Several other variables were hypothesized to have significant zero-order correlations which should be accounted for by the mediating effects of these central variables. The primary thesis is that contextual variables specify the correlates of personal protective behaviors. The hypothesis was tested through data collected by telephone interviews of 13 independently drawn samples in 10 neighborhoods in Philadelphia, Chicago, and San Francisco. A total of 2,695 interviews were completed. The major correlates of personal protective behaviors were found to be subjective estimates of personal danger and personal characteristics related to vulnerability, as measured by sex and age. Assessments of environmental danger, the interpersonal communication of crime information, and the perception of crime-related environmental cues are all related to personal protective behaviors, but only through their effects on estimates of personal danger. The comparative analysis indicated that special patterns of criminal victimization may affect the nature of the correlates of protective behaviors. After all of the major correlates of the dependent variable have been controlled, the residents of some of the neighborhoods studied still reported more protective behaviors than residents of other areas. Tabular data are provided, and the questionnaire is appended. A total of 70 references are listed.