In a standardized personal interview, elderly (65+) Dutch respondents (N = 233), were asked detailed retrospective questions about six physical activities like walking, cycling and their performance of household tasks. Surprisingly, the proportion of inadequate answers was small, suggesting that the interviewers-four professional research nurses-have done their very best to eventually obtain adequate answers. They used three different types of control of the interview process. Firstly, an optimal execution of the prescribed 'open loop control', that is, precisely following the text of the questionnaire. Secondly, via the 'feedback loop control' of repair: if nevertheless inadequate answers are given, further probing is performed until eventually an adequate answer is obtained. Thirdly, by decomposing a general question into its components, making the question easier to answer (partial questioning). Interaction analysis of transcripts of the interviews showed that the type of control exercised by the interviewers, strongly differed by topic of the question. It appeared that question topics requiring more complex cognitive activities of the respondents, not only lead to more (need for) repair, but also to larger differences between the interviewers concerning their readiness to use partial questioning as an alternative for the open loop control as designed by the researcher. © Springer Science+Business Media B.V. 2008.