Leadership is an active area of research in both the biological and social sciences. This review provides a transdisciplinary synthesis of biological and social-science views of leadership from an evolutionary perspective, and examines patterns of leadership in a set of small-scale human and non-human mammalian societies. We review empirical and theoretical work on leadership in four domains: movement, food acquisition, within-group conflict mediation, and between-group interactions. We categorize patterns of variation in leadership in five dimensions: distribution (across individuals), emergence (achieved versus inherited), power, relative payoff to leadership, and generality (across domains). We find that human leadership exhibits commonalities with and differences from the broader mammalian pattern, raising interesting theoretical and empirical issues. Leadership is an active research area in both biological and social sciences, but there has been limited synthesis within or across these areas; evolutionary theory can assist with such synthesis, but additional elements are needed for a robust comparative framework. Variation in leadership can be measured in multiple dimensions, including emergence (how does one become a leader?), distribution (how widely shared is leadership?), power (how much power do leaders wield over followers?), relative benefit (do leaders gain more or less than followers?), and generality (how likely are leaders in one domain, such as movement or conflict resolution, to lead in other domains?).A comparative framework based on these dimensions can reveal commonalities and differences among leaders in mammalian societies, including human societies.