In the last lecture, John Lucas suggested that considerable light was thrown on the nature of mind by considering a number of results in formal logic. You will have gathered that it is a matter of dispute among logicians and philosophers whether these results have the implications he thinks they have for philosophy of mind. But let us suppose that Lucas is right in this controversy. I think that the non-logician may feel certain rather human misgivings about his approach. It may be that John can show certain things to be true which no machine could prove. It may be that this shows that John Lucas is no machine—but what of the rest of us? After all, comparatively few human beings can follow Gödel's theorems or Church's theorem. Does this mean that we lack the indefinable property which sets minds above machines? I know that John would reply ‘Of course not, the Gödelian argument is merely a way of breaking down the extreme claim “all minds are machines”: once it has been accepted that some minds are not machines then there is very little reason for thinking that any are.’
Still, there remains something to be said for trying to find a way into the study of mind which travels via a less esoteric capacity than the ability to outwit computers by producing their Gödelian formulas. And that's what I hope to explore this evening under the title ‘To Mind via Syntax’. For there is reason to believe that considerable new light can be thrown on the human mind by a study of the skill which everyone in this room shares—indeed which every child in these islands over the age of six shares—the ability to frame and understand English sentences. So I want you to shift the focus of your attention from logic to linguistics, from Gödel to Chomsky.
The early work which brought Chomsky into public notice was predominantly within the specialised area of linguistics. But it has been Chomsky, I think, more than any other linguist, who has drawn the connections between the specialised area of linguistics and psychology in general. Especially since his famous hostile review of Skinner's Verbal Behaviour, Chomsky has stressed the implications of his linguistic work for psychology and philosophy of mind. I am no linguist, and therefore am not competent to evaluate his work in linguistics. Why then do I talk about him at all? I am emboldened by Chomsky's own hostility to the compartmentalisation of disciplines. I think he believes that fundamentally philosophy, psychology, and linguistics are a single discipline, and perhaps he is right.
If we study the phenomena of language-use, Chomsky believes, we are led to postulate for explanatory purposes abstract structures at four levels of abstraction.
First of all, in order to describe the grammatical structure of the sentences produced by native English speakers, we must postulate certain abstract mental patterns which underlie these sentences. These abstract mental patterns are called ‘deep structures’, by contrast with the surface structures of the audible and visible sentences.
Secondly, in order to explain how a language-user can produce from these deep structures the actual sentences of his language, we must postulate that he has internalised rules for certain grammatical transformations that map deep structures onto surface structures. And in order to explain the production of the deep structures themselves we must postulate that he has internalised rules for their generation. We must postulate that he has internalised a particular generative grammar. So that's the second level of abstract postulate—postulation of a particular grammar internalised within the speaker.
Thirdly, in order to explain how a child can learn language, that is to say, how a child can pick up the particular grammar of its language from the fragmentary linguistic data which are presented to it, we must postulate some innate organising principles of universal grammar which enable it to select a particular grammar as a hypothesis to explain the input data. So that is the third level of abstraction—universal grammar as an innate capacity.
Fourthly, if we take the learning of language as a typical example of the acquisition of knowledge by human beings, we may go on—and Chomsky does go on—to draw certain conclusions about human cognitive capacities in general, and thus to offer a general picture of the human mind as mirrored in language. Chomsky in particular postulates a particular hypothesis-forming ability—or rather a general hypothesis-forming ability—a faculty of abduction, he calls it, using an expression invented by the American philosopher, Charles Sanders Peirce. Now I daresay that to those of you who are not familiar with Chomsky's work these four levels of abstraction are pretty unintelligible, and what I want to do in the first part of my paper is to illustrate from Chomsky's writings what is meant by them.
First of all, deep structures. These are most commonly postulated in order to explain our understanding of ambiguous expressions and sentences. Take this sentence: ‘The police were ordered to stop drinking after midnight’. If you reflect on that you realise that it is capable of four different interpretations, not because of the ambiguity of any single word in the sentence but because of different possibilities of analysing it syntactically. Similarly, you can analyse in two different ways a sentence such as, ‘She is too old to marry’—is she the subject or the object of the marriage? In order to disambiguate particular occurrences of such sentences, we depend upon context and non-linguistic information of various sorts. But simply to know that they are ambiguous, as we all do, is part of knowing the language. And any adequate grammar has to assign a plurality of syntactic descriptions to such sentences.
