In this paper we provide a model of interviewer–respondent interaction in survey interviews. Our model is primarily focused on the occurrence of problems within this interaction that seem likely to affect data quality. Both conversational principles and cognitive processes, especially where they do not match the requirements of the respondent's task, are assumed to affect the course of interactions. The cognitive processes involved in answering a survey question are usually described by means of four steps: interpretation, retrieval, judgement and formatting. Each of these steps may be responsible for different overt problems, such as requests for clarification or inadequate answers. Such problems are likely to affect the course of the interaction through conversational principles which may cause, for example, suggestive behaviour on the part of the interviewer, which may in turn yield new problematic behaviours. However, the respondent may not be the only one who experiences cognitive problems; the interviewer may also have such problems, for example with respect to explaining question meaning to the respondent. Thus the model proposed here, unlike most of the other models which concentrate on the respondent, tries to incorporate cognitive processes and conversational principles with respect to both interviewer and respondent. In particular, the model looks at how cognitive processes and conversational principles affect both the interaction between interview participants and the quality of the eventual answers. Copyright © 2007 John Wiley & Sons, Ltd.