International Journal of Pediatric Otorhinolaryngology (2003) 67S1, S131—S135

Why did language develop?

J.F. Stein

University Laboratory of Physiology, Parks Road, Oxford OXI 3PT, UK

KEYWORDSGesture;Mirror neurones;Hemisphericspecialisation;Imitation;Chomsky;Vocalisations

Summary Language developed for communication, to facilitate learning the use oftools and weapons, to plan hunting and defence, to develop a ‘‘theory of mind’’ andthe tools of thought, and to attract and keep a mate. The adaptations required tookplace over many millions of years. The first important one was left-sided speciali-sation of the neural apparatus controlling involuntary emotional vocalisations thatbegan more than 200 million years ago. The next was the development in primates of‘‘mirror neurones’’ in the pre-motor cortex some 45 million years ago. These enabledthe imitation and voluntary control of previously involuntary manual gestures and vo-calisations. The third important adaptation was the descent of the larynx, 100,000years ago, which greatly increased the phonological range of vocalisations that couldbe made. Thus, language did not develop all at once as suggested by Chomsky, butevolved gradually building upon adaptations originally meeting quite different needs.© 2003 British Association for Paediatric Otorhinolaryngology. Published by ElsevierIreland Ltd. All rights reserved.

To ask why language developed might seem avery odd question; most people would say thatthe answer is blindingly obvious. Clearly, humansevolved language in order to be able to communi-cate with each other. Thinking about it long afterthe event, that seems trivially obvious to us. Butwe have to ask why more efficient means of com-munication than already existed in the form offacial expressions, gestures and vocalisations, wasso selectively advantageous to humans. Actuallylanguage seems to have evolved relatively recently,perhaps only 50,000 years ago. So, despite its ob-viousness to us, development of language commu-nication was not that overwhelmingly essential. Inthis paper I want to show that language and thenliteracy only evolved gradually, long after manyother things were in place. Contrary to Chomsky’ssuggestion [2], recent opinion is that language wasnot the result of a sudden mutation endowing uswith a new ‘‘neurologically encapsulated linguis-tic processor;’’ instead, it evolved gradually from

E-mail address: john.stein@physiol.ox.ac.uk (J.F. Stein).

anatomical and physiological adaptations that tookplace long before. The main reason, I believe, thisto be important is that it means that we can hopeto understand impairments in the development oflanguage and literacy, in terms of the basic biolog-ical processes that underlie them. Ontology doesto some extent mimic phylogeny, and understand-ing how basic motor, auditory and visual processesled to the development of language and literacycan help us to elucidate how they go wrong inconditions such as developmental dysphasia anddyslexia.

To understand how language evolved, we need toconsider not only the selection pressures that oper-ated to advantage it, but also the genetic variationsthat occurred by mutation that made it possible.First, therefore, I will provide a very brief overviewof ideas on why there were advantages to improvingcommunication by the development of language,and then an equally brief review of the evolutionof Homo Sapiens. Then, I will describe the evolu-tionary adaptations that were essential for the de-velopment of language before detailing how thesemay have mediated it.

0165-5876/$ — see front matter © 2003 British Association for Paediatric Otorhinolaryngology. Published by Elsevier Ireland Ltd. All rights reserved.doi:10.1016/j.ijporl.2003.08.011

S132 J.F. Stein

There are three strains of theory about whatkind of selective advantage linguistic communica-tion provided for hominids [1]. The first suggestionis that it enabled the shared use of tools, i.e. thedevelopment of technology. In this scenario, thedevelopment of language was required to explainhow tools should be used and to discuss the bestways of employing them. If these tools happenedto be weapons, then they could be used for huntingeither animals for food or indeed, enemy Nean-derthals. Sad to say, it seems that our ancestorssimply annihilated their main competitors.

It has been pointed out that chattering awaywould not lead to a very successful hunt; but, ofcourse, the conversations would have taken placein order to plan the hunt. The requirement for suchplanning gives rise to the second main suggestionabout the pressure to develop language, namelythat language enables you to discuss what is likelyto happen by predicting and representing it in ab-stract terms, i.e. language gives you the tools tothink about it.

This representation need not be confined to plan-ning a hunt, but could also be used as the basisof thinking about all manner of things. Perhaps themost important of these would be helping to deter-mine what others are thinking, developing a ‘‘the-ory of mind.’’ Cynics believe this to be the basisof ‘‘Machiavellian intelligence,’’ working out whatpeople are likely to be thinking, thus being in a bet-ter position to deceive them and gain advantage.But, I prefer to emphasise how developing a the-ory of mind would naturally lead to the inventionof the tools of thought: categorisation, quantifica-tion, abstraction, causality and logic.

