Microdialysis was used to assess the contribution to cocaine seeking of cholinergic input to the mesocorticolimbic dopamine system in ventral tegmental area (VTA). VTA acetylcholine (ACh) was elevated in animals lever pressing for intravenous cocaine and in cocaine-experienced and cocaine-naive animals passively receiving similar "yoked" injections. In cocaine-trained animals, the elevations comprised an initial (first hour) peak to approximately 160% of baseline and a subsequent plateau of 140% of baseline for the rest of the cocaine intake period. In cocaine-naive animals, yoked cocaine injections raised ACh levels to the 140% plateau but did not cause the initial 160% peak. In cocaine-trained animals that received unexpected saline (extinction conditions) rather than the expected cocaine, the initial peak was seen but the subsequent plateau was absent. VTA ACh levels played a causal role and were not just a correlate of cocaine seeking. Blocking muscarinic input to the VTA increased cocaine intake; the increase in intake offset the decrease in cholinergic input, resulting in the same VTA dopamine levels as were seen in the absence of the ACh antagonists. Increased VTA ACh levels (resulting from 10 microM VTA neostigmine infusion) increased VTA dopamine levels and reinstated cocaine seeking in cocaine-trained animals that had undergone extinction; these effects were strongly attenuated by local infusion of a muscarinic antagonist and weakly attenuated by a nicotinic antagonist. These findings identify two cholinergic responses to cocaine self-administration, an unconditioned response to cocaine itself and a conditioned response triggered by cocaine-predictive cues, and confirm that these cholinergic responses contribute to the control of cocaine seeking.