Strengths and Weaknesses of Memory in Autism Revealed By a Meta-Analysis
The literature on memory in Autism Spectrum Disorder (ASD) shows an inconsistent pattern of results.
To address this variability, we report the first ever meta-analysis of short-term (STM) and long-term (LTM) memory in ASD, evaluating the role of the type of material (verbal, visual, visuospatial, neutral faces), type of memory retrieval (free recall, cued recall and recognition), and the organization of items (serial, associative, semantically related).
We included 61 studies, comparing individuals with typical development (TD) and ASD or Asperger syndrome, confirmed by ADI and/or ADOS method for diagnosis, and/or DSM-4, DSM-5, or ICD-10 diagnostic criteria, corresponding to a total of 2788 participants (1373 with and 1415 without ASD). We computed effect sizes (Cohen’s d) using RevMan 5.3, and linear regressions using SAS 9.4.
Results revealed an overall memory decrement in ASD with a small to medium overall effect size (d= -0.42 [-0.52; -0.31], p<.001), with greater difficulties in STM (d= -0.50, [95%CI -0.65;-0.34], p<0.00001, I²=51%), compared to LTM (d= -0.36, [95%CI -0.49;-0.22], p<0.0001, I²=38%). Analyses on the type of material identified a relative preservation of verbal LTM (d= -0.15, p=0.09), contrasting with impaired LTM for visual material (d= -0.38, p=0.006) and neutral faces (d= -0.79, p<0.00001) in ASD compared to TD individuals, but any subgroup difference for STM. We found a general free recall impairment compared to cued recall and recognition (LTM, free recall: d= -0.31, p=0.0006, cued recall: d= -0.08, p=0.58, recognition: d= -0.15, p=0.08; STM, free recall: d= -0.56, p<0.00001, recognition: d= -0.33, p=0.07) in ASD compared to TD individuals. We identified diminished serial (d= -0.58, p<.00001) and non-serial STM performance, and associative (d= -0.38, p=0.0005) and non-associative LTM. Regression analyses revealed that memory impairment increased with the ADOS symptoms severity (β= -0.17, p=.014).
Overall memory difficulties in ASD suggest the involvement of common processes such as executive dysfunction associated with atypical connectivity. Based on neurobiological data in ASD, we suggest that alterations in low and high frequency oscillations may alter the maintenance and rehearsal processes associated with STM storage that may also impact LTM functioning. Dissociation in LTM suggest that intact verbal LTM, mainly conceptually-driven, may rely more on the global preservation of semantic knowledge in ASD, while difficulties in visual LTM may be due to abnormally elevated left-over-right connectivity ratio in ASD supporting a more local processing of visual items. Finally, the general free recall impairment may emerge from difficulties in adequately searching retrieval cues, associated with executive and electrophysiological specificities, limiting the memory search process. This hypothesis may also account for intact cued recall and recognition, in accordance with the Task Support Hypothesis, and argues in favor of using memory support.
Finally, we suggest that atypical oscillatory activity arising from neurophysiological abnormalities in ASD rather than a specific alteration of a brain region or a memory system may explain this pattern of results. Preserved verbal LTM and supported retrieval constitute important findings for therapists and caregivers of individuals with ASD, providing opportunities for memory rehabilitation.