clac CÍRCULO de

clac

lingüística
aplicada a la
comunica
ción

13/2003

RELEVANCE VERSUS CONNECTION:

DISCOURSE AND TEXT AS UNITS OF ANALYSIS

Joaquín Garrido

Universidad Complutense de Madrid

joaquin garridoATccinf ucm es

Abstract

The interpretation of speech can be dealt with in terms of grammar, if

grammatical categories are extended beyond the sentence. After briefly
reviewing speech acts, discourse acts and relevance, two units are thus
posited, the discourse and the text, and they are defined in strictly linguistic-
theoretical terms, by means of a principle of connection. Following this
principle, sentences are linked using additional information, which is not
explicitly represented in the sentence. The resulting units are structured
sequences of sentences, called discourses, which are, in turn, linked in a
structured way into larger units, called texts.

Key words: discourse, text, connection, grammar

© 2003 Joaquín Garrido

CÍRCULO de Lingüística Aplicada a la Comunicación (CIRCLE of Linguistics Applied to
Communication/ clac) 13, 13-22.
Universidad Complutense de Madrid. ISSN 1576-4737. http://www.ucm.es/info/circulo/
 14

1. Speech acts
As everybody knows, pragmatics is the standard solution to the limits of
sentence grammar1. In a way, the difference between grammar and pragmatics
is an updated version of the distinction between system and use, or between
langue and parole. A major contribution in the field of pragmatics is the
concept of the speech act. Speaking is not only saying things about reality, it is
also doing things about it, and there is even nothing but action in some cases.
There are two main arguments in favor of speech acts. One is the existence of
perfomative expressions, such as ‘I promise to come’. The second argument is
the existence of questions and orders or requests, which are not used to say
something about reality, but rather as tools to do something. The first point
results in the classification of speech acts in terms of the corresponding
performative expressions, for instance, commissive speech acts in terms of the
performative ‘I promise’. The second one leads to the distinction between
sentences (declarative, interrogative, and imperative) and utterances
(statements, questions, and requests), and poses the problem of the so-called
indirect speech act, where a request, for instance, may be made by means of a
declarative sentence, instead of an imperative one (which would be an
example of a direct speech act).
Actually, performative expressions are the result of a lexical property:
there are verbs that describe acts which are socially defined as being
performed by means of speech. For instance, promising is a social act,
whereas the linguistic phenomenon is the fact that a given expression (‘I
promise you to go’) may be understood as an act of promising. Performative
expressions are not different from others (such as ‘I will go’) in that they are
understood as actions, but rather, in the fact that they may refer to very action
which is performed by using them (performative expressions are self-
referring; see Garrido 1997: 204). We are then left with the other kind of
speech acts: the expressions that are understood as statements, requests, or
questions, whether they are declarative or imperative or interrogative (that is,
both direct and indirect speech acts).

1
This version of the plenary lecture given at the conference “European Union and its
multilingual dimension”, held at the Vysoká škola ekonomická in Prague in 1998, is part of
the research project PB97-0266 of the Spanish DGESIC [Dirección General de Educación
Superior e Investigación Científica]. Published in Hlavičková, Vlasta (ed.) 1998: EU a její
multilingvální dimenze [European Union and its multilingual dimension], Praga: VSE
[Vysoká škola ekonomická v Praze], pages 5-14, ISBN 80-7079-526-3.

clac 13/20033
 15

2. Discourse acts
Roulet (1995: 321) locates speech acts in a wider, hierarchical structure,
where dialogues are analyzed into one or more ‘exchanges’, and where
exchanges consists of ‘moves’, which are in turn linked to each other by “the
initiative and/ or reactive illocutionary functions which are generally attributed
to speech acts”, such as a question move followed by an information/ response
move. Each move “consists of a main ‘discourse act’, possibly accompanied
by exchanges, moves, or acts that are subordinated to it; those constituents are
linked by so-called interactive relations (preparation, justification,
reformulation)” (Roulet 1995: 321). These units may be extended to written
communication (see Kroon 1997). In the first version of this model (Roulet et
al. 1985, chapter 1), there are two other higher units, interactions and
transactions. Interactions are made up of transactions, and transactions consist
of moves.
The problem is, as Roulet (1995: 321) points out, that a discourse act
“does not necessarily coincide with a syntactic sentence”. There is no general
procedure that assigns a discourse structure to a given stretch of talk.
Moeschler (1996: 184) does posit “linking constraints” (“contraintes
d’enchaînement”), which determine the degree of cohesion and coherence of a
given sequence; but one has to know that a given constituent is an act, for
instance, in order to apply rules such as the one proposed by Moeschler (1996:
189): “For every sequence of acts <A1, A2> in a move M, project a category
mA [main act] and a category sA [subordinated act]”. Besides, it is not clear
how to decide which one of the two is to be the main act.
Problems in establishing the hierarchy of units in a given sequence lead
Moeschler (1996: 194) to substitute an approach based on relations between
propositional forms for this approach based on discourse constituents. Instead
of coherence as a result of discourse relations, Moeschler (1996: 214) applies
Sperber and Wilson’s (1986) relevance approach to relations between
propositional forms (propositions), as expressed by means of utterances and of
contextual information. Relevance applies to utterances as well as to
discourses. A discourse is a “non-arbitrary sequence of utterances” (“suite non
arbitraire d’énoncés”) (Reboul and Moeschler 1996: 90). An utterance is the
result of a specific use of a sentence (“phrase”), whereas a sentence is a
“complete grammatical sequence” (Reboul and Moeschler 1996: 64). We are
thus back to the standard approach, where language and speech are apart,
belonging to different components, grammar and pragmatics. The only
difference is that speech is allowed to have larger units than the utterance, but

