Using the National Sample Survey (NSS) household data for 2004, we analyse the determinants of individual participation in non-agricultural activities in rural and urban India. First, a multinomial logit estimation (MLE) is carried out to throw light on the determinants of (i) labour market participation of individuals as workers in agricultural and non-agricultural activities, and of (ii) self-employment in these two groups of activities. School and technical education, on the hand, and infrastructure (measured, for example, by road length per 100 square kilometres, number of telephone and mobile phone connections per 100 people), on the other, shape these choices, among other factors. Second, based on an MLE, the determinants of employment within non-farm activities (e.g. food processing, manufacturing and trading) are analysed. Again, school and technical education, and infrastructure have important roles. Other variables matter too. Those from socially disadvantaged groups (scheduled tribes or castes), for example, have lower probabilities of employment in manufacturing and trading. Finally, some forms of employment in non-agricultural activities have significant poverty reducing or welfare enhancing effects. Self-employment in non-agriculture is a case in point. Contrary to a widespread presumption in the development literature that non-farm activities expand easily to reduce poverty, it is emphasised that both poverty reduction and participation in non-agricultural activities depend on infrastructural development. Controlling for these and other effects, greater participation in non-agricultural activities reduces poverty.