Abstract This article explores different measurements of (sub-)national identities in survey research and examines to what extent they lead to different results. Using survey data from Belgium, where there is a long-standing tradition of (sub-)national identity surveys, three types of questions are scrutinized: the “hierarchical” question (asking respondents to which of a list of given identities they feel most closely related in first and second place), the “Linz-Moreno” question (asking respondents to situate their regional and national identities vis-à-vis each other), and the more recent “metric” question (asking respondents to situate themselves on distinct eleven-point scales for multiple identities). This article analyzes the extent to which respondents answer these questions consistently, how varying degrees of consistency can be explained, and what this tells us about the way social scientists measure (sub-)national identities. The results show that, depending on the question, only 39.4 percent to 69.2 percent of the respondents answered the three (sub-)national identity questions consistently. Differences in consistency are found to be not only related to respondents’ political knowledge and interest, but also to the question forms and wording, leading us to reflect on the validity of identity measurements.