Trial-by-trial Bayesian integration with attentional switching, rather than non-Bayesian switching heuristics, underlie perceptual estimation
Tamás Kovács, Central European University, Hungary; Máté Lengyel, University of Cambridge; Central European University, Hungary
Posters 1 Poster
Pacific Ballroom H-O
Thu, 25 Aug, 19:30 - 21:30 Pacific Time (UTC -7)
Whether humans behave as intuitive statisticians or use simple heuristics is a subject of ongoing debate in cognitive science. Perceptual estimation has been heralded as a staple paradigm in which the statistically principled Bayesian integration of prior expectations and current sensory evidence can be demonstrated. However, Laquitaine and Gardner (2018) has recently found that humans deviate from the Bayesian strategy in a paradigmatic motion direction estimation task and rely on a simple heuristic instead. Specifically, their results suggested that the apparent Bayesian integration of prior and evidence is an artefact of averaging participants’ responses across trials, and the bimodal distribution of trial-by-trial responses is instead best captured by a non-Bayesian model that stochastically switches between using only the prior or only the evidence on each trial. If true, these results call into question the most basic tenet of the Bayesian brain hypothesis. Therefore, we reanalyzed the data of Laquitaine and Gardner (2018), and fitted participants’ responses by twenty distinct Bayesian and non-Bayesian switching models, differing in their predictions both quantitatively and qualitatively. We found overwhelming evidence for Bayesian integration, with occasional lapses in attention that only affect the quality of sensory evidence but not its downstream integration with the prior. This model best accounted for the bimodality of responses as well as other, previously ignored diagnostic features of the data. These results demonstrate that humans perform trial-by-trial Bayesian integration of prior and evidence in perceptual estimation, rather than using a simple switching heuristic.