Haskins Laboratories Status Report on Speech Research1990, SR-101 /102, 14-27

Language Development from an Evolutionary Perspective*

Michael Studdert-Kennedyt

EVOLUTION AND DEVELOPMENT

PreliminaryIn his famous course of lectures at the

University of Geneva (1906-1911), de Saussuredistinguished langue, language as a system, acultural institution, from parole, language asspoken and heard by individuals: "...language isnot complete in any speaker; it exists only withina collectivity...only by virtue of a sort of contractsigned by members ofacommunity" (de Saussure,1966, p. 14). Language was thus seen as anabstract property of a group, related to its variableindividual speakers somewhat as a species is to itsvariable individual members.

Nineteenth century attempts to apply evolu­tionary principles to language did indeed vieweach language as a species. For example, Darwinwrote: "The formation of different languages andof distinct species...[is] curiously paralleI...We findin distinct languages striking homologies due tocommunity of descent, and analogies due to a sim~

ilar process of formation" (1871, pp. 465-466).Here, Darwin was following in the steps of AugustSchleicher (1821-1868), who drew on the earlierwork of Darwin himself and on the vast scholar­ship of 19th century European philology, to con­struct a taxonomy of Indo-European languages.Later work has continued to draw on Darwinianprinciples to construct evolutionary trees of lan­guage families (Lehmann, 1973).

However, a strict analogy between languagesand species is untenable for both linguistic andbiological reasons. For the historical linguist themost obvious difficulty is that a biological speciesis a reproductively isolated population: propertiesof one species cannot pass to another. Yetlanguages in contact clearly do influence oneanother, even after they have become discretesociopolitical entities. Often we cannot then knowhow far shared properties have arisen fromcontact, how far from common descent.

14

A more serious difficulty follows from twocentral tenets of neo-Darwinian theory (themodern synthesis of Mendelian genetics withDarwinian natural selection). First, in theevolution of a species, the principal unit ofvariation and selection is not the species, but theindividual organism. Individuals within a speciesdiffer in the number of offspring they produce;genes whose expression increases the relativenumber of an individual's offspring will increasetheir own relative number in the species'gene pool. The characteristics of a species aretherefore determined by competition amongindividuals of that species. And, by analogy, thecharacteristics of a language are formed bycompetition among speakers of that language. Sofar, so good.

The difficulty arises when we combine theprinciple of individual selection with the so-canedcentral dogma of modem biology, the "Weismannbarrier," insulating germ cells from body cells:genes alter body cells, but body cells cannot altergenes. In other words, biological evolution doesnot proceed by the transmission of acquiredcharacters across generations, and this isprecisely what an evolutionary model of languagechange requires. We must therefore distinguishthe cultural, or Lamarckian, evolution oflanguage, a concern of historical linguistics, fromits biological, or neo-Darwinian, evolution, aconcern of developmental biology.

This distinction was first clearly formulated byChomsky (1965, 1986), who recognized that deSaussure's definition of language as a property orproduct of a social group did not lend itself totreatment in biological evolutionary terms.Chomsky made the necessary move byreformulating the langue-parole distinction ascompetence (what a speaker-listener knows) andperformance (what a spe"aker-listener does inimplementing knowledge). He thus set the locus ofhuman language capacity in the individual's

Language Development from an Evolutionary Perspective 15

mindlbrain. He also set the stage for the modernstudy oflanguage acquisition.

Form and functionNonetheless, the competence-performance dis­

tinction as usually formulated also does not lenditself to evolutionary treatment. Competence isknowledge of a particular language, formed by in­teraction between an innate schema, the"universal grammar," and the grammar of thelanguage being learned. Universal grammar is"...a theory of the 'initial state' of the languagefaculty, prior to any linguistic experience"(Chomsky, 1986, pp. 3-4). Performance is thussaid to be a product of a partially innate compe­tence. In what follows, I shall argue that this pre­cisely reverses the true course of development.Universal grammar is not a prescription, or pro­gram, for development, but a partial and a poste­riori description of the phenotypic product of thedevelopmental system (cf. Oyama, 1985).Universal grammar is a consequence, not acondition of development.

What is at stake here is the relation betweenform (competence) and function (performance).Zoologists have traditionally stressed the harmo­nious match between form and function, asexpressed in an animal's mode of life, but theyhave disagreed on how the match comes about.From classical times down into the 19th century,the standard belief was that species and generahad fixed, unchanging forms or essences:departures from the species prototype were"unreal," and structure took precedence overfunction. For example, according to Mayr (1982),Georges Cuvier (1769-1832), the great Frenchzoologist, held that "...structure has primacy overfunction and habit, and ...only a change instructure might necessitate a change in function"(p. 367). Certain non-Darwinian French zoologistsstill hold such views: "...evolution originates inparent forms; if these are absent, new types oforganization never appear" (Grasse, 1977, p. 75).Thus, Chomsky and his structuralist forebears inlinguistics align themselves with the essentialisttradition in biology by asserting the primacy ofform over function.

For Darwin and modern evolutionary biologists,working within the British empiricist tradition,the form-function relation is reversed. Species arenot eternally fixed. A species is a geneticallyvariable population of individuals, adapted to aparticular ecological niche and thereby reproduc­tively isolated. No aspect of an animal's structure

determines a unique function. Rather, a structuredetermines an unbounded range of functions, tosome of which it is more nicely tailored thanothers. (We can use a screwdriver to drive screws,or as a dagger, a lever, a drumstick, a fork... ). Ifcertain members of a species are forced bycompetition with their fellows, or by an environ­mental change, into a new mode of life, a newhabitat, they may call on some hitherto unusedpotential function of their structure. The newmode of life then confers a reproductive advantageon those individuals whose structure is marginallybetter suited (due to small differences in theirgenetic make-up) to the new mode. Little by little,as with the beaks of Darwin's famous finches(Lack, 1947), the new selection pressures reshapethe old structure to its new function. Behavior isthus the great "pacemaker of evolutionary change"(Mayr, 1982, p. 612), the cause not the conse­quence of speciation and of species form.

In short, a commitment to gradual evolution bynatural selection entails a commitment to theprimacy of function over form. My central as­sumption, then, is that language competence (theneural substrate of language form) was shaped inphylogeny, and is still shaped in ontogeny, by lan­guage performance (function, behavior).

Behavior as the pacemaker ofdevelopment

If we extend Mayr's dictum on evolution todevelopment, we see development in a new light.We are freed from preoccupation with the "initialstate" as an index ofgenetic endowment, and fromthe nativist-empiricist controversy that hasdominated studies of language developmentalmost since their inception.

At birth, an organism suffers an abrupt changein the quality and quantity of environmentalconditions that may affect its growth. But thediscontinuity does not change the developmentalsystem. Gene action does not cease at birth, nordoes regulation of gene action by the cellularenvironment. Rather, the developmental circlewidens. The new external environment affords abroader range of stimulation, eliciting a broaderrange of response. Changes in the organism'senvironment and behavior now mediate, to asteadily increasing degree, changes in the cellularenvironment that controls genetic action.

The developmental course is not one of simplematuration (cf. Borer & Wexler, 1987). Languagedevelopment will not go forward merely becausethe environment meets general conditions of

16 Studdert-Kennedy

survival-air, food, other people, and so on.Specifically linguistic input and, probably, outputare called for so that if either is set to zero,development stops. Specific extralinguisticprocesses of perceptuomotor and cognitivedevelopment-a growing grasp on the physicaland social modes of being and acting thatlanguage represents-must also feed into thelanguage system. Thus, the proper study oflanguage growth is the sequence ofbehavioral andcognitive conditions, linguistic and extralinguistic,that precipitate language change. The task of this"post-natal embryology," as we might call it, is tochart the course by which perceptual and motoricfunctions induce structure, from undifferentiatedinfant performance to differentiated adultcompetence.

