An investigation of the persistence of mutual fund performance indicates that investors need more than past performance numbers to pick future winners. In this study, style analysis was used to separate fund total returns into style and selection components. Performance was defined in terms of total returns, selection returns, and information ratios (ratios of selection return to selection risk). For each fund type, two out-of-sample periods were established to investigate persistence of performance from one period to the next using regression analysis and contingency tables. The evidence supported persistence only for fixed-income fund performance. This persistence is beyond any effects of fund fees and expenses or data base survivorship bias. Unfortunately, this persistence edge cannot overcome the average underperformance of fixed-income funds resulting from fees and expenses.