We have discussed the fact that ideas or concepts occur to people, are the objects of various propositional attitudes, are parts of other ideas, are expressed by words, are acquired and possessed, represent things, and have contents or objects. Another characteristic property of ideas, one that lends itself to mathematical representation and experimental study, is their association with other ideas. We will observe that the association of ideas with sensory images and percepts is an important ingredient in the automatic use and understanding of language. We will also show that neural networks provide a good model and plausible basis for associative networks. But we will examine association mainly because it is a phenomenon distinctive of thought and ideation as opposed to belief and desire. Moreover, it is a causal relationship among ideas distinct from the structural relations obtaining among parts of the same thought. As such, the knowledge that ideas can be associated increases our understanding of the notion of an idea, as does recognition of the fact that ideational structure cannot be reduced to the relation of association.

Historically, philosophers and psychologists who gave ideas a central place in their theories tended to be associationists. Association psychology was the attempt to analyze and explain all mental phenomena in terms of the association of ideas. This resulted either in ignoring critically important mental phenomena or, more commonly, in applying the term “association” wherever two distinct elements were related in any way, which emptied the term “association” of all useful content.

To fully understand the thesis that basic word meaning is conventional speaker meaning or expression, and to avoid misguided criticism, we need to know what conventions are. Briefly, we may say that “convention” should be understood as denoting arbitrary social practices or customs. Constituting standards of correct usage, they are one type of rule. This chapter will be devoted to clarifying what conventions in this sense do and do not entail. We will develop Lewis's idea that conventions are regularities in action that are socially useful, self-perpetuating, and arbitrary. There is no requirement, on our definitions, that the regularities be nearly universal or mutually known. And of course there is no requirement that conventions result from agreements. We will take some pains to explain how word usage can be objectively correct or incorrect if it is arbitrary and conventional. The fact that conventional regularities may have exceptions allows languages to change over time, and the self-perpetuating character of conventions coupled with linguistic variation leads to evolving families of languages. The same facts make it difficult to assign precise boundaries to languages, as we will see in Chapter 11.

DEFINITION

It is often said that conventions are agreements. Indeed, in one sense, the word “convention” denotes an international agreement, and in another denotes formal meetings designed to secure agreements. But in the sense we are concerned with, most conventions, including linguistic conventions, are not and did not result from agreements.

We have defined cogitative speaker meaning in terms of the expression of thoughts or ideas, and cognitive speaker meaning in terms of the expression of belief. Communication is the effective expression of beliefs, thoughts, or other mental states, and reference is the verbal expression of a certain range of ideas. Word meaning or expression in living languages has been defined in terms of conventional speaker meaning or expression. We turn our attention now to the fundamental notions of thought and ideation.

Belief, desire, and intention have received considerable attention in the philosophical literature. Occurrent thought as a specific propositional attitude has been generally neglected. The notion of an idea has long been suspect, moreover, and ideational theories of meaning are widely regarded as having been thoroughly discredited. It is necessary, therefore, to clarify the relevant notion of thought and ideation, and to respond to well-known objections. I will survey the principal similarities and differences between thought and belief, and stress the distinctive causal role of thought. We will look at the range of terms that express thought in English. While I can present only part of the case here, I hope to make it evident that conceptually, thought is as primitive as belief and desire, while ontologically it is more general and more fundamental. The often mentioned “belief-desire psychology” is really the “belief-desire-thought psychology.”

THE COGITATIVE SENSE OF THOUGHT

The word “thought” is at least doubly ambiguous. Like “belief,” “desire,” and “intention,” it suffers from the act-object ambiguity.

Having defined speaker meaning in terms of expression and expression in terms of the intention to produce indications of belief or other mental states, we turn our attention now to communication and other concepts that entail communication. We have seen that the Gricean assumption that speaker meaning is the attempt to communicate distorted the Gricean analysis of speaker meaning in a number of ways, and we have suggested that his analysis would fit communicating, informing, and telling much better. We will now see that Grice's conditions are not completely appropriate even for these semantic acts. We will define communication in terms of expression and recognition, and then define informing and telling in terms of communicating. We will therefore adopt the Gricean strategy of reducing communication to intention, while again specifying different intentions from the Gricean ones. In the process, we will draw an important distinction between communicating to and communicating with, and reject the popular transmission model of communication.

Even though we have distinguished speaker meaning from communication, we will nonetheless go on to affirm in Part II that communication plays an important role in word meaning, as the common interest that sustains the conventions in virtue of which words have meaning. The use of words to express certain ideas is conventional, moreover, only when they are conventionally used to communicate them. So the idea that word meaning in living languages depends on communication in some way will be upheld.

