To understand caste in India we must explain the particularity of this mode of stratification, while avoiding an essentialism that isolates caste from other social forms. Stratification appeared in India long before an emic model of caste. Yet in contemporary India, I found actors concerned to place themselves in a rank order, even in ephemeral situations. I outline an Indian concept of the individual, characterised by relative immunity of the self to the social sphere, and argue that this immunity acts as a shield to keep ranking apart from the self. Indian actors show their social capability by adapting to situations rather than imposing a consistent personality across them. Local stratification is explained by actors in a historical mode, as if history was an interactional sequence played out between communities rather than individuals. Indigenous models of society provide an alternative explanatory mode, as when Brahmins claim the superior position. Their holistic model, however, is matched by king-centred and merchant-centred models. Subalterns have yet other views of Indian society. I argue that holistic models in India are constructions of the dominant, and should not be taken to represent 'Indian culture'. Indian society should rather be seen as plural, with several emic models used to describe and explain it. Still, the tendency to create rank in so many situations points to particular rules of interaction and discourse, which implies talking 'as if' hierarchy, in the Dumontian sense, was an objective reality.I Caste remains enigmatic, and a problem in understanding social stratification in India. Dumont (1966) proposed a holistic, hierarchical model centred on the Brahmins, while Quigley (1993) argued for a king-centred model.