In short, I am proposing that principles ofgrowth, generally accepted in the development ofthe peripheral anatomy, also apply to the centralnervous system. For example, writing of theplasticity of growing bones, D'Arcy Thomps.on(1917/1961) remarks "... the very importantphysiological truth that a condition of strain, theresult of a stress, is a direct stimulus to growthitself. This indeed is no less than one of thecardinal facts of theoretical biology. The soles ofour boots wear thin, but the soles of our feet growthick the more we walk upon them: for it wouldseem that the living cells are 'stimulated' bypressure, or by what we call 'exercise,' to increaseand multiply" (p. 238, italics in the original).Bosma (1975) invokes this principle of exercise indescribing the development of the vocalapparatus. Perhaps the development of the neuralrepresentations of speech and language is also aninstance of .....one of the cardinal facts oftheoretical biology."

Ontogeny <sometimes> parallelsphylogeny

Complex functions, and the physical structuresthat support them, arise in evolution by gradualdifferentiation from simpler forms. Everyevolutionary change is a change in developmentthat is preserved in later generations.Accordingly, when a complex function is anevolutionarily coherent, hierarchically developedsystem, ontogeny may parallel, or recapitulate,phylogeny. I should emphasize that I am notproposing to reinstate the discredited "biogeneticlaw" of Ernst Haeckel (1834-1919). I am merelydrawing attention to developmental facts well­known since the embryological studies of Karl

Ernst von Baer (1792-1876) in the early 19thcentury (Gould, 1977).

Of course, if we lack, as we largely do forlanguage, precursor forms that confirm thesequence, we cannot be sure that the order ofdevelopment we now observe was the actual orderof evolution. Such precursors as we mayestablish-for example, lateralized systems forneural control of body posture and manualfunction (MacNeilage, in press), or certaincapacities for symbolic representation (Savage­Rumbaugh, in press)-do not help us in thepresent context. Some "stages" may have beeninserted into the evolutionary sequence later thanothers that now follow them in development. Thismay be true, for example, of left hemispheresensitivity to speech at, or soon after, birthbecause it is unlikely that a specialized neuralsubstrate for speech evolved before speech itself.

Yet other processes may have evolved indepen­dently, parallel to processes with which they werelater integrated. For example, some form of seg­mental phonology, affording at least a modest lex­icon, would seem necessarily to have evolved be­fore syntax began to take shape, and we still ob­serve this sequence in development. But laterstages of phonology and syntax may have evolved,as they still seem to develop, more or less inde­pendently. Similarly, prosodic variations in pitch,amplitude and duration, characteristic of bothhuman and non-human systems of communica­tion,perhaps first followed an independent courseof evolution, to be modified and integrated into thelinguistic system only as longer utterances andmore finely differentiated syntactic functionsemerged. The double dissociation of right and lefthemispheres for emotional and linguistic uses ofprosody may reflect such a course of evolution.

Yet further limits on a phylogenetic inter­pretation of language ontogeny may seem to arisebecause the child is born into a community ofcompanions who already speak a language.Moreover, the child is born with a vocal tract anda pair of hands that soon mature into formsadaptable for spoken or signed language, and witha rich, plastic neural substrate fit for shaping bycognitive and linguistic function. The exogenousand endogenous conditions of developmenttherefore differ radically from those that musthave prevailed in a hominid community wherelanguage was not yet fully formed.

Nonetheless, language acquisition is notinstantaneous. The child does not have immediateaccess to the full adult language that surrounds it.

Language Development from an Evolutionary Perspective 17

The effective linguistic environment changes, stepby step, as the child comes into possession of newcognitive and linguistic capacities. The problem ofhow linguistic input is ordered so as to ensurecoherent development is solved by the child's ownincreasingly differentiated linguistic attention.With each new step the child finds the next stepwaiting, as it were, in the adult language becausethe adult language is adapted to the child no lessthan the child to the language. The reason for thisis simply that language, like every evolved form,is the product of successive ontogenies, itsstructure a record of its own evolution (cf. Locke,1983). Looked at in this way, language is not anobject, or even a skill, that lies outside the childand has somehow to be acquired or internalized.Rather, it is a mode of action into which the childgrows because the mode is implicit in the humandevelopmental system.

We conclude· that language, as a complex,hierarchical, behavioral structure with a lengthycourse of development, is a good candidate for(circumspect) study in a recapitulatory frameworkbecause its development is rich in sequentialdependencies: syllables and formulaic phrasesbefore phonemes and features (as I shall arguebelow), holophrases before words, words beforesimple sentences, simple sentences before lexicalcategories, lexical categories before complexsentences, and so on. Thus, if we assume that eachof the subsystems, phonology and syntax, evolvedhierarchically by repeated cycles of differentiationand integration, we may recover their course ofevolution by tracing the course of their growth.The general heuristic value of the assumption isthat it not only throws light on evolution, but alsopromises an understanding of development infunctional terms.

Development is not teleologicalIf we combine the principles of gradual evolution

and of a limited, functional recapitulation, we arefreed from the temptation to assign purpose todevelopment. At each point in its development anorganism is already complete, adapted andadapting, as best it can, to present conditions,internal and external. Just as earlier evolutionaryforms existed for themselves, not for any laterforms to which they might give rise, so the presentform of a developing organism has its own presentfunction. A child does not learn its first words sothat it may later combine them into sentences.First words have their own economy. Moregenerally, a child's grammar, at each stage ofdevelopment, is a possible adult grammar.

This last statement is limited exactly to theextent that the principle of recapitulation islimited. Some stages may have been inserted intothe developmental sequence relatively late inevolution. Others may now serve a new function,having lost the function for which they originallyarose. Such stages, perhaps particularly duringthe early months of rapid growth before linguisticfunctions have begun to differentiate, would havesurvived through successive generations becausethey facilitated later adaptive changes, while notinfringing on present functions. In this sense, anearly form may "preadapt" for a later. However, itwould be an error to see such preadaptations asliteral preformations. Development, like evolution,is a tinker, putting to present use whateverchance has laid to hand (cf. Jacob, 1977).

Individuals reach the same developmentalends by different routes and at different

ratesMany linguists believe that universal aspects of

language are "innate." Universals are said to bepurely genetically determined, and either to bepresent at birth (Chomsky, 1986) or to maturewithout contribution from the environment (Borer& Wexler, 1987). Aspects that vary acrosslanguages, though perhaps constrained to a fewvalues by principles of universal grammar, aresaid to be learned from, or "fixed" by, thesurrounding language. This is the model of"parameter setting" in syntactic theory (Chomsky,1981; Roeper & Williams, 1987) and, in effect, ofJakobson's (1968) account of phonologicaldevelopment (cf. Goad & Ingram, 1987).

I do not challenge the descriptive adequacy ofsuch theories. What I question is the proposed de­velopmental mechanism, the attribution of speciesinvariance to the genes, within-species variabilityto the environment. For, in fact, both invarianceand variability arise from both genes and envi­ronment. The point is important because individ­ual similarities in language development are oftentaken as support for a "biological" account of theprocess, while individual differences are seen as athreat (Goad & Ingram, 1987).

