The interaction between A. standleyanum and the Central-American agouti has become a classic example of a plant-rodent interaction in which animal and plant both benefit (Leigh 1998, Leigh & Rubinoff 2005, Moore 2001, Vander Wall 1990), particularly because the defleshing of palm seeds before caching was assumed to intercept an important insect seed predator, to the advantage of the agouti as well as the palm (Smythe 1989). Our study, however, shows that agoutis do not in fact intercept bruchid larvae. We found that the bruchid beetle infesting A. standleyanum seeds - now identified as Pachymerus bactris - attacks seeds in the canopy, before they leave the mother tree, and that the larvae have already entered the seed when the fruits fall to the ground. Pachymerus bactris does not infest seeds after seed fall. Monkeys and squirrels feeding on unripe fruit might still intercept beetle larvae, but fruit defleshing by agoutis after fruit fall no longer affects infestation. Also, prolonged exposure of palm seeds on the ground below the palm does not increase the fraction infested by bruchids. Thus, the original interception hypothesis is rejected.
The alternative hypothesis that seed defleshing by agoutis reduces predation of cached seeds by other insects and microbes (conservation hypothesis) also was not supported. Post-dispersal seed predation by invertebrates (i.e. scolytid beetles) was important. However, the infestation of seeds by scolytids was reduced by burial, as in Silva & Tabarelli (2001), but not by defleshing. Although we cannot rule out the possibility that agoutis reduce cues for seed finding by scolytid beetles further than we did, for example because they scatter and deflesh seeds better, we consider it unlikely that a strong effect was masked by any such bias. Fungal infestation was unimportant during the 4 mo of this experiment. This seems to be due to the fact that A. standleyanum seeds germinate slowly and have very thick endocarps and thus seem well-protected against fungi, at least until the germination pores open. In general, rodents create their long-term food reserves with well-protected, slow-germinating seeds with a long storage life (Jansen et al. 2006), for which fungal infestation may be less important.
We did find clear support for the hypothesis that defleshing by agoutis reduces cache robbery by mammalian food competitors (theft-prevention hypothesis). Experimentally cached seeds were pilfered at much lower rates when defleshed before caching. Our study site lacks white-lipped peccaries, which are known to forage for buried palm seeds (Kiltie 1981), but there is no reason to believe that the observed difference in cache survival was due to their absence. Previous studies suggest that olfaction plays an important role in cache pilferage (Stapanian & Smith 1978, Vander Wall et al. 2003). Thus, we believe that seed defleshing reduces the olfactory cue for food competitors finding and stealing the buried food.
Bruchid versus scolytid beetles
An additional finding from our study is that bruchid beetles were neither an important enemy for A. standleyanum nor an important food competitor for agoutis. Infestation percentages for seeds incubated in the laboratory were important, but bruchids appeared unimportant under field conditions. This was due to low survival of bruchid larvae under field conditions, not to the defleshing of the seeds by agoutis. We found that the major invertebrate seed predators of A. standleyanum are Coccotrypes scolytid beetles, not bruchid beetles. On BCI, up to 78% of the A. standleyanum seeds in the soil show signs of infestation by scolytid beetles (H. Sharrott & P. A. Jansen, unpubl. data). We found that bruchid presence in seeds was negatively correlated with the number of scolytid entrance holes, suggesting that scolytid beetles are responsible for the low survival of bruchid beetles; they either eat or outcompete bruchid larvae in palm seeds.
Bruchids also seem relatively unimportant to A. standleyanum and agoutis because seeds rarely contain more than one bruchid larva, which consumes only a limited portion of the endosperm, after which the seed can still germinate and establish (C. X. Garzon-Lopez & P. A. Jansen, unpubl. data). This contrasts with the sympatric palm Attalea butyracea, where a single larva will typically consume the entire seed (Gálvez & Jansen 2007). Additionally, recent studies in A. butyracea suggest that agoutis may consider bruchid larvae better food than palm endosperm (Gálvez & Jansen 2007, Silvius 2002). The fact that bruchid larvae spend a long time inside the endocarps - live larvae were present in endocarps after 4 and even 11 mo - implies that bruchid larvae are available to agoutis during August-March, when fruit abundance is relatively low and the rodents feed on cached A. standleyanum seeds. Thus, the selective pressure to prevent bruchid infestation might be negative for agoutis (Silvius 2002), and limited for A. standleyanum.
