids, body fluids/wastes, substances (e.g. aai ‘water’; ilem neuter 1
‘blood’, as ‘wood’)
places (e.g. am ‘house’, mon ‘old garden’, dafáb ‘summit’)
masses (e.g. afobèing ‘goods, property’, monî ‘money’)
body decoration (e.g. eit ‘decoration’, baasi ‘pig’s tusk’)
inanimate
weather phenomena (e.g. sók ‘rain’, ayung ‘mist’)
illnesses (e.g. kweim ‘fever’) neuter 2

intangibles/abstracts (e.g. āns ‘song’; kukub ‘way, custom’)

verbal nouns (e.g. fumin ‘activity of cooking’
some tools and weapons (e.g. káawa ‘steel axe’, mōk ‘stone adze’,
skemdâng ‘knife’)

Figure 14: The Neuter 1 and Neuter 2 genders in Mian (Fedden 2011: 172)

Compared with the first two genders, the two neuter genders are harder: count
nouns tend to be in the first neuter gender, and there are other semantic indica-
tors, but this is all less clear than in Bagwalal. Such systems are termed ‘predomi-
nantly semantic assignment systems’. While this is fine as a characterization of
the system as a whole, Mian suggests that the assignment to different gender
values can be rather different within the same system. Indeed, the type of split
we find in Mian is similar to that found in several languages of Daghestan.
A famous example of this type was Dyirbal, with four genders apparently
requiring complex semantic assignment rules. This system has been reanalyzed
by Plaster and Polinsky (2010) and we return to it shortly (in §4.2.1).

4.2 F ormal assignment systems

In many languages, the semantic rules assign the appropriate gender value to
many nouns but they also simply fail to apply for others. In Russian, and for
many other Indo-European languages, there are semantic assignment rules
for sex-differentiables (those where sex is salient or of importance to humans);
114       Greville G. Corbett

these are that nouns denoting males are masculine and those denoting female
are feminine. But these rules have nothing to say about the majority of the noun
inventory. In languages like Bagwalal, the nouns not assigned a gender value by
the semantic rules (the ‘remainder’ or ‘semantic residue’), all belong to a single
gender value. However, in languages like Russian they are found in more than
one gender, even in all the genders (as is the case in Russian). Here we find addi-
tional rules for assigning gender values to nouns, but according to their form.
Languages may use semantic assignment rules, or semantic and formal rules, but
not just formal assignment rules. In no language are genders assigned to nouns
by purely formal rules. An example would be a language in which there were two
agreement classes, and the nouns in the first all ended in a consonant cluster,
and those in the second did not, and there was no semantic regularity for the
distribution of nouns. I claim that this hypothetical type does not exist. Formal
assignment rules, which always operate alongside semantic assignment rules,
may appeal to two types of information: phonological and morphological.

4.2.1 Phonological

The clearest example of assignment depending on phonological information yet

found is provided by Qafar (Afar), an East Cushitic language spoken in north-
eastern Ethiopia and in Djibouti (data from Parker and Hayward 1985). I repeat
the evidence briefly here, since it is such a dramatic example, and then move on
to a more recently published analysis of a comparable system.
Qafar has semantic assignment rules which are unsurprising: for sex-differ-
entiable nouns, those denoting males are masculine and those denoting females
are feminine. We are interested in the nouns which these rules do not cover, the
semantic residue. These nouns are covered by phonological assignment rules:
nouns whose citation form ends in an accented vowel, like catò ‘help’, are femi-
nine; all others are masculine. Thus gilàl ‘winter’ does not end in a vowel and so
is masculine, as is tàmu ‘taste’, which ends in a vowel, but an unaccented one.
These rules, semantic and phonological, operate with few exceptions. The inter-
action in Qafar is particularly significant. Nouns denoting males and females typ-
ically fit with the phonological assignment rules too: bàqla ‘husband’ would be
masculine by either set of rules and barrà ‘woman, wife’ would be assigned femi-
nine gender by either. We might wonder whether we could do without semantic
assignment rules for Qafar, though typologically that would mean Qafar was quite
different from the languages we have discussed so far. However, there are key
examples where the two types of rule are in conflict. First consider abbà ‘father’:
the semantic assignment rules would predict masculine; given that it ends in an
 Gender typology       115

accented vowel, the phonological rules would predict feminine. It is masculine.

Similarly, gabbixeèra ‘slender-waisted female’, where the semantic prediction is
feminine, and the phonological predicition is masculine (the accent is non-final),
is feminine. When the semantic and phonological rules are in conflict, which is
relatively rare in Qafar, it is the semantic rules which take precedence. This is the
normal situation in gender assignment systems.
There are many more examples of phonological assignment, including for
instance Gujarati (Modi 2003), though few if any as clear as those of Qafar. At
the other extreme we find French, often claimed to have no system to its gender
assignment. Yet French has a phonological assignment system. For example,
of 938 nouns ending in /ɛ̃/, 99% are masculine, like le pain [pɛ̃] ‘the bread’ (for
details see Tucker, Lambert and Rigault 1977 and Hardison 1992).
As a less familiar example of a phonological assignment system, let us return
to Dyirbal (from Queensland, Australia). This is a famous gender system, first ana-
lysed by Dixon in his grammar of the language (1972: 44–47, 60–62, 306–312), and
taken up by many writers since. There are four genders, with these basic rules of
semantic assignment:

(37) Basic rules of semantic assignment in Dyirbal (from Dixon 1972: 306–312)

gender I: male humans, non-human animates
gender II: female humans, water, fire, fighting
gender III: non-flesh food
gender IV: residue

While these rules cover a good proportion of the nouns, many would be incorrectly
assigned. In addition, Dixon proposes principles of association. For example, in
mythology, the moon is the husband of the sun. The sun is in gender ii because
of the association with fire and light. Why then would the hairy mary grub be in
gender ii? According to this account, because the grub has a sting which, it is
said, feels like sunburn, so by association with the sun it is in gender ii (Dixon
1972: 310). The general idea was treated at length by Lakoff (1987), as an instance
of radial categories.
This famous analysis has been questioned on the basis of how the language
could be acquired. Plaster and Polinsky (2010) suggest that the age at which
children acquire gender is too early for them to have access to radial categories.
By looking more closely at the purely linguistic data, and taking account of the
now established typology of gender assignment, they propose an account using
just semantic and formal rules of assignment. Their assignment rules are as in
Figure 15.
116       Greville G. Corbett

Semantic assignment
1. nouns denoting males → gender I
2. females, fresh water, fire, stinging → gender II
3. edible → gender III
4. remaining inanimates → gender IV

Formal assignment (remaining animates)

1. nouns in bi-, gugu-, ma-, yi-, -gan → gender II
2. remainder → gender I

Figure 15: Gender assignment in Dyirbal (Plaster and Polinsky 2010: 135)

Plaster and Polinsky’s analysis accounts for 573 of the 597 documented nouns.
It does so without any need for radial categories, by appealing to semantic and
formal assignment rules. It is motivated in diachronic terms, since it suggests that
the system arose as a result of the collapsing of a classifier system. And it has
typological plausibility, since Dyirbal now fits into a well-attested type, requir-
ing just semantic and formal assignment rules, such as we see in many gender
systems; the elaborate radial categories are no longer required.12

4.2.2 Morphological

The morphological assignment system which has received the most attention
is probably that of Russian. Again, since it is such a fine example of its type, I
repeat the evidence briefly, and then go on to more recently discovered examples
of this general type. In Russian, as in many other Indo-European languages, for
sex-differentiable nouns, those denoting males are masculine and those denot-
ing females are feminine. But unlike the situation in languages like Bagwalal,
the nouns not covered by these rules  – the semantic residue  – are not simply
assigned to the neuter gender. Rather in Russian the residue is shared between
the three genders, with the neuter gender not even receiving the majority. This is
represented in Figure 16:

12 For an account of the gender system of the Daghestanian language Tsez, which also uses a
combination of semantic and formal factors, see Plaster, Polinsky and Harizanov (in press).
 Gender typology       117

masculine feminine neuter

Sex differentiables denot- Sex differentiables denot- Part of the semantic residue
ing males PLUS part of the ing females PLUS part of the
semantic residue semantic residue

Figure 16: Gender assignment in Russian

It seems unlikely that we are failing to spot additional semantic criteria; in the
past I have cited various ‘triplets’ of nouns with similar semantics but belong-
ing to different genders (for instance in Corbett 2007b). Here are some further
examples, to demonstrate that they are indeed numerous (Table 3):

Table 3: Russian nouns belonging to the semantic residue

masculine feminine neuter

xor ‘chorus’ pesnja ‘song’; solo ‘solo’

klarnet ‘clarinet’ skripka ‘violin’ fortep’jano ‘piano’
kostjum ‘suit’ jupka ‘skirt’ plat’e ‘dress’
nomer ‘hotel room’ gostinica ‘hotel’ foje ‘foyer’
braslet ‘bracelet’ cepočka ‘chain’ kol’co ‘ring’
jačmen’ ‘barley’ rož’ ‘rye’ pšeno ‘millet’
sir ‘cheese’ smetana ‘sour cream’ maslo ‘butter’
žurnal ‘magazine’ gazeta ‘newspaper’ pis’mo ‘letter’
lokot’ ‘elbow’ kost’ ‘bone’ koleno ‘knee’

If instead we consider the inflectional morphology of these nouns it becomes pos-

sible to find the assignment rules. Russian has four main inflection classes of
noun, each with thousands of members. The singular forms are given in Table 4
(the nouns given for illustration were included in the triplets in table 3 above):

Table 4: Inflection classes in Russian

i ii iii iv

nominative žurnal gazeta kost’ pis’mo

accusative žurnal gazetu kost’ pis’mo
genitive žurnala gazety kosti pis’ma
dative žurnalu gazete kosti pis’mu
instrumental žurnalom gazetoj kost’ju pis’mom
locative žurnale gazete kosti pis’me
gloss ‘magazine’ ‘newspaper’ ‘bone’ ‘letter’
118       Greville G. Corbett

It is important to stress that gender is not the same as inflection class.

Gender is accessible to the syntax (for agreement), while inflection class is
purely morphological. The two sometimes get confused in accounts of Indo-
European languages where there are relatively few inflection classes; in lan-
guages like Kuot, discussed below, the larger number of inflection classes con-
trasts more strongly with the number of genders and there is less likelihood of
confused analyses.
It is evident that the speaker must know the inflectional behaviour of a
noun (its inflection class), and given that information, the assignment rules are
straightforward. Gender can be assigned on the basic of inflection class, but not
vice versa. Nouns in inflection class I are masculine, those in classes II and III
are feminine, and those in IV are neuter. Since otec ‘father’ is in class I, while
sestra ‘sister’ is in class II, and mat’ ‘mother’ is in class III, we can see that many
sex-differentiable nouns would be assigned their gender value correctly by the
morphological assignment rules. And so, as we did with Qafar, we should ask
whether we could do without the semantic assignment rules. This would not
work because of nouns like deduška ‘grandfather’; this noun denotes a male
but is in class II, whose nouns are typically feminine; it is masculine. Nouns
like this show, once again, that we do not find languages where formal assign-
ment rules are sufficient. Further rules are required in Russian for indeclinable
nouns (like taksi ‘taxi’, which is indeclinable and neuter); for fuller details see
Corbett (1991: 34–43). The outline of gender assignment in Russian is given in
Figure 17.13
When the system of inflection classes is not reliably available, as is the case
with heritage speakers of Russian, the effect on the assignment system is dra-
matic, as Polinsky (2008) demonstrates convincingly.14 More generally, the way
in which semantic assignment rules take precedence is helpfully clarified in
Thornton (2009); see also Enger (2009).

13 The interesting question of how children acquire conflicting assignment rules is addressed
in an experimental study (Rodina 2007). Rodina investigated 25 Russian children 2;6–4;0, and
included Russian nouns like papa ‘daddy’ where semantic and formal assignment rules conflict.
She suggests that children first apply the formal rule, and gradually add the semantic rule. Thus,
according to this view, the semantic core to the gender system, which the typological evidence
shows is always key to assignment, is not acquired first. For recent work on the acquisition of the
Bantu language Sesotho, again showing the vital role of phonology, see Demuth and Wechsler
(2012).
14 And for a recent careful study of the decline of gender in Cappadocian Greek see Karatsareas
(2009).
 Gender typology       119

Semantic assignment (as before)

1. sex differentiable nouns denoting males → M
2. sex-differentiable nouns denoting females → F

Formal (morphological) assignment

1. nouns of inflection class I → M
2. nouns of inflection classes II & III → F
3. remainder → N

Figure 17: Gender assignment in Russian

So much is well established. Later work has helped to confirm and clarify this
position.15 Take, for instance, Lindström (2002: 147–164, 176–194) on Kuot, a lan-
guage isolate of New Ireland. In some respects Kuot is even more convincing than
Russian as an instance of morphological assignment, since it has few gender
values and more inflection classes. Not surprisingly, however, the system is less
well studied than that of Russian, and there are more uncertain instances.
Kuot has two gender values. Semantic assignment is as expected: sex differ-
entiables (humans and major animals) are assigned gender according to sex. Of
the semantic residue, a substantial number of nouns have their gender assigned
according to morphological class. Lindström gives useful statistics, based on a
list of 869 nouns. The inflection classes rest on number formation: nouns have
singular and plural, and then there is a dual, used mainly for animates, which is
for most nouns formed from the plural. Lindström gives eleven inflection classes,
determined by the final syllable of the singular, and its relation to the plural form.
For instance, nouns in the “ma-declension” form the plural using the normal
plural marker -p, dropping the -ma, for instance:

(38) Examples of the ma declension in Kuot (Lindström 2002: 153)

singular plural gloss
ima ip subclan; river
laukima laukip tooth
pipiduluma pipidulup bird (species)
adaima adaip dance mask

15 Morphological assignment systems are found widely in Indo-European. In Arabic too, gender
is assignable in the main according to morphology (Cowell 1964, 372–375); we now look at less
familiar data.
120       Greville G. Corbett

Lindström lists 142 nouns in this inflection class; they are all masculine except
bunima ‘last born’ which is covered by the semantic assignment rules (it takes its
gender according to the sex of the referent), and arəma ‘(species of) pandanus
fruit’, which is exceptionally feminine (plus eight for which the gender value is
not known). Statistics on the relations between inflection class and gender are
given in Table 5:

Table 5: Inflection class and gender in Kuot (Lindström 2002: 149)

f m f/m ? Total

plain 237 172 39 28 476

ma 1 132 1 8 142
na 3 29 2 2 36
bun 17 0 0 0 17
bu 8 1 2 0 11
uom 8 0 0 0 8
bam 21 0 0 3 24
nəm 35 0 0 4 39
nim 12 0 0 2 14
n 40 25 10 6 81
m 16 2 2 1 21
Total 398 361 56 54 869

In Table 5, ‘f/m’ indicates that either gender value is possible, or that Lindström
has conflicting information, while ‘?’ indicates that the gender value is not
known.
If we look first at the lower part of the Table  5, all but the ‘plain’ row, the
picture is fairly clear. The inflection class of a noun gives a strong prediction as
to its gender. The predictions are stronger than might first appear, because the
nouns covered by the semantic assignment rules have not been excluded: several
apparent exceptions to the morphological assignment rules are covered by the
semantic rules. The picture is more straightforward than that seen in Russian
because we have many inflection classes predicting just two gender values: there
is no question of the prediction being in the other direction.
On the other hand, the picture is less clear in another respect, as we see when
we consider the ‘plain’ inflection class. There are nouns which do not end in any
of the sequences listed for the ten other classes in the table; in the main, these
form the plural according to regular rules: the ending is -(i)p, with labialization
assimilation of /i/ to /u/ under certain conditions. It is in this plain class that
there is the least predictability of gender value. Lindström (2002: 183–190) consid-
ers various possible semantic groupings here, some of which are weakly predic-
 Gender typology       121

tive. For instance, most nouns denoting liquids are feminine, as are most abstract
nouns, including nominalizations.
To sum up, Kuot is a clear instance of a morphological assignment system.
In one respect it is more straightforward than Russian, in that the predictions go
from eleven inflection classes to two gender values; in another respect it is less
clear, since one substantial inflection class does not lead to a clear prediction of
gender value.
Another significant and interesting assignment system is found in Lavuka-
leve, a Papuan language of the Central Solomons language described by Terrill
(2003: 131–147). Lavukaleve has three gender values, with roughly equal numbers
of nouns in each. There are semantic assignment rules for each gender, not only
based on sex, but also, for instance, all mammals and turtles are masculine, while
abstracts are neuter. However, apart from the main rules based on sex, seman-
tic principles are more ‘weak tendencies’ (2003: 137). ‘Formal assignment prin-
ciples are more regular and transparent than semantic assignment principles.’
(2003: 137). These formal principles are mainly phonological, but also appeal
to derivational morphology. For comparison see also Terrill (2002) on gender
systems in Papuan languages of this area more generally.16
It is worth spending a moment on the assignment of gender to loanwords.
The simplest hypothesis is that they would be treated exactly as other nouns.
Often this is the case, and such instances confirm the validity of the assignment
rules. However, there can be respects in which loanwords are different from native
words, and here there are interesting issues. For careful discussion see Thornton
(2009), analyzing Italian data and Stolz (2009) working on Maltese. For the ways
in which gender assignment can change, a recent reference is Visser (2011), who
describes changes in West Frisian;17 changes over a long time stretch are docu-
mented for the Iroquoian family by Mithun (this volume).

4.2.3 Recategorization

We noted that assignment is something which partly sets gender apart. At least
in the more familiar languages, nouns have a single gender value; similarly per-
sonal pronouns have a single person value. On the other hand, nouns have access
to all (or many of) the case and number values. However, there are also instances

16 In addition to studies of gender systems in broad outline, others concentrate on the detail of
gender use: see for instance Stein (forthcoming) and references there for gender and its signifi-
cance in Biblical Hebrew texts.
17 See also Rovai (2012) for Latin.
122       Greville G. Corbett

of less rigid assignment or – to think of it in the context of other features – recat-
egorization. While in canonical gender systems, each noun is allotted to a single
gender value, there are some very interesting systems, often described only quite
recently, where there is more going on. Once again, there are intriguing parallels
with other morphosyntactic features. Let us begin with number. There are famil-
iar examples like two coffees please where coffees has a unit reading and is said
to show recategorization. This is easier to see with the plural, as here. But the
effect is more subtle. The non-count noun coffee is recategorized as a count noun,
which means that instead of lying below the threshold of number differentiabil-
ity, and so having only one number value, it is ‘moved up’ as it were to have a full
set of number values (singular and plural for English). Thus the recategorization
is to being a count noun, so that alongside the obvious plural there is also a unit
reading for the singular, as in a coffee please.
The most obvious analogy for gender systems is the threshold of sex-dif-
ferentiability. Languages may allow items below the threshold to be reclassi-
fied as being above it, and thus to have additional gender possibilities. We are
used to animals and even inanimate objects being personified in stories, and
then referred to with he and she. However, languages differ dramatically in how
ready they are to allow such recategorizations, and there are interesting grada-
tions. If we consider first the limits, we have languages like English where such
recategorization is easily accepted: if the rhyme in which the dish ran away
with the spoon were extended into their later lives, we can imagine one being
referred to with he and the other with she. At the other end of the scale, Comrie
(2005) reports on the Daghestanian language Tsez; in a story in which a rooster
has an affair with a frog, no recategorization is possible. Given the following
gender system: male humans  – gender i, female humans  – gender ii, nouns
denoting animals – gender iii, but nouns denoting inanimates distributed over
three genders, namely ii, iii and iv. Within this system, the rooster and the frog
are both treated as gender iii, as in normal situations, throughout the story.
Compare this with Bininj Gun-wok (Mayali), a language of the Gunwinyguan
family of northern Australia, where alwanjdjuk ‘emu’ is feminine, irrespective of
sex. However, in exceptional circumstances, as in an account of what happens
when emus divorce, and the wife emu marries another male emu, the use of
the masculine is possible (Evans, Brown and Corbett 2002: 130–131). Finally, in
certain Southern Bole-Tangale languages (part of West Chadic, spoken in north-
eastern Nigeria), with two-gender systems, Leger (1998) reports that personifi-
cation is possible, but only for a small list of animals. Thus in Kupto (Kutto),
animals are usually of feminine gender, but when personified ‘a few selected
animals’ are treated as masculine: these are túlúm ‘hyena’, yèɗɗè ‘dog’, kúngú
‘leopard’ and gàandùk ‘mouse’, but not mbòlè ‘dove’, yóŋné ‘elephant or kúglú
 Gender typology       123

‘tortoise’, which remain feminine, even if they are personified. In the related
Kushi, dǝ̀ɓɓò ‘hyena’, jàŋàní ‘leopard’ and ʔàkùmóm ‘hedgehog’ can change
their gender to masculine, but not júr ‘squirrel’ or jèy ‘porcupine’, which must
retain their feminine gender. Other nouns may be taken into the sphere of the
sex-differentiables; for instance, in Mian: ‘Cross-classification for inanimates
extends to body parts, which are generally (but not obligatorily) assigned to a
gender reflecting the sex of their owner …’ (Fedden 2011: 177).
It is not only sex-differentiability which can be the basis for recategoriza-
tion. Savosavo, which is the easternmost Papuan language, and is spoken in
the Solomon islands, has two genders, masculine and feminine. Nouns denot-
ing inanimates are by default masculine. However, recategorization is possible,
according to a diminutive/affective classification. Nouns which are normally
masculine (since inanimate) are occasionally made feminine; this may function
to indicate that the referent is small compared with normal expectation or that
it is in some way special (Wegener 2008: 65–67). The less that gender agreement
is fully determined by unique gender values, the greater its role in constructing
meaning. Other languages where this can happen include Lavukaleve (Terrill
2003: 140–141), and Walman (Brown and Dryer 2008a: 530; 2008b), where recat-
egorization according to a diminutive/affective classification is widespread, and
for which there is a set of diminutive gender agreement forms. Other interesting
examples include Mawng (a member of the Iwaidjan family, spoken to the east
of Darwin, Singer 2010) and Yawuru (a member of the Nyulnyulan family, non-
Pama-Nyungan, of the west Kimberley region of Australia, Hosokawa 1996).
What then of the other morphosyntactic features? Recategorization in person
is found in certain types of address; for instance, addressees “should” be treated
as second person, but they can be recategorized as having other options, as in
the doctor’s How are we today? And full noun phrases, which “should” be third
person, may be recategorized as having a person choice, as in Spanish. This was
pointed out, for instance, by Harmer and Norton (1957: 270). These examples were
found recently on the web (I am grateful to Enrique Palancar for the examples
and analysis):

(39) Los hombre-s no pode-mos hac-er

def.m.pl man(m)-pl neg can.pres.ind-1pl do-inf

dos cosa-s a la vez

two thing(f)-pl at def.f.sg time(f)
‘We men cannot do two things at the same time.’
124       Greville G. Corbett

(40) La-s mujer-es so-is muy rar-a-s

def.f-pl woman[f]-pl be.pres.ind-2pl very weird-f-pl
‘You women are very weird.’

These examples nicely show plural noun phrases recategorized as first person
(39) and as second person (40).
We do not often think of recategorization for case, but the phenomena of Dif-
ferential Object Marking and Differential Subject Marking show the availability of
choices in case marking, driven by semantic and information structure factors, in
place of uniquely determined syntactic government. Thus recategorization may
be found in gender systems as well as in number systems, and it has analogues
in person and case too.

4.2.4 Distribution of gender systems

Given this typology of gender systems, we can now ask how they are distributed
over the world’s languages. There is relevant information in the World Atlas of
Language Structures (WALS), which contains 142 world maps, produced by 40
authors. In a sample of 256 languages (Corbett 2005), somewhat over half (144)
have no gender system. Certain parts of the world are rather arid in terms of
gender: the areas where Sino-Tibetan and Austronesian languages are spoken
have few gender systems. Yet even here, some are found. Schapper (2010) docu-
ments the appearance of gender in several Austronesian languages, from contact
with Papuan languages in her view. To have a gender system requires two genders
at least, and two-gender systems are common, with 50 examples in this sample.
Three genders is around half as common (26 examples) and four genders, about
half as common again (12). Larger systems, with five or more genders, represent a
substantial minority (24 languages in the sample). Fula (a Niger-Congo language)
has around twenty genders, depending on the dialect.
If we ask about the semantic core of the gender system, we find that the major-
ity have an assignment system based on sex (84),18 but 28 languages in the sample,
notably in the Niger-Congo and Algonquian families, have systems based on
animacy. And as for the type of assignment system, taking strictly semantic and pre-
dominantly semantic assignment systems together, we find these in just under half
the languages (53), while a slight majority (59) have semantic and formal assign-
ment. The distribution of the systems across the world leads me to suggest that

18 It is noteworthy that sex as a constant property of referents is central in many gender systems;
in contrast, it is rare for the sex of the interlocutors to have a key role (see Dunn this volume).
 Gender typology       125

new systems are always semantically based; they may persist over considerable
periods, as in Dravidian languages, or evolve into semantic and formal systems.
Older semantic and formal systems may in turn become primarily semantic (as is
happening in various Germanic languages), but this is a later stage. To date there is
no evidence for original systems having partially formal assignment.

5 Conclusion
The typology of gender reveals considerable diversity. And data from languages
which, until recently, were less well studied, reveal that gender is even more
varied and interesting than was earlier believed. On the other hand, the typology
has been somewhat simplified since we have reduced the types of semantic rela-
tions permitted. Gender is indeed remarkable, but to appreciate this we need to
see it in the context of the other morphosyntactic features; the deviations we find,
as measured against the canonical construct, are challenging and interesting. In
many ways, however, they are comparable to non-canonical behaviour we can
find with other morphosyntactic features. To make further progress we need to
maintain both perspectives: gender being unlike other features, and gender being
one morphosyntactic feature among a small number of comparable features.

6 References
A working bibliography on canonical typology can be found at:
http://www.surrey.ac.uk/LIS/SMG/canonical_bibliography.htm.
Aikhenvald, Alexandra Y. 1994. Classifiers in Tariana. Anthropological Linguistics 36. 407–465.
Aikhenvald, Alexandra Y. 2000. Classifiers: A Typology of Noun Classification Devices. Oxford:
Oxford University Press.
Audring, Jenny. 2009. Reinventing pronoun gender. PhD thesis, Vrije Universiteit, Amsterdam.
Utrecht: LOT.
Audring Jenny. 2011. Gender. In Oxford Bibliographies Online: Linguistics,
http://www.oxfordbibliographiesonline.com/view/document/obo-9780199772810/
obo-9780199772810-0066.xml (accessed 4-Jan-2012).
Austin, Peter. 1981. A Grammar of Diyari, South Australia. Cambridge: Cambridge University
Press.
Baerman, Matthew, Dunstan Brown, and Greville G. Corbett. 2005. The Syntax-Morphology
Interface: A study of syncretism. Cambridge: Cambridge University Press.
Bakker, Peter. 1997. A Language of Our Own: The Genesis of Michif, the Mixed Cree-French
Language of the Canadian Métis. Oxford: Oxford University Press.
Bani, Ephraim. 1987. Garka a ipika: masculine and feminine grammatical gender in Kala Lagaw
Ya. Australian Journal of Linguistics 7. 189–201.
126       Greville G. Corbett

Bickel, Balthasar and Johanna Nichols. 2005. Inclusive-exclusive as person vs. number
categories worldwide. In Elena Filimonova (ed.) Clusivity: Typology and Case Studies of the
Inclusive–Exclusive Distinction (Typological Studies in Language 63), 49–72. Amsterdam:
John Benjamins.
Brown, Dunstan, Marina Chumakina and Greville G. Corbett (eds.) 2013. Canonical Morphology
and Syntax. Oxford: Oxford University Press.
Brown, Lea and Matthew S. Dryer. 2008a. The verbs for ‘and’ in Walman, a Torricelli language of
Papua New Guinea. Language 84. 528–565.
Brown, Lea and Matthew S. Dryer. 2008b. Diminutive as an inflectional category in Walman.
Ms. University at Buffalo. Available at: linguistics.buffalo.edu/people/faculty/dryer/
dryer/BrownDryerWalmanDimin.pdf. [Accessed 19.1.2009.]
Browne, Gerald M. 2002. Old Nubian Grammar (Languages of the World/Materials 330).
Munich: Lincom Europa.
Chumakina, Marina, Dunstan Brown, Harley Quilliam and Greville G. Corbett. 2007. Slovar’
arčinskogo jazyka (arčinsko-anglo-russkij) [A dictionary of Archi: Achi-Russian-English].
Makhachkala: Delovoj Mir. [Available as a WWW version on the SMG website].
Chumakina, Marina and Greville G. Corbett. 2008. Archi: the challenge of an extreme
agreement system. In A. V. Arxipov, L. M. Zaxarov, A. A. Kibrik, A. E. Kibrik, I. M. Kobozeva,
O. F. Krivnova, E. A. Ljutikova and O. V. Fëdorova (eds.) Fonetika i nefonetika: K 70-letiju
Sandro V. Kodzasova. 184–194. Moscow: Jazyki slavjanskix kul’tur.
Comrie, Bernard. 2003. When agreement gets trigger-happy. In Dunstan Brown, Greville
Corbett and Carole Tiberius (eds.) Agreement: a typological perspective. Special issue of
Transactions of the Philological Society 101(2). 313–337. Oxford: Blackwell.
Comrie, Bernard. 2005. Grammatical gender and personification. In Dorit Diskin Ravid and Hava
Bat-Zeev Shyldkrot (eds.) Perspectives on Language and Language Development: Essays
in Honor of Ruth A. Berman. 105–114. Dordrecht: Kluwer.
Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press
Corbett, Greville G. 2003. Agreement: Canonical instances and the extent of the phenomenon.
In Geert Booij, Janet DeCesaris, Angela Ralli and Sergio Scalise (eds.) Topics in
Morphology: Selected papers from the Third Mediterranean Morphology Meeting
(Barcelona, September 20–22, 2001). 109–128. Barcelona: Universitat Pompeu Fabra.
Corbett, Greville G. 2005. The canonical approach in typology. In Zygmunt Frajzyngier, Adam
Hodges and David S. Rood (eds.) Linguistic Diversity and Language Theories (Studies in
Language Companion Series 72). 25–49. Amsterdam: John Benjamins.
Corbett, Greville G. 2006. Agreement. Cambridge: Cambridge University Press
Corbett, Greville G. 2007a. Canonical typology, suppletion and possible words. Language 83.
8–42.
Corbett, Greville G. 2007b. Gender and noun classes. In Timothy Shopen (ed.) Language
Typology and Syntactic Description: III: Grammatical categories and the lexicon, second
edition, 241–279. Cambridge: Cambridge University Press.
Corbett, Greville G. 2009. Canonical inflectional classes. In Fabio Montermini, Gilles Boyé and
Jesse Tseng (eds.) Selected Proceedings of the 6th Décembrettes: Morphology in Bordeaux.
1–11. Somerville, MA: Cascadilla Proceedings Project. Available at:
http://www.lingref.com/cpp/decemb/6/abstract2231.html
Corbett, Greville G. 2011. The penumbra of morphosyntactic feature systems. Morphology
21. (= Jonathan Bobaljik, Uli Sauerland and Andrew Nevins (eds.) Markedness and
Underspecification in the Morphology and Semantics of Agreement) 445–480.
 Gender typology       127

