language
development
Elizabeth University Recent
of California,
research
not necessarily
San Diego, our ability
suggests that domain-specific.
be based on a relatively
That
and perceptual
functions.
includes
neural
simulations
brain
potential
studies
language development focal brain injury,
of normal
Evidence
of language
language
Opinion
of this conclusion
learning,
development,
event-related and studies
of subjects
and contrasting
of
suffering forms
of
retardation.
in Neurobiology
Introduction A critical analysis of recent research in language development requires consideration of two hotly contested issues. Proponents of innateness argue that our ability to acquire a language is determined by genetic factors, and mediated by a form of neural organization that is unique to our species. Proponents of domain specificity argue that this ability is also separate and ‘special’, constituting what Chomsky [l] has termed a ‘mental organ’. Because the first claim has to be true at some level of analysis (i.e. we are the only species that can acquire a language in its full-fledged form; see [a*]), the real debate revolves around the mental organ claim [3,4-l. h-e the mental structures that support language ‘modular’, discontinuous and dissociable from all other perceptual and cognitive systems? Does the brain of a newborn child contain neural structures that are destined to mediate language, and language alone? The domain-specificity view can be contrasted with an approach in which language is viewed as an innate system, but one that involves a reconfiguration of mental and neural systems that exist in other species [5*-7-l, and which continue to serve at least some non-linguistic functions in our own [W] In this review I will consider several areas of research on the biological foundations of language learning, focusing on the case for and against domain specificity.
and domain
but
appears to
that also serve other
in support
specific language impairment,
Current
USA
to learn language is innate,
is, language development
in several clinical populations mental
Innateness
La Jolla, California,
plastic mix of neural systems
cognitive
network
Bates
specificity
Two recent examples illustrate the confusion between innateness and domain specificity. Petitto and Marentette [9**] published an influential paper demonstrating that deaf infants exposed to sign language ‘babble’ with their hands, producing meaningless but systematic actions that are not observed in hearing children. Furthermore, this form of manual babbling occur:, around %lO months
1992, 2:180-185
of age, the point at which vocal babbling appears in the hearing child. The authors conclude that language learning involves innate abilities that are independent of modality (i.e. vocal or manual); they also claim that these abilities are specific to language, providing support for Chomsky’s mental organ claim [ 11. Their first conclusion is clearly supported by the evidence, but the second is not. We have known for more than 100 years that children begin to imitate novel actions (i.e. actions that are not already in their repertoire) at around E&10 months. The more systematic the adult input, the more systematic the child’s imitation is likely to be. Petitto and Marentette’s demonstration of babbling in the visual modality constitutes a particularly beautiful example of this interesting but well established fact. The kind of imitation that underlies babbling is undoubtedly based upon abilities that are innate, and particularly well developed in our species (human children imitate far better and more often than any other primate) [2*,10*], but proof of its existence does not, in itself, constitute evidence in favor of the notion that language is ‘special’. A somewhat different example appeared in a letter to Nuture by Gopnik [ 11-1 (for further details see [12*] ), describing preliminary results from a study of grammatical abilities in a family of individuals suffering from some kind of genetically based disorder (see also [ 13.1). Members of this family had difficulty with particular aspects of grammar, including regular verb inflections (e.g. the ‘-ed’ in verbs like ‘walked’ and ‘kissed’). By contrast, they had less trouble with irregular forms like ‘came’ or ‘gave’. This pattern was offered as an example of an innate and domain-specific disorder, termed ‘feature-blind dysphasia’, and has been cited as evidence in favor of Pinker’s claim that regular and irregular forms are handled by separate mental and perhaps neural mechanisms [ 14**,15*]. Shortly after Gopnik’s letter appeared, Nature published a rebuttal by Vargha-Khadem and Passingham [ 16**] (see also [17-l), who had studied the same family for a num-
Abbreviations MRI-magnetic
180
resonance
imaging; SLlLspecific @
Current
Biology
language impairment;
Ltd ISSN 0959-4388
WMGWilliams
syndrome.
language
ber of years. These authors point out that the members of this family suffer from a much broader range of linguistic deficits than one might conclude from Gopnik’s description. Their peculiar grammatical symptoms are only the tip of an iceberg, one by-product of a disorder with repercussions in many different areas of language and cognition, providing further evidence for innateness, but none for domain-specificity [ 1W]
language
learning
in neural
networks
There is a branch of language acquisition research called ‘learnability theory’ [19*], which uses formal analysis to determine the range of conditions under which different kinds of grammars can (in principle) be learned. Until recently, most of this research had been based upon the assumption that language learning in humans is similar to language learning in serial digital computers, where apriori hypotheses about grammatical rules are tested against strings of input symbols, based on some combination of positive evidence (‘here is a sentence in the target language’), and negative evidence (‘here is a sentence that is not permitted by the target language’). A famous proof by Gold [20] showed that a broad class of grammars (including generative grammars of the sort described by Chomsky) could not be learned by a system of this kind unless negative evidence was available in abundance, or strong innate constraints were placed upon the kinds of hypotheses that the system would consider. As we know that human children are rarely given explicit negative evidence, the learnability theory seems to require the conclusion that children have an extensive store of innate and domain-specific grammatical knowledge. In the past two years, this conclusion has been challenged by major breakthroughs in the application of a different kind of computer architecture (called neural networks, connectionism, and/or parallel distributed processing) to classic problems in language learnability. Because connectionism makes a very different set of assumptions about the way that knowledge is represented and acquired, Gold’s pessimistic conclusions about lan guage learnability do not necessarily apply. This new era began in 1986 with a simulation by Rumelhart and Mc Clelland [21] on the acquisition of the English past tense, showing that connection&t networks go through stages that are very similar to the ones displayed by children who are acquiring English (producing and then recovering from rule-like overgeneralizations like ‘corned’ and ‘wented’, in the absence of negative evidence). This simulation has-been severely criticized (see especially Pinker and Prince [22]). Since then, a number of new works have appeared that get around these criticisms, replieating and extending the Rumelhart-McClelland findings in several new directions [ 18**,23*“-270*]. The most rem cent example comes from Marchman [18**], who has ‘lesioned’ neural networks at various points during learning of the past tense (randomly eliminating between 2 %-44 % of the connections in the network). These simulations capture some classic ‘critical period effects in language learning (e.g. smaller earlier lesions lead to bet-
development
Bates
ter outcomes; later larger lesions lead to persistent problems in grammar), showing that such effects can occur in the absence of ‘special’ maturational constraints (compare with Newport [28*] and Elman [ 24**]). In addition, Marchman’s damaged systems found it more difficult to acquire regular verbs (e.g. ‘walked’) than irregulars (e.g. ‘came’), proving that the specific pattern of deficits described by Gopnik and by Pinker can result from nonspecific forms of brain damage in a general-purpose learning device. Such research on language learning in neural networks is still in its infancy, and we do not know how far it can go. But it promises to be an important tool, helping us to determine just how much innate knowledge has to be in place for certain kinds of learning to occur. / Electrophysiological
studies
.’ , ,L -.T; L,,_’
Z Two laboratories have begun to@produce exciting evidence on the event-related brain potentials associated with stages of language development in the normal child. Molfese [ .Z$P] has uncovered some electrophysiological correlates of the words that l-year-old children do and do not know. Mills, Coffey and Neville [30=*,31*=] have taken the method a step further, demonstrating changes in event-related potentials to familiar and unfamiliar words at different stages in early language development, controlling for chronological age. For example, there are components in the event-related potentials that distinguish lo-month-olds, who have begun to understand words, from IO-month-olds who are still unable to comprehend speech. Additional components in event-related potentials associated with comprehension of familiar words can distinguish between 13.montholds who can and cannot produce those words. In short, the neural systems that mediate language appear to change over time, and as a function of language learning in addition to chronological age.
