The first part of this chapter will discuss aspects of Veltman's chapter in this volume ‘Data semantics and the pragmatics of indicative conditionals’, and particularly his semantics for conditionals, while the second will comment on more general methodological issues having to do with relations between three seemingly disparate theories: (1) Grice's theories of meaning and conversational implicature, (2) the Bayesian theory of decision making, and (3) my own probabilistic theory of conditionals. The discussion of Veltman's chapter will presuppose familiarity with technical aspects of the theory presented there.

The central concept of Veltman's theory is that of a sentence being true in an information state in an information model, which is a ternary relation between sentences, information states, and information models. What we want to ask is how this ternary relation is related to the property of truth, which is what Tarski (1944) insists the sort of truth that satisfies Convention T must be.

INTRODUCTION

This chapter is about the initial flowering of conditionals, if-(then) constructions, in children's spontaneous speech. It is motivated by two major theoretical interests. The first and most immediate is to understand the acquisition process itself. Conditionals are conceptually, and in many languages morphosyntactically, complex. What aspects of cognitive and grammatical development are implicated in their acquisition? Does learning take place in the context of particular interactions with other speakers? Where do conditionals fit in with the acquisition of other complex sentences? What are the semantic, syntactic and pragmatic properties of the first conditionals?

Underlying this first interest is a second, more strictly linguistic one. Research of recent years has found increasing evidence that natural languages are constrained in certain ways. The source of these constraints is not yet clearly understood, but it is widely assumed that some of them derive ultimately from properties of children's capacity for language acquisition.

INTRODUCTION

1. (1) Kate 3;3 (pouring water on cement):

2. When you put water on it, it sparkles/

3. Adult: What?

4. Kate: If you put water on, it sparkles, see?

In a volume such as this, a reader might well ask, ‘Why is a chapter about toddlers and preschoolers included? Of what value can it be to scholars dealing with this complex and weighty topic?’ We hope to show that the process of child language acquisition presents a fertile resource for researchers interested in the basic nature of conditional sentences and their interaction with related language structures. As (I) demonstrates, children at an early age display knowledge of some of the interesting relationships of their language, such as the interchangeability of when and if in some contexts. The emerging cognitive and syntactic systems of preschoolers provide a different perspective on conditionals and allow us to see the basic building blocks of the adult system.

INTRODUCTION

Some arguments are logically valid but pragmatically incorrect. Others are pragmatically correct but logically invalid. Grice's Logic and conversation (1975) taught us to draw these distinctions, but unfortunately most of us draw them differently. What one calls a logically valid argument form with a few pragmatically incorrect instances is for another a logically invalid argument form with many pragmatically correct instances. For example, if you believe that indicative conditionals behave like material or strict implications, you will be ready to point out that the intuitively absurd argument.

1. (1) If Jones wins the election, Smith will retire to private life

2. If Smith dies before the election, Jones will win it

3. ∴ If Smith dies before the election, he will retire to private life

is just a pragmatically incorrect instance of the logically valid Hypothetical Syllogism:

1. (2) If B then C

2. If A then B

3. ∴ If A then C

OVERVIEW

The last chapter ended by listing 11 of the principal demands I claim functionalism makes of a theory of the mind. In this chapter I set about developing an account of belief that has grown up with functionalism, trying, as I go, to adjust it to those demands.

We can begin with a simple observation: belief comes by degrees. I believe more strongly that it will rain tomorrow than that President Carter will win the United States Presidential election in 1988. Our language is full of idioms that allow us to express the relative strengths of our beliefs. We say we think one thing more likely than another; or that we have more confidence that one thing will happen than another; or that we think something extremely likely or are highly confident that it will happen. And such differences in the strengths of our beliefs show up not only in what we say, but in what we do. People who believe only fairly strongly that aeroplanes are safe need a stronger motive for flying than they would if they had more confidence in them.

OVERVIEW

That is my claim. Once accepted, there is, as we have seen, good reason for thinking that the indicative conditional is not a material conditional. Indeed, since it is easy to see that no truth-function of two components has a probability equivalent to the conditional probability, there is good reason for thinking that the indicative conditional is not a truth-function at all. But the fact that it is not a truth-function is not, in itself, evidence that it has no truth conditions. The sentence-forming operator on sentences ‘I believe that …’ is not a truth-function: but there is no reason to doubt that it determines truth conditions, or to doubt that grasp of how those truth conditions are determined is part of a knowledge of what it means.

Since, in general,

it is tempting, given the strong evidence that Adams' assertibility rule is correct, to look for a set of truth conditions for ‘If A, then C’, such that the sentence is assertible iff it is believed to be probably true.

