Is the recent plateau in crude divorce rates due to compositional changes in the married population or to a fundamental change in the long-term trend of rising marital instability? I use refined measures of period divorce rates to show that the leveling of divorce rates appears to be real. Compositional factors do little to explain the end to the more than century-long pattern of rising divorce. Increases in cohabitation also fail to explain the plateau. New theories are needed to explain the determinants of divorce rates at the population level.