Volume 90, Issue 6 p. 1024-1032
Free Access

Pulp handling by vertebrate seed dispersers increases palm seed predation by bruchid beetles in the northern Amazon

Kirsten M. Silvius

Corresponding Author

Kirsten M. Silvius

Department of Zoology, University of Florida, PO Box 118525, Gainesville, FL 32611–8525, USA, and

Present address: Faculty of Environmental and Forest Biology, SUNY-ESF, 6 Illick Hall, 1 Forestry Drive, Syracuse, NY 13210, USA.

Kirsten M. Silvius (e-mail [email protected]).Search for more papers by this author
José M. V. Fragoso

José M. V. Fragoso

Department of Zoology, University of Florida, PO Box 118525, Gainesville, FL 32611–8525, USA, and

Museu Paraense Emilio Goeldi, Belém, Pará, Brazil

Present address: Faculty of Environmental and Forest Biology, SUNY-ESF, 6 Illick Hall, 1 Forestry Drive, Syracuse, NY 13210, USA.

Search for more papers by this author
First published: 11 December 2002
Citations: 51


  • 1

    The simultaneous use of fruits and seeds by invertebrate seed predators and vertebrate seed dispersers produces complex ecological interactions that reduce the predictability of seed fate.

  • 2

    Cocosoid palm seeds in the Neotropics are subject to high mortality by bruchid beetle infestation and such attack is the major cause of mortality for seeds of the palm Attalea maripa at our study site in the northern Brazilian Amazon.

  • 3

    The exocarp and mesocarp of 1400 fruits were manipulated in different ways to simulate handling by vertebrates. No eggs of the bruchid beetle, Pachymerus cardo, were laid on intact control fruits, while the highest numbers of eggs were received by fruits whose exocarp and mesocarp had been partially removed, as if by primates and rodents (mean of 15.9 and 18.9 eggs fruit−1, respectively, during the peak fruiting season). Fruits with intact mesocarp but no exocarp, and fruits with all mesocarp and exocarp removed, received low numbers of eggs (mean of 4.6 and 6.6 eggs per fruit, respectively, during the peak fruiting season). Thus both exocarp and mesocarp deter oviposition, and removal of these fruit structures increases fruit susceptibility to infestation.

  • 4

    Oviposition rates declined as the fruiting season progressed, but oviposition preferences remained the same. Seed mortality was high for any fruit on which eggs were laid.

  • 5

    Large rodents and primates, which have been considered among the most effective seed dispersers for large-seeded Neotropical trees such as palms, actually increased the susceptibility of seeds to bruchid beetle attack. Removal of (intact) seeds by other dispersers may be necessary to ensure seed survival.

  • 6

    These results indicate that the reliability of seed dispersers cannot be gauged without a complete understanding of variables that affect seed viability.


Studies of frugivory systems to date focus primarily on the interactions between plants and vertebrate fruit-eaters (Estrada & Fleming 1986; Fleming & Estrada 1993). Fewer researchers have focused on the interactions between fruits and invertebrates, including both seed-eaters and pulp-eaters (Janzen 1968; Herrera 1982, 1984; Traveset 1994; García 1998; García et al. 1999; Passos & Oliveira 2002). A few studies (Herrera 1989; Traveset 1992, 1993) have directly examined three levels of interactions, focusing on the simultaneous use of fruits by vertebrates and invertebrates. When vertebrate or invertebrate fruit consumers handle a fruit, they alter the remaining fruit parts, potentially making them more or less available to subsequent consumers. This three–way interaction may therefore affect the outcome of seed dispersal processes and influence plant demography and evolution.

Two kinds of plant–invertebrate–vertebrate interactions have been shown to affect the numbers of seeds that enter into the seed dispersal process. First, invertebrate pulp consumers can degrade pulp and make it unattractive to vertebrate pulp eaters that might otherwise disperse the seeds (Jordano 1987; Borowicz 1988; Bucholz & Levey 1990; Sallabanks & Courtney 1992; Cipollini & Stiles 1993; Traveset et al. 1995). Secondly, invertebrate larvae may increase the protein, lipid and other nutrient content of the fruit and therefore enhance its attractiveness to vertebrate dispersers or pulp eaters (Redford et al. 1984; Piper 1986; Valburg 1992a,b). Both these interactions posit that an initial action by the invertebrate is followed by the action of the vertebrate. This order can be reversed, however, as in cases when vertebrate pulp consumers improve fruit quality for invertebrate fruit predators and thus have the potential to greatly reduce seed survival. In such cases frugivorous seed dispersers may be analogous to disease vectors or pathogens that increase the seed's susceptibility to the final cause of mortality.