Consider, first, that identical phenotypes may beinduced either by a genetic change in anunchanged environment or by a changedenvironment acting on an unchanged genome. Forexample, a yellow (as opposed to grey) body in thefruit fly (Drosophila melanogaster) can be pro­duced either by mutation or by feeding the flylarva on fruit jelly impregnated with yellow dye(Dobzhansky, 1957). Similarly, the lack of a

18 Studdert-Kennedy

posterior cross-vein in fruit fly wings can resulteither from mutation or from subjecting the flypupa to a brief high-temperature shock early indevelopment (Waddington, 1975, Chapters 7 and8). Another example, not due to experimentalintervention, comes from Piaget (1978) whostudied fresh water snails in Swiss lakes duringthe late 1920's. He describes three species. Onespecies (Limnaea stagnalis), found in deep, calmwaters, has an elongated shell; two other species(L. lacustris and L. bodamica), found in shallow,rough waters, have almost indistinguishablestubby, contracted shells. If the two stubby speciesare bred in the calm water of a laboratoryaquarium, L. bodamica retains its stubby shellover many generations, while L. lacustrisgradually takes on the elongated form charac­teristic of L. stagnalis. Had Piaget simplydescribed the two stubby species, he might haveattributed their similarity to shared genes. But wecannot reliably infer genotype from phenotype, asdiscussions of language development oftenassume. Without controlled breeding studies, wecannot determine whether an invariant, species­typical form reflects genetic constancy orenvironmental constancy.

Notice, moreover, that even when we haveisolated genetic or environmental factors thatcontribute to some aspect of phenotypic variation,we have not demonstrated that the variable heldconstant has no effect. The snail species that takesan elongated form after generations in calmwater, a contracted form after generations inrough water, can adapt to the environmentalchange because it is genetically equipped to do so.The snail species that retains its contracted form,whether bred in rough waters or in calm, lacksthis genetic potential. But we cannot concludefrom this that the environment has no effect onthe form of its shell. We can conclude only that thetwo environments to which the snail was exposeddid not suffice to select for a change in formalthough other environments might have done so.

In short, phenotypic form is both genetically andenvironmentally determined, a fact of someimportance to language development. If we cannotassign phenotypic invariance, even in these"simple" examples, either to genes alone or toenvironment alone, we are surely not justified indoing so for language universals, or for the within­language constancies in linguistic developmentthat are often cited as evidence for an "innate"language competence..

By the same token, variability in developmentamong children learning the same language

cannot be assigned solely to the environment.Certainly, different children, even within thesame social class, must be exposed to widelyvarying patterns and frequencies of linguisticinput, and such differences are likely to affect thecourse and rate of development. At the same time,children differ genetically, and these differencestoo affect development (Locke & Mather, in press).The problem is not to assign variability to agenetic or environmental source, but to unders­tand how children resist the effects of genetic andenvironmental variation so as to arrive at acommon language.

Development is buffered against extremevariation

The problem of uniformity within and acrosslanguages is, in some respects, the obverse of theproblem of language diversity. Languages, likespecies, differ because individuals differ. If allmembers of a species were genetically identical,followed an identi.cal course of development, andarrived at an identical developmental term, newspecies could never arise. Genetic variation withina species, affording subtle individual differencesin adaptive response to new environments, is thebasis for the origin of species by natural selection.Similarly, if all speakers of a language followed anidentical course of development to an identicallinguistic term, new languages could never arise.

Of course, differences among speakers that leadto language change are unlikely to rest on geneticdifferences, because speakers of differentlanguages do not differ systematically in theirlanguage-related genetic endowment: childrenlearn any first language to which they areexposed. In fact, from a genetic point of view,language differences rest on commonalties ratherthan differences: the shared capacity to adapt toany linguistic environment and to learn thesurrounding language. In other words, it isprecisely because language is supported by (wemust presume) many thousands of genes whichdepend on the environment to trigger theirexpression, that languages can be learned-andthat languages differ. Subtle, environmentallyinduced individual differences in languagedevelopment, selected and transmitted to otherindividuals by sociocultural forces of which weunderstand at present, very little, are one sourceoflanguage diversity.

What then are the forces that resist change,guiding speakers into a common language? Theproblem is a special case, analogous to the generalproblem in developmental biology of how species

Language Development from an Evolutionary Perspective 19

maintain their identities across generations. Wecan gain conceptual leverage on the question byinvoking the principle of "canalization"(Waddington, 1957, 1975). The term is purelydescriptive, intended to capture "...a large numberof well-known facts in genetics and embryology,all of which are summarized in the statement thatthe development of any particular phenotypiccharacter is to some extent modifiable, and tosome extent resistant to modification, by changeseither in the genotype or in the environment"(Waddington, 1975, p. 72). In other words, bothgenetic and environmental variations are bufferedagainst extreme divergence, canalized towardequifinality by constraints from the developingsystem. Genes and environment act in reciprocal,mutually constraining concert to assure a stabletrajectory of growth.

To characterize the constraints on languagegrowth in any detail is well beyond the scope ofpresent knowledge. There would seem, however,to be two main sources of constraint, one in thechild itself, one in the physical, social andlinguistic environment. Within the child, eachstep in cognitive and linguistic development opensnew paths and closes others, an automaticconsequence of the selection, and increasingdifferentiation, of the neural structures thatsupport language and cognition, continuing theprocesses of tissue differentiation in the embryo.Once launched, the learning of a particularlanguage becomes increasingly constrained bystructural changes in the child's brain. Thesechanges open paths into the language beinglearned, and perhaps into other languages of thesame general type. Presumably, developmentalong these paths will cease if the child is abruptlytransferred to a· markedly different linguisticenvironment. (Studies of "savings," that is, of thehead-start or lack of it, displayed by children whohave switched languages or been exposed to asecond language, might then throw light oncognitive and linguistic commonalties between the

.languages and on their inferred neuralsubstrates.)

The external constraints on the child arise, first,from a general social context that invites the childto engage with its companions and to match itsbehavior to theirs. The second source of constraintis the language that the child's companions speak.Every language is a solution, one of anuncountable, but presumably limited, set ofsolutions to the problem of developing acommunicative system within the perceptuomotor,memorial and cognitive limits of humans (cf.

Lindblom, 1983). The child is then easily guidedinto its language because, as we have alreadyremarked, every language has evolved underconstraints very like those that limit the child.

Given the increasing strength of endogenousand exogenous constraints as the child grows, wewould expect phenotypic variation in phonologyand syntax, if not in lexicon, to diminish aschildren come into possession of their language.Nonetheless, individual variability, far from beingevidence against a biological account of languagedevelopment, is a hallmark of developingbiological systems.

In the following section, I will attempt to illus­trate some of the principles sketched above asthey might apply to the development ofphonologi­cal form over the first two years of life.

THE EMERGENCE OF PHONETICSEGMENTS

PerceptionThe first systematic studies of infant capacity to

perceive speech derived from the well-knownstudies of adult categorical perception. In aseminal set of experiments, the model for manyothers, Eimas and his colleagues (Eimas,Siqueland, Jusczyk, & Vigorito, 1971) demon­strated that 1- and 4-month old infants displayedthe same pattern of discrimination betweensyllables on a synthetic voice onset time (VaT)continuum as adults: they discriminatedsignificantly better between syllables belonging todifferent English phoneme categories thanbetween syllables belonging to the same phonemecategory. Much the same result came from laterinfant studies of stop consonant place ofarticulation, consonant manner, the [r]-[1]distinction, and almost every other syntheticcontinuum on which infants were tested (seeAslin, Pisoni, & Jusczyk, 1983, for a compre­hensive review).

From this accumulated wealth of evidenceEimas (1975) concluded that categoricalperception reflects the operation of innatemechanisms, specialized for processing speech,and that "...these early categories serve as thebasis for future phonetic categories" (p. 342). Thisconclusion is often cited by students of languageacquisition whose primary interest is in syntax.For example, Gleitman and Wanner (1982)summarily dismiss the problem of the origin ofdiscrete phonetic segments (a problem no lesssevere than the problem of the origin ofmorphemic units to which they devote much

20 Studdert-Kennedy

attention) by citing the work of Eimas and othersto support the claim that "...no learning apparatusis required for an initial segmentation of theacoustic wave into discrete phones. Thesegmentation has been provided in the nervoussystem" (p. 16). Exactly how this capacity forsegmentation came to be evolutionarily, or comesto be ontogenetically, "provided in the nervoussystem," they do not consider.

