«Indicators and procedures : nevertheless and but1 DIANE BLAKEMORE University of Salford (Received August ; revised January ...»
J. Linguistics (), –. Printed in the United Kingdom
# Cambridge University Press
Indicators and procedures : nevertheless and but1
University of Salford
(Received August ; revised January )
The paper aims to clarify the Relevance theoretic notion of procedural meaning (cf.
Blakemore , Wilson & Sperber ) through the analysis of but and nevertheless.
I show, ﬁrst, that a procedural analysis is able to account for diﬀerences between these expressions that cannot be explained in terms of the speech-act theoretic notion of non-truth conditional indicators, and, second, that these diﬀerences show that the conception of procedural meaning as a constraint on contextual eﬀects (cf. Blakemore ) is too narrow and must be extended to include all information about the inferential processes involved in utterance interpretation, including context selection.
. I The starting point for this paper is Grice’s () analysis of the non-truth conditional suggestions carried by expressions like but and nevertheless as
they are used in examples like the following :
() I have received the e-mail, but it’s in Dutch.
() I am sure he is normally very conscientious. Nevertheless the papers are missing.
Grice’s idea that non-truth conditional meaning should be analysed in terms of conventional implicature follows in the speech act theoretic tradition in which linguistic meaning is analysed either in terms of its contribution to the descriptive content of utterances or in terms of its role in indicating how utterances are to be interpreted. Rieber’s () re-analysis of conventional implicatures as tacit performatives follows the same speech act theoretic tradition. In this paper I argue that we should abandon a speech act theoretic approach to the distinction between describing and indicating in favour of an  This paper is a considerably revised (and expanded) version of papers I have delivered at the International Christian University Tokyo, Osaka University, the Relevance Theory Workshop at the University of Luton, and the University of Sheﬃeld. I am very grateful to members of those audiences for many useful comments and criticisms and to Bob Borsley, Richard Breheney, Robyn Carston and Deirdre Wilson for helping me realize that it was all a lot more diﬃcult than I thought it was. I would also like to thank the JL referees, whose comments and suggestions I found extremely useful in preparing my ﬁnal revisions.
For convenience, I have referred to the speaker as ‘ she ’ and the hearer as ‘ he ’.
approach concerned with the cognitive processes involved in utterance interpretation. The distinction that emerges from this change of perspective is a distinction between two ways in which linguistic meaning contributes to inferential pragmatic processes : on the one hand, it may encode constituents of or propositional representations which undergo these processes, while on the other, it may encode information about these processes.
In earlier work (e.g. Blakemore ) I suggested that this distinction is coextensive with the distinction between truth conditional meaning and nontruth conditional meaning. However, more recent research has shown that this is not the case. For example, Wilson & Sperber (), Ifantidou-Trouki () and Blakemore () have shown that while sentence adverbials, parentheticals and certain so-called apposition markers are non-truth conditional, they clearly encode concepts.
This raises the question of whether we should assume that the distinction between truth conditional and non-truth conditional meaning is indeed the fundamental distinction in a cognitively grounded theory of linguistic meaning. Within the Relevance theoretic approach to communication developed by Sperber & Wilson () it has been argued that there is a gap between linguistic meaning and the truth conditional content of the assumption explicitly communicated (see in particular, Carston ), and hence that the linguistic meaning of an utterance does not deliver a proposition with truth conditions. This suggests that linguistic semantics is concerned not with the relation between linguistic form and the external world, but with the relation between ‘ bits of linguistic form and the cognitive information they encode ’ (Carston ). In this picture, truth conditionality is not central to a theory of linguistic semantics : the question that matters is not whether a linguistic expression contributes to something with truth conditions, but rather what kind of cognitive information an expression encodes.
If this is right, then it is important to be able to give a precise account of what it means for a linguistic expression or structure to encode cognitive information, and in particular, a precise account of what it means for a linguistic expression or structure to encode either a concept or a pragmatic procedure. This paper aims to clarify the notion of procedural meaning through the analysis of but and its less well-known relative nevertheless. In particular, I shall show, ﬁrst, that a procedural analysis is able to account for diﬀerences between these expressions which are not explained by the speech act theoretic notion of a non-truth conditional indicator ; and, second, that these diﬀerences demonstrate that the conception of procedural information as a constraint on contextual eﬀects (cf. Blakemore , ) is too narrow, and must be extended to include all information about the inferential processes underlying utterance interpretation, including, for example, constraints on contexts.
. C : G The speech act theoretic distinction between describing and indicating was based on the claim that an utterance does not simply express a proposition, but is used to perform a range of speech acts. Thus although (a–c) would be said to have the same propositional or descriptive content, they are used to express diﬀerent speech acts.
() (a) Tom plays the trumpet.
(b) Does Tom play the trumpet ?
(c) Tom, play the trumpet !
This was taken to suggest that the role of the mood indicators in (a–c) is not to contribute to the propositional content of the utterances that contain them, but should be analysed in terms of their role in indicating or showing what kind of speech act is being performed. In other words, in this framework indicators are taken to encode rather than information.
