Interobserver agreement was determined between nurses and parents using a standard method of assessing diarrheal stools. The study population consisted of patients less than three years of age hospitalized at The Hospital for Sick Children, Toronto, Canada. Stool samples were independently categorized by observer pairs within minutes of being obtained from children with and without diarrhea as: watery--liquid, no solid elements; loose--liquid with solid elements; pasty--like a paste; formed--normal solid. Watery and loose stools were regarded as abnormal and indicative of diarrhea. Teaching sessions were conducted for nursing shifts, while parents were instructed prior to each observation. In the nurses' agreement study, each stool specimen was examined by the nurse providing care to the patient from whom the stool was obtained and a nurse not looking after the patient. Parents' assessments were also compared with nurses' assessments. Finally, parents' assessments were compared with each other. In the first group, agreement beyond chance for presence or absence of diarrhea measured by kappa was 0.78 (95% confidence intervals (CI) 0.55-1.0). The observed agreement on 148 pairs of observations between parents and nurses was 75% (kappa = 0.5; 95% CI 0.36-0.64). Between-parent agreement on 30 other paired observations was 77% (kappa = 0.54; 95% CI 0.24-0.84). Teaching parents about the four categories is a potentially useful adjunct for assessment of diarrheal stools in children.