Barberis et al. (J. Financial Econ. 49 (1998) 307), construct a model in which investors use the prevalence of past trend reversals as an indicator of the likelihood of future reversals. While such “regime-shifting” beliefs are consistent with a variety of psychological theories, other contrary predictions are consistent with the same theories. We report two experiments with MBA-student participants that strongly support the existence of regime-shifting beliefs. We conclude that regime-shifting models can provide a useful framework for understanding market anomalies, including underreactions to earnings changes and overreactions to long-term earnings trends.