Children
with focal brain
injury
It has been known for some time that children show better recovery from focal brain injury than adults with analogous lesions [32,33,34-l. As reviewed by Satz et al. [35-l, however, the pendulum has swung back and forth between claims of equipotentiality (i.e. the left hemisphere is in no way specialized for language) and specificity (i.e. left-hemisphere damage always leaves residual damage). New retrospective studies of older children with this etiology suggest that many do show complete recovery in language [36*]; others show relatively subtle deficits, which may be due to ancillary factors [37*]. By contrast, prospective studies of brain-injured infants in the first stages of language learning show significant delays with both left- and right-hemisphere damage [38=,39**,40*], but the pattern of deficits observed in the early stages does not map in any obvious way onto the brain-behavior correlations observed in brain-dam aged adults. For example, comprehension deficits occur
181
182
Cognitive
neuroscience
more often with right-hemisphere injury (and not with damage to Wernicke’s area); expressive deficits are more severe and/or more persistent with left posterior injury (and not with damage to Broca’s area). There is also no straightforward effect of lesion size. Indeed, there is some reason to believe that very large lesions lead to an early ‘compensatory switch’ to the undamaged hemisphere, and hence to better short-term outcomes (see also [31**]). My colleagues and I conclude that there is enormous plasticity in the brain regions that can support language, but that recovery occurs after an initial delay, and the areas responsible for language learning are not necessarily the same regions that mediate use and maintenance of language in adults.
Specific language impairment By definition, specific language impairment (SLI) refers to delays in receptive and/or expressive language development in children with no other known form of neurological or cognitive impairment. Recent studies of XI suggest that this definition may not be accurate. Although these children do not suffer from global forms of mental retardation, they show subtle impairments in aspects of cogmtion and/or perception that are not specific to language. For example, new findings [41-l corroborate claims by Tallal and her colleagues [42] that many children with XI experience difficulty in processing rapid transitions in acoustic information (including non-linguistic stimuli). This may help to explain new studies comparing SLI in English, Italian and Hebrew [43”,44*], showing that the specific areas of grammar that are most delayed vary from one language to another, and that the most vulnerable elements within each language appear to be those that are low in ‘phonological substance’ (i.e. salience). The subtle deficits associated with SLI may also transcend the acoustic modality, affecting certain kinds of manual gesture [45*]. The brain bases of this syndrome are still unknown; magnetic resonance imaging (MRI) studies of SLI show no signs of frank lesions, although subtle differences in symmetry have been uncovered [46~*]. Taken together, these studies suggest that SLI may not be a purely linguistic (or acoustic) phenomenon. Nor is it a congenital variant of the aphasias observed in adults with focal brain injury.
Williams
syndrome
and other ‘savants’
The strongest evidence to date in favor of domain specificity comes from rare cases in which language appears to be remarkably spared despite severe limitations in other cognitive domains. Etiologies associated with this unusual profile include spina bifida and hydrocephalus, and a rare form of mental retardation called Williams syndrome (WMS) [47**], The dissociations observed in WMS prove that language can ‘decouple’ from mental age at some point in development. Nevertheless, recent studies of WMS place constraints on the conclusion that
language is a separate mental system from the beginning. First, it is clear that language development is seriously delayed in infants and pre-school children with WMS, suggesting that certain cognitive prerequisites must be in place before language can be acquired [48]. Second, studies of older children with WMS demonstrate peculiar islands of sparing in some non-linguistic domains (e.g. face recognition), suggesting that their spared language may be based on some deviant form of information processing that operates beyond the boundaries of ‘language proper’. Comparisons of WMS and Down syndrome children matched for mental age suggest that the term ‘mental retardation’ is misleading. Each group shows a distinct and complementary profile of cognitive impairments, reminiscent in many respects of the contrasting profiles shown by adult patients with left- versus right-hemisphere dam age. In spite of this, MRI studies of these two groups yield no evidence of a left-right difference [49*]. Instead, there are group differences in brain morphology along an anterior-posterior axis, including differences in cerebellarqerebral ratios, and differences in the proportional size of neo-cerebellar versus paleo-cerebellar structures These landmark studies challenge traditional ideas about the neural structures that mediate language and related cognitive systems.
Conclusion The research of recent years has led to several interesting conclusions. First, neural network simulations of language learning have demonstrated that a great deal of language can be acquired by a general-purpose learning device. Second, electrophysiological studies have shown that the neural systems subserving language change with development, reflecting language level rather than chronological age. Third, studies of language acquisition in children with focal brain injury demonstrate a great deal of plasticity in the brain regions that can subserve language learning. Fourth, studies of specific language delay suggest that SLI is not so specific after all. Fifth, comparative studies of mental retardation challenge any simple view of the relation between language and cogmtion, and raise new questions about the neural substrate of linguistic and cognitive profiles. A great deal has been learnt in the past few years about the biological foundations of language development; evidence for innateness is good, but evidence for a domain-specific ‘mental organ’ is difficult to find. Instead, language learning appears to be based on a relatively plastic mix of neural systems that also serve other functions,
Acknowledgements This review was visiting scientist Research (CNR), riod came from
prepared while the author was in residence as a at the Institute of Psychology, National Council of Rome, Italy. Partial supportfor this sabbatical pethe CNR and from an NIDCD grant ‘Cross-linguistic
language development Bates studies of aphasia’. I am grateful earlier draft of the manuscript.
References
to V Volterra
1.
CHOMSKY N: Language
bridge,
Massachusetts:
on an
reading
within the annual period of re-
and Problems of Knowledge. MIT Press; 1988.
Cam-
2. .
GREENFIEU)P, SAVAGE-RUMBAIJGH E: Imitation, Grammatical Development and the Invention of Protogrammar by an Ape. In Biological and Behavioral Determinants of Language Development. Edited by Krasnegor N, Rumbaugh D, Schiefelbusch R, StuddertKennedy M. Hillsdale, New Jersey: Erlbaum; 1991: 235-262. Compares the acquisition of symbolic communication in human children and two species of chimpanzee. Performance by the pygmy chimpanzee (bonobo) bears a much closer resemblance to language acquisition in human children, including spontaneous (unreinforced) learn ing and comprehension well in advance of production. Argues for phylogenetic continuity in the evolution of language. 3.
FODOR J: Modularity Press; 1983.
ofMind
Cambridge,
Massachusetts:
MIT
4.
PINKERS, BLOOM P: Natural Language and Natural Selection. Behav Brain Sci 1990, 13:707-784. i review of evidence in favor of the ided that language originally evolved in the service of communicative functions, but eventually achieved the status of an autonomous (modular) cognitive system. 5. DEACONT: Rethinking Mammalian Brain Evolution. Am Zool . 1‘990, 30:629-705. An effort to bring speculation about the origins of primate intelligence into line with recent findings on the ontogeny of neural systems. Raises serious problems for the supposed discontinuity of language from other cognitive systems. 6. .
DEACON T: Brain-Language Coevolution. In 7& Ezlohtion of Human Latzguuges: Proceedings of the Santa Fe Institute Studies in the Sciences of Complexity. Edited by Hawkins JA, GeRMann M. San Francisco: Addison-Wesley; 1990. Arguments are presented in favor of the idea that language depends on pre-existing neural systems that continue to serve other cognitive functions (i.e. no truly ‘new’ neural structures evolved to subserve language).