A simple example will illustrate Chomsky's method of treatment. Consider the sentence: ‘I disapprove of John's drinking’. This can refer either to the fact of John's drinking or to its character. We can resolve the ambiguity in different ways in the two following sentences: ‘I disapprove of John's drinking the beer’, or ‘I disapprove of John's excessive drinking’. It is clear in a case like this, Chomsky says, that grammatical processes are involved. And he suggests that our internalised grammar assigns two different abstract structures to the sentence ‘I disapprove of John's drinking’; one of them is related to the structure which underlies ‘I disapprove of John's drinking the beer’, and the other is related to the structure which underlies ‘I disapprove of John's excessive drinking’. But, he says, it is at the level of deep structure that the distinction is represented. It is obliterated by the transformations that map the deep structures on to the surface form (Language and Mind, 27).
Again, consider the two following sentences: ‘John is eager to please’ and ‘John is easy to please’. These two sentences have similar surface structures. But corresponding to the first one we have the nominal phrase ‘John's eagerness to please’, but we can't form, corresponding to the second one, the nominal phrase ‘John's easiness to please’. Chomsky postulates that the first one has a deep structure which is close to the surface structure, whereas in the second example the deep structure is far removed. Its deep structure is something like ‘For one to please John is easy’. Now Chomsky goes on to postulate, to explain this, that a nominal phrase can be formed which corresponds to a base structure, but not to a surface structure; and he offers a large number of examples to illustrate this. This observation lends support to the assumption that abstract deep structures play a role in the mental representation of sentences. I quote from his paper in Language and Philosophy, ‘We find that when we study English grammar on the basis of this and related assumptions we are able to characterise quite readily the class of sentences to which there correspond nominal phrases of the sort under discussion. There is no natural way to characterise this class in terms of surface structure, since, as we have seen, sentences that are very similar in surface structure behave quite differently with respect to the formal processes involved in the construction of nominal expressions.’ (Language & Philosophy, 58–9).
Now we have already, as you may have noticed, moved from the first level to the second, with this postulation that nominalisations are permitted corresponding to deep structure but not to surface structure. This postulate is of course the postulation of a rule and it takes us to the second degree of abstraction. According to Chomsky, a native speaker of a language (each of us) has internalised a system of rules that relate sound to meaning in a particular way. The linguist, Chomsky says, when he constructs a grammar of a language is in effect proposing a hypothesis—a hypothesis concerning the internalised system of the language-user. And the linguist's hypothesis, if it is presented with sufficient explicitness and precision will have certain empirical consequences with regard to the form of utterances and their interpretation by the native speaker. (Language & Mind, 23). The rule which I mentioned about nominalisations is a syntactic rule, but of course if a grammar is to relate sound and meaning it must include a lot more than merely syntax. It will have to have a semantic component to attach meaning to deep structure, and it'll have a phonological component to assign a phonetic interpretation to a surface structure—syntax including both the deep and surface structure (Topics in the Theory of Generative Grammar, 16). In this evening's lecture I am going to concentrate on syntax, reserving the topic of semantics for tomorrow night when Longuet-Higgins will be dealing with it. So far as I understand, phonology is a topic that we are both going to leave out in the cold.
Now, at the third level one can search for explanatory theories of a deeper sort. A grammar is one level of explanatory theory of linguistic output. The native speaker has acquired a grammar on the basis of very restricted and degenerate evidence—the mouthings and cooings of parents, and so on. The grammar has empirical consequences that extend far beyond the evidence. That is to say, the speaker can understand and invent grammatical sentences that he has never heard. At one level, the phenomena with which the grammar deals are explained by the rules of the grammar itself. At a deeper level these same phenomena are explained by the principles that determine the selection of the grammar on the basis of this evidence. The principles that determine the form of a grammar, and select a grammar of the appropriate form on the basis of certain data constitute a subject that might, following a traditional usage, be termed ‘universal grammar’ (Language & Mind, 24).
So Chomsky postulates in the new-born or almost new-born child knowledge of universal grammar. Of course, the grammar attributed to the child is universal, not particular. We can't attribute knowledge of English to the child as an innate property, because we know that the child can learn Japanese as well as English (Language & Philosophy, 62). But we have the situation that a child who initially doesn't have knowledge of a language constructs knowledge of a grammar on the basis of certain data. The input is the data, the output is the child's knowledge of the language, and any scientist trying to study a black-box which had certain data as input and a grammar as output would conclude that any property of the output which went beyond the organisation of the input must in some way or other be attributed to the character of the device. And anything that was to be attributed to. The device on the strength of the study of this input/output relation would, according to Chomsky, be a universal of language (The Listener, 30/5/68).
