Earlier research on ethnic and religious conflict has identified fear as an important motivation. While theoretically sound, this expectation has never been tested at larger scales in ongoing episodes of political violence. Instead, conceptual progress has been made in lab experiments. Combining insights from observational research and stylized experiments, we predict that fear for personal safety due to witnessed violence causes prejudice against out-groups, enhanced internal cohesion, and support for extremist actors. To test these predictions, we conducted surveys in the Indian State of Uttar Pradesh with identical respondents in three waves starting in January 2017. The surveys continued during the tense Legislative Assembly elections in the Spring. The results largely corroborate the theoretical expectations and present a hard in-vivo test of long-standing conjectures.