This dissertation combines ethnographic, interview, and documentary data gathered in Ethiopia to analyze famine early warning systems - large-scale information systems bringing together climate, food production, market, and public health data, which have become a major focus in international famine prevention efforts since the 1970s. Famine early warning is part of a top-down international response that even many early warning experts themselves feel has been unable to solve the problems leading to famine. In Ethiopia, several different international, governmental, and nongovernmental organizations with different agendas have developed their own early warning systems. The information upon which humanitarian agencies act is a product of this complex political environment as well as particular understandings of famine that become fixed in the systems through the choice of indicators and analytic methods. The efforts of early warning practitioners and humanitarian agencies to separate technical aspects of early warning from their politics enables responses to famine that inevitably fail to address the needs of famine-affected people. Though the goal of early warning is to save lives by providing timely and accurate information, early warning experts have come to value objectivity, consensus, and transparency over accuracy because of constraints and pressures in their organizational environment. This, combined with donor governments' need for information they feel is credible, privileges certain kinds of expertise (that of early warning experts) over others (e.g. that of famine-affected people). Therefore, those officially responding to famine see those with the most direct experience of famine as lacking relevant knowledge and thus only capable of participating as passive aid recipients. The needs of famine-affected people can be better addressed by taking their experiences of and responses to famine into account and transforming the endeavors of famine early warning and response accordingly