A cohort of 363 rural children in Bilbeis, Egypt, were followed from birth from 1981 to 1983, with twice-weekly home visits made to detect diarrheal illness. Enzyme-linked immunosorbent assay was used for detection of rotavirus in stools collected during episodes of diarrhea. Rotavirus-associated diarrhea was detected once in 74 children and twice in 12 children. Using a technique not previously described, the authors calculated the age-specific incidence rates for initial episodes and second episodes of rotavirus-associated diarrhea to estimate the effectiveness of naturally acquired immunity. Assuming that the risk of exposure was the same before and after the first episode, the observed and expected numbers of second episodes of rotaviral diarrhea were equal (age-adjusted rate ratio = 1.01; 95 percent confidence interval 0.55-1.86), given the age-specific person-years at risk. The assumption of equal risk for reexposure to rotavirus appears to be invalid, however, since the children with one and two rotavirus-positive episodes appeared to be at greater risk for diarrheal illness of all causes (rate ratios of 1.42 and 1.78, respectively). The clinical illness may have been less severe in second episodes; emesis was reported more often with first rotavirus episodes than with second rotavirus episodes, and the only fatal case was in an initial episode. These data are compatible with the existence of partial immunity, since it appears that the risk of reexposure may be greater in children who experienced rotaviral gastroenteritis earlier in life. In four of seven children, rotavirus isolates from first and second episodes were of identical serotypes, indicating that even serotype-specific immunity for rotaviral diarrhea was incomplete.