Jason Lyall's Divided Armies asks, what affects battlefield performance in conventional wars? The book joins a long scholarly tradition focusing on non-material factors such as culture, human capital, regime type, and nationalism as potential determinants of battlefield performance. Divided Armies proposes that societies experiencing ethnic inequality, meaning discrimination and perhaps violence, exhibit lower levels of battlefield performance as compared with societies that avoid such inequality. Ethnic inequality undermines battlefield performance in several ways, including reducing trust between military units and undermining the willingness of aggrieved minority soldiers to fight. Divided Armies contains impressive empirical tests. Lyall and his research team built from the ground up a completely new data set of conventional wars, defined as military conflicts inflicting at least 500 battle dead in which belligerents engage in 'direct battle', as opposed to guerrilla or insurgency tactics. This data set, termed the Mars project, includes some colonial and civil wars, such as the American Civil War, in which conventional tactics were used. Usefully, it includes a substantial number of conflicts that are omitted from the mainstream Correlates of War lists of intrastate, extra-systemic, and interstate wars. The book tests its inequality thesis on the outcomes of these conventional wars, finding support. Beyond the large sample cross conflict tests, the book also presents several deeply researched case studies of battles in Africa, the Middle East and Europe. These studies present fine-grained evidence supporting the causal pathways described by the theory. Divided Armies is truly an extraordinary achievement, providing new ideas about the non-material determinants of battlefield performance, and presenting compelling quantitative and qualitative evidence. It will no doubt shape research agendas on war for many years to come.