According to the neo-Gricean analysis presented in Chapter 8, word meaning is conventional cogitative speaker meaning. That is, what an expression means is determined by what idea people conventionally use it to directly express. This thesis works very well for individual words, stock phrases, dead metaphors, and idioms. “Green” means green only because it is conventional for English speakers to use the word “green” to mean the color green. The phrase “kicked the bucket” means “died” only because people conventionally use it to mean that. The basic neo-Gricean analysis does not, however, account for the constructive and recursive character of the semantic rules of a language. That is, the analysis does not account for the way in which the meaning of a complex word, phrase, clause, or sentence is normally determined by the meanings of its components. In this chapter we will account for the compositionality of linguistic meaning by adding a recursive element to the neo-Gricean analysis. This element is provided by the fact that there are conventions whereby word structures are used to express idea structures. We will address objections that have been leveled against the notion of compositionality, and argue that projection rules need not be restricted to block generation of “anomalous” meanings.

THE PRODUCTIVITY PROBLEM

The deficiencies of the neo-Gricean analysis can be seen most starkly by reflecting on the fact that every natural language contains a large – indeed, infinite – number of meaningful sentences that have never in fact been uttered.

I began work on thought, belief, and desire shortly after I graduated from the University of Michigan in 1973, inspired by Alvin Goldman and his A Theory of Human Action, along with Stephen Stich, Arthur Burks, John Perry, and Jaegwon Kim. That work grew into my doctoral dissertation (Princeton University, 1977), directed by David Lewis, Gilbert Harman, and Richard Jeffrey. I remain indebted to these outstanding philosophers not only for key ideas but also for instilling a love of philosophy. The dissertation became a book-length manuscript entitled “Elements of Psychology: Belief, Desire, and Thought.” When a chapter on meaning took on the proportions of a book all by itself, I decided to first complete the present volume, Meaning, Expression, and Thought. Many of the ideas on thought presented in Part III were first developed in my dissertation and elaborated in “Belief, Desire, and Thought.” I use them here to provide the psychological foundations for the theory of meaning developed in the rest of this work. This book was delayed by my recent Implicature (1998), which explains why Grice's great “synthetic” project gets so much less attention here than his “analytic” project. I wrote Meaning, Expression, and Thought, furthermore, in tandem with my forthcoming Nondescriptive Meaning and Reference, which applies the expression theory of meaning to names, indexicals, and other special cases, develops the expression theory of reference in greater depth, and shows how referential semantics can be treated in the expression theory.

Word meaning is relative to a language. The neo-Gricean thesis that word meaning is recursive, conventional, cogitative speaker meaning holds only for living languages like English. Related theses hold for dead languages and idiolects. Word meaning is determined by stipulation rather than by convention in artificial languages (at least initially), by individual custom in idiolects, and by prior practice in dead languages. What we can say in general is that word meaning is “established” speaker meaning. This chapter distinguishes the different ways in which speaker meaning may be established by examining different types of languages. We will focus on the convention-dependence of living natural languages and on the linguistic lineages that their evolving conventions create. The self-perpetuation of conventions coupled with linguistic diversity has generated thousands of genetically related natural languages in much the same way that reproduction and individual variation have generated millions of genetically related biological species. Natural languages are discovered when a community is found that has a previously unknown set of linguistic conventions. What words mean now in a living natural language is determined by the current lexical and constructive conventions of those speakers whose conventions have evolved from the conventions of prior users of the language. This will enable us to avoid the apparent circularity inherent in saying that English is defined by the conventions of English speakers.

This chapter will define concept possession, and distinguish it from the stronger notions of understanding and mastery. To possess a concept, we will argue, is to have conceived it and remain capable of conceiving it. We will show that nominalist and information-semantic definitions of concept possession are wide of the mark, while recognition, knowledge, and inferentialist theories are too strong.

POSSESSING CONCEPTS

Fodor has remarked that “[i]t's a general truth that if you know what an X is, then you also know what it is to have an X,” and that this applies to concepts in particular (1998a: 2). There are many senses of “have” for which Fodor's generalization fails. A child may know what a baby is without yet knowing what it is to have a baby, and may know what a woman is without yet knowing what it is to have a woman. I believe that the having of concepts is a less obvious exception. I have defined concepts as parts of thoughts, but that does not tell us what it is to have a concept.

“Having” a belief is the same as believing the belief, and “having” a thought is the same as thinking it. To “have” an idea is to conceive it. Despite this pattern, having a concept, as this phrase is commonly used, must be distinguished from conceiving a concept. We have countless concepts that we are not currently conceiving. I have had the concept of neutrons since grade school.

We now turn our attention from speaker meaning to word meaning. One of the elementary facts about word meaning is that words have different meanings in different languages. The written word “rot” means red in German and decay in English. “Hood” means the engine cover in American English and the top of the car in British English. So we need to say something first about what languages are. As David Lewis (1975) has observed, languages must be distinguished from language. Languages are things like French, German, and English. French is a language, but is not language itself. Language is a human activity, in which languages are used. It includes speech, writing, and other types of symbol use. While linguists and philosophers sometimes debate as to which is primary, it should be clear that languages and language are complementary subjects.

