This article examines retirement outcomes in the first four waves of the 1992-1998 Health and Retirement Study (HRS). Measured retirement is seen to differ, sometimes substantially, with the definition of retirement used and among various groups analyzed. Moreover, those differences vary with the wave of the survey as respondents age. Retirement comprises a complex set of flows among states representing nonretirement, partial retirement, and complete retirement. Using the self-reported definition of retirement, 77 percent of transitions continue in the same or equivalent states between adjoining waves of the HRS, 17 percent involve a move from greater to lesser labor force participation, and 6 percent involve a move from lesser to greater labor force participation. Twenty-two percent of the sample report they were partially retired at some time in the first four waves, and by age 65, over a fifth of the population is partially retired. Altogether, 17 percent of the sample experienced a reversal in the course of the survey, moving from a state of less work to a state of more work. A comparison of retirement flows for men between the HRS and the 1969-1979 Retirement History Study (RHS) shows that the large spike in the population leaving nonretirement at age 65 observed in the 1969-1979 RHS has fallen from 18 to 11 percentage points in the HRS and that the share leaving nonretirement at 62 has increased from 13 to 20 percentage points over time. The results presented here should help researchers improve their understanding of the structure of the dependent variable in retirement studies. Incorrect or arbitrary measurement of the retirement variable may lead to a misunderstanding of how Social Security and related policies affect retirement outcomes. Thus, the improved understanding of retirement gained from this research will be helpful to those designing retirement policies as they attempt to understand the effects of those policies.