Content uploaded by Jacques Moeschler
Author content
All content in this area was uploaded by Jacques Moeschler on Dec 18, 2014
Content may be subject to copyright.
Speech act theory and the analysis of conversations.
Sequencing and interpretation in pragmatic theory
Jacques Moeschler
Department of Linguistics
University of Geneva
1. Introduction
Conversation has recently become a focus of interest for speech act theory and
several proposals have been formulated concerning the possible extension of
speech act theory to the analysis of conversation. This debate (cf. Searle et al.
1992) has to be interpreted as a reactive move rather than as a natural extension
of the domain of speech act theory. Nevertheless, this reaction, either sceptical
(cf. Searle 1992) or optimistic (cf. Dascal 1992, Vanderveken 1992 and 1994),
has brought interesting issues which contrast with the various attempts by
linguists at extending speech act theory to the domain of discourse
1
. The first
purpose of this paper is to explicit the divergence between philosophers and
linguists about the possible extension of speech act theory to discourse analysis.
This paper has another purpose : it also deals with the possible domain of
pragmatic theory with respect to discourse analysis. I shall argue that the main
purpose of discourse analysis is the definition of necessary and sufficient
2 MOESCHLER
conditions for sequencing and interpretating utterances in discourse. I claim that
these two aspects of discourse (sequencing and interpretation) are intrinsically
related and cannot be accounted for independently from each other. I claim
furthermore that speech act theory cannot give any insight into the sequencing
and interpretation problems, because speech act theory is neither a theory of
interpretation (it is a theory of meaning) nor a global theory of action. Finally I
show how a radical pragmatic theory (in the Gricean sense) accounts for the
sequencing and interpretation problems.
2
2. Speech act theory and conversation
There is a common sense argument shared by philosophers and linguists in
favour of the possible extension of speech act theory to discourse analysis. This
argument is the following :
Speech acts are not isolated moves in communication : they appear in more
global units of communication, defined as conversations or discourses.
Vanderveken (1994, 53) gives an explicit version of this thesis when asserting
that
speakers perform their illocutionary acts within entire conversations where
they are most often in verbal interaction with other speakers who reply to
them and perform in turn their own speech acts with the same collective
intention to pursue with success a certain type of discourse. Thus, above all,
the use of language is a social form of linguistic behavior. It consists, in
general, of ordered sequences of utterances made by several speakers who
tend by their verbal interactions to achieve common discursive goals such as
discussing a question, deciding together how to react to a certain situation,
negociating, consulting or more simply to exchange greetings and talk for its
own sake. For terminological convenience, I will call such ordered sequences
of speech acts conversations.
SPEECH ACTS AND CONVERSATION 3
The basis of this argument is that conversation is made of sequences of speech
acts. This certainly is a plausible theoretical claim
3
, but gives rise to a certain
number of objections, raised mainly by Searle (1992) in his skeptical argument.
These objections concern essentially the possible relations between questions
and answers in conversation, and can be stated as follows.
First of all, questions are defined in speech acts theory as requests for
information, and as such impose representative acts as replies. But this cannot be
correct, since a reply may have another illocutionary point (as a promise) if the
question is a request for a promise.
Secondly, certain questions require a directive as a reply, and not a
representative, when the question contains a modal auxiliary verb (cf. the
exchange : “Shall I marry Sally ?” - “Yes, do”/ “No, don’t” / “*Yes, you
shall” / “*No, you shall not”).
The third counter-example is given by indirect reponses, which do not
satisfy syntactic conditions, although the answer is pragmatically appropriate.
To these three arguments, we could add an even more embarrassing one :
answer is not a specific illocutionary force, which could be analysed by the seven
components of illocutionary force (cf. Searle & Vanderveken 1985). Answer is a
functional discursive qualification, but certainly not the semantic definition of a
speech act type.
These objections make explicit an important difference between the structure
of illocutionary acts and the structure of conversation. In speech act theory, and
more precisely in illocutionary logic, illocutionary force is decomposed into
seven components, which are all necessary conditions for the successful and non
defective accomplishment of illocutionary acts. These components (cf. Searle &
Vanderveken 1985, 12-20) are the illocutionary point, the degree of strength of
the illocutionary point, the mode of achievement of the illocutionary point, the
propositional content conditions of the illocutionary act, the preparatory
conditions of the illocutionary act, the sincerity conditions of the illocutionary act,
and finally the degree of strength of the sincerity conditions. That predictions
4 MOESCHLER
about the sequencing in conversation are difficult to come by follows from the
fact that the internal structure of illocutionary acts (and more specifically the set
of conditions for success) cannot determine the set of possible replies for any
type of illocutionary act.
By contrast, discourse analysis, while specifying sequential relations in
discourse between speech acts, does not constrain sequencing in conversation
depending on the set of possible components of illocutionary force. The
constraints are not structural, in the sense of speech act theory, they are on the
contrary functional. This means that the basic structures of conversation
(exchanges) are made of lower order conversational units (moves) which carry
functional properties. If speech act theory has been used so extensively within
this paradigm of discourse analysis
4
, it is because the functional properties
associated with speech acts as units of meaning have been exported to speech
acts as units of communication and discourse. This has several consequences for
the description of speech acts within discourse analysis.
The first consequence is that the structure of conversation is not only based
on a hierarchy of constituency, but is also functional. To take a classical
discourse model (cf. Sinclair & Coulthard 1975), discourse categories
(exchange, move, and act) are defined functionally. For instance, an act of
ELICITATION is part of a move of ELICITATION, which governs an exchange of
ELICITATION. Thus all discourse constituents receive a communicative function,
that is, an interactive meaning. But we are here far from the conventional and
semantic-meaning defining speech acts in speech act theory
5
.
As we have just noticed, discourse analysis supposes principles of
constituency which allow interpretive or functional inheritance. If we assume, as
above, that an ELICITATION is a two-place predicate relating utterance-units and
discourse-units, we must assume too that the functional properties of the smallest
discourse units (acts) are inherited by the larger constituents (moves and
exchanges). This principle is structurally identical to the projection principle in
generative grammar : a phrase is a maximal projection of a lexical head (for
SPEECH ACTS AND CONVERSATION 5
instance NP is a maximal projection of a N); in discourse, then, an exchange is
thus functionally a maximal projection of an act.
