The traditions of semantics in linguistics and philosophy have overlapped somewhat in recent decades, but earlier trends and treatments in each discipline have dealt with rather different aspects of meaning. In this essay we hope to unite some of these different threads. It is a special pleasure to contribute to a volume in honour of Sir John Lyons, who has made major contributions to a theory of meaning.

In Lyons' early work (1963, 1968) the meaning of a word was conceived of as its place in the lexical network of the semantic field to which it belonged. A theory of reference - to hook up language and the world - is also a necessary part of the semantic enterprise. We wish to explore how theories of sense and reference can be related.

A word on terminology

The word meaning is used in many different ways, in both philosophy and linguistics. It has been used to describe extralinguistic relationships, that is, between an expression and something in the external world, and intralinguistic relationships - between expressions within a language or between those in different languages. Given the ambiguity of the word meaning, we shall try to avoid using this word, instead employing reference or denotation for word-world connection, and sense, which Lyons has defined as the ‘place in a system of relationships which it contracts with other words in the vocabulary’ (Lyons, 1968: 427).

Introduction

In the section of Lyons' Semantics that deals with the distinction between homonymy and polysemy, he notes that a major criterion ‘is unrelatedness vs. relatedness of meaning … indeed, it is arguable that it is the only synchronically relevant consideration’ (1977: 551). However, he goes on to argue that all attempts ‘to explicate the notion of relatedness of meaning in terms of a componential analysis of the senses of lexemes … have so far failed’ (1977: 552–3). Although Lyons is sympathetic to a treatment of the lexicon that seeks to maximise polysemy at the expense of homonymy, he does not himself go on in that book to reconstruct the notion of relatedness of meaning that such a treatment requires.

Discussing polysemy some nine years earlier, Lyons was a little more explicit about what might be required: ‘Various … types of “extension” or “transference” of meaning were recognized by the Greek grammarians, and have passed into traditional works on rhetoric, logic, and semantics. Meanings that are more or less closely “related” in accordance with such principles are not traditionally regarded as being sufficiently different to justify the recognition of distinct words’ (1968: 406). Metaphorical extension is the only kind of extension explicitly discussed in connection with polysemy in Introduction to Theoretical Linguistics although it is clear from the quotation just given that it was not the only one that Lyons had in mind.

INTRODUCTION

So far we have not exploited to any significant degree the terminological distinction between ‘sentence’ and ‘utterance’ that was introduced in Chapter 1. Nor have we exploited the associated distinctions between Saussure's ‘langue’ and ‘parole’ and Chomsky's ‘competence’ and ‘performance’, which, as we saw in Chapter 1, need to be reformulated, as non-equivalent dichotomies within the system–process–product trichotomy, if we are to avoid the confusion that exists in the account that is given of these technical distinctions in most textbooks.

Much of the work that has been done in formal semantics (in so far as it has been applied to the analysis of natural languages) has been based on the view that languages are sets of sentences and that sentences are used primarily, if not exclusively, to make descriptive statements. Typically, therefore, no distinction is drawn in formal semantics between sentence-meaning and propositional (i.e., descriptive) content. This is clearly a very limited view of what a language is and (as we saw in Chapter 6) of sentence-meaning. It is a view that has been much criticized.

One of the most influential critics in recent years was the Oxford philosopher, J. L. Austin (1911–60), whose ideas have been much discussed, not only by philosophers, but also by linguists (and representatives of many other disciplines). In this chapter, we use Austin's theory of so-called speech acts as a departure-point for the analysis of utterance-meaning that follows in Chapters 9 and 10.

INTRODUCTION

This chapter follows on from the preceding one and looks at two historically important and highly influential theories of sentence-meaning which, since the mid-1960s, have been associated with the attempt to formalize the semantic structure of languages within the framework of Chomskyan and non-Chomskyan generative grammar.

The first is the Katz–Fodor theory of meaning, which originated in association with what we may now think of as the classical version of Chomsky's theory of transformational-generative grammar. The second theory is a particular version of possible-worlds semantics, initiated also in the late 1960s by Richard Montague, and, having been further developed by his followers, is now widely recognized as one of the most promising approaches to the truly formidable task of accounting for the propositional content of sentences in a mathematically precise and elegant manner.

The treatment of both theories is very selective and almost completely non-technical. I have been more concerned to explain some of the basic concepts than to introduce any of the formalism. At the same time, it must be emphasized that modern formal semantics is a technical subject, which cannot be understood without also understanding the mathematical concepts and notation that are a part of it. This chapter should definitely be read in conjunction with the more specialized introductions to formal semantics mentioned in the ‘Suggestions for further reading’.