Now what would these linguistics universals look like? What sort of thing would they be? Consider a simple example which Chomsky has used several times, namely the rule for forming questions from the corresponding statements. Take the English statement ‘The dog in the corner is hungry’. From this we can form the question ‘Is the dog in the corner hungry?’ Now what is the rule according to which this transformation is carried out? It is by no means easy to state it accurately, but as an approximation we might suggest that the occurrence of ‘is’ which follows the subject-noun phrase in a sentence of this kind is moved to the beginning of the sentence. Now this rule has a certain feature of great importance. It is what is called ‘structure-dependent’. That is, in order to apply the rule you have to consider not merely the sequence of elements that constitute the sentence but also their structure. You have got to be able to recognise a noun phrase for what it is. It would be very easy—mathematically it would be much simpler—to form structure-independent operations. For instance, one very simple structure-independent operation would consist in reversing the sentence and forming the question ‘Hungry is corner the in dog the’. This would be mathematically a much simpler thing to operate than the rule we have. A computer could do it much more easily than it could do what we do. Nonetheless, according to Chomsky, all formal operations in any known grammar are structure dependent, and this, as he says, is hard to explain on any grounds simply of utility or biological efficiency. One would expect the mathematically simpler operations to be chosen. So that is the kind of thing that he postulates as a universal of universal grammar, the rule that the rules of particular grammars must be structure-dependent. You can see that what is postulated is something highly abstract.
We can proceed, and as time goes on Chomsky seems to proceed further and further in this direction, to a fourth degree of abstraction. On Chomsky's account, as we have seen, the child's acquisition of knowledge is a kind of theory construction, which proceeds under strict limitations on what is to count as an admissible theory. And this is most readily understood if human theory construction in general is subject to limitations, if man's mind has a natural adaptation to imagining correct theories of certain kinds. The ability to learn language is, according to Chomsky, only one faculty of mind, and there are others. Logical and mathematical powers are quite distinct from the language faculty. The language faculty is the one that constructs the grammar out of the input, but we have many other faculties. Though human children learn languages which have only structure-dependent transformations, it may be that other minds, say Martians if there are any, have languages which have structure-independent transformations, and there is no reason, if we ever met them, why we shouldn't be able to puzzle out what they meant, as a sort of mathematical puzzle, using our mathematical abilities. But if Chomsky is right this would not be as easy as to learn a human language, constructed in the manner which he postulates is specific to the human species. (The Listener, 30/5/68).
Now, in postulating universal grammar as an innate structure, Chomsky is explicitly allying himself with a rationalist tradition going back to Descartes, and taking sides against the empiricist tradition, dominant, or certainly dominant until recently, in linguistics and philosophy. Chomsky has also praised Descartes for explicitly recognising something which more sophisticated linguists had ignored—namely, the creative aspect of language use: the distinctively human ability to express new thoughts and to understand entirely new expressions of thought. This creativity has three elements. First of all, language-use is innovative: much of what we say in the course of normal language is entirely new, and not mere repetition of anything we've heard before. One can understand without difficulty an astronomical number of sentences in one's native language, and indeed the use of language is potentially infinite in scope. The normal use of language, though, is not only innovative and potentially infinite in scope, it's also free from the control of detectable stimuli, either external or internal. But despite this, it is appropriate and coherent in a sense which is difficult to make precise, though clearly meaningful since we can distinguish coherent use of language from the ravings of a maniac, or the output of a computer with randomising element. (Language & Mind 123).
Chomsky's proposals, which I have summarised, are the subject of hot debate among linguists, psychologists and philosophers. Every step of the argument I have outlined has been challenged, right from the initial postulation of deep structures. Some linguists reject the need to go beyond surface structures, and to postulate the transformations from which transformational grammars take their name. As I've said, I'm not competent to assess the linguistic merits of these proposals. I can't judge whether these four abstract structures he postulates are either necessary or sufficient to explain the linguistic data adduced in evidence for them. As a philosopher, however, one doesn't have to wait for the results of the linguistic research in order to assess some of the consequences of Chomsky's proposals for the philosophy of mind. The philosopher can simply ask ‘Suppose that the linguistic facts were to turn out exactly as Chomsky supposes, what consequences would this have for the understanding of the nature of mind? Would it have the consequences Chomsky himself thinks it would have?’
