The rise of identity politics generates a tension between those who want to make particular identities the central and determining aspect of their social being and people who do not wish to do so or who refrain from doing so.1 The latter group, arguably the majority, mobilize such identi¿cations either only strategically (or tactically) or not at all.2 Formal identities, once accepted by the state, provide the basis for the politically implicated process of regular counting, i.e. censuses, and (eventually) the distribution of resources. Such identities are, therefore, the focus of much detailed attention from sociologists and other social scientists. This focus in the academic study of religion (or ethnicity) can easily give rise to a wellknown fallacy of misplaced concreteness: the fact that a category can be counted is wrongly taken to mean that a self-conscious group exists.3 Reality, however, rarely corresponds perfectly to academic theory and indeed often confounds expectations: for example, for many Nepalis with whom we have worked, the religious category into which they fall is by no means obvious and may depend on context. Furthermore, since the mobilization of categories is a process with status and political implications, in many cases there is a rather large gap between what people actually do in the privacy of their own shrine room and their declared identities when responding to census-takers or when participating in public meetings. ReÀecting on this gap, we have argued elsewhere that a fundamental

distinction is required for any adequate understanding of religious phenomena, namely that between category (census label) and practice (what people actually do) (Hausner and Gellner 2012). South Asian data bring out very clearly how very Àuid religious categories are and how fallacious it can be to assume that they correspond to distinct, homogeneous, and mutually exclusive social groups.