The principle of functional projection is not a necessary consequence of
discourse analysis. Another classical discourse model, the Geneva hierachical-
functional model (cf. Roulet et al. 1985, Moeschler 1985, Moeschler 1989a)
makes a different claim : functional values do not stand in a one-to-one
relationship with discourse structures. In this model, there is a basic difference
between rules of discourse formation and principles of functional interpretation.
The structural dimension is based on the following rules of formation :
R1 Units of type Exchange are made of units of type Move.
R1’ Exchanges are composed of at least two Moves.
R2 Units of type Move are made of units types Act, Move or Exchange.
R2’ Moves composed by a single Act are well-formed.
R2”Moves composed by an Act and another discourse-unit type (Move or
Exchange) are well-formed.
R2”’ Moves composed by a single Exchange are ill-formed.
Thus, the following discourse structures are well-formed :
(1) a. <E <M1 <A>, M2 <A>>>
b. <E <M1 <A>, M2 <E <M1 <A>, M2 <A>>>, M <A, M <A, A>>>>
c. <E < M1 <E <M1 <A, A>, M2 <A>, M3 <A>>, A>, M2 <A>, M3
<A>>>
where E = exchange, M = move, A = act
The structures in (1a-c) are the hierarchical representations corresponding to the
following short exchanges in (2)-(4):
(2) A Are you ready ?
B We can leave.
(3) A Are you ready ?
B Why ?
A We must leave now.
6 MOESCHLER
B Okay, but when I am in a hurry, I always forget something.
(4) A Are you ready ? Because we must leave now.
B Yes I am
A Good. Let’s go
B Let’s go
A Okay
We can represent the bracketting structures given in (1) by the following
tree-schemata :
(5)
Are you ready ?
We can leave.
Are you ready ?
Why ?
We must leave now.
Okay,
but when I am in a hurry,
I always forget something.
Are you ready ?
Because we must leave now.
Yes I am
Good
Let's go
Let’s go
Okay
AM1
M2
E
A
A
A
A
A
A
M1
M1
M2
M
E
M2
E
A
A
A
A
A
A
A
M1
M2
M3
E
M1
M2
M3
E
(a)
(b)
(c)
A
M
These structures mean that in (5a) the exchange is made of two moves both
composed of a single act, in (5b) the exchange is composed of two moves, the
second of which is made of an exchange with two moves, and a move composed
by an act and a move, and in (5c) the three-move exchange contains in the first
move an exchange made of three moves.
SPEECH ACTS AND CONVERSATION 7
What are the functional counterparts of the structural aspects of
conversational discourse ? There are two dimensions of functional properties
associated with the structural device : the first dimension is a restricted
inheritance principle, and the second, a general procedure for assigning
interpretation to discourse constituents.
The first principle is a principle of functional composition :
Principle of functional composition
(i) Constituents of exchanges bear illocutionary functions.
(ii) Constituents of moves bear interactive functions.
Definitions
(i) Illocutionary functions are of three types : initiative, reactive, and
reactive-initiative.
(ii) Interactive functions are of two types : directive, and subordinate.
The first move of an exchange (M1) is always initiative; the final move of an
exchange is always reactive. For instance M2 in the exchange <E <M1, M2>> is
the reactive move, and M1 is the initiative move. An inserted move (for example
M2 in the structure <E <M1, M2, M3>>) is a reactive-initiative move. A directive
(D) constituent is of the type move or act, and contains the act from which the
move receives its illocutionary function; a subordinate (constituent (of rank act,
move or exchange) is cancellable, and generally completes, argues for, or justifies
the main or directive constituent of the move.
We can now give the complete hierachical-functional structures given in (1)
and (5) as (6) and (6’) :
(6) a. <E <M1 <dA>, M2 <dA>>>
b. <E <M1 <dA>, M2 <sE <M1 <dA>, M2 <dA>>>, dM <As, dM <sA,
dA>>>>
c. <E <M1 <sE <M1 <dA, sA>, M2 <dA>, M3<dA>>, dA>, M2 <dA>,
M3 <dA>>>
8 MOESCHLER
where E = exchange, sE = subordinate exchange, M = move, sM =
subordinate move, dM = directive move, sA = subordinate act, dA =
directive act
(6’)
Are you ready ?
We can leave.
Are you ready ?
Why ?
We must leave now.
Okay,
but when I am in a hurry,
I always forget something.
Are you ready ?
Because we must leave now.
Yes I am
Good
Let's go
Let’s go
Okay
dAM1
M2
E
dA
dA
dA
sA
sA
dA
M1
M1
M
2
dM
sE
M2
E
dA
sA
dA
dA
dA
dA
dA
M1
M2
M3
sE
M1
M2
M3
E
(a)
(b)
(c)
dA
dM
The second functional counterpart of the stuctural device is a procedure of
interpretation assignment. It is not sufficient to have functional values assigned to
discourse constituents; required is also to have a procedure governing the
assignment of a functional interpretation to each constituent. In other words, the
types of structures given in (1), (5) or (6) are syntactic representations of
discourse; we need in addition a semantics, which can for instance assign to the
hierarchical-functional structures given in (6) the following functional
interpretations :
SPEECH ACTS AND CONVERSATION 9
(7) a. <QUESTION, ANSWER>
b. <QUESTION, <<QUESTION, ANSWER> ANSWER>
c. <<QUESTION, ANSWER, EVALUATION> PROPOSITION,
ACCEPTATION, EVALUATION>
Limited to the functions of the main moves, that is, moves which are constituents
of exchange, this very elementary assignment of functions shows that discourse
analysis needs something like a procedure for the interpretation of utterance-
units and their functional mapping onto discourse-units. In other words, we need
a theory of discourse interpretation. What is specific to discourse analysis is that
the criterion for assigning a functional value to a constituent is different from the
classical procedure used in speech act theory. Within speech act theory, the
procedure is mainly conventional: besides general cooperation principles and
background information, the procedure for the interpretation of e.g. indirect
speech acts is dependent on semantic rules (the conditions of success and
accomplishment of illocutionnary acts). By contrast, discourse analysis cannot
use a conventional procedure, because what defines a conversation is the
sequencing of acts, moves and exchanges, and also the dependance between
interpretation and sequencing. In conversation, a reactive move has two properties
: (i) its sequential dependance vis-à-vis the initiative move, and (ii) its capacity to
give retroactively an interpretation to the initiative move. The dependency relation
between sequencing and interpretation is the topic of the following section.