clac 13/2003
 16

the interpretation process, based on relevance, is the same for both utterances
and discourses. Before dealing with this double standard (of grammatical units
and pragmatic or discourse units), let us briefly examine the explanation based
on relevance.

3. Relevance
Sperber and Wilson (1986) offer a general theory of human cognition
and communication. When speaking, they argue, people ellicit conclusions
which are valid only for the specific case where they arise. In order to arrive at
those conclusions, people use data which are valid only for these specific
occasions. Both direct and indirect speech acts are explained in this way. For
instance, ‘It is rather late’ as an answer to ‘Do you want a drink?’ is to be
understood in terms of the contextual information that when it is late one does
not want a drink. The conclusion is that ‘It is rather late’ is to be interpreted
as a negative answer, in this particular situation.
There is a general principle, the principle of relevance, that explains not
only indirect but also direct cases. According to Sperber and Wilson, there is
always the need to look for additional information that will allow for
conclusions otherwise absent. This property of rendering additional
conclusions which compensate for the effort in bringing them about is called
relevance by Sperber and Wilson. They argue that relevance is guaranteed:
every utterance carries the presumption of being relevant. In order to be
relevant, an indirect speech act, that is, the corresponding utterance, must be
interpreted in these terms: something extra must be taken into account, so that,
for example, a request might be understood out of a declarative or an
interrogative sentence (‘I have no money left’ or ‘Do you have any money?’).
The problem of the second type of speech acts (statements, questions, and
requests, and their problematic indirect relation to sentence types, declarative,
interrogative, and imperative) is thus solved. There is always an interpretation
process, a more elaborate one in indirect cases. There remains the problem of
how this general interpretation takes place.

4. Two ways to the solution

The key to the process is taking into account the right additional
information. Both speaker and hearer must follow the same path in order to
arrive at the same interpretation. Wilson and Sperber (1993) (and Blakemore

clac 13/2003
 17

1995) argue that the non-propositional components of the utterance’s

linguistic form place constraints on the process of interpretation, that is, they
represent procedural information (versus the propositional information
encoded by the other components in the utterance’s linguistic form). But
rather than being a constraint on the interpretation, each aspect of the
linguistic expression contributing to its connection forces a given
interpretation, because connections are inevitable (see Garrido 1997: 242):
linguistic units are built up so that they are necessarily linked with additional
information. It is not that additional information may be provided, but rather
that units don’t work, that they are not understood, unless some specific
information is added. There is only a way to understand them, that is, a way to
connect them (when there is more than one, either ambiguity or
misunderstanding arise).
This amounts to disregarding the difference between sentence and
utterance. Sentences are made so that they fit in specific sequences or
discourses, that is, they require certain additional information which is either
available in the preceding sentences or it must be added to them. The way to
connect a sentence within the discourse where it belongs is marked by
grammatical means (see Garrido 1998a): in other words, the structure of the
sequence of sentences belongs to the grammar. This solution thus requires
both a grammatical unit larger than the sentence and a principle of
interpretation different from the principle of relevance. In fact, both are
different sides of the same coin.

