Language and evolution

by Berwick and Chomsky

MIT Press  ISBN 9-780262-034241  £17.95

reviewed by Alan Dent


The epigraph for this book might be Hilaire Belloc’s famous: 

Oh! let us never, never doubt
What nobody is sure about ! 

Berwick and Chomsky are not fans of certainty. They press at the margins of what we know and what it may be possible for us to know. There is a great deal of research and thought concentrated here and it’s difficult to do justice to it in a short review. Perhaps best then to go to the heart of what they are trying to dismiss in order to point to what, tentatively, they are suggesting. 

Language is exclusively human, in the sense of the Basic Property, “ a finite computational system yielding an infinity of expressions.” In addition, of course, and a necessary condition of the infinity, is the fact that language is syntactically hierarchical. It functions according to “minimal structural” rather than “minimal linear” distance. In other words, the speculations that human language may merely be a refinement of communication systems found in other animals is without serious evidence. The most mellifluous music of songbirds and the best efforts of apes to produce sentences are utterly impoverished compared to the linguistic abilities of a three-year-old child. 

Language appeared during a particular window, long in human terms, but relatively brief in terms of the age of the universe or even life. Modern humans appeared in Africa about 200,000 years ago, language, probably about 80,000 years ago. We can’t be entirely sure. So what was going on for 120,000 years ? Gradual accumulations of small changes which, little by little, in the standard way, resulted over a great span of time in something new ? This is what acceptance of the Darwinian dictum natura non facit saltum (nature does not make leaps) would imply.  There is evidence, however, that nature is not quite so stately in her progress as Darwin concluded. Language looks like a candidate for fairly sudden appearance. The mechanism might be “emergence” defined by Ian Tattersall as: “a chance combination of pre-existing elements results in something totally unexpected.” Thus, accounting for the arrival of language through unmodified Darwinism doesn’t do justice to what looks like its out-of-nowhere arrival.  

The essential computational operation of language is Merge, which selects any two syntactic elements and combines them into a hierarchical structure. This hierarchical structure faces two ways, in a manner of speaking: inward towards thought, outwards towards communication. Common sense, which so often leads us astray, tells us language is a means of communication, but look carefully and it’s hard to conclude that’s what language evolved for. If it had, minimal linear distance would make much more sense as its organizing principle. Minimal structural distance, at least, points to language having evolved as a means of thought. This idea is, of course, as old as Aristotle.  

Berwick and Chomsky are at pains to distinguish language from any faculty possessed by any animal on earth, now or since life began. It is helpful in this effort if it can be argued that it is not a product of natural selection working in its predicted way. They are not saying language isn’t a product of natural selection, rather that natural selection includes the surprise of emergence. It may be that in their urgency to defend language as a unique faculty, they over-egg the pudding of emergence. This is understandable given that language-as-communication is not merely the assumption of common sense, but is also peddled in numerous putatively scholarly papers, even though it would take the average undergraduate about five minutes to work out that  the hierarchical structure of language (which is beyond question) is hard to tally with a system which evolved principally so we can ask the way to Piccadilly Circus. 

The essential argument of this book is in favour of emergence and it is so, it seems, because if emergence is how language arrived, then it is, by that alone, distinguished from almost all other faculties and that is a powerful defence against those who wish to propose language as not different in kind from most other faculties and  principally a means of communication. 

If the theory of emergence is right, it should be able to withstand anything that is thrown at it. Perhaps one criticism could be that it rehearses what Berwick and Chomsky see as Alfred Russell Wallace’s crime: introducing divine intelligence to solve the riddle of language’s appearance. Yet isn’t emergence also somewhat miraculous ? Doesn’t it come close to saying that we just have to accept that in some way we can’t know, in a short time (in universal terms) language just sprang up out of the pre-existing faculties natural selection had formed? This may be right, but it’s noticeable that certain interesting references are absent from the book. 

Principal among the lacunae are mirror neurons. It’s well-known that the Broca’s and Wernicke’s are rich in these. As is the inferior parietal lobule, which may have some clues to offer as to how language arrived. For example, if we think of Merge, we must assume the existence of pre-existing syntactical elements: lexis, in essence. Where do these come from ? As they are what the brain must produce before it can begin assembling the syntactic structures which generate sentences (whether meaningful or otherwise) it seems sensible to posit some hypothesis about their origin. Perhaps Ramachandran’s famous and charming bouba-kiki effect has something to offer. The crucial point about this phenomenon is its illustration of the brain’s intrinsic capacity for abstraction: people associate a cloud-like shape with the sound bouba because their brains are extracting from each the abstract quality of softness. Linked to this, of course, is Saussure’s mistake in assuming the sounds of words are arbitrary, so that if a linguistic community agrees (just what that means is hard to say) that the sound “small” designates large size and the sound “big” absence of dimension, then the agreement will suffice. 

