Experiments were performed to investigate monthly variations in seed fate of Carapa procera (Meliaceae), a rodent-dispersed subcanopy tree species in French Guiana. A total of 600 thread-marked seeds were placed on the ground in mature forest under 20 adult trees during different months (March, April and May 1991) of the species' fruiting season. In 1991 the seed crop reached 1536 seeds in the study area, with a majority of seeds being produced in March and May. On average, seed removal rate steadily increased from March (23%) to May (96%). Of the seeds removed, the proportion eached almost doubled between March (28%)-April (25%) and May (48%), whereas the proportion eaten (gnawed) steadily declined between March (43%) and May (9%). Approximately 75–100% seed removal was therefore associated with intense scatterhoarding and low seed predation, especially in May. Greater seed dispersal rates occurred during the late wet season when fruit diversity decreased but when overall fruit biomass peaked because of fruiting occurrence of large-seeded species. The seasonal hoarding behaviour of rodents is discussed.

Maintenance of high plant diversity in tropical rain forests has been hypothesized to be related to intense seed predation, and studies have shown that the impact of seed-eaters varies according to seed size, forest type and level of disturbance. The hypothesis that seed survival and seedling recruitment depending on seed size, are related to forest type and the intensity of harvesting and stand refinement was therefore tested at Paracou, French Guiana, for three animal-dispersed tree species. Two forest areas, corresponding to two forest types, were selected based on the occurrence of a dominant species, Eperua falcata (Caesalpiniaceae). Two factors were tested: forest type (E. falcata-rich and E. falcata-poor) and level of harvesting (unlogged, low intensity of logging, low intensity of logging and thinning liberation by tree-poison girdling) within each forest type. Overall, there was no effect of species nor of logging treatment within forest type after 6 wk. There was only a significant effect of forest type on seed survival in the large-seeded Carapa procera, a rodent-dispersed species in the Meliaceae. In the two medium-sized seed species, Pradosia cochlearia and Manilkara bidentata, two frugivore-dispersed Sapotaceae, the effect of treatment was overall significant but was also highly variable between plots within treatments. After 1 y, C. procera showed little seedling recruitment due to low seed removal and dispersal, and high seed-seedling predation under all conditions. Neither forest type, nor treatment nested within forest type, affected seedling recruitment in P. cochlearia whereas M. bidentata seedlings were favoured in unlogged conditions at the E. falcata-poor forest type. These results suggest that species characteristics (such as seed size and palatability), or local tree diversity and resources for granivores select for contrasting patterns in tree recruitment between forest type in the rodent-dispersed tree species, and these interact with level of tree harvesting and stand refinement in the other frugivore-dispersed tree species.