Cowell, Mark W. 1964. A Reference Grammar of Syrian Arabic (based on the dialect of
Damascus) (Arabic series 7). Washington, DC: Georgetown University Press.
De Vos, Lien and Gunther De Vogelaar. 2011. Dutch gender and the locus of morphological
regularization. Folia Linguistica 45. 245–281.
Demuth, Katherine and Sara Weschler. 2012. The acquisition of Sesotho nominal agreement.
Morphology 22. 67–88.
Dešeriev, Ju. D. 1953. Bacbijskij jazyk: fonetika, morfologija, sintaksis, leksika. Moscow: AN SSSR.
Dixon, R. M. W. 1972. The Dyirbal Language of North Queensland. Cambridge: Cambridge
University Press.
Dixon, R. M. W. 1982. Where Have All the Adjectives Gone? and other essays in Semantics and
Syntax. Berlin: Mouton.
Dol, Philomena. 2007. A Grammar of Maybrat: a language of the Bird’s Head Penisula, Papua
Province, Indonesia (Pacific Linguistics, 586). Canberra: Pacific Linguistics, Research
School of Pacific and Asian Studies, Australian National University.
Donohue, Mark. 2001. Animacy, class and gender in Burmeso. In Andrew Pawley, Malcolm Ross
and Darrell Tryon (eds.) The boy from Bundaberg: Studies in Melanesian linguistics in
honour of Tom Dutton (Pacific linguistics 514). 97–115. Canberra: Pacific Linguistics.
Enger, Hans-Olav. 2009. The role of core and non-core semantic rules in gender assignment.
Lingua 119. 1281–1299.
Evans, Nicholas. 2003. Typologies of agreement: some problems from Kayardild. In Dunstan
Brown, Greville G. Corbett and Carole Tiberius (eds.) Agreement: a typological perspective
(Special issue of Transactions of the Philological Society 101(2)). 203–234. Oxford:
Blackwell.
Evans, Nicholas, Dunstan Brown and Greville G. Corbett. 2002. The semantics of gender in
Mayali: Partially parallel systems and formal implementation. Language 78. 111‑155.
Fedden, Sebastian. 2011. A Grammar of Mian. Berlin: De Gruyter Mouton.
Fortin, Antonio. 2011. The Morphology and Semantics of Expressive Affixes. Unpublished DPhil
thesis, University of Oxford.
Goddard, Cliff. 1982. Case systems and case marking in Australian languages: a new
interpretation. Australian Journal of Linguistics 2. 167–196.
Hardison, Debra M. 1992. Gender assignment to nonwords in French: Implications for the role
of the final syllable in lexical processing and organization of the mental lexicon. In Indiana
Linguistics Club Twenty-Fifth Anniversary Volume. 13–25. Bloomington, Indiana: Indiana
University Linguistics Club.
Harmer, L. C. and F. J. Norton. 1957. A Manual of Modern Spanish. London: University Tutorial
Press. [2nd edition, 1st edition 1935]
Haviland, John. 1979. Guugu Yimidhirr. In R. M. W. Dixon and Barry J. Blake (eds.), Handbook of
Australian languages, 27–180. Canberra: Australian National University Press.
Heuvel, Wilco van den. 2006. Biak: Description of an Austronesian language of Papua. PhD thesis,
Vrije Universiteit Amsterdam. Available at: http://www.lotpublications.nl/index3.html.
Hockett, Charles F. 1958. A Course in Modern Linguistics. New York: Macmillan.
Holisky, Dee Ann and Rusudan Gagua. 1994. Tsova–Tush (Batsbi). In Rieks Smeets (ed.)
Indigenous Languages of the Caucasus IV: The North East Caucasian Languages, Part 2.
147–212. Delmar, NY: Caravan Books.
Hosokawa, Komei. 1996. ‘My face am burning!’: quasi-passive, body parts and related issues
in Yawuru grammar and cultural concepts. In Hillary Chappell and William McGregor
128       Greville G. Corbett

(eds.) The grammar of inalienability: a typological perspective on body part terms and the
part-whole relation. 155–192. Berlin: Mouton de Gruyter.
Hyman, Larry M. 2009. How (not) to do phonological typology: the case of pitch-accent.
Language Sciences 3. (= Michael J. Kenstowicz (ed.) Data and Theory: Papers in Phonology
in Celebration of Charles W. Kisseberth) 213–128. Amsterdam: Elsevier.
Hyman, Larry M. 2011. In defense of prosodic typology: A response to Beckman and Venditti. UC
Berkeley Phonology Lab Annual Report. 200–235.
Igartua, Iván. 2006. Genus alternans in Indo-European. Indogermanische Forschungen 111.
56–70.
Jensen, Hans. 1959. Altarmenische Grammatik. Heidelberg: Carl Winter.
Karatsareas, Petros. 2009. The loss of grammatical gender in Cappadocian Greek. Transactions
of the Philological Society 107. 196–230.
Kibrik, Aleksandr E. 1977. Opyt strukturnogo opisanija arčinskogo jazyka: II: Taksonomičeskaja
grammatika. Publikacii otdelenija strukturnoj i prikladnoj lingvistiki 12. Moscow:
Izdatel’stvo Moskovskogo universiteta.
Kibrik, Aleksandr E. 1998. Archi (Caucasian – Daghestanian). In Andrew Spencer and Arnold
M. Zwicky (eds.) The Handbook of Morphology. 455–476. Oxford: Blackwell.
Kibrik, Aleksandr E. (ed.). 2001. Bagvalinskij jazyk: Grammatika: Teksty: Slovari. Moscow:
Nasledie. [co-editors K. I. Kazenin, E. A. Ljutikova and S. G. Tatevosov.]
Kibrik, A. E., S. V. Kodzasov, I. P. Olovjannikova and D. S. Samedov. 1977. Arčinskij jazyk: Teksty
i slovari. Publikacii otdelenija strukturnoj i prikladnoj lingvistiki 14. Moscow: Izdatel’stvo
Moskovskogo universiteta.
Klein, Jared. 2007. Classical Armenian morphology. In Alan B. Kaye (ed.) Morphologies of Asia
and Africa II. 1051–1086. Winona Lake, IN: Eisenbrauns.
Lakoff, George. 1987. Women, Fire, and Dangerous Things: What Categories reveal about the
Mind. Chicago: University of Chicago Press.
Leger, Rudolf. 1998. Grammatical gender in some Southern Bole-Tangale languages – Kwami,
Kupto, Kushi and Piya. In I. V. Sledzevskij and D. M. Bondarenko (eds.) Afrika: Obščestva,
kul’tury, jazyki. 204–216. Moscow: Institute of African Studies.
Lindström, Eva. 2002. Topics in the grammar of Kuot: a non-Austronesian language of New
Ireland, Papua New Guinea. PhD dissertation: Stockholm University.
Loporcaro, Michele and Tania Paciaroni. 2011. Four-gender systems in Indo-European. Folia
Linguistica 45. 389–433.
Maiden, Martin. 2011. Morphophonological persistence. In Martin Maiden, John Charles
Smith and Adam Ledgeway (eds.) The Cambridge History of the Romance Languages: I:
Structures. 155–215. Cambridge: Cambridge University Press.
Matasović, Ranko. 2004. Gender in Indo-European. Heidelberg: Winter.
Mel’čuk, Igor. 1986. Toward a definition of case. In Richard D. Brecht and James Levine (eds.)
Case in Slavic. 35–85. Columbus, OH: Slavica. [Revised version in: Igor Mel’čuk. 2006.
Aspects of the Theory of Morphology, edited by David Beck. 110–179. Berlin: Mouton de
Gruyter.]
Modi, Bharati. 2003. Gender in Gujarati. In Ritva Laury, Gerald McMenamin, Shigeko Okamoto,
Vida Samiian and K.V. Subbarao (eds.). Perspectives in Linguistics: Papers in Honor of P.J.
Mistry. 247–259. New Delhi: Indian Institute of Language Studies.
Nedelcu, Isabela. 2013. Three genders: masculine, feminine and neuter. In Gabriela Pană
Dindelegan (ed.) The Grammar of Romanian. Oxford: Oxford University Press.
 Gender typology       129

Nikolaeva, Irina and Andrew Spencer. 2008. Nouns as adjectives and adjectives as nouns.
Unpublished manuscript.
Parker, E. M. and R. J. Hayward. 1985. An Afar-English-French Dictionary (with Grammatical
Notes in English). London: SOAS, University of London.
Plaster, Keith and Maria Polinsky. 2010. Features in categorization, or a new look at an old
problem. In Anna Kibort and Greville G. Corbett (eds.) Features: Perspectives on a Key
Notion in Linguistics. 109–142. Oxford: Oxford University Press.
Plaster, Keith, Maria Polinsky and Boris Harizanov. In press. Noun classes grow on trees: Noun
classification in the North-East Caucasus. To appear in a Festschrift. Amsterdam: John
Benjamins.
Polinsky, Maria. 2003. Non-canonical agreement is canonical. In Dunstan Brown, Greville
Corbett and Carole Tiberius (eds.) Agreement: A Typological Perspective (special issue of
Transactions of the Philological Society 101(2)). 279–312. Oxford: Blackwell.
Polinsky, Maria. 2008. Gender under incomplete acquisition: heritage speakers’ knowledge of
noun categorization. Heritage Language Journal 6. 40–71.
Polinsky, Maria. and Ezra Van Everbroeck, E. 2003. Development of gender classifications:
Modeling the historical change from Latin to French. Language 79. 356–390.
Reid, Nicholas. 1997. Class and classifier in Ngan’gityemerri. In Mark Harvey and Nicholas Reid
(eds.) Nominal Classification in Aboriginal Australia (Studies in Language Companion
Series 37). 165–228. Amsterdam: John Benjamins.
Rodina, Yulia. 2007. Semantics and Morphology: The Acquisition of Grammatical Gender in
Russian. PhD thesis. University of Tromsø.
Rovai, Francesco. 2012. Between feminine singular and neuter plural: re-analysis patterns.
Transactions of the Philological Society 110. 94–121
Sagot, Benoît and Géraldine Walther. 2011. Non-canonical inflection: Data, formalization and
complexity measures. In Cerstin Mahlow and Michael Piotrowski (eds.) Communications in
Computer and Information Science 100. 23–45. Berlin: Springer.
Sansò, Andrea. 2009. Men, women, and birds. An embryonic system of noun classification in
Ancient Greek. Folia Linguistica 43. 95–133.
Schapper, Antoinette. 2010. Neuter gender in eastern Indonesia. Oceanic Linguistics 49. 407–435.
Seifart, Frank. 2005. The Structure and Use of Shape-based Noun Classes in Miraña (North West
Amazon). PhD thesis, Radboud University, Nijmegen.
Singer, Ruth. 2010. Creativity in the use of gender agreement in Mawng: How the discourse
functions of a gender system can approach those of a classifier system. Studies in
Language 34. 382–416.
Spencer, Andrew. 2005. Towards a typology of ‘mixed categories’. In C. Orhan Orgun and Peter
Sells (eds.) Morphology and the Web of Grammar: Essays in Memory of Steven G. Lapointe
(Stanford Studies in Morphology and the Lexicon). 95–138. Stanford: CSLI Publications.
Staden, Miriam van. 2000. Tidore: A Linguistic Description of a Language of the North
Moluccas. PhD thesis, Leiden University.
Stein, David E. S. Forthcoming. Gender representation in Biblical Hebrew. To appear in Geoffrey
Khan (ed.) Encyclopedia of Hebrew Language and Linguistics. Leiden: Brill.
Stolz, Christel. 2009. A different kind of gender problem: Maltese loanword gender from a
typological perspective. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud,
Thomas Stolz and Martine Vanhov (eds.) Introducing Maltese Linguistics: Selected papers
from the 1st International Conference on Maltese Linguistics, Bremen, 18–20 October, 2007
(Studies in Language Companion Series 113). 321–353. Amsterdam: John Benjamins.
130       Greville G. Corbett

Strom, Clay. 1992. Retuarã Syntax (Studies in the Languages of Colombia 3). Dallas, TX: Summer
Institute of Linguistics and The University of Texas at Arlington.
Stump, Gregory T. 2005. Word-formation and inflectional morphology. In Pavol Štekauer and
Rochelle Lieber (eds) Handbook of Word-formation (Studies in Natural Language and
Linguistic Theory 64), 49–71. Dordrecht: Springer.
Stump, Gregory T. 2006. Heteroclisis and paradigm linkage. Language 82. 279–322.
Stump, Gregory T. and Raphael Finkel. 2008. Stem alternations and principal parts in
French verb inflection. Paper presented at Décembrettes 6: Colloque International de
Morphologie, “Morphologie et classes flexionnelles”. December 4–5. 2008. Université de
Bordeaux, France.
Suthar, Babubhai Kohyabhai. 2006. Agreement in Gujarati. PhD Dissertation, University of
Pennsylvania.
Terrill, Angela. 2002. Systems of nominal classification in East Papuan languages. Oceanic
Linguistics 41. 63–88.
Terrill, Angela. 2003. A Grammar of Lavukaleve (Mouton Grammar Library 30). Berlin: Mouton
de Gruyter.
Thornton, Anna M. 2009. Constraining gender assignment rules. Language Sciences 31. 14–32.
Thornton, Anna M. 2011. Overabundance (multiple forms realizing the same cell): a non-canonical
phenomenon in Italian verb morphology. In Maria Goldbach, Marc-Olivier Hinzelin, Martin
Maiden and John Charles Smith (eds.) Morphological Autonomy: Perspectives from Romance
Inflectional Morphology. 358–381. Oxford: Oxford University Press.
Thornton, Anna M., Claudio Iacobini, and Cristina Burani. 1997. BDVDB: Una base di dati per il
vocabolario di base della lingua italiana, 2nd edition. Rome: Bulzoni.
Tucker, G. Richard, Wallace E. Lambert, and André Rigault, A. A. 1977. The French speaker’s skill
with grammatical gender: An example of rule-governed behavior. The Hague: Mouton.
Visser, Willem. 2011. Historical gender change in West Frisian. Morphology 21. 31–56.
Wegener, Claudia. 2008. A grammar of Savosavo: a Papuan language of the Solomon Islands.
(MPI series in psycholinguistics). PhD thesis, Radboud University, Nijmegen.
Zaliznjak, Andrej A. 1964. K voprosu o grammatičeskix kategorijax roda i oduševlennosti v
sovremennom russkom jazyke. Voprosy jazykoznanija 4. 25–40.
Zaliznjak, Andrej A. 1973. O ponimanii termina ‘padež’ v lingvističeskix opisanijax. In Andrej
A. Zaliznjak (ed.) Problemy grammatičeskogo modelirovanija. 53–87. Moscow: Nauka.
[Reprinted in: Andrej A. Zaliznjak. 2002. Russkoe imennoe slovoizmenenie: s priloženiem
izbrannyx rabot po sovremennomu russkomu jazyku i obščemu jazykoznaniju. 613–647.
Moscow: Jazyki slavjanskoj kul’tury.]
Marianne Mithun
Gender and culture
1 Language and culture
It is now generally recognized that grammatical categories develop in lan-
guages through use. Distinctions made most often by speakers as they speak
tend to become routinized over time in grammatical markers. Many grammati-
cal categories recur in language after language, no doubt because they reflect
common human interests. Most languages have grammatical causative con-
structions, for example, and many have tense systems. But other grammati-
cal markers reflect more specific environmental or cultural interests. Central
Alaskan Yup’ik, for example, contains a suffix -ir- ‘have cold N’. Added to
the noun root ‘ear’, it forms the stem ciuta-ir- ‘have cold ears’, as in the verb
ciuta-ir-tua ‘my ears are cold’ (George Charles, speaker p.c.). It is probably no
accident that this language is spoken in the Arctic. Central Pomo, a language
indigenous to California, has a prefix čh- ‘by gambling’. Prefixed to the verb
ley ‘exhaust, die off’, it yields the verb čh-léy ‘to lose everything in gambling’
(Frances Jack, speaker p.c.). California peoples are known for their long tradi-
tions of gaming.
Grammatical gender systems appear in languages distributed throughout
the world. They show many similarities cross-linguistically, but there are also
interesting differences. Comparisons among them raise questions about the
extent to which grammatical gender reflects cognitive universals versus cultural
specificity.

1.1 Mohawk

Mohawk, an Iroquoian language indigenous to northeastern North America,

shows a grammatical gender system that at first looks much like those of more
familiar European systems. Nouns contain gender prefixes.1

1 Examples are presented here in the official orthography in use in the Mohawk communities.
Vowels i, e, a, o have approximately IPA values. Nasalized vowels are represented by digraphs:
en [ᴧ̜] and on [u̜]. Glottal stop is represented by an apostrophe ’. The letter i represents a glide [j]
before vowels. Vowel length is represented by a colon :. Stress with rising tone is indicated with
an acute accent ó; stress with falling tone is indicated with a grave accent ò.
132       Marianne Mithun

(1) Noun prefixes

Masculine ra-ksà:’a ‘boy’
Feminine e-ksà:’a ‘girl’
Neuter ka-ièn:kwire’ ‘arrow’

1.2 Agreement

It has been proposed that a fundamental feature of grammatical gender systems is

the presence of agreement. Hockett defined gender in just this way: “Genders are
classes of nouns reflected in the behavior of associated words.” (1958: 231, cited
in Corbett 1991: 1). Corbett concurs. “While nouns may be classified in various
ways, only one type of classification counts as a gender system; it is one which is
reflected beyond the nouns themselves in modifications required of ‘associated
words’” (1991: 4). The Mohawk examples in (2) indicate that the language does
show grammatical gender in the strict sense.

(2) Mohawk matching prefixes (examples constructed for comparison)

a. Ra-ká:we-hs ne ra-ksà:=’a.
3m.sg.agt-paddle-hab the m.sg-child=dim
‘The boy is paddling.’

b. Ie-ká:we-hs ne e-ksà:=’a.
3f.sg.agt-paddle-hab the f.sg-child=dim
‘The girl is paddling.’

c. Ka-ká:we-hs ne kén-tsion.
3n.sg.agt-paddle-hab the n-fish
‘The fish is paddling.’

d. Rati-ká:we-hs ne rati-ksa’=okòn:’a.
3m.pl.agt-paddle-hab the m.pl-child=distr
‘The boys are paddling.’

(By regular rule, the vowel of the Neuter prefix ka- fuses with the vowel i of a fol-
lowing stem, yielding the nasal vowel en: ka-itsion > kéntsion ‘fish’.)
The verbal prefix system goes a step further. Verb prefixes represent both
arguments of a transitive clause. In some cases, the components corresponding
to the two arguments can still be discerned, but for the most part these prefixes
are highly fused forms.
 Gender and culture       133

(3) Mohawk prefixes in transitives (constructed for comparison)2

a. Ronwa-nòn:we’-s ne ra-ksà:=’a.
3f.sg>3m.sg-like-hab the m.sg-child=dim
‘She likes the boy.’

b. Iontate-nòn:we’-s ne e-ksà:=’a.
3f.sg>3f.sg-like-hab the f.sg-child=dim
‘She likes the girl.’ or ‘The girl likes her.’

c. Konwa-nòn:we’-s ne kén-tsion.
3f.sg>3n.sg-like-hab the n-fish
‘She likes the fish.’

d. Ronwati-nòn:we’-s ne rati-ksa’=okòn:’a.
3f.sg>3m.pl-like-hab the m.pl-child=dim
‘She likes the boys.’

Overt lexical nominals or free pronouns need not be present in every clause. Ref-
erence can be clear from the verbal prefixes.

2 Males
Not surprisingly, Masculine gender markers are used for all male persons and
certain animals whose gender is salient, such as bulls and roosters. Masculine
forms are used on occasion for other male animals as well, such as pet dogs or

2 The abbreviations used for grammatical terms are the following.

AGT Grammatical Agent INS Instrumental applicative
AUG Augmentative LK Linker
CISLOC Cislocative M Masculine
CONTR Contrastive N Neuter
DIM Diminutive NS Noun Suffix
DISTR Distributive PAT Grammatical Patient
DU Dual PFV Perfective
DV Duplicative PL Plural
F Feminine PROG Progressive
FACT Factual PRT Partitive
FI Feminine-Indefinite REP Repetitive
FUT Future REV Reversive
FZ Feminine-Zoic SG Singular
HAB Habitual aspect TRLOC Translocative
INCH Inchoative
134       Marianne Mithun

other significant individuals, particularly large ones. The excerpt in example (4)
is a free translation of a Mohawk account of an encounter with a polar bear. The
bolded English pronouns reflect the gender distinguished by the verb prefixes.

(4) Encounter with a polar bear: Josephine Kaieríthon Horne, speaker, p.c.
‘The bear (he) stood up. He was about twelve feet tall. When he opened
his mouth, the inside of it looked like a ball of fire next to his huge
white teeth. It took several shots to kill (him) the bear. The two brothers
were all excited. They quickly went back to camp and happily reported
that they had killed (it) a bear. But by the time they got back to the spot
where (it) the bear had been killed, it was nowhere to be found.’

The speaker used Masculine forms while the bear was alive, but Neuter forms
once he was dead.
Masculine dual and plural prefixes are also used for mixed groups of male
and female persons.

3 Objects and animals

Neuter forms are generally used for objects and animals. The most common Neuter
prefixes are ka-/ken- and o-, though there are a few others, and phonological pro-
cesses can obscure some forms. The kinds of nouns that occur with these prefixes
are not surprising: kèn-reks ‘wildcat, mountain lion’, ka-henta’kéha’ meadow-
lark’, ka-nó:tsot ‘muskie (fish)’, ka-ríhton ‘red oak’, ká-hi ‘fruit, berry’, kahén:ta’
‘field, meadow’, ka-honwé:ia’ ‘boat’, ká-tshe’ ‘bottle, jug, can’, ka-nhóha’ ‘door’ ;
o-hkwá:ri ‘bear’, o-hriò:ken ‘chipmunk’, o-tsí:non ‘flea, louse’, o-tsi’tèn:’a ‘bird’,
o-hkwé:sen ‘partridge’, o-iahè:ta’ ‘perch (fish)’, otsi’eróhta’ ‘crab’, ó-se’ ‘willow’,
ó-honte’ ‘grass’, otsì:tsa’ ‘flower’, o-hnennà:ta’ ‘potato’, ó’-:iente’ ‘wood’, o-’kèn:ra’
‘dust, soil’, o-nén:ia’ ‘rock’, etc.
For the most part, animals and inanimate objects are represented by the
same prefixes. Distinctions can be seen in a few contexts, however. Number is
distinguished for animates in verb prefixes, but not for inanimates.

(5) Number distinction for animates: Rita Konwatsi’tsaién:ni Phillips,

speaker p.c.
a. Awenhniserakwé:kon en-ka-hón:take’.
all day long it will grass eat
‘All day long it (a cow) would graze.’
 Gender and culture       135

b. Awenhniserakwé:kon en-konti-hón:take’.
all day long they (zoic.pl) will grass eat
‘All day long they (cows) would graze.’

(6) No number distinction for inanimates: Josephine Kaieríthon Horne,

speaker p.c.
a. Kanákta’ ka-kè:ron.
place it is laid out
‘A place was laid out (at the table).’

b. Kanakta’shòn:’a ka-kè:ron.
place here and there it is laid out
‘Places were laid out (at the table).’

Such a pattern is not uncommon cross-linguistically (Corbett 1991: 267–269, Enfield

2007: 78, and many others). Distinctions between animate Neuters (termed Zoic
in Iroquoian linguistics) and inanimate Neuters appear in a few other contexts to
be described later.
Not surprisingly, inanimate objects are sometimes personified, such as in
ceremonial contexts. In (7) the earth is categorized as Feminine, and in (8), the
medicines are categorized as Masculine.

(7) Thanksgiving address: Enos Williams, speaker

Wa’tiethinonhwerá:ton’ ne tsi
we hereby raise thanks to her the where
iohwentsiatátie’.
it earth extends along
‘We thank her, the earth.’ …
Teiethinòn:weron ahsonthenhnéhkha’ wenhnì:tare’.
we thank her nighttime it moon shines
‘We greet her, the nighttime moon.’

(8) Oratory: Frank Tekaronhió:ken Jacobs Jr., speaker

Ietshiiatahónhsatat ne ononhkwa’shòn:’a,
you all listen to them the various medicines
‘Listen to the medicines,
tsi nahò:ten’ rón:ton.
as such it is a kind they (m.pl) are saying
to what they (m.pl) are saying.’
136       Marianne Mithun

4 Female persons
So far, the Mohawk system looks much like those of European languages. It differs
in some intriguing ways, however. One is that the prefixes used for female persons
are also used as generics ‘one, they, people’, for persons of unspecified sex, and
indefinites. One domain in which generics show up routinely is in terms for tools
and other devices. Morphological verbs are often used as referring expressions,
and are often lexicalized as such. (The forms below are not restricted to utensils
typically used by women.)

(9) Feminine and generic usage

a. ie-hiá:tonhs ‘she, one writes’
ie-hiatónhkhwa’ ‘she, one writes with it’ = ‘pen, pencil’

b. ie-kòn:reks ‘she, one pounds’

ie-konhrékstha’ ‘she, one pounds with it’ = ‘hammer’

Additional examples of generic uses of these forms are in (10), all from spontane-
ous speech.

(10) Generic usage in speech

a. Ieióhe na-ie-nenhstaienthó:ko’.
it is time one should unplant the corn
‘It is time to harvest the corn.’

b. Ahshakoié:na’ ne iakaonkwe’táksen
he would catch one the one is a bad person
‘He would catch bad people.’

c. Ónhka’ iakotohetstonhátie’ aiakhinontará:nonte’.

someone one is passing by we would soup feed one
‘We would give soup to any strangers who might come our way.’

d. Iah ónhka’ teiakotahonhsatá:ton.

not someone not has one pricked one’s ears
‘No one listened.’

e. Iohsnó:re’ sahón:newe’ tsi nón:

it is fast they arrived back to place
ie-nakerénion’ nón:kwe.
one lives here and there the person
‘They soon returned to civilization.’
 Gender and culture       137

In combination with an indefinite pronoun such as ónhka’k ‘someone’, the

same forms are used as indefinites.

(11) Indefinite usage in speech

Ónhka’k wa-hshakó:-ken’ taiakawenonhátie’.
someone he saw one one is coming along
‘He saw someone coming.’  (Sex as yet undetermined)

This pronominal category is termed Feminine-Indefinite by Iroquoianists, abbre-

viated FI.
The use of the same forms for female persons, generics, and indefinites sug-
gests a link between grammar and culture. Iroquoian culture is characterized by
longstanding matrilineal and matrilocal traditions. Clan membership is inher-
ited through the mother. Land was traditionally passed down from mothers to
daughters. Women have always been highly respected, and they serve important
community roles. Chiefs were traditionally men, but they were appointed (and
potentially recalled) by the clan mothers. It is perhaps no surprise that generic
persons should be represented by Feminine gender forms.

5 Not so fast
In fact the picture is more interesting. Feminine-Indefinite (fi) prefixes are indeed
often used in reference to female persons. Just as often, however, a different set of
prefixes is used, the same Zoic prefixes used for animals. (This category is accord-
ingly termed Feminine-Zoic (fz) by Iroquoianists.)
The use of two categories for female persons immediately raises the ques-
tion of the difference between the two. Speakers are usually surprised when it
is brought to their attention. Some are horrified to realize that they are using the
same forms for women and animals. They report that they never thought about it,
and that “No one notices”. The difference in usage between the Feminine-Indef-
inite and Feminine-Zoic is in fact subtle and intriguing, and subject to variation
across communities, families, and even individuals.

5.1 Feminine-Indefinite

When speakers become interested in the difference between the two genders used
for women, and ask others to reflect on their own usage, they typically pose ques-
tions in terms of the emphatic/contrastive pronouns, akaónha ‘she, one’ (Feminine-
138       Marianne Mithun

Indefinite) and aónha ‘she, it’ (Feminine-Zoic): “Would you use akaónha or aónha
for X?” A point on which all speakers immediately agree is that akaónha (fi) is used
for grandmothers and mothers. This principle is often characterized in terms of
respect. One speaker generalized the principle as follows: “The baseline is to refer
to older relatives as akaónha (fi), younger as aónha (fz), but this can change accord-
ing to personality”. An elderly lady noted, “The older you get, the more respect
people have for you, so you’ll hear less and less aónha (fz).” Many volunteered that
akaónha (fi) is used “for refined, delicate, lady-like persons”. One speaker related
usage to the speaker: “Akáonha (fi) is used by people who are gentler.”
The Feminine-Indefinite is used not just for older women, however. One
speaker reported that she would use akaónha (fi) “always for little girls, but I
know some who would go by personality”. Another reported that she used
akaónha (fi) “for all my daughters, but I know some would do it differently”.
Several volunteered that Feminine-Indefinite forms have a “diminutive sense”
and indicate “endearment”. There are, for example, two forms translatable as
‘my daughter’. Speakers agree that “Both are definitely acceptable, and both are
used all the time.” Some noted a slight difference in connotation, however.

(12) Daughters
kheièn:’a ‘my daughter’ fi “more loving”
tièn:’a ‘my daughter’ fz “more frivolous”

Often mothers are initially shocked to realize that they use different gender pre-
fixes for different daughters. After some reflection, one observed, “Often the first-
born girl is aónha (fz), then succeeding girls are akaónha (fi).”

5.2 Feminine-Zoic

Speakers often remark that aónha (fz) is used for “louder, more powerful, more
assertive, more aggressive women”. Those referred to with aónha (fz) forms are
characterized as “tough women”, with “an outgoing nature”. The speaker cited
above who remarked that akaónha (fi) forms are used by gentler speakers also
noted that aónha (fz) forms would be “used more by someone who is really rough”.
Another speaker volunteered interesting comments on the two forms in (12).

(13) Pretty women

ie-ksa’tí:io ‘she is pretty’ fi “a bit more ladylike”
ka-ksa’tí:io ‘she is pretty’ fz “the one he’d like to go out with”
 Gender and culture       139

The forms in (12) could suggest a difference in degree of respect, and one
speaker made such an observation: “Aónha (fz) would be used for someone you
don’t respect: an animal or a stranger”. These forms are indeed common for non-
Mohawk women. The terms for ‘nun’ and ‘queen’ are both Feminine-Zoic.

