You are here

Seventh Lecture. The Origin of Language

Anthony Kenny

In the first four lectures, we discussed the development of mind in the individual human being. In this final batch of lectures we are considering the origin and future of the mind in the species.

Waddington attaches great importance to the notion of goals as defining mentality, and therefore when he was talking about the origin of mind he spoke in particular about the possible origins of goal-directed behaviour. In my first lecture, I offered a rather different definition of mind—the mind, I said, is the capacity to acquire intellectual abilities, and by intellectual abilities I mean activities involving operations with symbols. So now that it's my turn to talk about the origin of mind, what I have to discuss is how people can have begun to operate with symbols; I have to face the problem of origin of language.

No-one really knows anything about the origin of language. John Lucas reminded us that, in the book of Genesis, we're told that God fashioned all the wild beasts and all the birds of heaven and brought them to Adam to see what Adam would call them. Each of them was to bear the name that Adam would give it. I don't know if John takes this to be the literal truth about the matter. I imagine that probably few people do, but we haven't yet found an account to replace the Genesis myth.

‘The biological history of language cannot be revealed through a random comparison with animal communication’, says a biologist. ‘Reconstruction of the origin of language is impossible except for some very simple determinations. This is because of the following limitations: (1) the size and shape of the brain furnish no secure clue about the capacity for language; (2) given morphological peculiarities of the CNS do not bear a fixed relationship to behaviour; (3) even if we had direct knowledge of social structure or cultural complexity of the societies of various fossil men, we could not draw conclusions about language as we know it today. Different types of communication might have prevailed at those times.’ (Lenneberg, Biological Foundations of Language, p. 266.)

The linguists are similarly pessimistic. Lyons has written recently: ‘What one might call the “official” or “orthodox” professional attitude of linguists to evolutionary theories of the origin of language tends… to be one of agnosticism. Psychologists, biologists, ethologists and others might say, if they so wish, that language must have evolved from non-verbal communication; the fact remains that there is no actual evidence from language to support this belief.’ (Hinde, Non-verbal Communication, 76.)

Obviously, I'm not going to claim to know where language came from when the experts can only guess. On the contrary I'm going to suggest this evening that if anything Lenneberg and Lyons, in the passages I've quoted, underestimate the difficulty of explaining the origin of language by natural selection from animal communication systems. It isn't just that languages themselves don't evolve in the appropriate way, or that we don't know much about the brains of primitive men because brains don't fossilise. The difficulties are difficulties of principle: language as we know it has features which are prima facie inexplicable by natural selection.

I complained yesterday that Waddington didn't explain how language might have originated among a community that didn't have language. He did talk about the acquisition of language by children: but as his account presupposed that they learnt language from their parents, it offered no explanation of the origin of language in the population as a whole. However, in his first lecture he did make such a suggestion. He suggested that language might have developed by ‘the formation of a chreod surpassing the needs of integration which brought it into being’; as a floating animal with three stiff-jointed rods would greatly increase its stability if they formed a triangle, perhaps beyond anything actually required for protection.

‘It is not impossible’, Waddington went on to say, ‘that the universal generative grammar spoken of by Chomsky originated in a rather similar way; in some set of rules, which were at first favoured by natural selection because they made it possible to elaborate slightly on the degree of communication that could be carried out by ungrammatical grunts, shouts and so on, but which then turned out to be enormously more powerful than required by the immediate needs of the situation.’

As I remarked at the time, I think that in this passage Waddington somewhat misrepresents the nature of Chomsky's theory. He seems to treat universal grammar as differing from particular grammar rather as Esperanto differs from English. Now there are, as we shall see, difficulties in the idea that a grammar, like that of English or of Esperanto, might have evolved. But the difficulties in thinking that a universal grammar might have evolved are difficulties of a very special kind. I should like to insist once more on the highly abstract nature of the structure postulated by Chomsky. A child has to acquire the grammar of a language: he has to internalise the rules governing its use. To do so, Chomsky says, he must unconsciously frame a hypothesis about the nature and structure of the linguistic data fed him by his parents and other humans in the environment. Now the data available to him are too fragmentary, according to Chomsky, for him to be able to acquire the grammar unless he had some innate knowledge of the type of hypothesis that is acceptable. Universal grammar, then is itself not so much a grammar as a sieve through which candidate hypothetical grammars must pass. To see the relation between particular grammars and universal grammar, imagine that a Ministry of Sport sets up regulations governing the forming of rules of particular games. Particular grammars will then be related to universal grammar as particular rules are related to Ministry regulations. Neither the Ministry's law nor the set of provisions it will contain will be sufficient to enable one to play any game; the sort of provisions it will include will be ‘no rule of any game may be longer than fourteen words’, ‘no infringement of a rule of any game shall be punishable by death’, etc.

