Genetic Algorithms (GA) that �reward� particular observables (e.g., forward travel) and penalize others (e.g., metabolic cost) have been used successfully to synthesize gaits (Sellers et al., 2004). These applications of GA rely on physical simulation (to model muscle moments and inertial and gravitational forces). Distinct from these approaches, we use GA to evolve purely kinematic solutions, which are then compared with gaits observed with extant animals and as derived by other methods.