Study data.

Background
Previous studies reported that autistic adolescents and adults tend to exhibit extensive choice switching in repeated experiential tasks. However, a recent meta-analysis showed that this switching effect was non-significant across studies. Furthermore, the relevant psychological mechanisms remain unclear. We examined the robustness of the extreme choice-switching phenomenon, and whether it is driven by a learning impairment, feedback-related aspects (e.g., avoiding losses), or alternatively a different information sampling strategy.
Methods
We recruited an online sample of 114 US participants (57 autistic adults and 57 non-autistic). All participants performed the Iowa Gambling task, a four-option repeated choice task. Standard task blocks were followed by a trial block with no feedback.
Results
The findings replicate the extreme choice switching phenomenon (Cohen’s d = 0.48). Furthermore, the effect was found with no difference in average choice rates denoting no learning impairment, and was even observed in trial blocks with no feedback (d = 0.52). There was no evidence that the switching strategy of autistic individuals was more perseverative (i.e., that similar switching rates were used in subsequent trial blocks). When adding the current dataset to the meta-analysis, the choice switching phenomenon is significant across studies, d = 0.32.
Conclusions
The findings suggest that the increased choice switching phenomenon in autism may be robust and that it represents a distinct information sampling strategy and not poor implicit learning (or a bias in the sensitivity to losses). Such extended sampling may underlie some of the phenomena previously attributed to poor learning.