I want to go some way towards answering this question, but before doing so I have to introduce one more theoretical distinction. Chomsky distinguishes between what the speaker of a language knows implicitly, which he calls his competence, and what he actually does with his knowledge, his performance. He says a grammar in the traditional view is an account of a competence. It describes, attempts to account for, the ability of the speaker to understand an arbitrary sentence of his language and to produce an appropriate sentence on a given occasion. Performance provides evidence for competence, and competence provides a partial explanation for performance. But to explain performance fully, one needs much more than an understanding of competence. One must take into account a number of contingent factors, things like the shortness of memory span, variations of attention, changes of mind in mid-sentence, and so on—things which would make it impossible to predict a person's performance from even an exhaustive knowledge of his competence (Topics in Generative Grammar, 9).
Now, armed with this distinction, let us consider a very minor linguistic skill—something which is only a minute fraction of our competence. I refer to the ability to construct the names of the natural numbers: the ability, if you like, to count. This example is suggested by a letter of Descartes, which is one of that philosopher's fuller statements about the nature of language. In 1629, Father Mersenne posted to Descartes a proposal by an un-named author for a universal grammar with an international lexicon, and Descartes sent back several pages of destructive criticism. At the end he went on to say
‘I believe that it would be possible to devise a system to enable one to make up the primitive words and their symbols in such a language so that it could be learnt very quickly. Order is what is needed: all the thoughts which can come into the human mind must be arranged in an order like the natural order of the numbers. In a single day, one can learn to name every one of the infinite series of numbers, and thus to write infinitely many different words in an unknown language. The same could be done for all the other words necessary to express all the other things which fall within the purview of the human mind. If this secret were discovered, I am sure that the language would soon spread throughout the world.’
There seems something Utopian in the idea that all the objects of human thought could be ranged in an order comparable to that of the natural numbers, but Descartes was surely right that the ability to name numbers is itself a sufficiently remarkable one. It has many of the properties which Chomsky insists on in language-use. Knowledge of how to name numbers is innovative in the sense that we can all formulate and understand without difficulty, names of numbers which we have never before uttered or heard. Much of our linguistic use of numerals is also remarkably free from stimulus control: it can hardly be seriously suggested that the utterance of a certain number can be explained as a response stimulated by the presence of n perceptible objects, since at every moment of our lives we are presented with numerable objects, objects numerable in countless different and incompatible ways.
The third of the Cartesian criteria for the creativeness of language, the one which was hardest to make precise, was coherence and appropriateness. In the case of the use of numerals we have rather precise criteria of coherence, the rules of arithmetic; and we have extremely generous standards of appropriateness—to count things or to do mental arithmetic is by itself hardly ever a sign of madness. All this is rather striking in view of the fact that by comparison with other linguistic skills the ability to count appears the most mechanical and the least creative. Finally, as Descartes stresses, the ability to count is, paradigmatically, the ability to make infinite use of finite means. So much so, that to this day text books of transformational grammar point to the existence of an infinity of numerals as the best proof that language itself is infinite. On the other hand, the ability to count doesn't raise a number of the more technical linguistic issues which divide Chomsky from other grammarians.
Obviously, the ability to count has a semantic as well as a syntactic element, and that I shall leave to Professor Longuet-Higgins. I believe that he and his associates have done some interesting work on it. I just want to talk about the syntactic element in the knowledge of numerals, namely the ability to detect well-formedness and ill-formedness in numbers. Each of us here knows that of the two following strings the first is well-formed and the second is ill-formed, as the name of a number. ‘One million, nine hundred and thirty eight thousand, two hundred and nine’ is all right, whereas ‘Two hundred and nine, one million, and nine hundred thirty and thousand eight’ is not. Probably none of us have ever heard those particular strings before. There is literally no end to the number of such names of numbers that we could form and evaluate—at least, if we allow iteration as in ‘a billion billion billion and one’. Clearly, this is quite a remarkable ability, if one reflects on it, and it is one to which Chomsky's distinction between competence and performance clearly applies. When one acquires the competence, one acquires the ability to write infinitely many words. But clearly no-one, as a matter of performance, ever has or ever will write out an infinite set of names of numbers.