We will first examine what a language is, and then analyze what it is to use one. This will enable us to define what it is for a word to mean something on a given occasion (“applied” word meaning). We will be using “languages” in a very general sense, to denote symbol systems of all kinds, including codes, signal systems, sign languages, and artificial languages. The distinctive features of living, natural languages will be discussed in Chapter 11, culminating in a definition that implicitly defines what it is for a word or other linguistic unit to express or mean something in a living language.

I have elaborated the view – for both speaker meaning and word meaning, and for the vast bulk of expressions – that meaning consists in the expression of ideas, and that meaning is determined by the idea expressed. The word “dog” means something, for example, because it is conventionally used to express an idea. And it means “dog” rather than “cat” because it expresses the idea of a dog rather than the idea of a cat. A particular speaker means “dog” by a word when he or she uses it to express the idea of a dog. Hence the view that I have been developing is an ideational theory of meaning. I have acknowledged that for a relatively small set of expressions – interjections and the like – meaning consists in the expression of mental states other than ideas (see §2.5, §7.3). So, more generally, I have been developing a mentalistic theory of meaning. I have concentrated on words for which the relevant mental states are ideas, and shall continue to do so.

A major lacuna in classical ideational theories was that after the identification of meaning with the expression of ideas, little was said about expression. Yet without a detailed understanding of the relation of expression, the tautological character of formulas like “e means μ iff e expresses the idea μ” will wrongly suggest that ideational theories are vacuous.

Relativism in contemporary thought

Looking back a century, one can see a striking degree of homogeneity among the philosophers of the early twentieth century in terms of the topics central to their concerns. More striking still is the apparent obscurity and abstruseness of those concerns, which seem at first glance to be far removed from the great debates of previous centuries, between realists and idealists, say, or rationalists and empiricists.

The German philosopher Gottlob Frege (1848–1925) devoted his career to the foundations of mathematics and was rewarded with profound indifference from his fellow philosophers and mathematicians. The English philosopher Bertrand Russell (1872–1970) devoted his early work to exactly the same issue, culminating in Principia Mathematica, which was such an intellectual effort, he said, that it rendered him incapable of ever producing such detailed work again. In his early work the Austrian philosopher Ludwig Wittgenstein (1889–1951) focused even more narrowly on a critique of the work of Frege and Russell on the meaning of the logical constants. He also became aware of how much could be gained from such minute examination, saying that his work had widened out from the foundations of logic to the nature of the world. The Austrian philosopher Edmund Husserl (1859–1938) also started with the philosophy of geometry, before developing the phenomenological method, which was geared toward answering questions emerging from his earlier work.

Chapters 2–5 have examined relativism in a number of specific areas of philosophy. In Chapter 2 it was argued that it is incoherent to think of truth as being relative. The argument centred on the claim that the possibility of contradiction was essential to communication and argument. By holding to relative conceptions of truth, one rules out the possibility of contradiction and so endangers communication and argumentation. Because of the significance of contradiction, it furthermore became clear that the limits of logic are established in acknowledging it. One may have alternative systems of logic, but there is a presumption in favour of the dominance of systems holding to the law of non-contradiction. Any calculus that allows contradictions can only be used in a restricted way, and not a universal one. It has to be governed by a higher-level logic that obeys the law of non-contradiction.

Much of what motivates people to accept relativism can be accommodated by ontological relativism, as argued for in Chapter 3. This position attempts to overcome traditional dichotomies such as realism and idealism by rethinking and reconceptualizing the relationship of thought, language and world. The new picture allows for a variety of alternative sets of categories by which the world is mediated to us. None of these is held to be true of the world in itself. The argument goes that that way of thinking about the issues has been superseded.

There are a multiplicity of different positions to which the term epistemological relativism has been applied. However, the basic idea common to all forms denies that there is a single, universal means of assessing knowledge claims that is applicable in all contexts. Many traditional epistemologists have striven to uncover the basic process, method or set of rules that allows us to hold true beliefs. Think, for example, of Descartes's attempt to find the rules for the direction of the mind, Hume's investigation into the science of mind or Kant's description of his epistemological Copernican revolution. Each philosopher attempted to articulate universal conditions for the acquisition of true beliefs. Epistemological relativism can be best expressed negatively, just as ontological relativism was expressed in Chapter 3. It rejects an absolutist conception of epistemological justification: that always and everywhere there is a sole fundamental way by which beliefs are justified.

The multiplicity of positions labelled epistemological relativism arises due to the fact that the rejection of this absolutist view yields a variety of possible positions of varying degrees of strength. Philosophers have differed greatly in the amount of relativity they allow into the epistemological process. At a level that hardly anyone rejects, there is a degree of relativity in perceptual knowledge, depending on the perceptual apparatus and location of the individual. Yet perceptual relativism is often regarded as a primitive feature of cognition, below the level at which questions about the justification of belief arises and which is compatible with an absolutist conception of justification.