Some theoretical perspectives suggest people overestimate animals’ mental capacities (anthropomorphism), while others suggest the reverse (mind-denial). However, studies have generally not employed objective criteria against which the accuracy or appropriateness of people's judgments about animals can be tested. We employed memory paradigms, in which judgments are clearly right or wrong, in nine experiments (eight preregistered; n = 3,162). When tested shortly after exposure, meat-eaters’ memory about companion animals (e.g., dogs) but not food animals (e.g., pigs) showed an anthropomorphic bias: they remembered more information consistent with animals having versus lacking a mind (Experiments 1–4). Vegetarians' and vegans' memory, on the other hand, consistently showed an anthropomorphic bias regarding food and companion animals alike (Experiments 5 and 6). When tested a week after exposure, both those who eat meat and those who do not showed signs of shifting toward a mind-denying bias (Experiments 2, 3, and 6). These biases had important consequences for beliefs about animal minds. Inducing mind-denying memory biases caused participants to see animals as possessing less sophisticated minds (Experiments 7–9). The work demonstrates that memories concerning animals’ minds can depart predictably from reality and that such departures can contribute to biased evaluations of their mental capacities.