Why do subjects accept information that is typical but false? The present study analyses the effects of scripts on the creation of these false memories. Based on script information established in a normative data study, a brief narrative account of a mock mugging was elaborated with high- and low-typicality contents. Free recall turned out to be a synthesis of the event, centred more on typical actions than on low-typicality ones, and although there were few errors, they were actually script-consistent contents. In recognition the typicality of the sentences was systematically manipulated. There were more hits in high-typicality actions, but the accuracy rate was greater in low than in high-probability information due to the fact that the subjects accepted nearly half of the unstated typical actions. These false alarms received high confidence scores, showing that for the subjects these false memories represented real recollections. Furthermore, false memories, generated by activating our prior knowledge, were persistent after a week's delay and resistant even to explicit instructions warning against such errors.