Such interactions have been noted to occur in nature, although they have not been studied in detail. Some insect species preferentially lay their eggs into damaged areas of fruit pulp (Sallabanks & Courtney 1992). Janzen (1971) and Wright (1983) noted that bruchid beetles (probably Speciomerus giganteus) do not lay eggs on endocarps of Attalea butyracea in Central America until the husk and most of the pulp has been removed by vertebrates or fungus. The same was true for an Astrocaryum palm in Peru whose pulp was experimentally removed (Delgado et al. 1997). These observations suggest that both husk and pulp act as important deterrents to insect oviposition.

Predation by insects is often the greatest source of seed mortality or fruit abortion, especially in the tropics (Janzen 1982; Herrera 1986, 1989; Fleming 1991; Oyama 1991; Sallabanks & Courtney 1992; Terborgh et al. 1993; Forget et al. 1999). Alteration of fruit quality for invertebrate fruit predators may be a common occurrence in tropical forests, as most large vertebrates include a significant proportion of fruit in their diet (Terborgh 1986), and most plants produce animal-dispersed fleshy fruits (Levey et al. 1993). Primates, ungulates, rodents, carnivores and birds all handle fruits in different ways and remove differing amounts of mesocarp (pulp) and exocarp (husk) from fruits.

For the neotropical palm species Attalea maripa Martius, 97–100% of endocarps handled by rodents, ungulates other than tapirs (Tapirus terrestris Linnaeus) and primates remain beneath the parent tree canopy, where they are subject to 77% mortality due to infestation by bruchids (with additional mortality caused by other factors) (Fragoso 1997). A. maripa seeds dispersed by tapirs several kilometres away from the parent aggregation are more likely to escape infestation than seeds remaining near the parent tree, as evidenced by (i) counts of infested vs. un-infested seeds at parent trees and tapir latrines, and (ii) seedling abundances at parent trees, random sites and tapir latrines. Thus, endocarps defecated by tapirs at latrines located up to 2 km from the nearest palm clump had only 0.7% mortality from bruchid infestation, and six age classes of seedlings and saplings were significantly more abundant at tapir latrines than around parent trees. Fruit ingestion and subsequent endocarp defecation by tapirs therefore appears to be the most effective seed dispersal system for A. maripa at this site, and processes that increase the probability of occurrence of other seed fates will therefore decrease its chances of survival and recruitment. We tested for such effects by experimentally examining the impact of differential fruit handling by seed dispersing rodents, primates and ungulates on (i) oviposition choice by the bruchid beetle (Pachymerus cardo Fåhraeus) that parasitizes the seeds and (ii) subsequent survival rate of seeds that received differing numbers of eggs in response to different husk and pulp characteristics.

Study system

The study took place from June to December 1996 on Maracá Island, an 1100-km2 moist tropical forest reserve located in the northern Brazilian state of Roraima (3°25′ N, 61°40′ W). The site is protected from hunting, receives few human visitors, and therefore supports healthy populations of all vertebrates typical of Guyana Shield forests, including a full complement of large cats, ungulates and primates. Rainfall ranges between 1750 and 2300 mm annually, with a wet season from May to September and a dry season from October through to April. Milliken & Ratter (1998) describe the site in detail.

Neotropical cocosoid palms are parasitized by about 20 species of palm bruchid beetles (Coleoptera: Pachymerini; Nilsson & Johnson 1993; Delobel et al. 1995; Johnson et al. 1995; Silvius 1999). One of the most specific and widespread relationships is that between palms in the genus Attalea (including the genera Maximiliana, Orbygnia and Scheelea, as revised by Henderson 1995) and the beetles Pachymerus cardo and Speciomerus giganteus (Chevrolat). Interactions among these taxa have been documented from Costa Rica to Bolivia (Janzen 1971; Wilson & Janzen 1972; Anderson 1983; Wright 1983, 1990; Nilsson & Johnson 1993; Johnson et al. 1995; Fragoso 1997; Wright et al. 2000; Quiroga-Castro & Roldán 2001; Wright & Huber 2001). Most studies to date have been carried out at Barro Colorado Island, Panama, where rodent pressure on endocarps appears to be higher than at Maracá.