SERENO M: Language and the Primate Brain [Newsletter]. Ia Jolla: University of California, San Diego, Center for Research in language; 1990: 4.4. Questions the existence of so-called ‘association cortex’, and argues for a view of language as a system that is mediated by a succession of cortical maps of sensorimotor origin.
7. .
E, THAL D, MARCHMANV: Symbols and Syntax: a Darwinian Approach to Language Development. In Bio logical and Behavioral Determinants of I.anguage Dellelopmext. Edited by Ktasnegor N, Rumbaugh D, Schiefelbusch E, StuddertKennedy M. Hillsdale, New Jersey: Erlbaum; 1991: 2965. Various arguments are provided against linguistic autonomy in defense of the hypothesis that ‘language is a new machine that nature con strutted out of old parts.’
8. ..
10.
CHEVAL~EK-SKO~IKOFF S: Spontaneous Toot Use and Sensorimotor Intelligence in Cebus Compared with Other Monkeys and Apes. Behav Brain Sci 1991, 14368. A review of cognitive development in non-human primates, based on Piaget’s theoty of sensorimotor development. Emphasis is placed on similarities and differences between human ontogeny and the developmental course displayed by other primates. .
and recommended
Papers of particular interest, published view, have been highlighted as: . of special interest .. of outstanding interest
for comments
BATES
PETI’~TOL, MARENTETTEPF: Babbling in the Manual Mode: Evidence for the Ontogeny of Language. Science 1331, 251:1493-1499. Deaf infants ‘babble’ with their hands by 10 months of age, suggesting that language learning is based on innate systems that are inherently free of the modality in which language is expressed.
11.
G~PNIK M: Feature-Blind
Grammar and Dysphasia [Letter]. Nature 1990, 344715. .?btief description of a family of individuals who display specific problems with grammar, cited as evidence in favor of an independent neural system that is responsible for particular features of grammatical morphology. 12. GOPNIK M, CRAGO M: Familial Aggregation of a Developmen. tal Language Disorder. Cognition 1991, 39:1-50. Further details on the nature of a genetically-based language disorder with particular implications for aspects of grammar. * TAIUL P, ‘TOWNSENDJ, Cu~nss S, WULFEC~ B: Phenotypic Profiles of Language-Impaired Ctiudr&i- %sed on Genetic/Family History. Brain Lang J991;‘4i81-95. A comparison of language and cognitive &orders in language disordered children with and without a family history of language disorders. Language profiles are quite similar in the two groups, but children with a positive family history show a broader range of academic deficits. 13. .
14.
PINKERS: Rules of Language. Science 1991, 253:530-535.
yreview of evidence in favor of a dual-mechanism approach to Ian guage learning, suggesting that regular verb forms (e.g. walk+walked) are acquired by a rule-based mechanism, while irregular forms (e.g. come+came) are by some more general form of rote learning. KIMJ, PINK&XS, POUNCE A, SANDUPP: Why No Mere Mortal has 15. . Ever Flown Out to Center Field. Cogn Sci 1991, 15:175-218. Detailed evidence in favor of the dual mechanism view described in [14**1.
VARGHA-KIWEM F, PASSINGHAMR: Speech and Language De16. .. fects. Nature 1990, 346:226. A rebuttal of Gopnik’s [ 1 lo*] interpretation of genetically-based Ian guage disorders. The author, who has studied the same family for a number of years, suggests that these individuals suffer from a much broader range of linguistic deficits. FLETCHER P: Speech and Language Defects. Nature 1990, 346:226. cresents arguments against Gopnik’s [ 1 l**] interpretation of a family ial language disorder, suggesting that language delay typically involves a much broader range of symptoms. Points out that Gopnik’s family score below normal, but well above chance, on the grammatical features in question, performance that is incompatible with the term ‘feature blindness’. 17.
18. ..
MARCHMANV: Constraints on Plasticity in a Connectionist Model of the Acquisition of the English Past Tense [Technical Report]. Ia Jolla: University of California, San Diego, Center for Research in Ianguage; 1992: 9201. Describes simulations of language learning in neural networks that have been ‘lesioned’ (random removal of 2 “/o-44 96 of all connections) at different points in the learning process. Replicates many classic ‘critical period’ effects in the language learning development literature, without postulating a maturational change in the nature of the learning device. LIGHTFOOT D: The Child’s Trigger Experience-Degree-O 19. . Bebav Brain Sci 1991, 14:364. Learnability. Presents logical arguments for the innateness of language, based on the idea that key aspects of language are inherently ‘unlearnable’.
20.
GOU, E: Language 1967, 16:447-474.
21.
RUMELHART D, MCCLEUAND J, AND THE PDP RESEARCH GROUP: Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Cambridge, Massachusetts: MIT/Bradford Books; 1986, 1.
9. ..
Identification
in the
Limit.
Inf Control
183
184
Cognitive
22.
neuroscience
PINIER S, PRINCE A: On Language and Connectionism: an Analysis of a Paral!el Distributed Processing Mode! of Language Acquisition. Cognition 1988, 28:73-193.
ELWW J: Finding Structure in Time. Cogn Sci 1990, 23. .. 14:17+211. Describes models of language learning in recurrent networks whose task is to ‘guess the next word’. Demonstrates that key grammatical categories and the sequences that hold among them can be induced from the statistical dependencies that hold among words in a sentence, without prior knowledge about sentence structure. 24. ..
ELMANJ: Incremental Learning, or the Importance of Starting Small. In Proceedings of the Thirteenth Annuul Conference of the Cognitbe Science Society Hi!!sda!e, New Jersey: Erlbaum; 1991: 443-448. An extension of the evidence presented in [23**] to the learning of embedded sentences with long-distance dependencies. Shows that learn ing is possible if the system has a limited memoty span at the early stages, suggesting a new interpretation of critical period phenomena in first and second language learning. MAcWHINNFi
B: Implementations are not ConceptuaIizations: Revising the Verb Learning Model. Cognition 1991, 40:121-157. A detailed extension of the Rumelhart and McClelland [21 I simulation of past tense learning in a mode! that learns several different aspects of English verb morphology simultaneously. Provides a point-by-point response to Pinker and Prince’s [ 221 critique of the Rumelhart and McClelland model, &d successfully simulates many aspects of language leaming in English-speaking children. 25. ..
PL~JNKETT K, MARCHMAN V: U-Shaped Learning and Frequency Effects in a Multi-Layered Perceptron: Implications for Child Language Acquisition. Cognition 1991, 38:4>102. A response to Pinker and Prince’s [22] critique of the Rumelhart and McClelland mode!, showing that many critical aspects of grammatical leaming may depend on the statistical structure of the input (i.e. type and token frequency, phonological similarity, competition among dif~ ferent mappings between the present and the past tense).
26. ..
27. ..
PLUNKETI K, MARC~IMANV: From Rote Learning to SysModels: Prxeedings of the tern Building. In Connectionist 1990 Summer School. Edited by Touretzky DS, Elman J, Sejnowski T, Hinton G. San Mateo, California: Morgan Kaufman; 1991: 201-219. An extension of [26*-l in which new verbs are learned gradually, one at a time. Demonstrates a ‘critical mass’ effect in which a single sysys tern mimics the dual-mechanism approach described by Pinker [ 14**], starting with a memorization strategy and moving on to rule~like behave ior. 28. NEWPORTE: Maturational Constraints on Language Learning. . Cogn Sci 1990, 14:11-28. A review of evidence on first and second language learning, con!irm~ ing the idea that there is a critical period for the acquisition of certain a5pect.s of language. Raises the possibility that children are better at language learning because of their cognitive limitations (i.e. ‘less is more’). MOLFESED: Auditory Evoked Responses Recorded from 16Month-Old Human Infants to Words they Did and Did Not Know. Brain and Language 1990, 38~345~363. Demonstrates differences in the event-related brain potentials for words that the infants do and do not know. 29. .
