Many of people's beliefs about crime appear unrelated to their behavior. The more general and impersonal those beliefs, the less rooted they are in experience and surrounding conditions. Assessments of risk of victimization are more strongly linked to people's reports of what they do about crime. There is some evidence that assessments of risks of victimization are related to neighborhood conditions and the perception of personal vulnerability to victimization. Measures of fear of crime are particularly strong correlates of protective behavior. The relationship between perception of crime and behavior becomes clearer by examining alternative models of how the two might be linked. Except for Durkheim's 'naive sociological model,' none of the models assumes a simple fit between perception and action. Conklin's alternative to Durkheim emphasizes how perceived risks and fear make it more difficult for people to do much beyond withdraw from the community. The utilitarian model of behavior focuses on the costs of protective behavior and indicates that people endure risks at some acceptable level which takes into account these costs. The health-belief model emphasizes how features of countermeasures to crime shape the frequency of their adoption. The opportunity model points out that fear and perceived risk influence 'behavior intentions,' but a number of practical constraints inhibit the capacity of people to act on their intentions. Of the four categories of reactions to crime reviewed (suburban flight, community involvement, household protection, and personal precautions), only personal precautions seem directly encouraged by fear of crime. Tabular and graphic data and 29 references are provided. (Author summary modified)