The nonrandom nature and predictive efficiency of probabilistic exposure to risk in the construction of criminal victimization data are analyzed in a discussion of the etiology of criminal victimization. One paper contends that criminal justice studies employing behavioral indices can better gauge the effect of precautionary measures upon group crime rates by focusing more upon general behavioral dimensions, improved methodologies, realistic models of human processes, and research on the impact of crime prevention efforts. The extent to which victims are also offenders involved in serious assault is investigated in a followup survey of self-reported victimization subjects to determine whether findings support the thesis of a violent subculture. Another paper argues for a shift in victimization research from measuring the incidence of victimization to defining the environment in which victimization occurs and describing victim-offender interaction during the crime event. A general model of the causes and consequences of the fear of crime is presented, and components of the model are described in relation to research findings. The next paper reviews the evidence for multiple victimization, assesses mathematical models for measuring multiple victimization distribution, and advocates the use of social and economic measures in studying victimization. Three aspects of measurement error in victimization surveying are examined; the amount of error contained in survey-generated estimates of victimization, the net direction of that error, and the correlates of error. Lastly, the strengths and weaknesses of various sources of victimization data are identified and evaluated. Tables, graphs, charts, and chapter references and notes are included. For individual papers in this volume, see NCJ 76276-83.