What does convergence tell us about language impairment and bilingualism?

In a paper we published last year, we examined how bilingual children with and without language impairment performed on a repeated associations task. We’d found that children with impairment has lower semantic depth scores even after we controlled for their vocabulary size. Also, their conceptual scores were higher than single language scores which tell us that bilinguals make different associations within each of their languages for the same things. One thing we’d observed in the paper was that it seemed that children with language impairment came up with different items compared to those with typical development. Our recently published paper explores this further examining the original data more closely. 

A semantics association task is one where we ask a child to tell us a word that’s associated with a word we give them, such as “table.” A response might be “chair.” Then, we ask again, another response might be, “sit” or “eat” or “dinner.” This kind of task is often used with adults and like the adult task, it tells us about depth of semantic knowledge. A child who can come up with a lot of associated words has more depth. Typically, because of frequency in a language and exposure to it, there’s a certain amount of predictability with respect to what someone might say. Probably “chair” in response to “table” is very common, but “ice-cream knife” a Victorian standard probably not so much. We would expect most of children’s responses to converge—after all, one needs to learn a common set of words—otherwise how would we talk to each other. And we would expect individual variation as well in using less common words. The question however is if children with LI produce fewer of the common core set of words than children with typical development.

Basically, the answer is yes, children with LI produced fewer of the top 5 most common responses (based on a normative group of about 100 children) compared to their age- and language experience-matched controls. When we looked at whether children’s distribution by frequency was similar to that of the normative group, we found that generally, yes there were similarities. Both children with LI and those with typical development seemed to produce frequent items (compared to the normative group) frequently and infrequent items less frequently. But, the English correlations were weaker for the LI group.

These patterns of results seem to indicate that it’s a little of each. Children with LI produced frequent items less frequently than typical children. At the same time, the order of frequency was still similar to that of the comparison group. So, children seem sensitive to frequency, but maybe it’s that children with LI need even more exposures to specific words to learn them. Also, it’s important to note that these bilingual children (ages 6 to 9) are able to learn words in both their languages. The correlation differences were more (and significantly) pronounced in English and this may be a function of English being the second language for many of these children. I would predict that their distributions would “catch up” even if the overall number of associations (or depth) is lower.