Those of us who passed many (well- or ill-spent?) childhood hours reading tales of rockets and robots, androids and telepaths, galactic civilizations and time machines, know all too well that robots – hypothetical machines that simulate human behavior, often with an at least roughly human appearance – can be friendly or fearsome, man's best friend or worst enemy. When friendly, robots can be inspiring or pathetic – they can overawe us with their superhuman powers (and with their greater than human virtue as well, at least in the writings of some authors), or they can amuse us with their stupidities and naiveté. Robots have been ‘known’ to fall in love, go mad (power- or otherwise), annoy with oversolicitousness. At least in the literature of science fiction, then, it is possible for a robot to be ‘conscious’; that means (since ‘consciousness’, like ‘material object’ and ‘universal’, is a philosopher's stand-in for more substantial words) to have feelings, thoughts, attitudes, and character traits. But is it really possible? If it is possible, what are the necessary and sufficient conditions? And why should we philosophers worry about this anyway? Aren't the mind–body problem, the problem of other minds, the problem of logical behaviorism, the problem: what did Wittgenstein really mean in the private-language argument? (and why should one care?), more than enough to keep the most industrious philosopher of mind busy without dragging in or inventing the Problem of the Minds of Machines?

In this paper I want to discuss the nature of various ‘mentalistic’ notions in terms of a machine analog. In chapter 18, I tried to show that the conceptual issues surrounding the traditional mind–body problem have nothing to do with the supposedly special character of human subjective experience, but arise for any computing system of a certain kind of richness and complexity, in particular for any computing system able to construct theories concerning its own nature. In that paper I was primarily interested in the issues having to do with mind–body identity. In the present paper the focus will be rather in trying to shed light on the character of such notions as preferring, believing, feeling. I hope to show by considering the use of these words in connection with a machine analog that the traditional alternatives – materialism, dualism, logical behaviorism – are incorrect, even in the case of these machines. My objectives are not merely destructive ones; I hope by indicating what the character of these words is in the case of the machine analog to suggest to some extent what their character is in application to human beings.

One question which I shall not discuss, except for these remarks at the outset, is the question to what extent the application of such terms as ‘preference’ to Turing Machines represents a change or extension of meaning.

The innateness hypothesis

The ‘innateness hypothesis’ (henceforth, the ‘IH’) is a daring – or apparently daring; it may be meaningless, in which case it is not daring – hypothesis proposed by Noam Chomsky. I owe a debt of gratitude to Chomsky for having repeatedly exposed me to the IH; I have relied heavily in what follows on oral communications from him; and I beg his pardon in advance if I mis-state the IH in any detail, or misrepresent any of the arguments for it. In addition to relying upon oral communications from Chomsky, I have also relied upon Chomsky's paper ‘Explanatory models in linguistics’ (Chomsky, 1962), in which the IH plays a considerable role.

To begin, then, the IH is the hypothesis that the human brain is ‘programmed’ at birth in some quite specific and structured aspects of human natural language. The details of this programming are spelled out in some detail in ‘Explanatory models in linguistics’. We should assume that the speaker has ‘built in’ a function which assigns weights to the grammars G1, G2, G3, … in a certain class Σ of transformational grammars. Σ is not the class of all possible transformational grammars; rather all the members of Σ have some quite strong similarities. These similarities appear as ‘linguistic universals’ – i.e. as characteristics of all human natural languages. If intelligent nonterrestrial life – say, Martians – exists, and if the ‘Martians’ speak a language whose grammar does not belong to the subclass Σ of the class of all transformational grammars, then, I have heard Chomsky maintain, humans (except possibly for a few geniuses or linguistic experts) would be unable to learn Martian; a human child brought up by Martians would fail to acquire language; and Martians would, conversely, experience similar difficulties with human tongues.