The third kind of pressure that might havefavoured the development of language was the re-quirement to find a mate. This sexual pressure iscertainly the main driving force behind bird song.In fact, the left-sided brain structures that controlsong production by the syrynx actually increase insize during the breeding season; they grow newneurones under the influence of testosterone, andthen during the winter, lose these neurones andregress. In addition, because of the long childhoodof humans, females have the need to retain theirmate for several years to help bring up the chil-dren. This is, therefore, the only account that canexplain why females are in general slightly betterat communication than males.

These three suggestions are not mutually exclu-sive; probably all three are partly correct. But theyare almost impossible to confirm or refute. Hence,they are to a certain extent like Kipling’s Just SoStories: how the rhinoceros got his skin might havebeen because the Parsee put itchy raisins into it

when he’d taken it off to swim; but that is very un-likely. Our stories about the selective advantage ofcommunication by speech sound much more plau-sible, but since unprovable they might be just astotally off beam.

We must now consider the fossil record indicat-ing how Humans evolved, incomplete as it is. Mam-mals first appeared on the earth about 200 millionyears ago; primates around 65 million; the first An-thropoids about 45 million and Chimpanzees some12 million years ago. The first Hominoid, Ankarap-ithecus, split from the Chimp line about 9 millionyears ago, Australopithecus appeared around 3 mil-lion years ago, H. erectus at 2 million, Neanderthals250,000 and Homo Sapiens 200,000 years ago.

However, the important features of this historyfrom the point of view of language developmentwere that the left side of the brain became spe-cialised for the production of sounds many millionsof years ago. Then bi-pedalism developed in thefirst phase of hominoid evolution, and this freed thearms for more efficient food gathering, tool use andgesturing. Descent of the larynx then made the voic-ing of phonemes possible to increase their number,and finally, almost in modern time, the develop-ment of consistent right-handedness seems to havebeen important for the invention of writing.

The origins of left-sided cerebral specialisationcan actually be discerned from almost the very be-ginning of life in that the laevo-isomers of biolog-ical molecules are always favoured. The strongestasymmetries favouring the left are certainly foundin birds, but they are not seen so strongly again untilwe reach Homo Sapiens. It is surely no accident thatthe two groups that make the most use of acousticcommunication site the control system on the left.

However, walking on two legs only commencedwith Homo erectus, just 2 million years ago, un-less you count Kangaroos that evolved in Australiaseparately from the mammals. The third impor-tant adaptation that enabled the development oftrue speech was the descent of the larynx, but thisonly occurred about 50,000 years ago. Even thoughother evidence suggests that true speech did notevolve until much later, perhaps only 35,000 yearsago, descent of the larynx and the wider variety ofsounds that this enabled must have been sufficientlyadvantageous to outweigh the danger of inhalingfood.

The order in which these adaptations appearedis important because it sketches out the basis ofhow the capacity for speech may have developedinnately. For a long time there have been argu-ments about whether language is innate or learnt.In 7000 b.c., the Egyptian pharaoh Tsammtemichusis said to have had a child brought up in isolation

Why did language develop? S133

in a cage to find out whether he would speak with-out any teaching. His first word was the Phrygianword for bread; which he had clearly learnt from hisguards, so the experiment suggested that languagewas learnt. In Agra, in the 17th Century, a.d., MogulAkbar Khan ensured strict silence in his brutal ver-sion of the experiment by employing dumb nursesto rear 12 children in isolation. He found to his sur-prise that none of the children learned to speak atall, again implying that language had to be learntby example. In the 1960s, Genie was found in LosAngeles completely unable to talk because she hadbeen protected by her over religious parents fromalmost all sensory input that might have contami-nated her with evil, by being locked up in a darkcupboard. These and a host of other evidence haveshown clearly that language has to be learnt. TheEnglish learn English from their parents; the Frenchlearn French.

On the other hand, Johann, who had been aban-doned as a baby in Burundi, was brought up by chim-panzees and when found at the age of 5 was usingchimpanzee vocalisations and gestures to commu-nicate. This shows that our evolutionary past equipsus to learn languages but does not provide languageitself. Nor does it specify the language we learn.In Stephen Pinker’s words, our genome gives us an‘‘instinct’’ to communicate, but we have to learnthe means to express it.

What we now have to do is to sketch out the waysin which this came about [3]. Our starting point willbe the vocalisations that almost all mammals pro-duce. As in birds, these are controlled by a networkof neurones that shows a propensity to favour theleft-hand side. Why it is the left side that is chosenwe do not know, and even why one side is favouredis not entirely clear. Most features of animals arebilaterally symmetrical and perfectly satisfactorilycontrolled from both sides of a bilaterally symmet-rical brain because there are good cross communi-cations between the two sides.