One thing that Chomsky stresses about competence is that it isn't necessarily open to introspection. A person is not generally aware of the rules that govern sentence interpretation in a language that he knows. Nor in fact, he says, is there any reason to suppose that the rules can be brought to consciousness, or even necessarily the empirical consequences of these rules. I think Chomsky is right in saying that there is no paradox in the idea that one may know how to apply rules which one cannot formulate and indeed that one may not even realise without special instruction the different ways in which one can apply rules that one does possess. He has quite an interesting illustration of this, in the expression ‘I had a book stolen’. When one first hears this, it's hard to detect any ambiguity in it, yet in fact it can be interpreted in three totally different ways. Consider the following extensions: ‘I had a book stolen from me last night’, which is the natural one; ‘I hired a burglar and I had a book stolen for me’; ‘I had a book stolen, and half a dozen spoons, but before I could steal the rest of the library the police arrived’.
In the area of counting it isn't easy to illustrate unconsciousness of application of the rules, but the unconsciousness of the rules themselves is patent. We can all of us detect misplaced ‘ands’ in the formation of numbers in English, but even the most fluent English speaker, without reflection, is hard-put to give the rule for the placement of ‘and’ or the rule according to which ‘ninety-eight’ is well-formed and ‘eight-ninety’ is ill-formed.
When Chomsky distinguishes between competence and performance he appears commonly to have in mind the difference between competence and deficient performance—for instance, the incorrect enunciation of a numeral which is too long to carry easily in the head. But of course there is a difference, a philosophical difference of category, between a capacity and its exercise, which distinguishes a competence from even the most non-deficient performance. It isn't because of defects in the performance of counters that one cannot exhaust the infinity of the capacity to form numbers. And there is a categorial difference between a capacity and its exercise, even when the capacity is one that is exhausted by its performance—for instance, the capacity to commit suicide, if it is ever expressed in performance, is exhausted by a single successful performance. There is also another important categorial difference between a physical structure which exhibits or embodies a capacity, and that capacity itself—for instance, between the structure of a thermostat and the thermostat's ability to regulate the temperature. It seems to me doubtful whether Chomsky pays as much attention as he should to such categorial differences, and the value of his contribution to the understanding of mind may perhaps suffer in consequence.
There are, in Chomsky's writings, three claims of ascending strength, each of them made about each of these structures. The first is that these abstract structures exist; the second is that they exist in the mind, and the third is that they exist in the form of knowledge. And each of these three claims raises different and serious philosophical problems.
With regard to the first one, I think no particular philosophical difficulty arises in postulating deep structures, if these are regarded as processes which occur in time, related to the time in which surface sentences are uttered. I'm not very clear whether Chomsky does think of deep structures as processes in time. On the one hand, he often talks about them as if they were, and it seems that any empirical verification of his theory about them must involve there being processes in time. On the other hand, he insists that the generation of deep structures is a purely formal, descriptive, matter, and he regards it as an imperfection in Cartesian theory of language that the transformational operations relating deep and surface structures are actual mental operations. (Language & Mind, 16).
There is, as I say, no particular difficulty if they are processes in time because we do have a fairly clear notion of what is meant by the structure of processes in time. But there seems to be a great difficulty in understanding what is meant by the structure of a competence, if a competence is a capacity and not the physical basis for a capacity. A thermostat has parts which fit together in a certain way, but does its ability to regulate the temperature have parts? It doesn't seem to me at all clear that we can make sense of the notion of the structure of a capacity. When Chomsky does talk about competence and their structures, it is surely more likely that he means the structure of the material basis of the capacity, than a structure of the capacity itself—a sort of black box specification. The reason for talking of a structure of a capacity, such as a competence, rather than of its material basis, is that one might only be able to define the parts of the structure in terms of their function, without having any understanding of the nature of the material embodiment of the structure; just as one might well know that the body, since it regulates its own temperature, must contain a thermostat, and yet have no idea of the material structure of the thermostat.