Whatever difficulties there are about explaining by natural selection the internalisation of the rules of a particular language, it is surely doubly difficult to explain in that way the internalisation of a set of rules about rules. For the biological utility of having an internal sieve to assist in the mastery of grammars cannot antedate the existence of grammars themselves. To try to explain the origin of universal grammar by natural selection seems as least as difficult as trying to explain the utility of a light-sensitive cell to an animal living in an environment totally without light.

However, Chomsky's postulate of universal grammar is still a controversial hypothesis. If the innateness hypothesis is incompatible with explaining the origin of human language by natural selection, one may be inclined to say: ‘So much the worse for Chomsky's hypothesis’. So I pass to the second, and major objection that I wish to raise against evolutionary explanation of the origin of language. This is based not upon a dubitable hypothesis but upon a truism: the truism that language is a social, conventional, rule-governed activity. The importance of this truism does not seem to have yet been fully appreciated by all biologists and linguists, although it has, of course, been much stressed by philosophers in recent decades. Three philosophical works are worth mentioning in illustration: Wittgenstein's Philosophical Grammar (1933), Bennett's Rationality (1964), and Searle's Speech Acts (1967).

The most important difference between language and other media of communication is that language involves the use of symbols. Symbols differ from other signs (traces, clues, symptoms) in being conventional and not natural signs. When we say that language is conventional we do not mean that it was set up by some primeval linguistic contract on the style of Rousseau's social contract. Such a suggestion would be grotesque, not to say self-contradictory: obviously one cannot make a contract that a certain word is to mean X unless one already has the linguistic means of referring to X. Rather, when we say that language is conventional we mean that the behaviour of language users is rule-governed. This means, among other things, that language-behaviour is not as such governed by causal regularities: one can only keep a rule if one can also break it. There is no law of nature to the effect that our utterances have the meaning that they have.

There are several differences between rules governing behaviour and natural laws governing phenomena. One is that it is possible, and it often happens, that rules are violated. Short of a miracle it isn't possible for a natural law to be violated. If you find an apparent violation of a natural law you have an indication that the law has been wrongly stated. The occurrence of a violation of a rule, on the contrary, is no evidence that the rule has been wrongly framed.

Another difference is that if someone's behaviour is governed by rule he must be to some degree conscious of the rule. This does not mean that he must be able to formulate or enunciate the rule: it is notoriously hard even for a fluent language speaker to enunciate the phonological, syntactic and semantic rules he uses. What it means is that the user of a rule must be able to distinguish between correct and incorrect applications of the rule; he must know the difference between following and violating the rule if he can be said to be using the rule at all. In contrast, one can be operated upon by a natural law without having any consciousness of it at all. For it to be true that mammals are procreated as a consequence of sexual intercourse there is no need for mammals to have any knowledge of this fact; on the other hand if a community operates a certain transformation rule for interrogatives they must be capable of distinguishing between well-formed and ill-formed interrogatives.

Now the rule-governed nature of languages makes it difficult to explain the origin of language by natural selection. I shall say why in a moment: let me point out now that in my account of the nature of convention, I have not said anything about innateness. I contrasted conventional rules with natural laws, but I am not using ‘conventional’ to contrast with ‘innate’. Something might be innate in the sense of not being acquired by learning, and yet be conventional in the sense that it was an activity in accordance with rules and not a mechanical procedure.

The difficulty is this: the explanation by natural selection of the origin of a feature in a population presupposes the occurrence of that feature in particular individuals of the population. Waddington has earlier given us examples about the development of the length of legs in horses. One can very easily understand how natural selection might favour a certain length of leg; if it were advantageous to have long legs, then the long-legged individuals in the population might outbreed the others. Clearly, where such explanation of the occurrence of features is most obviously apposite, it is perfectly possible to conceive the occurrence of the feature in single individuals. There is no problem about describing a single individual as long-legged, or as having legs n metres long. (There may be a problem about the origin of the individual long-legged specimen, a problem which may or may not be solved by talking of random mutation; what I am concerned with is not the origin, but the conceivability, of the favoured individual specimen.)