Nevertheless, in about 95% of humans (includ-ing 70% of left-handers), parts of the left cerebralhemisphere are specialised for mediating the per-ception and production of speech. Probably one sideis chosen because if the two hemispheres both tryto control the same structure, they tend to com-pete with each other; hence, the majority of thecross connections between the hemispheres havebeen found to be inhibitory in order to prevent thetwo sides trying to do the same thing. Placing thecontrol of vocalisations predominately in the samehemisphere thus simplifies the control problem. An-other contributory factor may be that the lengthof axons joining sensory and motor language areaswithin one hemisphere would be slightly shorter

than requiring connections between the two hemi-spheres, thus cutting down on delays in a systemthat requires millisecond accuracy.

In lower mammals, including monkeys, vocalisa-tions are controlled by a medially placed system ofneurones that involves the cingulate cortex, basalganglia and hypothalamus. The composition of thissystem suggests strongly that it is primarily con-cerned with the expression of emotions, and it isnow clear that in primates other than man all vo-calisations are automatic, driven by the emotions.Thus, all attempts to teach primates to actuallytalk have failed; it is impossible to harness primatevocalisations for other kinds of communicationbecause they are not under the animals’ volun-tary control. Chimpanzees are actually intelligentenough to be taught to communicate to some ex-tent using their extensive repertoire of voluntarygestures, but never by vocalisations.

It is highly significant that this emotionally con-trolled medial system does not involve the monkeyhomologue of Broca’s speech area in the left lat-eral frontal cortex. Indeed, lesions in this area donot affect the ability of monkeys to make their vo-calisations at all. The development of the humanmotor speech area has followed a different course.This has been powerfully illuminated by the discov-ery of ‘‘mirror’’ neurones [5]. These unequivocallyplace the development of human language in theprovince of gesture and facial expression.

Mirror neurones are found in the ventrolateralfrontal lobe just in front of the face and arm repre-sentation in the primary motor cortex. Their impor-tant characteristic is that they fire not only whena monkey reaches out to grasp an object, but alsowhen the monkey observes somebody else doing thesame thing; however, not when the same goal isachieved in a different way, for instance using a pairof pincers. In humans, likewise, this area appearsto be activated both when the subject reaches tograsp something, but also when he imagines doing itand also when he sees somebody else doing it. Thus,mirror neurones could underlie how we learn to pro-duce speech by enabling us to imitate our parents’speech. In addition, they offer unexpected supportfor Lieberman’s motor theory of speech perception[4]. Mirror neurones would enable us to interpretspeech because the very same cells would be acti-vated by observing speech as those that we wouldemploy to make the same speech sounds.

In addition, Corballis and others have arguedvery convincingly that these mirror neurones showthat speech evolved from gesture and not fromvocalisations [3]. The argument runs as follows:by representing a particular gesture, mirror neu-rones enable other people to imitate it and thus to

S134 J.F. Stein

communicate by means of these gestures. Lieber-man, Studdert-Kennedy and their colleagues at theHaskins laboratory of speech science have long ar-gued that the basic elements of speech are not,as generally assumed, consonants and vowels, butrather the vocal gestures that generate them,namely the movements of the lips, tongue andlarynx [4]. Thus, mirror neurones in Broca’s areacould come to represent lip, tongue and laryngealgestures, and these could generate the phoneticelements of speech. In fact, most of us still gesturewith our hands when speaking and when we gestureour vocal production is synchronised with our handgestures. These voluntary movements of the vocaltract are mediated by this lateral system, and theyhave been superimposed on the older emotionalsystem for automatic vocalisations mediated bymore medial brain structures. The latter still supplythe basic intonation and prosody for sentences.

This account of the development of speech fromgesture has received further support from the re-cent discovery that the gestures of sign languageused by the deaf are controlled by the same lefthemisphere centres, particularly Broca’s area, thatspeech occupies in those who can hear. One caneven see in the way in which the order of ideas andsyntax is expressed in the trajectory of a sign lan-guage gesture, how language syntax and grammarmay also have been founded in the way that thestructure of a signed sentence is determined by theevolution of a gesture from shoulder to fingers.