A. maripa (formerly Maximiliana maripa) occurs throughout the Amazon and southern Orinoco basins (Henderson 1995). Individual trees produce one to three bunches of a few hundred to three thousand fruits, which drop intact to the ground when ripe. Many are also knocked down prior to abscission by primates and birds, often with the pulp only partly removed (Fragoso 1994, 1997; Silvius 1999). On the ground, rodents, ungulates and a variety of insects eat the pulp (Fragoso 1997; Silvius 1999). Mature fruits are 5–8 cm long. A dry, yellowish to orange brown husk less than 0.5 mm thick surrounds a centimetre thick layer of dense, slippery yellow to orange pulp, which in turn surrounds a 3–7 mm thick endocarp containing one to three hard, oily seeds.

P. cardo was the only bruchid species detected on A. maripa during this study, although S. giganteus does occur at low frequencies on the study site. Females lay multiple eggs on fallen fruits at night, and approximately 5 days after oviposition first instar larvae hatch and burrow through the endocarp, leaving tiny but visible entry holes (personal observation). Multiple larvae enter the endocarp, but only one survives to maturity per seed. Once inside, larvae feed on the endosperm and embryo, pupate and emerge as adults through exit holes carved in the endocarp by the last instar larva.


Oviposition rates and larval infestation rates were experimentally assessed on five A. maripa‘fruit types’ differing in the amount and pattern of pulp left attached on their endocarp after manipulation: Intact husk fruits abscised naturally or were dropped by primates without being opened. These were easily obtained by collecting intact ripe fruits from the ground or tree. These were the only fruits in the experiment with husk still attached. Intact pulp fruits imitate fruits whose husks were removed by primates, but were either rejected or accidentally dropped before pulp removal. Intact pulp fruits were obtained by peeling the husk of intact fruits, taking care not to gash the pulp. Gashed fruits are similar to those produced by squirrel monkeys (Saimiri sciurus Linnaeus), macaws and several ungulate species that leave strips of bare endocarp surrounded by a thick bed of pulp. Fruits were artificially gashed by tearing away four strips of pulp with a knife.

Rodent fruits are discarded after feeding by agoutis (Dasyprocta leporina Linnaeus) and pacas (Agouti paca Linnaeus), which squeeze most of the juicy pulp but leave a layer of moist fibres over the endocarp. Rodent-chewed fruits were obtained from the forest floor for the experiments, either from natural fruit fall or from fruits set out at an agouti's feeding site. Bare seed fruits, consisting of a seed surrounded by the endocarp, with no pulp or fibre attached, mimic those resulting from fruits of any of the above types that remain to rot under parent trees, or from endocarps defecated by spider monkeys (Ateles belzebuth Geoffrey) and tapirs. Other ungulates occasionally also remove all the pulp and fibre from endocarps. Bare endocarps were produced by scraping away most of the pulp from intact fruits with a knife, then soaking them in water for a few hours or overnight to remove the remaining fibre. Prior to placement in the forest, they were ‘tossed’ in a bag with pulp fruits to ensure they retained the scent of pulp. All fruit types except intact pulp are common on the forest floor, although their relative frequency at any one tree depends on which, if any, vertebrates have been feeding at that tree. Intact pulp fruits are produced primarily by spider monkeys, which tend to use only a few palm trees in their home range (personal observation).

oviposition rates

To test the hypothesis that the amount and pattern of husk and pulp removed by frugivores affects the number of eggs that are laid on the fruit, the five ‘fruit types’ were exposed to oviposition on the forest floor. Five closed-top ‘chicken wire’ exclosures (mesh size 1.5 × 1.5 cm, 5 cm in height and approximately 23 cm on each of four sides), one for each fruit type and each containing 10 fruits, were placed in a star pattern (with one corner touching) at the edge of a fruiting tree's natural fruit shadow (1–4 m from the base of the tree). All fruiting trees used had been dropping fruit for at least several days, and had at least 100 naturally fallen fruits in the fruit pile. Fruits were separated by 1–3 cm from each other within each exclosure and were set directly on the leaf litter.