This procedure is, no doubt, legitimate, given that we know so little of the neurophysiology of language. But notice that if this is a correct interpretation of Chomsky, he is appealing tacitly to the principle that in order to have an output of a certain complexity a structure must itself have at least as much complexity. This may be true, and it may be a good heuristic principle, but it isn't an a priori truth and it should be stated explicitly. If this interpretation is correct, why then does Chomsky say that the structures are in the mind, and not say ‘in the brain’? One reason, once given in conversation by Chomsky for not saying that the structures are in the brain is that we don't yet know enough to be sure where in the body the physical analogues of the capacities postulated will be located. As an admirer of Chomsky once put it to me, ‘For all his theory says about it, they may be in the left foot’. Without going so far as that one might say that for all we know the structures may be embodied in a system of properties still as unsuspected by us as electrical properties were by Descartes, so that our present neuro-physiological guesses will in time look as quaint as Descartes’ hydraulic hypotheses.
Now all this may be a good reason for not attributing linguistic abstract entities to the brain, but of course not being in the brain is not by itself a sufficient condition for being in the mind. We normally take it as a criterion for something's being in the mind that it's something about which its possessor can answer questions if asked. But this is not the only possible criterion, and I think Chomsky is right to play down the role of introspection. The reason for regarding these structures as mental—the most obvious one—is that they are structures postulated to explain an activity which is surely a mentalistic one, namely language. But this can't be the only reason for attributing such structures to the mind, because if this were all the explanation would be in danger of collapsing into an explanation of the same form as the theory that opium puts people to sleep because it has a dormitive power. The crucial point is surely that these postulates are postulates whose existence is verified by the occurrence of exhibitions of mental capacity—what Chomsky, not altogether happily, calls the ‘linguistic intuitions of the native language speaker’. Since the intuitions which are predicted by the theory are not the same phenomena as those which it's called on to explain, the ghost of vacuousness can be exorcised.
Thirdly, and finally, let's consider the postulation that these structures are present in the form of knowledge. This seems quite reasonable in the case of deep structures, which can be brought to consciousness by suitable instruction, like the case of the various interpretations of ‘I had a book stolen’.
But once we reach the stage of universal grammar, it's hard to see which of the criteria by which knowledge is ascribed to human beings still remain. It's occurred to many philosophers to ask why knowledge of universal grammar should be attributed to children, when knowledge of Kepler's laws is not attributed to planets. Chomsky's standpat reply is that the attribution of knowledge of Kepler's laws doesn't contribute to the explanation of planetary motion; that the attribution of a knowledge of the rules of grammar does contribute to an explanation of the speaker's ability to use a language. But this answer is inadequate, because even if it's been shown that the internalisation of grammar plays a part in the explanation of language-use, it hasn't been shown that any part at all is played by the assumption that the internalisation is present in the form of knowledge, and indeed the content of this claim has not been made clear, though of course the impact on an audience of the idea that a child knows a universal grammar, is usually quite striking. Chomsky's alternative reply, which is one which is likely to reduce the impact, is that whether this is knowledge or not is simply a sterile terminological matter. But this isn't so. On Chomsky's own account, knowledge of language differs from other forms of knowledge in that by definition a man cannot be wrong about his own language. Now if this is so, then this must make us hesitate not only whether to apply the word ‘knowledge’ to this, which is, as Chomsky says, comparatively trivial, but much more it must make us hesitate to use a man's grasp of his grammar as a paradigm for the understanding of human cognitive ability in general, which is what Chomsky does in moving from the third to the fourth degree of abstraction.
To sum up. Chomsky has three theses about these four abstract structures. From a philosophical point of view, it seems to me that as the theses get stronger, and as the abstractions get more abstract, the problems not only of verification but actually of content become more and more serious. The parallel with Freud is inescapable and is indeed invited by Chomsky himself. Freud's postulation of an unconscious structure of affection, of volition, is at its most comprehensible and plausible in the case of Freudian slips, where the unconscious motivation is capable of being revealed to normal consciousness by non-theory-laden procedures. Similarly, Chomsky's postulation of an unconscious intellectual structure is at its most comprehensible and plausible in the case of deep structures when it can be revealed by the striking pedagogic devices which Chomsky uses without demanding any commitment to theory in the listener. But as the level becomes deeper, in terms of the theory, and therefore more profound in importance, it becomes, it seems to me, more and more open to philosophical question. If my argument has been right, Chomsky's theories are philosophically most unobjectionable at the point at which they are linguistically most controverted—in particular, on the point in which transformational grammars differ from others. This is, perhaps, not an unexpected result, but I would be interested to know whether a linguist, starting the discussion of Chomsky from the other end, would perhaps come to the same conclusion.
