Go to Semiotic Review of Books Home Page
Go to SRB Highlights
Go to SRB Archives


SRB

SRB Archives

This article appeared in Volume 1 (2) of The Semiotic Review of Books.

Language and Genes

by Myrna Gopnik

It is well known that on occasion children, who are otherwise normal, don't start speaking until they are three or four or even five. When they do start speaking they are extremely hard to understand because their pronunciation and grammar are severely impaired. These children are classified as "dysphasics" or "developmental aphasics" or "language specific delayed" -- all of the terms mean the same: language impairment in the absence of retardation, deafness, psychosis or any other perceivable problem. This broad clinical category probably includes some diversity of syndromes. Until recently the etiology of this disorder has been unknown. It has been variously attributed to minimal brain damage, anoxia at birth, inadequate parental language input and, often by parents, simple stubbornness-the children could talk but they choose not. New data strongly suggest that dysphasia is caused by a single dominant gene. These data come from two different sources, a large single family and statistical studies of large populations. In the large family the pattern of occurrence of dysphasia is the same as that which would be produced by a single dominant gene. There is a problem, however, with this evidence from the point of view of genetic epidemiology. This family was singled out for study precisely because it displayed this pattern. It is possible, though given the size of the family extremely improbable, that what we are seeing here is just a statistical bleep. In order to rule out this ascertainment bias it is necessary to complement the single family study with a broader statistical study which first singles out normal individuals and affected individuals and then looks at the occurrence of dysphasia in their families. Two such studies have been done. Both show that there is a much higher incidence of dysphasia in families of dysphasics than in families of normals. Taken together these new data demonstrate conclusively that dysphasia runs in families and its distribution in families suggests that it is caused by a single dominant gene.

The question now is what precisely does this gene do? In order to answer this question we have to understand precisely what goes wrong with dysphasic children. They are not mute. They do learn to speak and when they reach adulthood their language does not appear on the surface to be significantly different from normal, though on more subtle tests it is clear that they have problems. In order to understand dysphasia it is essential to assume that the children are learning language, albeit imperfectly. This means that the testing and description of their output should be done in the light of what we know about the nature of language. For example, one of the most basic assumptions which must be made is that linguistic output is governed by an underlying grammar which consists of a set of rules operating on underlying representations. Individual surface forms must then be regarded merely as indicators of these underlying rules and representations. Lists of "correct" and "incorrect" surface forms cannot provide an adequate account of dysphasic language. In fact, it makes no sense to talk of "correct" forms. For example, a grammar which does not mark tense and which has only one form for the verb may indeed produce a form such as "I walk", but the verb in this sentence, though it resembles the normal form, is just as unmarked for tense and therefore just as wrong as the verb in "Yesterday I walk". Moreover it can be argued on formal grounds, using examples from Turing machines, that the fact that the output of two sets of rules shares some instances cannot be taken as evidence that the rules which produce them are similar. They may be, but they need not be. Therefore it is invalid to claim that different populations are "the same" merely because this or that "incorrect" surface form which is produced by dysphasics is also produced by young normals or by deaf children or by Broca's aphasics.

What must be demonstrated is that the underlying rules and representations which produce these surface forms are the same in some significant way. The construction of these underlying rules and representations from the occurring surface forms can only be accomplished in light of what is already known about the nature of language. Therefore the description and analysis of dysphasia must be informed by linguistic theory.

It was in this spirit that a detailed case study was conducted and the language of twenty-two members of the family, both normals and dysphasics, was tested. After looking at the data in a preliminary way I formed the hypothesis that a single level of grammar, syntastico-semantic features, was impaired. This was a very risky hypothesis. In order for it to be true I had to show that every such feature was impaired and that all of the parts of the grammar dependent on features were also impaired. Some surprising things may be predicted from this theoretically based hypothesis: "ed" and "s" to mark past and plural respectively should not occur, but aspect marking should occur in three different forms: be Ving (John was running.), be V (John was run.) and Ving (John running.); since tense is not marked, subject pronouns should be able to be dropped ("When they play get points."). All of the predictions from this theoretical hypothesis are confirmed by the data. Moreover these findings are consistent with reports in the literature, though these reports are sporadic and do not report on all of the forms which are significant for this hypothesis. The data show clearly that one of the things wrong with these particular dysphasics is that they cannot construct that part of grammar which handles feature marking: they are feature-blind. Yet other parts of the grammar, such as thematic relations are unimpaired.

But the situation is not so simple. In all of these subjects the phonology is also impaired, though there is as yet no good description of precisely what they do wrong. It may also be, though here the data are not quite so clear, that complex syntactic rules are also missing. The question is then what does all of this tell us about what the gene does. It may be that the gene affects several different parts of language independently or it may be that these diverse symptoms are connected in some way. We know that the level of feature specification seems to be fairly autonomous in all grammars. We also know that, though all languages have a phonology, not all have features and, among those that do, the complexity of the feature system varies widely. English has a fairly simple system and Inuktitut a much more complex one. Even if we believe that the fact that all of the subjects have both phonological and feature problems argues for a single underlying connected cause between them, we still have to find out what the cause is.