(14) Non-Mohawk women

io-ia’tatokénhti ‘nun’ fz (literally ‘she fz is bodily holy’)
ka-kwí:n ‘the Queen’ fz (Queen Elizabeth)

Speakers are quick to assert, however, that the forms “do not necessarily indicate
a lack of respect”. One woman noted, “I wouldn’t be insulted if someone used
aónha (fz) for me. It would mean they just don’t know me well. I do the same with
others.” Another volunteered, “This is not about respect, but about distance.”
Still another was quick to say that Feminine-Zoic forms “are not rude, never felt to
be rude.” One explained these forms as appropriate for “someone we don’t know,
who is not an elder”.
A slightly different angle was suggested by another speaker, and her comment
was met with unanimous agreement. “Aónha (fz) might be used for someone you
don’t like, though the form itself doesn’t say that”. It was explained that use of
Feminine-Zoic forms might convey dislike only if the speaker had always used
Feminine-Indefinite forms for an individual before.
Age is mentioned in discussions about aónha (fz) forms as well. Several
women have noted that aónha (fz) is used for “contemporaries, women your
own age”, and their own usage corresponds to their observations. The aónha (fz)
forms are used for good friends. They add that “older or younger women could be
either”. These assessments highlight the subtlety of the distinction, when they
are contrasted with the tone of endearment felt in the Feminine-Indefinite daugh-
ter terms seen in (11).
Speakers also observe that use of the two categories is very much a family
matter. When one woman first became conscious of the grammatical distinc-
tion, she asked herself, “Why do I say aónha (fz) for my aunt?” She then realized,
“Ah, I’m just imitating my mother.” Another noted that “A and her sisters use
aónha (fz) forms for each other, with no disrespect.” One hypothesized that “The
mother decides for a baby girl, and that categorization usually stays for life.”
Thinking about her own family, she made the following observations. “I am the
oldest daughter, so I am aónha (fz). My sisters are akaónha (fi). They always use
aónha for me, and I use akaónha for them. I always refer to all my aunts as aónha
(fz). But that may be because my mother was the youngest and has always been
akaónha (fi).”
140       Marianne Mithun

Usage also differs from one community to the next. Men in Kahnawà:ke gen-
erally refer to their wives as akaónha (fi), interpreted as a sign of respect. Men in
Ahkwesáhsne, however, approximately 50 miles upriver, generally refer to their
wives as aónha (fz), interpreted as a sign of affection. One speaker recounted a
story from the time she was being courted. Her family was originally from the
Ahkwesáhsne community. When she was a teenager, they moved downriver to
the Kahnawà:ke community. One day a young man came to the door, asking her
mother whether her daughter was at home, using Feminine-Indefinite forms. The
mother was perplexed, having no idea who he could be asking for. The daughter
he was interested in, the firstborn, had always been referred to with Feminine-
Zoic forms. Her younger sisters, referred to with Feminine-Indefinite forms, were
much too young to be courted.
The fi/fz distinction is thus intriguing but far from straightforward. The fre-
quently-cited features of age, respect, and familiarity first evoke European tu/vous
distinctions, but with closer scrutiny, it becomes clear that the distinctions are not
isomorphic. The Mohawk distinction is of course marked in third-person rather
than second. It also includes features not commonly associated with the European
categories, such as daintiness versus brashness. Feminine-Indefinite forms signal
endearment toward family members, but Feminine-Zoic forms signal familiar-
ity among close friends. The Feminine-Indefinite forms are said to mark respect,
but it is the Feminine-Zoic forms that are used for people one does not know well.
Nevertheless, like the European distinction, and more complex systems like that of
Lao described by Enfield (2007: 77–84), usage is a social and cultural matter, one
subject to subtle variation across communities, families, and individuals.

6 The real story

The associations between language and culture suggested by use of one category for
women and generics on the one hand, and a separate category for some women and
animals on the other, are not as direct as might be assumed. In fact the Feminine-
Indefinite category did not emerge from a view of women as quintessential or proto-
typical human beings, and the Feminine-Zoic category did not develop from a view
of women as akin to animals. The relationship between grammar and culture here is
more complex. This can be seen by examining the successive stages of development
of the two grammatical categories over time. Though there are no ancient written
records of the Iroquoian languages in which these stages are directly attested, paths
of development over time can be traced both from comparisons of the systems of
modern related languages and from internal clues within each language.
 Gender and culture       141

6.1 Comparative evidence

Relationships among the modern Iroquoian languages for which we have sub-
stantial documentation are shown schematically in Figure 1.

Proto-Iroquoian

Southern Iroquoian Northern Iroquoian

Five Nations Iroquois

Cherokee Tuscarora Huron Wyandot Seneca Cayuga Onondaga Oneida Mohawk

Figure 1: The Iroquoian Language Family

The family first separated into two major branches, Southern Iroquoian and
Northern Iroquoian, several thousand years ago. The only known member of the
Southern branch is Cherokee, spoken at first contact with Europeans in the 17th
century in the North American Southeast, over parts of what is now North Car-
olina, Virginia, South Carolina, Georgia, and Alabama. The next split occurred
within the Northern branch, when the ancestors of modern Tuscarora speakers
separated from the main group. The Tuscarora lived until the early 18th century
in what is now eastern North Carolina. After the Tuscarora Wars of 1711–1713,
they began moving north, eventually settling near the Five Nations groups in
western New York and southern Ontario. The Huron language was spoken in
the 17th century in what is now Ontario. After epidemics and wars with the Five
Nations Iroquois between 1647 and 1649, they were scattered. Some joined other
Iroquoian bands in the area and moved west and then south, ultimately settling
in Oklahoma, where they are now known as the Wyandot. Others, now known
as the Wendat, moved east toward their present location near Quebec City. The
Five Nations Iroquois groups lived, at the time of first contacts with Europeans
in the early 17th century, in communities extending across what is now New York
State. Their villages were located in the configuration shown in Figure 1, with the
Seneca at the western edge and the Mohawk at the eastern edge.
It is possible to reconstruct stages in development of the gender categories
by comparing the modern languages. Traits found in both branches of the family,
Southern Iroquoian (Cherokee) and Northern Iroquoian, can be hypothesized
142       Marianne Mithun

to have been present in their common ancestor, Proto-Iroquoian. On this basis

it appears that Proto-Iroquoian had one basic third person category with no
gender distinction, and one Indefinite category used only for generic and indefi-
nite humans. Number distinctions were made for all animates in the basic third
person category. This system has survived intact in modern Cherokee, where
basic third person prefixes are cognate with Mohawk ka- and o-, and the Indefi-
nite prefix is cognate with Mohawk ie-. Traces of this system remain in all of the
other languages.
Proto-Northern-Iroquoian shows the introduction of a new Masculine cate-
gory, present in all of the modern Northern Iroquoian languages (Chafe 1977). The
original basic third person category shrank accordingly to apply to just objects,
animals, and female persons. The original Indefinite category was still used only
for generics and indefinite persons. This system is preserved in modern Wendat
(Huron) and Wyandot.
Proto-Five-Nations shows the beginning of another change, still in progress
in some of the languages. Speakers began using Indefinite forms not just as gener-
ics and indefinites, but also as a sign of respect for certain female persons. This
change further diminished the range of the original basic third person markers
to just objects, animals, and other female persons. It is thus a residual category,
what is left of the original basic third person category after the removal of males
to the new Masculine category and some females to the Indefinite category. This
is essentially the situation found in modern Onondaga, Oneida, and Mohawk.
Chafe (p.c.) reports that in Onondaga, the Feminine-Indefinite category can signal
empathy, while the Feminine-Zoic can signal detachment. Abbott (1984) provides
a detailed description of the Oneida system. It largely parallels the Mohawk one,
with comparable variation across dialects and speakers, but it also shows subtle
differences. In his 1953 grammatical sketch of Oneida, Lounsbury reported that
age is the principal basis for the distinction, with a tendency among speakers to
restrict Feminine-Zoic forms to ‘adult, active, female persons’ (1953: 51). Abbott
noted that several decades later, some speakers, mainly male, concurred: Oneida
Feminine-Zoic forms signal greater age than Feminine-Indefinite forms, “with the
dividing line somewhat vaguely around adolescence” (1984: 128). He also found,
however that other speakers strongly disagreed about the overriding importance
of age. As in Mohawk, close relatives tend to be referred to with Feminine-Indef-
inite forms, but close friends with Feminine-Zoic forms. Abbott concludes that
the choice seems to ‘be based on age, size, gracefulness, kin relation, or even
empathy’ (1984: 128).
The extension of Indefinite forms to women has proceeded further in the
western Five Nations languages. In modern Seneca and Cayuga, they are now
used for all female persons. It is difficult to know precisely when these last devel-
 Gender and culture       143

opments occurred, and the extent to which they were driven by language contact.
The Five Nations groups constituted a strong political alliance, the League of
the Iroquois, and it is known that the members of the bands did interact and
intermarry to some extent. After the Tuscarora arrived in the area, they too were
brought into the League. Modern Tuscarora shows the same system as its immedi-
ate Seneca neighbors: reflexes of the original Indefinite forms are now used for
all female persons.

6.2 Internal evidence

There is also internal evidence within each of the Northern Iroquoian languages
of the extension of original Indefinite forms to female persons.
One indication is the form of traditional personal names. Naming practices
follow ancient, pre-contact traditions. Names were the property of clans, tradi-
tionally bestowed by the clan mother or some other knowledgeable elder. Often
a baby was given the name of a deceased relative of the same clan, but only one
person at a time could bear a traditional name. The names are old, so not all are
still interpretable, but most are still transparent. Some examples of traditional
Mohawk names are in (15).

(15) Some traditional Mohawk women’s names

Ka-ia’tanó:ron ‘She is body-precious’
Ka-nerahtenhá:wi ‘She is carrying leaves’
Ka-heráwaks ‘She shakes the cornstalk’

What is significant is that in all of the Five Nations communities and among the
Tuscarora as well, traditional names for women may contain Zoic or Neuter prefixes
(Ka-, W-), but never Indefinite prefixes, no matter who the bearer. This is true even
in communities where all women are now referred to with Feminine-Indefinite
forms. The names predate the extension of the Indefinite gender forms to women.
A second indication of the direction of development of the gender categories
involves kinship terms. The kinship terms are based on morphological verbs. As
noted earlier, verbal prefixes may be intransitive or transitive. They distinguish
three persons, three numbers, three genders, and two roles, grammatical Agent
and Patient. Where the relationship is reciprocal, as for cousins, intransitive
reciprocal forms are used with dual or plural prefixes (‘we two are cousins to each
other’ = ‘my cousin’). Where the relationship is asymmetrical, transitive prefixes
are used, with the senior kinsman marked by forms similar to grammatical Agents
in regular verbs, and the junior kinsmen marked like grammatical Patients.
144       Marianne Mithun

(16) Transitive verbs and kinterms

a. Verb
rák-ken-hs
3m.sg/1sg-see-hab
‘he sees me.’

b. Kinterm
rak-hsót=ha
3m.sg/1sg-be.grandparent.to=dim
‘he is grandparent to me’ = ‘my grandfather’

Most kinterms for female relatives have both Feminine-Indefinite and Feminine-
Zoic forms.

(17) ‘My younger sister’

a. khe-’kèn:’=a
1sg/3fi-be.older.sibling.to=dim
‘I am older sibling to her (fi)’ = ‘my younger sister’ fi

b. ke-’kèn:’=a
1sg/3fz.sg-be.older.sibling.to=dim
‘I am older sibling to her (fz)’ = ‘my younger sister’ fz

But not all kinship terms for women have two forms. For grandmothers and
mothers, there are only Zoic forms.

(18) Feminine-Zoic only

a. ak-hsót=ha
3fz.sg/1sg-be.grandparent.to=dim
‘she (fz) is grandparent to me’ = ‘my grandmother’ fz

b. ake-’nistén=ha
3fz.sg/1sg-be.mother.to=dim
‘she (fz) is mother to me’ = ‘my mother’ fz

What is surprising is that these are the very individuals speakers unanimously
agree should be referred to with Feminine-Indefinite verbal prefixes and indepen-
dent pronouns. The gaps begin to make sense when it is recognized that the use
of Indefinite forms as a sign of respect for elders is an innovation. Words for ‘my
grandmother’ and ‘my mother’ are among the earliest kinship terms learned by
children and the most frequently used. They are thus the most resistant to change.
 Gender and culture       145

The extension of Indefinite forms to certain women is only gradually making

its way through the regular verbal and kinship prefix paradigms. It is more
advanced in the verbal paradigms, but certainly not complete. There are, for
example, no separate Feminine-Indefinite dual or plural verb prefixes. Forms
used for referring to more than one woman are descended from the original basic
third person forms. The kinship prefix paradigms are even less advanced. The
progress of the innovation can be traced in Table 1. For each relationship (‘my
grandmother’, ‘your grandmother’, etc.) the table shows whether the innovative
Feminine-Indefinite form yet exists (fi) or not (–), and whether the original Fem-
inine-Zoic form still persists (fz) or not (–).

Table 1: Gradual extension of Indefinite forms to female kinsmen

‘my’ ‘your’ ‘her’ ‘his’

‘grandmother’ fz – fz – fz – fz –
‘mother’ fz – fz – fz – fz –
‘older sister’ fz – fz – fz fi fz fi
‘younger sister’ fz fi fz fi fz fi fz fi
‘daughter’ fz fi fz fi fz fi fz fi
‘stepdaughter’ fz fi fz fi fz fi fz fi
‘sister-in-law’ fz fi fz fi fz fi fz fi
‘daughter-in-law’ fz fi fz fi fz fi fz fi
‘niece’ fz fi fz fi fz fi fz fi
‘granddaughter’ fz fi fz fi fz fi fz fi
‘stepmother’ – fi – fi – fi fz fi
‘mother-in-law’ – fi – fi – fi – fi

The earliest learned, most frequently-used terms have been the most resistant
to change. The innovation has not yet reached any of the terms for grandmoth-
ers and mothers, nor those for ‘my older sister’ or ‘your older sister’. There are
now, however, Indefinite forms for ‘her older sister’ or ‘his older sister’. On the
other hand, for the terms which are probably the last to be learned and used,
the terms denoting those relationships in which it is most important to show
respect, the Feminine-Zoic forms have disappeared entirely. There are only
Indefinite forms for all mothers-in-law as well as ‘my stepmother’, ‘your step-
mother’, and ‘her stepmother’. A Feminine-Zoic term still persists for ‘his step-
mother’. (Additional developments in the pronominal system are discussed in
Mithun 2010.)
146       Marianne Mithun

7 Agreement reconsidered
The difference in rates of development between the regular verbal prefix para-
digms and the kinship prefix paradigms has lead to prefix mismatches within
clauses. Grandmothers and mothers, for example, can be referred to only with
Feminine-Zoic kinship terms, but only with Feminine-Indefinite verb prefixes.

(19) Grandmother mismatches: Josephine Kaieríthon Horne, speaker p.c.

a. Akhsótha enienenhstóhare’
ak-hsot=ha en-ie-nenhst-ohar-e’
3fz/1sg-be.grandparent.to=dim fut-fi.agt-corn-wash-pfv
my grandmother (fz) she (fi) will corn wash

tánon’
tanon’
and
and
‘My grandmother (fz) would wash (fi) the corn and

b. eniethe’serón:ni’ ohén:ton
en-ie-the’ser-onni-’ o-hent-on
fut-3fi.agt-meal-make-pfv n-lead-stative
she (fi) will grind before
grind (fi) it before

tsi niió:re’
tsi ni-io-r-e’
so prt-3n.pat-extend-stative
so it is far

c. kana’tarokhón:we
ka-na’tar-o-k=honwe
n-bread-be.in.water-cont=genuine
traditional cornbread
she (fi) made cornbread.’

eniena’tarísa’.
en-ie-na’tar-is’a-’
fut-3fi.agt-bread-finish-pfv
she (fi) will bread finish
 Gender and culture       147

It could be argued that this is typical of agreement systems in general. In

many languages, the gender markers that appear on different lexical items, such
as verbs, nouns, adjectives, demonstratives etc. need not have identical forms;
they must simply encode the same categories. But these Mohawk mismatches
cannot be generalized by lexical category. Other kinship terms for women have
both Zoic and Indefinite forms, and the prefix categories on these terms normally
match those on associated verbs.

(20) Zoic daughter: Sha’tekenhátie’ Marian Phillips, speaker

Tanon’ konwaièn:’a skaià:ta
tanon’ konwa-ien’=ah s-ka-ia’t-at
and 3fi/3fz.sg-have.as.offspring rep-3fz.sg-body-be.one
and she (fi) has her(fz) as child one (fz) body

Kahentinéhson’
ka-hent-ine-hson’
3fz.sg.agt-field-lead-distr
she (fz) leads the field
‘And her (fi) one (fz) daughter (fz), Kahentinéhson’ (fz)

è:neken’ nenkwá: she’s kanákere’.

è:neken’ nenkwa she’s ka-naker-e’
up that.side formerly 3fz.sg.agt-reside-stative
up that side formerly she (fz) resides
used to live (fz) up there.’

The terms for ‘grandmother’ and ‘mother’ could still be accommodated in various
ways theoretically. They could be listed as morphologically Zoic but syntacti-
cally Indefinite, for example. They could be likened to the Latin agricola ‘farmer’,
which appears to have a Feminine ending formally but is categorized as Mascu-
line for purposes of agreement.
These are not the only cases of mismatches. It was seen earlier that Zoic pre-
fixes on verbs distinguish number, while the usually similar Neuter prefixes do not.

(21) Number distinctions

a. Zoic verbal prefixes
ka- sg.agent io- sg.patient
keni- du.agent
konti- pl.agent ioti- duoplural patient
148       Marianne Mithun

b. Neuter verbal prefixes

ka- neuter.agent (sg/du/pl) io- neuter.patient (sg/du/pl)

(All of the prefixes exhibit considerable allomorphy.)

The prefixes on Zoic nouns, however, do not usually distinguish number. The
result is again mismatches between the nouns and verbs that are supposed to
agree.

(22) Bug mismatch? Dorothy Karihwénhawe Lazore, speaker p.c.

Kaniataratátie’ enkontohétstake’
ka-niatar-at-atie’ en-konti-ohetsts-ake’
n-river-extend-prog fut-3z.pl.agt-pass-cont
along the river they will pass by

katsi’niowá:nen’s (…)
ka-tsi’non-owan-en-’s
n-bug-be.large-stative-distr
it is bug large variously
‘Insects (neuter) will pass by (zoic.pl) along the river (that you
have never seen),

ok ò:ni’ ne karonhia’kéhshon’ enkontítie’,

ok ohni’ ne ka-ronhi-a’ke=hshon en-kont-itie-’
and also the n-sky-place=distr fut-3z.pl.agt-fly-pfv
and also the through the skies they will fly
and they will also fly (zoic.pl) through the skies

seken ne katsi’niowá:nen’s.
seken ne ka-tsi’non-owan-en-’s
also the 3n-bug-be.large-stative-distr
also the it is bug large variously
as well, those insects (neuter).’

This example reflects a common Mohawk pattern: many terms for animals
contain a Neuter prefix that does not distinguish number, but they co-occur
with verbs containing Zoic prefixes that do. The term for ‘cow(s)’ is literally ‘it
jowl protrudes doubly’, no matter how many cows are under discussion. The
form of the noun prefix does not change with number, but the form of the verb
prefix does.
 Gender and culture       149

(23) Cow mismatch: Rita Konwatsi’tsaién:ni Phillips, speaker p.c.

Ísi’ nónhskwati she’s niió:re’
yonder bridge beyond formerly so it is far

niekonnéhtha’ ne
n-ie-konn-e-ht-ha’ ne
prt-trloc-3z.pl.agt-go.with-hab the
there they (zoic.pl) habitually go the

teionhónhskwaron.
te-io-nhonhskwar-ont
dv-3n.pat-jowl-be.attached
it (neuter) doubly jowl protrudes
‘The cows used to pasture way over on the other side of the bridge.’

(The Duplicative prefix te- (dv) here reflects the fact that two jowls are involved.)
An obvious explanation might be that number is not marked in the noun
prefix morphology but it is marked in the verb prefix morphology. Singular, dual,
and plural Zoic prefixes on nouns have the same form, but those on verbs are
distinct. This hypothesis soon runs into exceptions, however. The term for ‘sheep’
has a Zoic plural prefix, no matter how many sheep are involved.

(24) Plural prefix on noun ‘sheep’

teiotina’karontòn:’a
te-ioti-na’kar-ont=on’a
dv-3z.pl.pat-horn-be.attached=distr
‘they are doubly horn attached’ = ‘sheep’ (any number)

The number of sheep under discussion is specified only in associated predicates

elsewhere in the clause. It is possible to specify that there is just one sheep, but in a
larger phrase. Even here the noun for sheep still contains the Zoic plural prefix ioti-.

(25) ‘one sheep’

skaià:ta teiotina’karontòn:’a
s-ka-ia’t-at te-ioti-na’kar-ont-on’a
rep-neuter-body-be.one dv-3z.pl.pat-horn-be.attached=distr
one body sheep

It is interesting to note that in the eastern Mohawk communities, where the

first language contact was with French, there is an alternate term for ‘sheep’:
timotón. This word is a transparent copy of a French term, but it is the plural
150       Marianne Mithun

form des moutons. Sheep are animals that usually appear in groups (flocks).
The unmarked number for sheep is plural. Such cases could be likened to the
‘unproblematic mismatches’ discussed by Corbett in his work on agreement,
mismatches ‘which arise because the controller lacks forms which would guar-
antee matching’ (2006: 144).
These mismatches between noun prefixes and the number of referents actu-
ally suggest a more general principle, that gender is not a syntactically active
property of nouns synchronically. In the account of the encounter with the polar
bear seen earlier, the noun for ‘bear’ retained its Neuter form o-hkwá:ri, whether
it was categorized in the verbal prefix as Zoic or Masculine.

(26) Masculine polar bear: Josephine Kaieríthon Horne, speaker p.c.

Wa’thá:ta’ne’ kí:ken ohkwá:ri.
wa’-t-ha-ta-’n-e’ kiken o-ahkwari
fact-dv-3m.sg.agt-stand-inch-pfv this neuter-bear
he stood up this bear
‘The bear (he) stood up.’

(27) Zoic polar bear: Josephine Kaieríthon Horne, speaker p.c.

Óksa’k sahiathrorià:na’
oksa’k sa-hi-at-hrori-ahna-’
immediately rep-3m.du.agt-tell-middle-andative-pfv
immediately they went back to tell
‘They quickly returned to camp to report

tsi wa’konwário’ ohkwá:ri.

tsi wa’-konwa-rio-’ o-ahkwari
that fac-3dp/3z.sg-kill-pfv neuter-bear
that they killed it zoic bear
that they had killed (it) a bear.’

It should be noted that this account was related by an excellent speaker. Other
speakers agree that these are appropriate forms.
Similarly, in the ceremonial speeches seen earlier where otherwise inanimate
objects are personified, it is only the verbal prefixes that show a category shift.
 Gender and culture       151

(28) Oratory: Frank Tekaronhió:ken Jacobs Jr., speaker

Ietshiiatahónhsatat ne ononhkwa’shòn:’a,
ietshii-at-ahonts-atat ne o-nonhkw-a’=shon’a
2pl/3hu.pl-middle-ear-prick the neuter-medicine-ns=distr
you all listen to them the various medicines
‘Listen to the medicines (neuter)

tsi nahò:ten’ rón:ton.

tsi n-a-h-o’ten-’ ron-aton
at prt-fact-n.-be.a.kind.of -pfv 3m.pl.agt-say
as such it is a kind they (m.pl) are saying
to what they (m.pl) are saying.’

It should be noted that the prefix o- on ‘medicines’ has approximately the same
form as the Neuter Patient prefix that appears on verbs, but the medicines are not
functioning as grammatical Agents here, as can be seen in the prefix ‘they’ on the
verb ‘they are saying’. The noun ‘medicine’ always begins with the prefix o- no
matter what its grammatical role in the clause.
Zoic forms differ from general Neuters not only in showing number distinc-
tions, but also in certain transitive prefix combinations. Basic neuters are not
represented unless they are the only core argument of the clause. The Masculine
plural agent prefix in intransitive verbs, for example is rati-/-hati‑, as in: wa-hati-
ká:we’ ‘they paddled’. The same form is used in transitives with Neuter patients
‘they/it or them’. Different forms are used in transitives with Zoic patients:
konwa- ‘they/ it or her’ and konwati- ‘they/them’.

(29) Zoics and Neuters

a. Zoic (animate) patients
Sa-konwa-ia’tisákha’.
‘They went back out to look for it (a cow).’

Sa-konwati-ia’tisákha’.
‘They went back out to look for them (cows).’

b. Neuter (inanimate) patients

Sahatisákha’. ‘They went back out to look for (it/them).’

But the agreement patterns are not always as might be expected. The choice
of Zoic or Neuter gender prefixes in the verb does not always depend on the
referent. The sentences in the next four examples come from an account of the
history of the old church bell at Kahnawà:ke, Quebec. The bell was sent by ship
152       Marianne Mithun

at the beginning of the 18th century as a gift from the King of France. It was
wartime, however, and the ship was captured and taken to Salem, Massachu-
setts. The cargo was sold, and the bell was purchased by the town of Deerfield.
When news of the sale reached the Mohawks, they set out for Deerfield. They
managed to find the bell, take it down from the church steeple there, and fasten
it to a beam to carry it back home. It was wintertime, however, and after strug-
gling some way through the deep snow, they decided to bury the bell and return
for it in the spring.
In the first clause in sentence (30), the ship was categorized in the verbal
prefix konwa- as Zoic. This is perhaps not surprising; many English speakers
refer to ships as ‘she’. But this same ship was categorized by the same speaker as
Neuter in the second clause of the same sentence.

(30) Zoic and Neuter ship: Josephine Kaieríthon Horne, speaker p.c.
Wa’konwaié:na’ ne kahonweia’kó:wa,
wa’-konwa-iena-’ ne ka-honweia’=kowa
fact-3pl/3z-catch-pfv the n-boat=aug
they caught it (zoic) the ship (neuter)
‘They captured it (zoic) the ship

iahatíhawe’ wastonhronòn:ke
i-a-hati-haw-e’ waston=hronon’=ke
trloc-fact-3m.pl.agt-take-pfv Boston=resident=place
they took it (neuter) United States
and took it (neuter) to the United States.’

Perhaps even more surprising is the categorization of the town of Deerfield as

Zoic in (31).

(31) Zoic town: Josephine Kaieríthon Horne, speaker p.c.

Deerfield tkonwá:iats, tsi nón:
Deerfield t-konwa-iat-s tsi ne=onwe
Deerfield cisloc-3fi/3z.sg-call-hab at the=place
Deerfield they call it (zoic) at the place

tioia’totarhè:’on.
t-io-ia’t-otarhe’-on
cisloc-3n.pat-bodily-hook-st
there it (neuter) was seized
‘It was seized at a place called (zoic) Deerfield.’
 Gender and culture       153

In (32) the bell was Neuter, as would be expected.

(32) Neuter bell: Josephine Kaieríthon Horne, speaker p.c.

Karontà:ke wahatihwánerenke’.
ka-ront-=a’ke wa-hati-hwanerenk-e’
neuter-log=place.at fact-3m.pl.agt-tie-pfv
log place they tied it (neuter)
‘They fastened it (neuter) to a beam.

Wahonnenhsà:ren’ …
wa-hon-nenhs-a-hren-’
fact-3m.pl.agt-shoulder-lk-set.on-pfv
They carried it (neuter) on their shoulders … ’

But in (33), this same bell was referred to as Zoic.

(33) Zoic bell: Josephine Kaieríthon Horne, speaker p.c.

Enkonwaia’táta’ ne iehwista’ékstha’.
en-konwa-ia’t-at’a-’ ne ie-hwist-a-’ek-hst-ha’
fut-3.pl/3z.sg-body-insert-p the 3fi.agt-metal-lk-hit-ins-hab
they will bodily insert it (zoic) the one metal strikes with it
‘[They decided] they would bury (it zoic) the bell.’

This is not a matter of stylistic options. Speakers agree that these prefixes are the
only acceptable ones in these sentences.
Similar alternations can be seen in the verbal prefixes associated with corn.
In (34) corn is Neuter, as would be expected.
154       Marianne Mithun

(34) Corn: Josephine Kaieríthon Horne, speaker p.c.

Nó:nen akwé: iostáthen
ne-onen akwék-on io-stath-en
the-when be.all-stative 3n.pat-be.dry-stative
when all it (neuter) is dry

nó:nenhste’, sók
ne=o-nenhst-e’ sok
the-neuter-corn-ns then
the corn then
‘When the corn was completely dry (neuter),

nòn:wa’ entsakwanenhstarón:ko’
ne=onhwa’ en-ts-iakwa-nenhst-a-ron-ko-’
the=now fut-rep-1excl.pl.agt-corn-lk-be.attached-rev-pfv
now we will corn remove
we would take the kernels off the cob.’

But in (35) the corn is referred to as Zoic. This speaker was commenting on what
she had noticed during a drive through the surrounding area.

(35) Corn: Watshenní:ne Sawyer, speaker p.c.

Ó:nenhste’ ken’k ni-konti-hnenié:son’s
o-nenhst-e’ ken’=ok ni-konti-hneni-es-on’s
n-corn-ns small=just prt-3z.pl.agt-length-be.long-distr
corn small so they (zoic.pl) are long variously
‘The corn are (zoic) very short.

tsi ní:io nakwé:

tsi ni-io-ht ne=akwek-on
as prt-n.pat-be.so the=all-stative
as so it is the all

thie-iotí:-ten.
th-ie-ioti-ten
contr-toc-3z.pl.pat-be.poor
they (zoic.pl) are just poor there
They all seem to be doing poorly (zoic).’

Other speakers concur that only Zoic prefixes could be used here: “‘No one would
ever say Ken’k niión:son’s for ‘it is short’ referring to corn.” (Kanerahtenhá:wi
Nicholas, p.c.).
 Gender and culture       155

The somewhat surprising choices of gender prefixes appear to be controlled

by the verb stem. Some additional examples of unexpected verb prefixes are in
(36). Verbs for growing and getting frostbitten require Zoic prefixes, which distin-
guish number. The verb for getting ripe requires a Neuter prefix, with no number
distinction. Speakers have no choices here.

(36) Some additional observed verbs

ioti-ká:ion ‘they are growing slowly’ (corn) pl
ioti-kenhé:ion ‘they are slow growing’ (corn, berries, potatoes, grass) pl
ioti-hniò:’on ‘they just grew’ (volunteer trees in the woods) pl
ioti-iénhton ‘they got frosted’ (berries) pl
ion-atonníson ‘they have finished growing, they’re ripe’ (potatoes) pl
io-hiá:ri ‘they are ripe’ (blackberries) sg

Relatively few verb stems can be used with either animate or inanimate argu-
ments. It appears that verbs for growing, catching, burying, and having a proper
name require grammatically animate patients, that is, they routinely occur with
Zoic Patient prefixes. Prototypical scenarios have been routinized to grammatical
requirements.
The prefixes on many nominals do not signal gender directly. The original
number markers have been frozen in the lexicalized nominals. Basic morpho-
logical nouns consist minimally of three parts: a noun prefix, a noun stem, and
a noun suffix. Nouns beginning with o- (or its allomorphs) are usually Neuter.

(37) Morphological noun

ó:nenhste’
o-nenhst-e’
neuter-corn-noun.suffix
‘corn’

A very large number of lexicalized referring expressions are actually morphologi-

cal verbs, without overt nominalizers. Some were seen earlier. Their prefixes may
match the gender category of their referents, as in the case of ‘nun’, or not, as in
the case of ‘pencil’, literally ‘one writes with it’.
156       Marianne Mithun

(38) Many syntactic nominals (referring expressions) are morphological

verbs.
a. ioia’tatokénhti
io-ia’t-a-tokenht-i
zoic.sg.pat-body-lk-be.holy-stative
‘she is body holy’ = ‘nun’

b. iehiatónhkhwa’
ie-hiaton-hkw-ha’
indefinite.agt-write-ins-hab
‘one writes with it’ = ‘pen, pencil’

There are in fact very few true morphological human nouns. The terms for ‘boy’
and ‘girl’ seen earlier, and for ‘man’ and ‘woman’, are ancient, but they are based
on the verb roots -ksa ‘be a child’ and -onkwe- ‘be a person’.

