Modelling children's negation errors using probabilistic learning in MOSAIC.

Abstract

Cognitive models of language development have often been used to simulate the pattern of errors in children’s speech. One relatively infrequent error in English involves placing inflection to the right of a negative, rather than to the left. The pattern of negation errors in English is explained by Harris & Wexler (1996) in terms of very early knowledge of inflection on the part of the child. We present data from three children which demonstrates that although negation errors are rare, error types predicted not to occur by Harris & Wexler do occur, as well as error types that are predicted to occur. Data from MOSAIC, a model of language acquisition, is also presented. MOSAIC is able to simulate the pattern of negation errors in children’s speech. The phenomenon is modelled more accurately when a probabilistic learning algorithm is used.