Three fundamentally different sorts of causes have been suggested: 1. physicalist accounts suggest that the primary problem is in a peripheral system, either auditory or oral-motor, 2. cognitive explanations argue that the deficit is in some general cognitive processing system and 3. linguistic accounts propose a deficit in an underlying specific linguistic process. The physicalist explanation would suggest that because the subject did not hear or could not move his mouth to form such sounds as "ed" or "s", they could not be incorporated into a grammar. Such explanations explicitly deny that innate grammars exist or at least that they have a complex internal structure. In these theories grammars are primarily the consequence of generalizations over inputs, either acoustic or motor. One of the problems with such hypotheses, even if we were willing to grant their initial assumptions, is that they do not provide an explanation for many of the observed data. They cannot account for the pattern of aspect, for example. Moreover, these subjects pass the standard tests for hearing and oral apraxia. If there is a defect it is a subtle one: when we see severe oral apraxia, as in persons with cerebral palsy, this pattern of feature-blindness does not occur. A cognitive explanation would suggest that both phonology and features are controlled by the same general cognitive capacity. We know from l.Q. tests-imperfect instruments though they may be--that these subjects are normal on non-verbal tests; it is only on the verbal tests that they show serious impairment. Conversely it has been demonstrated that subjects with extremely low intelligence have phonologically and grammatically perfect language. There is therefore no simple connection between general intelligence and language. Though someone might be able to find a specific cognitive impairment which also affected these aspects of language, no such hypothesis has yet been presented. The linguistic hypothesis does account for a significant; coherent subset of the grammatical data. No one yet has a- complete description, much less an explanation, for the phonological data. The linguistic stance suggests the hypothesis that the deficit is not in the surface sounds which the dysphasics produce, but in the ability to construct an underlying phonological system which produces these forms. The data also seems to be pointing in this direction. An intriguing question is whether the inability to construct an underlying phonological system. To date these two systems have been considered autonomous. If it could be shown that they had an underlying similarity under genetic control, it would profoundly change our notion of the organization of grammars.

The most intriguing, though necessarily speculative, questions which arise from this work concern the evolution of language. If there is a single gene which controls for features, then it is not unreasonable to suppose that a single recent mutation could have given rise to features in grammar. Though such evidence in one sense supports the innateness of language, in another sense it raises the question as to whether there is any single entity LANGUAGE. It may be best to think of language as a kind of Rube Goldberg machine, made up of bits and pieces which have separate and different evolutionary histories and which are controlled by different genes. Language emerged when all of these separate pieces, designed for different purposes at different times, were coordinated into one intricate, elegant and interconnected system.

Selected References

Clahsen, H. (1989). "The grammatical characterization of developmental dysphasia." Linguistics 27, pp.897-920. The Hague:Mouton de Gruyter.

Eisenson, J. (1984). Asphasia and Related Language Disorders In Children. New York: Harper and Row.

Grodzinsky, Y. (1986). "Language deficits and the theory of syntax." Brain and Language 27.

Hurst, J.A., Baraitser, M. Auger, E. Graham, F. and Norell, S. (1990). "An extended family with a dominantly inherited speech disorder." Developmental Medicine and Child Neurology. 32. (pp. 347-355).

Pinker, S. and Bloom, P. (in press). "Natural language and natural selection." The Behavioral and Brain Sciences. Cambridge University Press.

Tallal, P., Randal, R. & Curtiss, S. (1989). "Unexpected sex ratios in families of language/learning-impaired children." Neuropsychologia. 27. (pp. 987-998).

Tomblin, J.B. (1989). "Familial concentration of developmental impairment." Journal of Speech and Language Disorders. 54. (pp. 287-295).

Wyke, M.A. (1978). Developmental Dysphasia. New York:Academic Press.

Zangwill, O.L. (1978). "The concept of developmental dysphasia." Developmental Dysphasia (M. A. Wyke, ed.). New York: Academic Press.

Myrna Gopnik is Professor of Linguistics at McGill University (Montreal). Her recent publications include: A Rube Goldberg machine par excellence, Commentary on "Natural language and natural selection" by S. Pinker and P. Bloom in Brain and Behavioral Sciences, (in press); "Feature-blind grammar and dysphasia" Nature, 344, p. 715, April, 1990; "Reflections on Challenges Raised and Questions Asked", in Challenges to Developmental Paradigms:lmplications for theory assessment andpractice ed. P.R. Zelazo and R.G. Barr (Lawrence Erlbaum, Hillsdale, N.J.) 1989; "The Development of Text Competence", in Text and Discourse Connectedness ed. M. Conte, et al. (Benjamins, Amsterdam) 1989; Motor and Sensory Processes of Language ed. with E. Keller, (Lawrence Erlbaum, Hillsdale, N.J.) 1987; "The development of connexity in young children," in Text Connectedness from Psychological Point of View, ed. J.S. Petofi (Helmut Buske Verlag, Hamburg) 1986; "A featureless grammar in a dysphasic child," McGill Working Papers in Linguistics January 1986.


Go to Semiotic Review of Books Home Page
Go to SRB Highlights
Go to SRB Archives