This article reports on data collected from 8,276 clients by 22 interviewers on gender, race/ethnicity, and age, to assess the relationship between interviewer characteristics and disclosure of physical and sexual abuse by clients. Multilevel regressions that adjusted for the clustering of clients within interviewers were compared to unadjusted logistic regressions to determine the effect of response similarity within clusters. A study limitation was the fact that only one question each was asked for each type of abuse. Multilevel analyses that accounted for the interdependence of responses within interviewers were compared to logistic regression analyses that assumed independent response. Though clustering accounted for 2 to 5 percent of the unexplained variance, ignoring the clustering effect generated several misleading results. Adjusted models indicated that disclosure of physical abuse was more likely to occur to a Caucasian interviewer than to an African American interviewer, and more likely to female than male interviewers. It was found that disclosure of sexual or physical abuse did not increase when clients and interviewers were matched by gender, race, or age. In conclusion, it is emphasized that because these findings are specific only to substance users in the context of a brief interview, generalization of results is cautioned against. Tables, references