Sabcomeline, (SB-202026 [R-(Z)-alpha-(methoxyimino)-1 -azabicyclo [2.2.2] octane-3-acetonitrile]), a functionally selective muscarinic M1 receptor partial agonist, was tested in rats trained to perform a delayed, reinforced alternation task in a T maze, a test of short-term spatial memory. For comparison the cholinesterase inhibitor tacrine (THA-9-amino- 1,2,3,4-tetrahydroaminoacridine) and the non-selective muscarinic receptor agonist RS86 (2-ethyl-8-methyl-2,8 diazospiro [4.5]-decane-1,3-dione hydrobromide) were also tested and all three compounds were also compared using a conditioned taste aversion (CTA) task. Sabcomeline (0.001-1.0 mg/kg i.p.) significantly reversed the T-maze choice accuracy deficit induced by a 20-s delay at 0.03 and 0.1 mg/kg. RS86 (0.1-3.0 mg/kg i.p.) reversed the deficit at 1.0 mg/kg and THA (0.1-3.0 mg/kg i.p.) had no effect at any dose. All three compounds induced conditioned taste aversion with minimum effective doses (MED) of 0.3, 1.0 and 3.0 mg/kg, respectively. The results show that sabcomeline reverses delay induced deficits in T-maze choice accuracy in a rewarded alternation task at doses approximately 10 times lower than those required to induce conditioned taste aversion. RS86 was equipotent in both tests. These data support the findings of clinical studies which have shown that SB-202026 provides significant symptomatic improvement in patients with probable Alzheimer's disease at doses which do not induce cholinergic side effects.