Neurotypical humans readily optimize performance in ‘reversal learning’ games, but while mice learn the winning strategy, they refuse to commit to it, a new MIT study shows. The research provides a mathematical way to track the rodents’ more mixed tactics.