The analysis of all episodes of peritonitis occurring in a uniformly treated continuous ambulatory peritoneal dialysis (CAPD) population (N = 128) at one centre during a six-year period showed the following major findings. The initial white cell count (WCC) of the dialysate was less than 100 x 10(6)/L in 10% of the episodes and showed a predominance of mononuclear cells in 15%. The Gram stain results were consistent with the findings of the culture in 28% of the episodes and influenced the initial therapy in only 7% of the cases. Between 9% and 31% of all episodes would not have been classified as peritonitis if positive culture, a WCC of greater than 100 x 10(6)/L in the dialysate, or clinical symptoms had been required for the diagnosis. The proportion of negative dialysate cultures was 2% after the introduction of pre-culture membrane filtration. Tunnel infection as a cause of peritonitis was more frequent in episodes due to Staphylococcus aureus than in episodes due to coagulase-negative staphylococci (p = 0.009). Peritonitis caused by coagulase negative staphylococci were followed by a milder course than other organisms (p = 0.02). Of all episodes initially treated with cephradine only 62% were cured with this antibiotic (or cloxacillin) and 35% were followed by recurrency, protracted course or catheter loss, despite intermediate or full in vitro susceptibility. In only 53% of all episodes no complication was observed. Complications were more frequent in women and diabetics than in men (p = 0.01) and non-diabetics (p = 0.03), and were more common in episodes with clinical symptoms (p = 0.02). Peritonitis resulted in drop-out from CAPD in 6% of all episodes. Hospital care was needed in 68% of all episodes. We conclude that turbidity can be used as the sole criterion for the initial diagnosis of peritonitis, and that a first generation cephalosporin should not be used as a first line antibiotic in the treatment of CAPD peritonitis.