Acquisition and MIUS D, COFFEY S, NEWLLE H: Language Cerebra! Specialization in 20.Month-Old Children. J Cogn Neurosci 1992, in press. A study of event-related potentials associated with known and unknown words in children who are matched for chronological age but vaty in their language abilities. Specific and distinct components of the events related potential elicited by familiar words were a%ociated with levels of comprehension ability (independently assessed), and another set indexed differences in word production ability in the same children. Developmental changes in the ability to produce words were associated with increased regional specialization of the latter components (lateralized to the left, and localized over temporal and parietal regions). 30. ..
31. ..
MILL$ D, COFFEY S, NEVUE nization in Infancy During
H: Changes in Cerebra! OrgaPrimary Language Acquisition.
In Human BehaLlior and the Det’eloping Brain. Edited by Dawson G, Fischer K. New York: Guilford Publications; 1992: in press. An expansion of the data in [30**] including a discussion of eventrelated potentials in children with early foca! brain injury. 32.
HEC~\ENH: Acquired Aphasia p~&lo&z 1983, 21:581~587.
in Children:
Revisited.
Neuro
33.
ARAM DM: Language Sequelae of Unilateral Brain Lesions in Children. In Language, Commurzication and the Brain. Edited by Plum F. New York: Raven Press; 1988.
~VSKA MJ, LWSKA DJ, HORWITZ SJ, ARAM I>: Presentation, Clinical Course and Outcome of Childhood Stroke. Pediatric Neural 1991, 7:333%341. A review of factors that contribute to positive and negative outcomes in childhood victims of focal brain injury. 34. .
SAW P, STKAUSSE, WHITAKER H: The Ontogeny of Hemispheric Specialization: Some Old Hypotheses Revisited. Brain hng 1990, 38:59&614. A comprehensive review of the plasticity/specificity argument in the literature on recovety from focal brain damage in childhood. 35. .
VAR~jHA-KHADEM F, IsAAcs E, PAPAIELOGM H, POLKEY C, WIISON J: Development of Language in Six Hemispherectomized Patients. Brain 1991, 114:473-495. Compares outcomes in children who were hemispherectomized at different points in development, with children suffering early focal brain injury. The focal lesion sample is not signiIicant!y different from norma! on any language measures, but the hemispherectomized children show some restdua! damage depending on the age at which their problems first appeared and the age at which surgery occurred.
36. .
AR&MD, GIILESPIE L, YAM~~III’I‘A T: Reading Among Children with Left and Right Brain Lesions. Del’ Neuropxychl 1990, 6:27+290. Most children in the sample are indistinguishable from normal on measures of reading. A subgroup (especially those with subcortical involvement) do show deficits in reading, suggesting that a number of neurological factors may contribute to the outcomes of early brain injury. 37. .
MARCHMANV, MILLERR, BATES E: Babble and First Words in Children with Focal Brain Injury. Applied Psycholinguistics 1991, 12:1-22. Case studies of babbling and the passage into first words in five infants with different forms of focal brain injuty. AU children are initially delayed in babbling and meaningful speech; children with left posterior involvement show the slowest recovery from delay. 38. .
TILU DJ. MARCHMAN V, STILESJ, ARAMD, TRALINERD, NASS R, BATES E: Early Lexical Development in Children with Focal Brain Injury. Brain and Language 1991, 40:491-527. A prospective study of the first stages of receptive and expressive language in 27 children with different forms of focal brain injuty. A!! chi!~ dren are initially delayed in expressive language. Receptive de!ays are weakly associated with right hemisphere damage. Expressive de!ays are more protracted in children with left posterior damage. Suggests that the brain regions responsible for early language learning are not necessarily the same regions that mediate language use in adults. 39. ..
40. .
WULFECK
B, TKAUNERD, TALLALP: Neurologic, Cognitive and Linguistic Features of Infants after Focal Brain Injury. Peg diatric Neural 1991, 7:266269. Case studies of language and cognition in children with focal brain it+ juries acquired before the first year of life. Comprehension deficits tend to be associated with right hemisphere damage.
41. .
COHEN H, GEUNASC, LASSONDEM, GEOFFREYG: Auditory Lateralization for Speech in LI Children. Brain and Language 1991, 41:395-401. Children with specific language impairment display deficits in auditory perception, suggesting that linguistic deficits may have a perceptual base. 42.
TAUAL P, STARK R, MELLITSD: Identification of LanguageImpaired Children on the Basis of Rapid Perception and Production Skills. Brain Lang 1985, 25:316322.
language
LEONARD L, B~RTOUNIU, CA.SELLI M, MCCRECDRK, SABBADINI L: Two Accounts of Morphological Deficits in Children with Specific Language Impairment. Langqqe Acquisition 1992, in press. Reviews evidence for two competing explanations for the grammatical deficits displayed by children with specific language impairment. Concludes that a phonologically-based account is superior to theories based on a ‘pure’ grammatical deficit, providing further evidence for perceptual (non-linguistic) origins of the disorder.
43. ..
Deficits in Grammatical Morphology in SpeciIicaIIy Language-Impaired Children: Preliminary Evidence from Hebrew. Clin Linguisf Pbonet 1990, 4:93-105. The grammatical structure of Hebrew is radically different from other languages in which specific language impairment has been studied. Most of the grammar is relatively spared in tlebrew children with XL, although they still show problems with grammatical features that are low in phonological salience. 44. .
ROMA, LEONARD L: Interpreting
THAL D, TOBI.& S, MOIUUS~ND: Language and Gesture in Late Takers: a One-Year Follow-Up. J Speech Hearing Res 1991, 34:604-612. A follow-up of children who were seriously delayed in expressive language between 18-24 months of age. Recovery from delay is associated with good comprehension and symbolic gesture at initial testing. 45. .
46. ..
development
Bates
BELLIIGI IJ, BIHRLEA, N~vtttz H, JERN~CAN T, DOHERTYS: Language, Cognition and Brain Organization in a Neurodevelopmental Disorder. In Der~elopmental Behavioral Neuro science. Edited by Gunnar W, Nelson C. Ilillsdale, New Jersey: Erfbaum; 1991. A review of language, cognition, neuroanatomy and neurophysiology in older children and adolescents with WMS, a form of mental retardation characterized by remarkably spared language together with a few ‘islands’ of proficiency outside the linguistic domain.
47. ..
48.
TRIAL D, BATES E, BELL~ICIIJ: Language and Cognition in Two Children with Wiiiams Syndrome. J Speech Hearing Res 1989, 32:483-500.
Brain Morphology on Magnetic Resonance Images in Williams Syndrome and Down Syndrome. Arch New01 1990, 47:523_553. Although the contrasting behavioral protiles displayed by children with Williams versus Down syndrome resemble the contrasts associated with right versus left hemisphere damage, no left-right differences between these groups were uncovered on MRL Group cliffe~:ryces were obselved along an anterior-posterior dimension, inclttiding p?;portional differences in the structure of the cerebellum. II,, *;/ ., 49. ..
JERNGANT, BELISIGI U: Anomalous
JERNICANT, HESSEUNK MD, SOWELL E, TALLU P: Cerebral
Structure on Magnetic Resonance Imaging in Language and Learning-Impaired Children. Arch New-01 1991, 48:53!+545. Children diagnosed with specific language impairment at age 4 fail to display normal left-right aswntmetries in MRI after age 10.
E Bates, Departments of Psychology and Cognitive Science, University of California, San Diego, La Jolla, California 92093, USA.
