Research suggests that boards of directors select CEOs using signals of ability. However, little is known about how boards determine the combination of attributes that constitute a ‘good’ CEO, especially attributes without an ex ante clear impact on managerial quality, such as CEO optimism. I argue that boards will learn the optimal level of such attributes more quickly from past success, and empirical results support this. Boards, particularly those with high reputation/independence, are significantly more likely to select a moderately optimistic (optimal) successor following a moderately optimistic CEO departure. Robustness checks rule out alternate explanations and support this conclusion.