3. Sequencing and interpretation in conversation
Let me recall briefly what the conditions for discourse analysis are. The basic
notion of discourse analysis, as I have defined it in other occasions (cf.
Moeschler 1982, chapter 3; Moeschler 1985, chapter 3; Moeschler 1986;
Moeschler 1989b; and also Moeschler & Reboul 1994, chapter 17), is
appropriateness
6
. I have assumed that units of communication are evaluated in
10 MOESCHLER
terms of their degree of appropriateness. As units of communication are units of
discourse, two types of appropriateness can be distinguished: contextual
appropriateness and cotextual appropriateness. Let us describe these notions
and introduce the topic of this section, the sequencing and interpretation
problems.
Cotextual appropriateness depends on conditions of cotextual
appropriateness, which can be generally defined as sequencing constraints.
Conditions of cotextual appropriateness are imposed by initiative moves, and
have scope over reactive moves. These conditions of satisfaction (thematic
condition (TC), condition of propositional content (CPC), illocutionary condition
(IC) and condition of argumentative orientation (CAO)) impose on the reactive
move to share a common theme to the initiative move (TC), to be propositionally
related to the initiative move (by implication, contradiction or paraphrase) (CPC),
to bear an illocutionary force compatible with the illocutionary force of the first
move (IC), and to have a shared argumentative orientation, that is, an argumentive
co-orientation (CAO) (cf. Anscombre & Ducrot 1983). The relation between
conditions of satisfaction and cotextual appropriateness is a comparative one : the
more conditions the reactive move satisfies, the more cotextually appropriate it is.
In (9B1-B5), the degree of cotextual appropriateness increases, together with the
degree of satisfaction of the conditions of cotextual appropriateness:
(9) A Can you give me the time ?
B1 I have a serious headache. -TC
B2 The postman has just passed. +TC, -CPC
B3 Is it not already ten o’clock ? +TC, +CPC, -IC
B4 It is not yet ten o’clock. +TC, +CPC, +IC, -CAO
B5 It is ten o’clock. +TC, +CPC, +IC, +CAO
When the thematic condition, the condition of propositional content and the
illocutionary condition are satisfied, discourse is said to be coherent. If only the
thematic condition and the condition of propositional content are satisfied,
SPEECH ACTS AND CONVERSATION 11
discourse is said to be cohesive. So a coherent discourse is always cohesive,
whereas the reverse is false.
The converse notion is the notion of conditions of contextual
appropriateness. These conditions do not hold of the reactive, but of the initiative
move. For any initiative move, the degree of contextual appropriateness is
determined by the reactive move, and more precisely, by the degree of cotextual
appropriateness of the reactive move. We can formulate this dependency relation
between cotextual appropriateness and contextual appropriateness as stated by
the following principle (cf. Moeschler 1982 and 1989b) :
Principle of dependency
The more sequencing constraints the reactive move satisfies, the more the
initiative move is contextually appropriate; the less sequencing constraints
the reactive move satisfies, the more the initiative move is contextually
inappropriate.
The consequence of the principle of dependency is the following : sequencing
and interpretation in conversational sequences are closely related. This is so
because a dialogical sequencing, whether appropropriate or not, always gives an
image of the interpretation of the initiative move, and retroactively defines its
degree of contextual appropriateness. The following principle gives a more
precise definition of this relation:
Principle of dialogical interpretation
The interpretation of a move is dialogical, and results from the dialogical
sequencing to which it gives rise.
In (10), the degree of contextual appropriateness of the initiative move is a
function of the degree of cotextual appropriateness of the reactive move B1-B5:
(10) A Peter is a friend whom one can count on.
B1 By the way, what are you doing tonight ? -TC
B2 Would you call that a friend ? +TC, -CPC
B3 Do you forget he voted against your project ? +TC, +CPC, -IC
12 MOESCHLER
B4 He never inspired me confidence. +TC, +CPC, +IC, -CAO
B5 I think so too. +TC, +CPC, +IC, +CAO
This set of principles yields a presumably robust approach of discourse
sequencing. If we map the set of structural constraints defined in section 2 with
the set of sequencing constraints defined here, we should have a powerful theory
of discourse representation. An explicit version of this theory, containing a
syntax, a semantics, and a procedure of analysis, has been proposed in
Moeschler (1989a) and applied in computational linguistics for modelling
person-machine dialogue in Bilange (1992) and Pernel (1994).
So far so good. But we are here far from the projection philosophers of
language have made on conversation. We are also far from the basic principles of
speech act theory, in which speech acts are conventional units of meaning. What
is specific to the discourse theory presented so far is that illocutionary force is no
longer a complex unit of meaning made of seven components (what we generally
call a speech act), but is reduced to the functional and sequential properties of
moves. Beside the fact that in speech act theory, there is no reason to take
ANSWER as specific relational illocutionary force (the symetrical counterpart of a
question), while such reactive illocutionary function is required by the structural-
functional device, there is a major difference between initiative and reactive moves
that speech act theory cannot account for. This difference can be formulated as
stated in the following asymmetry postulate :
The asymmetry postulate of illocutionary functions
Whereas a reactive move is a function with two moves as arguments, an
initiative move is a function with a move as first argument, and a function as
second argument.
According to this postulate, the formal difference between a question (initiative
move A) and an answer (reactive move B) can be stated by means of the
following notation :
(11) ANSWER (B, A)
SPEECH ACTS AND CONVERSATION 13
(12) QUESTION (A, ANSWER (B, A))
This notation is consistent with the principle of dialogical interpretation : a move
is a question because it has been interpreted as a question by its answer, which
expresses the second argument of the question, that is, the function ANSWER.
The very optimistic proposal of this section thus contrasts with Searle’s
pessimistic view. But, as already mentioned, we are here far from an extension of
speech act analysis to conversation : what we have is a structural and functional
model for the analysis of conversation.