All fruits were procured or prepared on one day, then kept in a refrigerator to prevent decomposition until the experiment was set up in the afternoon of the following day. Fruits collected from different trees were mixed together and placed haphazardly into the different fruit categories. Because fruit size varies markedly among trees, and may affect oviposition choice by female bruchids, when different sizes had to be used in a single experimental set-up, sizes were apportioned equally among fruit types. In comparisons carried out at the end of the study, there were no significant differences in the size of endocarps assigned to each fruit type in any season (one-factor anova; F4,490 = 0.63, P = 0.64 in June; F4,486 = 2.17, P = 0.072 in July; and F4,366 = 1.77, P = 0.13 in September).

The experiment was repeated three times at different stages of the palm fruiting season: 10 set-ups in June (early peak), 10 in July (late peak) and eight in September (tail-end of season). Because few trees at the same fruiting stage were available at any one time, trees were used as they became available over a 1–2-week period. All trees were within 1.5 km of each other, and had several other A. maripa trees within 50 m.

All fruits were checked and eggs counted twice after each experiment was set up: once on the morning following the set-up, and again 7 days later. For the first five replicates in June, an attempt was made to count eggs after 14 nights of exposure, but was abandoned because hatching larvae and eggs slid off easily from the rotting pulp when handled. Only the eggs on intact pulp fruits could be successfully counted after 14 days. Intact husks, but not the other fruit types, were examined weekly for up to 9 weeks following set-up. At each count, endocarps were picked up, all eggs counted, and the endocarps returned to their original position. Gashed and rodent fruits were palpated carefully to feel for eggs hidden in the pulp; eggs on other fruit types were easily detected by sight alone.

larval infestation rates

The hypothesis that differences in oviposition rates on different fruit types translate to differences in larval infestation rates of seeds was tested by leaving the endocarps from the oviposition experiment in their exclosures in the field for approximately 3 months, the time period that both laboratory experiments and field observations indicated was sufficient for most larvae to reach full size and consume all the endosperm, but not sufficient to allow pupation and emergence.

After collection, all endocarps were measured along the long axis, and the number of visible entry holes counted (regardless of whether they penetrated to the endosperm or not). Entry holes are a measure of the proportion of eggs that survive to the first larval instar stage. Each endocarp was then split open with a machete, and the number and developmental stage of bruchids assessed.

A second experiment was established to determine whether seven nights of exposure to oviposition are sufficient to lead to high levels of seed infestation under the controlled conditions of the experiment, or whether seed infestation is due to oviposition after the first seven nights of exposure. Fifty gashed fruits were placed in each of two chicken wire exclosures and exposed to oviposition for seven nights; one exclosure was subsequently covered with mosquito meshing to exclude bruchid beetles. After 3 months the endocarps were collected and described as above.

data analysis

Oviposition rates

Data on oviposition rates were summarized as the mean number of eggs per fruit for each tree by fruit type combination, yielding 10 fruit type replicates in June and July, and eight in September. Because intact husk fruits had received no eggs, even after seven nights, data for this fruit type were examined separately. The number of eggs per fruit within each of the remaining four fruit types was not normally distributed, and variance was greater in June than in July and September. Data were normalized by log(1 + x) transformation, and analysed with a two-factor anova, with fruit type and season as factors. Data were not nested by tree because a plot of the raw data showed that differences in oviposition rates per fruit type were consistent from tree to tree (Silvius 1999). Post-hoc comparisons were made with Scheffe F-tests at the alpha = 0.01 level to detect significant differences between pairs of seed types within each season. The statistical package JMP ( SAS 1989) was used for all analyses of variance.

