Procedural Learning and Language Impairment: Evidence of a Deficit in Autism Spectrum Disorder with LI but Not in Specific Language Impairment

Background: In language development procedural memory has been implicated in learning and storing regularities and rule-based information, as required for phonology and grammar (Ullman, 2001). Deficits in procedural memory have been hypothesized to explain language impairment in Autism Spectrum Disorders (ASD) as well as in Specific Language Impairment (SLI) (Ullman, 2004). The hypothesized language-linked deficit in procedural memory would be most apparent in the subgroup of ASD children with concomitant language impairment. Yet, few studies have investigated procedural memory in language-impairedchildren with ASD, and none have compared their performance to that of children with SLI on the same measure.

Objectives: To clarify whether proposed procedural memory deficits are shared across children with language impairment or if deficits are disorder-specific we compared the performance of language-impaired children with ASD and children with SLI on a visual procedural learning task.  

Methods: We tested 14 language-impaired children with ASD and 14 children with SLI, matched on age (6-to 10-years-old) and NVIQ on a Serial Reaction Time task (SRT; Thomas & Nelson,2001). The SRT assesses implicit learning of a visual pattern presented on a computer screen. It involves five blocks of 40 trials: in blocks 2,3, and 5 the child sees a dog appear in one of four quadrants of the screen in a repeated 4-item sequence. In blocks 1 and 4 the dog moves randomly. The child is instructed to “catch” the dog as fast as possible by pressing a button corresponding to the quadrant of the screen where it appeared. Sequence learning is indicated by significantly faster reaction times in block 5 (sequence) in comparison to block 4 (random). This was measured by a sequence learning score (mean of block 4 – mean of block 5)/(mean of block 4 + mean of block 5). After the task we asked participants if they had noticed a pattern that helped them “catch the dog,” providing a measure of explicit knowledge of the sequence.  

Results: Groups showed similar ability to perform the SRT task as indicated by the number of errors in blocks 4 and 5; U=82, p=.47. However, the sequence learning score in the ASD group (Mdn=0.05) was significantly lower than in the SLI group (Mdn=.18); U=52.5, p=.04, r=-0.4. Within-group comparisons showed significantly faster responding in block 4 than 5 in the SLI group; T=0, p<.001, r=.88, but not the ASD group; T=30, p= .17, indicating that the SLI group learned the sequence while the ASD group did not (Figure1). Finally, similar proportions of children with ASD (5 of 14) and SLI (4 of 14) showed explicit knowledge of the sequence and were able to recreate the sequence they had learned.

Conclusions: Our results demonstrate diminished sequence learning in a language-impaired ASD group but intact learning in an SLI group matched on age and NVIQ.  This suggests that procedural learning deficits are not shared across forms of language impairment and that they are particulary implicated in ASD.  If replicated, procedural learning could be used to differentiate SLI from ASD and be integrated in ASD treatment.