People often perceive their in-groups as more heterogeneous than their out-groups. We propose an information sampling explanation for this in-group heterogeneity effect. We analyze a model in which an agent forms beliefs and attitudes about social groups from her experience. Consistent with robust evidence from the social sciences, we assume that people are more likely to interact again with in-group members than with outgroup members. This implies that people obtain larger samples of information about in-groups than about out-groups. Because estimators of variability tend to be right-skewed, but less so when sample size is large, sampled in-group variability will tend to be higher than sampled out-group variability. This implies that even agents that process information correctly – even if they are naive intuitive statisticians – will be subject to the in-group heterogeneity effect. Our sampling mechanism complements existing explanations that rely on how information about in-group and out-group is processed.