The Shape Bias: Investigations of Word Learning with Children with Autism

Background: The shape bias is a word learning mechanism that young children use to map the referent of a novel noun onto the shape of a novel object, rather than onto its color, texture, or size. Previous research has demonstrated that, unlike typical children, young children with autism up to four years of age do not use the shape bias as a word learning mechanism even though they have acquired a sizeable lexicon (Tek et al., 2008). Objectives: In the current study, we investigate whether these children developed a shape bias at a later age; moreover, we performed more detailed analyses of their shape bias performance. Methods: We tested 18 typically developing toddlers (mean age = 20.62 months at the onset of the study) every four months over four time points, and 17 children with autism (mean age = 33.13 months at onset) over six time points. The children had been matched on expressive vocabulary at Visit 1. We used the preferential looking paradigm, in which we presented novel target objects with two alternative test objects: one matching the target in shape and one in color. During the first set of trials, the NoName (baseline) audio asked “which one looks the same?” During the second set of trials, the target was given a novel label and the Name (test) audio asked “which one is (novel name)?” The dependent variables included (a) percent looking to the same-shape object, (b) the latency between the presentation of the test audio and subjects' looking to the shape-match, (c) the number of switches of attention between the shape-match, the color-match, and looking away, and (d) the direction of first look to either the shape match or the color match. We also created time-course graphs to investigate children's patterns of looking to the shape match, color match, and away during the span of every trial. Results: The typical group looked significantly longer at the shape match during the Name trials for Visit 2 through Visit 4 (p < .05), whereas the ASD group showed no such preference, even at the last visits. However, the ASD children did seem engaged in the task: Children in both groups had longer latencies of looking to the shape match during Name trials than during NoName trials (p < .05), showing that determining which of the test objects has the same label as the target requires longer and more conscious processing than simply determining perceptual similarity. The timecourse graphs revealed that, during Visits 3 and 5, the ASD group did tend to prefer the shape match during the last 3 seconds (50%) of each Name trial; however, this tendency disappeared during Visits 4 and 6. None of the other measures yielded significant results. Conclusions: These results support our previous finding that children with autism have difficulty using the shape bias as a word learning mechanism. The difficulty is not attributable, though, to lack of engagement in the task.