Bootstrapping the Hippocampus? Atypical Learning Characterizes Adolescents with Autism Spectrum Disorders

Background:  

Individuals with ASD are thought to exhibit intact semantic memory for facts, details, and routines, but manifest deficits in episodic memory and generalizing learning. The underlying neurobiology of this pattern, which may include impairments in the medial temporal lobe (MTL) and the prefrontal cortex (PFC), remains unclear. We have argued that young adults with ASD are more reliant on the hippocampus (HC) than on the PFC, given their cognitive control deficits.

Objectives:  

We investigate this contention in adolescents with ASD, using behavioral measurements that assess MTL, HC, and PFC contributions to encoding, retrieval, and generalization. 

Methods:  

Participants included 27 12-18 year olds with ASD (mean age = 14.8) and 25 age, gender, and IQ-matched participants with typical development (TYP) (mean age = 14.8). To assess MTL and HC contributions to learning and memory, we administered the Relational and Item-Specific Encoding task (RISE; Ragland et al., 2012), which requires respondents to make item-specific and relational judgments, and to provide response confidence estimates. Performance measures include accuracy rates and d'. Familiarity and recollection also were analyzed using ROC curves. To assess PFC contributions to learning and memory, participants completed California Verbal Learning Test-Children’s Version (CVLT-C; Delis, Kramer, Kaplan, & Ober, 1994). ANOVAs and t tests were performed using SPSS 22.

Results:  

On the RISE, ASD versus TYP showed poorer item-specific accuracy (t(49) =  2.2, p = .0.03) and d' (t(49) =  2.7, p = .009), but comparable relational encoding accuracy and d'. ASD also showed a reduced contribution of familiarity when making relational judgments (t(49) = 2.3, p = .025). This may suggest that ASD benefitted disproportionately from the deeper spatially-oriented encoding involved in the relational condition because they possess a relatively intact HC and/or other MTL regions that support recollection. On the CVLT-C, ASD versus TYP showed poorer list learning (t(52) = 2.2, p = .032) and free and cued recall at short [free recall: t(52) = 3.345, p = .002; and cued recall: t(52) = 3.32, p = .002] and long delays [free recall: t(52) = 4.74, p < .001; and cued recall: t(52) = 3.23, p = .002]. ASD also exhibited lower recall consistency (t(52) = 3.82, p < .001) with more perseverations (t(52) = 2.17, p = .034), suggesting they have PFC deficits which impede the deep encoding of semantic materials that supports generalization (Shohamy & Wagner, 2008). Both groups relied comparably on semantic and serial clustering strategies thought to require the HC. 

Conclusions:  

Findings are inconsistent with the view that ASD exhibit intact lower-level learning and memory and impaired higher level learning and memory. Instead, they suggest that ASD actually may be relatively worse at lower-level learning of items that is prefrontally-mediated or that involves areas of the anterior temporal system (Ranganath & Ritchey, 2012) including the lateral and orbito-fronal cortices, and the perirhinal cortex, and relatively better at relational versus item-specific encoding under conditions involving spatial processing, that is sub-served by a relatively more intact posterior medial system including the HC and the parahippocampal cortex.