Objective Posttraumatic stress disorder (PTSD) and depression are known to be highly comorbid. However, previous findings regarding the nature of this comorbidity have been inconclusive. This study prospectively examined whether PTSD and depression are distinct constructs in an epidemiologic sample, as well as assessed the directionality of the PTSD-depression association across time. Methods Nine hundred and forty-two Detroit residents (males: n = 387; females: n = 555) were interviewed by phone at three time points, 1 year apart. At each time point, they were assessed for PTSD (using the PCL-C), depression (PHQ-9), trauma exposure, and stressful life events. Results First, a confirmatory factor analysis showed PTSD and depression to be two distinct factors at all three waves of assessments (W1, W2, and W3). Second, chi-square analysis detected significant differences between observed and expected rates of comorbidity at each time point, with significantly more no-disorder and comorbid cases, and significantly fewer PTSD only and depression only cases, than would be expected by chance alone. Finally, a cross-lagged analysis revealed a bidirectional association between PTSD and depression symptoms across time for the entire sample, as well as for women separately, wherein PTSD symptoms at an early wave predicted later depression symptoms, and vice versa. For men, however, only the paths from PTSD symptoms to subsequent depression symptoms were significant. Conclusions Across time, PTSD and depression are distinct, but correlated, constructs among a highly-exposed epidemiologic sample. Women and men differ in both the risk of these conditions, and the nature of the long-term associations between them.