The Diagnostic Accuracy of Four Vocabulary Tests Administered to Preschool-Age Children

Abstract

This study examined the empirical evidence for using four vocabulary tests (Peabody Picture Vocabulary Test-III [Dunn & Dunn, 1997], Receptive One-Word Vocabulary Test [Gardner, 1985], Expressive Vocabulary Test [Williams, 1997], Expressive One-Word Vocabulary Test-Revised [Gardner, 1990]) to screen or identify specific language impairment (SLI) in preschool-age children. Tests were administered to 31 4- and 5-year-old children with SLI and 31 age-matched controls with normal language (NL). All children spoke General American English. Despite moderate to strong inter-test correlations, no test was a strong identifier of SLI. The group with SLI scored lower than the NL group on each test; however, the individual scores of children with SLI typically fell within the normal range. Vocabulary tests are frequently administered to determine whether a child’s language skills require further evaluation (screening), as a method of identifying SLI in children, or simply to describe aspects of language functioning. These purposes for administering a vocabulary test require various forms of empirical evidence in support of their use. Our data support construct validity for the four vocabulary tests examined, but do not support their use for identification purposes. Clinicians must apply a degree of sophistication in evaluating the evidence presented for test validity relative to the purposes for which the test will be administered. Unfortunately, although many test manuals offer inter-test correlations or statistically significant group differences as evidence of construct validity, they often omit data that would support common clinical uses, such as screening or identification.