Questionnaires were mailed to a proportional, stratified, random sample of 743 households in Washington State. Of the 584 households willing to participate in the survey, 457 produced usable questionnaires. Two versions of the questionnaire were constructed: one in which all hypothetical offenders were female and one in which all hypothetical offenders were male. Both versions of the questionnaire contained four vignettes of hypothetical offenses (shoplifting, embezzlement, selling marijuana, and second degree murder). Following each vignette, the questionnaires solicited respondents' perceptions of the offender's personal characteristics and motives, crime seriousness along with opinions on case disposition, preferred social distance from the offender, and familiarity with similar offenders. Regression analysis was used to test hypotheses and the multivariate models. Findings show that perceived similarity between the respondent and offender is neither a strong nor significant influence on recommended sentences; however, the perceived seriousness of the crime exerted a significant impact on recommended sentences. Other variables, most notably familiarity with like offenders, may contribute to variations in recommended sentences for specific crimes. Future research should address the differential explanatory power of those variables found significant for various types of crimes. Tabular data and 43 references are provided.