The final step in this history is to consider the in-vention of writing. This is truly a cultural invention,and probably not at all enshrined in our genomebecause it was only invented about 5000 years agoand was not common until the last century. Be-ing able to read and write carried no particularselective advantage; and, it is therefore most un-likely that we would ever find a gene or genes forreading, contrary to what is sometimes claimed.But, like speech before it, writing depends on prioradaptations, in particular the development of ar-ticulatory gestures controlled from Broca’s area.In addition, its invention probably depended onthe development of right-handedness. Despite thechoice in most animals of left-brain structures forthe control of both automatic emotional vocalisa-tions and voluntary speech, no lower animal showssuch strong right-handedness as humans do. Manyanimals choose to use either left or right hands forparticular tasks, but not even chimpanzees choosethe right so systematically.

Probably the main reason why right-handednesswas so important for the invention of writing is thathieroglyphs and letters are very impoverished vi-sual signals, and it matters greatly whether they

are pointing to the left or to the right. It is easiestfor us to agree which way round they should be ifwe all write from the same side. Ambidextrous ani-mals and children therefore have extreme difficultyknowing which way round bs and ds should go.

This is not to say that right-handedness onlyevolved 5000 years ago. Many of the Stone Agehand axes from 100,000 years ago show signs ofhaving been shaped for right-handers. Likewise thebeautiful cave paintings of Lascaux and Altamiraflowered at about the time that speech evolved;but most of the people depicted seem to havebeen right-handed. Like speech, therefore, writinghas clearly piggybacked on an adaptation that oc-curred a lot earlier for a different purpose, perhapsgesture.

Reading and writing are much more difficult tolearn than speaking is because the written worddoes not map so easily onto articulatory gestures asspeech does. As Lieberman and Studdert-Kennedypoint out, the phonemes that are represented byletters are artificial subdivisions of the articulatorygestures that generate them, and these subdivisionshave to be taught and learnt. Hence, a very largeproportion (some 10—20%) of humans never masterthis art properly; it is the most difficult thing thatmost of us ever have to learn.

Let us now return to Chomsky [2]. We can nowsee that his idea that the human acquisition of lan-guage and literacy resulted from a single mutationthat endowed us with a linguistic processor, all inone jump, is clearly inconsistent with recent dis-coveries. Speech and language evolved gradually,coat tailing on a series of adaptations that evolvedfor completely different purposes and occurred mil-lions of years earlier.

Nevertheless, Chomsky was not entirely wrong.I’ve always been a great admirer of him, andhis insights concerning phonology, deep grammarand syntax come out of our current concepts ofthe development of language rather well. Thepre-eminence of phonology follows from the de-scent of the larynx and the specialisation of Broca’sarea for the voluntary control of articulatory ges-tures. Syntax and grammar can be viewed as adirect consequence of the evolution of sentencesout of gestures that can involve the whole bodyfrom axial back muscles to distal finger-movers.In rather the same way that identical Chinese lo-gographs can represent totally different soundingwords in Chinese and Japanese, so the same deepstructure, the same flow of ideas conveyed by aparticular gesture, could be represented by differ-ent strings of articulatory gestures producing thethousands of different languages that have devel-oped throughout the world.

Why did language develop? S135

What makes this whole enterprise more than justcuriosity is that it provides new insights into whatcan go wrong with language. Because ontogeny re-peats phylogeny to some extent, elucidating howlanguage gradually evolved from archaic gesturesmeans that potentially we now have a new approachto understanding developmental disorders of lan-guage. Counter-intuitively, both speech productionand comprehension seem to depend on our mir-ror neurones being able to represent articulatorygestures. Hence, although comprehension clearlymakes greater demands on the auditory system,and speech production makes greater demands onBroca’s area and the motor vocalisation system,Broca’s area is engaged in both, and this is whatmodern imaging methods have shown clearly. Asexpected therefore, children with developmentaldysphasia are significantly worse at deciphering ar-ticulatory gestures, as in lipreading. Also, whereas

good speakers’ hearing of a phoneme is greatly al-tered if the speaker’s lips appear to be generatinga different one (the McGurk effect) (in develop-mental dysphasia, the mishearing of the phonemes)is much less apparent, because in them the mirrorsystem is not working as well as it ought to be.

References

[1] J.L. Bradshaw, Human Evolution, Psychology Press, Hove,1997.

[2] N. Chomsky, Reflections on Language, Pantheon, NY, 1975.[3] M.C. Corballis, From mouth to hand, Behav. Brain Sci., in

press.[4] A. Liberman, D. Shankweiler, M. Studdert-Kennedy, Percep-

tion of the speech code, Psych. Rev. 74 (1967) 431—461.[5] G. Rizzolatti, L. Fadiga, V. Gallese, L. Fogassi, Premotor

cortex and the recognition of motor actions, Cogn. BrainRes. 3 (1996) 131—141.