Now it does not seem at all plausible to suggest, in a precisely parallel way, that the human race may have begun to use language because the language-using individuals among the population were advantaged and so outbred the non-language-using individuals. This isn't simply because of the difficulty of seeing how spontaneous mutation could produce a language-using individual; it is the difficulty of seeing how anyone could be described as a language-using individual at all at a stage before there was a community of language-users. For consider; I am using language now in lecturing, and that I am doing so depends no doubt on decisions of my own and is conditioned on all kinds of ways by my own physiology; but whatever I did, whatever noises and gestures I made, they could not have the meaning my words now have were it not for the existence of conventions not of my making, and the activities of countless other users of the English language. If we reflect in this way on the social and conventional nature of language, then we begin to see something very odd about the idea that language may have evolved because of the advantages possessed by language-users over non-language-users. It seems as absurd as the idea that golf may have evolved because golf-players had an advantage over non-golf-players in the struggle for life, or that banks evolved because those people born with an innate cheque-writing ability were better off than those born without it.

We don't, of course, think of games like golf and institutions like banks as having been evolved; we think of them as having been invented. I don't wish to suggest that the origin of language can be explained in the same way. In order to be able to invent an instrument for a particular purpose you need to be able to conceive that purpose in advance and devise the invention as a means to it; whereas it doesn't seem that someone who didn't have a language could first of all conceive a purpose which language would serve, and then devise language as a means to serve it. Some human procedures were hit upon by accident, as in the legend that someone discovered the utility of roasting pork by having his house catch fire with his pig inside. It doesn't seem that language could originate in the same way, because such unexpected fortunate discoveries can surely occur only in the case of causal results, not of rule-governed activities like language. One can't conceive of a man's being the first accidentally to follow a set of linguistic rules as one can conceive of him being the first accidentally to set fire to his house.

The difficulty that I posed about the natural selection of language may seem to be one not peculiar to the present case. If the difficulty is that an individual's behaviour isn't linguistic behaviour except in the context of the behaviour of others, is not this true of all kinds of social behaviour, including some of which dumb animals are undoubtedly capable? A courtship ritual, one might say, would not be what it is without the responsive behaviour of the mate. Nor, for that matter, would sexual organs be sexual organs were it not for the corresponding parts of the anatomy of the opposite sex. For all that, there seems no special difficulty in postulating the evolution of sexual organs or courtship rituals.

Both of these parallels, I think, break down: to see why will take us a step further. Sexual anatomy, like long legs, is conceivable and describable without reference to the counterpart anatomy, though of course its usefulness for procreation cannot be brought out without reference to the counterpart. The difficulty concerns utility, not conceivability. It is not the sharing of rules between language-speakers that would provide the linguistic parallel, but such things as that an animal could not use for species-specific communication sounds that were inaudible to other members of the species.

The parallel between language and ritualised animal activity is a closer one, though it too breaks down. Ritualised activities, in fact, occupy an intermediate position between language and causally efficacious behaviour. They appear often to be displacement activities: that is, they are causally inappropriate to produce the particular effect currently desired. In this rituals resemble language in not being part of a causal mechanism operated to produce an effect. But unlike language they do not seem to be rule-governed, that is to say, they don't seem to be accompanied by any behaviour expressive of a conscious discrimination of violation. Moreover, the ritual activities can be given precise behavioural specification (e.g. ‘the two partners stand with their heads raised, the beak pointing downwards and the head turned away from the partner’, J. Maynard Smith, The Theory of Evolution, p. 169, etc.). Linguistic activity cannot be behaviouristically described in this way, because very many different behaviours may be the same linguistic act. For instance, I use the expression ‘No Smoking’ whether I say it or write it, but the behaviour involved in writing it is totally different from that involved in saying it. This does not mean that it is the invention of writing, rather than the origin of speech, that marks the boundary between human language and animal communication. It means rather that the way in which human language is conventionalised is quite different from the way in which animal behaviours are ritualised. For part of what we mean when we say that human language is conventional is that any kind of behaviour will do provided that it obeys the right rules. Not only speaking and writing but morse code, semaphore, American sign-language, etc., can be used to utter the same sentence, given the appropriate conventions. Indeed, it underestimates the crucial function of rules in language to say that language is governed by rules. Language is rather constituted by the rules. Take any human sign language: you can change anything in the sign apart from the rules governing its use and the language will remain the same. Courtship rituals provide no parallel to this.

