Available data on multiple victimization are limited by variations in degree of productivity from respondents in victimization surveys, the imprecision of 'series' victimization in which the respondent claims several victimizations over a certain time period but cannot remember precise details, and the difficulty of selecting an appropriate time period for analyzing victimization survey data. Assuming that an observed distribution of victimization with an excess of multiple victims is real, several probablistic models that are modifications of the basic Poisson model are described as possible explanations for repeated victimization. The Greenwood-Yule model is based on heterogeneity or differing degrees of fixed risk or 'proneness' in the population. However, individuals' risk of victimization are viewed as partly a function of variable personal and social characteristics. The Eaton-Fortin study on schizophrenic patients is advanced as a relevant model because it develops a skewed distribution of incidents that could be used to allow for variations in victimization 'proneness.' Another possible model is described that postulates decreased 'proneness' to victimization after each incident due to increased precautions that are taken by the potential victim. Six different ways are described in which the actions, attributes, or social situations of victims may help to explain variations in victimization rates. For example, precipitation of offender behavior by the victim's actions and vulnerability to crime due to a social situation are possible factors leading to victimization. Future research on multiple victims should be directed toward developing sufficient indicators of 'proneness' to victimization. Reference notes and 28 references are provided.