(39) Common nominals

raksà:’a rón:kwe
ra-ksa=’a r-onkwe
m.sg.agt-be.a.child=dim m.sg.agt-be.a.person
‘boy’ ‘man’

ieksà:’a iakón:kwe
ie-ksa=’a iak-onkwe
fi.agt-be.a.child=dim fi.agt-be.a.person
‘girl’ ‘woman’

These roots occur most often in the referring expressions above, but there is
ample evidence of their status as verb roots. The sets of noun and verb stems are
completely distinct in Mohawk. Only noun stems can serve as the foundation of
morphological nouns, and only verb stems can serve as the foundation of mor-
phological verbs. Mohawk shows extensive noun incorporation, whereby a noun
stem is compounded with a verb stem to form a new verb stem. Only morphologi-
cal noun stems can be incorporated. The verb roots -ksa ‘be a child’ and -onkwe
‘be a person’ must be overtly nominalized before they can be incorporated.

(40) Nominalized roots

a. ra-ksa-’t-í:io
m.sg.agt-be.a.child-nominalizer-be.good
‘He is a good boy.’
 Gender and culture       157

b. r-onkwe-’t-í:io
m.sg.agt-be.a.person-nominalizer-be.good
‘He is a good man.’

The morphological structures of noun and verb words are also completely dis-
tinct. The two verb roots ‘be a child’ and ‘be a person’ participate in verbal mor-
phological constructions, such as the Coincident.

(41) Coincident construction

Shikeksà:’a, …
shi-ke-ksa=’a
coincident-1sg.agt-be.a.child=dim
‘When I was a child …’

They are negated like other verbs as in (42), rather than like nouns as in (43).

(42) Clausal and verbal negation

Iáh teskaksà:’a.
iah te-s-ka-ksa=’a
not neg-rep-fz.sg.agt-be.a.child=dim
‘She is no longer a child.’

(43) Nominal negation

Iáh ó:nenhste’ tè:ken.
iah o-nenhst-e’ te’-ka-i
not n-corn-ns neg-3n.agt-be
not corn is it
‘It is not corn.’

Other terms for persons are also morphological verbs, such as akokstèn:ha
‘elderly lady’ and rokstèn:ha ‘elderly man’ based on the verb root -ksten ‘be old’;
nitiakoiòn:ha ‘young woman’, and nithoiòn:ha ‘young man’, based on another
root -akaion ‘be old’; ieià:tase’ ‘young lady’ (literally ‘she is new bodied’), and
ranekénhteron ‘young man’, a stative verb based on the verb stem -nekenhter- ‘be
a young man’. The same verb stem is used as a syntactic predicate: kenekénhteron
‘I am a young man’. The word rohsken’rakéhte’ ‘warrior, post-adolescent man’, is
literally ‘he carries a weapon (‘he rust-carries’). The word rahsennowá:nen ‘chief’
is a verb meaning literally ‘his name is great’. The word ratsiénhaienhs ‘council-
lor’ is literally ‘he lays the (council) fire’. As was seen earlier, kinship terms are
based on morphological verbs. Terms that characterize people in terms of their
158       Marianne Mithun

activities or professions are morphological verbs, such as rató:rats ‘hunter’, lit-

erally ‘he hunts’, or ra’swáhtha’ ‘fireman’, literally ‘he extinguishes’. One term,
owirà:’a ‘baby’, is a morphological noun based on the noun root ‘offspring’, but it
is marked grammatically as Neuter with its prefix o-.
The prefixes on referring expressions do not necessarily classify the refer-
ents of those terms directly. Many are simply relics of the original verbal con-
structions that were their sources. In some cases the sources are still very much
alive and alternatives exist: ra’swáhtha’ ‘fireman’, ron’swáhtha’ ‘firemen’ (‘they
extinguish’). In others, the terms have simply become labels, even if speakers can
retrieve the sources: teionhónhskwaron ‘cow(s)’.

8 Conclusions
The material seen in the preceding sections illustrates several points about rela-
tions between grammatical gender systems and culture.
The first is that such grammatical systems can be more intricate and inter-
esting than might first appear on the basis of elicited sentences. The sets of con-
structed Mohawk sentences seen at the outset would, at first, appear to show a
typologically prototypical, even European-type gender system, with Masculine,
Feminine, and Neuter gender categories marked on nouns and reflected in agree-
ment prefixes on verbs. The Masculine prefixes are used for male persons, the
Feminine prefixes for female persons, and the Neuter prefixes for objects and
animals. Examination of spontaneous speech quickly shows that the categoriza-
tion is more complex, however. Within the Neuter category a formal distinction
is made between inanimates and animates, with number distinguished only for
animates. Such a situation is not uncommon cross-linguistically. Mohawk Mas-
culine and Feminine prefixes are sometimes used for personified objects and
animals, again not a rare phenomenon. Somewhat surprising is the fact that the
Feminine forms used for certain female persons are also used to refer to unidenti-
fied persons or to people in general. Perhaps even more surprising is the fact that
the same Neuter (Zoic) forms used for animals are also used for referring to other
female persons. The factors governing choices between these two categories,
termed Feminine-Indefinite and Feminine-Zoic, are complex and variable across
communities, families, and individuals. These gender categories are not isomor-
phic with others found elsewhere in the world, and the factors underlying their
distribution do not match criteria for choice seen among categories elsewhere
either. Like gender systems in many other languages, however, they are subject to
variation across speakers and communities.
 Gender and culture       159

Second, the existence of agreement has often been cited as a fundamental

property of grammatical gender systems, and the first set of Mohawk examples
would appear to illustrate exactly this property. The nouns and associated verbs
in those examples contain similar sets of prefixes reflecting gender and number.
A closer look at spontaneous speech shows, however, that standard notions of
agreement do not fully capture the relationships between the forms that actually
occur. The prefixes on nouns may match the coreferential pronominal prefixes
on verbs in gender and number, but they often do not. The noun prefixes are not
synchronically active. They are an artifact of their diachronic sources and remain
unchanged no matter what their referent or syntactic role in the clause. The forms
of the noun prefixes may or may not match the referents of the nouns. Further-
more, the verbal prefixes are actually fully referential bound pronominal forms,
which reflect features of their referents, rather than noun classes.
Finally, a reconstruction of the development of the Mohawk gender system
shows that though grammatical gender categories may indeed reflect culture, the
relationship is not necessarily immediate and direct. The fact that the same gender
category is used for women and generic reference to human beings is not a result
of a conception of women as prototypical humans. It is, instead, the result of the
use of original indefinite forms to refer to certain women as a sign of respect. That
extension of the original indefinite category to certain female persons does of
course reflect an aspect of culture. The fact that the same gender category is used
for animals and certain other women is not a result of the conception of women as
akin to animals. That gender category is a residual category, one that originated
as a basic third person category undifferentiated for gender but was narrowed in
successive steps, first with the introduction of a separate Masculine category for
males, and then with the use of the original generic category for certain females.
Relations do indeed exist between grammatical gender categories and
culture, but these relations are often more complex and interesting than might
be assumed.

9 References
Abbott, Clifford. 1984. Two feminine genders in Oneida. Anthropological Linguistics 26(2).
125–137.
Chafe, Wallace. 1977. The evolution of third person verb agreement in the Iroquoian languages.
In Charles N. Li (ed.), Mechanisms of Syntactic Change. 493–524. Austin: University of
Texas Press.
Corbett, Greville. 1991. Gender. Cambridge, UK: Cambridge University Press.
Corbett, Greville. 2006. Agreement. Cambridge, UK: Cambridge University Press.
160       Marianne Mithun

Enfield, Nick. J. 2007. A Grammar of Lao. (Mouton Grammar Library.) Berlin: Mouton de Gruyter.
Hockett, Charles F. 1958. A Course in Modern Linguistics. New York: Macmillan.
Lounsbury, Floyd. 1953. Oneida Verb Morphology. New Haven: Yale University Press.
Mithun, Marianne. 2010. The search for regularity in irregularity: Defectiveness and its
implications for our knowledge of words. In Matthew Baerman, Greville Corbett, and
Dunstan Brown (eds.), Defective paradigms: Missing forms and what they tell us. Oxford,
UK: British Academy and Oxford University Press. 125–149.
Niels O. Schiller
Psycholinguistic approaches to the
investigation of grammatical gender
1 P
 sycholinguistic approaches to the
­investigation of grammatical gender in speech
production: An overview and new data
In some languages, the selection of grammatical features, such as number, case,
or gender (Corbett 1991) is mandatory to establish grammatical agreement in
noun phrase (NP) production. Grammatical gender is a morphosyntactic feature
that occurs in many languages of the world (see Corbett, this volume). It has been
shown in the past that grammatical gender may impact linguistic processing in
gender-marking languages. For instance, in Germanic languages like Dutch and
German, native speakers are faster to name an object in the presence of a distrac-
tor word with the same gender as the name of the object than when the distractor
has a different gender (Finocchiaro et  al. 2011; Heim et  al. 2009; La Heij et  al.
1998; Schiller and Caramazza 2003, 2006; Schiller and Costa 2006; Schriefers
1993; Schriefers and Teruel 2000; Van Berkum 1997).
Schriefers (1993) was the first to observe this so-called gender congruency
effect. He accounted for the effect in the following way: When a picture has to
be named with a determiner(Det)-adjective(Adj) NP in Dutch (e.g., het groene
boekneu ‘the green bookneu’),1 participants need to retrieve the necessary gram-
matical features of the noun to be able to produce the grammatically correct NP
because determiners and adjectives are gender-marked in Dutch and their gender-
marking has to agree with the gender of the head noun. According to Schriefers
(1993), the picture name activates its corresponding gender feature (e.g., neuter).
The gender feature has to be selected in order to activate the corresponding form
of the definite determiner (de or het in Dutch). This process may be disturbed
by the presence of a distractor word. When the distractor word has a different
gender than the target, it activates a different gender feature. The simultaneous
activation of two different gender features presumably results in competition. As
a consequence, the selection of the gender feature of the target is delayed com-
pared to the situation when both the target and distractor word activate the same

1 In this article, we will use the following abbreviations: neu for neuter, fem for feminine, mas for
masculine, and com for common gender.
162       Niels O. Schiller

gender feature, thereby boosting its activation and facilitating its selection. The
gender congruency effect in Germanic languages such as Dutch (La Heij et  al.
1998; Schiller and Caramazza 2003, 2006; Van Berkum 1997) and German (Heim
et al. 2009; Schiller and Caramazza 2003; Schiller and Costa 2006; Schriefers and
Teruel 2000) is a stable phenomenon that has been replicated many times in dif-
ferent laboratories.
However, Miozzo and Caramazza (1999) failed to replicate this effect with
Italian speakers. They noted that in Italian (as well as in other Romance lan-
guages) the selection of the determiner during the production of an NP does not
only depend on the gender of the noun but also on the immediate phonologi-
cal context. For instance, the masculine determiner in Italian is il as in il tavolo
(‘the table’) or il grande scienziato (‘the great scientist’) except when the word
following the determiner starts with a vowel, a consonant cluster of the type <s>
+ consonant or <gn>, or an affricate. If one of the latter conditions holds, the
determiner is lo as in lo strano tavolo (‘the strange table’) or lo scienziato (‘the
scientist’). Therefore, the selection of determiner forms (e.g., la, il or lo) in Italian
(as well as in other Romance languages) cannot be carried out at the level where
gender feature selection occurs but must wait instead until the phonological form
of the lexical item that follows the determiner has been selected – a fairly late
process in NP production.
Two scenarios for the absence of a gender congruency effect in Italian are pos-
sible. One possibility is that the selection of grammatical features is a competitive
process but a competition effect is masked by the late selection of the determiner
form in Italian. By the time the determiner is selected, any competition may have
been resolved. The other possibility is that grammatical feature selection is not
a competitive process, but an automatic consequence of lexical node selection.
Selection competition, on this view, is restricted to the selection of lexical forms,
including different forms of determiners. However, because determiners can only
be selected relatively late during NP production in Italian, the effects of a distrac-
tor word are not visible at the level of determiner selection in this language. This
latter hypothesis is in accordance with the results of a study by Costa et al. (1999)
who were not able to replicate the gender congruency effect in two other Romance
languages, namely Spanish and Catalan, which have similar characteristics with
respect to determiner selection as Italian. Moreover, Alario and Caramazza (2002)
failed to obtain the effect with determiners in French, another language in which
determiner selection is affected by phonological context.
In Germanic languages like Dutch or German, however, the form of a deter-
miner exclusively depends on the gender of the noun. As soon as the noun’s
gender information becomes available (along with information about “definite-
ness” and number), a determiner can be selected. We will refer to Romance and
 Psycholinguistic approaches to the investigation of grammatical gender       163

Germanic languages as “late” versus “early selection languages”, respectively,

to emphasize the observed differences concerning the point at which determiner
selection can take place.
Costa et al. (2003) found the gender congruency effect in Croatian, another
early selection language. Since Croatian is a Slavic language and does not have
determiners, they asked participants in their experiments to name objects with a
verb phrase (VP) including a gender-marked pronoun (‘I see itmas/fem’) and found
a gender congruency effect with pronouns, i.e. free-standing gender-marked
morphemes. However, they did not find a gender congruency effect when par-
ticipants were asked to produce gender-marked (possessive) Adj NPs (‘mymas/
fem <object name>mas/fem’). What could be the reason for the contrasting results?
Costa et al. (2003) considered two possibilities, both based on the fact that the
gender-marked items in the pronoun and Adj NP conditions differed in a crucial
respect: gender agreement for adjectives, but not for (clitic) pronouns, involves
the affixation of a bound morpheme. This means that the initial part of the adjec-
tive, i.e. the stem, in an Adj NP is the same for different genders (e.g., mojmas
krevetmas ‘my bed’ vs. mojafem trubafem ‘my trumpet’). One possible implication
of this fact is that since gender-suffixed variants of an adjective differ at the end
of the word, any effects of competition in the selection of affixes are not measur-
able with the currently used paradigms. Perhaps participants prepare for pro-
duction of the adjective stem and begin speaking as soon as they have enough
information about the noun, thereby masking any effects of competition that may
occur in the selection of the gender-marked suffix. The other possibility is that
the selection of morphophonological features, unlike that of free-standing mor-
phemes and stems, is not a competitive process but an automatic consequence of
specific grammatical properties.
However, before we can further consider these possibilities, we must first
confirm the reliability and generalizability of the results reported in Costa et al.
(2003) for Croatian. This is important because Schriefers (1993) found a gender
congruency effect with Adj NPs in Dutch (groenneu boekneu ‘green book’ vs.
groenecom tafelcom ‘green table’). The difference in results between the two studies
remains unresolved, and it could reflect differences between languages – Croa-
tian versus Dutch. It is important therefore to first investigate whether or not the
gender congruency effect is also found in the production of NPs in which the gen-
der-marked elements are purely inflectional features and not free-standing mor-
phemes or stems. We addressed this issue with three experiments in German and
Dutch. From earlier research (Heim et al. 2009; La Heij et al. 1998; Schiller and
Caramazza 2003, 2006; Schiller and Costa 2006; Schriefers 1993; Schriefers and
Teruel 2000; Van Berkum 1997) we know that the gender congruency effect with
free-standing morphemes is a stable phenomenon in these languages during the
164       Niels O. Schiller

production of Det (Adj) NPs. Therefore, these languages are particularly appro-
priate to test whether the selection of free-standing and bound morphemes are
subject to different processing constraints.
The method we use in all experiments of this study is the picture-word inter-
ference paradigm. In this paradigm, participants are presented with a picture
that must be named as fast as possible while ignoring a concurrently displayed
distractor word. This task is a variant of the Stroop (1935) task and it has been
widely used to investigate various aspects of speech production including lexical
access (for reviews see Glaser 1992 and MacLeod 1991). It has been shown, for
instance, that picture naming latencies are affected by specific properties of the
distractor word (Glaser and Düngelhoff 1984; Glaser and Glaser 1989; Lupker
1979, 1982; Schriefers, Meyer, and Levelt 1990; Starreveld and La Heij 1995, 1996).
In this study, we manipulated the gender of the target picture name and the dis-
tractor word.

2 E
 xperiment 1:
Indefinite Det NP production in German
In the first experiment, we tested (indefinite) determiner NPs. German distin-
guishes three grammatical genders – masculine, feminine, and neuter. Masculine
and neuter are marked by the indefinite determiner ein ‘a(n)’ (e.g., einmas Tischmas
‘a table’ or einneu Buchneu ‘a book’), the feminine form is eine (e.g., einefem Türfem
‘a door’). Participants were asked to name objects using an indefinite Det NP
while a gender-congruent or a gender-incongruent distractor word was visually
presented. Schriefers’ gender feature competition hypothesis predicts a gender
congruency effect in this situation. However, depending on whether ein/eine are
treated as morphologically simple or complex, different expectations may follow.
In the former case, we may expect a gender congruency effect because different
gender-marked morphemes compete for selection. In the latter case, however, we
may not necessarily expect such an effect because if ein/eine are derived from the
same stem ein, followed by affixation, the situation would be formally similar to
the case of moj/moja in Croatian (see above).
 Psycholinguistic approaches to the investigation of grammatical gender       165

2.1 Method

2.1.1 Participants

Twenty native German speakers participated in the study. Most of them were stu-
dents at the University of Münster in Germany. Participants were paid for their
participation in the experiment.

2.1.2 Materials

Ninety pictures corresponding to monomorphemic German nouns (30 mascu-

line, 30 feminine, and 30 neuter) were selected as targets for the experiment.
They were matched for frequency of occurrence (CELEX: Baayen, Piepenbrock
and Gulikers 1995) and word length (all F’s < 1). In addition, an equal number
of gender-congruent distractor words were chosen (again, all F’s < 1). Care was
taken to ensure that targets and distractors were not semantically or phonologi-
cally related. For each gender, the target items of the other two genders served as
gender-incongruent distractors. The complete list of target pictures and distractor
words can be found in Appendix A.
Pictures were black line drawings of everyday objects presented on a white
background. They were taken from the pool of pictures of the Max Planck Insti-
tute for Psycholinguistics in Nijmegen. Distractor words were displayed in their
singular form in black characters (font type and size: Geneva, 30 pts.) inside or
across the pictures. Pictures fitted into a 7 cm x 7 cm frame and appeared in the
center of the screen with the distractor words appearing around fixation.

2.1.3 Procedure

Participants were tested individually in a dimly lit testing room. They sat in front
of a computer screen at a viewing distance of approximately 80 cm. The experi-
menter sat in the same room to score potential errors. On each trial, a fixation
point appeared for 500 ms followed by the picture and the distractor word. Partic-
ipants were instructed to fixate the fixation point and to name the target picture
as quickly and as accurately as possible with the appropriate indefinite determiner
in German. At picture onset, a voice key connected to a microphone (Sennheiser
HME 25–1) was activated to measure the naming latencies. As soon as a response
was given and the voice key was triggered, picture and distractor word disap-
peared from the screen and after a short pause of one second the next trial started.
166       Niels O. Schiller

If no response was recorded within two seconds, the next trial started automati-
cally. The presentation of the trial sequences was controlled by NESU (Nijmegen
Experimental Set-Up). A response was considered invalid when it exceeded the
response deadline of two seconds, when it included a speech error, when a wrong
determiner or picture name was produced, or when the voice key was triggered
incorrectly. Invalid responses were excluded from the reaction time analyses.

2.1.4 Design

The experiment consisted of three parts. First, participants were engaged in a

familiarization phase. They saw each picture once on the computer screen to
become familiar with the pictures and learn the designated picture names (in case
alternative names were preferred by the participants). Each picture appeared on
the screen as a black-on-white line drawing and after two seconds the designated
name was added below the picture. Both remained in view for another three
seconds. Participants were asked to use the designated name for each picture.
After the familiarization phase, participants were engaged in a practice phase
during which each picture was presented once in the center of the screen pre-
ceded by a fixation point. Participants’ task was to name the picture as quickly
and as accurately as possible using the appropriate definite determiner and
picture name, e.g., der Tisch (‘the table’). This procedure was adopted to ensure
that participants knew the gender of all target picture names. After completion
of the practice phase, the experimenter corrected participants if they did not use
the designated name for a given picture. Then a second practice block consisting
of 30 randomly chosen trials was administered together with distractor words.
Participants were requested to name those pictures in the same way as in the first
practice block.
The naming phase proper began immediately after the practice phase. Stimuli
were presented in three blocks of 90 trials each. The stimulus-onset-asynchrony
(SOA) between target and distractor was 0 ms.
Three additional trials included at the beginning of each block served as
warm-up trials and were not included in the analyses. In each block, targets and
distractors of the three grammatical genders were represented approximately
equally often. Each target appeared once with a distractor word from each gender
(90 pictures x 3 genders = 270 trials + warm-up trials). Blocks were randomized
individually for each participant with the following constraints: (a) Before the
same object was presented again, at least four other objects appeared in-between;
and (b) targets could have the same gender on no more than two consecutive
 Psycholinguistic approaches to the investigation of grammatical gender       167

trials. Finally, the order of the blocks was varied across participants. The experi-
ment lasted approximately one hour.

2.2 Results

The data of two participants had to be excluded due to excessive error rates
(> 15%). Naming latencies shorter than 350 ms and longer than 1,500 ms were
counted as outliers (3.4% of the data). The mean naming latencies and error rates
are summarized in Table  1. Analyses of variance were run with Gender Condi-
tion (congruent or incongruent) and Gender of Target (masculine, feminine, or
neuter) as independent variables. Separate analyses were carried out with par-
ticipants (F1) and items (F2) as random variables.

Table 1: Mean naming latencies (in ms) and percentage errors (in parentheses) in Experiment 1.

Gender of Distractor

Masculine Feminine Neuter Mean

Gender of
Target
Masculine 613 (4.8) 602 (3.5) 621 (4.1) 612 (4.1)
Feminine 618 (5.4) 622 (6.5) 615 (6.5) 618 (6.1)
Neuter 610 (4.8) 611 (5.2) 617 (3.7) 613 (4.6)

2.2.1 Naming latencies

Picture naming latencies were 4 ms faster in the gender-incongruent condition

(613 ms) than in the gender-congruent condition (617 ms). The effect of Gender
Condition was not significant (F1(1,17) = 1.43, MSe = 147.85, ns; F2(1,89) = 1.49,
MSe  =  605.25, ns). Masculine targets were named fastest (612 ms), followed by
neuter (613 ms) and feminine targets (618 ms), but the effect of Gender of Target
was not significant (F1(2,34) < 1; F2(1,87) = 1.47, MSe = 616.22, ns) nor was the inter-
action between Gender Condition and Gender of Target significant (F1(2,34) =
3.23, MSe = 221.06, ns; F2(2,87) < 1).
168       Niels O. Schiller

2.3 Discussion

There was no difference in naming latencies between the gender-congruent and

the gender-incongruent condition. The gender feature competition hypothesis
predicted a gender congruency effect because the production of gender-marked
indefinite NPs makes the retrieval of gender features necessary. However, in this
experiment, we found no effect of Gender Condition (congruent vs. incongruent).
Note, that this result replicates an earlier finding using partially distinct materials
by Schiller and Costa (2006). Is there an explanation for the absence of a gender
congruency effect in this task? As noted in the introduction, one possibility is that
gender congruency effects are only found in the production of utterances where
the gender-marked item is a free-standing, morphologically simple morpheme
(see also Janssen, Schiller and Alario in press). Note that the gender-marked
indefinite articles in German, einmas/neu and einefem, are structurally similar to
the possessive adjectives in Croatian, mojmas and mojafem (see above), where no
gender congruency effect was found (Costa et al., 2003). Therefore, although ein/
eine are formally determiners, they seem to behave like adjectives with respect to
the way in which they are selected for production. That is, perhaps ein/eine share
a stem  – ein  – and differ only in the suffix, i.e.  – ø (zero morpheme) versus  –
e. Moreover, if we assumed that the process of stem retrieval is faster than affix
retrieval (plus subsequent affixation), or articulation of the stem can start before
affix retrieval and affixation have been completed, any gender congruency effect
would be masked, just as in Croatian. Alternatively, if we assumed that mor-
phophonological processes, such as affixation, are not selected competitively but
automatically, we would similarly not expect to find a gender congruency effect
in the production of phrases with the indefinite determiners ein/eine. On either
account, we should not find a gender congruency effect with Adj NPs in German,
even though, as already noted, Schriefers (1993) found such an effect in Dutch.
The absence of a gender congruency effect has been demonstrated in an Adj
NP experiment in German (Experiment 1b reported in Schiller and Caramazza
2003). The inflectional gender system in German is similar to its determiner
system: Adjectives take different suffixes in the singular to establish gender
agreement with the noun referent (e.g., grüne Tür (‘green door’, fem), grüner
Tisch (‘green table’, mas), or grünes Buch (‘green book’, neu)), but not in the
plural (e.g., the same suffix -e is used for all three genders – grüne Türen (‘green
doors’, fem), grüne Tische (‘green tables’, mas), or grüne Bücher (‘green books’,
neu)). If the gender congruency effect only occurs when a free-standing morpho-
logically marked word form (like a gender-marked definite determiner) has to
be selected, then no such effect should be found when the gender-marked mor-
pheme is bound to a word stem such as an adjective. Therefore, in a task in which
 Psycholinguistic approaches to the investigation of grammatical gender       169

participants were asked to produce Adj NPs, we expected that in both singular
and plural conditions there would be no gender congruency effects. If, however,
the effect occured also for bound morphemes such as gender-marked suffixes (as
reported by Schriefers, 1993), it should have been observable in singular Adj NPs
in German. However, we found no sign of a gender congruency effect in Adj NP
naming in German, neither in the singular nor in the plural.

3 Experiment 2:
Indefinite Det NP production in Dutch
The second experiment of this study is a replication of Experiment 1 in Dutch.
As mentioned above, Dutch distinguishes two genders, common and neuter. The
definite determiners for common and neuter gender are de and het, respectively.
However, there is only one indefinite determiner for both genders – een. Accord-
ing to the hypothesis that gender congruency effects reflect competition between
free-standing phonological forms (Caramazza et al. 2001; Miozzo and Caramazza
1999; Schiller and Caramazza 2003, 2006), no interference effect is expected
because the same determiner form is produced for both genders. However, the
gender feature competition hypothesis (Schriefers 1993) predicts a gender con-
gruency effect.
It may be argued that gender is not accessed at all in such a situation, i.e.
indefinite Det NP naming in Dutch. For instance, Levelt, Roelofs, and Meyer (1999)
proposed that gender features are only selected when they are necessary for the
encoding of an utterance. This assumption was made to account for the absence
of the gender congruency effect in bare noun naming in Dutch (La Heij et  al.
1998). Levelt, Roelofs, and Meyer (1999) argued that the reason for the absence
of the effect in this situation was that bare nouns could be named without the
selection of their gender features in Dutch. Hence, no gender congruency effect
occurred. This argument could also be applied to the indefinite determiner NP
case in Dutch. Since the indefinite determiner is invariably een, the production
system does not have to select the corresponding gender feature of the noun.
From different research, however, we know that this argument does not
hold: When German participants produced bare nouns in the singular and in the
plural, they showed a significant cost from singular to plural production for mas-
culine and neuter nouns (Tisch ‘table’ – Tische ‘tables’ and Buch ‘book’ – Bücher
‘books’, respectively), whereas for feminine nouns (Tür ‘door’ – Türen ‘doors’) no
such cost occurred (Schiller and Caramazza 2003). This shows that even when a
plural NP is produced and the selection of the gender feature is not logically nec-
170       Niels O. Schiller

essary, the gender of the noun is selected and the corresponding singular deter-
miner is activated. This leads to a competition effect in the case of masculine and
neuter nouns where the determiners are different in the singular and in the plural
(der – die and das – die, respectively), whereas in the case of feminine nouns the
two determiners are identical (die – die) and hence there cannot be form competi-
tion. Janssen and Caramazza (2003) obtained similar findings for Dutch diminu-
tive and plural NPs and Schriefers, Jescheniak, and Hantsch (2002) for German
plural NPs.
As in the original study by Schriefers (1993), different SOAs were included
in Experiment 2 because it is possible that the gender congruency effect is con-
tingent on SOA. When the distractor word is presented too early with respect to
picture onset, the activation of an incongruent gender feature may have already
decayed and thus would be too weak to influence the selection of the target’s
gender node. When the distractor is presented too late, the gender of the target
word may have already been selected and therefore immune to the activation of
an incongruent gender feature. Indeed, Schriefers (1993) obtained the largest
gender congruency effect at SOA 0 ms. The effect was only half as large at SOA
−200 ms in his Experiment 1, and at SOA +450 ms no significant congruency effect
was found. In his Experiment 2, the gender congruency effect was only signifi-
cant at SOA 0 ms. Since this is the SOA at which the semantic interference effect
is usually obtained, this was taken as evidence that the gender congruency effect
occurs at the same level as the semantic interference effect, i.e. the lexical node
level (but see Schriefers and Teruel 2000). We tested three SOAs in this experi-
ment, namely −100 ms, 0 ms, and +100 ms, in order to maximize the probability
of getting a gender congruency effect. The motivation for this manipulation was
to ensure that the distractor word’s gender feature was activated near the point in
time at which the target word’s gender feature was being selected.

3.1 Method

3.1.1 Participants

Sixteen native Dutch participants from the pool of participants of the Max Planck
Institute for Psycholinguistics in Nijmegen took part in Experiment 2 in exchange
for pay.
 Psycholinguistic approaches to the investigation of grammatical gender       171

3.1.2 Materials

Twenty-two target pictures corresponding to monomorphemic Dutch nouns were

selected for naming. Half of them had common gender; the other half had neuter
gender. The target pictures overlapped for the most part with the materials used
by Schriefers (1993) and La Heij et al. (1998). Each target picture was paired with
a gender-congruent, a gender-incongruent, a phonologically, and a semantically
related distractor word (see Appendix 8.2 for a complete list of target pictures
and distractor words). The reason for including the semantic and phonological
distractors was to obtain evidence for processing of the distractors in the poten-
tial absence of a gender congruency effect. In addition, there was a set of 32 filler
pictures paired with gender-congruent, phonologically and semantically unre-
lated distractors. Pictures were simple black line drawings of everyday objects
presented on white background. They were taken from the picture database of
the Max Planck Institute for Psycholinguistics in Nijmegen. Distractor words were
displayed as black characters (font type and size: Geneva, 30 pts) inside or across
the pictures. Pictures were presented in the center of the screen with the distrac-
tor words appearing around fixation.

3.1.3 Procedure

The procedure was the same as in the previous experiment.

3.1.4 Design

Again, there was a familiarization phase, a practice phase, and a naming phase.
The familiarization phase was as before. During the practice phase, participants
had to name each object with the appropriate definite determiner to make sure
participants knew the correct gender. The naming phase consisted of three blocks
(one for each SOA) of 120 trials. Except for the filler pictures, all pictures were
tested in all conditions (22 pictures x 4 conditions x 3 SOAs equals 264 trials +
32 filler pictures x 3 SOAs equals 360 trials altogether). There was a short break
between each naming block. The order of blocks was varied across participants
and the blocks were randomized individually for each participant with the same
constraints as in the first experiment. The whole experiment lasted approxi-
mately 45 minutes.
172       Niels O. Schiller

3.2 Results

Naming latencies shorter than 350 ms and longer than 1,500 ms were counted as
outliers (3.4% of the data). The mean naming latencies and error rates are sum-
marized in Table 2. Analyses of variance were carried out with Condition (gender-
congruent, gender-incongruent, semantically related, or phonologically related)
and SOA (–100 ms, 0 ms, or +100 ms) as independent variables. Separate analyses
were carried out with participants (F1) and items (F2) as random variables.

Table 2: Mean naming latencies (in ms) and percentage errors (in parentheses) in Experiment 2.