Let me say that I admire the vigor of Professor Massey's reply. Let me also say that all the references to Grünbaum in my paper were references to Grünbaum's pre-1970 publications, in particular to his reply to me. Pre-1970 Grünbaum is no figment of my imagination. Pre-1970 Grünbaum indicated in many places that he was using the expression ‘convention’ in a perfectly standard philosophical sense; for example he equated the statement that the choice of a metric is ‘conventional’ with the statement that the choice is just a matter of descriptive simplicity. ‘Descriptive simplicity’ is a term introduced by Reichenbach for those choices in science which do not affect the truth value of what is said, i.e. for those choices which are conventional in a perfectly standard philosophical sense of conventional. Massey sometimes seems to suggest that one cannot infer from the fact that something is ‘convention laden’ in Grünbaum's sense that it is conventional in the sense of any other philosopher. Also, as Massey himself recognizes, pre-1970 Grünbaum did hold that the term ‘relation of spatial equality’ has an ‘intension’ which fails to determine the extension of ‘relation of equality’ in the case of continuous physical spaces and space–time.

Now the question is: to what extent is the position of post-1970 Grünbaum a substantial improvement on the pre-1970 position Professor Grünbaum's post-1970 position is certainly more complicated than his pre-1970 position. One notion used by Professor Grünbaum, both early and late, and which I did not discuss in my paper, is the notion of a metric being intrinsic.

Language is the first broad area of human cognitive capacity for which we are beginning to obtain a description which is not exaggeratedly oversimplified. Thanks to the work of contemporary transformational linguists,† a very subtle description of at least some human languages is in the process of being constructed. Some features of these languages appear to be universal. Where such features turn out to be ‘species-specific’ – ‘not explicable on some general grounds of functional utility or simplicity that would apply to arbitrary systems that serve the functions of language’ – they may shed some light on the structure of mind. While it is extremely difficult to say to what extent the structure so illuminated will turn out to be a universal structure of language, as opposed to a universal structure of innate general learning strategies,‡ the very fact that this discussion can take place is testimony to the richness and generality of the descriptive material that linguists are beginning to provide, and also testimony to the depth of the analysis, insofar as the features that appear to be candidates for ‘species-specific’ features of language are in no sense surface or phenomenological features of language, but lie at the level of deep structure.

The most serious drawback to all of this analysis, as far as a philosopher is concerned, is that it does not concern the meaning of words.

General significance of the topic

In this paper I try to contrast realist theories of meaning with what may be called ‘idealist’ theories of meaning. But a word of explanation is clearly in order.

There is no Marxist ‘theory of meaning’ but there are a series of remarks on the correspondence between concepts and things, on concepts, and on the impossibility of a priori knowledge in the writings of Engels (cf. Engels, 1959) which clearly bear on problems of meaning and reference. In particular, there is a passage in which Engels makes the point that a concept may contain elements which are not correct. A contemporary scientific characterization of fish would include, Engels says, such properties as life under water and breathing through gills; yet lungfish and other anomalous species which lack these properties are classified as fish for scientific purposes. And Engels argues, I think correctly, that to stick to the letter of the ‘definition’ in applying the concept fish would be bad science. In short, Engels contends that:

1. Our scientific conception (I would say ‘stereotype’) of a fish includes the property ‘breathing through gills’, but

2. ‘All fish breath through gills’ is not true! (and, a fortiori, not analytic).

I do not wish to ascribe to Engels an anachronistic sophistication about contemporary logical issues, but without doing this it is fair to say on the basis of this argument that Engels rejects the model according to which such a concept as fish provides anything like analytically necessary and sufficient conditions for membership in a natural kind.

In this paper I want to explore the question: whether any sense can be made of the traditional view that a true assertion is one that corresponds to reality. Two opinions seem to be widespread: (a) that some sense can be made of the view, and that some sense is made of it (as much as can be hoped for) by Tarski's ‘Semantical Conception of Truth’; (b) that the view collapses as soon as one asks searching questions about the nature of the alleged ‘correspondence’. I shall try to show that both of these opinions are incorrect.

Tarski's conception of truth

According to Tarski, an adequate definition of ‘true as a sentence of English’ should have the feature that from it follows:

1. ‘Snow is white’ is true as a sentence of English if and only if snow is white.

2. ‘Grass is green’ is true as a sentence of English if and only if grass is green.

Tarski claims that this is a formalization of the ‘correspondence theory of truth’. But is it?

