We have been working on the question of how to best identify language impairment in bilinguals for a long time. One guideline that has been around for a long time is to test in both languages. In workshops and in presentations I often will repeat TEST IN BOTH LANGUAGES, test in both languages… But, how should results from two languages be combined?
There aren’t really any tested procedures for this. I think that generally we expect that children be low in both of their languages and it makes perfect sense. We don’t expect a child to demonstrate language impairment in only one language. That type of case would more likely be one of incomplete acquisition or language loss.
In a new paper by Peña, Bedore, & Kester, we tested two options for combining two languages using the semantics subtest of the BESA. One was to consider a total semantics score and the other was to use a 2-dimensional score. In both cases we used conceptual scoring– this is because the semantics subtest of the BESA has different items in each language. Both approaches worked pretty well.
For the total semantics score, we added the two standard scores. This allowed us to pool all the children regardless of their age. The mean for the typical bilinguals was 192 and the mean for the LI bilinguals was 135. The cut point was 163. This solution identified children with language impairment 93.3% correct (sensitivity) and identified typical children as typical with 85.7% accuracy (specificity).
The two-dimensional approach examined children’s scores in each language as a pair. If they were below the cut point in each language (88 for English, 85 for Spanish) then they were classified as having language impairment. This approach identified 93% of impaired children as impaired (sensitivity) and 97% of the typical children as typical (specificity). Below, you can see an illustration of the process.
All in all, we’re pretty satisfied with this result. It would be interesting to see how well this approach would work for other tests that are available in two languages.