Gender of Target

SOA Condition Common Neuter Mean

–100 ms
Congruent 620 (8.5) 617 (10.2) 618 (9.4)
Incongruent 635 (17.1) 625 (10.2) 630 (8.8)
Semantically related 647 (13.6) 634 (13.6) 640 (13.6)
Phonologically related 622 (6.8) 625 (11.4) 623 (9.1)
0 ms
Congruent 659 (11.4) 634 (8.5) 647 (9.9)
Incongruent 638 (10.2) 646 (9.1) 642 (9.7)
Semantically related 684 (17.0) 654 (24.4) 670 (20.7)
Phonologically related 659 (11.9) 638 (17.6) 649 (14.8)
+100 ms
Congruent 686 (6.3) 675 (8.0) 681 (7.1)
Incongruent 684 (8.0) 677 (9.7) 681 (8.8)
Semantically related 663 (20.5) 680 (21.0) 672 (20.7)
Phonologically related 641 (11.9) 643 (9.7) 642 (10.8)

3.2.1 Naming latencies

The main effect of Condition was marginally significant (F1(3,45) = 2.57, MSE =
1,106.59, p = .07; F2(3,63) = 4.28, MSE = 2,511.12, p < .01). Pictures were named
fastest in the phonologically related condition (638 ms), followed by the gender-
congruent (649 ms) and the gender-incongruent (651 ms) conditions; the seman-
tically related condition was slowest (660 ms). Considering the gender-congruent
condition as baseline, t-tests revealed that no condition was significantly dif-
ferent from the gender-congruent one (all p‘s > .07). Furthermore, pictures were
named fastest at SOA  – 100 ms (628 ms) followed by SOA 0 ms (651 ms), and
 Psycholinguistic approaches to the investigation of grammatical gender       173

slowest at SOA +100 ms (669 ms). This effect was significant (F1(2,30) = 7.79, MSE
= 3,576.75, p < .01; F2(2,42) = 32.13, MSE = 1,240.21, p < .01) as was the interaction
between Condition and SOA (F1(6,90) = 2.16, MSE = 751.70, p = .05, F2(6,126) =
2.62, MSE = 1,402.83, p < .05). Analyses of simple effects showed that Condition
yielded a significant result only at SOA +100 ms (F1(3,45) = 3.96, MSE = 1,057.08,
p < .05; F2(3,63) = 4.18, MSE = 1,908.85, p < .01), but not at SOA – 100 ms (F1(3,45)
= 2.09, MSE = 458.22, ns; F2(3,63) = 2.61, MSE = 1,211.95, p = .06) nor at SOA 0 ms
(F1(3,45) < 1; F2(3,63) = 3.17, MSE = 2,195.97, p < .05) – at least not in the analyses by
participants. Although the simple effects were not significant at SOAs – 100 ms
and 0 ms, we conducted pair-wise comparisons between the individual condi-
tion means. The difference between the gender-congruent and the gender-incon-
gruent condition was not significant at any SOA, but the difference between the
gender-congruent and the phonologically-related condition was significant at
SOA +100 ms (t1(15) = 2.82, SD = 48.02, p < .05, t2(21) = 3.63, SD = 49.55, p < .01). The
difference between the gender-congruent and the semantically related condition
was marginally significant at SOA – 100 ms (t1(15) = 1.94, SD = 36.19, p = .07; t2(21)
= 2.63, SD = 50.16, p < .05). These semantic and phonological effects showed that
the distractor words were processed and influenced naming in the absence of any
gender congruency effects.

3.3 Discussion

As predicted by the gender-marked free-standing morpheme congruency hypoth-

esis, there were no gender congruency effects at any SOA. This result is in contrast
with the gender feature selection hypothesis which predicted an effect – at least
at SOA 0 ms, the SOA at which Schriefers (1993) also found his gender congru-
ency effect. We can be sure, however, that participants processed the distractor
words because we found effects of phonological facilitation (at SOA +100 ms) and
semantic interference effects (though only marginally significant in the subject
analysis at SOA – 100 ms).
This result parallels the absence of a gender congruency effect obtained
with indefinite Det NP naming in German. In Dutch, just like in German, there
is no effect of a gender-incongruent distractor word on the naming of an object
when an indefinite Det NP is used. However, gender congruency effects have
been found with similar materials when a slightly different task, i.e. definite Det
NP production, was used (Schiller and Caramazza 2003, 2006). The reason for
the absence of a gender congruency effect is likely the absence of a difference in
form (eencom vs. eenneu). Reliable effects of gender congruency have been found in
Dutch, German, Croatian, and recently in Italian (Finocchiaro (2013)) whenever
174       Niels O. Schiller

free-standing gender-marked morphemes, such as definite determiners or clitic

pronouns, have to be produced.

4 Experiment 3:
Possessive Adj NP production in Dutch
The third experiment of this study is a replication of Experiment 2 using a differ-
ent utterance format. In Dutch, the form of the possessive adjective mijn (‘my’) is
invariable, i.e. independent of the gender of the noun referent. That is, it behaves
similarly to the indefinite article een (‘a’) tested in the previous experiment. For
instance, in Dutch one can say mijn tafelcom (‘my tablecom’) or mijn boekneu (‘my
bookneu’)  – the possessive adjective mijn has the same form for both genders.
Therefore, the predictions for a picture-word interference experiment are the
same as in the previous experiment: According to the gender feature competition
hypothesis (Schriefers 1993), a gender congruency effect is predicted. The alterna-
tive hypothesis, however, predicts that gender congruency effects are due to the
competition for selection between free-standing phonological forms (Caramazza
et al. 2001). Consequently, no gender congruency effect is expected according to
this latter view because the same form of the possessive adjective is used for both
common and neuter gender nouns. As in the previous experiment, we tested the
same stimuli under the same three SOAs, namely −100 ms, 0 ms, and +100 ms,
in order to maximize the probability of obtaining a gender congruency effect.
The motivation for this manipulation was to make sure that the distractor word’s
gender feature was activated near the point in time at which the target word’s
gender feature was being selected.

4.1 Method

4.1.1 Participants

Twenty-two native Dutch participants from the pool of participants of the Max
Planck Institute for Psycholinguistics in Nijmegen took part in Experiment 3 in
exchange for pay.
 Psycholinguistic approaches to the investigation of grammatical gender       175

4.1.2 M
 aterials, Procedure, and Design

Materials, Procedure, and Design were the same as in the previous experiment,
except that participants were requested to name each picture using a possessive
adjective NP of the format mijn (‘my’) + <picture name>.

4.2 Results

Naming latencies shorter than 350 ms and longer than 1,500 ms were counted as
outliers (4.2% of the data). The mean naming latencies and error rates are sum-
marized in Table 3. Analyses of variance were carried out with Condition (gender-
congruent, gender-incongruent, semantically related, or phonologically related)
and SOA (–100 ms, 0 ms, or +100 ms) as independent variables. Separate analyses
were carried out with participants (F1) and items (F2) as random variables.

Table 3: Mean naming latencies (in ms) and percentage errors (in parentheses) in Experiment 3.

Gender of Target

SOA Condition Common Neuter Mean

–100 ms
Congruent 506 (11.6) 502 (9.5) 504 (10.5)
Incongruent 491 (12.0) 503 (9.1) 497 (10.5)
Semantically related 519 (11.6) 512 (9.1) 516 (10.3)
Phonologically related 509 (14.5) 507 (12.0) 508 (13.2)
0 ms
Congruent 520 (10.3) 505 (9.9) 513 (10.1)
Incongruent 518 (9.5) 519 (9.1) 519 (9.3)
Semantically related 524 (14.0) 520 (9.9) 522 (12.0)
Phonologically related 505 (10.3) 503 (10.7) 504 (10.5)
+100 ms
Congruent 524 (9.5) 515 (8.3) 520 (8.9)
Incongruent 521 (11.2) 533 (9.5) 527 (10.3)
Semantically related 528 (15.3) 516 (9.1) 522 (12.2)
Phonologically related 522 (7.4) 519 (6.6) 514 (7.0)
176       Niels O. Schiller

4.2.1 Naming latencies

The main effect of Condition was significant (F1(3,63) = 3.19, MSE = 1,321.67, p <
.05; F2(3,63) = 3.41, MSE = 1,610.23, p < .05). Pictures were named fastest in the
phonologically related condition (508 ms), followed by the gender-congruent (512
ms) and the gender-incongruent (514 ms) conditions; the semantically related
condition was slowest (520 ms). Considering the gender-congruent condition as
baseline, t-tests revealed that only the phonologically-related condition was sig-
nificantly different from the gender-congruent one by subjects, but not by items
(t1(21) = 2.69, SD = 14.68, p < .05; t2(21) = 1.74, SD = 22.47, p < .10). Furthermore,
pictures were named fastest at SOA – 100 ms (506 ms) followed by SOA 0 ms (514
ms), and slowest at SOA +100 ms (521 ms). This effect was only significant in the
items analysis (F1(2,42) = 1.76, MSE = 5,158.86, ns; F2(2,42) = 10.75, MSE = 4,702.96,
p < .01) but the interaction between Condition and SOA was significant by both
subjects and items (F1(6,126) = 2.14, MSE = 769.00, p = .05; F2(6,126) = 2.19, MSE
= 940.26, p < .05). Analyses of simple effects showed that Condition yielded a
marginally significant result at SOA 0 ms (F1(3,63) = 2.56, MSE = 1,363.28, p = .06;
F2(3,63) = 3.17, MSE = 1,483.41, p < .05) and a significant effect at SOA – 100 ms
(F1(3,63) = 4.25, MSE = 911.82, p < .01; F2(3,63) = 5.33, MSE = 1,417.49, p < .01), but
not at SOA +100 ms (F1(3,63) = 1.51, MSE = 584.59, ns; F2(3,63) < 1).
Although the simple effects were not always significant, we conducted pair-
wise comparisons between all the individual condition means for each SOA. The
difference between the gender-congruent and the gender-incongruent condition
was not significant at any SOA, but the difference between the gender-congruent
and the phonologically-related condition was significant at SOA 0 ms by subjects
but not by items (t1(21) = 2.42, SD = 20.28, p < .05; t2(21) = 1.36, SD = 27.74, ns). The
difference between the gender-congruent and the semantically related condition
reached significance at SOA – 100 ms (t1(21) = 2.10, SD = 24.87, p < .05; t2(21) = 3.03,
SD = 19.83, p < .01). The semantic and the phonological effect (though the latter
only significant by subjects) showed that the distractor words were processed and
influenced naming in the absence of any gender congruency effects.

4.3 Discussion

As predicted by the gender-marked free-standing morpheme congruency hypoth-

esis, there were no gender congruency effects at any SOA. This result is in contrast
with the gender feature selection hypothesis which predicts an effect – at least
at SOA 0 ms, the SOA at which Schriefers (1993) also found his effect. We can be
relatively confident, however, that participants processed the distractor words
 Psycholinguistic approaches to the investigation of grammatical gender       177

in Experiment 3 (as in the previous experiments) because we found effects of

phonological facilitation (at SOA +100 ms, though only significant in the subjects
analysis) and effects of semantic interference (at SOA – 100 ms).
This result parallels the absence of a gender congruency effect obtained with
indefinite Det NP naming in German (Experiment 1) and in Dutch (Experiment 2).
In Dutch, furthermore, there is no effect of a gender-incongruent distractor word
on the naming of an object when an Adj NP is used. However, gender congru-
ency effects have been found with similar materials when a slightly different task
was used, i.e. definite Det NP production (Schiller and Caramazza 2003, 2006).
The reason for the absence of a gender congruency effect is likely the absence of
a difference in form (mijncom vs. mijnneu). Reliable effects of gender congruency
have been found in Dutch, German, Croatian, and Italian whenever free-standing
gender-marked morphemes, such as definite determiners or clitic pronouns, were
produced.

5 General Discussion
In three experiments reported above, we investigated the gender congru-
ency effect to further specify the circumstances under which this effect can be
obtained. In Experiment 1, German participants were asked to produce gender-
marked indefinite determiner NPs such as ein Tisch ‘a table’ or eine Tür ‘a door’
while gender-congruent or gender-incongruent distractor words were displayed
visually. No effect of gender congruency was obtained. Experiment 2 involved the
production of indefinite determiner NPs in Dutch. Again, there were no signs of a
gender congruency effect. Finally, Experiment 3 tested possessive adjective NPs
in Dutch, again without yielding a gender congruency effect. However, effects
of semantic and phonological relatedness in Experiments 2 and 3 demonstrated
that distractor words were being processed in the current study.
According to the gender feature competition hypothesis (Schriefers 1993), we
should have observed gender congruency effects in all three experiments. This
is because in each case gender has to be selected for the production of NPs and
the presence of a gender-incongruent distractor word should have interfered with
the selection of the target gender feature. As already noted, Schriefers (1993)
obtained results consistent with that hypothesis: He found that participants were
faster in Det+Adj and Adj NPs with gender congruent than incongruent distrac-
tors. However, the results reported here and other results in the literature con-
verge in support of an alternative hypothesis. There are three sets of data that are
relevant: the results on Romance languages, the results on plural NP production
178       Niels O. Schiller

in Germanic languages, and the results on Adj NPs in Germanic languages and in
Croatian.
In a series of experiments that investigated the gender congruency effect in
various Romance languages, Caramazza and collaborators systematically failed
to observe a gender congruency effect in Italian (Miozzo and Caramazza 1999;
Miozzo, Costa, and Caramazza 2002), Spanish and Catalan (Costa et al. 1999), and
French (Alario and Caramazza 2002). The failure to obtain a gender congruency
effect in these languages occurred in the context of experiments that manipulated
SOA, and had sufficient statistical power to reveal reliably semantic interference
and phonological facilitation effects. Recently, Finocchiaro (2013) reported a
gender congruency effect for Italian clitic pronouns demonstrating that the vis-
ibility of the effect does not depend on the language per se, but on the selection
properties of the specific condition.
Although various explanations could be entertained for the contrasting
results between Germanic and Romance languages, Caramazza and collaborators
(2001) proposed that the gender congruency effect is really a determiner congru-
ency effect that is observed only in early selection languages like German and
Dutch. That is, the effect does not reflect competition in the selection of gender
features but rather competition in the selection of determiners. The latter effects
are visible in experiments with early selection languages where the activation
of a competing determiner is sufficiently strong to be detectable in picture-word
interference experiments. Converging evidence for the determiner competition
hypothesis comes from experiments that directly compared the two hypotheses.
Schiller and Caramazza (2003, 2006; Schiller and Costa 2006) have shown
that a gender congruency effect is found in NP production experiments with
German and Dutch speakers only when the determiners associated with gender
incongruent nouns are phonologically distinct. In Dutch and German, plural Det
(Adj) NPs employ the same determiner for all genders (e.g., Dutch: de tafelscom
‘the tables’ – de boekenneu ‘the books’ – de rode tafelscom ‘the red tables’ – de rode
boekenneu ‘the red books’; and German: die Tischemas ‘the tables’ – die Türenfem
‘the doors’ – die Bücherneu ‘the books’ – die roten Tischemas ‘the red tables’ – die
roten Türenfem ‘the red doors’ – die roten Bücherneu ‘the red books’). As predicted
by the hypothesis that the gender congruency effect only occurs when different
determiner forms are associated with the target and distractor nouns, a gender
congruency effect was found for singular NPs but not plural NPs in German as
well as in Dutch (as evidenced in Experiments 1c and 4b of that study). Note that
this result is consistent with the results of our present Experiment 2 where no
gender congruency effect was found in the production of indefinite determiner
NPs in Dutch. In the latter case, as in the case of plural NPs in Dutch and German,
a single determiner form is used for both genders, precluding the possibility of
 Psycholinguistic approaches to the investigation of grammatical gender       179

determiner selection competition. Thus, the results from plural NP production

in German and Dutch and the results from indefinite determiner NP production
reported above support the view that the gender congruency effect is not due to
competition during the selection of grammatical features but due to competition
during the selection of gender-marked free-standing morphemes.
The third set of results, which is relevant for specifying the constraints on the
occurrence of a gender congruency effect, concerns the case of Adj NPs. Schrief-
ers (1993) reported a gender congruency effect with Dutch Adj NPs. Schiller and
Caramazza (2003) showed that they could not replicate this result in Dutch nor
in German. Furthermore, Costa et al. (2003) also failed to obtain a gender congru-
ency effect for Adj NPs (e.g., mojmas krevetmas ‘my bed’ or mojafem trubafem ‘my
trumpet’) in Croatian. As already noted, Croatian can be classified as an early
selection language and this was confirmed by the fact that Costa et  al. (2003)
obtained a clear gender congruency effect when participants were required to
produce utterances in which the gender-marked item was a free-standing mor-
pheme (e.g., vidim gamas ‘I see itmas’ [picture: krevetmas ‘bed’] or vidim jefem ‘I see
itfem’ [picture: trubafem ‘trumpet’]). Thus, the bulk of the evidence would seem
to favor the empirical generalization that the gender congruency effect is not
obtained in the production of Adj NPs that involve the “selection” of a gender-
marked, bound morpheme.
The pattern of results that has emerged provides a coherent picture of the
gender congruency effect: the effect is obtained in the production of NPs with
gender-marked free-standing morphemes for those languages where the selec-
tion of the latter morphemes can occur immediately after the selection of their
controlling nouns. Definite NPs in Dutch and German are prototypical exemplars
of these conditions. The effect is not obtained for late selection languages (e.g.,
Romance languages; but see Finocchiaro (2013) for an exception) or utterances in
which the gender-marked items are bound morphemes (e.g., Adj NPs in Croatian
and German). This pattern of constraints is most naturally explained by assum-
ing that gender feature selection is an automatic (non-competitive) consequence
of noun selection and that the gender congruency effect reflects competition at
the level of gender-marked, free-standing morpheme selection (Caramazza et al.
2001). This formulation of the hypothesis is more general than the previous pro-
posal by Caramazza et  al. who had stated it in terms of competition between
determiners in NPs (the only evidence available at the time concerned determin-
ers). The more general formulation is made necessary by the results reported by
Costa et  al. (2003) who found a gender congruency effect in the production of
utterances in which the gender-marked item was a pronoun. In other words, it
seems that the congruency effect depends on selection competition between gen-
der-marked, free-standing morphemes.
180       Niels O. Schiller

There are two issues that are closely connected to the question of grammati-
cal gender processing but which go beyond the limits of the current chapter, i.e.
gender congruency effects in bare noun naming and selection mechanisms in
closed- vs. open-class words. The first issue, i.e. bare noun naming, has been
briefly mentioned above and for more information we refer the interested reader
to the literature (Cubelli et al. 2005; Finocchiaro et al. 2011; Paolieri et al. 2010a,
2010b). The second issue, i.e. selection mechanisms in closed- vs. open-class
words, has recently gained a lot of attention following Schriefers, Jescheniak, and
Hantsch (2002; see also Schriefers, Jescheniak, and Hantsch 2005; but see Costa
et  al. 2003 and Schiller and Costa 2006) who claimed that different selection
mechanisms are at play for these two types of words. More recently, Lemhöfer,
Schriefers, and Jescheniak, (2006; see also Jescheniak, Lemhöfer, and Schriefers,
in press) claimed that two types of closed-class items, i.e. free-standing determin-
ers as well as bound inflectional morphemes, are both selected by a competitive
lexical selection mechanism. However, see Janssen, Schiller, and Alario (in press)
for a reassessment of the evidence.
Finally, let us have a look at data from different paradigms. In a recent study by
Heim and colleagues (Heim et al. 2009), not only behavioral effects of Det NP pro-
duction in German are reported, but also the neurocognitive correlates of potential
competition processes. Using functional magnetic resonance imaging (fMRI), these
authors found a steeper slope in the haemodynamic response function (HRF) for
picture naming (vs. rest) in the left Brodmann area (BA) 44 in the gender-congruent
relative to the incongruent condition. These data suggest the involvement of BA 44
in the selection of determiner forms for language production.
Using electroencephalography (EEG), Ganushchak, Verdonschot, and Schil-
ler (2011) demonstrated neurocognitive effects in a gender classification task.
Error-negativity (ERN) responses of Dutch-English bilinguals were significantly
higher not only in Dutch (L1) gender classification, but also in English (L2). This
result was interpreted as electrophysiological support for grammatical gender
transfer. Interestingly, this effect was demonstrated even though the L2, English
in this case, does not have grammatical gender in its nominal system.
One issue still remains unresolved. Why is it that we failed to find a gender
congruency effect for indefinite determiner NPs in German (Experiment 1; see
also Schiller and Costa 2006, Experiment 1A)? That is, why do German indefi-
nite determiners behave like adjectives and not like definite determiners in our
experiments? Does this result imply that the selection of the phonological form
of an indefinite determiner in German involves some type of morphophonologi-
cal transformation just like adjectives? This is an intriguing possibility that will
have to wait further experimental and theoretical investigation before it can be
resolved.
 Psycholinguistic approaches to the investigation of grammatical gender       181

In conclusion, the results reported in this study together with earlier results
(Alario and Caramazza 2002; Caramazza et al. 2001; Costa et al. 1999, 2003; Heim
et  al. 2009; La Heij et  al. 1998; Miozzo and Caramazza 1999; Schiller and Car-
amazza 2003, 2006; Schiller and Costa 2006; Van Berkum 1997) suggest that the
selection of grammatical features is an automatic, non-competitive process. The
selection of lexical nodes, however, is a competitive process. The current study
qualifies this last statement: a competition effect between lexical nodes is only
observed when these nodes correspond to free-standing morphemes.

6 Acknowledgements
The author would like to thank Pienie Zwitserlood and Jens Bölte (both Univer-
sity of Münster, Germany) for support in conducting Experiment 1, and Suzan
Kroezen, Frouke Hermens, Anne Jacobs, and Janneke van Elferen (formerly Max
Planck Institute for Psycholinguistics, Nijmegen, The Netherlands) for running
participants of Experiments 2 and 3. A slightly different version of this chapter
was read at the Workshop “The Expression of Gender”, Max Planck Institute for
Psycholinguistics, Nijmegen, 4 March 2011. The author would like to thank the
audience of this workshop for their comments.

7 References
Alario, François-Xavier and Alfonso Caramazza. 2002. The production of determiners: Evidence
from French. Cognition 82. 179–223.
Baayen, R. Harald, Richard Piepenbrock, and L. Gulikers. 1995. The CELEX lexical database
(CD-ROM). Linguistic Data Consortium, University of Pensylvania, Philadelphia, PA.
Caramazza, Alfonso, Michele Miozzo, Albert Costa, Niels O. Schiller, and François-Xavier Alario.
2001. A cross-linguistic investigation of determiner production. In Emanuel Dupoux
(ed.), Language, Brain, and Cognitive Development: Essays in Honor of Jacques Mehler,
209–226. Cambridge, MA: MIT Press.
Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press.
Corbett, Greville G. 2013. Gender Typology. In Greville G. Corbett (ed.), The Expression of
Gender, 87–130. Berlin/New York: Mouton de Gruyter.
Costa, Albert, Núria Sebastián-Gallés, Michele Miozzo, and Alfonso Caramazza. 1999. The
gender congruity effect: Evidence from Spanish and Catalan. Language and Cognitive
Processes 14: 381–391.
Costa, Albert, D. Kovacic, E. Fedorenko, and Alfonso Caramazza. 2003. The gender-congruency
effect and the selection of freestanding and bound morphemes: Evidence from Croatian.
Journal of Experimental Psychology: Learning, Memory, and Cognition 29. 1270–1282.
182       Niels O. Schiller

Cubelli, Roberto, L. Lotto, Daniela Paolieri, M. Girelli, and Remo Job. 2005. Grammatical
gender is selected in bare noun production: Evidence from the picture-word interference
paradigm. Journal of Memory and Language 53. 42–59.
Finocchiaro, Chiara. 2013. Facilitation effects of gender-congruency in the production of Italian
clitic pronouns. Journal of Cognitive Psychology 25. 24–29.
Finocchiaro, Chiara, François-Xavier Alario, Niels O. Schiller, Albert Costa, Michele Miozzo,
and Alfonso Caramazza. 2011. Gender congruency goes Europe: A cross-linguistic study
of the gender congruency effect in Romance and Germanic languages. Italian Journal of
Linguistics 23. 161–198.
Ganushchak, Lesya Y., Rinus G. Verdonschot, and Niels O. Schiller. 2011. When leaf becomes
neuter: event-related potential evidence for grammatical gender transfer in bilingualism.
NeuroReport 22. 106–110.
Glaser, W. R. 1992. Picture naming. Cognition 42. 61–105.
Glaser, W. R. and F.-J. Düngelhoff. 1984. The time course of picture-word interference. Journal of
Experimental Psychology: Human Perception and Performance 10. 640–654.
Glaser, W. R. and M. O. Glaser. 1989. Context effects in Stroop-like word and picture processing.
Journal of Experimental Psychology: General 118. 13–42.
Heim, Stefan, Angela D. Friederici, Niels O. Schiller, Shirley-Ann Rüschemeyer, and Katrin
Amunts. 2009. The determiner congruency effect in language production investigated with
functional MRI. Human Brain Mapping 30. 928–940.
Janssen, Niels and Alfonso Caramazza. 2003. Determiner selection in diminutive and plural
noun phrases in Dutch. Journal of Memory and Language 48. 635–652.
Janssen, Niels, Niels O. Schiller, and François-Xavier Alario. in press. The selection of
closed-class elements during language production: A reassessment of the evidence and a
new look on new data. Language and Cognitive Processes.
Jescheniak, Jörg D., Herbert Schriefers, and Kristin Lemhöfer. in press. Selection of freestanding
and bound gender-marking morphemes in speech production – A review. Language and
Cognitive Processes.
La Heij, Wido, Pim Mak, J. Sander, and E. Willeboordse. 1998. The gender-congruency effect in
picture-word tasks. Psychological Research 61. 209–219.
Lemhöfer, Kristin, Herbert Schriefers, and Jörg D. Jescheniak. 2006. The processing of free and
bound gender-marked morphemes in speech production: Evidence from Dutch. Journal of
Experimental Psychology: Learning, Memory, and Cognition 32. 437–442.
Levelt, Willem J. M., Ardi Roelofs, and Antje S. Meyer. 1999. A theory of lexical access in speech
production. Behavioral and Brain Sciences 22. 1–75.
Lupker, Stephen J. 1979. The semantic nature of response competition in the picture-word
interference task. Memory & Cognition 7. 485–495.
Lupker, Stephen J. 1982. The role of phonetic and orthographic similarity in picture-word
interference. Canadian Journal of Psychology 36. 349–367.
MacLeod, Colin M. 1991. Half a century of research on the Stroop effect: An integrative review.
Psychological Bulletin 109. 163–203.
Miozzo, Michele and Alfonso Caramazza. 1999. The selection of determiners in noun phrase
production. Journal of Experimental Psychology: Learning, Memory, and Cognition 25.
907–922.
Miozzo, Michele, Albert Costa, and Alfonso Caramazza. 2002. The absence of a gender
congruency effect in Romance languages: A matter of stimulus onset asynchrony? Journal
of Experimental Psychology: Learning, Memory and Cognition 28. 388–391.
 Psycholinguistic approaches to the investigation of grammatical gender       183

Paolieri, Daniela, L. Lotto, L. Morales, Teresa Bajo, Roberto Cubelli, and Remo Job. 2010a.
Grammatical gender processing in Romance languages: Evidence from bare noun naming
in Italian and Spanish. European Journal of Cognitive Psychology 22. 335–347.
Paolieri, Daniela, L. Lotto, D. Leoncini, Roberto Cubelli, and Remo Job. 2010b. Differential
effects of grammatical gender and gender inflection in bare noun production. British
Journal of Psychology 102. 19–36.
Schiller, Niels O. and Alfonso Caramazza. 2003. Grammatical feature selection in noun phrase
production: Evidence from German and Dutch. Journal of Memory and Language 48.
169–194.
Schiller, Niels O. and Alfonso Caramazza. 2006. Grammatical gender selection and the
representation of morphemes: The production of Dutch diminutives. Language and
Cognitive Processes 21. 945–973.
Schiller, Niels O. and Albert Costa. 2006. Different selection principles of free-standing and
bound morphemes in language production. Journal of Experimental Psychology: Learning,
Memory, and Cognition 32. 1201–1207.
Schriefers, Herbert. 1993. Syntactic processes in the production of noun phrases. Journal of
Experimental Psychology: Learning, Memory, and Cognition 19. 841–850.
Schriefers, Herbert, Jörg D. Jescheniak and Ansgar Hantsch. 2002. Determiner selection in noun
phrase production. Journal of Experimental Psychology: Learning, Memory, and Cognition
28. 941–950.
Schriefers, Herbert, Jörg D. Jescheniak, and Ansgar Hantsch. 2005. Selection of gender marked
morphemes in speech production. Journal of Experimental Psychology: Learning, Memory,
and Cognition 31. 159–168.
Schriefers, Herbert, Antje S. Meyer, and Willem J. M. Levelt. 1990. Exploring the time course
of lexical access in language-production: Picture-word interference studies. Journal of
Memory and Language 29. 86–102.
Schriefers, Herbert and E. Teruel. 2000. Grammatical gender in noun phrase production: The
gender interference effect in German. Journal of Experimental Psychology: Learning,
Memory, and Cognition 26. 1368–1377.
Starreveld, Peter A. and Wido La Heij. 1995. Semantic interference, orthographic facilitation,
and their interaction in naming tasks. Journal of Experimental Psychology: Learning,
Memory, and Cognition 21. 686–698.
Starreveld, Peter A. and Wido La Heij. 1996. Time-course analysis of semantic and orthographic
context effects in picture naming. Journal of Experimental Psychology: Learning, Memory,
and Cognition 22. 896–918.
Stroop, J. Ridley. 1935. Studies of interference in serial verbal reactions. Journal of Experimental
Psychology 28. 643–662.
Van Berkum, Jos J. A. 1997. Syntactic processes in speech production: The retrieval of
grammatical gender. Cognition 64. 115–152.
184       Niels O. Schiller

8 Appendices

8.1 Stimulus materials in Experiment 1

Target Gender Distractor word condition
picture name
Congruent Incongruent (mas) Incongruent (neu)

