Scatter-hoarding rodents play a crucial role in seed survival and seed dispersal. As one of the most important seed traits, seed size and its effect on rodent-seed interaction attract lots of attention. Current studies usually target one or a few species and show inconsistent patterns; however, few experiments include a large number of species although many plant species usually coexist in natural forest and overlap in fruiting time. Here, we tracked the dispersal and predation of 26 100 seeds belonging to 41 tree species in a subtropical forest for 2 years. Most species showed no relationships between seed size and rodent foraging preference, while the remaining species displayed diverse of patterns: monotonic decrease and increase trends, and hump-shaped and U-shaped patterns, indicating that a one-off study with a few species might give misleading information. However, the seed size effect across species was consistent in both years, indicating that including a large number of species that hold a sufficient range of seed size may avoid the aforementioned bias. Interestingly, seed size effect differed among rodent foraging processes: a negative effect on seed harvest, a hump-shaped effect on seed removal and removal distance, while a positive effect on overwinter survival of cached seeds, indicating that rodents may make trade-offs between large and small seeds both among foraging processes and within a single process, thus lead to a parabolic relationship between seed size and seed dispersal success, that is medium-sized seeds were more likely to be removed and cached, and transported with a further distance.