Individual Differences in Homograph Reading Amongst Hebrew-Speaking Autistic Children

Background:   One of the most consistently replicated and influential findings in research on autistic cognition is the poor performance of individuals with autism on tests of homograph reading. In the test, participants are required to read aloud sentences containing ambiguous words such as “tear” that are pronounced differently depending on their meaning. According to the weak central coherence account, poor performance reflects a failure of language comprehension: words are processed out of context, hence the appropriate meaning and pronunciation are not extracted. However, this account is not supported by evidence from other paradigms such as semantic priming and language-mediated eye-movements. Furthermore, significant group differences mask the fact that many participants with autism perform at ceiling on the test.

Objectives:   In the current study, we investigated individual differences in homograph reading amongst autistic children who spoke Hebrew as their native language. Hebrew is characterized by a high degree of orthographic ambiguity and provides many more suitable homographs than English. We were therefore able to obtain reliable individual differences and determine predictors of performance within the group of autistic children.

Methods:   Participants were 18 native Hebrew-speaking autistic children (16 boys) aged 8;3 to 17;6 years (mean = 11;7, SD = 1;8), rigorously diagnosed by a multi-disciplinary panel. In the homograph reading task, participants read aloud 26 sentences containing homographs whose correct pronunciation depended on the preceding sentence context. In addition, participants completed a battery of language and reading tests and teachers completed the Childhood Autism Rating Scale (CARS) for each child. Data were analysed using logistic regression with mixed random effects. 

Results:   There was, as expected, considerable variation in performance across participants, with scores ranging from 58 to 96% correct. Significant predictors of accuracy included age, CARS score, word reading (but not nonword reading), and paragraph reading speed and accuracy. However, the best predictor was picture naming. This accounted for unique variation beyond each of the other predictors. Indeed, the logistic regression model containing picture naming left no significant individual variation in homograph reading unaccounted for.

Conclusions:   The results confirm that there is considerable and reliable individual variation in homograph reading amongst autistic children. Moreover, they suggest an interesting new explanation for the homograph-reading difficulties. Performance was best predicted by picture naming which, like homograph-reading, involves the selection of one spoken form from amongst multiple possible candidates. Thus, difficulties in homograph reading faced by some autistic individuals may originate in the processes of speech production rather than in comprehension and the use of sentence context, as the central coherence account assumes. The results highlight the importance of investigating individual differences within the autism spectrum and demonstrate the novel insights this can bring to understanding of autistic cognition.