185
development
Elizabeth University Recent
of California,
research
not necessarily
San Diego, our ability
suggests that domain-specific.
be based on a relatively
That
and perceptual
functions.
includes
neural
simulations
brain
potential
studies
language development focal brain injury,
of normal
Evidence
of language
language
Opinion
of this conclusion
learning,
development,
event-related and studies
of subjects
and contrasting
of
suffering forms
of
retardation.
in Neurobiology
Introduction A critical analysis of recent research in language development requires consideration of two hotly contested issues. Proponents of innateness argue that our ability to acquire a language is determined by genetic factors, and mediated by a form of neural organization that is unique to our species. Proponents of domain specificity argue that this ability is also separate and ‘special’, constituting what Chomsky [l] has termed a ‘mental organ’. Because the first claim has to be true at some level of analysis (i.e. we are the only species that can acquire a language in its full-fledged form; see [a*]), the real debate revolves around the mental organ claim [3,4-l. h-e the mental structures that support language ‘modular’, discontinuous and dissociable from all other perceptual and cognitive systems? Does the brain of a newborn child contain neural structures that are destined to mediate language, and language alone? The domain-specificity view can be contrasted with an approach in which language is viewed as an innate system, but one that involves a reconfiguration of mental and neural systems that exist in other species [5*-7-l, and which continue to serve at least some non-linguistic functions in our own [W] In this review I will consider several areas of research on the biological foundations of language learning, focusing on the case for and against domain specificity.
and domain
but
appears to
that also serve other
in support
specific language impairment,
Current
USA
to learn language is innate,
is, language development
in several clinical populations mental
Innateness
La Jolla, California,
plastic mix of neural systems
cognitive
network
Bates
specificity
Two recent examples illustrate the confusion between innateness and domain specificity. Petitto and Marentette [9**] published an influential paper demonstrating that deaf infants exposed to sign language ‘babble’ with their hands, producing meaningless but systematic actions that are not observed in hearing children. Furthermore, this form of manual babbling occur:, around %lO months
1992, 2:180-185
of age, the point at which vocal babbling appears in the hearing child. The authors conclude that language learning involves innate abilities that are independent of modality (i.e. vocal or manual); they also claim that these abilities are specific to language, providing support for Chomsky’s mental organ claim [ 11. Their first conclusion is clearly supported by the evidence, but the second is not. We have known for more than 100 years that children begin to imitate novel actions (i.e. actions that are not already in their repertoire) at around E&10 months. The more systematic the adult input, the more systematic the child’s imitation is likely to be. Petitto and Marentette’s demonstration of babbling in the visual modality constitutes a particularly beautiful example of this interesting but well established fact. The kind of imitation that underlies babbling is undoubtedly based upon abilities that are innate, and particularly well developed in our species (human children imitate far better and more often than any other primate) [2*,10*], but proof of its existence does not, in itself, constitute evidence in favor of the notion that language is ‘special’. A somewhat different example appeared in a letter to Nuture by Gopnik [ 11-1 (for further details see [12*] ), describing preliminary results from a study of grammatical abilities in a family of individuals suffering from some kind of genetically based disorder (see also [ 13.1). Members of this family had difficulty with particular aspects of grammar, including regular verb inflections (e.g. the ‘-ed’ in verbs like ‘walked’ and ‘kissed’). By contrast, they had less trouble with irregular forms like ‘came’ or ‘gave’. This pattern was offered as an example of an innate and domain-specific disorder, termed ‘feature-blind dysphasia’, and has been cited as evidence in favor of Pinker’s claim that regular and irregular forms are handled by separate mental and perhaps neural mechanisms [ 14**,15*]. Shortly after Gopnik’s letter appeared, Nature published a rebuttal by Vargha-Khadem and Passingham [ 16**] (see also [17-l), who had studied the same family for a num-
Abbreviations MRI-magnetic
180
resonance
imaging; SLlLspecific @
Current
Biology
language impairment;
Ltd ISSN 0959-4388
WMGWilliams
syndrome.
language
ber of years. These authors point out that the members of this family suffer from a much broader range of linguistic deficits than one might conclude from Gopnik’s description. Their peculiar grammatical symptoms are only the tip of an iceberg, one by-product of a disorder with repercussions in many different areas of language and cognition, providing further evidence for innateness, but none for domain-specificity [ 1W]
language
learning
in neural
networks
There is a branch of language acquisition research called ‘learnability theory’ [19*], which uses formal analysis to determine the range of conditions under which different kinds of grammars can (in principle) be learned. Until recently, most of this research had been based upon the assumption that language learning in humans is similar to language learning in serial digital computers, where apriori hypotheses about grammatical rules are tested against strings of input symbols, based on some combination of positive evidence (‘here is a sentence in the target language’), and negative evidence (‘here is a sentence that is not permitted by the target language’). A famous proof by Gold [20] showed that a broad class of grammars (including generative grammars of the sort described by Chomsky) could not be learned by a system of this kind unless negative evidence was available in abundance, or strong innate constraints were placed upon the kinds of hypotheses that the system would consider. As we know that human children are rarely given explicit negative evidence, the learnability theory seems to require the conclusion that children have an extensive store of innate and domain-specific grammatical knowledge. In the past two years, this conclusion has been challenged by major breakthroughs in the application of a different kind of computer architecture (called neural networks, connectionism, and/or parallel distributed processing) to classic problems in language learnability. Because connectionism makes a very different set of assumptions about the way that knowledge is represented and acquired, Gold’s pessimistic conclusions about lan guage learnability do not necessarily apply. This new era began in 1986 with a simulation by Rumelhart and Mc Clelland [21] on the acquisition of the English past tense, showing that connection&t networks go through stages that are very similar to the ones displayed by children who are acquiring English (producing and then recovering from rule-like overgeneralizations like ‘corned’ and ‘wented’, in the absence of negative evidence). This simulation has-been severely criticized (see especially Pinker and Prince [22]). Since then, a number of new works have appeared that get around these criticisms, replieating and extending the Rumelhart-McClelland findings in several new directions [ 18**,23*“-270*]. The most rem cent example comes from Marchman [18**], who has ‘lesioned’ neural networks at various points during learning of the past tense (randomly eliminating between 2 %-44 % of the connections in the network). These simulations capture some classic ‘critical period effects in language learning (e.g. smaller earlier lesions lead to bet-
development
Bates
ter outcomes; later larger lesions lead to persistent problems in grammar), showing that such effects can occur in the absence of ‘special’ maturational constraints (compare with Newport [28*] and Elman [ 24**]). In addition, Marchman’s damaged systems found it more difficult to acquire regular verbs (e.g. ‘walked’) than irregulars (e.g. ‘came’), proving that the specific pattern of deficits described by Gopnik and by Pinker can result from nonspecific forms of brain damage in a general-purpose learning device. Such research on language learning in neural networks is still in its infancy, and we do not know how far it can go. But it promises to be an important tool, helping us to determine just how much innate knowledge has to be in place for certain kinds of learning to occur. / Electrophysiological
studies
.’ , ,L -.T; L,,_’
Z Two laboratories have begun to@produce exciting evidence on the event-related brain potentials associated with stages of language development in the normal child. Molfese [ .Z$P] has uncovered some electrophysiological correlates of the words that l-year-old children do and do not know. Mills, Coffey and Neville [30=*,31*=] have taken the method a step further, demonstrating changes in event-related potentials to familiar and unfamiliar words at different stages in early language development, controlling for chronological age. For example, there are components in the event-related potentials that distinguish lo-month-olds, who have begun to understand words, from IO-month-olds who are still unable to comprehend speech. Additional components in event-related potentials associated with comprehension of familiar words can distinguish between 13.montholds who can and cannot produce those words. In short, the neural systems that mediate language appear to change over time, and as a function of language learning in addition to chronological age.
