In standardized survey interviews with closed-ended questions it quite often occurs that respondents do something else than mentioning their choice from the set of response categories presented together with the question. In these cases their initial response is inadequate and interviewers have to initiate repair activities. By analyzing 804 question-answer sequences, stemming from parts of a survey with elderly people (N=201), we found that 322 (41%) of the initial responses were inadequate. In 74% of these 332 sequences the interviewers made an attempt at repair. Of all repair attempts, 75% were performed correctly, that is, in a nondirective way. Most of the repair attempts (86%) result in an adequate final response of the respondent. The result of the repair activities by the interviewers is that the number of sequences with an adequate response increased from 472 to 688, thus an increase of about 45%. There does not exist 'one best way' of repairing. The most important thing is that some repair initiative has to be taken by the interviewer and that the repair is done in a nondirective way. © Springer 2006.