So far, my optimism was actually purely strategic, in order to discuss one of
the possible extension of speech act theory for the analysis of conversation. The
following section is much more pessimistic: I show the descriptive inadequacy of
the so-called model of conversation, and in a more radical argumentation that
both the sequencing and the interpretation problems can be solved in a non
discourse-based approach to meaning, which belongs to radical pragmatics.
4. The interpretation and sequencing problems revisited
Let us first reformulate the interpretation and sequencing problems in a more
general way.
The interpretation problem can be reduced to the following question: What
should hearers do in order to understand what speakers intend to
communicate? Strictly speaking, this problem is not different from the one
pragmatic theory tries to answer. The main task of pragmatics is to explain, via
non linguistic principles, procedures by which hearers can understand what
speakers want to communicate. If discourse analysis were limited to the
interpretation problem, one could conclude that there is no need for a specific
domain of investigation like discourse analysis, since pragmatics and discourse
analysis would be about the same set of facts, that is, the interpretation problem.
Nevertheless, the question which remains open is whether interpretative facts are
14 MOESCHLER
better understood within pragmatics or within a conversational theory. As we
have seen in the preceeding section, discourse analysis covers another domain,
the sequencing problem, which implies that the interpretation problem cannot be
solved independently. The conclusion is that if we can show that the sequencing
problem can be solved outside discourse analysis, the interpretation problem
would no longer be a conversational problem and could be treated within
pragmatic theory. Before pursuing the argument, let us reformulate properly the
sequencing problem.
The sequencing problem follows from the sequential and dialogical nature
of discourse. It can be summarized as in the following question : Are there rules
or principles which guarantee the well-formedness of discursive and
conversational sequences ? This formulation presupposes first that some
discourse productions can be evaluated as well-formed, and others as ill-formed,
and second that defective and non defective aspects of discourse sequences result
from the satisfaction and non satisfaction of sequencing rules or principles. This
formulation implies that the sequencing problem is identical with the problem of
discourse coherence. But it has been argued (cf. Charolles 1988 for a synthetic
review of research on text grammar and coherence) that discourse coherence is
basically an interpretation problem, and we can conclude that the sequencing
problem can be limited to the interpretation problem. If this assumption is
correct, then there is no need for a conversational theory, since the sequencing
problem is part of the interpretation problem, i.e. the object of a pragmatic theory.
So far, we have discussed two pairs of antinomic theses, belonging
respectively to conversational pragmatics (what I called discourse analysis) and
to pragmatic theory :
The theses of conversational pragmatics
T1 The interpretation problem is part of the sequencing problem.
T2 Conversational pragmatics is independant of pragmatic theory and
makes different predictions.
The theses of pragmatic theory
SPEECH ACTS AND CONVERSATION 15
T’1 The sequencing problem is part of the interpretation problem.
T’2 Conversational pragmatics is dependent of pragmatic theory and does
not make different predictions.
Now, I must demonstrate that principles of conversational pragmatics are
falsifiable, and that pragmatic aspects of discourse must be accounted for by
general pragmatic principles. The principle of dialogical interpretation constitutes
the Achilles’ heel of conversational pragmatics, as I have defined it until now.
We can raise three objections against this principle.
First objection. If the principle of dialogical interpretation is valid, it is
impossible to assign an interpretation to the last move of an exchange, since there
is no reactive move to base this interpretation on. From a technical point of view,
the heuristics given by the principle of dialogical interpretation is only partial, and
we must admit that other principles than sequential principles affect the
interpretation of moves in conversation.
Second objection. As Trognon & Brassac (1982, 85) have shown, a reactive
move is not an interpretation of the initiative move, but an interpretation in act. So,
principles which make the accomplishment of an act come true do not rely on a
theory of sequencing, but belong to illocutionary logic. For instance, the axiom
of illocutionary logic stating that accomplishment implies success is just what we
need to explain sequential relations between moves. In other words, sequencing
in conversation is a trace of the accomplishment of illocutionary acts in
conversation.
Third objection. The third objection is much more radical. It states that the
principle of dialogical interpretation is either non informative or circular.
According to this principle, it is possible to assign an illocutionary function to the
initiative move depending on the reactive move. But we face here a dramatic
choice: either the reactive move is predictable via sequencing rules, and then it
says nothing about the interpretation of the iniative move, or the reactive move is
not predictable. In the latter case, it becomes impossible to say anything about the
relation between moves via sequencing rules, since they have no predictive
16 MOESCHLER
power. The situation is even worse, though: if the interpretation is given by the
reactive move, and reactive moves are directed by sequencing constraints, how
can these sequencing constraints be accounted for when there is no possibility of
determining a priori the function of the initiative move ? In any case, the
principle of dialogical interpretation leads to undesirable conclusions : it is either
uninformative or circular.
We conclude that the theses of conversational pragmatics are falsified, and
that we must abandon them. It also seems that we have not made much progress :
we have spent a long time arguing in favour of a structural and sequential model
of discourse within conversational pragmatics, and we now arrive at the
conclusion that the interpretation problem is not part of the sequencing problem.
So, what possibilities remain for the analysis of conversations ? It seems that
what we need is a pragmatic theory which is a true theory of interpretation. I will
now discuss two approches to meaning and communication which could be good
candidates, namely illocutionary logic and Relevance theory, and I will defend an
approach of conversation within the latter type of pragmatic theory.
5. Illocutionary logic and conversation
Recent work by Alain Trognon and Christian Brassac
7
offers a good illustration
of how the sequencing problem can be treated within speech act theory, and more
specifically illocutionary logic. Trognon & Brassac (1992), for instance, propose
a general procedure of interpretation and sequencing for indirect speech acts and
conversational implicatures. If we take as a prototypical example the indirect
request Can you pass the salt?, their analysis proceeds as follows :
By passing the salt, the interlocutor satisfies the request, which entails its
success, which entails the truth of the proposition that the hearer can pass the
salt (preparatory condition of requests), which entails the satisfaction of the
SPEECH ACTS AND CONVERSATION 17
question, which entails its success. (Trognon & Brassac 1992, 89; the
translation is mine)
To make a long story short, we have the following chain of entailments :
(13) SATISFACTION(REQUEST) → SUCCESS(REQUEST) → SATISFACTION
(QUESTION) → SUCCESS(QUESTION)
The element which determines the satisfaction of the primary illocutionary act
(the request) is the passing of the salt, that is the action that should be obeyed
under the illocutionary point of the directive act. I do not intend to discuss here
the principle of illocutionary logic under which satisfaction implies success, but
this analysis calls for the following remarks.
