Abstract
Imagine being asked to select from a color wheel the hue of the last stop sign you saw. Your choice would likely be biased to a prototypical red, even if the actual last seen sign happened to be off-color. Such ‘priors’, or pre-existing beliefs about reality, are known to affect perceptual decision-making. While these effects are often assumed to occur in a Bayesian manner, little is known about its underlying mechanisms. Since stimulus expectations has been hypothesized to bias pre-stimulus visual activity, one could posit that priors would interact with incoming perceptual information in a similar way. If true, the effect should depend on the prior being available during or near to stimulus encoding. Alternatively, priors could be affecting post-perceptual decision-making. If this is the locus of priors, then Bayesian effects should not significantly depend on whether the prior was made available before or after stimulus encoding. With a delayed color-matching task, we demonstrate that while both pre and post-stimulus priors result in Bayesian-predicted effects – such as perceptual estimation reports being biased to the prior (N=41, both ps< .01 compared to control) – the effect of these two priors were not significantly different from each other. Incorporation of priors can occur post-stimulus, suggesting more of a decisional locus than a perceptual one, at least in situations where the ‘prior’ is explicitly provided and constantly changing. However, even when priors were fixed, presumably allowing perceptual changes to accrue over time, the effects were not significantly different from having a prior that changes trial-by-trial (ps>.5 for both bias and precision, N=16). This reinforces the notion that changes in early perception contributes little to the effect of priors, at least in comparison to later downstream changes. Future research should carefully consider the method and the loci of effects when expectations are altered.
Acknowledgement: NYUAD Research Enhancement Fund REF-175