Using longitudinal data from the Survey of Natural and Social Scientists and Engineers (SSE), I investigate whether what appears to be a glass ceiling in cross-sectional analyses of the gender earnings gap for engineers results instead from a cohort effect. Two findings are reported: (1) For each of three cohorts, within-cohort effects of being female remained relatively constant over the seven years of the SSE (1982-1989). This suggests that earnings penalties for women engineers result from cohort rather than glass-ceiling effects-that the earnings penalty to women is more a matter of when they started their careers than of how long they have worked. (2) In absolute terms, the earnings penalties for younger cohorts of women are essentially zero. These findings indicate that multi-cohort longitudinal designs should be used in investigations of the glass-ceiling hypothesis, in particular, and of women's progress in management and the professions, in general.