Both specific and non-specific factors may play a role in acupuncture therapy for pain. We explored the cerebral consequences of needling and expectation with real acupuncture, placebo acupuncture and skin-prick, using a single-blind, randomized crossover design with 14 patients suffering from painful osteoarthritis, who were scanned with positron emission tomography (PET). The three interventions, all of which were sub-optimal acupuncture treatment, did not modify the patient's pain. The insula ipsilateral to the site of needling was activated to a greater extent during real acupuncture than during the placebo intervention. Real acupuncture and placebo (with the same expectation of effect as real acupuncture) caused greater activation than skin prick (no expectation of a therapeutic effect) in the right dorsolateral prefrontal cortex, anterior cingulate cortex, and midbrain. These results suggest that real acupuncture has a specific physiological effect and that patients' expectation and belief regarding a potentially beneficial treatment modulate activity in component areas of the reward system.