The current language eligibility guidelines provided by TSHA, and many districts across the state, consider students with a language disorder eligible for Speech Language Pathology services in the schools when their standard scores on a language battery fall 1.5 standard deviations below the mean. This is consistent with eligibility criteria for many other states, as well as many researchers in the field (Spaulding, Plante, and Farinella, 2006); yet, how do we know that children with language disorders will have a score that low or lower on any given language test? This is the question that Spaulding, Plante and Farinella asked in their 2006 article in Language, Speech and Hearing Services in the Schools. They were interested in 2 things in their study of a large number of language tests: (1) Did children with language disorders typically score at the low end of “normal” for those on whom the test was normed? (2) Which tests provided information on specificity (i.e., percentage of typical children diagnosed as having typical language) and sensitivity (i.e., percentage of disordered children diagnosed as having a language impairment) in their test manuals.
So, what did they find? After looking at 43 tests, they found that only 10 reported score differences for children with language impairments greater than 1.5 standard deviation. The average mean difference between typical and disordered children’s scores in the norming samples was 1.34 SD. In 9 of the tests, most of the children with language impairment scored within 1 SD of the mean! Are there students with language impairments that aren’t made eligible simply because of the test we selected to administer?
Research indicates that using sensitivity and specificity is a better way to determine the accuracy of a test for the purposes of identification of a disorder (Spaulding et al, 2006). Specificity is how accurate the test is at identifying typical children as typical, and sensitivity is how accurate the test is at identifying disordered children as disordered. Specificity and sensitivity information was provided in 9 of the tests’ manuals. Only 5 of those tests had sensitivity and specificity reported in their manuals that would be good enough to support their use in identifying language disorders: CELF-4, PLS-4, Test of Narrative Language, Test of Early Grammatical Impairment, and Test of Language Competence – Expanded Edition. I frequently see the CELF and PLS in use in the schools, but the others are not quite so frequently used; in fact, the TLC-E hasn’t been updated since 1989!
So what does this mean for us? The authors of this study recommend that if we are using a test to determine eligibility for services, or to identify a disorder, we should make sure that the specificity and sensitivity data from the test support its use in that way. Further, we should not consider that 1.5 standard deviation score to be a “cut off” for services without gathering other data to support our decision. If everything says the child is disordered, but the test does not, maybe the problem is with the test.
Spaulding, T.J., Plante, E., and Farinella, K.A., (2006). Eligibility Criteria for Language Impairment: Is the Low End of Normal Always Appropriate? Language, Speech and Hearing Services in the Schools, 37, 61-72.