I have tried in this book to use the general account of the nature of belief, given in Part I, to ground an account of meaning, given in Part II; and to use the theory of meaning, and, in particular, the notion of assertibility that it entails, in Part III, to explain the semantics of some indicative conditionals. I have produced only a small part of the full theory of beliefs: nothing about second-order probabilities; little about the reference relation between singular representational elements and the world; mere intimations – promissory notes – about quantification. The account of meaning, too, needs developing: to deal with speech-acts other than assertion, for example, and, more generally, with conversational and conventional implicature. And there is much more to be said about conditionals: subjunctives, in particular. All these are matters well worth pursuing. And, I conjecture, the study of each of the unresolved problems in the theory of meaning will benefit, as I think the study of conditionals has, from an approach that begins with the nature of beliefs and moves out to the semantics of sentences that express them.

The main argument for the theory I have developed is that we need its features to do the job we do every day of explaining each other's action … including, of course, out utterances. I claim, in fact, that, except for a technical precisification and, no doubt, some errors of mine, it is your theory also.

SUBJUNCTIVE AND INDICATIVE CONDITIONALS

‘If's and ‘iff's abound in philosophical analysis, but the philosophical analysis of ‘if's and ‘iff's is a highly controversial matter. Still, I believe that, at least for some conditionals, the problem of analysis is now approaching a solution. In the next four chapters I say what I think that solution is for one class of conditional, the class that is usually called indicative.

That there are more conditionals in English than these is, I think, plain enough. In 3.3 I gave my main reason for thinking this: I suggested that there were two distinct jobs to be done by conditional beliefs, the states most naturally expressed in English by sentences beginning with ‘if’. One job, that of the other major class, the subjunctive conditionals, was in deciding what to do, which shows up in the theory as computing expected utility. The other job, that of the indicatives, was in changing our minds, which shows up in the theory as conditionalisation. That these two jobs must be done by different beliefs follows from the existence of the class of Newcomb problems that led to the reformulation of decision theory in its causal version.

OVERVIEW

In Part I of this book I was concerned almost exclusively with minds. The last chapter was concerned very largely with language and relatively little with mind. In this chapter I want to give my own view of assertion, a view which connects language and mind. I want to show how, given our theory of the nature of beliefs, it is relatively easy to give an account of declarative meaning. This task will seem to some of our contemporary anti-psychologists – such components of ‘psychologism’ in semantics as Dummett (1973: passim) – absurd. It is not possible, they have held, to carry out this project because, in Davidson's (1975) words, ‘without speech … we cannot make … fine distinctions between thoughts’. My project, from this point of view, is circular. For I assume that we can ascribe beliefs and then go on to say what terms in the agent's language mean; whereas they hold that I need meanings to ascribe the beliefs.

Let me mention three reasons why I have ignored their position here. First, and importantly, I can simply admit some of what Davidson is claiming.

OVERVIEW

I turn now from an account of certain general features of the mind to an account of declarative meaning. In doing so I begin by examining the main stream of thought about meaning; what I believe to be the central theory in philosophical semantics. I mean what Michael Dummett, amongst others, has called ‘realism’.

Realism, in semantics, as I characterise it, is the claim that the meaning of a central class of declarative sentences can be given by stating their truth conditions. This way of putting the matter leaves much open. Which class of sentences is central? Do we give meanings to token sentences or to types? And, above all, what, for sentences, are truth conditions? Without answers to these questions realism is just a scheme for a theory of meaning; a scheme we might acquiesce in without being clear what it entailed. The service Donald Davidson has done us is to offer a way of filling out the scheme; a way that allows us to ask and to answer these and many other questions. But it is just one way. And if we end by rejecting it, that is not the end of realism. Some other way of filling out the scheme might do the job.

OVERVIEW

In this chapter I wish to draw together conclusions about the way the conventional notion of a truth condition fits into the framework of a functionalist theory as I have set it out. I want to consider whether, as Loar puts it, the truth conditions of beliefs are determined by a functionalist theory; and to draw some consequences of the fact that the resources available to a functionalist theory are more limited than those which classical accounts of truth conditions take for granted.

A functionalist theory of the mind looks to individuate mental events by their functional roles: by their causal antecedents and consequences within and without the mind. Mental states are characterised by the way in which they determine the functional roles of mental events. When we approach an account of the sentential attitudes – those psychological states, like belief, desire, hope, and so on, the English verbs for which take sentential complements in ‘that’-clauses – we are left with a problem. For the sentential attitudes seem to have the general form of a relation between an agent and what philosophers have called a proposition.