Children
with focal brain
injury
It has been known for some time that children show better recovery from focal brain injury than adults with analogous lesions [32,33,34-l. As reviewed by Satz et al. [35-l, however, the pendulum has swung back and forth between claims of equipotentiality (i.e. the left hemisphere is in no way specialized for language) and specificity (i.e. left-hemisphere damage always leaves residual damage). New retrospective studies of older children with this etiology suggest that many do show complete recovery in language [36*]; others show relatively subtle deficits, which may be due to ancillary factors [37*]. By contrast, prospective studies of brain-injured infants in the first stages of language learning show significant delays with both left- and right-hemisphere damage [38=,39**,40*], but the pattern of deficits observed in the early stages does not map in any obvious way onto the brain-behavior correlations observed in brain-dam aged adults. For example, comprehension deficits occur
181
182
Cognitive
neuroscience
more often with right-hemisphere injury (and not with damage to Wernicke’s area); expressive deficits are more severe and/or more persistent with left posterior injury (and not with damage to Broca’s area). There is also no straightforward effect of lesion size. Indeed, there is some reason to believe that very large lesions lead to an early ‘compensatory switch’ to the undamaged hemisphere, and hence to better short-term outcomes (see also [31**]). My colleagues and I conclude that there is enormous plasticity in the brain regions that can support language, but that recovery occurs after an initial delay, and the areas responsible for language learning are not necessarily the same regions that mediate use and maintenance of language in adults.
Specific language impairment By definition, specific language impairment (SLI) refers to delays in receptive and/or expressive language development in children with no other known form of neurological or cognitive impairment. Recent studies of XI suggest that this definition may not be accurate. Although these children do not suffer from global forms of mental retardation, they show subtle impairments in aspects of cogmtion and/or perception that are not specific to language. For example, new findings [41-l corroborate claims by Tallal and her colleagues [42] that many children with XI experience difficulty in processing rapid transitions in acoustic information (including non-linguistic stimuli). This may help to explain new studies comparing SLI in English, Italian and Hebrew [43”,44*], showing that the specific areas of grammar that are most delayed vary from one language to another, and that the most vulnerable elements within each language appear to be those that are low in ‘phonological substance’ (i.e. salience). The subtle deficits associated with SLI may also transcend the acoustic modality, affecting certain kinds of manual gesture [45*]. The brain bases of this syndrome are still unknown; magnetic resonance imaging (MRI) studies of SLI show no signs of frank lesions, although subtle differences in symmetry have been uncovered [46~*]. Taken together, these studies suggest that SLI may not be a purely linguistic (or acoustic) phenomenon. Nor is it a congenital variant of the aphasias observed in adults with focal brain injury.
Williams
syndrome
and other ‘savants’
The strongest evidence to date in favor of domain specificity comes from rare cases in which language appears to be remarkably spared despite severe limitations in other cognitive domains. Etiologies associated with this unusual profile include spina bifida and hydrocephalus, and a rare form of mental retardation called Williams syndrome (WMS) [47**], The dissociations observed in WMS prove that language can ‘decouple’ from mental age at some point in development. Nevertheless, recent studies of WMS place constraints on the conclusion that
language is a separate mental system from the beginning. First, it is clear that language development is seriously delayed in infants and pre-school children with WMS, suggesting that certain cognitive prerequisites must be in place before language can be acquired [48]. Second, studies of older children with WMS demonstrate peculiar islands of sparing in some non-linguistic domains (e.g. face recognition), suggesting that their spared language may be based on some deviant form of information processing that operates beyond the boundaries of ‘language proper’. Comparisons of WMS and Down syndrome children matched for mental age suggest that the term ‘mental retardation’ is misleading. Each group shows a distinct and complementary profile of cognitive impairments, reminiscent in many respects of the contrasting profiles shown by adult patients with left- versus right-hemisphere dam age. In spite of this, MRI studies of these two groups yield no evidence of a left-right difference [49*]. Instead, there are group differences in brain morphology along an anterior-posterior axis, including differences in cerebellarqerebral ratios, and differences in the proportional size of neo-cerebellar versus paleo-cerebellar structures These landmark studies challenge traditional ideas about the neural structures that mediate language and related cognitive systems.
Conclusion The research of recent years has led to several interesting conclusions. First, neural network simulations of language learning have demonstrated that a great deal of language can be acquired by a general-purpose learning device. Second, electrophysiological studies have shown that the neural systems subserving language change with development, reflecting language level rather than chronological age. Third, studies of language acquisition in children with focal brain injury demonstrate a great deal of plasticity in the brain regions that can subserve language learning. Fourth, studies of specific language delay suggest that SLI is not so specific after all. Fifth, comparative studies of mental retardation challenge any simple view of the relation between language and cogmtion, and raise new questions about the neural substrate of linguistic and cognitive profiles. A great deal has been learnt in the past few years about the biological foundations of language development; evidence for innateness is good, but evidence for a domain-specific ‘mental organ’ is difficult to find. Instead, language learning appears to be based on a relatively plastic mix of neural systems that also serve other functions,
Acknowledgements This review was visiting scientist Research (CNR), riod came from
prepared while the author was in residence as a at the Institute of Psychology, National Council of Rome, Italy. Partial supportfor this sabbatical pethe CNR and from an NIDCD grant ‘Cross-linguistic
language development Bates studies of aphasia’. I am grateful earlier draft of the manuscript.
References
to V Volterra
1.
CHOMSKY N: Language
bridge,
Massachusetts:
on an
reading
within the annual period of re-
and Problems of Knowledge. MIT Press; 1988.
Cam-
2. .
GREENFIEU)P, SAVAGE-RUMBAIJGH E: Imitation, Grammatical Development and the Invention of Protogrammar by an Ape. In Biological and Behavioral Determinants of Language Development. Edited by Krasnegor N, Rumbaugh D, Schiefelbusch R, StuddertKennedy M. Hillsdale, New Jersey: Erlbaum; 1991: 235-262. Compares the acquisition of symbolic communication in human children and two species of chimpanzee. Performance by the pygmy chimpanzee (bonobo) bears a much closer resemblance to language acquisition in human children, including spontaneous (unreinforced) learn ing and comprehension well in advance of production. Argues for phylogenetic continuity in the evolution of language. 3.
FODOR J: Modularity Press; 1983.
ofMind
Cambridge,
Massachusetts:
MIT
4.
PINKERS, BLOOM P: Natural Language and Natural Selection. Behav Brain Sci 1990, 13:707-784. i review of evidence in favor of the ided that language originally evolved in the service of communicative functions, but eventually achieved the status of an autonomous (modular) cognitive system. 5. DEACONT: Rethinking Mammalian Brain Evolution. Am Zool . 1‘990, 30:629-705. An effort to bring speculation about the origins of primate intelligence into line with recent findings on the ontogeny of neural systems. Raises serious problems for the supposed discontinuity of language from other cognitive systems. 6. .
DEACON T: Brain-Language Coevolution. In 7& Ezlohtion of Human Latzguuges: Proceedings of the Santa Fe Institute Studies in the Sciences of Complexity. Edited by Hawkins JA, GeRMann M. San Francisco: Addison-Wesley; 1990. Arguments are presented in favor of the idea that language depends on pre-existing neural systems that continue to serve other cognitive functions (i.e. no truly ‘new’ neural structures evolved to subserve language).