Assume for the moment that Tarski's Criterion of Adequacy (roughly described above) is correct. Even so, it seems to refer to purely interlinguistic aspects of the usage of ‘true’. This can be seen, for example, by observing that if the meta-language, say, ML, is only partially interpreted, we might still be able to certify a definition of ‘true as a sentence of L’ to be ‘adequate’ by checking that Tarski's Criterion of Adequacy was conformed to (all biconditionals of the form ‘S is true as a sentence of L ≡ S’ are theorems of ML where S is an arbitrary sentence of L and S is the name – or Gödel number – of S), even though the extra-logical constants of L are totally uninterpreted.

Once upon a time there was a tough-minded philosopher who said, ‘What is all this talk about “minds”, “ideas”, and “sensations”? Really – and I mean really in the real world – there is nothing to these so-called “mental” events and entities but certain processes in our all-too-material heads.’

And once upon a time there was a philosopher who retorted, ‘What a masterpiece of confusion! Even if, say, pain were perfectly correlated with any particular event in my brain (which I doubt) that event would obviously have certain properties – say, a certain numerical intensity measured in volts – which it would be senseless to ascribe to the feeling of pain. Thus, it is two things that are correlated, not one – and to call two things one thing is worse than being mistaken; it is utter contradiction.’

For a long time dualism and materialism appeared to exhaust the alternatives. Compromises were attempted (‘double aspect’ theories), but they never won many converts and practically no one found them intelligible. Then, in the mid-1930s, a seeming third possibility was discovered. This third possibility has been called logical behaviorism. To state the nature of this third possibility briefly, it is necessary to recall the treatment of the natural numbers (i.e. zero, one, two, three…) in modern logic. Numbers are identified with sets, in various ways, depending on which authority one follows.

About a century ago, Charles Sanders Peirce asserted that the meaning of an ‘intellectual conception’ is identical with the ‘sum’ of its ‘practical consequences’ (cf. Peirce, 1958). And he thought this idea sufficiently important that he made it the primary maxim of the philosophy he called Pragmatism. This is nothing but an early statement of the Verifiability Theory of Meaning. And Pragmatism was the first philosophy dedicated to the proposition that theory of meaning can solve or dissolve the traditional problems of philosophy.

Today the Verifiability Theory of Meaning has been pretty well abandoned, not, alas!, because the fundamental intuition behind it has been universally conceded to be erroneous, but simply because there are formidable technical objections to the doctrine. The fundamental intuition behind the doctrine has two closely related components. One component is the Open Question Argument: what more does the expression mean, if it means more than that we will have certain experiences? The other component is the argument, often found in nineteenth century literature (and still occasionally found today) that to believe in things that are not concepts (or at least mental entities) is to believe in things that are inconceivable.†

The reply to the Open Question argument is that the Verificationist philosopher is here speaking as if he had succeeded in the enterprise of translating our commonplace talk of things into a sensationalistic vernacular.

The question which troubles laymen, and which has long troubled philosophers, even if it is somewhat disguised by today's analytic style of writing philosophy, is this: are we made of matter or soul-stuff? To put it as bluntly as possible, are we just material beings, or are we ‘something more’? In this paper, I will argue as strongly as possible that this whole question rests on false assumptions. My purpose is not to dismiss the question, however, so much as to speak to the real concern which is behind the question. The real concern is, I believe, with the autonomy of our mental life.

People are worried that we may be debunked, that our behavior may be exposed as really explained by something mechanical. Not, to be sure, mechanical in the old sense of cogs and pulleys, but in the newer sense of electricity and magnetism and quantum chemistry and so forth. In this paper, part of what I want to do is to argue that this can't happen. Mentality is a real and autonomous feature of our world.

But even more important, at least in my feeling, is the fact that this whole question has nothing to do with our substance. Strange as it may seem to common sense and to sophisticated intuition alike, the question of the autonomy of our mental life does not hinge on and has nothing to do with that all too popular, all too old question about matter or soul-stuff.

