Common Ravens (Corvus corax) in the Mojave Desert of California, USA are subsidized by anthropogenic resources. Large numbers of nonbreeding ravens are attracted to human developments and thus are spatially restricted, whereas breeding ravens are distributed more evenly throughout the area. We investigated whether the spatial distribution of risk of predation by ravens to juveniles of the threatened desert tortoise (Gopherus agassizii) was determined by the spatial distribution of (1) nonbreeding ravens at human developments (leading to "spillover" predation) or (2) breeding individuals throughout developed and undeveloped areas (leading to " hyperpredation"). Predation risk, measured using styrofoam models of juvenile desert tortoises, was high near places attracting large numbers of nonbreeding ravens, near successful nests, and far from successful nests when large numbers of nonbreeding ravens were present. Patterns consistent with both "spillover" predation and "hyperpredation" were thus observed, attributed to the nonbreeding and breeding segments of the population, respectively. Furthermore, because locations of successful nests changed almost annually, consistent low-predation refugia for juvenile desert tortoises were nearly nonexistent. Consequently, anthropogenic resources for ravens could indirectly lead to the suppression, decline, or even extinction of desert tortoise populations.