We document an annual cycle in U.S. Treasuries, with variation in mean monthly returns of over 80 basis points from peak to trough. This seasonal Treasury return pattern does not arise due to macroeconomic seasonalities, seasonal variation in risk, cross-hedging between equity and Treasury markets, conventional measures of investor sentiment, the weather, seasonalities in the Treasury market auction schedule, seasonalities in the Treasury debt supply, seasonalities in the Federal Open Market Committee (FOMC) cycle, or peculiarities of the sample period considered. Rather, it is correlated with a proxy for variation in risk aversion linked to seasonal mood changes. Such a model can explain more than sixty percent of the average seasonal variation in monthly Treasury returns. The White (2000) reality test suggests this is not data snooping.