The finding that ascending cholinergic systems are severely degenerated in Alzheimer's disease has driven the search for a cholinomimetic therapy. Adverse effects observed with cholinesterase inhibitors and high-efficacy muscarinic agonists led us to design compounds with an improved profile. SB 202026 (R-(Z)-(+)-alpha-(methoxyimino)-1-azabicyclo[2.2.2] octane-3-acetonitrile) displaced [3H]-oxotremorine-M from muscarinic receptors in the rat brain with high affinity (IC50 = 14 nM), a potency similar to that of oxotremorine-M itself (IC50 = 13 nM), but exhibited low affinity for cholinergic nicotinic receptors and other neuroreceptors. In studies using cloned human muscarinic receptors, SB 202026 possessed approximately equal affinity in displacing [3H]-quinuclidinyl benzilate from all muscarinic receptor subtypes. In functional models in vitro, SB 202026 caused maximal depolarization of the rat superior cervical ganglion at low concentrations (300 nM) (M1-mediated effect), while producing a lower maximal effect than the high-efficacy agonists oxotremorine-M and carbachol on M2-mediated release of ACh and M3-mediated smooth muscle contraction (guinea pig ileum), respectively. The functional selectivity and partial agonist profile seen in vitro were reflected in vivo through potent cognition-related activity (M1-induced increase in hippocampal EEG power) combined with low efficacy, compared with arecoline or oxotremorine, on induction of bradycardia (M2-mediated response), hypotension (via M3-mediated vasorelaxation) and tremor (thought to be mediated by M3 receptors). The foregoing profile of SB 202026 predicted that cognition-enhancing activity would be achieved at doses below those that initiate undesirable side effects, and this has subsequently been demonstrated in rodents, marmosets and humans.