Recast density and acquisition of novel irregular past tense verbs

Children with specific language impairment (SLI) lag behind children with typical language (TL) in their grammatical development, despite equivalent early exposure to recasts in conversation (M. E. Fey, T. E. Krulik, D. F. Loeb, & K. Proctor-Williams, 1999) and the ability to learn from recasts in intervention as quickly as do children with TL (K. E. Nelson, S. Camarata, J. Welsh, L. Butovsky, & M. Camarata, 1996). This experiment tested whether this apparent paradox could be attributed to variations in the density of recasts in conversation versus intervention.

Thirteen children (7-8 years of age) with SLI and 13 language-similar children (5-6 years of age) with TL were exposed to 3 recast densities of novel irregular past tense verbs (none, conversation-like, intervention-like) over 5 sessions. Outcomes were based on spontaneous conversational productions and a post-test probe.

As predicted, at conversation-like densities, children with TL more accurately produced the target verbs they heard in recasts than in nonrecast models (d = 0.58), children with SLI showed no differences, and children with TL produced the verbs more accurately than did children with SLI (d = 0.54). Contrary to expectations, at higher intervention-like recast densities, the SLI group did not improve their accuracy, and the TL group performances were significantly poorer (d = 0.47).

At conversational levels, recasts facilitated greater verb learning than models alone but only in the TL group. Increasing recast density to the modest levels in this brief intervention experiment did not benefit children with SLI and led to poorer learning for children with TL. To optimize learning, efficiency of recast distribution as well as rate must be considered.