1. An [3H]-adenine pre-labelling methodology was employed to assay cyclic AMP generation by adenosine analogues in Chinese hamster ovary (CHO.A2B4) cells, transfected with cDNA which has been proposed to code for the human brain A2B adenosine receptor, and in guinea-pig cerebral cortical slices. 2. Adenosine analogues showing the following rank order of potency in the CHO.A2B4 cells (pD2 value): 5'-N-ethylcarboxamidoadenosine (NECA, 5.91) > adenosine (5.69) > 2-chloroadenosine (5.27) > N6-(2-(4-aminophenyl)-ethylamino)adenosine (APNEA, 4.06). The purportedly A2A-selective agonist, CGS 21680, failed to elicit a significant stimulation of cyclic AMP generation at concentrations up to 10 microM in CHO.A2B4 cells. In the guinea-pig cerebral cortex, NECA was more potent than APNEA with pD2 values of 5.91 and 4.60, respectively. 3. Of these agents, NECA was observed to exhibit the greatest intrinsic activity in CHO.A2B4 cells (ca. 10 fold stimulation of cyclic AMP), while, in comparison, maximal responses to adenosine (32% NECA response), 2-chloroadenosine (61%), and APNEA (73%) were reduced. 4. Antagonists of NECA-evoked cyclic AMP generation showed the rank order of apparent affinity (apparent pA2 value in CHO.A2B4 cells: guinea-pig cerebral cortex): XAC (7.89: 7.46) > CGS 15943 (7.75: 7.33) > DPCPX (7.16: 6.91) > PD 115,199 (6.95: 6.39) > 8FB-PTP (6.52: 6.55) > 3-propylxanthine (4.63: 4.59). 5. We conclude that, using the agents tested, the A2B adenosine receptor cloned from human brain expressed in Chinese hamster ovary cells exhibits an identical pharmacological profile to native A2B receptors in guinea-pig brain.