Polarization measures that are used in examining the empirical relationship between ethnic divisions and violent conflict, heavily rely on mechanisms of group identification and often use somewhat arbitrary divisions of a society into ethnic groups. In this paper we construct two new measures of polarization, one that accounts for differences in linguistic policies across localities during the colonial era and one that accounts for the differences over time and across localities in the experience of violence throughout the conflict episode. By examining the protracted war in Sri Lanka and applying these indices (and their combination) to a data set describing victims of the civil conflict by district and year, we are able to better identify the effect of ethno-linguistic polarization on the civil conflict in the country. We find that, for each of our polarization indices, there is a positive effect on the conflict. The historical underpinnings of our indices allow us to demonstrate in a quantitative and concrete way the relevance of historical processes for understanding episodes of civil conflict.