Differentiating children with and without language impairment based on grammaticality.

PURPOSE This study compared the diagnostic accuracy of a general grammaticality measure (i.e., percentage grammatical utterance; PGU) to 2 less comprehensive measures of grammaticality--a measure that excluded utterances without a subject and/or main verb (i.e., percentage sentence point; PSP) and a measure that looked only at verb tense errors (i.e., percentage verb tense usage; PVT)--in differentiating children with and without language impairment. METHOD Two groups of 3-year-olds, 17 with language impairment and 17 with typical language, participated in a picture description task. PGU, PSP, and PVT were computed. Receiver operating characteristic curve analyses were conducted to determine the best cutoff value for each measure. RESULTS All 3 measures demonstrated a sensitivity of 100%. PGU showed a specificity of 88%, and both PSP and PVT showed a specificity of 82%. In addition, PGU showed a larger positive likelihood ratio than the other 2 measures. CONCLUSION PGU, PSP, and PVT were all sensitive to language impairment. However, PGU was less likely than PSP and PVT to misclassify children with typical language. The resultant diagnostic accuracy makes PGU an appropriate measure to use to screen for language impairment.