‘Empirical realism’ is the position that the existence of the external world is supported by experience in much the way that any scientific theory is supported by observational data. The empirical realist reply to skepticism has recently been extended by Paul Ziff from skepticism about material objects to skepticism about other minds (Ziff, 1965). I do not suggest that Ziff was unaware of the need for the various qualifications that have to be made in the realist position if it is to be tenable. However, I am not happy with the way in which Ziff states the arguments. Ziff's statements are very brief, and it may be that the features I shall object to are ones that he would have eliminated in a longer and less aphoristic presentation. However, here they are.

There are two parts to Ziff's argument: what he calls the via negativa, and the citation of positive support. I take them up in turn.

The via negativa amounts to this: if I accept the hypothesis that I alone have a mind, then I must, according to Ziff, suppose that I differ from other human beings in some other respect, presumably a physiological respect. I can't differ from other human beings in just this one way, that I have a mind and they don't.

I shall discuss conventionalism in Quine's writings on the topic of radical translation, and in the writings of Reichenbach and Grünbaum on the nature of geometry.

Let me say at the outset that Quine and Reichenbach are the two philosophers who have had the greatest influence on my own philosophical work. If Quine's ideas have not had the full influence they deserve, it may be in part because of the intensely paradoxical nature of the doctrines put forward – or seemingly put forward – in Word and Object. The doctrines of Word and Object – in particular the impossibility of radical translation – look wrong to many philosophers. Since these doctrines are thought by Quine himself to follow from the doctrines put forward in ‘Two dogmas of empiricism’, they cast doubt on ‘Two dogmas’ itself. My contention here will be that the impossibility of radical translation does not follow from the critique of the analytic-synthetic distinction. I believe that Quine is right in his critique of the analytic-synthetic distinction, but wrong in his argument for the impossibility of radical translation. By showing that one does not have to accept Quine's arguments for the impossibility of radical translation, I hope, therefore, to clarify just what it is in Quine's work that is true and important.

Similarly, I think that Reichenbach understood the importance of non-Euclidean geometry for epistemology as it has not been understood by philosophers in general to the present day.

For over a hundred years, one of the dominant tendencies in the philosophy of science has been verificationism: that is, the doctrine that to know the meaning of a scientific proposition (or of any proposition, according to most verificationists) is to know what would be evidence for that proposition. Historically, verificationism has been closely connected with positivism: that is, at least originally, the view that all that science really does is to describe regularities in human experience. Taken together, these views seem close to idealism. However, many twentieth-century verificationists have wanted to replace the reference to experience in the older formulations of these doctrines with a reference to ‘observable things’ and ‘observable properties’. According to this more recent view, scientific statements about the color of flowers or the eating habits of bears are to be taken at face value as referring to flowers and bears; but scientific statements about such ‘unobservables’ as electrons are not to be taken as referring to electrons, but rather as referring to meter readings and the observable results of cloud chamber experiments. It is not surprising that philosophers who took this tack found themselves in a certain degree of sympathy with psychological behaviorism. Just as they wanted to ‘reduce’ statements about such unobservables as electrons to statements about ‘public observables’ such as meter readings, so they wanted to reduce statements about phenomena which, whatever their private status, were publicly unobservable, such as a person's sensations or emotions, to statements about such public observables as bodily behaviors.

At this point, they found themselves in a certain bind.

In the last decade enormous progress seems to have been made in the syntactic theory of natural languages, largely as a result of the work of linguists influenced by Noam Chomsky and Zellig Harris. Comparable progress seems not to have been made in the semantic theory of natural languages, and perhaps it is time to ask why this should be the case. Why is the theory of meaning so hard?

The meaning of common nouns

To get some idea of the difficulties, let us look at some of the problems that come up in connection with general names. General names are of many kinds. Some, like bachelor, admit of an explicit definition straight off (‘man who has never been married’); but the overwhelming majority do not. Some are derived by transformations from verbal forms, e.g. hunter = one who hunts. An important class, philosophically as well as linguistically, is the class of general names associated with natural kinds – that is, with classes of things that we regard as of explanatory importance; classes whose normal distinguishing characteristics are ‘held together’ or even explained by deep-lying mechanisms. Gold, lemon, tiger, acid, are examples of such nouns. I want to begin this paper by suggesting that (1) traditional theories of meaning radically falsify the properties of such words; (2) logicians like Carnap do little more than formalize these traditional theories, inadequacies and all; (3) such semantic theories as that produced by Jerrold Katz and his co-workers likewise share all the defects of the traditional theory.

