The collection conditions for a survey are not neutral as regards the quality of the information collected. In order to reduce biases, surveyors are instructed to ensure that interviews are conducted on a person-to-person basis, without the presence of third parties. However, they cannot impose the interview conditions and, in many cases, one person or another may take part in all or part of the interview. We thus also describe as accurately as possible the interview conditions for each interview in order to be able to evaluate bias in the responses. The methodology used for the Study of Family and Intergenerational Relationships (ERFI) carried out by INED and INSEE in 2005 allows us to describe the interview conditions more accurately than in other surveys. A quarter of interviews began or ended in the presence of a third person and comings and goings during interviews are not rare. Furthermore, when a person is present, that person is most often the interviewee's spouse. The importance attached to describing these interview conditions is justified in terms of the declarative biases that they can bring about. The example of the division of household tasks between a couple appears to indicate that the presence of the interviewee's spouse plays a "controlling" role on responses while their absence fosters greater intervention by the interviewee, giving rise to a smoother division which is more in line with a fairly egalitarian norm. Moreover, the spouse's presence seems to boost the workings of the interviewee's memory when it comes to dating a number of events in marital life. While these results appear to suggest that it is ultimately preferable for interviews to take place with the spouse in attendance, contrary to the instructions given to interviewees, the presence of a third party ­ particularly that of the interviewee's spouse ­ gives rise to a lesser enumeration of past marital events, mainly by the male partner.