Purpose

Research on the diagnostic accuracy of different language measures has focused primarily on English. This study examined the sensitivity and specificity of a range of measures of language knowledge and language processing for the identification of primary language impairment (PLI) in French-speaking children. Because of the lack of well-documented language measures in French, it is difficult to accurately identify affected children, and thus research in this area is impeded.

Method

The performance of 14 monolingual French-speaking children with confirmed, clinically identified PLI (M= 61.4 months of age, SD= 7.2 months) on a range of language and language processing measures was compared with the performance of 78 children with confirmed typical language development (M age= 58.9 months, SD= 5.7). These included evaluations of receptive vocabulary, receptive grammar, spontaneous language, narrative production, nonword repetition, sentence imitation, following directions, rapid automatized naming, and digit span. Sensitivity, specificity, and likelihood ratios were determined at 3 cutoff points:(a)− 1 SD,(b)− 1.28 SD, and (b)− 2 SD below mean values. Receiver operator characteristic curves were used to identify the most accurate cutoff for each measure.

Results

Significant differences between the PLI and typical language development groups were found for the majority of the language measures, with moderate to large effect sizes. The measures differed in their sensitivity and specificity, as well as in which cutoff point provided the most accurate decision. Ideal cutoff points were in most cases between the mean and− 1 SD. Sentence imitation and following directions appeared to be the most accurate measures.

Conclusions

This study provides evidence that standardized measures of language and language processing provide accurate identification of PLI in French. The results are strikingly similar to previous results for English, suggesting that in spite of structural differences between the languages, PLI in both languages involves a generalized language delay across linguistic domains, which can be identified in a similar way using existing standardized measures.