The paper tests and finds strong support for the hypothesis that in the nonunion sector of the economy, turnover is negatively selective on a worker's job performance. At establishments with about 17 employees, a worker who is one standard deviation (21 percent) less productive than average during the first few months on the job is 11 percentage points more likely to be laid off or fired and 7 percentage points more likely to quit during the succeeding year. At large nonunion establishments and in small labor markets, productivity has very large effects on risks of an involuntary separation but almost no effect on the propensity to quit. Productivity appears to be positively related to layoffs and quits at unionized establishments.