Entry holes and larval infestation rates

Kruskal–Wallis tests were used to analyse the entry hole data (transformation did not achieve normality) and infestation rate data (this was proportion rather than count data). Endocarps derived from intact husk fruits were included in both these analysis. Infestation rate data were further compared with chi-square tests. The statistical package Statview (Abacus Concepts 1986) was used for all non-parametric analyses.


oviposition rates

Intact husk fruits did not receive any eggs during the first week of exposure. Subsequent weekly monitoring in June showed that egg-laying on intact husk fruits peaked after 6 weeks of exposure, with a mean of 1.2 eggs per fruit (range 0–8). Egg-laying on intact husk fruits occurred only when the husk had rotted sufficiently to split, or to lift away from the pulp, opening a gap at the fruit's proximal end between husk and endocarp. Eggs were laid on both the husk and the pulp at this stage.

Patterns in fruit preference were the same for the first and second egg counts, although the total number of eggs was lower after one night; therefore only data from the second count are presented here. Gashed and rodent fruits received the highest number of eggs in all seasons. Intact pulp fruits consistently received the lowest number of eggs in all seasons, and bare seeds were intermediate (Fig. 1). The two-factor anova, with fruit type and season as factors and not including intact husk fruits, revealed a significant model effect (F11,100, P < 0.0001). There were significant effects of both fruit type (F3,100 = 39.94, P < 0.0001) and season (F2,100 = 43.74, P < 0.0001), and a marginally significant interaction between season and fruit type (F6,100 = 2.26, P = 0.0437). Post-hoc comparisons (Scheffe F-test) at the P = 0.001 level indicated that in all seasons, gashed and rodent fruits are statistically indistinguishable from each other, as are bare seed and intact pulp fruits. In June and July, both gashed and rodent fruits received significantly more eggs than both bare seed and intact pulp fruits (Fig. 1), but in September the significance is retained only by gashed fruits (Fig. 1).

Details are in the caption following the image

Mean number of eggs after seven nights on all fruits at all trees for each fruit type, with standard error, showing the high oviposition intensity in June, and the low oviposition intensity in July and September. Intact husk fruits received no eggs after seven nights and are not included in this graph.

The total number of eggs and the variation among fruit types in number of eggs decreased markedly during the study, both within and among seasons (Fig. 1). In June significantly more eggs were laid than in either July or September, but the latter two are statistically indistinguishable from each other.

entry holes

In the second experiment, after 3 months there was no significant difference in the number of entry holes on endocarps exposed to beetles for seven nights (15.3 ± 6.3) and those exposed for the full 3 months (16.1 ± 5.7; t= 0.42, P = 0.338), suggesting that either no more eggs were laid after 7 nights, or that only early eggs were successful. Both endocarp types had 100% larva-induced mortality rates.

In the fruit type exclosures, the number of entry holes was significantly different among fruit types in all seasons (Kruskal–Wallis, H = 280.78, 135.64 and 75.33, respectively, in June, July and September, P = 0.0001 in all cases). The trends in entry holes followed those expected from the number of eggs for all fruit types in all seasons, with the exception of intact pulp fruits. This fruit type received the lowest number of eggs in all seasons. However, in June it had more entry holes (6.2 ± 3.3, n = 100 endocarps) than bare seeds (3.6 ± 3.8 holes, n = 100), and in July and September it had the same number of entry holes (4.0 ± 2.5, 2.0 ± 1.5) as rodent fruits (3.8 ± 3.2, 1.9 ± 1.5) and more than bare seeds (1.2 ± 1.5, 1.1 ± 1.2). These discrepancies suggest that once the pulp is degraded after 1 week in the field, beetles continue laying on intact pulp fruits, and this latter round of egg-laying produces larvae that successfully enter the endocarp. Therefore, as for intact husks, the number of eggs counted after seven nights does not correspond to the actual egg-laying intensity eventually experienced by the fruits.

For the three other fruit types, the seven-night egg count should be a good index of the larval cohort that will attempt the first entry to the seed. Excluding the intact husk and intact pulp fruits for the reasons explained above, the ratios of entry hole number : egg number in June, July and September are 0.539, 0.367 and 0.390 for bare seed; 0.826, 0.754 and 0.610 for gashed; and 0.788, 0.692 and 0.504 for rodent. These larval success ratios reflect the laying preferences of beetles; fruit types with the highest egg success ratios in each season received the highest number of eggs. Additionally, the seasonal decrease in egg : entry hole ratio indicates increasing egg mortality from June to September.

