The following is a transcript of the presentation video, edited for clarity. Presentation slides are available for download via the PDF button in the toolbar.
The following is a transcript of the presentation video, edited for clarity. Presentation slides are available for download via the PDF button in the toolbar.
Originally presented at the ASHA Convention (November 2013) as part of the session Development of a Bilingual Test for Spanish-English Children: A Long and Winding Road. Videos in this series are:
Challenges in Assessing Bilingual Populations
(Elizabeth D. Peña, University of Texas at Austin)
(Elizabeth D. Peña, University of Texas at Austin)
Bilingual Phonology Assessment Design
(Brian A. Goldstein, La Salle University)
Bilingual Pragmatics Assessment Design
(Aquiles Iglesias, Temple University)
Bilingual Semantics Assessment Design
(Lisa M. Bedore, University of Texas at Austin)
Bilingual Morphosyntax Assessment Design
(Vera F. Gutierrez-Clellen)
What are the challenges of assessing bilingual kids? Grosjean has said, and this has been quoted very often, that bilinguals are not two monolinguals in one.
When we’re testing bilingual kids, we probably need to do something different than what we would do for assessment of a monolingual. We can’t necessarily give the same thing that you would do for a monolingual Spanish child or a monolingual English-speaking child in two languages.
Measuring L1 and L2
One of the challenges in looking at bilingualism is that it’s highly variable at an individual level. The amount of contact that children have with each of their two languages is going to vary tremendously among the children you might be working with.
What are ways that we might think about measuring their first and their second language? There are a number of things that are recommended that we looked at in the literature.
Things like, the age of their second language exposure. At what point did they start learning English? We have some kids who start learning English — and Spanish — from birth. And some kids who have their first English exposure in preschool. And others when they’re in kindergarten.
What is their current use and exposure? How much do they get English input and Spanish input in a typical day?
Then we can do direct testing. When we do direct testing, what do we test? Do we test vocabulary to make a determination of their language proficiency or their language dominance? Do we test grammar? Do we test their narratives? The kinds of measures that you test can also vary tremendously.
Language Dominance
One of the things we have been looking at — because we think it’s really important to disambiguate — is the level of dominance in a language from language ability.
What we’re trying to do as speech-language pathologists is identify whether this is characteristic of an impairment in language or an impairment in speech. So we want to find measures that might help us disambiguate that from the exposure they have to each of the two languages.
One of the things we’ve looked at is how much exposure they have. How much do they use Spanish? How much do they use English? And how much do they hear each? We’ve looked at this against, for example, semantics.
In this slide you can see that children who use English 60% of the time or more were also heavily English-dominant on a test of semantics.
Over here, children who used Spanish 60% or more of the time were also heavily Spanish-dominant of a test of semantics in Spanish.
But the bilinguals who use Spanish and English more or less the same, between 40% and 60% which clusters around 50/50, there’s a small number of them who also showed balanced performance in semantics. So we have these differences here, on the ns. In terms of exposure, we have 122 kids who have balanced exposure, 80 children who show balance in semantics, but 39 who show balance on both of those measures.
For these bilingual kids, we get more dispersion around the middle. These bilingual kids are much more variable in what they hear, and how they perform on different measures.
We take the same kids, and we do the same thing on morphosyntax. Again, at the two ends of the continuum — the kids who get exposed mainly to English do really well in English, the kids who get exposed mainly to Spanish do really well with Spanish. But those kids in the middle show some differences. In morphosyntax, we have an even smaller number of kids who show balanced exposure in Spanish and English and balanced performance in Spanish and English morphosyntax. We have these differences here represented by these boxes. So 122 who have equal exposure, and 33 who are balanced in their performance on morphosyntax. So you get a lot of dispersion around the middle for those bilingual kids.
Then last we look at the crosstabs between morphosyntax and semantics, and we get that same kind of pattern. This is again with that same group of kids.
The story I’m trying to emphasize here is that bilinguals are variable on a number of these different variables.
True Language Impairment vs. Outcomes of Divided Input
A challenge, then, is separating what’s a true language impairment and what is simply an outcome of having divided input and divided use of each of your two languages.
Have you not had enough time to master the semantic system yet, and that’s why the scores are low? Or is it really language impairment? Have you not mastered the grammatical system of your second language? Or are you still in the process of learning it? And how do we disambiguate those two?
Because of this mixed input, sometimes second language acquisition looks like primary language impairment. And also first language loss might also look like language impairment.
When you’re going through that shift of acquiring a second language and maybe the first language is in attrition, or there’s not as much exposure to the first language, you might start making errors in that first language that look very much like the errors that you see in monolingual children that have language impairment.
This is what we have to disambiguate.
We know also that in different kinds of domains, children have different kinds of experiences. So we might see sociocultural experiences also play a role in how kids perform on standardized tests, especially tests that are normed on a monolingual population.
Kids may have gaps. They may know some things in one language and have lexicalized things in another language. It’s really common for preschool kids to have really good home vocabulary in Spanish, and really good school vocabulary in English, but they don’t necessarily have translation equivalents of all those things.
And you’re going to see this across the different domains. So depending on what the demands are — and those linguistic demands are tied to a specific language — you might see some variability across their two languages.
So what do we do, in terms of thinking about test development for this population that is highly variable, that has divided time in each of their two languages. Where age of acquisition of their first and second language is going to vary tremendously. And where familiarity of content is going to vary.
We need to do something more than what you would do for a monolingual test.
When you’re developing a test for a monolingual population, these aren’t variables that you necessarily take into account. Maybe familiarity of content — but age of acquisition of L1 and L2 isn’t an issue. And divided time between two languages isn’t an issue.
Taking Bilingual Variables into Account
So how do you take these into account?
Usually, when you develop a standardized test, your goal is to classify language impairment, or phonological impairment, and you want to develop items that are challenging for your clinical group. So those items are sensitive items for phonological impairment. Or for language impairment.
At the same time, you don’t want those items to be sensitive to different levels of language experiences. You want those items to be insensitive, as much as possible, to variation that might come from things that are not related to the clinical marker that you’re looking for.
So, what we have been really focusing on is looking at items, and looking at the item-level data, in terms of development of a test.
We used an item-analysis approach where we compare item-level differences between children with typical development and children with language impairment or phonological impairment. Items that did a good job at differentiating those two groups are items that we want to keep.
At the same time, we also want to make sure that those items still work at these different levels of linguistic exposure. If you vary on this other dimension of L1 and L2 exposure or on the age at which children starting learning a second language, do these items still differentiate between children with typical development and children with language impairment or phonological impairment?