I conclude then that the difficulty in using the principles of natural selection to explain the origin of language is a special one, affecting language in a unique way by comparison with other phenomena natural selection is called on to explain. I do not say that it is an insuperable difficulty; but it is a difficulty which, so far as I know, has not yet been seriously faced.

The difficulty I have raised remains a difficulty whether one conceives language, in the individual, as a totally learnt phenomenon or as one involving an innate component. But as I have said the difficulty is compounded if one accepts Chomsky's view that the facts of language learning in the individual can only be explained if we postulate an innate mental structure, or universal grammar.

Christopher is fond of reminding us that in evolution ontogeny recapitulates phylogeny. In the case of the origin of language it seems that this cannot be the case. However far we go with Chomsky in postulating innate mechanisms we cannot deny that what children acquire, when they learn language, is the language of their parents: it is a language spoken in their environment. Whatever it was that the first language-users acquired, it was not a language taught them by their parents. Learning, of course, in general, does not demand teaching as a correlate: a rat may learn its way around a maze without a teacher by trial and error. But language could not originate by trial and error learning; because the notions of trial and error presuppose stable goals which successive attempts realise or fail to realise (the acquisition of the food pellet, etc.). But there is no independently specifiable goal to which language is a means: the communication of thoughts cannot be regarded as such a goal, because there are so many thoughts which can only be expressed in language. In particular, one can't have the goal of acquiring a language because one needs a language to have that wish in.

In 1967 Lenneberg remarked that it was a weakness of the theory that human language was continuous to animal systems of communication, that the examples cited in support of the theory were drawn from all over the animal kingdom, birds, insects, fishes, mammals, in complete disregard for phylogenetic proximity to man. It might be thought that his argument has been weakened since he wrote because of the successes of Washoe and Sarah: chimpanzees, reasonably close to man phylogenetically, who have shown remarkable abilities to mimic language. Now there is some dispute as to how far these two chimps have really mastered syntax; but let us waive this and accept for the moment that their linguistic skills are beyond reproach. The results still seem irrelevant to the prehistory of language for two reasons. The first, and less important, is that the modalities in which they have learnt language are very different from any known to any but very recent generations of humans, thereby making any phylogenetic continuity improbable. The second, more important one, follows from the point I have just been making: Sarah and Washoe were undoubtedly taught by the human families with whom they lived; and the demonstration of the ability to be taught a language in the ancestors of human beings would not go any way at all to explain the origin of language itself. For our problem was one that would arise even if there had been men around for generations before language started.

Striking as the successes of Sarah and Washoe have been, they do not in themselves have anything to tell us about the origin of language. The possibility of teaching human language to a chimpanzee no more proves that human language evolved from a pre-human communication system than the possibility of teaching a chimpanzee to ride a bicycle shows that bicycles evolved from pre-human transportation systems.

The difficulties in explaining the origin of language by natural selection are masked by the language we use to describe the phenomena. First, we tend to group together quite different human skills and performances; second, we tend to use abstract descriptions of human and animal communications in ways which conceal their differences.

An example of the first difficulty occurs in Popper's paper Clouds and Clocks, as Chomsky pointed out in Language and Mind. Popper had argued that the evolution of language passed through several stages including a lower stage in which vocal gestures were used for the expression of emotions, and a higher stage in which articulated sound was used for the communication of thought. Chomsky remarks: ‘His discussion of stages of evolution of language suggests a kind of continuity, but in fact he establishes no relation between the lower and higher stages and does not suggest a mechanism whereby transition can take place from one stage to the next. In short he gives no argument to show that the stages belong to a single evolutionary process… In fact it is difficult to see what links these stages at all… There is no more of a basis for assuming an evolutionary development of “higher” from “lower” stages in this case than there is for assuming an evolutionary development from breathing to walking. The stages have no significant analogy, it appears, and seem to involve entirely different processes and principles.’

