Property Extension in ASD: Categorically but Not Perceptually Restricted

Background: Difficulties with how words are represented, organized, and accessed have been observed in individuals with Autism Spectrum Disorder (ASD). Category formation and generalization abilities are also often atypical in ASD, with school age children and adolescents frequently performing inconsistently on categorical induction tasks. For example, they often restrict properties assigned to specific instances of a named category, to the taught instance itself rather than extending to new instances of that named category (Kelley et al., 2006). What is still unclear is whether their category restrictions apply to all new instances vs. just instances that differ visually/perceptually from what was taught.

Objectives: The current study compares TD children and children with ASD during their early school years on their extension of stated properties to new instances when these instances were (a) identical except for the pose, (b) from the same category but different perceptually, and (c) perceptually similar but from a different category.

Methods: Typically developing (TD) children (N=29, MA =5.55 years (0.33), 5 girls) and children with ASD (N=16, MA= 6.35 years (0.59), 2 girls) participated. The children’s Test for Auditory Comprehension of Language (TACL-3) vocabulary raw scores were similar (M(TD)=36.66(4.78), M(ASD)=32.31(5.47) although the TD group was more advanced (t(43)=2.81,p = 0.008). Pictures of eight natural kinds, both animate and inanimate, were presented. Children were asked to predict whether a known property of one member of a named category would be applicable to another member of the same category. For example: “This white rabbit eats grass. Does this brown rabbit/white squirrel/white-rabbit-in-different-pose/cat also eat grass?” The four test pictures included a target (same natural kind but different color or form), perceptually similar but not the same category, same natural kind with slight modifications in pose, and a distractor.

Results: Most TD children and children with ASD successfully extended the attribute to the slightly modified item of the same natural kind at above chance levels (TD=93%; ASD=88%); however, the TD group performed more consistently than the ASD group (M(TD)=7.24(1.15); M(ASD)=5.88(1.59); t(43)=3.32, p=0.002). With the critical target item, the TD group endorsed the category extension significantly more often than the ASD group: M(TD)=4.72(1.51), M(ASD)=3.56(1.86) (t(43) =2.27 p=0.028). Many in the TD group performed above chance unlike most in the ASD group (TD=59%, ASD=25%). These group differences held when TACL vocabulary scores were entered as a covariate (Fs >4.76, ps<0.035). Finally, for the physically similar but not of the same natural kind stimuli, there were no observed differences between the groups t(43)=-1.773, p=0.09, who both endorsed the physical similarity foil at levels significantly below chance.

Conclusions: Both the TD and ASD groups successfully extended the property to instances that only varied in pose from the taught instance; hence, property extension is not completely restricted in children with ASD. However, when property extension involved perceptual differences within a named category, the children with ASD performed at chance, unlike the TD children. Thus, extension on the basis of categories seems to be a major challenge for children with ASD.