Some studies indicate that placebo analgesia is stronger when pre-conditioning with effective analgesic treatments is performed, thereby suggesting that the placebo response is a learning phenomenon. Here we further tested this hypothesis in order to better understand when and how previous experience affects the placebo analgesic response. To do this, we used a conditioning procedure whereby the intensity of painful stimulation was reduced surreptitiously, so as to make the subjects believe that an analgesic treatment was effective. This procedure induced strong placebo responses after minutes, and these responses, albeit reduced, lasted up to 4-7 days. In addition, in a second group of subjects we repeated the same conditioning procedure 4-7 days after a totally ineffective analgesic treatment, and found that the placebo responses were remarkably reduced compared to the first group. Thus we obtained small, medium and large placebo responses, depending on several factors, such as the previous positive or negative experience of an analgesic treatment and the time lag between the treatment and the placebo responses. We also ran extinction trials, and found that these effects did not undergo extinction in a time span of several minutes. These findings indicate that placebo analgesia is finely tuned by prior experience and these effects may last, albeit reduced, several days. These results emphasize that the placebo effect is a learning phenomenon in which many factors come into play, and may explain the large variability of the placebo responses that is found in many studies.