Testing bilinguals in English: Do we get any useful information?

I think the title says it all, and it’s an important question. There’s no doubt that the best way to determine if a child has a language impairment is to test them in their language. For bilinguals, that usually means two languages– in the US context, their home language and English. But, there aren’t enough bilingual speech-language pathologists available nationally; and bilingual speech-language pathologists don’t speak every language represented by the clients in their area. So, one of the questions we’ve had is whether we can get any useful information from testing children who are bilingual in English only. The quick answer is yes, no, and it depends. The longer answer follows.

We have a new paper in press in JSLHR called Identification of Specific Language Impairment in Bilingual Children, Part 1: Assessment in English. In this paper we followed 167 kids from our Diagnostic Markers study for two years after we screened them before kindergarten. These kids were selected into the study from the original 1200 because they used both Spanish and English at least 20% of the time, and because they scored below the 30th percentile on one screening measure in Spanish and one in English. So, we had a group of kids some of whom were likely to have language impairment; and others who had mixed dominance, but typical development, but who might be likely to be referred because they demonstrated patterns of strengths and weaknesses that were inconsistent.

We tested this group of kids on a battery of measures in English and Spanish, and in this study we present the English results. Of the group of children, 21 were identified with language impairment by 3 very experienced bilingual SLPs. We used this as the gold standard measure against to test how well a battery of measures that included the TOLD-P subtests and the TNL together differentiated bilingual children with and without language impairment.

Using cutscores set for monolingual English speakers, we found 95% sensitivity and 45% specificity. This means that the criteria identified most of the kids with LI as having LI, and a bunch of the typical kids as LI too. So, here the answer is no, English testing doesn’t do a very good job at identification of LI in bilinguals.

We then looked at whether we could set new cut points. That led to improvment, but we still over identified bilinguals with LI. Then we used regression modeling to optimize the classification and using these optimized curves were able to obtain improved classification to 81% sensitivity and 81% specificity. So, yes, we do get some useful information. In the paper, we provide the formula for calculating the probability that the child has language impairment based on the TNL and the TOLD-P:3 subtests. Using this formula is an improvement over a cut-off score approach. Also, we recommend that children need to have at least 30% regular exposure to English. Kids who have less exposure to English than that are likely to be misdiagnosed. So, here the “yes, we can get good information from English testing” is moderated by “it depends.”

It’s better, but 81% correct classification still means there’s 19% error, so we’d like to improve on that. Our next study looks at whether testing in both languages can improve our diagnostic accuracy even further. I’m betting it can– that will be part 2.

About these ads

, , , , ,

  1. #1 by Ron Gillam on August 6, 2013 - 12:22 pm

    Just how good is 81% sensitivity and 81% specificity? In the Spaulding, Plante, and Farinella (2006) paper in LSHSS, the average sensitivity of 18 English tests administered to monolingual English speakers was .76 and the average specificity was .86. Only 8 of 18 English tests for which sensitivity and specificity could be assessed had values that both exceeded .80 for identifying SLI in monolingual English speakers.

    On top of that, our sensitivity and specificity values yielded likelihood ratios that were very close to the likelihood ratios from studies that used Spanish tests to diagnose LI in Spanish-English bilingual children (Dollaghan and Horner, 2011). All those studies employed two-gate designs that were likely to inflate diagnostic accuracy.

    As Liz said, we might find sensitivity and specificity values that are somewhat higher when we combine the results from Spanish and English testing. However, it will be interesting to see if the increases in sensitivity and specificity values from bilingual testing are large enough to justify the expense of testing in two languages. Even if they are, given the variety of languages that are used in US schools, the lack of tests in most of those languages and the lack of bilingual examiners, there are many instances in which testing in two languages is nearly impossible. If we can adjust the cut scores from English tests in ways that yield reasonably good sensitivity and specificity, we’ll be doing bilingual children and the SLP’s who serve them a big favor.

  2. #2 by Elizabeth D. Peña on August 6, 2013 - 12:33 pm

    absolutely, this is a huge improvment over using currently available tests for bilinguals; but remember, it’s not just about re-setting the cut points– that didn’t do as good a job as the formula based on a optimized cuts- and kids should be fairly fluent in English as well. So, use the formula!

  3. #3 by St Paul Bankruptcy Attorney on August 8, 2013 - 2:30 pm

    learning two or more languages is more interesting.

  4. #4 by Carpet Cleaner Anchorage on September 5, 2013 - 6:46 am

    Your post is very interesting and informative.

  1. Testing bilinguals in English: Do we get any useful information? | on raising bilingual children

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 1,303 other followers

%d bloggers like this: