We investigate the relationship between violence and economic risk preferences in Afghanistan combining: (i) a two-part experimental procedure identifying risk preferences, violations of Expected Utility, and specific preferences for certainty; (ii) controlled recollection of fear based on established methods from psychology; and (iii) administrative violence data from precisely geocoded military records. We document a specific preference for certainty in violation of Expected Utility. The preference for certainty, which we term a Certainty Premium, is exacerbated by the combination of violent exposure and controlled fearful recollections. The results have implications for risk taking and are potentially actionable for policymakers and marketers.