The various issues and puzzles that make up the traditional mind–body problem are wholly linguistic and logical in character: whatever few empirical ‘facts’ there may be in this area support one view as much as another. I do not hope to establish this contention in this paper, but I hope to do something toward rendering it more plausible. Specifically, I shall try to show that all of the issues arise in connection with any computing system capable of answering questions about its own structure, and have thus nothing to do with the unique nature (if it is unique) of human subjective experience.

To illustrate the sort of thing that is meant: one kind of puzzle that is sometimes discussed in connection with the ‘mind–body problem’ is the puzzle of privacy. The question ‘How do I know I have a pain?’ is a deviant† (‘logically odd’) question. The question ‘How do I know Smith has a pain?’ is not all at deviant. The difference can also be mirrored in impersonal questions: ‘How does anyone ever know he himself has a pain?’ is deviant; ‘How does anyone ever know that someone else is in pain?’ is non-deviant. I shall show that the difference in status between the last two questions is mirrored in the case of machines: if T is a Turing machine (see below), the question ‘How does T ascertain that it is in state A?’

Although this is primarily a textbook written for a specific purpose, I said in the Preface that it ought to be useful to ordinary people in the ordinary course of their lives—that is, not just to those who face a general paper, or have to do a course in philosophy. This is not just a pious hope: but it may seem rather a forlorn one, because the gap between philosophy and ordinary life is horrifyingly large. Consequently it may be useful to say something about the way in which the techniques illustrated in this book come into philosophy, and the way in which philosophy may come into ordinary life. Of course this is an immense subject, and I cannot do it justice: but I hope at least to show that the ordinary person may justifiably be more optimistic about the relevance of philosophy than perhaps some philosophers have led him to expect.

Everything turns on the business of philosophy. One view, perhaps still the most popular, is that philosophy is directly and immediately concerned with a way of life and with the truth about reality. It has to do with what people are, what they do, and what they feel: with their behaviour, their emotions, their beliefs and moral judgements. By this account a man's philosophy is a sort of blend between his motives, his behaviour, and his values.

This is a comparatively short chapter: I have not given a very great number of passages for criticism, nor very many questions of concept to be answered. For this there are several reasons. First, inasmuch as the book is used in sixth forms and for the benefit of any students who face examinations, their teachers will be primarily concerned with the particular kind of general paper relevant to the needs of their particular students: and of course, apart from the fact that they all include questions of concept, these papers vary very widely. Teachers will naturally want to make use of past papers printed by universities and colleges, and direct the attention of their pupils to the sort of passages and questions which these include. Secondly, those who read this book without any examination paper in view are likely to be interested in one field of thought rather than others: thus some will be more concerned with religion, others with politics, others again with morals, and so on. These specific interests are important, because they give an extra incentive for the analysis of concepts: someone seriously concerned with religion is likely to do more justice to the concepts involved in a passage dealing with religion than to those involved in passages dealing with other matters.

WHAT IS CONCEPTUAL ANALYSIS?

This book is designed to give the reader mastery over certain skills and techniques. Half the battle is won if you can get a clear idea of exactly what these skills and techniques are, and what purpose they serve: so we shall have to begin by spending a lot of time over this point. Techniques like being able to solve quadratic equations, doing Latin prose, or translating German into English are difficult to master: but at least we have the advantage of knowing just what it is that we are supposed to be doing, even if we do not always do it very well. These techniques and many others have for a long time been placed under different headings: they are what schools call ‘subjects’—mathematics, Latin, German, and so on. Often we can look up the right answers to questions in these subjects, by referring to a dictionary, or a grammar, or an authoritative textbook. But none of this applies to the techniques outlined in this book. That is partly because they are new techniques: we have only become fully conscious of them in the last twenty or thirty years. But it is chiefly because of the nature of the techniques themselves, and the general purpose which they serve.

What are these techniques like? They are not like ‘subjects’ such as Latin or mathematics, which have clear-cut and well-defined rules, and in which answers are indisputably right or wrong.