What does 'likely' mean, when respondents estimate the risk to become a victim of crime? We apply prospect theory to derive hypotheses of framing-effects on perceptions of risk. Victimization risks can either be interpreted as gains (?being spared of offences?) or as losses (?becoming a victim of crime?). Because losses are perceived as more severe, respondents will state lower subjective victimization probabilities in the loss-frame, compared to the gain-frame. We demonstrate such a framing-effect with data from an experimental survey. Furthermore, we show that the meaning of vague quantifiers varies with the frequency and the severity of the event. Respondents assign to the same vague quantifiers (e.g. 'unlikely') higher likelihoods in terms of percentages for frequent and for less severe events than for infrequent and for severe events. Because respondents do not use vague quantifiers consistently, it is problematic to compare subjective risks for different victimizations.