Ramachandran’s experimental works suggests this is mistaken. The sound “small” engages the lips and tongue in small movements, just as the sound “big” does the opposite. This correspondence between the movement of the organs of speech and the meaning of words doesn’t hold for function words, of course; or if it does, it’s very hard to discern. However, it is easy to demonstrate relatively convincingly for words which designate essential qualities. or those with a strong emotional overtone. When we say “come” our lips close and pull inwards. When we say “go” our lips pout.  

Darwin observed than when cutting with a pair of scissors, we may unconsciously clench and unclench our jaw in rhythm with the movement of our hand. The hands and mouth are mapped onto the brain contiguously. It wouldn’t be surprising if content were leeched or lent from one area to the other. Ramachadran calls this synkinesia. The capacity to link concepts across brain maps, which may have evolved slowly through the usual tinkering method of natural selection, may be the clue to how lexis appeared. Grunts and calls in primates originate in the anterior cingulate. In our ancestors the simultaneous production of a manual gesture, some form of essentially emotional utterance and mimicry of what was intended by the organs of speech might have given rise to lexis  through cross-translation between the anterior cingulate and areas controlling the movement of the hands and organs of speech.  

That doesn’t get us very far because language by definition is the hierarchical organization of these essential elements into structures which obey rules of minimal structural distance. That lexis may have evolved without a great leap forward doesn’t imply language did.  
Language, of course, is not a thing but a relationship. What we have evolved with is language competence, but to be engaged, or perhaps expressed would be a better term, this competence requires a social trigger ( an instinct to acquire an art). Part of Berwick and Chomsky’s definition of language is that it is “internal to an individual”. This is a truism but isn’t it also slightly misleading? We know definitively that a child deprived of human society doesn’t acquire language. The Wild Boy of Aveyron didn’t emerge from the woods speaking fluent French peppered with educated use of the subjunctive. He, like all of us, was born with the biological endowment to assimilate language. Universal Grammar was his inheritance; but he hadn’t internalised that variant of it known as French because since being a child who hadn’t yet learned to speak, he’d had no human contact. 

Let’s try a thought experiment: suppose a modern-day Crusoe, shipwrecked at the age of twenty-one. Imagine he spends the next six decades alone on his island with only birds, lizards, insects and one or two man-eating predators for company. Would his rescuers find him still fluent in English ? Would he be making use of English for all those internal processes for which language capacity evolved ? Isn’t it more likely that not only acquisition requires other minds, but maintenance of language capacity too ? Wouldn’t we expect Crusoe’s language and thought to be somewhat impoverished ?  We can’t perform the experiment, but we do know that prolonged and severe isolation destroys the mind. Don’t people who suffer this often display not merely reduced language skills but also confusion of thought, magical ideas, paranoia-like fantasies ?  

Berwick and Chomsky often refer to language capacity as “computational”. This too is a truism, but isn’t there a caveat here  ? If we eliminated every computer in the world bar one, wouldn’t it still be able to compute ? If some catastrophe left only one person on earth, would they still have language, in the essential sense of an internal faculty which facilitates thought ?  Rather, left alone, would their brain maintain that capacity or would it atrophy like an unused muscle ? 

If language evolved to facilitate thought, as the evidence suggests, might the essence of that thought not be mind-reading ? Isn’t the ability to work out, from minimal input, what is going on in someone else’s head, a significant survival advantage ? Not only we can do this, but other primates. A monkey sees its neighbour reach for a peanut and  neurons neighbouring those which fire to make the intention and action possible in the actor, are also activated in the observer. Can it be accidental that the Broca’s and Wernicke’s are rich in mirror neurons ? 

This is not to return to the Saussurean position of language as a “sort of contract”. Saussure didn’t conceive language as a biological endowment, clearly a mistake. Rather it is to suggest that mirror neurons might have evolved to facilitate mutual intuition of motivation through minimal input and by “a chance combination of pre-existing elements” gave rise to the Basic Property.  

