Is Cross-Situational Word Learning Impaired in Children with Autism Spectrum Disorder?

Poster Presentation
Thursday, May 10, 2018: 11:30 AM-1:30 PM
Hall Grote Zaal (de Doelen ICC Rotterdam)
C. Hartley1, L. A. Bird2 and P. Monaghan1, (1)Lancaster University, Lancaster, United Kingdom, (2)Lancaster University, Lancaster, United Kingdom of Great Britain and Northern Ireland

Word learning is of crucial importance to language acquisition. The challenge of mapping sound-to-meaning is often characterised by referential ambiguity; there might be multiple potential referents for a newly-heard word. In the absence of ostensive cues, typically developing (TD) children can infer the meanings of new words from cross-situational statistics alone. By contrast, many children with autism spectrum disorder (ASD) experience impaired language acquisition and word learning. Here we explore whether children with delayed linguistic development have specific difficulty tracking word-referent co-occurrences.


The objective of this study was to examine the ability of children with ASD (and concomitant language delay) to identify, retain, and generalise the meanings of unfamiliar words from cross-situational statistics.


Participants were 15 children with ASD (M age: 8.78 years) and 20 TD children (M age: 6 years) with similar receptive vocabulary age equivalents (ASD M: 5.7 years; TD M: 6.5 years). Populations differed on non-verbal IQ (TD M: 101.7; ASD M: 83.73; t = 3.99, p < .001). Using a touch-screen tablet, participants were presented with pairs of unfamiliar objects and were instructed to identify the referent of a novel word. These exposures were intentionally ambiguous (i.e. there was no cue to which object was the correct referent), but cross-trial statistics enabled children to disambiguate correct word-object pairings. Participants learned six new word-object relationships over 48 trials, divided into two blocks. After a 5-miute delay, children completed 24 test trials assessing label retention and generalisation to differently-coloured exemplars.


Performance on learning and test trials was analysed via generalised linear mixed effects models. Learning was significantly more accurate in Block 2 (ASD: 71.7%; TD: 74%) than Block 1 (ASD: 64.4%; TD: 68.5%; Z =2.93, p =.003), and populations did not differ. Both populations demonstrated similar accuracy on retention (ASD: 64.4%; TD: 65%) and generalisation test trials (ASD: 64.5%; TD: 67.5%) that significantly exceeded chance (all ps < .001). For children with ASD, both learning and test trial accuracy were predicted by receptive vocabulary (Z = 3.78-4.5, p < .001) and chronological age (Z = 3.09-3.4, p =.002-.001). For TD children, both learning and test trial accuracy were predicted by non-verbal IQ (Z = 2.01-3.12, p =.04-.002) and chronological age (Z = 2.32-2.35, p =.02).


In comparison to TD controls matched on receptive vocabulary, the accuracy of cross-situational word learning in linguistically-delayed children with ASD is unimpaired. Both groups were able to identify, retain, and generalise novel word meanings from statistical associations alone. Thus, delayed language development in ASD may not be attributable to specific difficulties tracking word-referent co-occurrences. However, ASD may impact the mechanisms that scaffold this ability. Performance was predicted by receptive vocabulary in children with ASD and non-verbal IQ in TD children, suggesting that cross-situational word learning may be influenced by language-specific mechanisms in the former population and domain-general mechanisms in the latter. We recommend that follow-up analyses examine potential differences in reaction times and explore whether test trial accuracy can be enhanced by attentional cues during learning.