The effects of a variety of adenosine A1 and A2 antagonists on N6-((R)-phenylisopropyl)adenosine (R-PIA)- and scopolamine-induced amnesias were investigated in rodents in order to clarify the role of adenosine receptors in learning and memory. Some of the selective adenosine A1 antagonists exhibited antiamnesic activities at several doses where they did not induce an increase of spontaneous locomotion. These results suggest that the blockade of A1 receptors is more important than that of A2 receptors in learning and memory. Detailed studies of structure-activity relationships of adenosine A1 antagonists in two amnesia models demonstrated that there were three types of adenosine A1 antagonists: (A) Compounds 3-5 (8-substituted 1,3-dipropylxanthines) ameliorated the shortened latency in both models. (B) Compounds 7-11 (8-substituted 1,3-dialkylxanthines) and 19-21 (imidazo[2,1-i]purin-5(4H)-one derivatives) ameliorated the shortened latency in the (R)-PIA-induced amnesia model but not in the scopolamine-induced amnesia model. (C) Compounds 14-16 ameliorated the shortened latency in the scopolamine model but not in the (R)-PIA model. Aminophenethyl-substituted compounds C did not exhibit adenosine A1 antagonism in vivo presumably due to rapid metabolism. The dramatic change in the activities of A and B could not be explained by their simple pharmacokinetic differences because both types of compounds showed clear blockade of central adenosine A1 receptors in the (R)-PIA model. 8-(3-Dicyclopropylmethyl)-1,3-dipropylxanthine (5) (KF15372) was chosen for further studies and is currently under preclinical development as a cognition enhancer.