The role of lexical frequency, telicity & phonological factors on past tense production in children with SLI & their typically developing peers

Limited research is available about how lexical and phonological verb properties interact with past tense production by children. Frequency of the inflected form and phonotactic probability might serve as input-driven alternatives to previously-studied factors such as lexical aspect and coda composition.

Archival elicited production data from 4-9 year old children with typical language (N = 24) and specific language impairment (N=14) using 108 two-clause complex sentences/85 different verbs were analyzed for past tense use, coda composition, telicity, phonotactic probability (Vitevitch & Luce, 2004), and lexical frequency (CHILDES; MacWhinney, 2000).

Several regression models were considered, including one with only categorical factors (e.g. obstruent/continuant ending), one with only continuous factors (e.g. average biphone probability), one with only phonological factors, one with only lexical factors, and several mixed models.

Diagnostic status and verb regularity accounted for the majority of the variance. The combination of lexical frequency of the inflected form with residuals of stem lexical frequency was the best lexical model. Place and manner information for the final consonant of the stem comprised the best phonological model. These two models combined into a final overall predictive model.