First, the retroactive procedure is close to the principle of dialogical
interpretation. The analysis goes backwards, that is, moves from the satisfaction
of the primary illocutionary act to the success of the secondary illocutionary act.
As a natural procedure of interpretation, it seems very strange, for at least two
reasons: it is contradictory to the basic principles of speech act theory, which
proceedes from the derived illocutionary act to the literal illocutionary act; once
the illocutionary point is obtained, it seems odd to go on processing until the
source of the derived illocutionary point is found. This leads naturally to the
second objection.
Second, the analysis is counter-intuitive, and does not constitute an
interpretation procedure. It implies that in speech act theory, as well in
illocutionary logic, the literal meaning of an utterance like Can you pass the salt?
has the illocutionary force of a question. But the illocutionary point of this
utterance is not that of a request for information; thus, the question is a
secondary act, and it conveys a primary illocutionary act. Speech act theory
predicts that the illocutionary point is a directive, because the utterance questions
a preliminary condition of directives; so, the literal meaning is a question, and the
derived meaning (which corresponds to speaker’s meaning) is a request. The
following interpretation procedure can be stated in contrast to (13) :
18 MOESCHLER
(14) a. literal meaning : request for information
b. propositional content of the request for information : ability of the
hearer to pass the salt
c. condition of success of directives : as a preparatory condition, the
hearer is able to accomplish the requested action
d. generalisation on indirect directives : to ask for a preparatory
condition of a directive is a way to realise an indirect directive (as a
primary illocutionary act)
e. inference : the speaker accomplishes as primary illocutionary act a
request via a secondary illocutionary act of request for information.
The conclusion is very simple : Trognon & Brassac’s proposal within
illocutionary logic is another version of the principle of dialogical interpretation,
for which the interpretation problem is part of the sequencing problem. The
revised version of their analysis proposed in (14) under the most classical
version of speech act theory has the advantage of giving a coherent procedure for
utterance interpretation. Admittedly, it has the disavantage of saying nothing
about the sequencing problem: I claim that a radical pragmatic theory of utterance
interpretation makes correct predictions about sequencing in conversation
without having to formulate any sequential constraints on interpretation.
6. Relevance theory and sequencing in conversation
In previous work (cf. Moeschler 1989c, 1993, 1994, to appear, chapter 11 and
12), I have argued for a radical pragmatic treatment of the sequencing problem
within Relevance theory. The argument developed in these papers is that
sequencing explicited via discourse connectives
8
cannot be explained by
principles of discourse structure or discourse sequencing, because discourse
connectives often contradict either their conventional meaning or the predictable
discourse structure they should make explicit. The meaning of discourse
SPEECH ACTS AND CONVERSATION 19
connectives is what Wilson & Sperber (1993) call procedural encoding, and
refers to the nature of context (as a cognitive construct), and to the possible
contextual inferences.
9
I will give a brief overview of this type of analysis with an
example of conversational use of the French connective parce que (‘because’).
Let us take the following example, drawn from a phone call (cf. Schmale-
Buton & Schmale 1984, 190-191), translated here from French (where S = the
secretary of the practice and P = the patient) :
(15) S1 you should come at the very beginning of the afternoon
P1 at what time
S2 well at two o’clock but not later because just after I don’t know if
he visits as he has no appointment
P2 yes
P2’ he will be there just before two o’clock or
S3 yes
P3 (parce que) because if I come a little before two o’clock he will be
there
S4 he will be there yes yes yes
This example raises two types of problems for conversational pragmatics :
the first problem is structural and sequential, and bears on the discourse function
of connectives like parce que; the second problem is interpretive, and bears on
the possibility of the standard causal reading of parce que. I will discuss briefly
these two aspects of (15), which will demonstrate that a discourse oriented
analysis is hopeless. I will then turn to an alternative interpretation, within a few
postulates of Relevance theory.
The first problem is structural. Very informally, the sequences P1-S4 is
composed by three exchanges, as described in (16) :
20 MOESCHLER
(16) a. <E1 <QUESTION (P1), ANSWER (S2), EVALUATION (P2)>>
b. <E2 <QUESTION (P2’), ANSWER (S3)>>
c. <E3 <QUESTION (P3), ANSWER (S4)>>
These exchanges are related, and one of the interpretation is to represent the
integration of forward-oriented exchanges : there is a consecutive relation
between E1 and E2, and E2 and E3. It is because the answer in E1 that P
questions S in E2, and so on. So the hierarchical-functional representation
should be something like (17) :
(17)
P1
S2
P2
P2'
S3
P3
S4
M1
M2
M3
dA
M2
dM
M2
sE
M1
sE
M1
E
Unfortunately, it is no longer possible to describe discourse connectives as
markers of interactive functions (as Roulet et al. 1985, chapter 2).
Conventionally, parce que should introduce a subordinate move (introducing a
cause or an explanation), and thus, if the move is initiative, a subordinate
exchange. This solution is structurally possible, as shown in (18), but raises a
new problem, which is interpretive.
