Does mentioning "some people" and "other people" in a survey question increase the accuracy of adolescents' self-reports?

A great deal of developmental research has relied on self-reports solicited using the "some/other" question format ("Some students think that… but other students think that…"). This article reports tests of the assumptions underlying its use: that it conveys to adolescents that socially undesirable attitudes, beliefs, or behaviors are not uncommon and legitimizes reporting them, yielding more valid self-reports than would be obtained by "direct" questions, which do not mention what other people think or do. A meta-analysis of 11 experiments embedded in four surveys of diverse samples of adolescents did not support the assumption that the some/other form increases validity. Although the some/other form led adolescents to think that undesirable attitudes, beliefs, or behaviors were more common and resulted in more reports of those attitudes and behaviors, answers to some/other questions were lower in criterion validity than were answers to direct questions. Because some/other questions take longer to ask and answer and require greater cognitive effort from participants (because they involve more words), and because they decrease measurement accuracy, the some/other question format seems best avoided.