1995/03/01 by James L. Morgan, Katherine M. Bonamo, Lisa L. Travis · 2 citations
Psychology · Computer Science · #Language Development and Disorders #Machine Learning and Algorithms #Natural Language Processing Techniques
paper · doi:10.1037/0012-1649.31.2.180
openalex publication_date 1995/03/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/06/11
Previous work has shown that recasts may be contingent responses to children's early ungrammati-cally. On this basis, it has been claimed that recasts provide negative evidence, thereby offsetting the need for linguistic constraints in theories of acquisition. This study explores whether children exploit negative evidence putatively provided by recasts by examining whether parental recasts are associ-ated with children's recovery from particular overgeneralization errors. Data from longitudinal in-vestigations of 2 common syntactic errors reveal that recasts are related to children's subsequent grammaticality. However, contrary to what would be expected if recasts serve as corrections, the data show that recasts are negative leading indicators of grammaticality. Finally, correction and negative evidence axe examined and are shown to be nonequivalent. Therefore, corrections in whatever form they might exist can offset only a limited subset of proposed innate constraints on language acquisition. In the course of development, children may at times adopt grammars that appear to be overgeneral with respect to the lan-guage they are acquiring. For example, children learning En-glish may alternate between using and omitting articles preced-