Cues to Pronominal Reference Resolution In Children with and without Autism Spectrum Disorders

Background:  

Children with autism spectrum disorders (ASD) have long been reported to have problems with production of pronouns (Fay, 1979; Lee, Hobson, & Chiat, 1994), but few studies have investigated comprehension of pronouns in this population (e.g., Lee, et al., 1994). This area merits further exploration as misinterpreting pronouns could contribute to more global impairments in comprehension. Third-person pronouns are relatively ambiguous in nature, as they can be used to refer to any individual in the world. In order to determine an appropriate coreferent, the listener needs to incorporate additional information from sources such as the syntax, semantics, and prosody.

Objectives:  

By exploring several types of cues in the same group of children, we can determine which cues emerge first in English-speaking children, and which are relatively more difficult for children with ASD.

Methods:  

Nineteen children with ASD (mean age: 98 months, range: 60-127) and 19 typically-developing children (mean: 78 mo.; range: 61-105) participated in several experiments to assess pronoun comprehension. Diagnostic status was confirmed using the ADOS and groups were matched for receptive language using the Test for Reception of Grammar (TROG).  For each experimental trial, the participant heard sentences played from speakers and used a button-box to select a matching picture from two choices displayed on a computer screen. The first experiment explored children’s ability to use the lexical gender (he vs. she) of a pronoun (1a) and gender- and age-based stereotypes about people (1b) to disambiguate pronouns.  Experiment 2 used the context of a previous phrase (2a) or sentence (2b) to disambiguate a pronominal referent.  In Experiment 3, prosodic cues were manipulated to shift a referent. 

Results:  

In the typically-developing group, the highest “passing” rate (defined as scoring at least 75% correct on a task) was on the task using age stereotypes (child-biased vs. adult-like activities) to disambiguate the referent, followed by the semantic tasks, then using lexical gender, and finally using prosodic cues.  In contrast, the group with ASD performed best on the lexical gender task, followed by age stereotypes, semantics, and prosody.  The unexpectedly low success rate on the lexical gender task by the typically-developing group may be attributed to use of gender stereotypes overriding the lexical gender of the pronoun, possibly an artifact of the younger age of this group.

Conclusions:  

Overall, when children with and without ASD are matched for receptive grammar abilities, their scores on tests of pronominal reference resolution are comparable; however, age differences between the groups may explain superior performance on some tasks by the ASD group.  Further, different factors may impact performance on individual tasks within each group. Explorations of which participant characteristics are associated with success on these tasks will be discussed and may give us insight into how these different types of cues are learned.