Egg mortality from at least one source could be ascertained: unidentified small ants opened bruchid eggs and fed on the contents during the second week after egg-laying. Although mortality rates were not quantified in this study, eggs killed by ants were noted on 11, 13, 13 and 17 occasions on intact pulp, gashed, rodent and bare seed fruits, respectively.

larval infestation rates

An endocarp was considered infested if it contained a live or dead larva or adult beetle, parasites of bruchid beetles, the track of a larva through the endosperm, or larval frass. Multiple-seeded endocarps with at least one larva were classified as infested. Despite the decline in oviposition rates from June to September, all fruit types in all seasons had a nearly 100% larval infestation rate with the exception of intact husk fruits (mean infestation rate 60 ± 32%) in June (Table 1). The difference in survival rate among fruit types was significant in June (Kruskal–Wallis test, H = 16.08, P = 0.0001; based on mean survival rate per seed type per tree), but not in July and September (P = 0.07 and 0.4, respectively). The seed survival rate for intact husk fruits was significantly higher in June than in July and September, and was not different between July and September (χ2 = 67.3, P = 0.0001, d.f. = 8).

Table 1. Seed mortality per fruit type and season, presented as proportion of endocarps with evidence of bruchid infestation. Numbers in parentheses indicate number of seeds in sample. Initial sample size was 100 fruits per type in June and July and 80 in September; note reduction in sample size for intact husk fruits in all seasons and for intact pulp fruits in July and September, due to fruit removal by mammals and scarabid beetles
Fruit type Season/replicate
June July September
Intact husk 60% (95) 97% (94) 94% (59)
Intact pulp 98% (100) 93% (97) 93% (73)
Gashed 95% (100) 99% (100) 97% (78)
Rodent 96% (100) 92% (100) 99% (79)
Bare seed 86% (100) 98% (99) 99% (79)


general patterns

The presence of an exocarp, the amount of pulp remaining on endocarps, and the pattern in which the pulp was arranged all had significant effects on the number of eggs laid on fruits by bruchid beetles. Fruits simulating partial fruit removal by mammals (‘gashed’ or ‘rodent’) received more eggs than those with intact husk, intact pulp, or no pulp at all. Differences in oviposition rates translated into differences in larval entry rates in all seasons. However, they translated into differences in infestation rates (and hence seed survival rates) only in June, when one-third of intact husk fruits remained uninfested. Note, however, that eggs laid after pulp and husk degradation occurred are probably responsible for infestation of intact husk and intact pulp fruits. Had bruchids been excluded from further oviposition after seven nights, infestation rates of intact fruits would probably have been lower due to low oviposition rates or low entry success of first instar larvae.

Although the number of eggs per fruit declined from June to September, at no time did fruits other than intact fruits stop receiving eggs (Fig. 1), and the numbers they received were sufficient to cause nearly 100% infestation rates (Table 1). Fruits produced very early or late in the season, or out of season, all received large numbers of eggs (Silvius, personal observation), in contrast to the situation observed for Speciomerus giganteus laying on Attalea butyraceae fruits on Barro Colorado Island (Wright 1990; Forget et al. 1994), where fruits exposed late in the season or out of season received no eggs. The interaction observed between seasonality and fruit type in numbers of eggs received may be caused by changing pulp and endocarp quality as rainfall and moisture levels decrease during the course of the fruiting season (see below).

implications for survival through dispersal

The fruits with the highest probability of receiving eggs are those that have been handled by vertebrates usually considered to be good dispersers of large-seeded fruits (rodents and primates) and also by procyonids and psittacines that handle fruits similarly to primates. Rodent and gashed fruits always received eggs during their first night on the ground. The longer a fruit remains without eggs after falling to the ground, the more likely it is to survive infestation, because during this window of opportunity it may be protected from oviposition by removal by tapirs, burial by trampling or burial by scarabid beetles (see below). Intact husk and intact pulp fruits are ingested by tapirs in large amounts and carried long distances from the parent tree (Fragoso 1997). Because fruits with intact husks do not receive eggs until several weeks after falling, they are uninfested at the time of ingestion by tapirs. After defecation, endocarps buried in faeces are protected from oviposition, although those that are exposed on the surface of the faecal pile do receive bruchid eggs (personal observation), indicating that passage through the gut and contamination by faeces are not sufficient to deter oviposition. Quiroga-Castro & Roldán (2001) also documented high escape rates from P. cardo infestation for Attalea phalerata endocarps in tapir faeces in Bolivia, and conclude that escape was due to mechanical protection by faecal matter. In their study, seedling establishment at latrines could not occur because latrine sites were seasonally flooded. At Maracá, however, tapirs defecate in upland latrines that never flood, and seedling establishment is high (Fragoso 1997; Fragoso & Huffman 2000).