Another instance of the second difficulty occurs in the celebrated ‘design features’ isolated by Hockett for the comparative description of human and animal communication systems. In two papers (1960, 1968 with Altmann) Hockett isolated sixteen characteristic features of human language with a view to investigating which of them could be found also in animals. The sixteen are: (1) vocal-auditory channelling; (2) broadcast transmission and directional reception; (3) rapid fading; (4) interchangeability—speakers can be hearers and vice versa; (5) complete feed-back—the speaker hears what he says; (6) specialisation—the linguistic signals don't do any non-linguistic work; (7) semanticity; (8) arbitrariness—there is no resemblance between elements in language and their denotation; (9) discreteness—the signals are digital rather than analogue; (10) displacement—the signals refer to things remote in time and space; (11) openness—new messages are freely coined and easily understood; (12) tradition—the communication system can be passed on by teaching; (13) duality of patterning—combinations of meaningless elements can be meaningful; (14) prevarication—the ability to lie or mislead; (15) reflexiveness—the ability to talk about language; (16) transfer of learnability to other language-systems.

These features obviously provide a very useful framework for investigation. But in the light of what we have been saying, the most striking thing about them is that rule-governedness, the most characteristic feature of human language, is not mentioned at all. It is possible that rule-governedness is implicit in some of them, e.g. ‘prevarication’, ‘displacement’ and ‘semanticity’, but Hockett does not seem to be aware of this (he defines ‘semanticity’ in terms of associative ties).

It is possible that the use of the language of communication engineering in biology is a source also of the masking of difficulties. At an earlier meeting we discussed in what sense one could speak literally of a genetic code, for instance. The reason for calling it a code is that there is no known law specifying which group of bases in a strand of DNA determines which protein (string of amino acids). The correlation seems quite arbitrary. But that is not enough to make it a code in the sense of a language: for that to be so the operation of the DNA would have to be by means of rules (like a chain of command) and not by causal mechanisms (like that of a template). To see the difference think of a piano and a pianola. The holes in a pianola roll are correlated with the notes the pianola plays just as the notes on the page of the score correlated with the notes the concert pianist plays. But the holes in the pianola roll are not symbols in the way that the notes on the page are.

In general, the fact that X is isomorphic to and merely arbitrarily connected with Y does not make X mean Y or be a symbol for Y. In addition it is necessary that X should be linked with Y by convention, and that demands the intervention of voluntary agents acting intentionally according to rule.

If this were not so, the world would be full of symbols. Given suitable conventions, the fact that this glass is on the table might express the fact that I am sitting on that chair. That is, if the glass stands for me, and the table stands for the chair, and the relation of being on top of stands for itself, we might conceive the glass on the table as a false sentence saying that I am sitting in the chair. The situation has the right logical multiplicity to represent, and the connection between the two situations is entirely arbitrary. But of course that is not enough. The glass on the table isn't a sentence, and that is because we haven't set up the appropriate conventions: we don't use it as a symbol.

For something to be a linguistic representation of a state of affairs it not only needs to have the appropriate abstract structure, it needs to consist of elements conventionally correlated with elements of the structure to be represented. And conventions can only be set up by those who can use symbols: in the normal case, by human beings with the parts and passions of human beings. It is for this reason that I have denied, and deny, that the elements in a computer bearing particular information mean that information and are symbols for that information; and thus I have maintained that the output of a computer is only meaningful and symbolic to the extent that the designers and programmers of the computer have set up the appropriate conventions.

I conclude that there are two things which language must have: (1) an abstract structure capable of translation into diverse behavioural modalities; (2) a user who is capable of giving meaning to that abstract structure, by using it in rule-governed behaviour of the appropriate sort. Non-human animal systems of communication in the natural state appear to me to lack both of these features. If human language has evolved from non-human communication systems then at some time the activity of following rules must have been produced by natural selection. And there seem to be profound difficulties of principle in seeing how the practice of following rules could have originated in such a way.



I agree with Tony that a great deal remains mysterious about the evolution of language. You have heard enough of my ideas already to know that I do not think that ‘natural selection of random gene mutations’ is a magic formula which clears up all the problems. But in connection with Tony's arguments, I first ought to remind you that the evolution of a great many other things besides language remains very mysterious too. We cannot give a full evolutionary explanation of the origins of any of the major phyla. How did the worms originate? How did the insects, or the vertebrates? We do not really know in any detail. It seems that the evolution of these major changes of general organisation happens very rapidly; the system goes very quickly over from one strongly chreodic organisation into another, so rapidly indeed that very few intermediates were ever formed, and one hardly ever finds them in fossils. In a few places we have one or two intermediates, for instance between reptiles and birds, but as a general rule the intermediate steps in major changes in evolution are very little known.

