The existing literature on Indian growth finds no evidence of convergence across states. This represents a puzzle given the relatively free flows of capital, labor and commodities across state borders. We use a new data set of district level income and socio-economic data to explore the hypotheses of conditional convergence, using distance as an indicator of internal geographical trade and migration costs. We find evidence of conditional convergence for Indian districts but at a rate that is only half of Barro's "Iron Law". The results suggest that district level differences in trade and transport costs, infrastructure, and literacy rates, as well as state-level effects, have all contributed to the lack of absolute convergence in India.