Despite considerable attention from the survey research community, the effect of rapport between interviewer and respondent on survey responding remains a topic of considerable debate. Some researchers suggest that rapport is good for survey interviews because it motivates participants to help interviewers by giving honest responses, even on sensitive questions. Others suggest that rapport is bad for survey interviews because it causes respondents to attempt to ingratiate themselves to interviewers by distorting their responses, especially on sensitive questions, in order to appear more favorably to their interviewers. We propose an alternative explanation. While rapport has previously been considered a unified concept and has yielded different, often contradictory results, we argue that there are multiple types of rapport, some that motivate respondents to answer honestly and others that cause respondents to ingratiate themselves to their interviewers. In this talk, we will discuss our research on the different components of rapport in the survey interview. In our research, rather than depending on post-hoc analyses of interview behavior, or the more-or-less successful performance of trained survey interviewers, we have used prior research in social psychology and our own research on face-to-face interaction to derive a parameterized model of rapport and its effects. That model drives the performance of a series of virtual humans that differ in the kind of rapport they demonstrate. We then used those virtual humans to analyze the effects of different kinds of rapport on socially desirable responding in survey interviews containing sensitive questions.