Using computational modeling in language acquisition research

Abstract

Language acquisition research is oen concerned with questions of what, when, and how – what children know, when they know it, and how they learn it. eoretical research traditionally yields the what – the knowledge that children attain. For instance, this includes how many vowel phonemes the language has, how the plural is formed, and if the verb comes before or aer the object. ese and many other questions must be answered before the child can speak the language natively. is linguistic knowledge is the child's goal. Experimental work traditionally provides the when – at what point in development the child attains particular knowledge about the language. Of course, there is a certain logical trajectory. It would be dicult to discover how the past tense is formed before being able to identify individual words in uent speech. Still, this logical trajec-tory does not oer precise ages of acquisition. Experimental work can, for example, pinpoint when word segmentation occurs reliably and when English children correctly produce past tense forms. is gives us the time course of language acquisition. e child can segment words reliably by this age, and apply regular past tense morphology by that age, and so on. en, there is the how – how the child learns the appropriate what by the appropriate when. is is the mechanism of language acquisition, which includes what knowledge is required to reach the adult knowledge state at the appropriate time. Computational modeling can be used to examine a variety of questions about the language acquisition process, because a model is meant to be a simulation of the relevant parts of a child's acquisition mechanism. In a model, we can precisely manipulate some part of the mechanism and see the results on acquisition. If we believe the model accurately reects the child's language acquisition mechanism, these manipulations and their eects inform us about the nature of that mechanism. Importantly, some manipulations we can do within a model are dicult to do with children. e modeling data are thus particularly useful because of the diculty of getting those

Cite this paper

@inproceedings{Pearl2010UsingCM, title={Using computational modeling in language acquisition research}, author={Lisa Pearl}, year={2010} }