SPEECH ACTS AND CONVERSATION 21
(18)
P1
S2
P2
P2'
S3
P3
S4
M1
M2
dA
M1
dA
M1
M2
sE
M2
sE
M3
E
Let us turn to the interpretation problem and suppose that parce que
(because) triggers the following logical elimination rule (cf. Blakemore 1987, 43)
:
(19) Input P parce que Q
Output (a) P
(b) Q
(c) Q is the cause of P
Let us reformulate the following variables in this use of parce que :
(20) P = QUESTION (the doctor will be there just before two o’clock)
Q = if the speaker comes before two o’clock, the doctor will be there
The rule (19c) cannot apply here, because it would yield a very counter-intuitive
interpretation, as (21) shows:
(21) (if the speaker comes before two o’clock, the doctor will be there)
CAUSE (QUESTION (the doctor will be there just before two o’clock))
In others words, it does not make any sense to say that the conditional assertion
of a fact (the doctor will be there) causes the request for information (will the
doctor be there ?). The coherent relation suggested by the connective parce que
is in fact the opposite, as shown in (22) :
22 MOESCHLER
(22) (QUESTION (the doctor will be there just before two o’clock)) CAUSE
(if the speaker comes before two o’clock, the doctor will be there)
To be more explicit, what happens is that the reverse causal relation is no longer a
relation between propositional contents, but a relation between illocutionary acts,
as (23) indicates:
(23) (QUESTION (the doctor will be there just before two o’clock)) CAUSE
TO SAY (if the speaker comes before two o’clock, the doctor will be
there)
The interpretation of parce que which imposes here a reverse causal (so to
speak) relation implies that the use of parce que is here identical with the
semantic meaning of donc (‘thus’). If this interpretation is correct, the
problematic structural interpretation given in (17) is accounted for: donc
introduces a directive act (or move). So we face here a very strange case for
discourse analysis : the interpretive procedure imposing a possible structural
interpretation of the discourse, rather than the discourse structure imposing an
interpretation. Sequential relations in discourse are thus determined by pragmatic
constraints on interpretation. If we adopt Blakemore’s terminology, we can say
that we face here a very typical semantic constraint on relevance. The
consecutive meaning of parce que (or, which is the same, its inverse causal use)
is the only possible pragmatic interpretation because it is the only interpretation
consistent with the principle of relevance. What this means is that the consecutive
interpretation is the first interpretation which comes to mind and which creates a
sufficient effect capable to balance the cognitive effort necessary for the treatment
of the utterance. What is this effect ? The effect is minimal, but consistent with
the principle of relevance, which states that the utterance is the most relevant
possible one in the circumstances : the effect concerns the plausibility of whether
a fact true of X can be true of Y. So, the demand of confirmation uttered in P3
and introduced by parce que is a relevant utterance as far as it questions the non
trivial fact that if it is true that the doctor is at a certain place at a certain time for
X, this fact is also true for speaker P.
SPEECH ACTS AND CONVERSATION 23
The consecutive meaning of parce que is not part of the semantics of the
connective, but part of its procedural encoding. This “change of logical
meaning” is not an exception, and constitutes a typical way of reasoning in
Gricean pragmatics. As examples, one can mention the exclusive reading of or
via a scalar implicature and its inclusive meaning (cf. Gazdar 1979), the so-called
bi-conditional use of if P, Q, which by invited inference (cf. Geiss & Zwicky
1974), conveys if non-P, non-Q or also the interpretation of negative utterances
(cf. Moeschler 1991). For instance, (24) does not communicate (25a), but (25b) :
(24) If you are quiet, we will go to the movie.
(25) a. If we will not go the movie, then you are not quiet.
b. If you are not quiet, we will do not go to the movie.
In this section, I have tried to give a solution to the sequencing problem of
parce que within an interpretive framework. I pointed to some consequences of
the analysis for the function of connectives in discourse, and as a result, it
appears that this description does not require a theory of discourse. But I still
have to say how sequencing with indirect speech acts or conversational
implicatures could be described and explained within Relevance theory.
7. Sequencing and implicatures in Relevance theory
I have argued in section 5 that a possible application of illocutionary logic to the
analysis of conversation is identical with a model of discourse based on the
principle of dialogical interpretation. I have also assumed that an interpretive
approach within speech act theory will necessarily explain inferential and
inductive reasoning from the secondary to the primary illocutionary act. I have
claimed further that speech act theory says nothing about the sequencing
problem, as I have tried to show in section 2. But I have suggested too that
Relevance theory, which is a pragmatic theory of utterance interpretation, has
something to say about the sequencing problem.
24 MOESCHLER
What is paradoxical is the assertion that a theory of interpretation is capable
of predictions about discourse. The main reason why this is possible is that
discourse coherence is interpreted in Relevance theory as an effect of relevance,
and not as something specific to discourse. For a theory of interpretation, the
sequencing problem is equivalent to the interpretation problem. The issue we
should adress is whether conversation, as a type of communication, exhibits
particular levels of organisation which would play a role similar to notions like
frame (Fillmore), script (Shank) or background (Searle). I have arged in
Moeschler (1993) that in certain types of conversation, such as those examined
in this paper, i.e. phone calls to a medical surgery, it is necessary to refer to
general scripts, which relate to the sequential organisation of conversation. For
instance, move (26) is typical of a script associated with a phone call to a surgery
(27), which constitutes a chunk of information easily accessible in such
circumstances :
(26) I am calling because my daughter is a patient of Dr. R*
and she’s had a high temperature since yesterday
now her temperature is over 40 degrees Celsius
yesterday night it was about 39
10
(27) Script associated with a phone call to a surgery
1. When a patient calls a surgery she generally gives the reason for her
call. The principal reasons in this context are :
(a) beeing a client of the doctor (generally a necessary, but not a
sufficient condition for the call);
(b) indication of the state of the patient (high temperature, pain, etc.).
2. When the principal reason (b) has been given (for instance high
temperature), further details can be given depending on the informative
intention of the speaker (for instance, if the previous information was
vague, it could be further specified by giving quantitative, qualitative,
temporal, spatial references).
SPEECH ACTS AND CONVERSATION 25
We see that in (26) each act refers to one proposition of the script, and that the
sequential organisation of the discourse perfectly reflects the hierarchy of
information presented in the logical structure (28) :
(28)
call,
∧
state, v
temperature, ∧
yesterday today
… …
patient
We face here the problem raised by Searle (1992) in the third part of his
paper on conversation, the default of common intentionality :
The reason that conversations do not have an inner structure in the sense that
speech acts do is not (as sometimes claimed) because conversations involve
two or more people, but because conversations as such lack a particular
purpose or point. (Searle 1992, 20)
This means two things: first, that conversation has no structure in the sense that
illocutionary acts have
11
; and second, that the cause is that there is no shared
intentionality. At this point, we have the choice between two positions: the first
position would require a particular type of discourse to be defined in which these
conditions whould be satisfied; the second position would just ask why it should
be so, that is, why speakers should share intentionality in conversation ?