In any event, many reasons to doubt Eimas'sinterpretation of the infant data have been givenby Jusczyk (1981), Kuhl (1987), Studdert-Kennedy(1986), and Walley, Pisoni and Aslin (1981). I willnot rehearse the reasons in detail, but here brieflynote three. First, we now know that categoricalperception is peculiar to neither speech .noraudition (see Hamad (1985) for several relevantpapers). Second, Kuhl and her colleagues havedemonstrated categorical effects on labial,alveolar and velar VOT continua in chinchillas(Kuhl & Miller, 1975), and on a place continuumin macaques (Kuhl & Padden, 1982, 1983). Thus,whatever the bearing of categorical studies on thedevelopment of speech perception, they clearly donot reflect a perceptual specialization. Third, theclaim that categorical perception reveals the basisfor future phonetic categories confuses two typesof category. The categories mimicked by asynthetic series vary along a single acousticdimension in a fixed context; they comprise, atmost, the random variations that we mightobserve in a single syllable, spoken repeatedlywith identical stress and at an identical rate bythe same speaker. However, the phonologicalcategories that a child must form are equivalenceclasses of intrinsic and extrinsic allophonicvariants, formed by execution of a particularphoneme in a range of phonetic contexts, spokenwith varying degrees of stress, at different rates,and by different speakers.

For study of the infant's potential grasp on theseequivalence classes, we must turn to another bodyof research. In a systematic series of studies Kuhland her colleagues have demonstrated that 3- to6-month old infants can learn to recognize theequivalence of: (i) isolated vowels spoken by amale, a female or a child; (ii) syllables with thesame consonantal onset before different vowels;(iii) syllables with a particular consonantalacoustic pattern in initial, medial or final position;(iv) syllables that share an initial "feature" (stop,nasal). (For a comprehensive review, see Kuhl,1987). Yet we should be cautious in interpretingeven these studies as evidence that infants c·an

recognize phonetic (as opposed to acoustic)invariants, if only because Kluender, Diehl andKilleen (1987) have successfully trained Japanesequail to form equivalence classes across syllablesspoken with different pitch contours, by differentspeakers or with the same consonant beforedifferent vowels. In short, we have no grounds forclaiming that perceptual analysis of the acousticstructure of a spoken syllable, and the formationof categories from the resulting components,engage distinctively human capacities, whetherinfant or adult.

In fact, these infant studies seem to have nomore than a general bearing on the specialized de­velopment of language. They are psychophysicalstudies, demonstrating that infants at, or soonafter, birth have the capacity to discriminate andcategorize certain acoustic patterns which occur inspeech. They provide detailed support for a gen­eral observation, suggested by the fact that speechsounds are concentrated in the few octaves of theacoustic spectrum to which humans (and manyother animals) are most sensitive: spoken lan­guage has evolved and develops within the con­straints of prelinguistic auditory capacity. Surely,it would be surprising if this were not so. We donot expect an animal to have a communicationsystem that is not matched to its sensory capaci­ties.

However, the most serious objection to the stan­dard interpretation of these studies is that byassuming the child to be innately endowed withsensitivity to the smallest phonetic units of whichspoken utterances are composed, they implicitlyadopt a view of development as proceeding fromthe specific to the general rather than the reverse.Yet, as we have already seen, it is a general ruleof both phylogeny and ontogeny that complexstructures evolve by differentiation of smallerstructures from larger. Accordingly, we should notexpect children to build words from phonemes asadults do; rather, we should expect phonemes toemerge from words. We may note, in passing, thata similar principle must apply to the developmentof word classes and syntactic structures, a fact notgenerally recognized in developmental psycholin­guistics.

From this vantage we can take a different viewof infant speech perception than has hithertoprevailed, one that emphasizes development offunction rather than psychophysical capacity. Oneaspect of this work will entail charting the processof "attunement" to the surrounding language, thecourse by which the infant learns to perceive

Language Development from an Evolutionary Perspective 21

speech, bringing to bear capacities evidenced inthe laboratory on the speech it hears at home.Work along these lines has, in fact, already begun,with studies of infants' loss of sensitivity tophonetic contrasts not deployed in their nativelanguage (e.g., Best, McRoberts, & Sithole, 1988;Werker & Tees, 1984). Other studies by Hirsh­Pasek and her colleagues (e.g., Hirsh-Pasek &Golinkoff, in press; Hirsh-Pasek, Nelson, Jusczyk,Cassidy, Druss, & Kennedy, 1987) are tracing thedevelopment of sensitivity to prosodic patternsthat specify clausal units, perhaps the thin end ofan infant wedge into syntactic structure.

Here, however, I wish to pursue another aspectof perceptual function: its role in guiding thedevelopment of production. For I assume that thedevelopment of speech depends not only on thematuration and use of the vocal motor system, butalso on the infant's gradual discovery of structurein the speech it hears: By learning how to listenthe infant learns how to speak. The speech signaltherefore comes to specify for the infant theactions-or, more exactly, the motoric componentsof articulatory action-by which speech isproduced. Yet neither phonemes nor features, theperceptual units typically posited in infantstudies, can be defined either acoustically ormotorically. They are abstract units beyond thereach of an infant who does not yet know alanguage. If we are to understand speechdevelopment, we must couch our descriptions andframe our experiments in terms of auditory andmotoric units to which an infant might reasonablybe expected to have access. In what follows I willbriefly sketch a speculative account of the processby which infants gradually harness prelinguisticmotoric elements to communicative use.

ProductionWe may discern at least four broad stages in the

early development of speech, successive cycles ofdifferentiation and integration that carry theinfant from prelinguistic cries and mouthings tothe emergence of phonetic segments and thebeginnings of phonology in early words. Thestages form a necessary hierarchical sequencefrom the general to the specific, from the non­linguistic to the linguistic. However, they are notsharply delimited: processes that begin in onestage may continue more or less unchanged intoseveral later stages-for example, a child maycontinue to babble long after it has begun toproduce words. Also, we must be cautious incharacterizing a stage as differentiative or

integrative for two reasons. First, becausedifferentiation at one level (for example, breakinga syllable into its component gestures) entailsintegration at another (coordinating movements toform gestures), and vice versa. Second, becauseeven within a level both differentiation andintegration go on during a single stage: forexample, integrating gestural patterns intoconsonants and vowels entails simultaneouslydifferentiating a syllable into its segmentalcomponents.

Nonetheless, since even a coarse and tentativetaxonomy of development may lend insight into itsprocess, let us consider the following four stages:(1) early vocalizing: differentiation of respiratoryand vocal tract activities into patterns ofsoundmaking associated with different nonspeechactions (0-7 months); (2) canonical babbling:integration of nonspeech movement patterns intorhythmic syllabic structures (7-10 months); (3)variegated babbling and early words: differen­tiation of the syllable into its component gestures(10-15 months); (4) integration of recurrentgestural patterns into canonical phonetic seg­ments (15-24 months). The suggested time periodsare, of course, approximate, because childrendiffer widely in the time courses of theirdevelopment.