That leaves, however, the puzzle of “a chance combination”. Obviously, there is no lack of contingency in the universe.  Ian Tattersall argues, because of the half million years between the development of a vocal tract capable of speech and the emergence of language that: “We have to conclude that the appearance of language and its anatomical correlates was not driven by natural selection.” However, if language evolved as a means of thought rather than communication, why would the development of a vocal tract be crucial ? Language, as an internal property, the Basic Property, facilitating thought, may have existed  before speech. What seems clear is that 80,000 years ago modern humans in Southern Africa showed signs of symbolic behaviour. Perhaps, if Ramachandran’s speculation about the origin of lexis has some validity, a basic repertoire of words might have kick-started the process of abstraction by which they were attributed to categories to be selected from and combined. Employed primarily as a facilitator of thought, the rapidly developing capacity of Merge might have had a head start on the secondary function of communication, which could, nevertheless, have emerged failure quickly in evolutionary terms.  

To return to Ramachandran’s bouba-kiki effect: it reveals a capacity for abstraction which depends on the inferior parietal lobule. The upper part of the IPL, the supramarginal gyrus, if damaged, produces apraxia (the inability to mime simple actions because the capacity to generate an internal picture which can fire the appropriate motor neurons is missing). Apraxia appears to be mediated by a failure of mirror neurons. How did the IPL evolve as quickly as it seems to have done ? Ramachadran speculates that “the need to achieve an exquisitely refined, fine-grained interaction between vision and muscle and joint position while negotiating branches on tree tops” might have been the driver. What has this to do with the question of how the Basic Property arose ? The accidental splitting of the IPL (probably by gene duplication) created the angular gyrus which enhanced the capacity for abstraction. Isn’t abstraction the essence of how the Basic Property works ? Merge permits stored elements to be selected and combined, in ever more hierarchical patterns, to generate meaning (whatever that is). The selected items are not stored randomly but by category (verbs, nouns, adverbs, prepositions etc). In other words, prior to Merge a process of abstraction has already taken place.  

We know also that Broca’s or Wernicke’s aphasia result in different effects: in the former, lack of syntax but semantic coherence; in the latter, competent syntax but gobbledygook. This separateness is not so complete as it first appears, but leaving that aside, it’s clear syntax does not depend on semantics, as Chomsky’s famous “Colorless green dreams sleep furiously” demonstrates. Nevertheless, syntactic structure and semantic coherence tend to combine. The Broca’s may be more or less a “syntax box” but mostly it operates in conjunction with the Wernicke’s to produce meaning. Meaning is the capacity to abstract. Even the simplest operation of Merge (say read/book) depends on the ability to abstract from a potentially infinite array of nouns and verbs.  

Is it the case that the Broca’s is a “syntax box” which evolved separately from any other part of the brain ? Is Stephen Jay Gould on the right path in suggesting that language isn’t “a specialized mechanism based on brain modules” but rather the “specific implementation of a more general mechanism which evolved for  other reasons, namely thinking” ? Ramachandran thinks Gould is wrong in proposing that thinking evolved first. Can we disentangle thought from language ? The famous experiment with female capuchin moneys indicates we can. Deprived of the better reward for the same task as their neighbours the less favoured monkeys effectively go on strike, refuse the task and protest. They certainly don’t have the Basic Property but what must be going on their brains is at least a recognition that they have been deprived and that a refusal to behave as expected is the right way to respond. It is, of course, perilous to use language to speculate on what is going on in a monkey’s brain, but this, and many more examples, point to a rudimentary form of thought.  

If capuchin monkeys, dogs and rabbits can think, wouldn’t our ancestors have been able to do so before they had the Basic Property? Language didn’t make thought possible, it made it highly abstract. In a sense, language is metaphor. Perhaps when we ask, How did language evolve ? we are asking How did the capacity for abstraction evolve ? Perhaps Ramachandran’s speculation about the IPL from which the angular gyrus was born will prove fruitful.  

Further, his speculation that “an area very close to what we now call the Broca’s area originally evolved in tandem with the IPL…There was a subsequent  duplication  of this ancestral area , and one of the  new substrates  became further  specialized  for syntactic structure…in other words, it became Broca’s area” may offer a basis for greater clarity.  

