We examined the potential for some of the abnormalities of vascular endothelium found in diabetes mellitus to cause neuropathic changes. Non-diabetic rats were treated for 2 months with the cyclo-oxygenase inhibitor flurbiprofen (5 mg . kg-1 . day-1) to reduce prostacyclin production, the nitric oxide synthase inhibitor N(G)-nitro-L-arginine (5 or 25 mg . kg-1 . day-1), or combined treatment. There were dose-dependent reductions in sciatic motor and saphenous sensory conduction velocity. The two inhibitors acted synergistically, thus, the 5-6 % motor conduction deficits (p < 0.01) produced by either flurbiprofen or N(G)-nitro-L-arginine (5 mg . kg-1 . day-1) increased to 17 % (p < 0.001) for combined treatment. With N(G)-nitro-L-arginine (25 mg . kg-1 . day-1) and flurbiprofen, motor and sensory conduction velocity were reduced by 23 % (p < 0.001) and 12 % (p < 0.001), respectively, matching the deficits following 2-month streptozotocin diabetes. N(G)-nitro-L-arginine (25 mg . kg-1 . day-1) and flurbiprofen together produced a 13 % prolongation of the time taken for 80 % hypoxic conduction failure in vitro (p < 0.05) and a 10 % reduction in sciatic capillary density. A second investigation tested an alternative hypothesis that overproduction of nitric oxide was responsible for vascular-related complications in diabetes, the prediction being that N(G)-nitro-L-arginine (5 mg . kg-1 . day-1) would prevent nerve dysfunction. However, rather than prophylaxis during 2-month streptozotocin diabetes, treatment exacerbated nerve abnormalities. Thus, N(G)-nitro-L-arginine worsened (8 %, p < 0.001) the motor conduction deficit, there was an 11 % increase in hypoxic conduction failure time (p < 0.01) and an 11 % reduction in endoneurial capillary density (p < 0.01). We conclude that overproduction of nitric oxide is unlikely to be involved in the aetiology of experimental diabetic neuropathy. However, endothelial dysfunction resulting in impaired nitric oxide and prostacyclin synthesis could make a substantial contribution.