Horses have been associated with human societies for millennia, and for many have come to symbolize wildness, power, resilience, and freedom. Although equids were extirpated from North America 10 000-12 000 years ago, descendants of domestic horses now roam freely in the USA and 17 other countries across six continents. In landscape-scale and experimental investigations, free-roaming horses (Equus caballus) have been shown to induce numerous alterations to native-ecosystem components and processes through influences on soil, water, plants, native ungulates, small mammals, birds, reptiles, and other aspects of biodiversity. However, we argue that the management of free-roaming horses both in the U.S. and globally has been complicated by “socio-ecological mismatches.” These three mismatches stem from discrepancies between: 1) steady-state (horse-management) goals vs. a stochastically variable environment; 2) different methods to evaluate evidence, resolve conflicting beliefs, and achieve objectives, in scientific vs. political processes; and 3) differing spatial and temporal scales between policy and management, and values of science vs. values of the general public. Such mismatches arise from an inability to reconcile conflicting processes and functions in a social-ecological system, often reflecting differences in the spatio-temporal scales at which diverse components operate. Here, we describe three types of mismatches, and illustrate how the ecological dynamics of aridlands generally fit poorly with existing approaches to horse management and policy. Such mismatches complicate cost-effective management of free-roaming horses and the ecosystems they inhabit, and reduce the palette of potential solutions.