SERENO M: Language and the Primate Brain [Newsletter]. Ia Jolla: University of California, San Diego, Center for Research in language; 1990: 4.4. Questions the existence of so-called ‘association cortex’, and argues for a view of language as a system that is mediated by a succession of cortical maps of sensorimotor origin.
7. .
E, THAL D, MARCHMANV: Symbols and Syntax: a Darwinian Approach to Language Development. In Bio logical and Behavioral Determinants of I.anguage Dellelopmext. Edited by Ktasnegor N, Rumbaugh D, Schiefelbusch E, StuddertKennedy M. Hillsdale, New Jersey: Erlbaum; 1991: 2965. Various arguments are provided against linguistic autonomy in defense of the hypothesis that ‘language is a new machine that nature con strutted out of old parts.’
8. ..
10.
CHEVAL~EK-SKO~IKOFF S: Spontaneous Toot Use and Sensorimotor Intelligence in Cebus Compared with Other Monkeys and Apes. Behav Brain Sci 1991, 14368. A review of cognitive development in non-human primates, based on Piaget’s theoty of sensorimotor development. Emphasis is placed on similarities and differences between human ontogeny and the developmental course displayed by other primates. .
and recommended
Papers of particular interest, published view, have been highlighted as: . of special interest .. of outstanding interest
for comments
BATES
PETI’~TOL, MARENTETTEPF: Babbling in the Manual Mode: Evidence for the Ontogeny of Language. Science 1331, 251:1493-1499. Deaf infants ‘babble’ with their hands by 10 months of age, suggesting that language learning is based on innate systems that are inherently free of the modality in which language is expressed.
11.
G~PNIK M: Feature-Blind
Grammar and Dysphasia [Letter]. Nature 1990, 344715. .?btief description of a family of individuals who display specific problems with grammar, cited as evidence in favor of an independent neural system that is responsible for particular features of grammatical morphology. 12. GOPNIK M, CRAGO M: Familial Aggregation of a Developmen. tal Language Disorder. Cognition 1991, 39:1-50. Further details on the nature of a genetically-based language disorder with particular implications for aspects of grammar. * TAIUL P, ‘TOWNSENDJ, Cu~nss S, WULFEC~ B: Phenotypic Profiles of Language-Impaired Ctiudr&i- %sed on Genetic/Family History. Brain Lang J991;‘4i81-95. A comparison of language and cognitive &orders in language disordered children with and without a family history of language disorders. Language profiles are quite similar in the two groups, but children with a positive family history show a broader range of academic deficits. 13. .
14.
PINKERS: Rules of Language. Science 1991, 253:530-535.
yreview of evidence in favor of a dual-mechanism approach to Ian guage learning, suggesting that regular verb forms (e.g. walk+walked) are acquired by a rule-based mechanism, while irregular forms (e.g. come+came) are by some more general form of rote learning. KIMJ, PINK&XS, POUNCE A, SANDUPP: Why No Mere Mortal has 15. . Ever Flown Out to Center Field. Cogn Sci 1991, 15:175-218. Detailed evidence in favor of the dual mechanism view described in [14**1.
VARGHA-KIWEM F, PASSINGHAMR: Speech and Language De16. .. fects. Nature 1990, 346:226. A rebuttal of Gopnik’s [ 1 lo*] interpretation of genetically-based Ian guage disorders. The author, who has studied the same family for a number of years, suggests that these individuals suffer from a much broader range of linguistic deficits. FLETCHER P: Speech and Language Defects. Nature 1990, 346:226. cresents arguments against Gopnik’s [ 1 l**] interpretation of a family ial language disorder, suggesting that language delay typically involves a much broader range of symptoms. Points out that Gopnik’s family score below normal, but well above chance, on the grammatical features in question, performance that is incompatible with the term ‘feature blindness’. 17.
18. ..
MARCHMANV: Constraints on Plasticity in a Connectionist Model of the Acquisition of the English Past Tense [Technical Report]. Ia Jolla: University of California, San Diego, Center for Research in Ianguage; 1992: 9201. Describes simulations of language learning in neural networks that have been ‘lesioned’ (random removal of 2 “/o-44 96 of all connections) at different points in the learning process. Replicates many classic ‘critical period’ effects in the language learning development literature, without postulating a maturational change in the nature of the learning device. LIGHTFOOT D: The Child’s Trigger Experience-Degree-O 19. . Bebav Brain Sci 1991, 14:364. Learnability. Presents logical arguments for the innateness of language, based on the idea that key aspects of language are inherently ‘unlearnable’.
20.
GOU, E: Language 1967, 16:447-474.
21.
RUMELHART D, MCCLEUAND J, AND THE PDP RESEARCH GROUP: Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Cambridge, Massachusetts: MIT/Bradford Books; 1986, 1.
9. ..
Identification
in the
Limit.
Inf Control
183
184
Cognitive
22.
neuroscience
PINIER S, PRINCE A: On Language and Connectionism: an Analysis of a Paral!el Distributed Processing Mode! of Language Acquisition. Cognition 1988, 28:73-193.
ELWW J: Finding Structure in Time. Cogn Sci 1990, 23. .. 14:17+211. Describes models of language learning in recurrent networks whose task is to ‘guess the next word’. Demonstrates that key grammatical categories and the sequences that hold among them can be induced from the statistical dependencies that hold among words in a sentence, without prior knowledge about sentence structure. 24. ..
ELMANJ: Incremental Learning, or the Importance of Starting Small. In Proceedings of the Thirteenth Annuul Conference of the Cognitbe Science Society Hi!!sda!e, New Jersey: Erlbaum; 1991: 443-448. An extension of the evidence presented in [23**] to the learning of embedded sentences with long-distance dependencies. Shows that learn ing is possible if the system has a limited memoty span at the early stages, suggesting a new interpretation of critical period phenomena in first and second language learning. MAcWHINNFi
B: Implementations are not ConceptuaIizations: Revising the Verb Learning Model. Cognition 1991, 40:121-157. A detailed extension of the Rumelhart and McClelland [21 I simulation of past tense learning in a mode! that learns several different aspects of English verb morphology simultaneously. Provides a point-by-point response to Pinker and Prince’s [ 221 critique of the Rumelhart and McClelland model, &d successfully simulates many aspects of language leaming in English-speaking children. 25. ..
PL~JNKETT K, MARCHMAN V: U-Shaped Learning and Frequency Effects in a Multi-Layered Perceptron: Implications for Child Language Acquisition. Cognition 1991, 38:4>102. A response to Pinker and Prince’s [22] critique of the Rumelhart and McClelland mode!, showing that many critical aspects of grammatical leaming may depend on the statistical structure of the input (i.e. type and token frequency, phonological similarity, competition among dif~ ferent mappings between the present and the past tense).
26. ..
27. ..
PLUNKETI K, MARC~IMANV: From Rote Learning to SysModels: Prxeedings of the tern Building. In Connectionist 1990 Summer School. Edited by Touretzky DS, Elman J, Sejnowski T, Hinton G. San Mateo, California: Morgan Kaufman; 1991: 201-219. An extension of [26*-l in which new verbs are learned gradually, one at a time. Demonstrates a ‘critical mass’ effect in which a single sysys tern mimics the dual-mechanism approach described by Pinker [ 14**], starting with a memorization strategy and moving on to rule~like behave ior. 28. NEWPORTE: Maturational Constraints on Language Learning. . Cogn Sci 1990, 14:11-28. A review of evidence on first and second language learning, con!irm~ ing the idea that there is a critical period for the acquisition of certain a5pect.s of language. Raises the possibility that children are better at language learning because of their cognitive limitations (i.e. ‘less is more’). MOLFESED: Auditory Evoked Responses Recorded from 16Month-Old Human Infants to Words they Did and Did Not Know. Brain and Language 1990, 38~345~363. Demonstrates differences in the event-related brain potentials for words that the infants do and do not know. 29. .
