When participants view and then reproduce simple objects that vary along a continuous dimension such as length, shade, or emotional expression their estimates tend to be biased toward the average value of the presented objects. This phenomenon has been modeled as the result of a Bayesian combination of prior category knowledge with an imprecise memory trace (Huttenlocher, Hedges & Vevea, 2000). Whereas previous work analyzed data aggregated across participants, here we examined individual differences in strategy. Thirty-six participants viewed and reproduced 496 morphed face stimuli that ranged from angry to happy. We found substantial variation in the bias patterns participants produced. Individuals’ estimates were well fit by a model that posited attraction toward three categories, one at the happy end of the range, one at the angry end, and one that captured the entire range of presented stimuli, allowing the weight given to each category to vary by participant.