Page 88 - Encyclopedia of Philosophy of Language

Language and Mind
children—historical oddities such as The Wild Boy of Aveyron and others, who have been restored to normal human contact after a period of living wild—attest to this obvious point (Brown 1958: 189). But the charac- ter of input language to children, and its role, had not been extensively studied. It quickly became clear that CDS, at least among middle-class English-speaking mothers, the usual subjects of enquiry, constitutes an identifiable language variety. It has phonological modifications, nonsegmental and segmental, and specific grammatical characteristics. So, for example, fundamental frequency is higher than comparable speech to adults, and pitch range wider. Pronunciation is said to be more careful and precise than is usual in adult-to-adult conversation. Adults speak to children in short, grammatical sentences. It would be plausible to assume that such modifications would assist the child in the language learning task. It would beequally plausible, however, to interpret the modifications as ways adults have (or learn) to make themselves com- prehensible to small children with limited language capacity. Studies which have attempted to resolve this issue by correlating the effects of variation in syntactic input on development, have had mixed success. One result which has been replicated, originally established by Newport, et al. (1977), concerns the effect of utter-
ance-initial auxiliaries on the child's development of this category. There does appear to be a positive cor- relation. The more a mother uses auxiliaries like can, will, shall, in her utterances to her 18-month-old child, the more likely she is to hear them from the child six months later. While similar examples of established effects of syntactic variation in CDS on language growth in the child are few, research on the effect of discourse modifications—expansions or extensions by the interlocutor which pick up on the child's topic and expand or extend it—suggests that these strategies by mothers may be effective in facilitating language development. What is not clear is how any feature of CDS which turns out to be facilitative in language development actually achieves its effect. It is still necessary to hypothesize a learning mechanism, or a component of it, which can use the relevant input to advance its learning.
2. MethodologyandTheory:TheCurrentPhase
groups of different children at different ages (e.g., a group of three-year olds, one of five-year olds, and a third of seven-year olds) provide the language samples. It is assumed that the linguistic changes one finds between, say, the members of the three- and five- year-old group are similar to those one would find if the same children were sampled at three and again two years later.
2.1 Data Archiving and Theoretical Debates
By the end of the 1970s the field of child language was served by two academic journals, the Journal of Child Language and First Language, entirely devoted to research in the area. It was (see below) about to resume its close relationship with linguistic theory. It was in one sense rich in data, with nearly 20 years of data collection behind it. Data was not, however, readily accessible to researchers other than those associated with the particular project that had generated it, and there seemed no obvious way of aggregating data from different projects. At this point, in a project funded by the MacArthur Foundation, Catharine Snow and Brian MacWhinney founded the Child Language Data Exchange System (MacWhinney and Snow 1985; MacWhinney 1995). This is a computer archive of child language transcript data, with material from more than 20 projects on English (including the Brown and Wells data), and data in addition from Afrikaans, Danish, Dutch, French, German, Hebrew, Hungar- ian, Spanish, Tamil, and Turkish. The archive is available on request to any researcher and affords a facility to test hypotheses systematically against large bodies of quantitative data. This methodological advance leads to new theoretical insights. So, for example, Marcus, et al. (1992) were able to assess the validity of the assumption that the child's over- regularization of past tense represents 'U-shaped' development. The reference is to learning which pro-
ceeds from initial correct forms (irregular pasts such as came,sang, hit) to a period in which, because of the acquisition of the regular past tense rule, the irregular past forms are substituted for by regularized forms (corned, singed, kitted). After this period of uncer- tainty, the child establishes essentially the adult system, with regulars and irregulars correctly differ- entiated. Marcus, et al. review an extensive range of data from the CHILDES database, and establish that, while overregularization rates do vary across children, these forms are in a relatively small minority—usually under 10 percent of all forms. So in reality there is no period of marked U-shaped development. (The U is to be imagined as a graph, with the first tail representing correct performance on irregulars, the trough rep- resenting a large number of errors, and then the
The recording technology of the late 1950s allowed
Brown to make permanent records of the speech of
the children he was investigating. Apart from the
major study by Wells that followed a decade after
Brown, there have been dozens of other studies,
initially on English and then on other languages,
which have used tape-recorded language samples as
the basis for the investigation of grammar construc-
tion. Some of these studies involve longitudinal second tail showing the child's recovery to correct
sampling, where the same children serve as subjects at all stages of the research. Other studies have involved cross-sectional simulations of language development:
performance on both regulars and irregulars). Marcus, et al. interpret the new data as indicating a process of development which advances from rote
66