Well, I must say that I agree with all the presuppositions which He behind Kenny's brilliant analysis of Chomsky's philosophy. In particular, with the view that language is one of the central keys to the study of the mind. But there are three questions.
I'd like to take up with him. They are quite minor questions.
I'm wondering if he has been quite fair to Chomsky in discussing the concept of competence. If one puts the words ‘the structure of a competence’ together, one certainly gets something which looks very strange. But I haven't ever seen that phrase in Chomsky's writings: I don't know whether you have? I always understood that the concept of competence was adduced really for the purpose of defining what a language is. And it could be argued that competence is something that might be judged by an examiner, for example, who was reading a student's answer to a question in an examination paper on Logic, to see whether the student had in fact produced a formally correct proof of a proposition which was supposed to be proved, using all the right rules. And I invite your comments on that. The question of how students actually do set about constructing proofs when they are answering examination papers is of course a different question entirely from the question of whether the proof is correct once constructed.
Well now, a second point. You were talking about the naming of the integers, and indeed I think it's a very interesting subject and we have thought about it, as you know. You mentioned that it exhibited the feature which Chomsky called ‘creativity’ in that people can name new numbers without being stimulated to do so in any obvious way, and certainly that's true. But if I say ‘one hundred and twenty three’ I say ‘one hundred and twenty three’ on the basis of a number which I have in mind, possibly in visual form. I may have in mind a row of three digits, 123, and read it off as ‘one hundred and twenty three’. If one is not careful one can seem to be suggesting that creativity in language is saying things for no reason whatever, and I would wish to put a large question mark against that, although I'm sure that's not what you meant. And indeed, the whole question of saying what we mean, which I shall be trying to take up tomorrow, is very relevant here.
And thirdly, an empirical question—I don't know what Chomsky's view on this would be. Perhaps you have read enough of his writings to be able to answer. The acquisition of language is, of course, one of the most extraordinary and remarkable tasks that human beings ever perform. And the question is, what conditions are really necessary for the acquisition of a language? If one reads Chomsky one sometimes gets the impression—at least I do—that it is thought to be possible for a child to learn a language by simply receiving samples of that language in sufficient numbers. Now of course our lives and our experience as children are very rich and almost all our experience is non-linguistic. I wonder if you have any comments as to what Chomsky's philosophy might seem to imply about that?
Briefly, I think that Chomsky is committed to the notion that competences have structures, in that he thinks of a universal grammar as a competence, and a universal grammar as an innate structure of the mind. He says both of these things several times. And by simple logic it seems to follow that there are competences which have structures.
With regard to your point about creativity, I don't agree that in order to enunciate the name of a number, one first of all has to conjure it up in the mind in arabic numerals and read it off. But even if one has to, this seems to me merely to put the creativity one stage back in the conjuring it up. Though there are differences between the way in which arabic numerals are linguistic, and the way in which the actual English words are linguistic, I think with regard to creativity there is no important difference. The fact that one can make up long numerals, and read long numerals that one hasn't before met is, in the sense in which Chomsky defines, creative. Now this may be taken two ways: as showing that perhaps creativity as defined by Chomsky is not such an exciting thing after all; or, in the way in which I wanted to take it, as showing that just the ability to count is itself something rather striking.
With regard to the third point, I think Chomsky doesn't want to deny that non-linguistic experience is very important in the learning of language. He thinks that linguistic data, non-linguistic experience and innate structures are three independent factors which interrelate together to produce, as an output, the grammar of a language. He thinks that non-linguistic experience plus linguistic data alone, without the innate capacities, would not be enough to explain the output.
