In his 1967 book "Biological Foundations of Language" Eric Lenneberg argued for a critical period in language acquisition on the basis of the evidence from the following populations:
Subsequent supporters of the notion of critical period in language acquisition, such as Pinker, have relied on similar evidence. Not all of the cases cited by Lenneberg are clear cut. Bialystok and Hakuta provide a trenchant critique of Lenneberg's analysis. They point out that the neurologists studying children's recovery of language after brain damage are generally not linguists; they may well miss more subtle alterations in grammatical ability subsequent to brain damage. When this qualification is paired with the observation that the most convincing cases of recovery occur with damage incurred before the age of five, rather than before puberty, this line of evidence becomes questionable as an indicator of a critical period lasting until puberty. This questioning of the puberty cut off also bears on Lenneberg's hypothesis that the language development of children with Down's Syndrome halted at puberty.
The "wild children" evidence has generated the most interest and discussion. The early cases cited by Lenneberg were discussed quite fully in your reading for the last lecture by Sacks. The most complete study of language learning in a previously isolated child is the case of Genie. This poor girl was discovered in Los Angeles at age 13. She had been isolated since she was a few months old. She was tethered to her bed or potty and not spoken to at all. After her rescue she became the focus of great interest on the part of linguists and received intensive speech therapy. Despite the enormous amount of effort invested in her language teaching she was not able to construct grammatical sentences. Interpretation of this case is clouded by the brain damage that resulted from the physical and mental abuse endured by Genie.
Pinker cites a case that he finds more telling: a woman known in the literature as "Chelsea" whose deafness was only discovered at the age of 31. Unlike Genie, Chelsea had not been abused. Nevertheless her language acquisition after hearing aids were fitted was quite similar to Genie's pattern of significant vocabulary acquisition but little grasp of syntax. The implication is that Chelsea had passed the critical period during which acquisition of grammatical structure can occur. However, doubts about Chelsea's case spring to mind. A series of questions occurs to me: Why was Chelsea's deafness not detected?, Did she not make the attempts to communicate that other profoundly deaf children manifest?, Did she not develop any home signs? How well did her hearing aids work? Bialystok and Hakuta direct our attention to the dangers inherent in generalizing from an individual case: the case may be complex and not representative. Instead they take the same tack as Sacks and draw our attention to more wide ranging studies of congenitally Deaf children.
American Deaf children acquire ASL at different ages due to the extent of their exposure to other signers. Elissa Newport (1990) studied the effects of age of acquisition of ASL on the grammatical competence of signers. She found that native learners of ASL, the children of Deaf parents, used more grammatically sophisticated constructions than early learners, who had acquired ASL between 4 and 6 years old. The early learners were, in turn, more grammatically sophisticated than the late learners who had not acquired ASL until after puberty. Consistent with their stance against the notion of a critical period, Bialystok and Hakuta, are quick to point out that some grammatical competencies, such as the treatment of word order, were shown by all the ASL signers studied regardless of the age of acquisition. So there is evidence for some advantage of early learning of a language for subsequent use of grammar, but it is not the case that grammatical competence entirely shuts down after a critical period, as Pinker suggests in chapter 9 of "The Language Instinct".
Bialystok and Hakuta question this strong interpretation of Johnson and Newport's study. They reanalyzed the data and found that it was better described by dividing the group at the age of twenty rather than puberty. When they did this the correlation between age and grammaticality was still high for the under twenty group, but the correlation for the above twenty group increased significantly, showing an age effect subsequent to the supposed point of shut off. Further, they attribute the difference between early and late learners of English in this task to differences in vigilance with age; the task was long, repetitive and made heavy demands on attention, thus Bialystok and Hakuta postulate that the later learners are substantially older and more subject to lapses. These academic squabbles about how to categorize and explain the data illustrate that the simplistic notion of a critical period where the brain is in qualitatively different states of linguistic readiness before and after puberty is too restrictive. Early learners of a language do routinely perform better on tests that tap complex aspects of grammatical competence, but older learners are not incapable of acquiring this sophistication. The large conceptual leap from data such as Johnson & Newport's to a brain-based critical period is not justified. A critical period necessitates the existence of a specialized learning capacity during a fixed period. Whatever age-related decline there might be in language learning is progressive not abrupt.
"The Critical Period hypothesis seems more convincing when applied only to motor aspects of language acquisition. There is a quite considerable body of conclusions based on research on second language acquisition and second/foreign language learning which shows that younger learners are better at acquiring a native-like accent than older ones."Bialystok and Hakuta discuss the work of Oyama (1976) who studied male Italian immigrants who had come to New York between the ages of six and thirty years. She played tapes of them speaking English to native speakers of English and asked them to rate "accentedness" of the speech using a five-point scale, with 1 representing native speech. Ages 6-10 scored 1.3, ages 11-15 scored 2.4 and ages 16-20 scored 3.6. The relationship between age of arrival and overall approximation to a native accent is one of progressive decline, rather than a sharp cut-off. In the title of her paper, Oyama uses the terminology "sensitive period" rather than "critical period"; perhaps this is more appropriate because it doesn't have the same all-or-none connotations.
Other studies have been more specific about the aspects of phonology that are difficult to emulate as a non-native speaker. A key pattern in the findings is that sounds that are similar to those in the native language are more difficult to learn than those that are novel. This is consistent with some of the work on phonemic categorization that I discussed in the beginning of the section on language. The categorical nature of speech sounds that is language specific by the age of one year is retained in the perception of a second language. In other words the categorical distinctions established for the native tongue interfere with the perception and formation of new but similar sounds in the second language. If a child learns two languages simultaneously, the prototypical case of bilingualism, the sound systems are thought to be acquired largely separately. Instances of blends have been noted, such as 'shot' as a blend of 'chaud' and 'hot' from a two-year-old English-French bilingual. Apart from some initial mixing and a few blends, bilingual children acquire the language system of each language side by side and seem to develop a feeling for their distinctiveness.