Although pooled human serum diluent is advocated in the serum dilution test, its use may compensate for protein binding defects in patients and yield nonrepresentative titers. To test this hypothesis, comparison was made of serum ultrafiltrate (molecular weight cutoff .ltoreq.30,000) serially diluted into either pooled serum ultrafiltrate or Mueller-Hinton broth with patient serum samples diluted into pooled human serum in 111 assays from 55 patients and 6 volunteers. Of 111 bactericidal titers in ultrafiltrate and/or Mueller-Hinton broth, 101 were within a single twofold dilution of titers in pooled human serum. Nine of 10 discordant titer involved highly bound drugs and were usually higher in ultrafiltrate than in pooled human serum. In seven additional volunteers with renal failure, titers in ultrafiltrate and in each volunteer''s serum were higher than those diluted in pooled human serum (P = .002). Recommended methods using pooled serum diluent may not accurately predict actual bactericidal titers in patients with abnormal protein binding.