A field experiment was conducted in the Nouragues Reserve, an undisturbed lowland rain forest in French Guiana, to test possible selective pressures towards higher seed value of Carapa procera by scatter-hoarders, and to quantify survival probabilities of seeds harvested and cached by red acouchies (Myoprocta exilis). The following hypotheses were also tested: (1) large seeds are more likely to be harvested by acouchies than small seeds; (2) large seeds are harvested by acouchies more quickly than small seeds; (3) large seeds harvested by acouchies have a higher probability than small seeds to be cached rather than eaten; (4) large seeds are cached further away and in lower densities than small seeds; (5) large seeds are recovered from caches and consumed at lower rates than small seeds; and (6) large seeds have higher survival probabilities than small seeds. All seeds were removed within 1.5 days, except for three seeds, which were removed 1-2 days later, thus hypothesis 1 was not confirmed. A total of 362 seeds were located after dispersal wherein 84% was cached and relatively few seeds (16%) were eaten, which confirms hypothesis 3. The distance at which acouchy caches were found immediately after dispersal was highly variable, both within and among plots. The nearest cache was found <1 m from the plot, from which the farthest was 124 m away. Cache distance increased with seed size, as predicted by hypothesis 4. There was also great variation in spatial isolation of acouchy caches, both within and among plots. Recovery of seeds did not necessarily lead to seed consumption since many seeds were reached or were not found again. Larger seeds had a higher probability of being recached rather than eaten while consumed seeds were found much closer to the cache site than recached seeds. These results support hypothesis 5. Seed mass strongly affected the fate pathways of seeds and how long seeds were kept in stock. The ultimate probability of survival increased with seed mass, as posited by hypothesis 6.