This idea has been applied to the analysis of explicit performatives (Austin ), parentheticals (Urmson ), mood indicators (Bach & Harnish ), sentence adverbials (Bach & Harnish ) and evidentials (Palmer ). Clearly, the validity of these analyses depends on the validity of the underlying assumption that the interpretation of an utterance crucially involves the identiﬁcation of the illocutionary act it performs.# However, this paper is not so much concerned with the speech act theoretic assumptions underlying these analyses as it is with the question of how they can be applied to the analysis of discourse connectives such as but and so. In what sense can expressions like but and so be regarded as communicating information about the illocutionary force of an utterance ?
The answer according to Grice () is that while mood indicators communicate information about the or - speech act performed by an utterance, discourse connectives like but and so communicate information about a - or - speech act which comments in some way on the interpretation of the central speech act.
For example, in () the speaker performs a central speech act by which she makes a ground-ﬂoor statement that Tom is here and he has forgotten his trumpet, and at the same time a non-central speech act by which she is drawing a contrast between the two conjuncts of the ground ﬂoor statement.
() Tom is here, but he has forgotten his trumpet.
The role of but, according to this analysis, is to signal the performance of this non-central speech act. Since the truth or falsity of the speaker’s words is  For further discussion, see Sperber & Wilson () ; Wilson & Sperber () ; Clark (, ).
determined by the relation of the ground-ﬂoor speech act to the world, a misperformance of a higher-order speech act will not aﬀect the truth value of
the utterance, although it may constitute a ‘ semantic oﬀence ’ (Grice :
In this way, Grice is able to distinguish between those aspects of linguistic meaning which contribute to by a speaker in making an utterance U, and those aspects of linguistic meaning which do not. W is restricted to the propositional truth conditional content of the central speech act. The information carried by but, however, is a part of what is rather than part of what is said, since it is information about the performance of a non-central speech act.
If but signals the performance of a non-central speech act, then this speech act, like all speech acts, must have a propositional content. And indeed, it seems that Grice would want to say that the speech act whose performance is signalled by but in an utterance like () has the content in (), and hence that this is the conventional implicature carried by but.
() There is a contrast between the statement that Tom is here and the statement that Tom has forgotten his trumpet.
But then it is not clear exactly what the higher-order speech act signalled by but is. It cannot be the act of contrasting since the fact that there is a contrast is represented by the propositional content of the act. More generally, it is not clear whether ‘ contrasting ’, ‘ adding ’, or ‘ explaining ’, which, according to Grice, are associated with but, moreover and so respectively, are speech acts in the sense made familiar by classical speech act theory (cf. Austin , Searle ). Perhaps one could say that this higher order act is simply an act of commenting. But then one would have to say that acts signalled by the non-truth conditional discourse connectives – for example, but, so, moreover and after all – are individuated not by their illocutionary properties but by their propositional content, or locutionary properties. That would mean that these expressions are not after all being analysed as illocutionary force indicators, but are being treated as indicators of something propositional.
And this would leave us with the task of specifying what exactly the relationship between an expression like but and the proposition in () is. For it is not clear that whatever is encoded by but appears in this proposition in the same way that, say, what is encoded by trumpet or forgotten appears in the propositional content of the ‘ ground-ﬂoor ’ speech act performed by ().
. C : R ’ It seems that these questions might be answered by Rieber’s () modiﬁcation of Grice’s analysis of conventional implicature, which, he claims, addresses these questions. According to his analysis, expressions like but are . Thus he argues that () should be analysed as ().
() Sheila is rich but she is unhappy.
() Sheila is rich and (I suggest that this contrasts) she is unhappy.
(Rieber : ) As we shall see, it is not clear that Rieber’s assumption that all utterances containing but express a conjoined proposition can be maintained. This section, however, is not concerned with Rieber’s analysis of but in particular, but with the general assumptions underlying his tacit performative analysis of conventional implicature. In particular, it aims to make sense of Rieber’s account of what it means for a linguistic expression to indicate or signal information.
The classical speech act theoretic argument that performatives do not have truth values is no longer universally accepted. Thus according to Recanati () they are self-verifying declarations and hence must be regarded as having a value ‘ true ’. However, Rieber claims that his analysis is compatible with either analysis. Since, according to the classical approach, the performative in () has no truth value and simply ‘ indicates ’ that the propositional content has the force of a suggestion, the truth value of () does not depend on whether there is a contrast between wealth and unhappiness.
At the same time, since according to Recanati’s () analysis, the performative in () is a self-verifying declaration that the speaker is suggesting there is a contrast, its truth is not aﬀected by whether there is a contrast. Thus it would seem that on either approach Rieber’s analysis ‘ gets the truth conditions right ’ ( : ).
However, the question is whether in getting the truth conditions right Rieber has also explained what expressions like but communicate and how they do it.
According to Rieber, his analysis is one in which ‘ what is non-truth conditionally expressed by the discourse connective is part of their meaning ’ ( : , my emphasis). At ﬁrst sight it is not clear whether meaning can be both tacit and conventional. It would seem that to say that the conventional meaning of but is analysed in terms of a performative of the form I suggest that P is to say that there is a linguistically determined relation between but and the information that its utterance constitutes the performance of the act of suggesting that P. On the other hand, to say that this performative is tacit would seem to suggest that the utterance does not actually contain an expression which identiﬁes the act being performed.
It seems that the apparent contradiction here stems from the assumption that means , and hence that in the absence of an actual performative verb as in I suggest, the hearer must the information that the speaker is performing the act of suggesting on the basis of the context and
() There’s a bomb in that car.