Does conceptual scoring increase classification accuracy for vocabulary tests?

We’ve (as a field) have known for about 20 years that single word vocabulary tests whether they are receptive or expressive tests are poor indicators of developmental language disorders (DLD). At the same time, these tests are very often used by SLPs as part of a diagnostic. They are easy to give, quick, and highly reliable. It’s hard to make an error in  administration or scoring on these tests. But, reliability is not enough (neither are the other reasons). Even if it only takes 5 minutes and the score is a perfect representation of what the child can do it doesn’t mean that a low score indicates impairment or that a high score indicates typical development. As far as domains of language go– children with DLD do pretty well with vocabulary at the single word level. It’s semantics (connections among words) that they have difficulty with. Still, what about bilinguals? There has been some really nice work looking at bilingual vocabulary acquisition relative to children who have exposure to only one language. When bilingual children are tested in one language they often  look like they know less vocabulary than their monolingual peers.  But, this approach doesn’t  take into account Across a number of studies, when children are credited for knowing a word in EITHER  language their vocabulary size is comparable to that of their monolingual peers. There is some variation based on age where different ways of counting makes a difference.

For younger kids a total word count is more comparable. Total words is where you would count all the words whether they are translation equivalents or not. There’s a logic to this, especially with young children. They might use each word in different ways and  they use each word in different contexts. A word is a word so why not count it?

For older kids it seems that conceptual scoring will do the trick. Here, words known in each language are counted, but translation equivalents are counted only once– the focus is on concepts. Once  kids are in school they learn vocabulary related  to that as well as continuing to learn words in their home  language. So counting words in only one language may underestimate what they know especially if they are learning different things in each of their languages.

But, what about bilingual children with DLD?  Well, the short answer is: if vocabulary testing doesn’t work for monolinguals why would it work for bilinguals. And that’s essentially what happens.

We did a study comparing Spanish-English bilingual children with and without language impairment on the EOWPVT-3rd edition. Yes yes, we know there’s a 4th, but this version allows us to see translation equivalents at the same point on the test.  We also  tested 14 items past the ceiling  to account for the item order differences observed in the 4th edition, so if we had the inclination (and so  far no one has) we could rescore the item level responses according to the newer version  (I  don’t think it will  make a difference).

We had 247 children participate. Of these, 38 had DLD. We scored each of the tests (the Spanish version and English versions) monolingually,  Then we scored conceptually. We did the conceptual score different ways. First we did conceptual scoring within the test. So, if the child responded in Spanish during English we had a within-test conceptual score (same for English responses on the Spanish version.  Then we did  between-test conceptual scoring. Here, we looked across the two versions of the test item by item. We added it into the Spanish test if they used it on the English test, and we added it into the English test if they used it on  the Spanish test.

First, we found that the between-test version of the test resulted in  the highest scores. We think this represents best what children know. And it’s probably another blog.  On the question of whether ANY of  these ways of scoring accurately classified children with and without language impairment. The answer is no–  our results are consistent with what’s been found for monolingual English speakers.

You can see in the table below, that pretty much no matter what we did or how we scored, sensitivity and specificity is around 60% to 65%. This is NOT GOOD. It’s not even fair classification. This is poor classification. English only has a sensitivity of 81.6% but it’s at the cost of way overidentifying (around 48%) typical as impaired. So a low score doesn’t mean anything.  There is no way that you can make an accurate decision about DLD based on this measure even if you use conceptual scoring.

That doesn’t mean the measure is useless. I do use vocabulary tests and I think they tell us about what kids know. That can help us make intervention  decisions. But, we can’t use these measures to make accurate diagnostic decisions.

Measure Language Cutoff Overall Correct Classification Sensitivity Specificity
Single-language Spanish 88.43 65.2 57.9 66.6
English 66.02 57.1 81.6 52.6
Within-test Conceptual Spanish 94.12 66.0 65.8 66.0
Conceptual English 67.20 59.1 76.3 56.0
Across-test Conceptual Spanish 104.66 68.8 65.8 69.4
Conceptual English 80.25 63.6 68.4 62.7

, , , ,

  1. Leave a comment

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 )

Google photo

You are commenting using your Google 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 )

Connecting to %s

%d bloggers like this: