Psychiatric epidemiological research has focused disproportionate attention on traditionally female disorders such as depression. This paper shifts epidemiological gears to elaborate and test an etiological model of a traditionally male disorder, that of alcohol abuse. We argue that social-relational deficits (narcissistic orientations) lead to abuse of alcohol for stress reduction purposes, given interpersonally oriented stressors and the incapacity to form social supports. The model was tested in the context of training for a traditionally male occupation encompassing both social-relational demands and limited social supports. A cohort of medical students was surveyed from medical school entrance through a portion of clinical training. Time 1 social-relational deficits were predictive of time 3 alcohol abuse, partially as a function of social support deficits and, to a lesser extent, patient care-related stressors. Moreover, there was an initial gender difference in social-relational deficits and a trend-level gender difference in alcohol abuse that disappeared when social-relational deficits were held constant. By time 3, women did not differ from men in social-relational deficits or alcohol abuse.