Agoutis are considered key dispersers for large-seeded neotropical trees (Hallwachs 1986; Smythe 1989; Forget 1990; Forget & Milleron 1991; Emmons & Feer 1997; Peres & Baider 1997; Peres et al. 1997; Forget et al. 1998). However, agoutis on Maracá do not actively scatterhoard endocarps during the A. maripa fruit-fall period (Fragoso 1997). Rather, they strip the pulp and leave most of the endocarps lying by the parent tree, then later feed on the bruchid-infested endocarps (Silvius 2002). To the extent that infested endocarps are easier to open or more nutritious than intact endocarps, it may actually be to the advantage of agoutis not to cache A. maripa endocarps when they are abundant. Seasonality and species specificity in caching intensity by agoutis has been recorded in several studies (Hallwachs 1986, 1994; Forget & Milleron 1991; Forget et al. 1998).

The only evidence we have for effective secondary dispersal by any organism near the parent tree is fruit burial by the scarabid beetle Oxysternon festivum. Three exclosure fruits (see Table 1) buried to depths of 10 cm by O. festivum after they had received from one to three bruchid eggs all developed larvae; four fruits buried with no eggs survived intact (this suggests that burial of endocarps with recently laid eggs does not prevent infestation; the same would be true of endocarps cached by agoutis after oviposition). On two occasions, O. festivum was observed burying intact pulp Attalea fruits shallowly in a network of tunnels under fruit piles, in a manner similar to that of shallow burial of dung balls for adult feeding by other dung beetle species. A beetle captured at a fruit pile and held in the laboratory for a few hours buried five intact pulp fruits to a depth of 5 cm, the maximum depth of soil provided in its container. Throughout the five fruit type experiment described in this paper, fruits were found within the exclosures shallowly buried in the O. festivum fashion. Because Oxysternon is a common species in the study area (F. Vaz de Mello, personal communication), it may be an effective ‘planter’, if not mover, of palm seeds; many of the seedlings that germinate near the parent plant on Maracá (Fragoso 1997) may be the result of burial by beetles rather than by rodents. Given the high mortality from rooting by white-lipped and collared peccaries (Tayassu pecari and Tayassu tajacu) experienced by seedlings near the parent tree (Fragoso 1997), however, removal to a tapir latrine becomes even more advantageous.

how do husk and pulp protect the seed?

Several mechanisms may account for bruchid oviposition preferences. Bruchids might choose to lay their eggs on fruits that have already been handled by frugivores (i.e. rodent and gashed fruits) if this reduces the probability that eggs will be removed from the fruit during subsequent handling by other frugivores. This is unlikely, however, because on Maracá most larvae-containing endocarps remaining near the parent tree are eventually opened by rodents or crushed by white-lipped peccaries (Silvius 2002). Laying on rodent-gnawed or primate-gashed fruits thus does not reduce the likelihood of subsequent handling by vertebrates and consequent death of the larva or adult.

The most striking differences in both oviposition and seed survival rates occurred between intact husk fruits and all other fruit types. Larvae may thus be unable to penetrate the husk due to its fibrous composition or the presence of abrasive macromolecules. Protection by pulp, on the other hand, presents a more complex picture. The general trend in egg laying preference was that endocarps with some pulp were preferred above endocarps with no pulp or with too much pulp. It is possible that eggs do not attach well to or are harmed by some component of the pulp. Alternatively, first instar larvae may be unable to penetrate the pulp. Female bruchids often avoided pulp by laying eggs on pieces of litter stuck onto intact pulp fruits, on indentations in the pulp made by pulp-eating invertebrates, or at the proximal end of the seed, where the pulp layer is thinnest. Eggs on intact pulp frequently fell off when handled or after heavy rainfall. For Old World legume bruchid beetles, lipids applied to seed crops have been extensively documented as an important egg mortality factor (Don-Pedro 1989; Credland 1992; Parr et al. 1998). Lipids in A. maripa (and other palm species) pulp may therefore serve as an oviposition deterrent as well as an attractant/reward for seed-dispersing frugivores.