5. Discourse and text as units

Sentences are linked in sequences which may be called ‘discourses’.
These units have a semantic representation, a “discourse representation
structure”, in terms of Kamp and Reyle (1993: 59), which is incremental: “in
order to make sense of a discourse-internal sentence the interpreter must
connect it with the interpretation he has assigned to the sentences preceding it”
(Kamp 1995: 254). In turn, discourses are organized into larger structures, up
to texts. And texts are not defined in terms of coherence (as proposed by
Beaugrande and Dressler 1981) or relevance (Blass 1990; see also Wilson
1998), but by the connection existing between their component discourses.
Similarly, a discourse is not defined in terms of contextual information, but it
is the result of connecting its component sentences. In order for this
connection to take place, additional information must be included. This

clac 13/2003
 18

happens whenever a unit is connected to others. Connection is the general

process that accounts for syntactic structure itself. In every complex unit there
is information that is not represented by the component units, but rather added
to them so that the result is the complex unit itself. Frequently this additional
information is added as default, but it may also originate in preceding units.
Thus words are linked to each other into phrases, phrases into clauses, clauses
into sentences, sentences into discourses, and discourses into texts. Text types,
or genres, account for the way discourses are connected; styles, or discourse
types, organize in general ways the connection between sentences within
discourses.
When we examine an example, we reverse the real situation. We
interpret every sentence, adding contextual information to it. In real life
situations, every sentence makes salient the piece of available information that
is required in order to link the sentence interpretation (that is, to link the
sentence itself) to the discourse and the text it belongs to. In the following
example, the first speaker A says its first sentence, A1, whereas the second
speaker, B, says two sentences, B1 and B2 (notice that a sentence may or may
not include a clause structure):
A1: ¿you want a drink?
B1: no
B2: it’s late
The only way to make sense out of the three sentences is to link them in two
steps: B1 and B2 are the answer to A1. There are two discourses, to be
interpreted as an invitation and as a refusal to the invitation, respectively: D1
(consisting of A1) and D2 (consisting of B1 and B2).
A different possibility would be the following:
A1: ¿you want a drink?
B1: no
A2: it’s late
Here, the first speaker accepts the refusal by giving a possible explanation for
the refusal. There are three one-sentence discourses, and the three of them
may be followed by another piece of conversation (and they be preceded by
another one, for that matter). If it were so, there would be the preceding
discourse D1, the actual discourse D2 (consisting of three one-sentence
discourses just mentioned, D3 (A1), D4 (B1) and D5 (A2)), and the following
discourse, D6 (where discourses are given arbitrary numbers, D1, D2, and so
on).
A third possible discourse structure would be:

clac 13/2003
 19

A1: ¿you want a drink?

B1: no
B2: let’s go
Here, the second sentence that B says is to be linked to the total preceding it.
Thus, A1 and B1 make up D1 (a complex discourse consisting of the one-
sentence discourse D2, the invitation, and the one-sentence discourse D3, the
refusal). B2 is the continuation to D1, that is, it makes up another discourse,

┌ D2: A1: ¿you want a drink?

D4.

┌D1:┤
┤ └D3: B1: no
└D4: B2: let’s go
D1 and D4 make up another discourse, a larger one, which may be linked to
other parts of the conversation (which itself is the text unit). Or the whole
conversation may consist of D1 and D4, that is, the whole text might be this
short sequence.
The point is that, for every example, there is only one way to link the
sentences. If there were other ways, there would be either ambiguity or
misunderstanding, or both.

6. Connection
Instead of two components, a grammatical one and a pragmatic one,
there is just one, up to texts. Instead of syntax and relevance, there is only
connection. Connection may be very generally described by the following
principle of connection: every unit is connected to the others by means of
additional information, which is not explicitly represented. Instead of every
utterance being relevant, every sentence, in fact every unit, from the word up
to the text, is guaranteed to be understandable, that is, except for possible
ambiguities or misunderstandings, the interpretation of every unit has to be
integrated in a wider cognitive unit, according to the unit’s position in the
corresponding wider syntactic structure. The sentence’s inner structure
establishes its outward links, that is, every sentence is designed so that its
inner structure fits the outer, discourse structure where the sentence is located
(see Garrido 1998b).
This approach differs from relevance theory in two ways: first, instead
of utterances as uses of sentences, discourses are made up of sentences (and in
turn make up texts); secondly, instead of warranting relevance for the