INTRODUCTION

As we saw in the preceding chapter, it is generally agreed that the words, phrases and sentences of natural languages have meaning, that sentences are composed of words (and phrases), and that the meaning of a sentence is the product of the words (and phrases) of which it is composed.

But what is a word? And do all natural languages, in fact, have words? These questions are not as easy to answer as they might appear to be at first sight. One reason is that the term ‘word’ is ambiguous, both in everyday usage and also as it is employed technically by linguists. Words may be considered purely as forms, whether spoken or written, or, alternatively, as composite expressions, which combine form and meaning. To complicate matters further, the term ‘form’ is employed in several different, though related, senses in linguistics. One of my principal aims in this chapter is to sort out these different senses of ‘word’ and ‘form’ and to establish notational and terminological conventions for avoiding ambiguity and confusion.

Another reason why it is not as easy to say whether something is or is not a word as non-linguists might think – or to say whether all natural languages have words – is that several different criteria come into play in the definition of words, both as forms and as expressions, and these criteria are often in conflict.

A PRINCIPLED BASIS

Introduction

My concern in this chapter is to reject semantic holism and defend molecular semantic localism. The methodology described in the last chapter will play a central role.

I have pointed out (1.1) that holism is usually accompanied by a no-principled-basis consideration along the following lines:

Many who are not sympathetic to holism are impressed by the no-principled-basis consideration. Lepore and Fodor (1993) are striking examples. Quine has shown that an epistemic criterion like aprioricity cannot provide the principled basis. Lepore and Fodor think that the possibility of a nonepistemic criterion providing the basis is near enough empty:

So, if we accept that some inferential properties constitute meanings, we have no principled basis for denying that they all do. Molecular localism is untenable. Fodor, at least, concludes that no inferential properties count, thus embracing an atomistic localism.

Three important questions get insufficient attention in semantics. What are the semantic tasks? Why are they worthwhile? How should we accomplish them? The central purpose of this book is to answer these “methodological” questions and to see what semantic program follows from the answers.

It is troubling that much semantic theorizing proceeds with inexplicit reliance on apparently ad hoc views of the semantic tasks. Thus it is common to take for granted that semantics is concerned with truth and reference. I think that this view is right, but why is it right? What can we say to someone who disagrees, claiming that semantics should be concerned with, say, warranted assertability or “use”? Furthermore, it is troubling that, in attempting to accomplish the semantic task, we all go in for “intuition mongering,” even those of us who are naturalistically inclined and skeptical of the practice (e.g., Jerry Fodor 1990: 169). Broadly, it is troubling that we seem to lack a scientifically appealing method for settling the disputes that bedevil semantics. In Chapter 2, I propose a view of the semantic tasks by looking at the purposes we attempt to serve in ascribing meanings. And I propose a way of accomplishing them. This methodology has a place for intuitions, but it is the same limited place that they have elsewhere in science. I think that applying this methodology will help with all semantic issues. In this book I shall use it in the hope of settling some, including some of the most notorious.

A by-product of this methodological discussion is a naturalistic account of the thought experiments characteristic of “armchair” philosophy.

A REPRESENTATIONALIST PROGRAM

Representationalism

In the last chapter I have argued against the holistic threat to semantic localism. I have claimed that we ascribe localistic properties to our words for semantic purposes (descriptive), that we ought to do so (normative), and hence that these properties are meanings (basic). This raises the question: Which localistic properties do we and ought we to ascribe for semantic purposes? That question will be my main concern in this chapter. Certain wellknown arguments for eliminativism and revisionism will be set aside until the next chapter. My aim is to present a localistic program rather than a detailed theory.

The program I shall be presenting will be Representationalist. Representationalism is the view that meanings are entirely constituted by “representational” properties (3.11). So the meaning of a sentence is exhausted by the properties that determine its truth conditions} And the meaning of a word is exhausted by properties that determine its reference. Representationalism made an appearance in Chapter 1 in the guise of the Fregean assumption, but it played only a minor role in the critique of the case for holism. In Chapter 3, I argued that Representationalism counted decisively against holism, but I did not rest my case for localism on it. In the rest of this book, with holism rejected, I shall be arguing for Representationalism.

INTRODUCTION

Semantic Holism and Semantic Localism

At its most extreme, semantic, or meaning, holism is the doctrine that all of the inferential properties of a token in language or thought constitute its meaning. This doctrine is opposed by semantic localism, which, at its most extreme, denies that any of the inferential properties of a token constitute its meaning.

