Working Memory Across the Adult Lifespan: Do Individuals with and without Autism Show Differential Age-Related Decline?

Background: Working memory (WM), the ability to maintain and manipulate information for guiding goal-directed behavior, is characterized by large individual and pronounced age-related differences. In typical development, WM performance increases across childhood, decreases constantly during adulthood, and shows accelerated decline in old age. In autism spectrum disorder (ASD), WM development in childhood seems to be delayed; there is inconsistent evidence whether it reaches comparable levels in adulthood, and it is largely unknown how it develops in old age. This study aims at bridging this gap by studying WM performance in individuals with ASD over the entire adult lifespan.  

Objectives: To investigate age-related patterns in WM performance across adulthood in participants with and without ASD, as well as inter-individual differences therein.

Methods: WM performance was assessed with a N-back task (three load-level: 0, 1, 2-back) in 111 adults with a clinical diagnosis of ASD (IQ>80) and in a comparison group (COM) of 164 adults without an ASD diagnosis between 19 and 79 years old. Dependent measures were accuracy (the proportion of correct responses) and reaction times (RT). Using regression trees, we explored whether demographical variables, comorbidities, or executive functioning could predict inter-individual differences in change over age.    

Results: There were no differences in accuracy between participants with and without ASD across all load levels. However, people with ASD exhibited longer RTs overall. Regression analyses revealed a linear decline of WM performance over age. Adding a quadratic age term, and its interaction with group, explained considerably more variance and revealed group-by-age interactions. Whereas performance change with age within both the ASD and COM group was well described by a linear trend, the decline was steeper in the COM group. Moreover, the quadratic fit was more suited for the COM group, suggesting an accelerated decline in older age. Exploratory regression trees revealed that IQ and interference control were predictors of inter-individual differences in age-related WM decline. Higher IQs (>109) were associated with better WM performance and stronger WM decline over age in comparison to lower IQs (<=109). Within participants with lower IQs, better interference control was associated with slightly better WM performance but not with differential age-related decline.

Conclusions: Although the current study does not provide evidence for a WM performance deficit across adulthood in ASD, the effect of age on WM is differentially expressed in adults with and without ASD. While increasing age is associated with marked, and finally accelerated, decline in individuals without ASD, increasing age had only a slightly detrimental effect on individuals with ASD. These findings provide initial insights into how ASD modulates cognitive aging, but also underlie the need for analyzing individual change trajectories.