clac 13/2003
 20

utterance, each sentence is part of a discourse, and its semantic representation

is a part of the discourse representation structure. As a result, instead of adding
contextual assumptions to the utterance, each sentence is built so that it is
integrated into the discourse by means of additional information. This process
of adding information is governed by the discourse structure, that is, by both
the internal structure of the component sentences and the discourse structure
itself. The explanation based on connection is not based only on the hearer’s
interpretation, but it accounts for the way the speaker designs an interpretation
for each sentence in the discourse, and, when the speaker succeeds, for the
way the hearer is forced to construct an interpretation following the path
designed by the speaker.
Let us review the kind of information that is always arrived at in every
category. Adding whatever information may be demanded in the process, the
connection of two sentences requires that a speaker (and possibly, an
addressee) of these sentences be established (for instance, the same speaker
for both sentences or a different one for each sentence). Similarly, linking two
discourses requires the specification of its component relation, such as the
rhetorical relations specified by Lascarides, Copestake and Briscoe (1996: 48).
For instance, a discourse may make up a question, and the following discourse
may serve as an answer to that question. Notice that discourses may consist of
just one sentence. Discourses are recursively integrated into larger discourses,
and those into text structures.
Text structures are patterned after text types, which are usually said to
have a social function, “a shared public purpose” (Swales 1987: 13). Actually,
text structure reveals in linguistic terms what is generally considered to be
extralinguistic, that is, this “public function” (for an example of press news,
see Garrido 1998c). Mann, Matthiesen and Thompson (1992: 41) posit three
different levels of structure: “holistic” or text structure, “relational” structure
(where rhetorical relations obtain) and “syntactic” structure, or sentence
structure. Instead of this three-layer structure, there is only a hierarchical one,
from text to discourse, from discourse to sentence, or, rather, from sentence
structure to discourse structure, and from discourse structure to text structure.
As we have seen, each sentence is interpreted by being integrated into a
discourse representation structure, in the same same way each discourse is
intepreted as it is connected into the text representation structure. Actually,
sentences and discourses are designed so that they are interpretable in this
way. Similarly, each text is designed so that it is integrated into a cognitive
representation of an action, since texts are part of wider structures, social

clac 13/2003
 21

actions. A contract, for instance, is a part of a larger social institution, of

buying and selling property and services. When a text is so connected into the
wider social unit, additional information plays a role, too. We understand a
text when we are able to put its intepretation into a wider cognitive
representation, that of the action of which the text is a part. There is nothing
extralinguistic in texts, but rather, texts, as linguistic objects, are linguistic
parts of non-linguistic objects, human actions.

References

Beaugrande, Robert de, and Dressler, Wolfgang, 1981, Introduction to text

linguistics, London, Longman.
Blakemore, Diane L., 1995, “Relevance theory”, in Verschueren et al.,
443-452.
Blass, Regina, 1990, Relevance relations in discourse, Cambridge, Cambridge
University Press.
Connolly, J. H., Vismans, R. M. and Butler, Christopher S., eds., 1997,
Discourse and pragmatics in functional grammar, Berlin, Mouton de
Gruyter.
Kamp, Hans, and Reyle, Uwe, 1993, From discourse to logic, Dordrecht,
Kluwer.
Kamp, Hans, 1995, “Discourse representation theory”, in Verschueren et al.,
253-257.
Kroon, Caroline, 1997, "Discourse markers, discourse structure and
Functional Grammar", in Connolly et al., 17-32.
Garrido, J., 1997, Estilo y texto en la lengua, Madrid, Gredos.
Garrido, J., 1998a, “Discurse structure in grammar”, Estudios Ingleses de la
Universidad Complutense, 6, 33-47.
Garrido, J., 1998b, “Syntaxe de la phrase dans le discours”, Travaux de
Linguistique 36, 37-46.
Garrido, J., 1998c, “Diskurs zwischen Satz und Text am Beispiel von
Pressentexten”, in U. Helfrich and H. Klöden, eds., Mediensprach in der
Romania, 131-147, Egert, Wilhelmsfeld.
Lascarides, Alex, Copestake, Ann, and Briscoe, Ted, 1996, “Ambiguity and
coherence”, Journal of Semantics 13, 41-66.

clac 13/2003
 22

Mann, William C., Matthiessen, Christian M. I. M., and Thompson, Sandra

A., 1992, “Rhetorical structure theory and text analysis”, in Mann, W.
and Thompson, S. A., eds., Discourse description, 39-78. Amsterdam,
Benjamins.
Moeschler, Jacques, 1996, Théorie pragmatique et pragmatique
conversationnelle, Paris, Armand Colin.
Reboul, Anne, and Moeschler, Jacques, 1996, “Faut-il continuer à faire de
l'analyse de discours?”, Hermes 16, 61-92.
Roulet, Eddy, 1995, “Geneva school”, in Verschueren et al., 319-323.
Roulet, Eddy, Auchlin, A., Moeschler, J., Rubattel, C., and Schelling, M.,
1985, L'articulation du discours en français contemporain, Bern, Lang.
Sperber, Dan, and Wilson, Deirdre, 1986, Relevance, Oxford, Blackwell.
Swales, J. M., 1990, Genre analysis. Cambridge, Cambridge University Press.
Verschueren, J. F., Östman, J.-O. and Blommaert, J. eds., 1995, Handbook of
pragmatics, Amsterdam, Benjamins.
Wilson, Deirdre, 1998, "Discourse, coherence and relevance: A reply to
Rachel Giora", Journal of Pragmatics 29, 57-74.
Wilson, Deirdre, and Sperber, Dan, 1993, “Linguistic form and relevance”.
Lingua 90, 1-26.

clac 13/2003