Longitudinal studies of acute and persistent diarrhea in 677 children less than 3 years old were conducted for 27 months in an underprivileged, periurban community near Lima, Peru. Incidence rates and accurate durations of diarrhea were obtained by twice-weekly community-based surveillance and clinical and laboratory features of diarrheal episodes were documented. Study children had an overall incidence of diarrhea of 8.1 episodes per child-year and an incidence of persistent (greater than 14 days) diarrhea of 0.25 episodes per child-year. Episodes of longer duration were associated with young age (0-5 months) and more severe illness (greater than or equal to 6 diarrheal stools per day) in the first week. None of the laboratory tests performed in the first week of illness (fecal leukocytes, blood, reducing substances, pH, or fat) proved of value to identify episodes that would become persistent. Although steatorrhea was commonly present in the acute phase of the illness, it became less frequent by the third or subsequent week of illness in the persistent diarrheas. These results suggest that there are no clinical or laboratory features of acute diarrhea that are strongly predictive of the subset of diarrheas that persist. Thus, it is important that all diarrheal episodes should have appropriate fluid and dietary management and follow-up to detect the persistent diarrheas that may need special intervention.