There are some people, of course, who take this as definite evidence that evolution is not the whole story, and that there must have been Creation as well. In recent years there seems to have been, in some parts of the world, an increasing tendency to accept such arguments. You have probably read that the State of California has passed a law that all school textbooks of biology must include an account of Creation as an explanation of the diversity of the living world, alternative to evolution, and, the State legislature claims, equally intellectually respectable. I somehow doubt whether Tony would wish his argument to be pressed quite so far.

However, I think our knowledge of the evolution of language is in a rather similar state to our knowledge of other major evolutionary changes, in the sense that we have very little actual evidence of intermediate stages. I am no expert on comparative linguistics, but from what I am told there are no such things in the world as really primitive languages. Apparently you can say anything equally well in any of the known languages. Some of them may be a bit clumsier in handling certain kinds of subject matter, and others may be especially appropriate in other connections, but all languages seem to be essentially equivalent in their efficiency as means of communication. But to an evolutionist what this implies is not that there never were any intermediate stages, but rather that the steps leading to fully developed languages were passed through quickly, and the intermediate stages have disappeared. I think that Tony finds it difficult to account for the evolution of language because he takes it for granted that he has to account for the evolution of a fully developed language at one step. For instance, he says: ‘Whatever it was that the first language-users acquired, it was not a language taught them by their parents’. Because by definition the parents (people before the first language-users) could not use language. But surely this is just the old paradox of which came first, the chicken or the egg? I should argue that the parents may have had a primitive sort of language, which the children learnt and improved on; in fact, improved on quite quickly in evolutionary terms, until what had started as only semi-articulate grunts became converted into a fully-fledged language.

I would now like to pass on to some further, more particular aspects of Tony's argument. One point he made was that human language is conventionalised in quite a different way from that in which animal behaviour is ritualised. He argued that when we say that human language is conventional, we mean that many types of behaviour will be adequate, provided they obey certain general rules. We can utter the same sentence using several different modes of expression; not only speech, but writing, the morse code and several others. I should agree that this is true of many languages as they exist today, but I doubt if it is very relevant when we are considering the evolutionary origin of language. After all, writing was certainly invented long after language was first expressed in speech. There are many languages spoken in the world today for which no conventional form of writing exists, and I suppose about half of the world's population is illiterate, although they can all speak some language or other quite adequately. The other modalities, such as the morse code, semaphore and so on, have for the most part been invented only in the last one or two hundred years. I see no reason to suppose that at the time of its evolutionary origin language could be expressed in anything more than the modality of speech (I include in that the bodily gestures that usually accompany speech).

Tony goes on to argue that animal communication is of quite a different kind to language, in that it does not use conventional forms which convert certain things into symbols. He gave as an example the types of behaviour involved in sexual display. However, I think there are other types of animal behaviour which come much closer to using symbolic forms in communication. Consider for instance intra-specific fighting, that is to say fighting between two members of the same species, such as two dogs. It is very common to find that such fights are terminated before any great damage has been done. Possibly the under-dog rolls over on its back and exposes its belly or makes some other similar move, but this is not taken by the upper-dog as an opportunity to get in a really wounding bite, but rather is accepted as a signal that the other chap is throwing in the towel, and calling the fight off. The evolutionary mechanisms for producing such types of behaviour have recently been discussed by Maynard Smith in On Evolution. His discussion is in rather elaborate terms taken from Games Theory, but I am not sure whether this is really necessary. I think a main, and perhaps a sufficient, point is that when such a conventional ‘throwing in the towel’ signal becomes widely accepted, it will be very useful not only to the under-dog, but even to the upper-dog, because after all who can know that the upper-dog today will not find himself the under-dog in another fight tomorrow? I do not see any impossibility in supposing that there could be a gradual evolution of a more and more definite acceptance of various forms of behaviour as symbols in a system of communication, rather than as incidents in a causal sequence of processes. This might apply whether the types of behaviour accepted as symbols were movements such as the dog rolling on its back, or were sounds accepted as words.

