Some scripted interview questions, built as multi-unit turns, generate a response before they reach their prescribed completion. Such ‘untimely’ interruption and premature production of responses may generate flawed data. In this article I will describe how untimely interruption also presents a problem with regard to non-scripted follow-up questions – or ‘probes’ – to closed survey items. I will try to show that some probing formats are more vulnerable than others to being interrupted before the listing of a second response option. I shall also consider the more general implications that probing format has for respondents’ understanding of ‘the ground rules’ for participating in a standardized interview. Current practices and recommendations for probing will be discussed in the light of the objective of collecting valid and reliable data.