If pulp deters oviposition, why do endocarps completely devoid of pulp also receive few eggs? On bare seed fruits, females usually laid eggs on the underside of the endocarp. Wright (1990) observed the same behaviour for a different palm bruchid species in Panama. This behaviour suggests that moisture may be a key factor in egg-laying choice: eggs on bare endocarps would be more likely to dry out, because there is no pulp to retain moisture. Fruits gnawed by rodents and gashed by ungulates or monkeys provide a secure attachment site on the bare endocarp or on the juice-free fibre, while providing moisture-retaining pulp or fibre. On gashed fruits, eggs were nearly always laid on the endocarp but close to or under the edge of the pulp.

Alternatively, higher predation rates by ants on eggs laid on bare endocarps may affect oviposition choice by females, and may increase the value of intact pulp fruits, as eggs appear to be less likely to be attacked on intact pulp fruits than on other types. The texture of A. maripa palm pulp may deter visits by small ants, although it does not deter large ants and spiders, some of which visit the fruits to feed on insect larvae developing in the pulp or to ambush ovipositing insects (personal observation). Additionally, bruchid larvae are parasitized by brachonid wasps (Silvius and Fragoso, unpublished data), and oviposition choice may be affected by fruit characteristics that reduce access by wasps. Thus, the complexity of fruit–invertebrate interactions in this system suggests that factors other than the physical ability of the larva to reach the seed may affect bruchid oviposition choice.

multiple functions of pulp

Although it is difficult to determine the evolutionary origins (original function) of a plant structure, a consideration of alternative current functions may broaden our understanding of dispersal ecology. Current ecological theory considers fleshy mesocarp and other soft layers around the seed, such as sarcotesta and arils, primarily as adaptations to encourage seed removal and/or ingestion by frugivores, while dry husks and hard or spiny endocarps are considered adaptations to prevent ingestion, digestion or parasitism of seeds (Janzen 1968; van der Pijl 1982; Tiffney 1984; Estrada & Fleming 1986; Wilson & Whelan 1990; Fleming & Estrada 1993; Rodgerson 1998). In this study we present some evidence that pulp as well as husk protects seeds against insect predation, an important source of seed mortality in the tropics (Sallabanks & Courtney 1992). If we view a fruit as a series of layered structures that present different kinds of barriers to seed access by seed consumers, then mesocarp could be interpreted as one more protective layer, along with the epicarp and endocarp.

Given that fleshy fruits evolved independently multiple times (Tiffney 1986; Dilcher 2000; Mack 2000), it is possible that in some plant lineages pulp or other fleshy structures may initially have played a protective role by deterring seed-eating insects (Stebbins 1970; Mack 2000). Alternatively, selection pressure for insect deterrence may act in parallel with selection pressure for traits that attract dispersers, producing structures with multiple functions. Pulp may thus act as a medium in which to place secondary compounds, which have been shown to play a deterrent role against seed-eating insects (Herrera 1982; Cipollini & Levey 1997). It may also act as a simple physical barrier impeding the passage of larvae or preventing the attachment of eggs, as in this and other studies on palms (Delgado et al. 1997). As more alternative hypotheses for the evolution of fruit traits are examined, more instances of multiple or alternative selection pathways for fruit and seed traits may become apparent.


Funding for this project was provided by NYZS/The Wildlife Conservation Society and the Lincoln Park Zoo's Scott Neotropical Fund. Field assistance was provided by Toby Benshoff, Rosildo dos Santos and Fabio Bonatto. John Kingsolver identified all beetles in this project. The staff of Maracá Island Ecological Reserve and the World Doctors Yanomami Mission provided essential logistic support.