It seems to me that human language might have evolved in a way very similar to the evolution of these conventional gestures in animals. We know that early man often hunted game much larger than himself, for instance mammoths. Presumably the men hunted in co-operating groups, and surely methods of communication would have been very useful. It would have been advantageous not merely to make a loud noise and point, but to be able to convey things like ‘It's not behind that bush, you silly clot! It's in that bunch of reeds. And it isn't a lion, it's a buffalo.’ This would probably require only quite a small vocabulary, and a relatively primitive mental apparatus, but it would definitely be a beginning from which language might have developed, and it would also be related to some of the things which we know that animals can do. Birds of the same species, but from different localities, may sing slightly different songs, which their fellows seem to recognise. Some birds have different alarm cries, which indicate whether the source of alarm is a hawk or a snake. These cries are, of course, not names in a fully symbolic linguistic sense, but they do convey instructions of some moderate degree of precision. As we have seen in previous talks, language is to be thought of rather as instructions than as statements or descriptions, and these types of animal behaviour involve the conveying of instructions and therefore provide, it seems to me, a basis from which language might have evolved.

I admit that the basis for language provided by these observations on animal behaviour is not very sturdy. We can find some clue as to how the symbolic expression of instructions might have evolved, but it is more difficult to see the origin of the structural grammar by which separate instructions are related to one another. We know very little about this, and I rather doubt whether many people have studied those aspects of animal behaviour which look most likely to provide suggestive information. I should like to know, for instance, much more about anything analogous to grammar underlying the structure of communications which a shepherd may have with a sheep-dog. If one watches the dogs being controlled at a sheep-dog trial, it is quite clear that precise and complicated instructions are being communicated, although the communication is non-verbal. It looks, at first sight at least, as though it must have a structure of a kind analogous to grammar. I wonder if anyone has really studied this fully in the light of modern ideas about linguistics. I think it is in this sort of context that one should look for the evolutionary origins of the grammatical structure of language.

Thus, although I agree that we still have an enormous amount to learn about the evolution of language, I do not think it is impossible to get some general clues as to the ways in which it might have happened. Tony seemed to argue not only that we do not know what actually happened in the evolution of language, but that it is in principle impossible to conceive of language as having come into being by a process of evolution. I believe that his difficulty arises because he was thinking of language as having evolved as a single sudden jump, whereas I should argue that it must have been acquired in a series of steps, probably rapid steps, after which man did what Wittgenstein described as ‘throwing away the ladder after he has climbed up on it’.


The question in Tony's interesting talk which struck me as the most challenging one was, where could universal grammar have come from, how could it have originated. I don't suppose he would want to put up a barrier, and say it was no good for us to continue to ask this question, and so I want to try and pursue it for a minute or two.

Quite plainly, as Wad's just pointed out, the grammars of the sorts of language we have now are much more complicated, to say the very least, than the grammars of the earliest languages, and one must suppose therefore that the universal grammar which determines our approach to the noises that our parents make is also probably a good deal more complicated now than it once was. The most simple rule that could have any claim to be taken seriously as a rule of universal grammar, is one according to which any human language must have a phrase structure grammar. We know that human languages have more than just phrase structure grammars, but phrase structure is the most important obvious element in the structure of all human languages. In the book by Miller, Galanter & Pribram called Plans and the Structure of Behaviour, the authors consider how we might try and assign structure, in an abstract sense, to different sorts of behaviour. They make a very plausible case for saying that much of our rational behaviour has a structure which is like phrase structure. Let me try and indicate what I mean. In making a plan, one may decide to do A and then B in order to carry out the plan. But in order to do A, one has to do X and then Y, and to do B one has to do Z; so if one is going to carry out the whole plan one first does X and then Y, and then one turns to the next stage in the plan and proceeds in the same manner. Now. a linguist will recognise that the order X, Y, Z which arises from this structure (see figure) is closely analogous to the order of the words in a structured sentence. Now, I'm not suggesting for a moment that our languages as we find them are simply characterisable by saying that they have phrase structure, of course not; but none the less, the most up-to-date theories of language take phrase structure as the basic component of the grammar, and any subsequent transformations are applied to this structure. I just wonder whether along such lines we mightn't hope to understand a little more about the origin of universal grammar. Perhaps universal grammar took over for symbolic communication a set of ways of thinking which were already useful in the formulation of plans; and presumably the ability to formulate plans must have considerably predated the appearance of language as we know it.