Tür (‘door’) feminine Glocke (‘bell’) Fuß (‘foot’) Krokodil

(‘crocodile’)
Sonne (‘sun’) feminine Feder (‘feather’) Koffer (‘suitcase’) Faß (‘barrel’)
Schlange (‘snake’) feminine Banane (‘banana’) Mund (‘mouth’) Klavier (‘piano’)
Gans (‘goose’) feminine Palme (‘palm tree’) Schuh (‘shoe’) Lasso (‘lasso’)
Kerze (‘candle’) feminine Sonne (‘sun’) Besen (‘groom’) Pferd (‘horse’)
Leiter (‘ladder’) feminine Kasse (‘cash register’) Knopf (‘button’) Herz (‘heart’)
Bombe (‘bomb’) feminine Ente (‘duck’) Tisch (‘table’) Regal (‘shelf’)
Hose (‘pants’) feminine Tür (‘door’) Teller (‘plate’) Sofa (‘couch’)
Birne (‘pear’) feminine Pfeife (‘pipe’) Frosch (‘frog’) Zebra (‘zebra’)
Palme (‘palm tree’) feminine Tasse (‘cup’) Magnet (‘magnet’) Kreuz (‘cross’)
Pfeife (‘pipe’) feminine Gabel (‘fork’) Affe (‘monkey’) Schwein (‘pig’)
Nase (‘nose’) feminine Lupe (‘magnifying Stern (‘star’) Bett (‘bed’)
glass’)
Gabel (‘fork’) feminine Hose (‘pants’) Baum (‘tree’) Kissen (‘pillow’)
Brille (‘glasses’) feminine Leiter (‘ladder’) Finger (‘finger’) Schaf (‘sheep’)
Flasche (‘bottle’) feminine Birne (‘pear’) Helm (‘helmet’) Blatt (‘bottle’)
Vase (‘vase’) feminine Brille (‘glasses’) Zahn (‘tooth’) Messer (‘knife’)
Banane (‘banana’) feminine Tasche (‘bag’) Kamm (‘comb’) Kamel (‘camel’)
Blume (‘flower’) feminine Schlange (‘snake’) Löffel (‘spoon’) Kanu (‘canoe’)
Ente (‘duck’) feminine Kette (‘chain’) Rock (‘skirt’) Ruder (‘oar’)
Feder (‘feather’) feminine Ziege (‘goat’) Korb (‘basket’) Zelt (‘tent’)
Glocke (‘bell’) feminine Puppe (‘puppet’) Hammer (‘hammer’) Fenster (‘window’)
Kasse (‘cash feminine Flasche (‘bottle’) Hund (‘dog’) Brot (‘bread’)
­register’)
Kette (‘chain’) feminine Gans (‘goose’) Schlitten (‘sled’) Floß (‘raft’)
Lupe (‘magnifying feminine Mauer (‘wall’) Tiger (‘tiger’) Kabel (‘cable’)
glass’)
Mauer (‘wall’) feminine Trommel (‘drum’) Schrank (‘closet’) Glas (‘glass’)
Puppe (‘puppet’) feminine Blume (‘flower’) Vogel (‘bird’) Auto (‘car’)
Tasche (‘bag’) feminine Vase (‘vase’) Stuhl (‘chair’) Rad (‘wheel’)
Tasse (‘cup’) feminine Nase (‘nose’) Kreis (‘circle’) Schiff (‘boat’)
Trommel (‘drum’) feminine Bombe (‘bomb’) Sattel (‘saddle’) Bein (‘leg’)
Ziege (‘goat’) feminine Kerze (‘candle’) Ofen (‘stove’) Schwert (‘sword’)
Schlitten (‘sled’) masculine Zahn (‘tooth’) Gabel (‘fork’) Kanu (‘canoe’)
Knopf (‘button’) masculine Vogel (‘bird’) Palme (‘palm tree’) Lasso (‘lasso’)
Teller (‘plate’) masculine Föhn (‘hair dryer’) Leiter (‘ladder’) Schiff (‘boat’)
Fuß (‘foot’) masculine Hammer (‘hammer’) Puppe (‘puppet’) Blatt (‘leaf’)
 Psycholinguistic approaches to the investigation of grammatical gender       185

Target Gender Distractor word condition

picture name
Congruent Incongruent (mas) Incongruent (neu)

Tisch (‘table’) masculine Fuß (‘foot’) Glocke (‘bell’) Krokodil

(‘crocodile’)
Affe (‘monkey’) masculine Kamm (‘comb’) Tasche (‘bag’) Brot (‘bread’)
Frosch (‘frog’) masculine Baum (‘tree’) Mauer (‘wall’) Kissen (‘pillow’)
Hund (‘dog’) masculine Teller (‘plate’) Birne (‘pear’) Faß (‘barrel’)
Helm (‘helmet’) masculine Korb (‘basket’) Sonne (‘sun’) Zelt (‘tent’)
Hammer masculine Rock (‘skirt’) Tür (‘door’) Ruder (‘oar’)
(‘hammer’)
Schuh (‘shoe’) masculine Mund (‘mouth’) Banane (‘banana’) Klavier (‘piano’)
Koffer (‘suitcase’) masculine Helm (‘helmet’) Brille (‘glasses’) Sofa (‘couch’)
Löffel (‘spoon’) masculine Schlitten (‘sled’) Gans (‘goose’) Floß (‘raft’)
Magnet (‘magnet’) masculine Knopf (‘button’) Kasse (‘cash Herz (‘heart’)
­register’)
Kamm (‘comb’) masculine Löffel (‘spoon’) Hose (‘pants’) Schaf (‘sheep’)
Korb (‘basket’) masculine Hund (‘dog’) Flasche (‘bottle’) Schwein (‘pig’)
Ofen (‘stove’) masculine Schuh (‘shoe’) Trommel (‘drum’) Fenster (‘window’)
Schrank (‘closet’) masculine Affe (‘monkey’) Blume (‘flower’) Auto (‘car’)
Stuhl (‘chair’) masculine Tiger (‘tiger’) Vase (‘vase’) Messer (‘knife’)
Rock (‘skirt’) masculine Stern (‘star’) Tasse (‘cup’) Bett (‘bed’)
Baum (‘tree’) masculine Finger (‘finger’) Pfeife (‘pipe’) Zebra (‘zebra’)
Besen (‘broom’) masculine Frosch (‘frog’) Schlange (‘snake’) Kamel (‘camel’)
Finger (‘finger’) masculine Tisch (‘table’) Ente (‘duck’) Regal (‘shelf’)
Kreis (‘circle’) masculine Besen (‘groom’) Ziege (‘goat’) Pferd (‘horse’)
Mund (‘mouth’) masculine Stuhl (‘chair’) Kette (‘chain’) Kabel (‘cable’)
Sattel (‘saddle’) masculine Magnet (‘magnet’) Lupe (‘magnifying Kreuz (‘cross’)
glass’)
Stern (‘star’) masculine Koffer (‘suitcase’) Feder (‘feather’) Rad (‘wheel’)
Tiger (‘tiger’) masculine Sattel (‘saddle’) Bombe (‘bomb’) Bein (‘leg’)
Vogel (‘bird’) masculine Kreis (‘circle’) Nase (‘nose’) Glas (‘glass’)
Zahn (‘tooth’) masculine Ofen (‘stove’) Kerze (‘candle’) Schwert (‘sword’)
Brot (‘bread’) neuter Schaf (‘sheep’) Leiter (‘ladder’) Finger (‘finger’)
Bein (‘leg’) neuter Zebra (‘zebra’) Pfeife (‘pipe’) Frosch (‘frog’)
Zebra (‘zebra’) neuter Glas (‘glass’) Nase (‘nose’) Stuhl (‘chair’)
Schaf (‘sheep’) neuter Kabel (‘cable’) Feder (‘feather’) Koffer (‘suitcase’)
Messer (‘knife’) neuter Bett (‘bed’) Bombe (‘bomb’) Stern (‘star’)
Bett (‘bed’) neuter Faß (‘barrel’) Schlange (‘snake’) Hammer
(‘hammer’)
Pferd (‘horse’) neuter Kreuz (‘cross’) Lupe (‘magnifying Magnet (‘magnet’)
glass ')
Kamel (‘camel’) neuter Sofa (‘couch’) Brille (‘glasses’) Helm (‘helmet’)
Schiff (‘ship’) neuter Rad (‘wheel’) Vase (‘vase’) Tiger (‘tiger’)
Zelt (‘tent’) neuter Schwert (‘sword’) Kerze (‘candle’) Ofen (‘stove’)
Lasso (‘lasso’) neuter Fenster (‘window’) Tür (‘door’) Schrank (‘closet’)
Faß (‘barrel’) neuter Blatt (‘leaf’) Puppe (‘puppet’) Baum (‘tree’)
186       Niels O. Schiller

Target Gender Distractor word condition

picture name
Congruent Incongruent (mas) Incongruent (neu)

Klavier (‘piano’) neuter Floß (‘raft’) Gans (‘goose’) Schlitten (‘sled’)

Glas (‘glass’) neuter Brot (‘bread’) Birne (‘pear’) Knopf (‘button’)
Regal (‘shelf’) neuter Auto (‘car’) Blume (‘flower’) Affe (‘monkey’)
Kissen (‘pillow’) neuter Pferd (‘horse’) Ziege (‘goat’) Besen (‘broom’)
Auto (‘car’) neuter Kamel (‘camel’) Tasche (‘bag’) Kamm (‘comb’)
Blatt (‘leaf’) neuter Regal (‘shelf’) Ente (‘duck’) Tisch (‘table’)
Fenster (‘window’) neuter Ruder (‘oar’) Kette (‘chain’) Rock (‘skirt’)
Floß (‘raft’) neuter Zelt (‘tent’) Sonne (‘sun’) Korb (‘basket’)
Herz (‘heart’) neuter Kissen (‘pillow’) Kasse (‘cash Kamm (‘comb’)
­register’)
Kabel (‘cable’) neuter Schiff (‘boat’) Palme (‘palm tree’) Schuh (‘shoe’)
Kanu (‘canoe’) neuter Schwein (‘pig’) Flasche (‘bottle’) Hund (‘dog’)
Kreuz (‘cross’) neuter Herz (‘heart’) Hose (‘pants’) Teller (‘plate’)
Krokodil neuter Lasso (‘lasso’) Mauer (‘wall’) Sattel (‘saddle’)
(‘crocodile’)
Rad (‘wheel’) neuter Kanu (‘canoe’) Gabel (‘fork’) Zahn (‘tooth’)
Ruder (‘oar’) neuter Klavier (‘piano’) Banane (‘banana’) Mund (‘mouth’)
Schwein (‘pig’) neuter Messer (‘knife’) Trommel (‘drum’) Kreis (‘circle’)
Schwert (‘sword’) neuter Bein (‘leg’) Tasse (‘cup’) Vogel (‘bird’)
Sofa (‘couch’) neuter Krokodil ­(‘crocodile’) Glocke (‘bell’) Fuß (‘foot’)
 Psycholinguistic approaches to the investigation of grammatical gender       187

8.2 Stimulus materials in Experiments 2 and 3

Target Gender Distractor word condition
picture name
Congruent Gender Incongruent Gender

poes (‘cat’) common kerk (‘church’) common blad (‘leaf’) neuter

wortel (‘carrot’) common muis (‘mouse’) common stuur (‘wheel’) neuter
stoel (‘chair’) common jas (‘jacket’) common plein (‘square’) neuter
vork (‘fork’) common zoon (‘son’) common touw (‘rope’) neuter
mond (‘mouth’) common zeep (‘soap’) common nest (‘nest’) neuter
tafel (‘table’) common kers (‘cherry’) common glas (‘glass’) neuter
gitaar (‘guitar’) common ladder (‘ladder’) common strand (‘beach’) neuter
zaag (‘saw’) common klap (‘bang’) common papier (‘paper’) neuter
trein (‘train’) common schoen (‘shoe’) common hoofd (‘head’) neuter
taart (‘cake’) common auto (‘car’) common bureau (‘desk’) neuter
fles (‘bottle’) common staart (‘tail’) common wiel (‘wheel’) neuter
konijn (‘rabbit’) neuter stuur (‘wheel’) neuter kerk (‘church’) common
kasteel (‘castle’) neuter blad (‘leaf’) neuter jas (‘jacket’) common
hemd (‘shirt’) neuter wiel (‘wheel’) neuter muis (‘mouse’) common
been (‘leg’) neuter strand (‘beach’) neuter auto (‘car’) common
geweer (‘rifle’) neuter bureau (‘desk’) neuter staart (‘tail’) common
schaap (‘sheep’) neuter nest (‘nest’) neuter kers (‘cherry’) common
schip (‘ship’) neuter glas (‘glass’) neuter zeep (‘soap’) common
brood (‘bread’) neuter plein (‘square’) neuter klap (‘bang’) common
paard (‘horse’) neuter touw (‘rope’) neuter zoon (‘son’) common
bed (‘bed’) neuter papier (‘paper’) neuter ladder (‘ladder’) common
raam (‘window’) neuter hoofd (‘head’) neuter schoen (‘shoe’) common
poes (‘cat’) common hamster common poets (‘trick’) common
(‘hamster’)
wortel (‘carrot’) common asperge common worm (‘worm’) common
(‘asparagus’)
stoel (‘chair’) common bank (‘couch’) common stoep (‘pavement’) common
vork (‘fork’) common lepel (‘spoon’) common vonk (‘spark’) common
mond (‘mouth’) common neus (‘nose’) common monnik (‘monk’) common
tafel (‘table’) common kast (‘cupboard’) common tabak (‘tobacco’) common
gitaar (‘guitar’) common cello (‘cello’) common giraf (‘giraffe’) common
zaag (‘saw’) common hamer (‘hammer’) common zaak (‘thing’) common
trein (‘train’) common bus (‘bus’) common trede (‘step’) common
taart (‘cake’) common koek (‘cake’) common taal (‘language’) common
fles (‘bottle’) common kan (‘jug’) common fluit (‘flute’) common
konijn (‘rabbit’) neuter lam (‘lamb’) neuter koren (‘corn’) neuter
kasteel (‘castle’) neuter huis (‘house’) neuter katoen (‘cotton’) neuter
hemd (‘shirt’) neuter pak (‘suit’) neuter hek (‘fence’) neuter
been (‘leg’) neuter oor (‘ear’) neuter beeld (‘statue’) neuter
geweer (‘rifle’) neuter kanon (‘gun’) neuter gewicht (‘weight’) neuter
schaap (‘sheep’) neuter hert (‘deer’) neuter schaak (‘chess’) neuter
188       Niels O. Schiller

Target Gender Distractor word condition

picture name
Congruent Gender Incongruent Gender

schip (‘ship’) neuter veer (‘ferry’) neuter schild (‘shield’) neuter

brood (‘bread’) neuter ei (‘egg’) neuter brein (‘brain’) neuter
paard (‘horse’) neuter varken (‘pig’) neuter paleis (‘palace’) neuter
bed (‘bed’) neuter rek (‘rack’) neuter beest (‘animal’) neuter
raam (‘window’) neuter luik (‘hatch’) neuter rag (‘cobweb’) neuter
 Psycholinguistic approaches to the investigation of grammatical gender       189

8.3 Stimulus materials in Experiments 2 and 3 (continued)

Filler picture name Gender Distractor word Gender

bijl (‘axe’) common haan (‘rooster’) common

bril (‘glasses’) common kwast (‘brush’) common
eend (‘duck’) common jurk (‘dress’) common
fiets (‘bike’) common tand (‘tooth’) common
hoed (‘hat’) common bal (‘ball’) common
schaar (‘scissors’) common tijger (‘tiger’) common
vis (‘vis’) common klomp (‘clog’) common
hand (‘hand’) common zon (‘sun’) common
klok (‘clock’) common schaats (‘skate’) common
bloem (‘flower’) common maan (‘moon’) common
tent (‘tent’) common noot (‘nut’) common
trompet (‘trumpet’) common kaart (‘card’) common
kaars (‘candle’) common trui (‘sweater’) common
aap (‘monkey’) common emmer (‘bucket’) common
peer (‘pear’) common wolk (‘cloud’) common
kleed (‘carpet’) neuter circus (‘circus’) neuter
masker (‘mask’) neuter bord (‘plate’) neuter
mes (‘knife’) neuter dak (‘roof’) neuter
net (‘net’) neuter blik (‘can’) neuter
oog (‘eye’) neuter web (‘web’) neuter
penseel (‘brush’) neuter geld (‘money’) neuter
spook (‘ghost’) neuter lint (‘ribbon’) neuter
zwaard (‘sword’) neuter blok (‘block’) neuter
harp (‘harp’) neuter robot (‘robot’) neuter
bot (‘bone’) neuter laken (‘sheet’) neuter
kruis (‘cross’) neuter schort (‘apron’) neuter
slot (‘lock’) neuter kalf (‘calf’) neuter
fornuis (‘stove’) neuter vest (‘vest’) neuter
potlood (‘pencil’) neuter monster (‘monster’) neuter
anker (‘anchor’) neuter zadel (‘saddle’) neuter
boek (‘book’) neuter spel (‘game’) neuter
hart (‘heart’) neuter orgel (‘organ’) neuter
Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller
Plural as a value of Cushitic gender:
Evidence from gender congruency effect
experiments in Konso (Cushitic)
1 Introduction
The grammatical features gender and number are related in a complex way in
Cushitic languages (see Hayward 1979; Corbett and Hayward 1987; Mous 2008).
The relation between gender and number becomes apparent in the so-called
“plural” gender nouns. Interestingly and arguably, a third value of grammati-
cal gender in addition to masculine and feminine has been identified for some
Cushitic languages (Hayward 1979; Mous 2008). In contrast  to  other languages
that have three-way gender distinction systems, this third value is not neuter in
Cushitic. In terms of agreement, this third gender value requires the same agree-
ment pattern as the third person plural. As a result, it is called “plural” gender in
many studies of Cushitic languages. In order to avoid confusion, the term “plural”
is used to refer to this gender value and “multiple reference” to refer to the multi-
plicity of number, with the abbreviations (m), (f) and (p) to stand for masculine,
feminine and plural gender values in that order, and (m.r.) and (s.r.) to represent
multiple reference and singular reference nouns, respectively, throughout this
chapter, after Hayward (1984) and Mous (2008). We also used the term “feature”
to refer to gender and number constructs, and “values” for sets of values within
these features (e.g. masculine, feminine and plural for gender) following Corbett
(see Corbett’s chapter in this volume). This complex interrelatedness between
grammatical gender and number sometimes paves ways to different, often con-
flicting analyses of these features in different Cushitic languages.
Two conflicting hypotheses have been put forward to analyze the gender
systems of Cushitic languages. The first one comes from Corbett and Hayward
(1987), in which only two gender values are recognized and the third value is
analyzed as part of the number feature. This analysis is applied to Bayso, a
Lowland East Cushitic language. The argument is that this value has a small
membership and should be analyzed with features indicating irregularity in
number agreement, marked as irregular nouns taking plural agreement (Corbett
and Hayward 1987). Corbett (2012: 223–233) explains that taking the value plural
from the number system as a value in the gender system as well runs counter to
the general principle of exclusiveness that one value belongs to just one feature.
This principle is at the basis of both typology and theoretical studies. A language
192       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

system in which plural operates as a value for gender undermines this central
claim and any alternative analysis of such a gender and number system should
be scrutinized. This is exactly what Corbett (2012: 224–233) does for Bayso in great
clarity and detail. The proposed analysis for Bayso has the added advantage that
it explains anomalies in the system at the cost of marking only a limited number
of nouns with lexical features for exceptional behaviour. If we can show that the
plural value of the gender feature is like the masculine and feminine value of this
feature in terms of psycholinguistic relevance, this poses an important challenge
to the general principle that a value can only belong to one feature, provided that
the plural values for gender and number are indeed instances of the same value.
The second one is the position taken in most descriptive studies on Cushitic
languages (Hayward 1979 for Bayso, Pillinger and Galboran 1999 for Rendille,
Savà 2005 for Ts’amakko, and Orkaydo 2013 for Konso) and argued for in Mous
(2008). The argument is that gender and number are two independent agreement
systems and adjectives show agreement for both features independently (Mous
2008). If the third value of gender is taken to be plural [multiple reference], a
situation can arise in which adjectives show conflicting values for number and
gender agreement with one and the same head noun. For instance, a word in
Iraqw that is of multiple reference and plural in gender has two different agree-
ment markers on the adjective. One agreement system (gender) has low tone on
the final syllable for (f) and (p) head nouns and high tone for (m) head nouns
irrespective of number; the second agreement system has a different form of the
adjective for multiple reference nouns (for examples, see Mous 2008: 156).
The analysis that Corbett (2012) proposes for Bayso introduces extra chal-
lenges when applied to Konso. First of all, in the case of Konso the number of
underived (p) gender nouns is much larger than in Bayso. Based on a count of the
appendix of nouns in Orkaydo’s (2013) grammar of Konso, 96 underived nouns
are (p), against 135 (f) and 245 (m). It becomes less satisfactory to treat all (p)
nouns as exceptional. Secondly, in Konso, gender and number display two sepa-
rate agreement systems. An adjective agrees in number with the head noun by
initial reduplication and in gender by a final suffix, for example filaa-sini’ poor-aa
/comb-def.p [sg]black-p/‘the black comb’ against orra-si’ ka-kapp-a /people-
def.m/f pl-fat-m/f/ ‘the fat people’ Orkaydo (2013: 79). If this (p) head noun were
analysed as not having a gender value but showing plural number agreement,
the morphological analysis would be /comb-def.pl [sg]black-pl/ with compet-
ing values of number on the two agreement slots of the adjective. Moreover the
agreement on the definite marker and the final suffix of the adjective would be
according to gender for one set of nouns but according to number for another set
of nouns – note that multiple reference words like ‘people’ can be masculine in
gender. The Cushitic languages that are mentioned in Corbett (2012: 233) do not
 Cushitic gender: experiments in Konso       193

have these additional challenges because they do not show independent number
agreement (Bayso, Kambaata), or do not have a (p) gender value (Sidamo, Kam-
baata). The analysis for Bayso taking certain nouns as exceptional and similar to
pluralia tantum nouns had the added advantage of explaining anomalies in the
system. The analysis of (p) as a value for gender has the disadvantage of down-
playing a number of anomalies of the Konso gender and number system. Nouns
that are derived for number have predictable gender properties: singulatives are
either masculine or feminine (depending on the formative, not on the gender of
the base as is the case in Bayso) and never (p) in gender; pluratives, derived plural
nouns, are all (p) in gender; there are seven different plural formations in Konso
and all impose (p) gender (Orkaydo 2013: 94–99). Thus, there is ample indication
in Konso for an association of (p) words and plurality in number.
This chapter aims at providing experimental, psycholinguistic evidence
to shed light on the processing of grammatical gender in Cushitic languages
and possibly decide whether the third value (plural gender) is a proper gender
value or rather belongs to the number system in Cushitic languages that have
this value. We employed the picture-word interference paradigm, a commonly
used paradigm in the study of lexical access. In this paradigm, participants are
asked to name a picture while ignoring a distractor word presented with it. It
has been shown that the naming time of the picture is affected by the relation-
ship between the to-be-named and the to-be-ignored words. When the words are
semantically related, for example, interference appears (slower naming times),
however, facilitation (faster naming times) occurs when the two words are pho-
nologically related. Whereas the interference effect is due to competition at the
level of lexical node selection, the facilitation effect is due to the priming at the
level of the phonological form activation (Levelt et al., 1999). In the same way, the
so-called gender congruency effect, i.e. naming times of a picture are faster when
the to-be-named and the to-be-ignored words have the same gender compared
to when they have a different gender, has been proposed for gender processing
mechanisms (see Schiller’s chapter in this volume for a detailed discussion of
psycholinguistic work on gender).
The discussions of the gender congruency effect in psycholinguistic research
have begun after 1993 following Schriefers’ work. Schriefers (1993) found faster
reaction times (RTs) when the target picture and the distractor word had the same
gender compared to when they had a different gender in Dutch noun phrase (NP)
production. He interpreted the effect as demonstrating the competition for selec-
tion of a word’s syntactic features (in this case gender).
According to Bordag and Pechmann (2008), many features inhibit the gender
congruency effect and its interpretation might be more complicated than sug-
gested by Schriefers (1993). They point out that the gender congruency effect
194       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

has been found to be language-specific, which leads to the question of cross-

linguistic differences in gender representation and NP production mechanisms.
Another point of debate is whether the effect is really a gender congruency
effect or rather a determiner congruency effect (Schiller and Caramazza 2002),
which leads to the question of whether it reflects competition at the grammati-
cal (abstract gender nodes) or at the phonological level (phonological forms).
Finally, Bordag and Pechmann mention the presence of inconsistent evidence in
relation to the magnitude of the gender congruency effect. Thus, Schiller and Car-
amazza (2003) found the effect only when an NP in the form of a free morpheme
(article or other determiner) + noun was produced, whereas Schriefers (1993) also
obtained the effect in the production of NPs consisting of an adjective ending
in a gender-marked inflection (a bound morpheme) + a noun. Moreover, La Heij
et al. (1998) observed the effect only when participants named the pictures with
gender-marked NPs, but not when bare nouns were produced in Dutch, implying
that the gender feature may not be always selected. Cubelli et al. (2005), on the
other hand, obtained an interference effect due to gender congruency when bare
nouns were produced in Italian.
This chapter, therefore, attempts to contribute to filling the gap of cross-lin-
guistic confirmation from non-western languages in all these areas, specifically
on the magnitude of gender congruency effect, that is, whether or not the effect
is found in naming bare nouns (Experiment 1), and in naming a noun + a gender-
marked inflection (a bound morpheme) (Experiment 2) in addition to addressing
the issue of deciding whether or not the third value is a proper gender value in
Cushitic languages.
Cushitic languages that arguably have a three-way gender distinction system
with the third being plural besides feminine and masculine are Bayso, Konso,
Dirayta, Ts’amakko, Rendille, and Boni of the Southern Lowland as well as Iraqw,
Alagwa, and Burunge of Southern Cushitic languages (Mous 2008). The language
we investigate is Konso, a Lowland East Cushitic language. Konso is spoken in
the south-west of Ethiopia by 250,000 people (Central Statistics Agency of Ethio-
pia 2009), see the map in Figure 1 at the end of the chapter for its location. The
language has no standardized writing system, proper dictionary or published
grammar book yet. However, Ongaye Oda Orkaydo has recently finished his PhD
thesis on the grammar of Konso. Most of the materials for the experiments reported
in this chapter are taken from him and his work. Konso was chosen for several
reasons: first, it has a relatively large number of underived plural gender nouns;
second, an in-depth analysis of the grammar of the language is available (Orkaydo
2013); third, Ongaye Oda Orkaydo, a native-speaker linguist was available to help
with the selection of the stimuli; fourth, carrying out experiments is feasible in
Konso because the population is large enough and relatively easily accessible.
 Cushitic gender: experiments in Konso       195

2 Gender system of Konso

Konso is said to have a three-way gender distinction system, and the third gender,
besides feminine and masculine, is plural gender (Orkaydo 2013). According to
Orkaydo, for the majority of the nouns neither the form of a noun nor its meaning
predicts its gender value. Nouns in Konso can be classified into three gender
values on the basis of their agreement on the verb: those nouns that show the
same agreement as the third person masculine singular subject (which is not mor-
phologically marked) are masculine; those nouns that show the same agreement
as the third person feminine singular subject (marked on the verb by the suffix –t)
are feminine; and those nouns that show the same agreement as the third person
plural subject (marked on the verb by the suffix –n) are plural. Note that gender
on the verb is realized as long as nouns function as non-focused subjects.
For example, nouns that are semantically masculine, such as sakoota
‘coward’ may have feminine gender agreement on the verb. Similarly, nouns
that are semantically feminine, for example ʔokkatta ‘cow’, may have masculine
gender agreement; also, nouns that are semantically singular, filaa ‘comb’ as
in (c) of Table 1, may have a plural gender agreement. Likewise, nouns that are
semantically undetermined for sex (as feminine or masculine) may have mascu-
line, feminine or plural gender agreement on the verb.

Table 1: Konso subject gender agreement on the verb (Orkaydo p.c.)

(a) sakoota-siʔ ʔi=ɗey-t-i

coward.DEF.F/M 3=come-3SF-PF
‘The coward came.’

(b) ʔokkatta-siʔ ʔi=pat-ay

cow-DEF.F/M 3=disappear-PF(.M)
‘The cow disappeared.’

(c) filaa-siniʔ ʔi=pat-i-n

comb-DEF.P 3=disappear-PF-PL
‘The comb disappeared.’

(d) ʔiskatta-siʔ parre ʔi=ɗey-a

women-DEF.F/M tomorrow 3=come-IPF.FUT(.M)
‘The women will come tomorrow.’

The examples in Table 2 show that the assignment of definite marking on nouns
is determined by the gender of the noun. Thus, nouns that show the same gender
196       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

agreement as the third person masculine or feminine subject take the singular
definite suffix –siʔ while those that show the same agreement as the third person
plural subject on the verb take the definite suffix -siniʔ.

Table 2: Assignment of definite marking on nouns in Konso (Orkaydo 2013: 77–78)

(a) ʛimayta-siʔ ʔi=kutiʔ-ay

old.man-DEF.F/M 3=sit.down-PF(.M)
‘The old man sat down.’

(b) ʔalleeta-siʔ ʔi=piʔ-t-i

hut-DEF.F/M 3=fall-3SF-PF
‘The hut fell.’

(c) ʔinnaa-siniʔ ʔi=muk-i-n

child-DEF.P 3=sleep-PF-PL
‘The child slept.’

(d) lahaɗɗaa-siniʔ ʔi=kat-am-a-n

rams-DEF.P 3=sell-PASS-IPF.FUT-PL
‘The rams will be sold.’

Taken together, there are two divergent ways of analyzing the so-called plural
gender in Cushitic languages that are characterized by this value. On the one
hand, only two gender values, namely masculine and feminine, are recognized
and the third value is analyzed as part of number feature (Corbett and Hayward
1987). On the other hand, three gender values are recognized and the third is
treated as a proper gender value (Mous 2008; Savà 2005; Orkaydo 2013). We want
to investigate whether words like ʔinnaa ‘child’ which are (p) in gender are repre-
sented like words such as karmaɗaa ‘lions’ which are plural in number (support-
ing Corbett’s analysis) or are treated like words such as furaa ‘key’ which are also
(p) in gender (supporting Mous’ analysis) in Konso.

3 Gender congruency effect experiments in Konso

Two gender congruency effect experiments that measured and compared naming
times were conducted in Konso for the first time in the language and in the area.
The first experiment was on the production of bare nouns and the second was on
the production of definite nouns. Both experiments were carried out at a Konso
high school, where the majority of the students speak Konso natively.
 Cushitic gender: experiments in Konso       197

3.1 O
 verview of the experiments

3.1.1 Participants

Forty-six pre-university students, aged between 17 and 25 years, took part in the
two experiments. Twenty-two and 24 students participated in the first and in the
second experiments, respectively. All participants were native speakers of Konso
and they all had normal or corrected-to-normal vision. They were all paid for par-
ticipating in the study.

3.1.2 Materials

Targets were selected from various semantic categories and were presented as
black line drawings on a white background. Names of different gender nouns
were equally represented in both target and distractor conditions. Each target
picture was presented with gender-congruent and gender-incongruent words as
well as with gender-neutral pink noise. The reason to use a neutral distractor
condition (pink noise) was to see if participants processed the distractor words.
Targets and distractor words were not related in terms of meaning (they did not
belong to the same or related semantic categories) and sound (their initial sounds
were not the same and they did not share more than three phonemes). The target
pictures were presented in the center of a 15.6 inch laptop screen accompanied
by one of the distractor words, which was auditorily presented via headphones.

3.1.3 Procedure

Participants were tested individually in a quiet room. They sat in front of a laptop
screen at a viewing distance of around 60 cm. Reaction times (RTs) were measured
from the onset of the target stimulus to the beginning of the naming response
using a voice key. Participants were instructed that they would see a picture and
hear a word or pink noise and were asked to name the target picture while ignor-
ing the word/pink noise they heard.
Each trial began with a fixation point (+) presented in the centre of the screen
for 500 ms followed by the target picture along with the auditorily presented dis-
tractor word until response or for maximally 2,000 ms. Then the asterisk sign (*)
in the centre of the screen was shown before the presentation of the next trial. The
experimenter registered errors and malfunctioning of the voice key.
198       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

The whole experiment was conducted in three phases; familiarization, prac-

tice and test. During the familiarization phase, participants were presented with
all the target pictures along with their intended names twice so as to encourage
the use of the designated picture names of each target. In the practice phase,
each picture was put on the laptop screen for 350 ms accompanied by pink noise.
Participants were asked to name the picture and corrected when they produced
a noun different from the intended one. In the last phase, the test proper, par-
ticipants were asked to name the picture as quickly and as accurately as possible
while ignoring simultaneously presented auditorily distractor words/pink noise.
The Stimulus Onset Asynchrony (SOA) was 0 as the target picture and the distrac-
tor element were presented at the same time. The stimuli were presented in dif-
ferent blocks with a break between them. Under each block, trials were pseudo-
randomized using a Latin square so as to avoid the subsequent appearance of
each distractor condition in a row.

