The U.S. stock market’s return during the first month of a quarter correlates strongly with returns in future months, but the correlation is negative (positive) if the future month is (is not) the first month of a quarter. These effects offset, leaving the market return with its weak unconditional predictive ability known to the literature. The pattern accords with a model in which investors extrapolate announced earnings to predict future earnings, not recognizing that earnings in the first month of a quarter are inherently less predictable than in other months. Survey data support this model, as does out-of-sample evidence across industries and international markets. These results seriously challenge the Efficient Market Hypothesis and advance a novel mechanism of expectation formation.