We compared the inter-method differences shown by control materials and by patients' sera for the measurement of some plasma proteins in the same pair of analytical systems. Sets of 100 to 110 samples of patients' sera and of 18-19 control materials, including the recently available CRM 470, were assayed with up to five automatic analytical systems, in two different experiments. About 5500 values were produced and assessed statistically. Materials (either patients' sera or control materials) were considered non-commutable (i.e. exhibiting significantly different inter-method behaviour) when their distance from the regression line in a stated pair of methods exceeded 3 standard deviations. According to this criterion, less than 1.5% of the patients' sera, and an even lower proportion of control materials were non-commutable. However, the inter-method behaviour of control materials was usually slightly different from that of patients' sera. Some systematic inter-method difference in the measurements on patients' sera may therefore exist, even though inter-method equivalence has been demonstrated with control materials.