Chapter 1
What is a theory of meaning?
The central task of the philosophy of language is to illuminate or explain the concept of meaning as it applies to language. One way in which we might do this is to reflect on and possibly offer an analysis of our concept of linguistic meaning. Another is to reflect on our use of the term âmeaningâ in application to linguistic items. Dummettâs recommendation (and he takes himself to be following Davidson here) is that we do best to examine the concept by trying to get clear about the form that a theory of meaning for a particular language, a meaning-theory,1 should take. Weâll need to probe in much more detail Dummettâs ideas about the nature of a meaning-theory. However, for the moment, letâs characterize a theory in terms of what it should achieve. A theory of meaning for a particular language will provide the means systematically to generate specifications (or characterizations) of the meaning of each expression in the language. Our only model of how such specifications might be systematically generated as part of a theory is to conceive of the theory as a deductively connected set of propositions; axioms will provide the basis of the process of generation and the specifications will emerge as theorems of the theory. In other words, we shall have a basic set of truths about the language, truths which capture the meanings of primitive expressions and which explain how the meanings of complex expressions depend upon the meanings of their components. Given some expression of the language we shall then be able to infer the specific theorem which relates to its meaning. This sketchy characterization of a meaning-theory doesnât, of course, tell us very much about the nature of such a theory but it provides a sufficient contrast with other approaches for us to begin to weigh them up.
Let us contrast the case of knowledge with that of meaning, as Dummett does. Questions concerning the nature of knowledge can be pursued by asking questions about ascriptions of knowledge: when are these justified and what content do they have? But questions about meaning cannot be focused in this way. We cannot examine the nature of linguistic meaning merely by confining our attention to sentences of the forms: x means that y; or, x means the same as y. There are, it seems to me, two reasons for this. The first is that it is contentious that direct specifications of meaning (specifications of the first form) will, in general, be a possibility or that specifications of sameness of meaning provide all we want to capture in an explanation of meaning. Secondly, even if these queries could be quelled, what we are interested in is generally applicable conditions for determining the correctness of specifications of meaning. That is, we donât actually want to possess a set of clauses specifying meaning; we already understand our own language and are not engaged in the business of translating a foreign language. Rather we want to know how these clauses would be generated and what ratifies them as correct. So what we are interested in is something general, namely, a method for producing specifications of meaning and for testing the acceptability of the results. But now we seem to have regenerated Dummettâs approach. For these constraints will be given content in terms of how they govern the form taken by a meaning-theory: that is, a general specification of meanings.2
An additional point is that the nature of an expressionâs meaning is at least as much a function of the sorts of explanation we use to teach its use and of features of that use, as it is a function of our use of the expression in contexts involving explicit discussion of meaning. So were we to focus on speakersâ use of the term âmeaningâ we might be ignoring substantial information about the concept, information which would form part of the data for a meaning-theory. Conversely, it seems that speakers of a language lacking in vocabulary for talk of meaning would nevertheless possess that concept (otherwise they couldnât communicate or succeed in teaching language). Thus it seems that our concept of meaning is illuminated more thoroughly by reference to relevant uses of language rather than by the explicit use of the concept itself. The way to get round the potentially impoverished use of the term âmeaningâ is to think about how the concept would feature in an attempt, quite generally, to specify the meanings of expressions, that is, to investigate the form of a meaning-theory.
The subject matter of a meaning-theory
How would one set about constructing a meaning-theory? Clearly the theory will begin with linguistic behaviour and so, in some sense, will have to fit with it. So, on the one hand, we shall talk of ascending from facts about linguistic behaviour to the mechanics of the theory and, on the other, of descending from the inner workings of the theory to an account of linguistic behaviour. However, the terms in which linguistic behaviour is given to us, and the nature of the fit between that and the theory, need to be explained. That explanation will depend on what we want the theory for and what sort of theory we think it is possible to construct.
