Most versus least: A meta-analysis of stimulus overselectivity

Abstract

Stimulus overselectivity (OS) occurs when a person's behaviour is controlled by only a limited subset of available cues. This phenomenon is thought to contribute to learning and social challenges in individuals with autism spectrum disorder (ASD). This thesis presents a multilevel meta-analysis quantifying the magnitude of OS in individuals with ASD compared to typically developing (TD) peers. Analyses of 15 studies, producing 41 unique effects, showed a statistically significant population gap (p = .036). The average level of OS was 35.16% for ASD groups and 20.96% for TD groups. Both groups exhibited a general tendency for OS, but the effect was more pronounced in ASD. Notably, this difference was fragile: when the oldest study (Gersten, 1983) was excluded, the gap was no longer statistically significant (p = .129). A major limitation was that only 15 out of 158 relevant studies provided enough data for quantitative analysis, highlighting a widespread reporting gap. These findings were consistent with stimulus control hierarchies, suggesting that the systematic narrowing of control observed in individuals with ASD may be sensitive to non-linear diagnostic shifts within the evolving diagnostic landscape. For practitioners, the findings emphasise the importance of assessing which cues guide a learner’s responses and using targeted strategies to broaden cue control in educational settings. For researchers, these results highlight the need for consistent, standardised reporting of element-level data and variability to advance theoretical understanding of overselectivity across populations.