Survey questions about sleep duration: does asking separately about weekdays and weekends matter?

There is no standard way to ask about sleep duration in surveys, and little is known about the comparability of different questions. This article reports on a randomized survey experiment designed to test 1 type of variation: 1 question about usual sleep hours versus 2 questions separately asking about weekday and weekend sleep. Mean sleep duration was significantly shorter (7.03 hr vs. 7.28 hr) for the single question than a weighted average from 2 questions, and race and ethnicity contrasts differed by question format. Correlations between sleep duration and reported sleep need were significantly higher for the 2-question format. These results demonstrate that question wording can affect reported sleep duration in complicated ways, making it difficult to compare studies that use different questions.