The empirical finance literature documents a puzzling phenomenon: investors tend to adjust their portfolios insufficiently in response to changes in their own forecasts of asset returns. This paper proposes an investment experiment with information treatments to better understand the mechanisms behind this puzzle.