Humans continuously learn the properties of their environment, allowing them to predict and prepare for future (sensory) events. Bayesian predictive processing posits that these expectations play a major role in sensory processing, perception being the result of their integration with sensory evidence in this framework. Placebo hypoalgesia (when expectations of pain relief are sufficient to relieve pain) is probably the best example of expectations biassing perception. In this experiment, we manipulated pain expectations using a classic probabilistic reversal learning task (2 visual cues associated with the probability of the next stimulus being painful/non-painful) and assessed whether the participants' predictions (used as a proxy for their expectations) would bias their ability to identify the stimuli as being painful/non-painful and the intensity ratings they provided for these stimuli. Fifty healthy volunteers were administered the task twice (160 trials each time), at a one week interval. Data analysis was performed using mixed effects generalised linear models, grouping by participants and sessions.The analysis showed that the participants could predict and identify the stimulus quality above chance level and that the intensity ratings were significantly different for the two types of stimuli (all p<0.001). Furthermore, participant’s predictions biassed their perception: stimuli were perceived more often as painful and rated higher if the stimulus was predicted to be painful (p<0.001 and p=0.045).Therefore, in line with predictive processing, this study provides new evidence that participants’ expectations significantly influence their perception of pain during a classic probabilistic reversal learning task.