Scatter: Quantifying a Qualitative Vocabulary Difference in Adults with ASD

Background: Semantic emptiness and odd or imprecise use of vocabulary, in combination with accurate use of unusual words, have been reported in adults with ASD who do not also have language impairment. Such features are not captured by total scores on standardized vocabulary tests, but might be captured by item scatter. A scattered or inefficient pattern of responses occurs when early test items are missed and later items are defined correctly, in contrast to a gradual decrease in scores as items become progressively more difficult to define. Scatter is thus consistent with the coexistence of semantic errors and sophisticated word use that has been informally observed in ASD.

Objectives:  Whether the uneven quality of vocabulary use in ASD could be quantified using an alternative scoring method of standardized tests was explored here. If adults with ASD demonstrated more scatter in their pattern of responses on a vocabulary test as compared with controls, the result would support observations that have been made both in ASD case studies and in the present sample (Stothers & Oram Cardy, 2016). No significant difference by group in total mean scores for expressive vocabulary was expected. The ASD group was expected to have a significantly higher Scatter score than the control group.

Methods: Adults aged 19–44 years completed the Vocabulary subtest of the Wechsler Abbreviated Scale of Intelligence (WASI), with 21 participants in both the ASD and Control groups (N = 42). Two scoring methods were used. The conventional method followed the WASI scoring protocol. In the Scatter method, original scores of 0 or 1 were coded as failed, and scores of 2 as passed. Item weights were calculated as the percentage of control participants who passed the item. Scatter was equal to the sum of the weights of the items that participants missed, multiplied by the peak item answered correctly.

Results:  Hypotheses were supported. Group means were not significantly different for the total Vocabulary score, t (40) = .35; p = .78, with a negligible effect size, d = .11. All scaled scores were within or above the average range in both groups. The ASD group had a significantly higher Scatter score than the control group, t (24) = 2.40; p = .02, with a moderate to large effect size, d = .72.

Conclusions:  Qualitative differences previously reported for this sample were captured by a quantitative measure. Scatter was shown to be an effective measure of difference, separating the control and ASD groups in the context of equivalent overall scores for Vocabulary. Results supported using this method, in addition to the standard scoring protocol, to more sensitively characterize performance of individuals with ASD without language impairment on standardized tests of word knowledge.