Seasonality in U.S. stock prices is investigated using monthly average data on the Dow Jones Industrial Average over the period 1970-2005. By estimating a dummy variable model using OLS and rolling regressions, the results reveal the presence of a significant January effect except in the most recent period, 1990-2005, when a strong negative July effect surfaced. This finding is confirmed by using a more sophisticated structural time series model with an autoregressive structure. Some explanations are suggested for the disappearance of the January effect and the surfacing of the July effect.