The argument underlying the second position, within Relevance theory, is
that relevance is defined relatively to an individual. More precisely, “An
assumption is relevant to an individual at a given time if and only if it is relevant
in one or more of the contexts accessible to that individual at that time” (Sperber
26 MOESCHLER
& Wilson 1986, 144). This means that relevance is a question of accessibility of
context, and not of shared knowledge.
12
We face here one of the most important
ideas of Relevance theory : communication is a highly risky task, and a not
always successful one. If we admit that the use of language in verbal
communication is never accompagnied by a warranty of successful
communication (whereas verbal communication implies a warranty of relevance),
the second position becomes acceptable. There is no evidence or necessity for
shared intentionality because communication does not imply mutual knowledge.
If these proposals are correct, one important consequence for the analysis of
conversations is that it is not possible (except for very standard and canonical
processes likes those illustrated in (26)) to define conversation in terms of a
theory of complex actions. If this is correct, we are again in a position to argue
for a non discourse-based solution to the sequencing problem. And if there is no
possibility for a discourse theory of sequencing, we are back to the interpretation
problem. So, why should we not accept the classical analysis for indirect speech
acts and implicatures, implied by speech act theory, as presented in section 5 ?
The main criticism I would address to a theory of meaning like speech act theory
is that it is basically a theory of literal meaning. Indirect speech acts, metaphors,
irony, are analysed within speech act theory (cf. Searle 1979) as derived speech
acts. Within Relevance theory, we can make a different claim. It is not assumed
that any non literal speech has to be understood against its literal meaning,
because linguistic meaning is seldom sufficient for the hearer to understand the
utterance. Non linguistic knowledge is nearly always necessary, which implies
that the interpretation of an utterance triggers two types of processes: the
development of the logical form of the utterance (that is, its explicature) and the
detection of implicatures (either the implicated premisses necessary for
inferences to take place, or the implicated conclusions).
Implicatures do no longer follow from the use or the exploitation of maxims
of conversation, contra Grice (1975) : they are just contextual implications
resulting from an inferential process based on contextual premisses, and on the
SPEECH ACTS AND CONVERSATION 27
logical form of the utterance. As the context is constructed, and not given,
communication succeeds when the intersection of the set of the contextual
implications drawn by the hearer and the set of the conclusions entailed by the
propositional form representing the speaker’s thought is not empty.
Communication fails when the intersection of these two sets yields the empty set.
But the situation where the two sets are the same, i.e. in literal utterances, obtains
seldom.
What about indirect speech acts ? Even if Sperber & Wilson do not propose
any explicit treatment of indirect speech acts (mainly because the classical
analysis is based on a taxonomy of speech acts rejected within Relevance
theory), we can suppose that indirect speech acts, when they have the properties
of generalised conversational implicatures, are no longer indirect: they are direct
speech acts. But what they communicate (for instance for the classical Can you
pass the salt ?
13
) is that some fact is sufficiently manifest for the question to be
relevant in that context. It means that the explicature of the utterance would be
that of a directive : “the speaker wants the salt”.
8. Conclusion
In this paper, I have tried to show how speech act theory could be extended to the
analysis of conversations. I have argued that one of the possible extensions,
which belongs to discourse analysis, makes different predictions on conversation
than speech act theory, and that the meaning of speech act changes: starting as a
unit of communication, it becomes a unit of discourse. Discourse analysis leads
to specific problems, that is, the interpretation and the sequencing problems. I
have discussed the classical solutions within discourse analysis and illocutionary
logic, arguing that both approaches meet the same type of objections. Finally I
have defended a non discourse-based solution for the sequencing problem within
Relevance theory and proposed a solution within the same framework for the
28 MOESCHLER
interpretation problem, and more specifically for indirect speech acts and
conversational implicatures.
NOTES
1. I take discourse to refer to any type of speechs consisting of sequences of exchanges,
moves or speech acts, either monological or dialogical, following here the Geneva tradition of
discourse analysis developed in Roulet et al. (1985). Cf. also Moeschler (1985), Moeschler
(1989a) and Moeschler (forthcoming) for further developments.
2. See Moeschler (1993) for a formulation of the sequencing and interpretation problems
within Sperber & Wilson’s Relevance theory (1986), and Moeschler (1992a) for a more general
discussion on the domain of discourse analysis.
3. From the following extract of Searle & Vanderveken (1985, 11), we are authorised to
look for sequencing principles defining the sequences of speech acts in conversation : “The key
to understanding the structure of conversations is to see that each illocutionary act creates the
possibility of a finite and usually quite limited set of appropriate illocutionary acts as replies”.
I shall present below a coherent and complete model of sequencing rules for conversation
analysis, which is explanatorily adequate, but unfortunately descriptively inadequate.
4. One should remember here the classical opposition formulated by Levinson (1983)
between discourse analysis (DA), which is based on speech act theory, and conversation
analysis (CA), which comes from the ethnomethodological paradigm.
5. Cf. Stenström (1984) for an exhaustive analysis of questions and answers within a
functional model, and Moeschler (1986) for a critical review.
SPEECH ACTS AND CONVERSATION 29
6. Cf. van Dijk (1977) for a general definition of pragmatics and discourse via the criterion
of appropriateness.
7. See especially Brassac (1992), Brassac (1994), Trognon & Brassac (1992), Trognon &
Larrue (1994), Trogon & Ghiglione (1993), Trognon (1994),; see also Moeschler (1992b) for a
general discussion of Trognon & Brassac (1992).
8. Following Blakemore’s (1987) terminology.
9. I refer here to Luscher (1994) for a very detailed analysis of pragmatic connectives within
Relevance theory.
10. For simplicity, I present every utterance-act as a separate line.
11. See Vanderveken (1992) and (1994) for a different analysis, and a very global proposition
of discourse taxonomy in terms of direction of fit.
12. See Sperber & Wilson (1982) for a detailed criticism of the notion of mutual knowledge,
and Moeschler (1994) for a general discussion of warranty of uptake in conversation.
13. See Groefsema (1993) for an anlysis of can and more specifically of Can you pass the
salt? within Relevance theory.
References
Anscombre, Jean-Claude and Oswald Ducrot
1983 L’argumentation dans la langue. Bruxelles: Mardaga.
Bilange, Eric
1992 Modélisation du dialogue oral finalisé personne-machine par une approche
structurelle. Théorie et réalisation. Paris: Hermès.
