Referential Generalization In Children with Autism Spectrum Disorder

Background:  Previous findings suggested that children with autism spectrum disorders (ASD) rely mainly on perceptual salience to map novel labels to novel objects. Unlike typically-developing children (TD) they did not rely on a speaker’s social cues (i.e., pointing, gazing) alone to identify referents, when these conflicted with salience (Preissler & Carey, 2005; Parish-Morris et al., 2007). However, Luyster and Lord (2009) found that when ASD and TD groups were matched on both expressive language and age, these higher-functioning children with ASD were able to use social cues to map novel labels to objects regardless of salience.

Moreover, recently learned labels need to be generalized to other objects of the same kind. Labels can be extended associatively, linking with one particular object, or referentially, linking to the object’s symbolic category. Preissler (2008) reported that TD children understand word-picture relations referentially, while children with ASD understand them associatively.

Objectives: Our first aim was to directly test whether higher language ability is related to the use of social cues for word-mapping in children with ASD. Our second aim was to investigate whether children with ASD display referential understanding of novel labels, and whether this is also related to higher language ability.

Methods:  Sixteen children with ASD (16-72 months) were individually matched on receptive language using the Mullen with 16 TD children (14-31 months). Our word-mapping task followed Parish-Morris et al. (2007): a pair of novel objects was presented, one more perceptually salient than the other. Speakers provided a novel label in two conditions: coincident, where social cues and salience indicated the same object, and conflict, where social cues indicated the less salient object.

Each word-mapping trial was followed by a photograph task where participants viewed four photographs: color and black-and-white images of both objects just seen. Participants were asked to indicate any of the “novel label” they saw. We defined referential generalization as extending the novel label to both photos of the named object. Data was included only if the child passed the word mapping task, and was analyzed for the coincident condition alone.

Results: No significant group differences were found on the word-mapping task in selecting the correct referent in either condition, replicating Luyster and Lord (2009). However, whereas the TD children performed equally well across conditions, the ASD group performed better when social cues matched salience, consistent with Parish-Morris et al. (2007). For both groups word-mapping success was significantly correlated with receptive and expressive language ability.

In addition, a similar number of ASD (6/13) and TD children (4/13) displayed referential generalization (p=.42). However, in the ASD but not the TD group, referential generalization was displayed only by children with high rather than low receptive and expressive language, based on a median split of scores.

Conclusions: Our findings demonstrate that, contrary to early reports, some children with ASD display similar word learning processes to TD children; they are able to learn words by following social cues and display referential understanding of novel labels. These children have higher language abilities.