3.1.4 Analysis

Responses were excluded from further analyses in light of the following crite-
ria: (a) unintended name for the picture; (b) non-verbal sounds that triggered the
voice key; (c) unregistered responses including responses given after 2,000 ms;
and (d) reaction times shorter than 250 ms and longer than 1,900 ms. Analyses of
Variance (ANOVAs) were conducted on the filtered correct responses based on the
aforementioned criteria to compare reaction times in different distractor condi-
tions. The significance level used in both experiments was p=0.05. All ANOVAs
that are reported in this chapter were performed on the means per subject (F1)
and the means per item (F2).

3.2 Experiment 1: Production of bare nouns in Konso

In this experiment, participants were asked to name the picture by producing

the noun only (the bare noun). The aim was to be able to decide whether the
gender value “plural” (p) is a proper gender value or rather a value inherent to
the number feature. The logic behind this experiment was that a picture should
be named relatively faster when a word with the same gender is used as a dis-
tractor than the use of a word with a different gender. If (p) is a proper gender
value in Konso, we should find a gender congruency effect in the production
of plural nouns, that is, naming a picture with (p) value should be relatively
faster when a gender-congruent distractor word is presented as compared to the
 Cushitic gender: experiments in Konso       199

presentation of a gender-incongruent distractor word. If it does not show these

properties, this may be due to the value (p) not belonging to the gender feature
and hence probably belonging to the number feature instead. Such a finding
would also have an implication on the analysis of current models of language
production (see below).

3.2.1 Materials

A total of 60 target pictures corresponding to underived Konso nouns were

selected. Twenty of the nouns were masculine, 20 feminine and 20 plural gender
nouns. For each target picture, in addition to neutral pink noise, a gender-con-
gruent and two gender-incongruent (labeled as incongruent I and incongruent
II) distractor words were selected. A total of 240 (20 [target pictures] x 3 [target
genders] x 4 [distractor conditions] = 240) trials were used in this experiment.
The materials were presented in four blocks of 60 items. In each block, each
pictured appeared only once, either with a congruent, one of the two incongru-
ent, or the neutral (pink noise) distractor conditions. Each picture was presented
four times to each participant throughout the experiment, once in each condition
and in each block.

Table 3: Example of the experimental items used in Experiment 1

Target Gender Distractor word conditions*

picture name
Congruent Incongruent I Incongruent II Neutral

kaawwata feminine eetota muutiya ɲupuraa pink noise

‘mirror’ ‘dinner/supper’ ‘worm’ ­‘component of a
set of weaving’

tuyyuuraa masculine ammaʔitta paakkota ‘span χoffaa ‘groin’ pink noise

‘airplane’ ­‘breakfast’ (measurement)’

kupaʔtaa plural marʄaa hiparaata hallaka pink noise

‘tortoise’ ‘hip’ ‘bat’ ‘fat’

* For (m) target, (f) is incongruent I and (p) is incongruent II; for (f) target, (m) is incongruent I
and (p) is incongruent II; and for (p) target, (f) is incongruent I and (m) is incongruent II

Table 3 shows the design for the first experiment. Each target picture was pre-
sented four times, each time associated with different distractors. For the mascu-
line target tuyyuura ‘air plane’, for example, we had a gender-congruent distrac-
200       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

tor word ammayitta ‘breakfast’ in one trial, a gender-incongruent feminine word

paakkota ‘span (measurement)’ in another trial, another gender-incongruent
(incongruent II) plural gender xoffaa ‘groin’ and a neutral ‘pink noise’ as a control
in different trials. The same was true for the other targets as well. The list of items
used in Experiment 1 can be found in Appendix A.

3.2.2 Results and discussion

From a total of 5,280 observations, 982 (19%) were discarded from the analysis
as they were incorrect responses and 207 (3.92%) were labeled as outliers, RTs
shorter than 250 ms and longer than 1,900 ms. Thus, a total of 4,091 (77.48%) data
points were included in the statistical analysis.

Table 4: Reaction Times (RTs) in ms and error in percentage (%e)

Distractor Congruent Neutral Incongruent* Congruency

­conditions effect

RTs (%e) 1010 (5%) 908 (3%) 1029 (11%) 19 (6%)

* Represents a combined effect of incongruent I and incongruent II conditions

Table 4 shows the mean RTs in the three distractor conditions (congruent, neutral
and incongruent). The first overall ANOVA was performed with Distractor Con-
dition (congruent, neutral and incongruent) and Target Gender (masculine,
feminine and plural) as independent factors. This analysis showed a significant
effect of the factor Distractor Condition in both the subject and the item analy-
ses (F1(2,42) = 25.098, p < .0001; F2(2,110) = 45.193, p < .0001). It took the partici-
pants slightly more time to produce a noun in the gender-incongruent condition
(picture-word pairs having a different gender value) than to produce a noun in
the congruent (picture-word pairs having the same gender value) or in the pink
noise (neutral) conditions. The fastest RTs in the pink noise condition reveal that
participants indeed processed the distractors, as processing words interferes
more with naming a picture than processing pink noise. The factor Target Gender
reached significance in the subject analysis (F1 (2,42) = 13.340, p < .0001) but not
in the item analysis (F2 (2,55) = 1.552, p < .221). The interaction between Target
Gender and Distractor Condition, however, failed to reach significance in both
subject and item analyses (F1 (4,84) = .858, p < .497; F2 (4,110) = .586, p < .674).
In the bare noun production, the mean RTs in the gender-congruent condi-
tion were 19 ms faster than naming latencies in the incongruent condition. To
 Cushitic gender: experiments in Konso       201

examine the effect of gender congruency (congruent versus incongruent), a

separate ANOVA was performed on the mean reaction times per subject (F1) and
per item (F2) with Distractor Condition (congruent and incongruent) and Target
Gender (masculine, feminine and plural) as independent factors. This analy-
sis showed a significant effect only on the subject analysis of the factor Target
Gender (F1 (2,42) = 12.072, p < .0001) and was close to reaching significance on
the subject analysis of the factor Distractor Condition (F1 (1,21) = 3.848, p < .063).
One could speculate, however, that the overall slow RTs, the nature of the lan-
guage (gender is not marked on the nouns) and methodological issues might be
possible causes for the inhibition of the effect in the variables that did not reach sig-
nificance level. Compare the 982 ms mean RT of our study to that of less than 700 ms
in previous studies (e.g. La Heij et al. 1998). Moreover, the comparatively weak per-
formance of participants reflected in the relatively high rate of incorrect responses,
i.e. 982 (19%), compared to less than 10% in previous studies. This means that par-
ticipants of this language require more training than set by the standard, as they are
inexperienced in participating in this sort of experiments and working with some of
the equipment. Additional experiments with better training of participants ensur-
ing faster and better performance, may give us a better picture of the issue.
The fact that gender is not marked on the nouns themselves in Konso might
also inhibit the effect from being revealed. Involving gender-marking elements
such as verbs as part of the experiment might help here. Although all possible
precautions were taken in the absence of norms in the language to control for
factors such as frequency, familiarity, typicality and age of acquisition, it could
be the case that our result is inconclusive because of some methodological flaws
such as the use of badly selected words in the experiment (see below).

Table 5: Mean RT by target gender and distractor condition

Target gender Distractor condition

Congruent Incongruent Neutral Congruency effect

Masculine  978  998 872 20

Feminine 1047 1048 923  1
Plural 1007 1042 932 35

Table  5 shows that 20 ms congruency effect in masculine noun production and

35 ms in the plural gender noun production but almost nothing in feminine noun
production (1 ms). In order to see why we failed to obtain a congruency effect in
feminine noun production in the presence of a comparable effect in masculine and
plural gender nouns, we conducted item-by-item analysis of both the target and dis-
202       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

tractor words. Our investigation revealed that participants performed very slowly in
naming certain nouns. Table 6 shows that the majority of the nouns that have rela-
tively slow RTs are feminine nouns (6 out of 10) followed by plural gender nouns (3
out of 10) and only one masculine noun had slow RTs. This might be due to semantic
interference effects between different targets and the clarity of the pictures used (see
Table 6). Only further research will exclude the possibility of the role of methodolog-
ical flaws for the absence of the congruency effect on feminine noun production.

Table 6: Targets that show relatively slow RTs with possible reasons

Target Gender* Potential reasons for slow RT

1. muklaa ‘bangle’ (p) Less familiar word

2. tuuyyata ‘pig’ (f) Less familiar word and semantic interference with
kaharta ‘ewe’ (used as a feminine target)
3. hirribaa ‘eyelash’ (p) Less clear picture
4. napahata ‘ear’ (f) the preferred word is kurra (s.r.) (m) ‘ear’
5. xashitta ‘shoulder’ (f) Less clear picture and semantic interference with
kessa ‘arm’ (used as a masculine target)
6. mulaketa ‘frog’ (f) Less clear picture and semantic interference with
kupataa ‘tortoise’ (used as a plural target)
7. paala ‘feather’ (m)
8. hiipta ‘lip’ (f) Semantic interference with afaa ‘mouth’ (used as a
plural target)
9. loqta ‘leg’ (f) Semantic interference with kessa ‘arm’ and xashitta
‘shoulder’
10. hashallaa ‘leaf ’ (p)

* 3 (p) words, 6 (f) words and 1 (m) word

Thus, explaining the present result in the light of current models of language
production is not straightforward. Recall, however, that La Heij et al. (1998) failed
to find a gender congruency effect in bare noun production in Dutch. Moreover,
most current models of language production suggest that syntactic properties
such as gender are not selected in bare noun naming (Caramazza 1997; Levelt,
Roelofs and Meyer 1999).
According to Caramazza’s (1997) Independent Network (IN) model, a word’s
grammatical gender is activated after the selection of the related lexical node
and hence its selection is an automatic (non-competitive) process. Although the
WEAVER++ model of Levelt et  al. (1999) assumes that the activation of a noun’s
gender takes place before the selection of the lexeme node, it also assumes that the
activation of gender has no effect on the activation level of the nodes related to nouns
with the same gender. This model, therefore, predicts no gender congruency effect
 Cushitic gender: experiments in Konso       203

in bare noun naming as the selection of gender is expected only in the production
of NPs that involve the selection of gender marked elements such as determiners.
Cubelli et  al. (2005), however, found interference effect of grammatical
gender in the production of bare nouns with Italian speakers. They interpret
the result as showing the obligatory selection of grammatical gender, and thus
gender is always selected whenever its noun has to be named, which is in oppo-
sition to the prediction of the WEAVER++ model. Moreover, they argue that the
selection of the noun’s gender is a competitive process in contrast to the assump-
tion of IN model. In order to explain the variation of their data with the assump-
tion of both WEAVER++ and IN models, they came up with the Double Selec-
tion model, which assumes the independent and competitive selection of both
lexical-semantic and lexical-syntactic information prior to the selection of the
phonological form of a word.
In sum, the overall naming latencies in the gender incongruent condition
were 19 ms slower than the congruent condition. This result, however, failed
to reach significance except in the subject analysis of the factor Target Gender.
Moreover, a 35 ms congruency effect observed in naming plural gender nouns
could be taken as a sign for recognizing plural as a proper gender value in Konso.

3.3 Experiment 2: Production of definite nouns in Konso

The gender of nouns also determines the assignment of definite marking on

nouns in Konso (Orkayda 2013). Plural gender nouns take the definite suffix -siniʔ
(e.g. ʔinnaa-siniʔ ‘the child’). Nouns that show the same gender agreement as the
third person masculine or feminine subject, however, take the singular definite
suffix -siʔ (e.g. ʛimayta-siʔ ‘the old man’). In Experiment 2, we investigated the
gender congruency effect in naming nouns with a suffixed definite marker (noun
+ -siʔ [masculine or feminine definite nouns] versus noun + -siniʔ [plural gender
definite noun]). If (p) is a proper gender value in Konso, we should find a gender
congruency effect in the production of plural gender definite nouns.

3.3.1 Materials

Forty target line drawings were selected from Experiment 1, i.e. 20 (noun + -siniʔ
[plural gender definite noun]) and 20 (noun + -siʔ [masculine or feminine] non-
plural gender definite nouns). A total of 120 (20 [pictures] x 2 [genders] x 3 [condi-
tions] = 120) trials in three blocks were used in this experiment. The list of items
used in Experiment 2 can be found in Appendix B.
204       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

Table 7: Examples of the experimental items used in Experiment 2

Target gender Target picture Distractor conditions

Congruent Incongruent Neutral

Plural gender filaasiniʔ oytaasiniʔ tuubutasiʔ pink

definite noun ‘the comb’ ‘the upper part of ‘the false banana noise
the compound’ bread’
Non-plural (m/f) kaawwatasiʔ eetutasiʔ pakaannaasiniʔ pink
gender definite ‘the glass’ ‘the dinner’ ‘the root crop’ noise
noun

3.4 Results and discussion

From a total of 2,880 observations, 491 (17%) were eliminated from the analy-
sis, as they were incorrect responses and 137 (4.76%) were labeled as outliers,
RTs outside of 250 to 1,900 ms. Thus, a total of 2,252 (78.19%) data points were
included for statistical analysis.

Table 8: Reaction Times (RTs) in ms and error in percentage (%e)

Distractor Congruent Neutral Incongruent Congruency

­condition effect

RT (%e) 1210 (7%) 980 (4%) 1223 (6%) 13 (-1%)

Table 8 shows the mean RTs in the three distractor conditions (congruent, neutral
and incongruent). An ANOVA was performed with Distractor Condition and
Target Gender as independent factors. This analysis showed a significant effect
of the factor Distractor Condition in both the subject and the item analyses (F1
(2,46) = 129.733, p < .0001; F2 (2,76) = 62.975, p< .0001). It took the participants
slightly more time to produce a noun in the gender-incongruent condition than
to produce a noun in the gender-congruent and the pink noise condition. The
factor Target Gender reached significance only in the item analysis (F2 (1,38) =
64.489, p < .002). The interaction between Target Gender and Distractor Condi-
tion, however, failed to reach significance in both subject and item analyses (F1
(2,46) = 1.241, p < .298; F2 (2,76) = .160, p < .853). Although we obtained a 13 ms
congruency effect, ANOVAs comparing only the gender-congruent and incongru-
ent conditions did not reach significance for any variable.
 Cushitic gender: experiments in Konso       205

To examine the effect of gender congruency between plural-congruent versus

plural-incongruent and non-plural congruent versus non-plural incongruent in
the distractor conditions, separate ANOVAs were performed on these data. These
analyses only showed a significant effect in the subject analysis between plural-
congruent versus plural-incongruent (F1 (1,23) = 4.395, p < .047) but not between
non-plural congruent versus non-plural incongruent. The congruency effects
were 32 ms and −6 ms, respectively, as can be seen in Table 9. Similar to the first
experiment that showed no effect of gender congruency in feminine noun pro-
ductions, the non-plural gender definite nouns (in which half of them are femi-
nine definite nouns taken from Experiment 1) had a non-significant reversed con-
gruency effect.

Table 9: Mean RTs by target gender and distractor condition

Target gender Distractor conditions

Congruent Incongruent Neutral Congruency effect

Plural 1197 1228 977 32

Non-plural 1223 1217 982 −6

Studies investigating the issue of whether the competition for selection hypoth-
esis applies to the retrieval of gender-marked inflections or restricted to only free-
standing morphemes like determiners are few and present contradictory results
(see Schiller’s chapter in this volume).
Schriefers (1993) obtained a congruency effect when participants produced
NPs in the form of adjective + noun in Dutch (gender is marked in the adjec-
tives suffixed as bound morphemes, e.g. groenneu boekneu ‘green book’ versus
­groenecom tafelcom ‘green table’), suggesting that either the gender features
compete for selection, or else there is competition for selection of the bound
morphemes associated with the gender inflection of the adjective.
Schiller and Caramazza (2003), however, failed to replicate this result in
both German and Dutch (see also Schiller’s chapter in this volume). To account
for the failure of replicating Schriefers’ (1993) result in the production of adjec-
tive + noun that involve the selection of a gender-marked bound morpheme, the
assumption that the gender congruency effect reflects competition at the level
of gender-marked, free-standing morpheme selection is hypothesized (see Schil-
ler’s chapter in this volume.
The significant 32 ms congruency effect in plural gender definite noun
productions shows the effect could be obtained in gender-marked bound mor-
phemes in Konso. This result is in line with the gender feature selection hypothesis
206       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

(Schriefers 1993). In light of the present result, one could also argue that the so-
called plural gender nouns that take –siniʔ as their definite suffix belong to the
proper gender feature in Konso.

4 General discussion
This chapter attempted to address the issue of deciding whether the so-called
plural gender is a proper gender value or a value inherent to the number system
in Cushitic languages using a picture-word interference paradigm in two experi-
ments. Besides, the psycholinguistic investigations of gender are restricted to a
limited number of Germanic and Romance languages. We believe that field-based
psycholinguistic investigations in less studied languages are important to provide
additional empirical and cross-linguistic evidence as well as to broaden our
knowledge of language processing in general and the cognitive representation of
gender features in particular. In this regard, we took a new step to fill the gap of
cross-linguistic confirmation from non-western languages and to introduce psy-
cholinguistic approaches into the study of Cushitic languages by tackling Konso.
The two experiments reported in this chapter attempted to shed more light
on the magnitude of the so-called gender congruency effect in psycholinguistic
research, particularly whether or not the effect is obtained in naming bare nouns
(Experiment 1), and in naming a noun + a gender-marked inflection (a bound
morpheme) (Experiment 2).
In the bare noun production (Experiment 1), we found 35 ms and 20 ms
congruency effects for the production of the so-called plural gender nouns and
masculine nouns, respectively, though we found no effect of congruency for the
feminine nouns. As far as the issue of plural gender in Cushitic is concerned, the
35 ms congruency effect observed in naming plural gender nouns could be taken
as a preliminary sign for recognizing plural as a proper gender value in Konso
although the 19 ms overall congruency effect is significant only in the subject
analysis of the factor Target Gender. Mention was made of the effect in the other
factors (that were not significant), which might be masked by the overall slow
RTs, lack of overt gender markers on the nouns in the language and methodologi-
cal issues in relation to the selection of stimuli.
In Experiment 2, we investigated the gender congruency effect in naming
nouns with a suffixed definite marker with the prediction that a gender congru-
ency effect in the production of plural gender definite nouns should be observed
if (p) is a proper gender value in Konso. We found a significant 32 ms congruency
effect in the plural gender definite noun productions. This result provides an indi-
 Cushitic gender: experiments in Konso       207

cation for the so-called “plural” gender nouns, which take –siniʔ as their definite
suffix marker, to belong to the proper gender value in Konso. It also shows that
gender congruency effect could be obtained in the production of gender-marked
bound morphemes, which confirms the prediction of gender feature selection
hypothesis (Schriefers 1993).
The non-significant overall 13 ms congruency effect and −6 ms congruency
effect in the non-plural gender definite noun production of the present result,
however, pose the question whether gender congruency effects are found in a gen-
der-marked bound morpheme production at all. Note that the gender-marked free-
standing morpheme congruency hypothesis (see Schiller’s chapter in this volume)
predicts no effect of gender congruency in a gender-marked bound morpheme pro-
duction. The negative congruency effect observed in the non-plural gender definite
nouns might also be due to the overall slow RTs, lack of overt gender marking on
the nouns in the language and methodological issues in relation to the selection of
stimuli. This is because half of the stimuli in the non-plural gender definite noun
group are the feminine nouns that were also used in Experiment 1.
Taken together, the overall results in both experiments fail to reach robust
significance levels and hence it is difficult to make any strong generalizations.
Parts of the results of the two experiments (i.e. the presence of gender congru-
ency effects in the plural gender nouns in both experiments), however, tend to
suggest that the so-called “plural” gender is a proper gender value. Nevertheless,
it is emphasized that there is an urgent need for replicating both experiments
by giving better trainings to participants, by involving gender marking elements
such as verbs as part of the experiment and by replacing part of the stimuli that
are identified as problematic with better ones.

5 Acknowledgements
The authors would like to thank Ongaye Orkaydo for his assistance in selecting the
materials, and Anto Arkato for recording the experimental stimuli. The authors
would also like to express their gratitude to the Leiden University Fund (LUF) for
sponsoring Mulugeta’s field trip to Ethiopia that was carried out between January
and April, 2012. The paper resulted from the project ‘Plural as a value of Cushitic
gender: a psycholinguistic study’, which is part of the Leiden University Center
for Linguistics (LUCL) ‘Language Diversity in the World’ research profile area.
Various parts of this paper were presented at the Language and Cognition Group
(LACG) meetings, Leiden Institute for Brain and Cognition (LIBC). The authors
would like to thank the audience of this group for their valuable comments.
208       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

Figure 1: The location of Konso in Ethiopia (map by Ian Agnew: taken, with permission, from
http://www.geog.cam.ac.uk/research/projects/konsoethnography/)
 Cushitic gender: experiments in Konso       209

6 References
Bordag, Denisa and Thomas Pechmann. 2008. Grammatical gender in speech production:
Evidence from Czech. Journal of Psycholinguistic Research 37. 69–85.
Caramazza, Alfonso. 1997. How many levels of processing are there in lexical access? Cognitive
Neuropsychology 14. 177–208.
Central Statistics Agency of Ethiopia (CSA). 2009. Statistical Tables for the 2007 Population and
Housing Census of Ethiopia.
Corbett, Greville G. 2012. Features. Cambridge: Cambridge University Press.
Corbett, Greville G. 2013. Gender Typology. In Greville G. Corbett (ed.) The Expression of
Gender. 1–39. Berlin/New York: Mouton de Gruyter.
Corbett, Greville G. and Richard Hayward. 1987. Gender and number in Bayso. Lingua 73. 1–28.
Cubelli, Roberto, L. Lotto, Daniela Paolieri, M. Girelli, and Remo Job. 2005. Grammatical
gender is selected in bare noun production: Evidence from the picture-word interference
paradigm. Journal of Memory and Language 53. 42–59.
Hayward, Richard J. 1979. Bayso revisited: Some preliminary linguistic observations II. BSOAS
42(1). 101–132.
Hayward, Richard J. 1984. The Arbore language: A first investigation including a vocabulary
[Kuschitische Sprachstudien 2]. Hamburg: Helmut Buske.
La Heij, Wido, Pim Mak, J. Sander, and E. Willeboordse. 1998. The gender-congruency effect in
picture-word tasks. Psychological Research 61. 209–219.
Levelt, Willem J. M., Ardi Roelofs, and Antje S. Meyer. 1999. A theory of lexical access in speech
production. Behavioral and Brain Sciences 22. 1–75.
Mous, Maarten. 2008. Number as an exponent of gender in Cushitic. In Zygmunt Frajzyngier
and Erin Shay (eds) Interaction of Morphology and Syntax: Case Studies in Afroasiatic
(Typological Studies in Language 75), 137–160. Amsterdam: John Benjamins.
Orkaydo, Ongaye Oda. 2013. A Grammar of Konso. Utrecht: LOT publications.
Pillinger, Steve and Letiwa Galboran. 1999. A Rendille dictionary including a grammatical
outline and an English-Rendille Index. Cologne: Rüdiger Köppe.
Savà, Graziano. 2005. A grammar of Ts’amakko [Kuschitischen Sprachstudien 22]. Cologne:
Rüdiger Köppe.
Schiller, Niels O. and Alfonso Caramazza. 2002. The selection of grammatical features in word
production: The case of plural nouns in German. Brain and Language 81. 342–357.
Schiller, Niels O. and Alfonso Caramazza. 2003. Grammatical feature selection in noun phrase
production: Evidence from German and Dutch. Journal of Memory and Language 48.
169–194.
Schiller, Niels O. this volume. Psycholinguistic approaches to the investigation of grammatical
gender. In Greville G. Corbett, The Expression of Gender. Berlin/New York: Mouton de
Gruyter.
Schriefers, Herbert. 1993. Syntactic processes in the production of noun phrases. Journal of
Experimental Psychology: Learning, Memory, and Cognition 19. 841–850.
210       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

7 Appendix

Appendix A: Stimulus materials in Experiment 1

Target gender Distractor word conditions*

picture name
congruent Incongruent I Incongruent II

kaawwata ‘mirror’ feminine eetuta ‘dinner/ muutiya ‘worm’ ɲupuraa ‘compo-

supper’ nent of a set of
weaving’
lawaʃeeta ‘mouse’ feminine fillayyaata ‘flea’ saalpataa ‘belt’ pahnaa ‘example’
harreeta ‘donkey’ feminine ʄureeta ‘dirt’ toma ‘bowl’ pakataa ‘wide
shield’
piirtuta ‘sun’ feminine ʄooʛʛita ‘mud’ hoppatta ‘guts’ miɗaa ‘cabbage
leaves’
hiɓta ‘lip’ feminine ʄileeta ‘stick used kaɓa ‘canal for payraa ‘type of
by old women’ irrigation’ farm tool’
pottaata ‘pumpkin’feminine haaruta ‘revenge’ hoʃʃa ‘cliff’ saaraa ‘poem’
ʛapaleeta feminine ɗalta ‘seed’ χaʔtiya ‘fly’ pohaa ‘contribu-
‘monkey’ tion, tribune’
ʄifeeta ‘a ring’ feminine ohta ‘cloth (worn kanta ‘sub- piʃaa ‘water’
in the night)’ village’
napahta ‘ear’ feminine haaɗita ‘load, kasirayta ‘tick paankaa ‘machete’
burden’ (parasite)’
irroota ‘mountain’ feminine hoollata ‘sheep hiiɓa ‘meat soup’ peeʛaa ‘quarrel’
skin’
kaharta ‘ewe’ feminine kaaɓtuta ‘farm hawla ‘grave, mookkaa ‘cassava’
tool’ tomb’
tika ‘house’ feminine koorita ‘type of sataʔta ‘heart’ leʛaa ‘a loan
cloth’ (money)’
muukuta ‘frog’ feminine kanaʔta ‘palm’ kappaa ‘wheat’ teepaa ‘rope’
lafta ‘bone’ feminine koromta ‘heifer’ karayta ‘tributary’ urmalaa ‘market’
loʛta ‘leg’ feminine kulleeta ‘hood; ɗaammaa ‘flour’ pakaannaa ‘root
cap’ crop’
farta ‘horse’ feminine kusumta ‘navel’ ʃenʛera ‘hook’ χallaa ‘kidney’
tuuyyata ‘pig’ feminine leemmuta kaasa ‘horn’ paarkaalaa
‘bubble’ ‘enemy’
oχinta ‘fence’ feminine kuuɲata ‘gnat’ kuuɲata ‘gnat’** ɗooʃʃaa ‘sarcasm’
χampirteeta ‘bird’ feminine kannoota ilkitta ‘tooth’ koofinaa ‘lung’
‘calabash to drink
from’
taaltaallata feminine kawwatta ‘terrace’ orritta ‘devil’ masaanaa
‘giraffe’ ‘autumn’

* For (f) target, (m) is incongruent I and (p) is incongruent II

** Has (f) gender value and also used as a congruent distractor for oχinta ‘fence’ (f)
 Cushitic gender: experiments in Konso       211

Appendix A: Stimulus materials in Experiment 1 (continued)

Target Distractor word conditions*

gender
picture name congruent Incongruent I Incongruent II

arpa ‘elephant’ masculine hoofa ‘hole’ χoraa ‘fine, pun- kawlaa ‘metal
ishment’** tool for ginning’
kessa ‘chest’ masculine alkitta ‘sisal’ mooluta ‘bald’ tiraa ‘liver’
paala ‘feather’ masculine ekerta ‘olive’ nooɗɗuta ‘bribe’ kolkaa ‘food
without cabbage’
harka ‘arm’ masculine ɗila ‘field’ moonta ‘sky’ toorraa ‘opposi-
tion’
murkuʄaa ‘fish’ masculine ʄalaʛʛitta ‘flat pokkeeta ‘short tiyyaa ‘dispute’
stone’ (with pockets)’
tuyyuuraa ‘air masculine ammaʔitta paakkuta ‘span χoffaa ‘groin’
plane’ ‘breakfast’ (measurement)’
mottooʛaa ‘truck’ masculine ukkaʃʃa ‘husk'’ hakayta ‘second toʄaa ‘water
round harvest’ droplet’
sookitta ‘salt’ masculine hallaka ‘fat’ paallata ‘piece of torraa ‘speech,
clay to fetch fire talk’
with’
okkatta ‘cow’ masculine urratta ‘cloud’ pooyta ‘mourn- ʃaaɓɓaa
ing, cry’ ‘stretcher’
ʛoyra ‘tree’ masculine irɲa ‘gum’ poʛoota ‘lower ilmaamaa ‘tears’
jaw’
tuuma ‘onion’ masculine ʄapara ‘rig’ furoota ‘type of sinɗaa ‘urine’
bead’
lukkalitta ‘chicken’ masculine ditiitaa ‘sweat’ ʛaawuta ‘cough- χapnaa ‘forest’
ing’
ʛayranta ‘leopard’ masculine ɗaʔta ‘butter’ mateʔta ‘upper χaaʃaa ‘reed’
millstone’
ʛupitta ‘finger’ masculine ʄaɓɓaa ‘weed’ ʃaloota ‘cotton kaaʄaa ‘money’
thread’
kuta ‘dog’ masculine ɗakaa ‘stone’ taamta ‘branch’ elalaa ‘cowrie
shell’
oraayta ‘hyena’ masculine ɗapna ‘temple’ kalaʔta ‘spider’ kasaraa ‘dread-
locks’
χaʃʃitta ‘shoulder’ masculine ɗamayta ‘wind’ talteeta ‘she- ʄaɓɓeernaa ‘belt
goat’ for bullets’
parʄuma ‘stool’ masculine ɗikla ‘elbow’ keeʔuta ‘belch- eennaa ‘vacant
ing’ homestead’
pora ‘road’ masculine ʛayya ‘smoke’ kaankita ‘mule’ koottaa ‘but-
tocks’
karmaa ‘lion’ masculine ɗuttana ‘belly’ lanɗeeta ‘spleen’ ʄiiʄaa ‘curse’

* For (m) target, (f) is incongruent I and (p) is incongruent II

** has a (p) gender value
212       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

Appendix A: Stimulus materials in Experiment 1 (continued)

