Various subsets of pharmacological activity indices of benzodiazepines, of 8-quinolinol derivatives, and of rifamycin B amides were decomposed into mutually independent components by using principal component analysis. The activites not included into the subset were considered as the dependent variables. In three out of the six cases with a fair correlation coefficient (r less than or equal to 0.9) between pairs of primary pharmacological indices, the main component obtained by the decomposition procedure showed significantly higher correlation with the dependent variable than any of the original pharmacological activity indices. Factors, explaining a rather low portion of the total sample variance of the subset, may still account for important secondary effects.