The study interviewed and examined 470 consecutive assault victims admitted to an inner city accident and emergency service in 1986. Their reported intake of alcohol in the 12 hours before the assault was recorded by trained interviewers. Blood ethanol concentrations corresponded to this alleged intake in the 362 patients who consented to testing. The rate of admission to a hospital was not significantly different in patients who consented to the investigation and those who did not. An index of injury severity was compiled, and this was related to the need for admission to a hospital by using a logistic regression model. Patients who reported abstinence tended to have more minor and fewer severe injuries than drinkers who had consumed more than 10 units of alcohol. This finding may result because aggression and loss of judgment induced by alcohol in the victims prolongs violence or reduces victims' capacity to protect themselves during an assault. 1 table, 4 footnotes.