The effects of a warning banner, informing patrons of the randomness of Video Lottery Terminal (VLT) outcomes, on gambling behaviour and beliefs were tested in a field setting using a mixed-model 2 x 3 design over a six-week period with 27 problem and 27 non-problem gamblers recruited from bars in a Canadian city with a population of 85,000. Overall, self-reported VLT gambling behaviour decreased significantly during the three, two-week periods. More specifically, gambling behaviour during the banner period was significantly less than during the baseline period. In addition, for only problem gamblers, significant decreases in faulty gambling beliefs were found across the three periods. This suggested that decreases in gambling behaviour were mediated by decreases in faulty gambling beliefs incurred by the banner of our study and can have practical importance in the prevention of problem gambling.