Toddlers with Autism Do Not Show Evidence of Categorization in a Novel Word Learning Task

Background:  It has been shown in literature that individuals with autism spectrum disorders (ASD) can categorize rule-based information, but have difficulty abstracting prototypes or relationships among object features  for natural objects, artifacts, and faces (Gastgeb et al., 2006). Because infants can abstract prototypes around 3 months of age (Mareschal & Quinn, 2001), a difficulty with generalization or prototype formation can be an early sign of autism in young children.  

Objectives:  To investigate whether children with ASD would generalize from simple geometric shapes to real-life objects in a novel word learning task.

Methods:  We tested 16 children with ASD (mean age = 28.33 months) and 18 TD children (mean age = 28.66 months) in a novel word learning task (Son et al., 2008). Children were first presented with a simple geometrical novel object paired with a novel name: a “dax.” In the following trial, children were presented with the same object with a distractor, and were asked to identify the dax. This was the memory trial, because the test object was the same as the one in the previous trial. In the next trial, children were presented with real-life versions of objects that were presented in the memorization trial, and were asked to identify the dax. This was the generalization trial, since the children were expected to generalize the novel name that they learned during the memory trial with the simple objects to the real-life objects. Children in both groups were also administered the Mullen Scales of Early Learning (MSEL; Mullen, 1995).

Results:  We compared children’s performance on the memory trials to their performance on the generalization trials, and then their performance on both trials to the 50% chance. The performance of TD group on the memory trials was similar to their performance on the generalization trials, which were both above 50% chance. The ASD group performed significantly better on the memory trials compared to the generalization trials, and their performance on the memory trials was above chance, whereas their performance on the generalization trials was not.

Since it has been shown that children who have larger vocabulary perform better in a novel word learning task than children with limited vocabulary (Son et al., 2008), we conducted the analyses with the ASD group with and without expressive language delay. Eight children were classified as “average-language group” (ASD-AL) whose standard sores on the MSEL Expressive Language were within the average range, and eight children were classified as language-delayed group (ASD-LD)  whose scores were at least one standard deviation below the mean. ASD-AL group did significantly better on the memory trials as compared to the generalization trials, and their performance on the memory trials was above chance, whereas their performance on the generalization trials was not.  ASD-LD group performed at the chance level on both memory and the generalization trials.

Conclusions:  Children with ASD, both those with average expressive language and those who have delayed language, do not seem to generalize from simple to complex objects in a novel word learning task.