The interplay between difficulty and reward during experiential and observational category learning

Perceptual categorization requires mapping highly variable signals onto the same category. Some items are clearer examples of the category, being more typical or more distinct from contrasting categories. As a result, individual items differ in the difficulty required to categorize them correctly – easier exemplars are more quickly and accurately identified compared to more difficult exemplars. Here, we ask how associating stimulus difficulty with reward impacts novel category learning. We examine rule-based auditory category learning with stimuli varying in the difficulty of categorization. Three conditions varied in how difficulty was aligned with reward – higher reward for harder stimuli, higher reward for easier stimuli, and random reward regardless of difficulty. We compared results when participants actively learned the categories through experience and when participants learned from observing the actions and outcomes of someone else. We found that aligning reward with either easier or harder stimuli was not beneficial for learning and that participants learned the categories and successfully generalized to novel instances regardless of whether they learned through experience or observation. However, learners demonstrated a biased expectation about the relationship between difficulty and reward. When reward was random, participants defaulted to an expectation that categorizing harder stimuli should yield higher rewards. Further, only participants who learned through experience could learn that easier stimuli yielded higher rewards; observational learners relied on the default expectation. Overall, learning and generalization of category knowledge is possible through either experiential or observational learning but adjusting expectations about the relationship between difficulty and reward requires learning from experience.