Reproducibility of capsular serotypes of 55 consecutive clinical isolates of Klebsiella pneumoniae was evaluated by an indirect immunofluorescent-antibody technique previously described by Riser et al. (J. Clin. Pathol. 29: 296-304, 1976) Five colonies per specimen were examined for colony-to-colony variation, day-to-day variation, and reader-to-reader variation. Seventy-two reference strains were tested with each of 18 pools and 72 specific antisera prior to the clinical specimens to determine antiserum specificity and cross-reaction patterns. Lot-to-lot variation was examined with the reference strains. There was minimal lot-to-lot variation among the antisera tested. Ten antiserum pools required supplementation with individual antisera. The patterns of supplementation may vary from lot to lot. Colony-to-colony variation in intensity of immunofluorescence occurred, but there was no variation in serotype. These findings differ from previously reported colonial variation which occurred when API 20E biotypes were determined for individual colonies of K. pneumoniae directly from clinical specimens. Eighteen percent of clinical isolates studied gave cross-reactions when tested with the indicated specific antiserum. All but one of the cross-reactions were resolved with further dilutions. Day-to-day and reader-to-reader variations were minimal. The immunofluorescent-antibody technique is a reliable and reproducible method for capsular serotype determination. Capsular serotypes are less variable than API biotypes since colony-to-colony variation of serotype does not occur.