Finding relations between input and outcome in language acquisition

James L. Morgan
1996 Developmental Psychology  
and M. Mark (1996) raised numerous points of disagreement with the analyses of J. L. Morgan, K. Bonamo, and L. L. Travis (1995) concerning children's use of parental recasts, responses potentially providing negative evidence. Claiming to show that the bivariate time-series analyses used by Morgan et al. were inherently flawed, Bohannon et al. argued that training studies offer a preferable means for examining contributions of language input. In this article, multiple weaknesses of training
more » ... es of training studies are noted; at best, such studies may yield information on the sufficiency of aspects of language input but can yield none on their necessity. The failure of Bohannon et al.'s time-series analyses to distinguish among varying models of recast function is shown to be attributable to confounding of parameters and idiosyncratic assumptions adopted in generating simulated data from these models. Contrary to Bohannon et al.'s assertions, bivariate time-series analyses of observational data may provide invaluable tools for discerning signs and magnitudes of relations among variables in language development. A child cries out at birth and would doubtless in any case after a time take to gurgling and babbling, but the particular language he learns is entirely a matter of environment. . . . The child learns to speak like the persons round him. (Bloomfield, 1933, p. 43) Bloomfield's (1933) observations form part of the foundation of every theory of language acquisition. Because children learn only those languages to which they are exposed, denning the nature of this exposure is a necessary component of any account of how acquisition proceeds. But it may also be observed that, among organisms, only children succeed in learning language. Determining the nature of the perceptual, representational, computational, and linguistic capacities that enable children to perform this feat is thus also a necessary component of any theory of acquisition. It is clear that language develops as a result of interactions between structure in the environment and structure in the language learner. A central problem for theories of language acquisition concerns how to apportion explanatory responsibility for the structure apparent in acquired grammars to these two sources. Over the past several years, theorizing has perhaps tended to focus lopsidedly on defining linguistic constraints that might subserve acquisition. Several recent developments, however, augur increasing attention to the role of language input. These include the widespread availability of databases, including both parental and child speech (e.g., MacWhinney, 1991; also, Bloomfield, 1933, noted, "Almost nothing is known because observers report what the child says, but not what it has heard" [ p. 512 ]), the advent of rich, multidimensional conceptions of information available in input (e.g., Morgan, Shi, & Allopenna, 1996) , and the development of computational models for assessing possible contributions of such information to learning
doi:10.1037/0012-1649.32.3.556 fatcat:hfo3m6sq45ednflyvhfwfuickm