Labor markets have improved considerably since the Great Recession. However, one indicator of labor market health, the mean duration of unemployment, remains stubbornly high. The mean duration of unemployment measures the number of weeks the average unemployed worker has been without a job. While the unemployment rate has nearly returned to pre-recession levels, the mean duration of unemployment is still more than 1.6 times higher than it was on the eve of the Great Recession. But the mean duration of unemployment does not fully capture how long most unemployed workers should expect to remain unemployed. Empirical evidence has consistently shown that those unemployed for a longer time-a year, for example-are less likely to find a job in a given month than those unemployed for a shorter time-say, a few weeks. As a result, those who do not find a job in the first few months of unemployment are likely to remain unemployed for a long time, and these long-term unemployed workers increase the mean duration of unemployment beyond what most unemployed workers will experience. This relationship between the length of time a worker has been unemployed and the likelihood of finding a job is known as duration dependence.This article reviews the evidence for duration dependence in job-finding rates and its implications for the unemployment duration distribution. The authors document duration dependence and show that it exists within nearly every demographic subgroup. Then, they examine the implications of duration dependence on unemployment duration, emphasizing that a uniform job-finding rate that does not incorporate duration dependence understates unemployment duration. Finally, they explore a composition-based approach to duration dependence, where they solve for the distribution of preexisting heterogeneity that is consistent with observed duration dependence.