Differentiation: Early vocalizingOver the first half year of life infant sounds

progress systematically from clearly nonspeechvocalizations to canonical babbling that invitesphonetic transcription. The changing forms arepresumably determined, at least in part, bygeneral maturation and by exercise of the vocalapparatus and of its neural control structures (forreview, see Kent, 1981). Stark (1986) divides thedevelopment of soundmaking over the first 7-8months of life into three stages: reflexive cryingand vegetative sounds (0-8 weeks), cooing andlaughter (8 to 20 weeks), and vocal play (16-30weeks). These stages are of interest in the presentcontext because they reflect changes in thetopography of vocal tract activities and in theforms of the resulting sounds by whichprelinguistic elements are differentiated andmarshalled for protolinguistic use (cf. Oller, 1986).

Reflexive crying, like the distress calls of otheranimals, has, of course, a communicative (thoughnon-linguistic) function. Cries, executed with arelatively unconstricted vocal tract, arepredominantly voiced, often with the formantstructure of low to mid front vowels. Vegetative

22 Studdert-Kennedy

sounds-the grunts, sighs, clicks, stops and pops,associated with breathing and feeding-may be.either voiced or voiceless, formed with either anopen, vowel-like or a constricted, consonant-likeconfiguration of the vocal tract. The cooing, orcomfort, sounds of Stark's second stage tend tooccur in a series of 3-10 segments, each of about500 ms, separated by voiceless intakes of breathand glottal stops. Their energy is concentrated infrequencies below about 1500 Hz, a pattern thatOller (1980) has termed a "quasi-resonant"nucleus, having the form of a nasal consonant ornasalized vowel.

Thus by the end of the fifth month, the infant'ssound repertoire already contains a varietyprotoconsonantal and protovocalic elements. Theimportant changes in the next stage are not somuch in the size and quality of the repertoire as inits function and organization. First, sounds beginto lose their original functional moorings: theybecome "...divorced from their previous cry,vegetative or comfort sound contexts, and are usedin a variety of communicative situations" (Stark,1986, p. 159). Second, sounds become longer (700­1500 ms), and form longer, more complexsequences. The infant emits repetitious strings ofconsonant-like clicks, trills, friction noises,syllabic nasals with a constriction at the front ofthe mouth, and lip smackings. Vowel-like sounds,now with a "fully resonant" nucleus (Oller, 1980)often carrying extreme pitch glides, are executedwith increased variation in tongue height andfront-back placement.

However, the key change during this stage iscombinatorial: the infant begins to superimposemovements of tongue, jaw and lips on thelaryngeal actions associated with cry (Koopmans­Van Beinum & Van der Stelt, 1986), so that theproportion of supraglottal to glottal articulationsgradually increases (Holmgren, Lindblom,Aurelius, JaIling, & Zetterstrom, 1986.) Even atthis early stage, children may differ quite sharplyin their preferred types of vocalization (Stark,1986). Toward the end of the stage increasinglylong and complex combinations of tractconstrictions and openings appear, formingsequences that Oller (1980) terms marginalbabQle.

The functional value of the differentiation ofearly soundmaking into these diverse patterns isnot obvious. Teleologically, of course, exercise ofthe vocal apparatus must contribute to its neuraland anatomical development, laying the basis forlater integration of sounds into syllables.However, the immediate function may simply lie

in the increased range of emotional expressionthat it affords, with a consequent tightening ofdyadic social bonds (cf. Stern, 1985; Trevarthen &Marwick, 1986). Interestingly, MacNeilage(personal communication) has noted a possibleprecursor of such differentiation in the repetitivegirneys (lip smacks, accompanied by a low mur­mur) exhibited by Japanese macaques in intimate,affiliative situations (Green, 1975).

In any event, I have dwelt on this early stagebecause it is here that we can see most clearly thetinkering together of disparate non-linguistic pat­terns of vocal tract activity into the beginnings ofphonological structure (cf. Bates et aI., in press).

Integration: Canonical babbleWith the onset of canonical babble, often a

sudden event over a few days in the 7th or 8thmonth, the infant begins to integrate patterns ofvocal tract constriction and opening into unitary,cohesive syllables. Rhythmic, reduplicated soundsequences are common: [bababa], [nenene],[dIChdI]. However, the convenient use of phonetictranscription should not mislead us into supposingthat the infant has independent control oversegments within a syllable. In fact, rhythmiclowering and raising of the jaw in canonicalbabble seems to occur with little or noindependent movement of the tongue (Davis &MacNeilage, 1990). This rhythmic jaw oscillationoften begins about the same time as rhythmicmovements of the legs and arms (Thelen, 1981; inpress), and perhaps facilitates the integration ofvocal tract activities into cohesive syllabicpatterns.

We owe the clear distinction between early vo­calization and canonical babble to Oller (1980,1986). He describes some half dozen acousticproperties of structures that listeners recognize ascanonical, adult consonant-vowel (CV)(constricted-to-open) or CVC syllables. Of theseproperties the most important in the present con­text (because they are the only ones that appearrarely, if ever, in early vocalizations) are tempo­ral. The canonical syllable has a duration of noless than 100, no more than 500 ms; onset (and,when present, offset) formant transitions displaysmooth changes in frequency and amplitude withdurations between 25 and 120 ms. These acousticpatterns reflect the infant's increasing skill in in­tegrating the closing and opening phases of jawmovements.

Patterns of movement in babble are not random.During the const"riction phase of the syllable,complete closure, as in stops, is favored over

Language Development from an Evolutionary Perspective 23

partial closure, as in fricatives. Points of closureare biased toward the front of the mouth,engaging lip and tongue tip muscles active insucking. During the open phase, the favoredtongue positions are those associated with lowfront vowels, indicating that the tongue tends toride up and down on the jaw with little or noactive movement of its own (Davis & MacNeilage,1990). The glottis is typically approximatedthroughout the syllable, giving the impression ofvoiced consonants at syllable onset.

These "phonetic" biases have been reported forinfants growing in a number of languageenvironments (Locke, 1983). The biases are alsopresent in many adult languages, perhapsreflecting the infant proclivities from whichlanguages have evolved (Locke, 1983; cf.Lindblom, 1989). Here we have a tangle thatcross-linguistic data on babble are still too sparseto resolve. How far do the perhaps universalbiases of infant babble reflect the naturationalstate of the infant vocal apparatus, and how far dothey reflect the surrounding language? We mayask much the same question concerning individualdifferences in babbling repertoire within alanguage: do they reflect differences in thedevelopment of the vocal apparatus or differencesin the language patterns which the infantshappen to have heard? Whatever the answers tothese questions, recent work with deaf infants hasshown that the emergence of canonical babble isnot purely an effect of maturation.

Deaf infants were once said to exhibit the samebabbling patterns as hearing infants, at least overthe first year of life (Mavilya, 1969; cf. Locke,1983). However, recent comparisons havedemonstrated that canonical babble does notappear on schedule in deaf infants (Oller & Eilers,1988; Oller, Eilers, Bull, & Carney, 1985; Stoel­Gammon & Otomo, 1986). Whether this is due tothe lack of auditory input from the infant's ownvocalizations, from those of its adult companions,or both, we do not know. If self-stimulation werethe only essential, a purely maturational accountcould still hold. To the extent that communicativeinterchange, or the impulse to imitate the actionsof conspecifics plays a role, the onset of canonicalbabble would be determined, at least in part, byexperience of a surrounding language. Persuasiveevidence for the role of the surrounding languagecomes from reports that deaf children exposed tosign language begin to "babble" with their fingersat about the same age as hearing infants begin tobabble with their mouths (Newport & Meier, 1985;Laura Petitto, personal communication).