Smythe (1989) based his conclusion that agoutis intercept bruchid larvae on a comparison of results from three experiments. In one experiment 100 fruits were collected the second morning after they had fallen and stored on trays. Seven days later, exocarp and mesocarp were removed and the defleshed seeds were incubated in jars. After 4 mo, 34 bruchid beetles had emerged. In contrast, in two other experiments, no bruchid exit holes were found in 125 seeds that had been defleshed immediately and left buried in germination bags in the field for several months. Smythe concluded that the larvae entered the endocarps during the 8 d following seed fall unless timely defleshing removed them. Our results suggest that differences in bruchid emergence in Smythe (1989) were due not to the defleshing treatment but rather to the different environments in which the seeds were placed; with and without protection of bruchid larvae from scolytid beetles.
Bruchids as pre-dispersal seed predators
Johnson & Romero (2004) distinguish three oviposition guilds in bruchid beetles: guild A includes species that lay eggs on intact fruits when still on the plant, guild B includes species that lay eggs directly on mature seeds when still on the plant, and guild C includes species that lay eggs on fruits or seeds that have fallen to the ground. All palm bruchids (subfamily Pachymerinae) were believed to be in guild C, infesting endocarps only after fruit have fallen to the ground (Johnson & Romero 2004). Removal of the exocarp and mesocarp facilitates subsequent oviposition and enhances seed predation by guild C species, for example in Speciomerus giganteus attacking Attalea butyracea in Costa Rica and Panama (Janzen 1971b, Wright 1983), in P. cardo attacking A. maripa in Brazil (Silvius & Fragoso 2002), and in Caryoborus serripes attacking Astrocaryum chambira in Peru (Delgado et al. 1997). Exocarps protect seeds against guild C species, and fruit handling by mammals gives bruchid beetles access to the endocarps (Silvius 2005).
Our study is the first to identify a palm bruchid belonging to guild A, infesting endocarps while still on the parent plant, hence acting as a pre-dispersal seed predator (cf. Johnson & Romero 2004). This implies that P. bactris does not require fruit handling by mammals to oviposit successfully. Forget (1991) reported that an unidentified Pachymerus infested A. paramaca while still in French Guiana, and unidentified bruchids from guild A have also been observed in Attalea maripa in Brazil (J. Fragoso, pers. comm.). Pre-dispersal seed infestation by bruchids is more important than is currently appreciated.
Implications
Seed defleshing has become standard preparation for propagating A. standleyanum in nurseries (Potvin et al. 2003) because it was believed to prevent seed infestation by bruchids. Our study suggests that this tedious process will in fact not reduce seed predation by microbes or invertebrates. The only demonstrated effect of defleshing - reduced loss of buried seeds to mammals - is irrelevant in protected nurseries. Whether seed defleshing enhances survival and germination ratios in other ways remains to be shown.
Mammal defaunation could interact with oviposition behaviour of beetles (guild) to have important implications for seed fate. The decline of frugivorous mammals in forest systems reduces access to endocarps for guild C bruchids, which cannot penetrate the exocarp, but not guild A bruchids. Defaunation may therefore differentially affect palm species depending on the oviposition guilds of the associated bruchids (Wright & Duber 2001, Wright et al. 2000).
ACKNOWLEDGEMENTS
We thank Pieter van Eijk, Dumas Galvez, Carol Garzon-Lopez, Maaike Renard and Nina Verkerk for help in the field, the Smithsonian Tropical Research Institute for facilities, Jan den Ouden and Pierre-Michel Forget, Tad Theimer and two anonymous reviewers for comments, and the Netherlands Foundation for the Advancement of Tropical Research (Grant W84-584) and the Eliasen-Uytenboogaart Foundation for funding.