Berwick and Chomsky pose two fundamental questions, among others: why did language capacity evolve and why are there so many languages ? Is the question of why the capacity for abstraction evolved easier to answer than why language evolved? The IPL may help us understand how the capacity for abstraction appeared (not, it seems, as a result of a mutation in a single individual). Once it was in place, it would have quickly spread because of its survival advantage. Thought is not abstraction. The latter is the ability for open-ended symbol manipulation, the capacity which makes syntactic structures possible. Might the Basic Property be a specific, unique variant of a generic capability ? 

Language may have evolved rapidly from a more general capacity for abstraction because of its metaphorical power: the world is inordinately more comprehensible through metaphor than through non-metaphorical thinking. To produce as internal language the word “dog” is to select from a potentially infinite array of possibilities which itself is made possible by the ability to assign linguistic items to abstract categories. In addition, of course, this one, simple item can have a multiplicity (theoretically an infinity) of abstract associations.  

Why are there so many languages ? Because if there weren’t there wouldn’t be any. UG provides the basis for a multiplicity of routes to meaning. The differences are superficial. Whether your brain selects “dog”, “chien”, “Hund” or “perro” is neither here nor there. It’s the capacity to make the selection from abstract categories and to engage Merge which matters. So long as the rules of syntax, rules embedded in our biological inheritance, were adhered to, there was no imperative for the brain to evolve neural pathways to close down on any choice but, say, “dog”. 

Language is ours and only ours, but we didn’t create it, any more than we created our hearts or livers. Children assimilate it effortlessly and unconsciously. We engage Merge in our thinking all day long, but we don’t do so wilfully. Among the characteristics which mark us a human, language is arguably prime. 

This ought to give us pause. Vital to point out that what follows is not discussed in this book at all, but language capacity makes nonsense of theories such as IQ. Its proponents, and there are many in spite of it being utterly intellectually discredited, claim it is an inherited, testable quantity unamenable to education. Stephen Jay Gould in The Mismeasure of Man, makes clear just how much cruelty has been perpetrated and excused on the back of IQ theory. The 11-plus in Britain, pioneered by Cyril Burt in London, was founded on the acceptance of the theory and on Burt’s fraudulence. Millions of youngsters were tested for the putative quantity, millions found wanting and sent to second-rate, underfunded, over-crowded schools where, surprise of surprises, their academic failure confirmed the theory. In America, Dr Prosser’s destructive prescriptions for the education system were similarly reliant on IQ theory. IQ theory divides humanity: some are born with a big lump, others with a lesser. It follows, of course, that the former should be granted power and wealth, while the latter should have little and do as they’re told.  

The theory of UG unites us. We have a common biological inheritance and the differences between us are superficial and trivial. This is not to argue against individual uniqueness or freedom; on the contrary. As our unity is guaranteed by biology we have no reason to fear individuals enjoying the greatest degree of liberty to live as they choose. Hence Rabelais’ s injunction: Do what you like . 

Paradoxically, the proponents of the so-called freedom of the individual are commonly authoritarians who believe in the right of the State to close down on some group or other: blacks, Jews, homosexuals, leftists, artists, Muslims, doctors who perform abortions. The ideology of the freedom of the individual is usually the outgrowth of the belief that the differences between us are fundamental: blacks are just not like whites, Jews are just not like Christians, homosexuals are ill, Muslims are beyond the pale.  

No culture explicitly upholds the negative capacities of our endowment. They must always be insinuated behind an excusing ideology.  Hitler had to make the elevation of the “master race” the excuse for war and holocaust, Stalin had to conceal his viciousness behind the liberation of the proletariat; Isis must sugar its sadism with claims of adherence to high, moral, religiously derived principles. Our shared humanity is as inescapable and indefeasible as UG. Hence the need to excuse what offends against it. 

Injustice is always founded on division: white against black, lord against peasant, bosses against workers, male against female. Injustice is always a denial of our common inheritance. We are equals in UG. Nor does the superior linguistic capacity of Shakespeare or Whitman undermine that. The difference between them and a child who struggles to read is superficial by comparison to the unity of UG. Injustice works be elaborating, on the basis of inconsequential distinctions, institutions which dehumanize and humiliate. It rests, if you like, on Freud’s “narcissism of small differences”. The putative great differences on which injustice rests are in fact trivial. UG is one of the facts of our nature which makes this obvious. 

This book is a lucid defence of the theory of UG. Many questions remain to be explored and may never be answered. We understand little about ourselves. A good reason to set aside the arrogance of power.