Dummett is quite clear that we donât want an empirical theory that will enable us to predict linguistic behaviour. Predictive success is neither necessary nor sufficient. It is not sufficient since a theorist equipped only with a predictively successful theory would not be able to exploit the knowledge she thereby possesses in order to pursue her own ends in conversation. She may be able to pass unnoticed in the community but she could only exploit her knowledge of the theory to do such things as to express her view or ask for information if she conjoined that knowledge with an antecedent grasp of the significance of certain sorts of linguistic acts. But then that knowledge should properly be articulated in the theory. Predictive success is not necessary since a theorist might have all the abilities just mentioned and have them in virtue of her knowledge of the theory yet be unable to predict linguistic behaviour in the way a good empirical theory might. A theory of this sort would convey all the knowledge necessary to illuminate the concept of meaning; it would be a meaning-theory. So we donât want a theory that accounts for linguistic behaviour from an external point of view (and which is thus constrained only by predictive success), we want a meaning-theory that makes explicit the knowledge implicated in linguistic behaviour. Thus a theory of meaning is a theory of understanding. The connection between the knowledge thus articulated by the theory (the knowledge possessed by the theorist) and the knowledge possessed by actual speakers will emerge as a delicate matter to be tackled below.
Systematicity, compositionality and meaning-theories
The meaning of an utterance is a product of that of its component words and the way in which they are combined in the utterance. This apparent platitude would appear to explain a speakerâs capacity to understand novel utterances: the utterance is understood by deploying oneâs grasp of the meanings of the component words and of the effect of the manner in which they are combined in the utterance. Since words and modes of combination make repeatable contributions to the meanings of utterances, the meaning of an utterance is determined systematically by that of its parts and their mode of combination. An adequate portrayal of the meanings of utterances will exhibit these systematic dependencies; an account of meaning will be systematic and so will deserve the accolade âmeaning-theoryâ.
What assumptions â what views of language â are implicated in this, not unnatural, train of thought? What shape would a systematic account of meaning take? What relation would it bear to speakersâ understanding of language? What is the relation between the (supposed) systematicity of the meaning-theory and the (supposed) compositionality of language? What justification is there for insisting either on systematicity or on compositionality? This and the next chapter attempt to make some progress on Dummettâs answers to these questions. I begin with a contrast between two conceptions of language.
Interpretation and convention
Suppose someone makes an utterance. To an appropriately qualified audience the utterance has a certain significance. It may convey information about the world, about the speakerâs state of mind or about what the speaker requires of the hearer, and may convey mixtures of these sorts of thing. Linguistic meaning might be seen as a product of the process of interpretation whereby the hearer attaches a certain significance to the utterance. On the other hand, linguistic meaning might be seen as a resource exploited in the process of interpretation. Davidson exemplifies the former, interpretivist outlook. In contrast, Dummett argues that linguistic meanings are products of conventions that enable interpretation (where necessary) to take place. The basis of systematicity in the theory of meaning is intimately connected with the rejection of the interpretivist in favour of the conventionalist approach. The debate between Dummett and Davidson is very illuminating here.
An utterance, a speech act, may be â to give an inexhaustive list â an assertion, a command or an enquiry. That is, utterances come in a number of varieties depending on what the speaker is attempting to achieve by means of the utterance: whether she is trying to convey information, request information, require that something be done. Utterances can be distinguished by what, following Frege, has come to be called the force that attaches to them. As social creatures we are adept at picking out the force with which an utterance is made, and doing so is an essential element in communication: one couldnât learn language if one failed to be sensitive to the force of an utterance, and a translation manual would be little help in communicating with a foreign people were one blind to matters to do with force. So, to be a competent speaker, one must be sensitive to the force of utterances, but the constitution of this sensitivity needs to be explained.
Force then is a property of utterances. But utterances involve the use of sentences, and sentences themselves come in a number of forms. Depending on word order, punctuation and so on, a sentence might be indicative, imperative, optative or interrogative. Consider âThe apple is peeledâ, âPeel the appleâ, âWould that the apple were peeledâ and âIs the apple peeled?â. That is, a sentence might have one or another mood.
Now one way of explaining a speakerâs sensitivity to force would be to reduce it to a sensitivity to mood. If, say, an assertion is an utterance of an indicative sentence then a sensitivity to the mood of the sentence will deliver a sensitivity to the force of the utterance (and similarly for the other moods and forces). In âMoods and Performancesâ Davidson (1984) puts this simplistic scenario under stress and thus questions the underlying thought, namely that a sensitivity to force is based on a sensitivity to mood. If I read him aright his thought is to reverse the direction of dependency here: mood is a product of a theory of interpretation whose object is the interpretation of utterances.3 So interpreting an utterance as having a certain force is conceptually prior to the assignment of moods to sentences. The point relates to the issue between the interpretivist and conventionalist. For, if mood is prior to force, it is plausible to think of hearers interpreting speakers by availing themselves of conventions governing the use of sentences. Conversely, if force is prior to mood, then the assignment of a certain significance to utterances of sentences of a certain form will purely be a product of the process of interpretation.4
Dummett concedes that Davidsonâs argument shows there is no simple connection between the force of an utterance and the mood of a sentence. However, he thinks that each type of force must be explained by reference to a given practice. The mood of a sentence then may fit it for inclusion in one or another such practice, so it may have one or another force. But the force that an utterance possesses will be (very largely) determined by conventions governing the use of sentences in appropriate moods. So although mood does not determine force, force is determined by a certain conventional use being made of a sentence in a given mood. The point of making such an utterance will then determine answers to further questions â those questions that most occupy Davidson â about the role of the utterance.