Acquisition and MIUS D, COFFEY S, NEWLLE H: Language Cerebra! Specialization in 20.Month-Old Children. J Cogn Neurosci 1992, in press. A study of event-related potentials associated with known and unknown words in children who are matched for chronological age but vaty in their language abilities. Specific and distinct components of the events related potential elicited by familiar words were a%ociated with levels of comprehension ability (independently assessed), and another set indexed differences in word production ability in the same children. Developmental changes in the ability to produce words were associated with increased regional specialization of the latter components (lateralized to the left, and localized over temporal and parietal regions). 30. ..
31. ..
MILL$ D, COFFEY S, NEVUE nization in Infancy During
H: Changes in Cerebra! OrgaPrimary Language Acquisition.
In Human BehaLlior and the Det’eloping Brain. Edited by Dawson G, Fischer K. New York: Guilford Publications; 1992: in press. An expansion of the data in [30**] including a discussion of eventrelated potentials in children with early foca! brain injury. 32.
HEC~\ENH: Acquired Aphasia p~&lo&z 1983, 21:581~587.
in Children:
Revisited.
Neuro
33.
ARAM DM: Language Sequelae of Unilateral Brain Lesions in Children. In Language, Commurzication and the Brain. Edited by Plum F. New York: Raven Press; 1988.
~VSKA MJ, LWSKA DJ, HORWITZ SJ, ARAM I>: Presentation, Clinical Course and Outcome of Childhood Stroke. Pediatric Neural 1991, 7:333%341. A review of factors that contribute to positive and negative outcomes in childhood victims of focal brain injury. 34. .
SAW P, STKAUSSE, WHITAKER H: The Ontogeny of Hemispheric Specialization: Some Old Hypotheses Revisited. Brain hng 1990, 38:59&614. A comprehensive review of the plasticity/specificity argument in the literature on recovety from focal brain damage in childhood. 35. .
VAR~jHA-KHADEM F, IsAAcs E, PAPAIELOGM H, POLKEY C, WIISON J: Development of Language in Six Hemispherectomized Patients. Brain 1991, 114:473-495. Compares outcomes in children who were hemispherectomized at different points in development, with children suffering early focal brain injury. The focal lesion sample is not signiIicant!y different from norma! on any language measures, but the hemispherectomized children show some restdua! damage depending on the age at which their problems first appeared and the age at which surgery occurred.
36. .
AR&MD, GIILESPIE L, YAM~~III’I‘A T: Reading Among Children with Left and Right Brain Lesions. Del’ Neuropxychl 1990, 6:27+290. Most children in the sample are indistinguishable from normal on measures of reading. A subgroup (especially those with subcortical involvement) do show deficits in reading, suggesting that a number of neurological factors may contribute to the outcomes of early brain injury. 37. .
MARCHMANV, MILLERR, BATES E: Babble and First Words in Children with Focal Brain Injury. Applied Psycholinguistics 1991, 12:1-22. Case studies of babbling and the passage into first words in five infants with different forms of focal brain injuty. AU children are initially delayed in babbling and meaningful speech; children with left posterior involvement show the slowest recovery from delay. 38. .
TILU DJ. MARCHMAN V, STILESJ, ARAMD, TRALINERD, NASS R, BATES E: Early Lexical Development in Children with Focal Brain Injury. Brain and Language 1991, 40:491-527. A prospective study of the first stages of receptive and expressive language in 27 children with different forms of focal brain injuty. A!! chi!~ dren are initially delayed in expressive language. Receptive de!ays are weakly associated with right hemisphere damage. Expressive de!ays are more protracted in children with left posterior damage. Suggests that the brain regions responsible for early language learning are not necessarily the same regions that mediate language use in adults. 39. ..
40. .
WULFECK
B, TKAUNERD, TALLALP: Neurologic, Cognitive and Linguistic Features of Infants after Focal Brain Injury. Peg diatric Neural 1991, 7:266269. Case studies of language and cognition in children with focal brain it+ juries acquired before the first year of life. Comprehension deficits tend to be associated with right hemisphere damage.
41. .
COHEN H, GEUNASC, LASSONDEM, GEOFFREYG: Auditory Lateralization for Speech in LI Children. Brain and Language 1991, 41:395-401. Children with specific language impairment display deficits in auditory perception, suggesting that linguistic deficits may have a perceptual base. 42.
TAUAL P, STARK R, MELLITSD: Identification of LanguageImpaired Children on the Basis of Rapid Perception and Production Skills. Brain Lang 1985, 25:316322.
language
LEONARD L, B~RTOUNIU, CA.SELLI M, MCCRECDRK, SABBADINI L: Two Accounts of Morphological Deficits in Children with Specific Language Impairment. Langqqe Acquisition 1992, in press. Reviews evidence for two competing explanations for the grammatical deficits displayed by children with specific language impairment. Concludes that a phonologically-based account is superior to theories based on a ‘pure’ grammatical deficit, providing further evidence for perceptual (non-linguistic) origins of the disorder.
43. ..
Deficits in Grammatical Morphology in SpeciIicaIIy Language-Impaired Children: Preliminary Evidence from Hebrew. Clin Linguisf Pbonet 1990, 4:93-105. The grammatical structure of Hebrew is radically different from other languages in which specific language impairment has been studied. Most of the grammar is relatively spared in tlebrew children with XL, although they still show problems with grammatical features that are low in phonological salience. 44. .
ROMA, LEONARD L: Interpreting
THAL D, TOBI.& S, MOIUUS~ND: Language and Gesture in Late Takers: a One-Year Follow-Up. J Speech Hearing Res 1991, 34:604-612. A follow-up of children who were seriously delayed in expressive language between 18-24 months of age. Recovery from delay is associated with good comprehension and symbolic gesture at initial testing. 45. .
46. ..
development
Bates
BELLIIGI IJ, BIHRLEA, N~vtttz H, JERN~CAN T, DOHERTYS: Language, Cognition and Brain Organization in a Neurodevelopmental Disorder. In Der~elopmental Behavioral Neuro science. Edited by Gunnar W, Nelson C. Ilillsdale, New Jersey: Erfbaum; 1991. A review of language, cognition, neuroanatomy and neurophysiology in older children and adolescents with WMS, a form of mental retardation characterized by remarkably spared language together with a few ‘islands’ of proficiency outside the linguistic domain.
47. ..
48.
TRIAL D, BATES E, BELL~ICIIJ: Language and Cognition in Two Children with Wiiiams Syndrome. J Speech Hearing Res 1989, 32:483-500.
Brain Morphology on Magnetic Resonance Images in Williams Syndrome and Down Syndrome. Arch New01 1990, 47:523_553. Although the contrasting behavioral protiles displayed by children with Williams versus Down syndrome resemble the contrasts associated with right versus left hemisphere damage, no left-right differences between these groups were uncovered on MRL Group cliffe~:ryces were obselved along an anterior-posterior dimension, inclttiding p?;portional differences in the structure of the cerebellum. II,, *;/ ., 49. ..
JERNGANT, BELISIGI U: Anomalous
JERNICANT, HESSEUNK MD, SOWELL E, TALLU P: Cerebral
Structure on Magnetic Resonance Imaging in Language and Learning-Impaired Children. Arch New-01 1991, 48:53!+545. Children diagnosed with specific language impairment at age 4 fail to display normal left-right aswntmetries in MRI after age 10.
E Bates, Departments of Psychology and Cognitive Science, University of California, San Diego, La Jolla, California 92093, USA.
185
