Migration is among the best studied of animal behaviors, yet few empirical studies have tested hypotheses explaining the ultimate causes of these cyclical annual movements. Fretwell's (1980) hypothesis predicts that if nest predation explains why many tropical birds migrate uphill to breed, then predation risk must be negatively associated with elevation. Data from 385 artificial nests spanning 2,740 m of elevation on the Atlantic slope of Costa Rica show an overall decline in predation with increasing elevation. However, nest predation risk was highest at intermediate elevations (500-650 m), not at lowest elevations. The proportion of nests depredated by different types of predators differed among elevations. These results imply that over half of the altitudinal migrant bird species in this region migrate to safer breeding areas than their non-breeding areas, suggesting that variation in nest predation risk could be an important benefit of uphill migrations of many species.