Blakemore, Diane
1987 Semantic Constraints on Relevance. Oxford: Blackwell.
30 MOESCHLER
Brassac, Christian
1992 S’engager en conversation. Thèse de doctorat. Nancy: Université de Nancy 2.
1994 “Actes de langage et enchaînement conversationnel”. Cahiers d’épistémologie
9401. Montréal: UQAM.
Charolles, Michel
1988 “Les études sur la cohérence, la cohésion et la connexité textuelle depuis la fin
des années 1960”. Modèles Linguistiques 10/2. 45-66.
Dascal, Marcelo
1992 “On the pragmatic structure of conversation. In John R. Searle et al., (On)
Searle on Conversation. Amsterdam: Benjamins.35-56.
Dijk, Teun A. van
1977 Text and Context. Explorations in the Semantics and Pragmatics of
Discourse. London: Longman.
Gazdar, Gerald
1979 Pragmatics. Implicature, Presupposition and Logical Form. New York:
Academic Press.
Geis, Michael L. and Arnold Zwicky
1971 “On invited inferences”. Linguistic Inquiry 2. 561-6.
Grice, H.Paul
1975 “Logic and conversation”. In Peter Cole and Jerry L. Morgan (eds.), Syntax
and Semantics 3: Speech Acts. New York: Academic Press. 41-58.
Groefsema, Marjolein
1993 “‘Can you pass the salt ?’: A short-circuited implicature?”, Lingua 87 (1/2).
137-167.
Levinson, Stephen C.
1983 Pragmatics. Cambridge: Cambridge University Press.
Luscher, Jena-Marc
1994 “Les marques de connexion: des guides pour l’interprétation”. In Jacques
Moeschler, Anne Reboul, Jean-Marc Luscher and Jacques Jayez, Langage et
pertinence. Référence temporelle, anaphore, connecteurs et métaphore.
Nancy: Presses Uiversitaires de Nancy. 175-227.
Moeschler, Jacques
1982 Dire et contredire. Pragmatique de la négation et acte de réfutation dans la
conversation. Berne: Peter Lang.
1985 Argumentation et conversation. Eléments pour une analyse pragmatique du
discours. Paris: Hatier.
1986 “Answers to questions about questions and answers”. Journal of Pragmatics
10. 227-253.
1989a Modélisation du dialogue. Représentation de l’inférence argumentative.
Paris: Hermès.
SPEECH ACTS AND CONVERSATION 31
1989b “La problématique des règles d’enchaînement et d’interprétation revisitée”. In
Christian Rubattel (ed.), Modèles du discours. Recherches actuelles en
Suisse Romande. Berne: Lang. 61-84.
1989c “Pragmatic connectives, argumentative coherence, and relevance”.
Argumentation 3. 321-339.
1991 “The pragmatic aspects of linguistic negation : Speech act, argumentation and
pragmatic inference”. Argumentation 6. 51-76.
1992a “Foundations of dialogue analysis”. In Sorin Stati and Ewa Weigand (eds.),
Methodologie der Dialoganalyse. Tübingen: Niemeyer. 66-74.
1992b “Théorie pragmatique, acte de langage et conversation”. Cahiers de
Lingusitique Française 13. 108-124.
1993 “Relevance and conversation”. Lingua 90. 149-171.
1994 “How do we know that what we mean is understood? Hypothesis and warranty
of uptake in conversation”. In Hermann Parret (ed.), Pretending to
communicate. Berlin: de Gruyter. 33-47.
(forthcoming) Théorie pragmatique et pragmatique conversationnelle. Paris, Armand
Colin.
Moeschler, Jacques and Anne Reboul
1994 Dictionnaire encyclopédique de pragmatique. Paris: Seuil.
Pernel, Didier
1994 Gestion des buts mutiples de l’utilisateur dans un dialogue homme-machine
de recherche d’informations. Thèse de doctorat, Orsay: LIMSI.
Roulet, Eddy et al.
1985 L’articulation du discours en français contemporain. Berne: Peter Lang.
Schmale-Buton, Elisabeth and Günther Schmale
1984 Conversations téléphoniques. Bielefeld: Université de Bielefeld.
Searle John R.
1969 Speech acts. Cambridge: Cambridge University Press.
1979 Expression and meaning. Cambridge: Cambridge University Press.
1992 “Conversation”. In John R. Searle et al., (On) Searle on conversation.
Amsterdam: Benjamins.7-29.
Searle, John R. and Daniel Vanderveken
1985 Foundations of Illocutionary Logic. Cambridge: Cambridge University Press.
Sinclair, John McH. and R.M. Coulthard
1975 Towards an Analysis of Discourse. The English used by teachers and pupils.
Oxford: Oxford University Press.
Sperber, Dan and Deirdre Wilson
1982 “Mutual Knowledge and relevance in Theories of Comprehension”. In Neil V.
Smith (ed.), Mutual Knowledge. New York: Academic Press. 61-85.
32 MOESCHLER
1986 Relevance. Communication and Cognition. Oxford: Blackwell.
Stenström, A.-B.
1984 Questions and responses in English conversation. Lund, CWK Gleerup.
Trognon, Alain
1994 “Sur la construction interactive du quotidien”. In Alain Trognon et al., La
construction interactive du quotiden. Nancy: Presses Universitaires de
Nancy. 7-52.
Trognon, Alain & Christian Brassac
1992 “L’enchaînement conversationnel”. Cahiers de Linguistique Française 13.
76-107.
Trognon, Alain & Jeanine Larrue
1994) Pragmatique du discours politique. Paris: Colin.
Trognon, Alain & Ghiglione, Rodolphe
1993 Où va la pragmatique ? De la pragmatique à la psychologie sociale.
Grenoble: Presses Universitaires de Grenoble.
Vanderveken, Daniel
1992 “La théorie des actes de discours et l’analyse de la conversation”. Cahiers de
Linguistique Française 13. 9-61.
1994 “Principles of speech act theory”. Cahiers d’épistémologie 9202, Montréal:
UQAM.
Wilson, Deirdre and Dan Sperber
1993 “Linguistic form and relevance”. Lingua 90. 1-25.
