Recently, I posted in my lab blog (or is it on my lab blog? I don’t know) about the challenges in developing a test for bilingual children. In collaboration with Aquiles Iglesias, Vera Gutierrez-Clellen, Brian Goldstein, and Lisa Bedore, I worked on development of the Bilingual English Spanish Assessment (BESA)– a test for Spanish-English bilinguals designed for identification of language impairment. The challenge that we faced when we began this 7 year project (in 1998) is that there was very little data on markers of language impairment in other languages. In fact some of this information had just begun to emerge for language impairment in English speakers.Development of such a test for bilinguals of course poses a more complex challenge. First, there is a great deal of variability in terms of the amount and type of language exposure children have to the home and the school language. Even if the exposure to each language is evenly split the nature of the input is often different. So the kinds of tasks that are appropriate for inclusion on a test for L1 vs. L2 are different.
Second, depending on where children live, how many children are in a particular geographic area who speak the same home language, and the specific curricular options available in that area children may have exposure to school in the home language, English, or both. Some children learn to read and write in the home language in districts (and states) where there is a lot of support for the home language. This means that there is not just one profile of bilingualism in terms of L1 and L2 exposure, knowledge, and skills.
Next, we needed to get a handle on whether bilingual children with language impairment make similar errors as monolingual children with language impairment or if they make different kinds of errors. The answer is yes. And, typical bilingual children make the same kinds of errors of monolingual children with LI. So, we had to look at the degree of error and the proximity to the target in order to differentiate bilingual children with and without language impairment.
What does this have to do with gold? Well, the way to test a new test is to compare its classification accuracy against that of a gold standard. The problem is that we didn’t HAVE one. We had to create one and we did so based on the literature at the time. We collected language samples in both languages (when we could), we interviewed teachers and parents, and documented clinical impressions. These data were used as converging information in order to base our gold standard and to identify children with and without language impairment. The next step (which we are in process of completing) has been to compare our BESA against the classifications based on all this information we have about individual children. Some of these studies on the morphosyntax, semantics, and phonology subtests of the BESA have been published. And we are continuing to study these questions.
In all, we’ve made some progress– there’s still a lot to do. But, we know more now than we did when we started.