I'm going to find myself in the position of being much more of a naturalist than Tony. His position depends on a false dilemma between the language which he says is constituted by conventions, and the other source of signs which are based not on convention but on displacement activity; and he wants to drive a very sharp wedge between these in order to produce a gap over which evolution cannot leap. But conventions can be natural; and if you start by having a convention which is natural then it can become more and more conventional thereafter. For instance, we often now in university circles see languages in very abstract terms. But if you come next year to read what we've said today and read it in cold print, all the words will be there but you'll find it's very heavy reading. You depend immensely for understanding on context; in our case the way our hands work and all sorts of other things, in order to get the full sense of what we are saying. Even in our sophisticated languages we are still very much dependent on a large number of other forms of behaviour which are only partly conventional, and partly natural. Or take a second example: earlier today I was in Edinburgh Castle which was redolent with the symbols of Royalty—Crowns, Thistles, Orbs and Sceptres; there was a Sword of State, and this now is a convention of the civil power; but once it wan't a convention—it was a real threat; when Bruce wielded that sword, you obeyed. Even now the Keepers of the Castle recognise this point, and next to the Crown Jewels of Scotland, they have the RAF room, and the battle-colours and the guns.

This explains how the jackpot phenomenon could have occurred. You start with various natural conventions, the dog giving his underflank to indicate its willingness to be killed—this is absolutely natural—and then because it is natural, it becomes understood, and because it becomes understood, it becomes in its own right a symbol. It seems to me perfectly unexceptionable that some account of the sort that Wad has given should bridge the gap between what are at first obvious responses to the situation, a purely natural form of activity, and later, more conventional, behaviour. We move in our own society in a very similar way—from the original generosity to the conventional glass of sherry or conventional hand-shake; when I shake your hand, it no longer means that I haven't got a sword in my hand, and so, that I will not kill you—it is an entirely conventional sign.

This is one of the difficulties that Tony put forward. There's another one, however, which was about how to get an abstract grammar, and this is largely answered by Wad who points out that you can start with a much lower-grade one, and then work up to the abstract one by having a series of translation rules—you can go from speaking to writing, to Morse, to all sorts of other symbols, and this is what underlies the concept of abstractness. Then there's one other biological point, namely that even the birds, I'm credibly informed, have some sort of abstract grammar. Konrad Lorenz and the other ethologists brought up some birds in human company and the birds made a great mistake; they thought that they were human beings, and that Professor and Mrs Lorenz were their parents. It is clear that in this case the birds had an innate abstract grammar in the sense of a correlation between parents and the appropriate sort of behaviour, which has then been filled in with the wrong filler. The birds' mistake lay in their specification of this abstract grammar into a particular one, due to their misidentifying their parents. They therefore tried to feed their parents with worms—and as they found they couldn't get them into their mouths, they put them into their ears.


I don't want to claim, and I don't think I did claim, that it is in principle impossible to explain the origin of language by natural selection. I said that there was a difficulty of principle which had not been faced up to, that is, a different difficulty from the general difficulties which Wad reminded us of, that crop up all over the evolutionary scale. In the case of the origin of the birds, what we are lacking is knowledge of the existence of the intermediate cases which the theory would postulate. The special difficulty with regard to the origin of language, is the difficulty not of providing an intermediate case, but conceiving exactly what is an intermediate case between linguistic and non-linguistic behaviour. In spite of the erudition and ingenuity of the other symposiasts I'm not convinced that this evening produced something which is at all like an intermediate.

I don't deny that animal systems of communication are very efficient in communicating a lot of information to one animal about another, but you can do that without having a language as you can best see if you study human non-verbal communication of the kind which John mentioned. If you overhear a conversation in a neighbouring railway compartment in a language of which you don't understand a word, you can get a lot of information about the age, the sex, the temper, and so on, of the people. You may realise, for example, that there is an old woman who is very angry indeed with a small child who is being very naughty. This is the sort of thing which is a close parallel to the information that is communicated in the bird's system of communication. The examples, the human analogues which John gave were, it seems to me, all non-linguistic cases. If one is going to use the word ‘grammar’ so broadly that even putting worms in somebody's ear is an exercise of a linguistic ability, then all serious discussion of language must come to an end.

From the book: