|
|
Long-term generalization is a central goal of language intervention. What does infant artificial grammar learning contribute to understanding how to achieve this goal? Previous research from my lab and many others has demonstrated that infants who are typically developing generalize most linguistic rules with apparent ease. But does this generalization persist? And does determining what persists and what does not contribute to our understanding of atypical learners? My talk will attempt to grapple with these questions, and it will consist of three parts. First I’ll provide an overview of different types of non-linguistic learning tasks that were described over half a century ago by Shepard, Hovland, and Jenkins (1961). These tasks include generalizing from input examples based on single features shared by all examples vs. family resemblance relations, as well as other bases. I will attempt to map different types of linguistic learning onto these Shepard et al. tasks. Next I’ll present work on infant linguistic rule learning. In this part, I’ll show how at least some infant learning appears to follow rational principles, is extremely rapid, but doesn’t appear to persist for even a few seconds. In contrast, other learning may occur via a different mechanism, takes somewhat longer, and appears to persist over time. Finally, I’ll suggest some new directions for infant artificial grammar learning research and how these directions may apply to the treatment of affected components of linguistic and non-linguistic processing in childhood language disorders.
|
|