Statistical Learning in Specific Language Impairment and Autism Spectrum Disorder: A Meta-Analysis.

Rita Obeid,Jarrad A G Lum,Kristen Gillespie-Lynch,Patricia J Brooks,Kasey L Powers

doi:10.3389/fpsyg.2016.01245

Abstract

Impairments in statistical learning might be a common deficit among individuals with Specific Language Impairment (SLI) and Autism Spectrum Disorder (ASD). Using meta-analysis, we examined statistical learning in SLI (14 studies, 15 comparisons) and ASD (13 studies, 20 comparisons) to evaluate this hypothesis. Effect sizes were examined as a function of diagnosis across multiple statistical learning tasks (Serial Reaction Time, Contextual Cueing, Artificial Grammar Learning, Speech Stream, Observational Learning, and Probabilistic Classification). Individuals with SLI showed deficits in statistical learning relative to age-matched controls. In contrast, statistical learning was intact in individuals with ASD relative to controls. Effect sizes did not vary as a function of task modality or participant age. Our findings inform debates about overlapping social-communicative difficulties in children with SLI and ASD by suggesting distinct underlying mechanisms. In line with the procedural deficit hypothesis (Ullman and Pierpont, 2005), impaired statistical learning may account for phonological and syntactic difficulties associated with SLI. In contrast, impaired statistical learning fails to account for the social-pragmatic difficulties associated with ASD.

Highlights

Statistical learning of complex rules or patterns is thought to play a crucial role in the development of language, social-cognitive, and motor skills (Perruchet and Pacton, 2006; Frith and Frith, 2008; Romberg and Saffran, 2010; Ruffman et al, 2012
We report a series of meta-analyses conducted on studies of statistical learning in Specific Language Impairment (SLI) and Autism Spectrum Disorder (ASD) in order to evaluate whether the procedural deficit hypothesis provides an adequate account of impairments in SLI and ASD
A mixed-effects meta-analysis addressed the first aim of the study, to determine whether statistical learning is impaired in SLI, by extending Lum et al.’s (2014) findings using a larger dataset that was not restricted to the Serial Reaction Time (SRT) task

Summary

INTRODUCTION

Statistical learning of complex rules or patterns is thought to play a crucial role in the development of language, social-cognitive, and motor skills (Perruchet and Pacton, 2006; Frith and Frith, 2008; Romberg and Saffran, 2010; Ruffman et al, 2012). Lum et al (2014) used meta-analysis to evaluate whether impairments in statistical learning, as assessed using the Serial Reaction Time (SRT) task, constitute a core deficit in SLI. If learning of the fixed sequence of stimuli occurs, reaction times (RTs) will be significantly faster for trials in sequenced as compared to random blocks Basing their methodology on a prior meta-analysis of learning deficits in the SRT task in patients with schizophrenia (Siegert et al, 2008), Lum et al (2014) calculated effect sizes by assessing the difference between the mean RTs in the final sequenced block vs the first random block. The aims of the current meta-analysis were (1) to examine whether impairments in statistical learning are a shared challenge for individuals with SLI and ASD; and (2) to examine whether task modality and age moderated effect sizes. Learning is typically measured by calculating the number of correct responses (learning the association between the cue and the outcome) across trials (Mayor-Dubois et al, 2015)

MATERIALS AND METHODS

RESULTS

DISCUSSION