A technique of numerical classification was applied to 13 soil series belonging to 5 great soil groups from Bellary, Bijapur and Dharwar districts of Karnataka [India] and the Union Territory of Goa. Data on 32 soil characteristics were collected. The qualitative characteristics were numerically coded and then the characteristics were converted to a uniform scale of 0 or 1-100, adjusting the intermediates proportionately. Taxonomic similarities between different soil series were calculated by computing coefficients of resemblance such as similarity indices, correlation coefficients and Fisher's Z-values for all possible combinations (13C2 = 78) of 13 soil series. Cluster analysis was later employed for grouping of these soils. The results, though generally confirming earlier groupings of these soils, indicated the extent of dissimilarities within the groups. This technique enabled measurement of the affinity among different soils in quantitative terms examination of the stages at which soils joined together to form major groups and thus reduced their number (groups) at desired level of similarity.