Finally, what is the function of canonical bab­ble? "None" has been the answer of some (e.g.,Jakobson, 1968; Lenneberg, 1967) who viewedbabble as random mouthing, sharply discontinu­ous from truly linguistic utterance. However,Locke and Pearson (1988) have recently reported asevere (though not irremediable) delay in the de­velopment of speech in a tracheostomized child,deprived of the opportunity to babble from 5 to 20months. This suggests that babble is a necessarystep in normal linguistic development. Moreover,several studies have now shown that babblemerges smoothly into early words and that thephonetic structure of a child's first words tends toreflect its babbling preferences (Locke, 1983;Oller, Wieman, Doyle, & Ross, 1975; Vihman,Macken, Miller, Simmons, & Miller, 1985; Vihman& Miller, 1988). The immediate function wouldseem then to be to continue the "imitative" processof aligning the infant's communicative skills moreclosely with those of its adult companions.

Differentiation: Variegated babble andearly words

Toward the end of the first year full syllablereduplication fades. The infant begins todifferentiate the closing and opening gestures ofsuccessive syllables so that "variegated" sequences(Oller, 1980) appear, giving the impression ofvariations in consonant and/or vowel (e.g., Inem/,Im';jnw, Idedi/). Also, at about this time the childproduces its first recognizable attempts at adultwords or phrases (All gone, What's that?), us~ally

(in English children) single syllables ordisyllables. The two modes of output then proceedconcurrently, often for many months, with wordsgradually coming to predominate. Over this periodthe child is gradually forging links between itsperceptual and productive capacities.

The perceptual ground for this developmentseems to be laid by the child's growing attention towords in the surrounding language. Data on earlycomprehension are scarce, but children evidentlyaccumulate sizeable receptive lexicons beforeattempting their first words. Benedict (1979) hasreported a longitudinal study of comprehensionand production in eight children from the age of 9or 10 months to a point where they had achieved aproductive lexicon of 50 words (between 15 and 22months). In every child comprehension comfort­ably outstripped production. On the average, thechildren understood more than 60 words by thetime they could produce 10, with a range from 30to 182 words; and on the average, their receptivelexicon over the period of the study was three

24 Studdert-Kennedy

times their productive lexicon (Benedict, 1979,Table I).

The gap between perception and productiondemonstrates that a perceptuomotor link is not"innate," as some have proposed (Liberman &Mattingly, 1985), but must develop. Many studieshave shown that the phonetic forms of early wordsare similar to those of concurrent babble (seeVihman, in press, for review). A child selects forimitation words that match "vocal motor schemes"(McCune & Vihm~, 1987) already present in itsbabble and avoids words that do not match (Menn,1983). Thus, the child initially grows into alexicon, as it were, by discovering correspondencesbetween adult words and auditory feedback fromits own babbled output. Because the structure ofauditory feedback must correspond, in somefashion, to the structure of motor controls thatproduced it, the child's recognition of babble-to­word correspondences presumably facilitatesgrowth of perceptuomotor links, and their gradualextension to new words.

Each word (or formulaic phrase) seems to be aprosodic unit (Macken, 1979), its productionplanned as a whole (Menn, 1983). Evidence forthis comes from gestural interactions. A childoften fails to execute two different places or man­ners of articulation within the same word, thusmaintaining some of the reduplicative tendenciesof canonical babble: dog [gag], lady [jeiji],duck [tAt]. Closing and opening gestures also in­teract. For example, Davis and MacNeilage (1990)report an extensive study of a child's concurrentbabbling and speech over the period from 14 to 20months of age. Their data are replete with in­stances not only of consonant, but of vowel andeven consonant-vowel assimilation. The latter isrevealed by the child's preference for high frontvowels following alveolar closures and for low,front-central vowels following labial closures. Atthe same time, these authors also report an in­verse relation between consonant and vowel redu­plication: where the child succeeds in combatingassimilation in the open phases of a disyllable, sheoften fails to do so in the closing phases, and viceversa. This demonstrates an incipient segregationof consonants and vowels into phonetic classes.

The study by Davis and MacNeilage isparticularly important because the child deployedan unusually large lexicon, growing from about 25words at 14 months to over 750 words at 20months. Evidently, a child may have a substantiallexicon long before it has fully masteredsegmental structure. The principal phonologicalachievement of this period, then, is internal

modification of the integrated syllable bydifferentiation of its gestural components.

Integration: From gestures to phonemesThe final step in the path from mouth sounds to

segments is the integration of gestural patterns ofsyllabic constriction and opening into the coherentperceptuomotor structures we know as consonantsand vowels (Studdert-Kennedy, 1987). As is wellknown, the status of the segment is problematic.For, on the one hand, we can neither specify theinvariant articulatory-acoustic properties sharedby all instances of a particular consonant or vowel,nor isolate any given segment as a discretearticulatory-acoustic entity within a syllable. Onthe other hand, across-word metathetic errors inadult speaking ("Spoonerisms") attest to thefunctional role of segments in the planning andexecution of an utterance (Shattuck-Hufnagel,1983). Moreover, such errors typically entailexchanges between consonants and vowels thatoccupy corresponding slots in their respectivesyllables. Since the exchanging elements may bephysically quite disparate, it is evident that theirexchange is premised on shared function (onset,nucleus, coda) in the formation of a syllable.

There are therefore two aspects to theemergence of segments as elements of wordformation in a child's lexicon. First is the groupingof all instances of a particular gesture-soundpattern into a single class presumably on the basisof their perceptuomotor, or phonetic, similarity(e.g., grouping the initial or final patterns of dad,dog, bed, etc. into the class /dJ). Second is thedistributional analysis and grouping of thesegesture-sound patterns into higher-order classes(consonants, vowels) on the basis of their syllabicfunctions. These processes of phonetic categoryformation are perhaps analogous to thoseproposed by Maratsos and Chalkley (1980) for theformation of syntactic categories.

Evidence from across-word metathetic errors forthe formation of these classes in young children issparse. The only systematic data known to mecome from Jeri Jaeger (personal communication).She reports her daughter's first across-wordmetathesis as occurring in her 27th month: ummytakes for tummy aches. This was followed in her30th month by fritty pace for pretty face, sea tet fortea set, and Bernie and Ert for Ernie and Bert.Jaeger did not report data on the size of herdaughter's lexicon at this time, nor on thecomplexity orher multi-word utterances. But thecollection of errors suggests that both were welladvanced.

Language Development from an Evolutionary Perspective 25

Two possible selection pressures may precipitateformation of consonant and vowel classes. Onepressure is toward economy of storage as thelexicon increases in size. We have seen that achild may accumulate an appreciable lexicon ofsome 750 words without showing signs of indepen­dent segmental control (Davis & MacNeilage, inpress). But this lexicon is roughly one hundredthof the size that it will eventually become; and itseems reasonable to suppose that, as the lexiconincreases, words should organize themselves onthe basis of their shared gestural and soundpropertieso Recurrent patterns of laryngeal andsupralaryngeal gesture would thus formthemselves into unitary classes of potential utilityfor recognition and activation of lexical items (cf.Lindblom, 1988; Lindblom, MacNeilage, &Studdert-Kennedy, 1983).

A second possible selection pressure is towardrapid lexical access in the formation of multi-wordutterances. Several authors (e.g., Branigan, 1979;Donahue, 1986) have argued that the form of earlymulti-word combinations may be constrained bythe child's limited ability to organize and executethe required articulatory sequences. One suchconstraint might be a child's inability to producetwo successive words with different initial placesof articulation. Donahue (1986), for example,describes two strategies adopted by her son in hisfirst two-word utterances. One strategy was toattempt only those words that conformed to hispreexisting rule of labial harmony: big book, bigbird, big ball were all attempted, but big dog andbig cooky were "adamantly refused" (p.215). Thesecond strategy was to circumvent the consonantharmony rule by adopting vocalic words lackingconsonants (e.g., where [ej~], want [wac.]) as pivotsthat could be comfortably combined with many ofthe words already in his vocabulary. Such findingsimply that the integration of gestures intoindependent phonemic control structures, orarticulatory routines (Menn, 1983), may serve toinsulate them from articulatory competition withincompatible gestures and so facilitate their rapid,successive activation in multi-word utterances'.

