AbstractRun‐to‐run sample separation reproducibility has been compared on two commercial high performance capillary electrophoresis units which differ in the mode by which the capillary temperature is thermostatted. Three standard analytes, differing dramatically in molecular character and size, were used for the analysis: benzoic acid, a 14 amino acid peptide from human chorionic gonadotropin, and ribonuclease A represent, respectively, small stable organic molecules, small peptides with little or no secondary structure, and proteins with secondary structure. These standards were evaluated with regard to reproducibility of migration time, peak area, and peak height. The analyses, performed in buffers of optimum pH for the separations, demonstrated that the liquid and forced air convection thermostatted systems both performed extremely well. The reproducibility, as judged by the percent coefficient of variance (% CV) of replicate analyses, was generally found to be less than 1 % (migration time); the reproducibility decreased in the order migration time > peak height > peak area. Whereas the absolute % CV values for MTrel (migration relative to a standard) observed with the liquid thermostatted system were 2‐ to 4‐fold lower than those observed with the forced air convection thermostatted system, there was little statistically significant difference between the two. As expected, the data indicated a reduction in reproducibility as the complexity of the analyte increased, perhaps as the result of an increased potential for wall interactions. Comparing separations in which low (≈︁1 watt/meter [W/m] of capillary) and high (>5 W/m) Joule heat was generated by altering the sodium chloride content of the buffer revealed few statistically significant differences in the reproducibility obtained from the two systems. With these particular standard analytes and their respective buffer systems, there appears to be little difference between forced air convection and liquid thermostatting of the capillary.