I am here less interested in trying to decide between these points of view than in exhibiting the consequences each has for systematicity in the theory of meaning. Since Dummett distinguishes between what is said by means of an utterance and the point of saying it he can relegate the latter to an account of the general manner in which we set about estimating the intentions of others. This is not an essentially linguistic process and so does not fall within the scope of a meaning-theory for a language. If one refuses to acknowledge Dummettâs distinction (as Davidson does) then the nature, and thus significance, of an utterance will not be a matter of convention. Linguistic meaning then will not be construed simply as a product of convention but will always be, in part, a product of a speakerâs purposes and intentions. So the subject matter for a systematic theory of meaning vanishes. In contrast, Dummettâs distinction combined with the further thought that what is said is determined by (linguistic) conventions allows for the systematic portrayal of these conventions in a meaning-theory. So the Dummettian perspective is a prerequisite for systematicity in the theory of meaning.
This may appear to be an odd claim given Davidsonâs predilection for a theory of meaning as a systematic theory of truth. But there is really no tension here since for Davidson the object of the meaning-theory is not a public language such as English or a dialect thereof. Rather a hearer will construct a meaning-theory in an effort to make sense of a speakerâs utterances (on a given occasion). The theory is subservient to the process of interpreting the speaker; it is an instrument in this process. So, for our purposes, it is irrelevant that Davidson thinks that the most efficient way of engaging in this process of interpretation is via the construction of a systematic theory of a certain form. The main point is that he eschews a systematic meaning-theory for a language in favour of a notion of interpretation. Indeed, in denying that languages, as conceived of by philosophers and linguists, exist, Davidson precisely denies the object of such theories.
Sense and force
If we accept Dummettâs view then the difference between mood and force becomes less important. I want now to discuss the relation between the sense of an utterance and its force. What drives the idea that there is a common element between, say, the utterances of âThe window is closedâ, âIs the window closed?â and âClose the windowâ, that this common element is a shared content or sense and, finally, that each of these utterances attaches a different force to the same content? In other words, what, if anything, inclines one to the view that an utterance can be dissected into a sense used with a certain force? What would be the consequence of noting that these sentences have very different uses and thus of being sceptical that there is a general and systematic set of relationships here to be described? That is, how essential is the common content view to an insistence on systematicity in the theory of meaning?
There is an intuitive plausibility to the common content view. It does appear that our three sentences above share a content and that the first puts this content forward as true, the second asks whether it is true and the third commands that it be made true. English clearly includes a procedure by which indicatives, quite generally, can be transformed into interrogatives and imperatives (by means of changes in word order and punctuation). Moreover it seems that an understanding of this procedure combined with a general grasp of the practice of using commands allows one to move from an assertion that one understands to a command that one understands (and vice versa, with similar remarks applying for the other sorts of force). But this last point about understanding is an addition to the banal observation about English. There is nothing surprising or thought-provoking in the idea that we can manufacture an imperative from an indicative and vice versa. What should give us some cause for thought is the idea that in doing so we are performing an operation on a given content (which is preserved). The additional thought about understanding goes hand in hand with this about content. For if the content of an imperative as used in a command can be seen as a product of an operation on the content of an indicative as used in an assertion then a grasp of the significance of the operation together with a grasp of the content of the indicative will deliver a grasp of the content of the imperative. Conversely, if an understanding of the imperative can be derived from a grasp of the significance of the grammatical transformation, indicative to imperative, and an understanding of the corresponding indicative then the content of the imperative is a function of the content of the indicative. So far we have a point about the structure of understanding. Given that structure, one might then think that there is a further question about the relative priorities of elements in that structure: should we see grasp of an imperative as derived from a grasp of the indicative or vice versa?
The following line of thought not merely supplies a reason for the claim about the structure of understanding but does so by addressing the question of priority. An imperative requires a certain action of its addressee. To understand the imperative is to know what sort of action is required, what constitutes compliance with the comman...