104 cAMP analogs, most of them modified in the adenine moiety, were tested as activators of cAMP-dependent protein kinase I (from rabbit or rat skeletal muscle) and kinase II (from bovine heart or rat skeletal muscle). When tested singly, only 2-phenyl-1,N6-etheno-cAMP showed a considerably (sevenfold) higher potency as an activator of kinase II than of kinase I. Analogs containing an 8-amino modification preferentially activated kinase I, some being more than 10-fold more potent as activators of kinase I than kinase II. When two analogs were combined, the concentration of one (complementary) analog required to half-maximally activate each isozyme was determined in the presence of a fixed concentration of another (priming) analog. Analogs tested in combination had been analyzed for their affinity for the intrasubunit binding sites (A, B) of isozyme I and II. The degree to which complementary analogs preferentially activated one isozyme was plotted against the mean site selectivity, i.e. (affinity A/B isozyme I X affinity A/B isozyme II) 1/2. This plot produced a straight line, the slope of which reflected the ability of the priming analog to discriminate homologous sites on the isozymes. This means that the isozyme discriminating power of an analog pair can be quantitatively predicted from the affinity of the analogs for site A and B of the two enzymes. It also means that a systematic analysis of those features of analogs imparting a high mean site selectivity or the ability to discriminate between homologous isozyme sites will facilitate the synthesis of new even more isozyme-selective analogs.