In 1984, an outbreak of gastroenteritis occurred at a school with 1,860 students in Brooklyn, NY. In a single-stage cluster sample of 375 students, 129 (34%) had illnesses that met our case definition of vomiting or diarrhea. The mean incubation period was 26 hours, and the mean illness duration was 24 hours. All case students had eaten in the cafeteria on at least one day between Nov 13 and 16, compared with 174/214 (81%) noncase students (P = 10(-8), Fisher exact test). Foods implicated were french fries (relative risk 1.7, 95% confidence limits 1.4, 2.0) and hamburgers (relative risk 1.6, 95%, confidence limits 1.2, 2.1). Two cafeteria employees had served those foods while affected by diarrhea. By a recently developed blocking enzyme-linked immunosorbent assay, six of 11 (55%) case students showed fourfold antibody increases between acute- and convalescent-phase serum samples for Snow Mountain agent, a Norwalk-like virus, compared with one of ten (10%) noncase students (P = .04, Fisher exact test). We strongly suspect, but cannot document conclusively, that the Snow Mountain agent was spread to students on a vector of hot foods contaminated by ill food handlers. Implicated foods conferred low relative risks and could only have accounted for 74% of cases of illness. The strong association between cafeteria exposure and illness, therefore, suggests that additional modes of spread occurred.