Theories of radicalization make implicit predictions about variation among attitudes in the communities from which radicals are drawn. This article subjects some popular theories to empirical testing. The precise process by which individuals come to sympathize with, provide material support for or actually engage in political violence is fully comprehensible only by longitudinal analysis, but much existing work tries to reconstruct the process by looking only at one part of its outcomes: those who have become radicalized. The result is a large number of theories and mechanisms, with little compelling empirical support. A cross-sectional snapshot of an at-risk community cannot definitively support a particular theory of radicalization, but it can rule out those whose predictions about attitudes are at variance with the empirical observations. We designed a survey instrument to measure attitudes to issues widely believed to be relevant to radicalization and deployed it among Muslim communities in Ottawa. The results are remarkably inconsistent with patterns of variation in attitudes predicted by popular theories of radicalization. Instead, they show variation of attitudes along three independent dimensions: social/economic/political satisfaction/dissatisfaction, moral/religious satisfacton/dissatisfaction, and a dimension that appears to be associated with radicalization. This suggests that governments may have less policy leverage to mitigate radicalization than generally supposed. © 2012 Canadian Political Science Association.