I’d mentioned last week that I was starting to learn more about codeswitching through collaborative research with Kai Greene. We have a new paper in Child Language Teaching & Therapy where we explore the use of code-mixing in children with and without language impairment. We were interested in how many kids switched to their other language during testing, if their switching was related to language dominance, and how successful they were when they did switch. We examined the semantics data from the Bilingual English Spanish Oral Screener (BESOS) that was given to 1200 children. In this study, we examined only data from the 5 year olds from the first two years of data collection: 606 children. The semantics screener has 11 items in English and 12 items in Spanish and roughly half of these are expressive items, which were the focus of the analysis.
Across the whole set of kids, about half did not switch from Spanish to English or from English to Spanish. 19% of the kids codeswitched from English to Spanish at least once; another 25% codeswitched from Spanish to English at least once. This means that when kids do codemix or codeswitch, it’s mainly asymmetrical. A small number (7%) codeswitched at least once in both English and Spanish), these symmetrical codemixing children overall tended to be balanced bilinguals.
We were also interested in whether children with and without risk for language impairment were able to codemix successfully. We defined success as answering the semantics item correctly in the other language. In English, both children with and without risk codesmixed, but the children with risk for language impairment were more likely to respond incorrectly. So, in that case mixing did not help them. In contrast children without risk for language impairment switched between languages and got the answer correct. Testing in Spanish had a different pattern. There were no patterns associated by risk, but there were patterns associated by dominance. Children who were Spanish dominant were able to codemix to English and answer the question accurately, while English dominant children codemixed to English but tended to respond inaccurately. Balanced children were in the middle.
So, overall codemixing doesn’t mean that children can’t do the task, most of the time they switch or mix to respond to the question in whatever language they can access. Children with risk however are less accurate overall, even when they codemix. I think this is one reason why allowing and counting codeswitched utterances during assessment can provide important diagnostic information.