The rivers of South Africa differ in water chemistry because of differences in geology and climate, and in the nature of the terrestrial vegetation. Thus the riverine biotas also differ in their water quality requirements. This paper describes the use of multivariate analytical techniques on a large inorganic chemical database as a means of dividing the country's rivers into regions of like water chemistry for the purposes of water quality management of aquatic ecosystems. Data were used from about 500 usable sites for the three earliest hydrological years (October 1980–September 1983 inclusive) for which information is available for most of the sites. In some cases separate winter (May–September) and summer (November–March) analyses were performed. Those observations in which orthophosphate-phosphorus was greater than 0.1 mg 1-1 and/or nitrate- plus nitrite-nitrogen was greater than 0.5 mg 1-1 and/or conductivity > 500 mS m-1, were excluded. Some of the analyses were run on the subset of records for which the reading of weir height (as a surrogate for discharge) was within 10% of the mean. Chemical variables comprised conductivity and pH, and the concentrations of chloride, total alkalinity (TAL), sodium, calcium, potassium, magnesium, chloride, sulphate, fluoride, silicate, nitrate- plus nitrite-nitrogen, ammonium-nitrogen and orthophosphate-phosphorus, as well as the ratios Cl-:(Cl- + TAL), and Na+:(Na+ + Ca2+). Multivariate techniques included principal components analysis, detrended correspondence analysis and cluster analysis. Eigenvalues are low (< 0.05) for detrended correspondence analyses but high (> 0.7) for principal components analyses. In all cases, [silicate] correlates most strongly with the first axis, and conductivity, [K+] and/or with the second. Alkalinity correlates inversely with [H+], while [Ca2+] and [Mg2+] are closely correlated with each other and with alkalinity. [Na+] and [Cl-] are correlated with each other and weakly with conductivity. Cluster analysis of secondary drainage regions is used together with biogeographic and physiographic information to produce a map of South Africa divided into five major regions for the management of water quality for riverine ecosystems.