Uremic plasma and dialysis fluid were separated by high speed gel filtration followed by gradient elution ion exchange chromatography into several UV-absorbing fractions containing middle molecules (MM) (7a, b, c, etc.), which were quantitated by integrating the peak areas of the chromatograms. Dialyzer clearance of MM was determined in vivo for two dialyzers, Gambro Optima and RP-6, but no difference could be found in spite of far better in vitro MM clearance of RP-6 than of Gambro Optima. Nor was there any difference in post-dialysis decrease in MM between the two dialyzers, or between them and two dialyzers with larger surface areas (Gambro Major and Dow HK-5). In 45 determinations of MM in blood obtained before and after dialysis, using different dialyzers, the mean decrease in MM was significantly larger than for creatinine, suggesting a much smaller distribution volume immediately available for exchange. Ultrafiltration without dialysis in 7 patients induced a significant decrease in plasma MM (7a, b, c, and g) in spite of stable osmolality, urea, creatinine, and sodium concentrations in plasma, suggesting that the reduction in body water affected the net production rate of these MM. In 5 studies, using RP-6 and a 75 l recirculating dialyzate, a rapid post-dialysis rebound was observed for MM in plasma, sometimes followed by a secondary decrease in plasma concentration. The volume of distribution of dialyzed MM, assessed from pre- and post-plasma concentrations and concentrations in the dialyzate, appeared to be much smaller than for urea and creatinine. In a stable dialysis population of 17 patients, significant negative correlations were found between pre-dialysis plasma concentrations of some MM (7a, 7b, 7f, 7g) and dialysis index or endogenous renal creatinine clearance. No such correlations were found for peaks 7c and 7d. The good correlation between 7b and dialysis index (r = 0.73) suggests that plasma determination of 7b may be used as a measure of the efficiency of MM removal.