Fig. 1

Theoretical framework and tasks. (a) Graphical depiction of how the Bayesian inference predicts the internal response and optimal decision criterion during a categorization task. An observer is deciding between two possible categories (Category A or Category B). We obtain the expected cost of each decision (E_A and E_B) by multiplying the sensory uncertainty and prior corresponding to each stimulus and then summing the costs associated with the two possible categories (here, we assume equal cost). (b) Illustration of the sequence of events within a trial in all experiments. For each trial, participants were asked to categorize the Gabor (Category A or Category B) depending on its orientation. The two sets of Gabor provide examples of possible orientations for each category, and different levels of contrast —from top to bottom, 0.004, 0.18, and 0.72. (c) Stimulus orientation distributions for each category in Task 1 (Experiment 1) and Task 2 (Experiment 2). (d) Illustration of the internal representation of the category distributions. In Experiment 1 (top graphic), d’ represents the sensitivity or ability to separate the two categories, and c represents the adjustment of the decision criterion when the prior favors Category A. In Experiment 2, (bottom graphic), the distributions with vivid colors represent the internal representations of the categories when the sensory noise is low, and the faded colors when the sensory noise is high. s represents the internal noise, and k represents the decision boundaries, shifting outwards when the sensory noise is increasing. (e) Illustration of the number of trials for each category per block in Experiment 1. Each block contained 320 trials. Category B could appear with a probability of 75% (Block 1), 50% (Block 2), or 25% (Block 3). The block order was randomized across participants. In Experiment 2, each block contained the same number of trials (i.e., 320), and the two categories had an equal probability