I am not going to say anything very polemical now, but I want to draw out one point about Chomsky. The key word is the word ‘abduction’ on the bottom of the blackboard, which I think is important from the philosopher's point of view, and is important in showing why Chomsky is of philosophical interest: a point which I've always found very difficult. I went to some lectures he gave in Oxford. Enormous crowds of people listening to rather repetitive attacks on Professor Quine, and at last I got what it was, what the mixture was that gave this charisma; it was a mixture of extreme Left Wing politics, and extreme Right Wing metaphysics. And the—Right Wing metaphysics, about which he has absolute honesty, he's pro-Descartes, down with Locke: Locke's theory of innate ideas is all wrong. Nevertheless, he conceals a certain ambiguity. You could take it, as I think Chomsky himself takes it, and as his followers take it, that this is a piece of work in the rationalist tradition. We show the limitations of empiricism, and in Chomsky's latest book, which Kenny lent me on the ‘plane this morning, he was working it out as an attack on Nelson Goodman's extreme empiricist reconstruction of induction. And the message that comes over here, then, is that there is more to the human mind than can be reduced to any little formula. This is a message which I find very congenial, but it is one that could be taken a different way, and was taken a different way, for instance by David Hume. Hume also came to the conclusion that the workings of the human mind couldn't be rationally reconstructed according to the principles which he took as being essential for satisfactory reconstruction of the human mind. Sometimes he will have it that you can't even give a rational justification of induction. We don't talk about ‘induction’ any more, but something rather grander, ‘abduction’. On other occasions, Hume will allow induction, but then comes to the conclusion that our moral reasonings are ones which are not really reasons at all, but only are a matter of sentiment, a question too nice for the operation of human reason. We have to settle it by our sentiment or our emotion. And the problem which I think now comes to be decided, on philosophical evidence, is whether, supposing Chomsky is right about our linguistic abilities, is this something which is only for Waddington to be interested in. Curious things, human beings: not only are they featherless bipeds, but they've got deep structure, universal grammar. This is a matter of great interest, but only to the biologists who find it very interesting why human beings come this way rather than the other. Or, is it, as once Chomsky, I think Kenny, and certainly Lucas, want to maintain, this shows something very important about human beings. Not merely featherless bipeds, but rational agents.
I've got very few remarks to make in this connection, because of course, as a biologist I'm mainly concerned with organisms which do not indulge in language, and some of them have what I should be willing to call ‘minds’. They can learn, they can do a variety of performances, but they don't actually talk. However, we'll come back to that consideration later. I've never really felt I had any firm understanding of the Chomsky scheme, certainly when you get down to the level of universal grammar. But I just want to ask a question. We've had this talk about naming, for instance forming correctly formed names of integers. What is the grammar like in languages that don't have, for instance, a zero? Recently I came across some South Africans, in whose native language you can count up only to five; then it is ‘many’, and no distinction amongst the many. And no zero; The absence of zero, I should think, would make a very considerable difference to all grammatical structures. Is this part of a universal grammar, and does the universal grammar contain a zero or not? Related to this is the question ‘Is the universal grammar related to the experienced world?’ Chomsky says it's now innate, but I'm an evolutionist, and things become innate through processes of evolution, usually because they are in some way useful. Is this what is supposed to be the case with universal grammars? And if so, what produces a zero in a universal grammar that for a long time lacked a zero? We are told that it was invented by the Indians—a Sanskrit invention, I believe. But how does it come to be invented? What is the status of invention of new items in universal grammar?
I think Lucas has wrongly claimed Chomsky as an ally in thinking that the innateness of universal grammar is something which marks human beings off as rational agents. Chomsky believes, I think, that there could be perfectly rational Martians who didn't have our universal grammar innate and for all I know had no universal grammar innate, but learnt language as we learn mathematics. (I'm not quite sure why Chomsky excludes mathematics from depending on an innate component, as he appears to do.) But the Martians, so far as I can tell from Chomsky's brief descriptions of them, could very well turn out to be people about whom all the theses of B. F. Skinner were true, and therefore, from Chomsky's point of view, turn out to be very unpleasant creatures indeed.
To turn from the Martians to the non-human terrestrial creatures whom Waddington reminds us of—I think that Chomsky has had very little to say about the evolution of language-users from animals who did not have the use of language, though he does have a number of rather caustic remarks about people who think that natural selection is the name of a theory with any content. What lies behind these remarks I do not know. But he has always by-passed this question by saying ‘well, the first thing to decide is whether or not it's correct that it's innate; only if it is innate do we have to go back and ask the question when, if ever, did it become innate?’
I would like to clarify something in my paper which I think Waddington and perhaps others misunderstood, which was the function of the example about counting. This wasn't meant to illustrate all of Chomsky's theses, and it was certainly not meant to suggest that Chomsky thought that the English or Arabic system was itself innate. It was meant primarily to illustrate the distinction between performance and competence and what Chomsky meant by creativity in language. In fact, the principles of universal grammar are very much more abstract than the rules for counting. They are such principles as ‘grammars must contain only structure-dependent transformations’, and so to the question ‘is there a zero in universal grammar?’, I think the answer is definitely ‘no’.