Multi-level models permit more direct tests than previously possible of the importance of "proximity" relative to other concepts in the model, such as "exposure." Within opportunity theory, the concept of proximity is the intuitive notion that if one routinely conducts activities near motivated offenders, the risk of crime will increase. Domain-specific models of victimization predict categories of crime defined by activity at the time of the incident, rather than defined exclusively by the nature of the criminal act. By restricting attention to crimes occurring in a particular domain, such as the home, a more rigorous empirical test of opportunity theory is possible. Results of this study indicate that the concept of proximity should be given preeminence in the opportunity framework, and subdivisions of the concepts of guardianship, attractiveness, and exposure should be identified. Models should be explicitly multilevel, incorporating measures of activity concepts at various levels of aggregation rather than emphasizing only the individual level. Implications of the research for opportunity theory and for policy are also discussed. 25 references, 30 tables.