We report an evaluation of the MANTRELS clinical score in predicting appendicitis in a prospectively studied pediatric population presenting with abdominal pain. One hundred eighty-nine independent episodes from 187 children 2 to 17 years old were studied. For the groups as a whole and for the individual groups less than 16 years old, the score failed to satisfactorily discriminate those with appendicitis from those without. Had the MANTRELS score been used to determine observation and laparotomy in our patients, 21 additional patients would have been unnecessarily hospitalized, and 16 would have been subjected to unnecessary laparotomies. One patient would have received appropriate intervention earlier. In the 40 children 16 and 17 years old, the MANTRELS score adequately distinguished the two groups. A seven-variable discriminant function, derived from stepwise discriminant analysis, performed slightly better but showed essentially the same findings as the MANTRELS score. We believe the MANTRELS score failed to predict appendicitis in younger children because it does not contain variables that allow for separation of appendicitis from the numerous other conditions mimicking it in the pediatric population. The clinician remains the best judge of the acute abdomen in the pediatric age group.