Despite its prima facie implausibility, semantic holism is ubiquitous. It has, as Jerry Fodor says, “something of the status of the received doctrine in the philosophy of language” (1987: 57). And it is urged, or taken for granted, in psychology and artificial intelligence. In this chapter, I shall look critically at the case for semantic holism.

The case can always be made to fit the folio wing “basic” argument:

1. Some of a token's inferential properties constitute its meaning.

2. If some of a token's inferential properties constitute its meaning then they all do.

3. So, all of a token's inferential properties constitute its meaning.

Fodor is an extreme “atomistic” localist: He resists this argument by rejecting premise (1) (pp. 73–95). Fodor's major reason for rejecting (1) is quite clear: He thinks that it leads inexorably to holism, which he regards as “a crazy doctrine” (p. 60) threatening Life As We Know It. He thinks that (1) has this unfortunate consequence because he accepts (2). Indeed, he is as committed to (2) as the most fervent holist.

I agree with Fodor's view of the holistic conclusion but think that he is quite wrong about (2). My aim in this chapter is to reject the case for (2).

Two things led to this book. The more immediate, but less important, cause was my concern about semantic, or meaning, holism. Holism has, as Jerry Fodor says, “something of the status of the received doctrine in the philosophy of language” (1987: 57). And it is urged, or taken for granted, in psychology and artificial intelligence. Yet it seemed to me, as it did to Fodor, clearly false (“crazy” was his word). So, in 1989, I set out to show this.

First, I had to show that the arguments for holism were no good. The main argument stems from Quine: The localist idea that some but not all inferential properties of a token constitute its meaning (or content) is alleged to yield an analytic-synthetic distinction with epistemologically objectionable consequences. You can accept this argument without becoming a holist, of course, if you are prepared to adopt an “atomistic” localism according to which no inferential property ever constitutes the meaning of a token. That is Fodor's path. However, atomism strikes me as implausibly extreme. Very likely, the meanings of some tokens are atomistic, but surely the meanings of others – perhaps ‘bachelor’ is an example – are not. I want to defend a “molecular” localism, according to which a few of the inferential properties of a token may constitute its meaning. I think that I can have what I want because I reject the Quinean argument: Molecular localism does not have epistemologically objectionable consequences unless it is saddled, gratuitously, with an epistemic thesis. I also reject other arguments against there being a “principled basis” for the molecular localises distinction among inferential properties.

INTRODUCTION

The Problem

Semantics is a veritable Balkans of the intellectual world. Localists war with holists, truth conditionalists with verificationists, deflationists with substantivists, direct-reference theorists with Fregeans, one-factor theorists with two-factor theorists, and so on. An army of enthusiasts for narrow content have occupied the territory formerly held by the proponents of wide content. Finally, no settlement of these disputes seems to be in sight.

One sound stands out in these battles: the clash of semantic intuitions. Indeed, sometimes that is the only sound to be heard. Intuitions are almost always aired in “thought experiments.”

This reliance on intuitions may be untroubling from some perspectives because it seems to exemplify the characteristic method of “armchair” philosophy. Yet it is surely troubling from the naturalistic perspective that I favor. According to naturalism, semantics is an empirical science like any other. Intuitions and thought experiments do not have this central role elsewhere in science. Why should they in semantics?

This question leads to the general ones that are the main concern of this chapter: How should we get to the truth in semantics? How should we go about settling semantic disputes? What is the right methodology for semantics?

A naturalistic approach to these questions can only hope for modest answers. We cannot expect to make more progress with the methodology of semantics than has been made with scientific methodologies in general. And we know how limited that progress is. It has turned out to be very difficult to say how we should get to the truth and settle disputes in science. My hope is only to bring semantic methodology close to other scientific methodologies.

INTRODUCTION

What Are Eliminativism and Revisionism?

The view of meanings that has emerged in the last chapter is realist and conservative: Thoughts and utterances have meanings, and meanings are, minor revisions aside (4.12), the properties that we already ascribe for semantic purposes. And meanings are, as many suppose, truth referential, “Representationalist.” But I have not yet seriously considered the radical alternatives of eliminativism (or nihilism) and revisionism. In this chapter, I shall do so. We need to start by clarifying both these alternatives. Eliminativism. Eliminativism about F's is the doctrine that there are no F's. It is important to note that this eliminativism needs to be accompanied by a background assumption about what it would be like for there to be F's, about what is essential to being an F, about the nature of F-hood. For, it is not sufficient simply to say there are no F's; one needs an argument. And that argument will have the following form:

The first premise is the background assumption. A possible realist response is then to deny the assumption: F's are not essentially G. How do we settle this disagreement? That is precisely the methodological issue discussed earlier (2.10). We saw then that it may be difficult to settle the issue even by the “ultimate” method.