CONCLUSIONI have proposed that the study of language

development might be fruitfully cast in anevolutionary and recapitulatory framework, as asort of post-natal embryology. Language univer­sals (like other species characters) are theendpoint of development (and evolution), aconsequence, not a condition, of learning a naturallanguage. Language function thus determines

language form, and form is viewed as an aposteriori description of, rather than an a prioriprescription for, development. We are thus freedfrom the habit of viewing universals as innate,language-specific properties as learned. Both setsof characters are the product of a species-specificdevelopmental system, in which genetic andenvironmental conditions cannot be separated.

I have attempted to illustrate the approach witha sketch of the process by which the universalphonological categories of consonants and vowelsmight emerge. They are viewed as deriving bysuccessive cycles of differentiation and integrationfrom prior non-linguistic perceptual and motorcapacities.

REFERENCESAslin, R. N., Pisoni, D. B., & JU5Czyk, P. W. (l983). Auditory

development and speech perception in infancy. In M. Haith &J. Campos (Eds.), Infancy and the biology of development.Carmichael's manual ofchild psychology, Vol. II (4th ed., pp. 220­284). New York: Wiley.

Bates, E., Thal, D., & Marchman, V. (in press). Symbols andsyntax: A Darwinian approach to language development. In N.Krasnegor, D. Rumbaugh, R. Schiefelbusch, & M. Studdert­Kennedy (Eds.), Language acquisition: Biological and behavioraldeterminants. Hillsdale, NJ: Lawrence Erlbaum Associates.

Benedict, H. (l979). Early lexical development: Comprehensionand production. Journal of Chl1d Language, 6, 183-200.

Best, C. T., McRoberts, G. W., & Sithole, N. W. (l988).Examination of perceptual reorganization for nonnative speechcontrasts: Zulu click discrimination by English-speaking adultsand infants. Journal of Experimental Psychology: Human Perceptionand Performance, 14, 345-360.

Borer, H., & Wexler, K. (1987) The maturation of syntax. In T.Roeper & E. Williams (Eds.), Parameter setting (pp. 123-172).Dordrecht, Holland: D. Reidel.

Bosma, J. F. (1975). Anatomic and physiologic development of thespeech apparatus. In D. B. Tower (Ed.), The nervous system(Vol.3): Human communication and its disorders. New York:Raven.

Branigan, G. (l979). Some reasons why successive single wordutterances are not. Journal ofChild Language, 6, 411-421.

Chomsky, N. (l965). Aspects of the theory of syntax. Cambridge,MA: M.LT. Press.

Chomsky, N. (l972). Language and mind (rev. ed.). New York:Harcourt Brace Jovanovich .

Chomsky, N. (l981). Lectures on government and binding: The Pisalectures. Dordrecht, Holland: Foris.

Chomsky, N. (l986). Knowledge of language. New York: PraegerSpecial Studies.

Darwin, C. (l871). The descent of man. New York: The ModernUbrary.

Davis, B. L., & MacNeilage, P. F. (l990). Acquisition of correct.vowel production: A quantitative case study. Journal of Speechand Hearing Research, 33,17-27. .

de Saussure, F. (1966). Course in general linguistics (Wade Baskin,Trans.). New York: McGraw Hill.

Dobzhansky, T. G. (1957). Human genetics. New York: ColumbiaUniversity Press.

Donahue, M. (1986). Phonological constraints on the emergence oftwo-word utterances. Journal of Child Language, 13, 209-218.

26 Studdert-Kennedy

Eimas, P. D. (1975). Auditory and phonetic coding of the cues forspeech: Discrimination of the /r-l/ distinction by younginfants. Perception and Psychaphysics, 18, 341-347.

Eimas, P. D., Siqueland, E. R., Jusczyk, P., & Vigorito, J. (1971).Speech perception in infants. Science, 171, 303-306.

Gleitman, L. R., & Wanner, E. (1982). The state of the state of theart. In E. Wanner & L. R. Gleitman (Eds.), Language acquisition(pp. 3-48). New York: Cambridge University Press.

Goad, H, & Ingram, D. (1987). Individual variation and itsrelevance to a theory of phonological acquisition. Journal ofChild Language, 14, 419-432.

Green, S. (1975). Variation of vocal pattern with social situation inthe Japanese monkey (Macaca fuscata): A field study. In L. A.Rosenblum (Ed.), Primllte behaTJior (Vol. 4, pp. 1-102). New York:Academic Press.

Gould, S.J. (1977). Ontogeny and phylogeny. Cambridge, MA:Belknap Press.

Grasse, pop. (1977). E'Dolution of li'Ding organisms. New York:Academic Press.

Harnad, S. (Ed.) (1987). Categorical perception. New York:Cambridge University Press.

Hirsh-Pasek, K., & Golinkoff, R. M. (in press). Languagecomprehension: A new look at old themes. In N. Krasnegor, D.Rumbaugh, R. Schiefelbusch, & M. Studdert-Kennedy (Eds.),Language acquisition: Biological and beha'Dioral determinants.Hillsdale, NJ: Lawrence Erlbaum Associates.

Hirsh-Pasek, K., Kemler Nelson, D. G., Jusczyk, P. W., Cassidy, K.W., Druss, B., & Kennedy, L. (1987). Causes are perceptualunits for young infants. Cognition, 26, 269-286.

Holmgren, K., Lindblom, B., Aurelius, G., Jalling, B., &Zetterstrom, R. (1986). On the phonetics of infant vocalization.In B. Lindblom & R. Zetterstrom (Eds.), Precursors ofearly speech(pp. 51-63). New York: Stockton Press.

Jacob, F. (1977). Evolution as tinkering. Science, 196,1161-1166.Jakobson, R. (1968). Child language, aphasia and phonological

uni'De1'sals (A. R. Keller, Trans). The Hague: MoutonJusczyk, P. W. (1981). Infant speech perception: A critical

appraisal. In P. D. Eimas & J. L. Miller (Eds.), Perspectiues on thestudy ofspeech. Hillsdale, NJ: Lawrence Erlbaum Associates.

Kent, R. D. (1981). Sensorimotor aspects of speech development.In R. N. Aslin, J. R. Alberts, & M. R. Petersen (Eds.),Deuelopment of Perception Volume 1 ( pp. 161-189). New York:Academic Press.

Kluender, K. R., Diehl, R. L., & Killeen, P. R. (1987). Japanese quailcan learn phonetic categories. Science, 237, 1195-1197.

Koopmans-van Beinum, F. J., & Van der Stelt, J. M. (1986). Earlystages in the development of speech movements. In B.Lindblom & R. Zetterstrom (Eds.), Precursors ofearly speech (pp.37-50.) New York: Stockton Press.

Kuhl P. K., & Miller, J. D. (1975). Speech perception by thechinchilla: Voiced-voiceless distinction in alveolar plosiveconsonants. Science, 190, 69-72.

Kuhl, P. K., & Padden, D. M. (1982). Enhanced discrimination atthe phonetic boundaries for the voicing feature in macaques.Perception and Psychaphysics, 32, 542-550.

Kuhl, P. K., & Padden, D. M. (1983). Enhanced discriminability atthe phonetic boundaries for the place feature in macaques.Journal of the Acoustical Society ofAmerica, 73,1003-1010.

Kuhl, P. K. (1987). Perception of speech and sound in earlyinfancy. In P. Salapatek & L. Cohen (Eds.), Handbook of infantperception (VoL 2). New York: Academic Press.

Lack, D. (1961). Darwin's finches: An essay on the general biology ofroolution. New York: Harper.