Target Distractor word conditions*

gender
picture name congruent Incongruent I Incongruent II

innaa ‘boy’ plural hiippaa ‘a riddle’ taammata ‘desert hikkitta ‘star’

bee’
ukukkaa ‘egg’ plural ararsaa ‘local beer tampoota ‘tobacco’ arrapa ‘tongue’
made for sale’
furaa ‘key’ plural aataa ‘culture’ tolloʔta ‘hump’ kirra ‘river’
filaa ‘comb’ plural oytaa ‘upper part of
tuuɓuta ‘false kolalta ‘acacia’
the compound’ banana bread’
haaʃʃullaa ‘leaf’ plural makkaa ‘sickness’ tulluppaata ‘wood kokaa ‘skin, hide’
boring beetle’
ɲirfaa ‘hair’ plural aannaa ‘milk’ keltoota ‘cattle koɗaa ‘work’
louse’
uwwaa ‘dress’ plural ellaa ‘spirit’ furoota ‘type of hakalaa ‘cabbage’
bead’**
fulaa ‘door’ plural ikkaamaa ‘selected χaayyata ‘night- keltayta ‘baboon’
seeds’ mare’
rikaa ‘a tooth plural erkannaa yoʔta ‘greed’ kawsa ‘chin, beard’
brush’ ‘message’
timɓaa ‘drum’ plural ipsaa ‘light’ χarinta ‘horizon- ɗankaa ‘throat’
tally placed fence
bar’
siinaa ‘nose’ plural olsaa ‘dream’ yaakata ‘bead’ karitta ‘belly’
χolmaa ‘neck’ plural unʛulaa ‘grain kaaʄʄata ‘t’eff’ mura ‘forest’
store from bamboo’
χopaa ‘shoe’ plural ʄorrooʛaa ‘eye moossuta ‘(piece leya ‘month’
discharge’ of) bread’
ɲaaɲɲaa ‘tomato’ plural kaariyyaa ‘devil faroota ‘luck’ kittayyaa ‘bed bug’
(ghost)’
kiʔsaa ‘cricket’ plural utaa ‘faeces’ aχawuta ‘roasted tokkayta ‘porcupine’
grain’
afaa ‘mouth’ plural fuuraa ‘fear’ puulluta ‘fer- kilpa ‘knee’
mented dough’
hirriiɓaa ‘eyelash’ plural ʛolfaa ‘bark of χompalta ‘cactus’ fanʛala ‘splinter’
trees’
akataa ‘sugar plural ɗarɗaa ‘lie, ɗuusuta ‘fart’ roopa ‘rain’
cane’ untruth’
muklaa ‘bangle’ plural hanʄufaa ‘saliva’ uffaata ‘balloon’ ɗaltayta ‘relative’
kupaʔtaa ‘tortoise’ plural marʄaa ‘hip’ hiparaata ‘bat’ hallaka ‘fat’***

* For (p) target, (f) is incongruent I and (m) is incongruent II

* * it has also been used as a distractor for the target tuuma ‘onion’ (m)
* * * it has also been used as a distractor for the target sookita ‘salt’ (m)
 Cushitic gender: experiments in Konso       213

Appendix B: Stimulus materials in Experiment 2

Target Gender Distractor word conditions*

picture name
congruent Incongruent I

innaasiniʔ ‘the boy’ plural hiippaasiniʔ ‘the riddle’ taammatasiʔ ‘the desert
bee’
ukukkaasiniʔ ‘the plural ararsaasiniʔ ‘the local beer tampootasiʔ ‘the tobacco’
egg’ made for sale’
furaasiniʔ ‘the key’ plural aataasiniʔ ‘the culture’ tolloʔtasiʔ ‘the hump’
filaasiniʔ ‘the comb’ plural oytaasiniʔ ‘the upper part tuuputasiʔ ‘the false banana
of the compound’ bread’
haaʃʃullaasiniʔ ‘the plural makkaasiniʔ ‘sickness’ tulluppaatasiʔ ‘the wood
leaf’ boring beetle’
ɲirfaasiniʔ ‘the hair’ plural aannaasiniʔ ‘the milk’ keltootasiʔ ‘the cattle louse’
uwwaasiniʔ ‘the plural ellaasiniʔ ‘the spirit (e.g. furootasiʔ ‘the type of bead’
dress’ of well)’
fulaasiniʔ ‘the door’ plural ikkaamaasiniʔ ‘the χaayyatasiʔ ‘the nightmare’
selected seed’
rikaasiniʔ ‘the tooth plural erkannaasiniʔ ‘the yoʔtasiʔ ‘the greed’
brush’ message’
timɓaasiniʔ ‘the plural ipsaasiniʔ ‘the light’ χarintasiʔ ‘the horizontally
drum’ placed fence bar’
siinaasiniʔ ‘the nose’ plural olsaasiniʔ ‘the dream’ karittasiʔ ‘the belly’
χolmaasiniʔ ‘the neck’ plural unʛulaasiniʔ ‘the grain murasiʔ ‘the forest’
store from bamboo’
χopaasiniʔ ‘the shoe’ plural ʄorrooʛaasiniʔ ‘the eye leyasiʔ ‘the month’
discharge’
ɲaaɲɲaasiniʔ ‘the plural kaariyyaasiniʔ ‘the devil kittayyaasiʔ ‘the bed bug’
tomato’ (ghost)’
kiʔsaasiniʔ ‘the plural utaasiniʔ ‘the faces’ arrapasiʔ ‘the tongue’
cricket’
afaasiniʔ ‘the mouth’ plural fuuraasiniʔ ‘the fear’ kilpasiʔ ‘the knee’
hirriiɓaasiniʔ ‘the plural ʛolfaasiniʔ ‘the bark of keltaytasiʔ ‘the baboon’
eyelash’ trees’
akataasiniʔ ‘the sugar plural ɗarɗaasiniʔ ‘the lie’ koɗaasiʔ ‘the work’
cane’
muklaasiniʔ ‘the plural hanʄufaasiniʔ ‘the saliva’ kawsasiʔ ‘the chin, the
bangle’ beard’
kupataasiniʔ ‘the plural marʄaasiniʔ ‘the hip’ ɗankaasiʔ ‘the throat’
tortoise’

* Half of the incongruent distractors are masculine and the other half are feminine nouns
214       Mulugeta T. Tsegaye, Maarten Mous and Niels O. Schiller

Appendix B: Stimulus materials in Experiment 2 (continued)

Target Gender Distractor word conditions

picture name
Congruent Incongruent

harreetasiʔ ‘the feminine ʄureetasiʔ ‘the dirt’ peeʛaasiniʔ ‘the quarrel’

donkey’
hiɓtasiʔ ‘the lip’ feminine ʄileetasiʔ ‘the stick pakataasiniʔ ‘the wide
used by old shield’
women’
irrootasiʔ ‘the moun- feminine hoollatasiʔ ‘the sheep miɗaasiniʔ ‘the cabbage
tain’ skin’ leaves’
ʄifeetasiʔ ‘the ring’ feminine ohtasiʔ ‘the cloth piʃaasiniʔ ‘the water’
(worn in the
night)’
kaawwatasiʔ ‘the glass’ feminine eetutasiʔ ‘the dinner’ pakaannaasiniʔ ‘the root crop’
lawasheetasiʔ‘the feminine fillayyaatasiʔ ‘the flea’ saaraasiniʔ ‘the poem’
mouse’
napahtasiʔ ‘the ear’ feminine haaɗitasiʔ ‘the load, paankaasiniʔ ‘the machete’
burden’
piirtutasiʔ ‘the sun’ feminine ʄooggitasiʔ ‘the mud’ χapnaasiniʔ ‘the forest’
pottaatasiʔ ‘the feminine haarutosiʔ ‘the kaaʄaasiniʔ ‘the money’
pumpkin’ revenge’
ʛapaleetasiʔ ‘the feminine ɗaltasiʔ ‘the seed’ pohaasiniʔ ‘the contri-
monkey’ bution, the
tribune’
arpasiʔ ‘the masculine hoofasiʔ ‘the hole’ tiraasiniʔ ‘the liver’
elephant’
harkasiʔ ‘the arm’ masculine ɗilasiʔ ‘the field’ pahnaasiniʔ ‘the example’
kessasiʔ ‘the chest’ masculine alkittasiʔ ‘the sisal’ torraasiniʔ ‘the speech,
the talk’
mottooʛaasiʔ ‘the truck’ masculine ukkaʃʃasiʔ ‘the husk’ toʄaasiniʔ ‘the water
droplet’
murkuʄaasiʔ ‘the fish’ masculine ʄalaʛʛittasiʔ ‘ the flat kawlaasiniʔ ‘the metal tool
stone’ for ginning’
okkattasiʔ ‘the cow’ masculine urrattasiʔ ‘the cloud’ ʃaaɓɓaasiniʔ ‘the stretcher’
paalasiʔ ‘the masculine ekertasiʔ ‘the olive’ tiyyaasiniʔ ‘the dispute’
feather’
ʛoyrasiʔ ‘the tree’ masculine irɲasiʔ ‘the gum’ χoffaasiniʔ ‘the groin’
tuumasiʔ ‘the onion’ masculine ʄaparasiʔ ‘the rig’ ɲupuraasiniʔ ‘the compo-
nent of a set
of weaving’
tuyyuuraasiʔ ‘the air masculine ammaʔittasiʔ ‘the break- χaaʃaasiniʔ ‘the reed’
plane’ fast’
Author index
Aalto, Penti 50 Cabral, Ana Suelly 58
Abbott, Clifford 142 Camden, William G. 53
Abe, Hideko Nornes 43 Cameron, Deborah 17
Adam, Lucien 60 Campbell, Lyle 58
Adelaar, Willem 60 Caramazza, Alfonso 162, 169–170, 173–174,
Adriaen, Monique 12 176, 178–179, 194, 202, 205
Alario, François-Xavier 162, 168, 180 Cassidy, Kimberley 75
Allan, Keith 75 Caudwell, Sarah 92
American Dialect Society 20 Chafe, Wallace 142
Annand, Joseph 54 Charlie, Bella 50–51
Audring, Jenny 104–105 Chatterjee, Anjan 83
Austin, Peter 111 Chumakina, Marina 98, 107–108
Clark, Michael J. 44
Baker, John R. 53 Comrie, Bernard 122
Ballard, William L. 41 Connors, Kathleen 12
Barbin, Hercule 17 Cooper, William E. 69–73, 75–78, 84
Bargh, John A. 79 Corbett, Greville G. 4, 5, 9, 87, 89, 91, 94, 96,
Baron, Naomi 24 100–102, 105, 107–109, 112, 117, 124,
Barrett, Rusty 26 135, 150, 161, 191–192
Basilico, David 83 Costa, Albert 162–163, 168, 178–179
Beardsley, Elizabeth Lane 22 Coward, Rosalind 34
Bell, Eleanor O. 75 Cowell, Mark W. 119n
Bem, Sandra L. 70 Cresson, Edith 12
Benor, Sarah B. 73, 76, 83 Cubelli, Roberto 194, 203
Bertinetti, Marco 97 Curzan, Anne 6–8, 20, 22
Birdsong, David 72–73, 83 Cutler, Anne 81
Birtalan, Agnes 50
Black, Maria 34 De Lauretis, Teresa 84
Bloomfield, Leonard 33–34, 36 De Vogelaar, Gunther 104
Bock, J. Kathryn 72 De Vos, Lien 104
Bodine, Anne 69, 71–72, 84 Deaux, Kay 77, 79
Bogoras, Waldemar 55 Demuth, Katherine 118n
Bordag, Denisa 193–194 Dešeriev, Ju. D. 95
Boroditsky, Lera 5 Diekman, Amanda B. 77, 79
Bradley, John 50–51 Dixon, R. M. W. 105, 115
Breton, Raymond 58 Dol, Philomena 101
Brown, Dunstan 112 Donohue, Mark 101–104
Brown, Lea 123 Drude, Sebastian 61–62
Browne, Gerald M. 96 Dryer, Matthew S. 123
Burani, Cristina 108 Dunn, Michael 54
Burr, Elisabeth 11–13
Butler, Judith 84 Eagly, Alice H. 77, 79
Eckert, P. 40
Ekka, Francis 49
216       Author index

Enfield, Nick 135 Iacobini, Claudio 108

Enger, Hans-Olaf 118 Inoue, Miyako 43
Evans, Nicholas 112 Institut National de la Langue Française 12
Everett, Daniel L. 44, 59–60
Jack, Frances 131
Fausto-Sterling, Anne 16–17 Janssen, Nils 168, 170
Fedden, Sebastian 112–113, 123 Jescheniak, Jörg D. 170, 180
Fenk-Oczlon, Gertrude 73, 75, 81 Johnson, Mark 9
Finocchiaro, Chiara 173, 178–179 Jospin, Lionel 11–12
Fiske, Susan T. 77
Flannery, Regina 56 Karatsareas, Petros 118n
Fleming, Luke 46 Keil, Frank C. 72
Fletcher, Katie 77 Kelly, Michael H. 72–73, 75
Fortune, David L. 62 Khosroshashi, Fatemah 30
Fortune, Gretchen 62 Kibrik, Aleksandr E. 107, 110
Foucault, Michel 17 Kimball, Geoffrey 47, 57
François, Alexandre 52 King, Ruth 12
Frank, Francine W. 33 Kirton, Jean 50–51
Klein, Jared 93
Gagua, Rasudan 96 Kölligan, Daniel 93n
Ganushchak, Lesya Y. 180 Krifka, Manfred 10
Gelfer, Marylou Pausewang 44 Kroskrity, Paul V. 46
Glück, Helmut 42, 55–56 Kulick, Don 17
Goddard, Cliff 101
La Heij, Wido 169, 194, 202
Haas, Mary R. 42–43, 47, 57 Labov, William 40, 47
Hall, Kira 17–20 Lakoff, George 115
Hantsch, Ansgar 170, 180 Leger, Rudolf 122
Harmer, L. C. 123 LeMaster, Barbara 49
Haviland, John 100 Lemhöfer, Kristin 180
Hay, Jennifer 75, 77 Lemieux, Anthony F. 83
Hayward, Richard J. 114, 191–192 Levelt, Willem J.M. 169, 193, 202
Healy, Bernard 26 Levy, Roger 73, 76, 83
Hegarty, Peter L. 77–80, 83 Lewis, Laurie L. 79
Heim, Stefan 180 Lieberson, Stanley 75
Henry, Victor 60 Lindström, Eva 119–121
Hillenbrand, James M. 44 Livia, Anna 14–15, 17, 24
Hilton, Denis J. 76 Lounsbury, Floyd 142
Hock, Hans Henrich 47 Lunau, K. 41n
Hockett, Charles F. 89, 132 Luthin, Herbert 47
Hoff, Berend J. 58
Hoffman, Curt 77 Maass, Anne 83–84
Holisky, Dee Ann 96 McConnell-Ginet, Sally 6, 27, 31, 40, 71, 79,
Hough, Carole 75 84
Hurst, Nancy 77 McDonald, Janet L. 72–73, 76, 83
McQueen, Grant 77, 83
Major, Brenda 77
 Author index       217

Malkiel, Yakov 69–71, 74–78, 83–84 Roberts, Marjorie 26

Matasović, Ranko 87–88 Rochefort, Charles de 39, 58
Mathiot, Madeline 26 Rodina, Yulia 118n
Matossian, Lou Ann 30 Roelofs, Ardi 169, 202
Mel’čuk, Igor 92 Rosch, Eleanor 9, 72
Mervis, Carolyn B. 72 Rose, Françoise 60
Meyer, Antje S. 169, 202 Ross, John Robert 69–73, 75–78, 84
Mikelson, Kelly S. 75 Roudy, Yvette 12
Mikos, Victoria A. 44 Rovai, Francesca 121n
Miozzo, Michele 162, 169, 178 Rudes, Blair A. 26
Mithun, Marianne 4, 47, 83, 121, 145 Russo, Aurore 83
Modi, Bharati 115
Mollin, Sandra 69, 73–74, 76, 83 Sag, Ivan 9
Motschenbacher, Heiko 42, 56 Sans, Pierric 60
Mous, Maarten 191–192, 195 Sapir, Edward 47
Muchnik, Malka 36 Saville-Troike, Muriel 47, 57
Muysken, Pieter 60 Schapper, Antoinette 124
Scheffler, Harold W. 41
Naito, Maho 53 Schiller, Niels O. 104, 168–169, 173, 176,
Neuberg, Steven L. 77 179–180, 194, 205, 207
Nevalainen, Terttu 3 Schriefers, Herbert 161, 163–164, 169–170,
Newman, Michael 30 173–174, 176–177, 179–180, 193–194,
Noël, Georgine 14–16 205–207
Norton, F. J. 123–124 Seifart, Frank 105–106
Sen, Sukamara 41
Ochs, Elinor 47 Sharoni, Lee’at J. 75
O’Donovan, Veronica 18 Sherzer, Joel 40
Orkaydo, Ongaye Oda 193–195, 203 Simpson, Adrian P. 44
Orne, Martin 76 Skorik, Piotr Ja. 40
Southwood, Helen 83
Palancar, Enrique 123–124 Spivak, Michael 25
Pandharipande, Rajeshwari 47 Stein, David E. S. 121n
Parker, E. M. 114 Stolz, Christel 121
Parry, Sir William Edward 44 Suitner, Catarine 84
Pastre, Genviève 17, 26
Pechmann, Thomas 193–194 Taylor, Allan R. 56
Pinker, Steven 72–73, 83 Taylor, D. 58
Plaster, Keith 113, 115–116 Terrill, Angela 99, 121, 123
Polinsky, Maria 113, 115–116, 118 Thornton, Anna M. 108, 110, 118, 121
Pollard, Carl 9 Treichler, Paula, A. 33
Potter, Simeon 74 Tryon, Darrell T. 52
Pratto, Felicia 79 Twain, Mark 4

Queen, Robin 26 Vallejos Yopán, Rosa 58–59

Verdonschot, Rinus G. 180
Raumolin-Runberg, Helena 3 Visser, Willem 121
Ribeiro, Eduardo 62
218       Author index

Wagner, Gunter 41 Wickler, W. 41n

Watson, Nila 77 Wright, Saundra 75–78, 81, 83
Wechsler, Sara 118n
Wegener, Claudia 123 Yokoyama, Olga 41
Whissell, Cynthia 75
Whitehead, Neil L. 58 Zaliznjak, Andrej A. 90, 92
Whittaker, G. 47
Language index
Afar (Qafar) 114–115, 118 Five Nations languages 142–143
Alagwa 194 French 5–6, 9, 11–12, 14–17, 26, 91, 95–96,
Algonquian languages 111, 124 115, 162, 178
Alutor 54–55
Arabic 119n German 4–6, 9–10, 161–170, 173, 176,
Arawak 58 178–180
Archi 98–99, 107–108 Germanic languages 125, 178
Austro-Asiatic languages 111 Gros Ventre 45, 56
Austronesian languages 124 Gujarati 115
Awetí 45, 61–62 Guugu Yimidhirr 100–101, 104

Bagwalal 110–111, 114, 116 Hebrew 36

Batsbi 91, 95–96 Hijra Boli 19
Bayso 191–194 Hindi 17–20
Bengali 41 Huron 141–142
Bésɨro (Chiquitano) 60
Biblical Hebrew 121n Indo-European languages 4, 8, 116, 119n
Bininj Gun-wok 112, 122 Iraqw 192, 194
Boni 194 Irish Sign Language 43, 45–46, 48–49, 63
Burmeso 101–104 Iroquoian languages 1, 121, 141–143
Burunge 194 Island Carib 46, 57–58
Italian 97–98, 108–109, 121, 162, 173, 176,
Cappadocian Greek 118n 178, 203
Catalan 162, 178
Cayuga 142–143 Japanese 3, 43
Central Pomo 131
Cherokee 141–142 Kala Lagaw Ya 111
Chiquitano 60 Kalmyk 50
Chukchi 1, 39, 43–46, 54–56 Kannada 111
Chukotko-Kamchatkan languages 54 Karajá 62–63
Classical Armenian 92–94, 96 Kerek 54
Croatian 163–164, 168, 173, 176, 178–179 Khoisan 111
Cushitic languages 2, 191–194 Koasati 45, 47, 57
Kokama-Kokamilla 45–46, 58–59
Dirayta 194 Konso 2, 191–208
Diyari 111 Koryak 54
Dravidian languages 111, 125 Kuot 118–121
Dutch 104–105, 161–163, 169–180, 193–194 Kupto 122–123
Dyirbal 113, 115–116 Kurux 45, 49–50
Kushi 123
Emegul 47
Emesal 47 Latin 108–109, 121n
English 1, 5–11, 20–37, 69–86, 180 Lavukaleve 99–100, 121, 123
220       Language index

Macedonian 108 Romanian 91, 93–96

Maltese 121 Russian 55–56, 89–91, 113, 116–119
Mawng 123
Maybrat 101 Sanskrit 47
Mele-Fila 101 Savosavo 123
Mian 105, 112, 123 Seneca 142–143
Miraña 105–106 Sesotho 118n
Mohawk 1, 4, 83, 131–160 Southern Bole-Tangale languages 122
Southern Iroquoian 141–142
Ngan’gityemerri 105 Spanish 123, 162, 178
Niger Congo languages 111, 124 Sumerian 47
Northern Iroquoian 141–143 Swedish 25

Oceanic languages 53 Tamil 1, 111

Old English 22 Tangoan 45–46, 52–54, 63
Old Nubian 96–97 Tariana 105
Oneida 142 Thai 43
Onodaga 142 Tidore 105
Ts’amakko 194
Palana Koryak 55 Tsez 122
Papuan languages 121, 123–124 Tsova Tush 91
Parji 111 Tuscarora 141, 143
Pidgin Carib 58 Tutuba 52–53
Pirahã 44–46, 59–60
Portuguese 62 Walman 123
Prakrit 47 Wendat 141–142
Proto-Five-Nations 142 West Frisian 121
Proto-Iroquoian 141–142 Wyandot 142
Proto-Northern-Iroquoian 142
Proto-Oceanic 53 Yana 47
Yanyuwa 45–46, 50–52
Qafar 114–115, 118 Yawuru 123
Yuchi 41
Rendille 194 Yup’ik 54, 131
Retuarã 105
Subject index
accusative case 93 classification, gender as 89
adjectives 108–109 classifiers 105–106
age and gender assignment 10 cognitive associations 27–28
agency 77 combined case system 101, 104
agents 143–144 combined gender system 104
agreement conceptual baggage 27–28, 79
––feminine forms 146–158 conservatism, origins of gender dialects 46
––gender systems 132–133 correlates of gender 47
––and referential pressures 8–11 cross-expressing 17
agreement classes 90–91, 96, 102–104 culture and language 131
Agreement Hierarchy 9
agreement targets 3–4, 6 definite nouns, Konso experiments 203–206
anaphora 8 deictic reference 8–10
androcentrism and word order 70–71 demonstrative pronouns 8
animals 134–135 determiner competition hypothesis 178
animate first rule 72–73 dialect merger 46
anyone 29–30 dialect, use of term 40
areal variation 52 dialect variation see gender dialects
assumptions, about gender 27–28 Differential Object Marking 124
avoidance language 50 Differential Subject Marking 124
awareness of gender 1 diversification, origins of gender dialects 46
division of labor, and gender
bare noun naming 169, 180, 198–203, 206 assignment 11–13, 36
behavioral effects of Det NP production 180 double-gender 12, 36
binomial phrases drift, origins of gender dialects 46
––androcentrism 70–71
––change over time 74 E/e 25
––effects of stereotyping 78–80 em/eir/eirs/eirself 25
––empirical studies 72–74 enforced gendering 21–25
––flexibility 73–74 epicene forms, neologisms 24–25
––frequency of occurrence 73 epicene nouns 5, 11
––name order 74–78 euphemism 45–50
––phonology 72–76, 81–82 everyone 29
––proximity 75 expressive usage 13–20, 26–27
––semantics 72–73, 76–78, 80 Ey/ey 25
––situation-specific 74
––word order preferences 69–71 false generics 34–35
birdsong 41n familiarity and name order 78
boys’ use of female dialect 60 feminine alternatives 36
British National Corpus 73 feminine forms 136–143, 146–158
feminine generics, enforced gendering 
ceremonial speech 135, 150 31–33
children’s acquisition of rules 118n feminine-indefinite form 137–138, 142–145
class markers 106 feminine-zoic form 137–140, 142
222       Subject index

feminism 11–13, 35 ––derivation 45
feminization 12–13, 36 ––distribution map 48
feminizing affixes 35 ––documentation 40
formal assignment 113–125 ––lexical differences 45
formant frequency 44 ––morphological differences 45
France, linguistic politics 11–13 ––origins 46
French literature, minority identities 14–17 ––structural typology 44–45
frequency (hz) 44 ––typology 42–43
frequency of occurrence ––usage 42–43
––and name order 74–76 gender disagreement 9–10
––and word order 73 gender equity 36–37
frozen binomials 69–71 gender feature competition hypothesis 164,
168–169, 174, 177
Garifuna people 39 gender feature selection
gay men 17, 26 hypothesis 205–206
gender gender ideologies 30
––correlates of 47 gender-incongruence 62
––derivation of term 89 gender marking 19, 35
––use of term 3 gender-neutrality 22–25, 36
gender and number 191–196 gender separation 63
gender assignment 4–5, 9–10 gender stereotypes 27–28, 32, 77–80
––biological and social 40–42 gender systems
––and characteristics 5–6 ––agreement 132–133
––and division of labor 11–13, 36 ––Cushitic 191–193
––formal assignment 113–125 ––distribution 124–125
––human beings 8 ––grammatical vs. natural 3–8
––loanwords 121 ––interactions with sex and sexuality 6
––morphological assignment 116–121 ––reform strategies 35–37
––phonological assignment 114–116 ––see also gender typology
––recategorization 121–124 gender typology
––semantic assignment 110–113 ––analysis problem 89–90
gender attribution, avoiding 22 ––assignment 110
gender congruency effect 161–164, 193–194 ––distinction across lexemes 106–110
––discussion of experiments 177–181 ––distribution 124–125
––indefinite determiner NPs experiment, ––evidence for 90–91
Dutch 169–174 ––formal assignment 113–125
––indefinite determiner NPs experiment, ––morphological assignment 116–121
German 164–169 ––phonological assignment 114–116
––Konso experiments 196–208 ––recategorization 121–124
––possessive adjective NP production ––semantic assignment 110–113
experiment, Dutch 174–177 ––unique distinction 97–100
––stimulus materials 184–189 ––word classes 100–106
gender dialects 30 ––see also gender systems
––acquisition 43 gender variation, phonetic expression 44
––areal variation 46 genderlect 41–42
––bidialectalism 43 ––see also gender dialect
––categorical 40 generics 13, 31–33, 105, 136
 Subject index       223

genus alternans 94 kinship terms 41, 143–146

grammatical categories, development 131
grammatical gender 4–6, 21 labeling 17, 36
grammatical gender systems, vs. natural  ladies and gentlemen, word order 74–75
3–8 language
greetings cards, name order 77 ––and culture 131
Gricean pragmatics 76 ––and social status 47
––written 47
he generic use 29, 71 language ideologies 30
hen 25 lesbians 17, 26
heterosexuality, presumptions of 28 lexical differences, gender dialects 45
hijras 17–20 liminal identity 14, 17–18
Histoire naturelle et morale des Iles linguistic politics 11–13, 25
Antilles 39 linguistic processing, effects of grammatical
hit neuter pronoun 22 gender 161–162
human beings 8 linguistic strategies 35
humanity, as male 34–35 loanwords 55–56, 62, 87, 121
hybrid nouns 9–10
Mädchen gender agreement 10
ideologies 30 male-first hypothesis 70
inanimates 134–135 male humanity, as humanity 34–35
indefinite determiner NPs experiment, -man suffix, masculinity 36
Dutch 169–174, 177 maps
––stimulus materials 187–189 ––distribution of gender dialects 48
indefinite determiner NPs experiment, ––Konso 208
German 164–169, 177 markedness 73
––stimulus materials 184–186 masculine forms 133–134
indefinite forms, extension 143–145 masculine generics 29–30
Independent Network (IN) model 202 me-first principle 70–71, 77–78
inflection classes 102, 117–118, 120 minority identities
inflectional morphology 89 ––gay men 17
inquorate gender values 96 ––hijras 17–20
interference effects 169–170, 193–194, ––intersexualism 16–17
202–203 ––lesbians 17
internal change, origins of gender ––transsexuals 14–16
dialects 46 morphological assignment 116–121
internal evidence, feminine-indefinite morphological differences 157
form 143–145 ––gender dialects 45
internet, name order 77
intersexualism 16–17 name order
it in reference to humans 22–23 ––effects of stereotyping 78–80
––phonology and frequency 74–76
job titles 11 ––and popularity 81–82
Journal of a second voyage for the discovery names
of a north-west passage from the Atlantic ––form of 143
to the Pacific (Parry) 44 ––as gendered 22
––number of syllables 75, 81
224       Subject index

––phonological characteristics and psycholinguistics 2

popularity 82 ––behavioral effects of Det NP
––statistical differences 75 production 180
––structure, and word order 81 ––discussion of experiments 177–181
natural gender ––further research 206
––in English 20–21 ––gender congruency effect 161–164
––as myth 18–19 ––indefinite determiner NPs experiment,
natural gender systems Dutch 169–174
––as potentially misleading 7–8 ––indefinite determiner NPs experiment,
––vs. grammatical 3–8 German 164–169
neologisms, gender-neutral 24–25 ––neurocognitive effects in gender
neuter forms 134 ­classification task 180
non-autonomous values 96–97 ––possessive adjective NP production
notional gender systems 3, 5–7, 21, 79 experiment, Dutch 174–177
noun classifiers 105–106 ––stimulus materials 184–189
noun phrase (NP)
––any/every 29 quantifiers 29–31
––pronoun choice 21
nouns, assignment of gender 1, 4–5 recategorization 121–124
number 122, 134–135, 151 referential pressures and agreement 8–11
number and gender 191–196 referentially linked pronouns 5
reform strategies 35–37
parent, gender assignment 32
patients 143–144 selection mechanisms 180
personal pronouns 8 selection of grammatical features 181
phonetic expression, gender variation 44 selection of lexical nodes 181
phonological assignment 114–116 self-attribution 13–17, 20
phonological factors, sex identi- semantic assignment 110–113
fication 44–45 semantics, word order 72–73, 76–78, 83
phonology separation of genders 63
––name order 74–76 sex
––word order 72–73, 81–82 ––as permanent personal characteristic 23
physical separation of genders 63 ––use of term 3
picture-word interference paradigm 163 sex-differentiability, threshold 122–123
plural masculine forms, as generic 13 sex identification, phonological
portraits 83–84 factors 44–45
possessive adjective NP production sex reassignment 14
experiment, Dutch 174–177 sexually mixed groups, gender
––stimulus materials 187–189 assignment 13
power source first rule 77 shamanism 55
prescriptive grammar, pronoun choice 5 she 20–21
prestige in gender dialects 47, 53 significance of gender divisions 3
presumptive leaps 27–29 singular feminine generics 31–33
pronoun choice 5, 6–9, 21 situation-specificity, word order 74
pronoun switching 26–27 social status, and language 47
prototypical first rule 71–72, 78 sociocultural gender 3, 8–11
proximity and word order 75
 Subject index       225

sociolinguistic variables, stigmatized and WEAVER++ model 202–203

prestige 47 wifman 22
Spivak pronouns 25 word order 75
stereotypes 27–28, 77 ––androcentrism 70–71
suffixes 35–36 ––change over time 74
syllables, number and word order 72, 75, ––effects of stereotyping 78–80
81 ––empirical studies 72–74
––flexibility 73–74
taboos 50, 55 ––and frequency of occurrence 73
“The Awful German Language” (Twain) 4 ––frozen binomials 69–70
they 22, 29–31 ––names 74–78
third sex, hijras 17–20 ––phonology 72–76, 81–82
transsexuals 14–16, 23 ––preferences 69–71
––proximity 75
United States, promotion of gender- ––semantics 72–73, 76–78, 83
neutrality 36 ––situation-specific 74
unknown sex 29 ––and word popularity 81–82
World Atlas of Language Structures
variable gender agreement 9 (WALS) 124
variations in gender assignment 5 written language 47
verb forms, gender-marked endings 19
visibility, males and females 34 ze/zir/zirs/zirself 25