We studied patterns of removal and predation on seeds of ten plant species over two years in four abandoned pastures adjacent to forest fragments in Costa Rica. We hypothesized that: (1) removal within pastures would be greater nearest forest fragments and decline at farther distances; (2) removal rates would be greater on smaller-seeded species; and (3) removal rates would differ among pastures. We placed seeds of two species in three pastures in 1997 and eight species in two pastures during 1998. In each pasture, removal was monitored at three distances from the forest edge. Rates of seed removal in 1997 and 1998 were significantly greater 1–5 m from the forest than at 20 m distances, but removal rates at the 40 m distance did not differ from the 1–5 m distance. Rates of removal for both species were low in 1997, although removal was significantly greater for the smaller-seeded species Sorocea trophoides than for the large seeds of Nectandra sp. Removal rates differed significantly among pastures for both species in 1997 and among pastures for five small-seeded species in 1998. No removal of three large-seeded species was detected in 1998. Interspecific differences in seed removal rates were consistent with the hypothesis that larger seeds are removed less in pastures than smaller seeds. While some seed removal patterns appear predictable, species-specific variation in removal within and among sites was substantial. Predator population densities and other among-pasture differences may also influence patterns of spatial variation in seed predation.