Lehmann, W. P. (1973). Historical linguistics. New York: Holt,Rinehart and Winston.

Lenneberg, E. H. (1967). The biological foundations of language. NewYork: John Wiley.

Liberman, A. M. & Mattingly, 1. G. (1985). The motor theory ofspeech perception revised. Cognition, 21, 1-36.

Lindblom, B. (1983). Can the models of evolutionary biology beapplied to phonetic problems? In A. Cohen & M. P. R. Broecke(Eds.), Proceedings of the Tenth International Congress of PhoneticSciences. Dordrecht, Holland: Foris.

Lindblom, B. (1989). Some remarks on the origin of the "phoneticcode." In C. von Euler, 1. Lundberg, & G. Lennerstrand, (Eds.),Brain and reading (pp. 27-44). Basingstoke, England: MacMillan.

Lindblom, B., MacNeilage, P., & Studdert-Kennedy, M. (1983).Self-organizing processes and the explanation of languageuniversals. In B. Butterworth, B. Comrie, & O. Dahl (Eds.),Expla1Ultions for language uni'Dersals (pp. 181-203). The Hague:Mouton.

Locke, J. (1983). Phonological acquisition and change. New York:Academic Press.

Locke, J., & Pearson, D. M. (1988). Linguistic significance ofbabbling: Evidence from a tracheostomized infant. Report fromthe Neurolinguistics Laboratory at Massachusetts General Hospital(Whole issue).

Locke, ]. L., & Mather, P. L (in press). Genetic factors in theontogeny of spoken language: Evidence from monozygotic anddizygotic twins. Journal ofChild Language.

Macken, M. A. (1979). Developmental reorganization ofphonology: A hierarchy of basic units of organization. Lingua,49,11-49.

MacNeilage, P. F. (in press). The "postural origins" theory ofprimate neurobiological asymmetries. In N. Krasnegor, D.Rumbaugh, R.Schiefelbusch, & M. Studdert-Kennedy (Eds.),Language acquisition: Biological and beha'Dioral determinants.Hillsdale, NJ: Lawrence Erlbaum Associates.

Mayr, E. (1982). The gruwth of biological thought. Cambridge, MA:The Belknap Press.

McCune, L., & Vihman, M. (1987). Vocal motor schemes. Papersand Reports on Child Language Deuelapment, 26, 72-79

Menn, L. (1983). Development of articulatory, phonetic andphonological capabilities. In B. Butterworth (Ed.), Languageproduction (Vol. II, pp. 3-50). London: Academic Press.

Newport, E. L, & Meier, R. P. (1985). The acquisition of AmericanSign Language. In D. 1. Slobin (Ed.), The crosS-linguistic study oflanguage acquisition, Volume I: The data (pp. 881-938). Hillsdale,N]: Lawrence Erlbaum Associates.

Oller, D. K. (1980). The emergence of the sounds of speech ininfancy. In G. Yeni-Komshian, J. F. Kavanagh, & C. A. Ferguson(Eds.), Child phonology, Volume 1: Production (pp. 93-112). NewYork: Academic Press.

Oller, D. K. (1986). Metaphonology and infant vocalizations. In B.Lindblom, & R. Zetterstrom (Eds.), Precursors ofearly speech (pp.21-35). New York: Stockton Press.

Oller, D. K., & Eilers, R. E. (1988). The role of audition in infantbabbling. Child Deuelapment, 59,441-449.

Oller, D. K., Eilers, R. E., Bull, D. H., & Carney, A. E. (1985).Prespeech vocalizations of a deaf infant: A comparison withnormal metaphonological development. Journal of Speech andHearing Research, 28, 47-63.

Oyama, S. (1985). The ontogeny of information. New York:Cambridge University Press.

Piaget, J. (1978). Beha'Dior and e'Dolution. Chicago: ChicagoUniversity Press.

Roeper, T. & Williams, E. (Eds.) (1987). Parameter setting.Dordrecht, Holland: D. Reidel.

Savage-Rumbaugh, E. S. (in press). Language learning in theBonobo: How and why they learn. In N. Krasnegor, D.

Language Development from an Evolutionary Perspective 27

Rumbaugh, R. Schiefelbusch, & M. Studdert-Kennedy (Eds.),Language acquisition: Biological and behavioral determinants.Hillsdale, NJ: Lawrence Erlbaum Associates.

Shattuck-Hufnagel, S. (1983). Sublexical units and suprasegmentalstructure in speech production planning. In P. F. MacNeilage(Ed.), The production of speech (pp. 109-136). New York:Springer-Verlag.

Stark, R. E. (1986). Prespeech segmental feature development. InP. Fletcher & M. Garman (Eds.), Language acquisition (2nd ed.,pp. 149-173). New York: Cambridge University Press.

Stem, D. N. (1985). The interpersonal world of the infant. New York:Basic Books.

Stoel-Gammon, c., & Otomo, K. (1986). Babbling development ofhearing-impaired and normally hearing subjects. Journal ofSpeech and Hearing Disorders, 51, 33-41

Studdert-Kennedy, M. (1986). Sources of variability in earlyspeech development. In J. S. Perkell & D. H. Klatt (Eds.),Invariance and variability of speech processes (pp. 58-76). Hillsdale,NJ: Lawrence Erlbaum Associates.

Studdert-Kennedy, M. (1987). The phoneme as a perceptuomotorstructure. In A. Allport, D. MacKay, W. Prinz, & E. Scheerer(Eds.), Language perception and production (pp. 67-84). London:Academic Press.

Thelen, E. (1981). Rhythmical behavior in infancy: An ethologicalperspective. DerJelopmental Psychology, 17, 237-257.

Thelen, E. (in press). Motor aspects of speech. In N. Krasnegor, D.Rumbaugh, R. Schiefelbusch, & M. Studdert-Kennedy (Eds.),Language acquisition: Biological and behavioral determinants.Hillsdale, NJ: Lawrence Erlbaum Associates.

Thompson, D. W. (1917/1961). On growth and form (abridged eeL).New York: Cambridge University Press.

Trevarthen, c., & Marwick, H. (1986). Signs of motivation forspeech in infants, and the nature of a mother's support fordevelopment of language. In B. Lindblom & R. Zetterstr6m(Eds.), Precursors of early speech (pp. 279-308). New York:Stockton Press.

Vihman, M. M., Ferguson, C. A., & Elbert, M. (1986). Phonologicaldevelopment from babbling to speech: Common tendenciesand individual differences. Applied Psycholinguistics, 7, 3-40.

Vihman, M. M., & Miller, R. (1988). Words and babble at thethreshold of lexical acquisition. In M. D. Smith & J. L. Locke(Eds.), The emergent lexicon: The child's development ofa linguisticrocabulary (pp. 151-183). New York: Academic Press.

Waddington, C. H. (1957). The strategy of the genes. New York:MacMillan.

Waddington, C. H. (1975). The evolution ofan evolutionist. Ithaca,NY: Cornell University Press.

Werker, J. F., & Tees, R. C. (1984). Cross-language speechperception: Evidence for perceptual reorganization during thefirst year of life. Infant Behavior and Development, 7, 49-63.

Walley, A. c., Pisoni, D. B., & Aslin, R. N. (1981). The role of earlyexperience in the development of speech perception. In R. N.Aslin, J. Alberts, & M. R. Petersen (Eds.), Development ofperception, Volume 1 (pp. 219-255). New York: Academic Press.

FOOTNOTES-In N. Krasnegor, D. Rumbaugh, R. Schiefelbusch, & M. Studdert­

Kennedy (Eds.), Language acquisition: Biological and behavioraldeterminants. Hillsdale, NJ: Lawrence Erlbaum Associates, inpress.

tAlso University of Connecticut and Yale UniverSity.