Coordination games admit two types of equilibria: pure equilibria, where all players successfully coordinate their actions, and mixed equilibria, where players frequently experience miscoordination. The existing literature shows that under many evolutionary dynamics, populations converge to a pure equilibrium from almost any initial distribution of actions. By contrast, we show that under plausible learning dynamics, where agents observe the actions of a random sample of their opponents and adjust their strategies accordingly, stable miscoordination can arise when there is heterogeneity in the sample sizes. This occurs when some agents make decisions based on small samples (anecdotal evidence) while others rely on large samples. Finally, we demonstrate the empirical relevance of our results in a bargaining application. Final pre-print of a manuscript accepted for publication in American Economic Journal: Microeconomics.