Paper detail

Biased perception leads to biased action: Validating a Bayesian model of interception

We tested whether and how biases in visual perception might influence motor actions. To do so, we designed an interception task in which subjects had to indicate the time when a moving object, whose trajectory was occluded, would reach a target area. Subjects made their judgments based on a brief display of the object's initial motion at a given starting point. Based on the known illusion that slow contrast stimuli appear to move slower than high contrast ones, we predict that if perception directly influences motion actions subjects would show delayed interception times for low contrast objects. In order to provide a more quantitative prediction, we developed a Bayesian model for the complete sensory-motor interception task. Using fit parameters for the prior and likelihood on visual speed from a previous study we were able to predict not only the expected interception times but also the precise characteristics of response variability. Psychophysical experiments confirm the model's predictions. Individual differences in subjects' timing responses can be accounted for by individual differences in the perceptual priors on visual speed. Taken together, our behavioral and model results show that biases in perception percolate downstream and cause action biases that are fully predictable.