Ecology, 86(2), 2005, pp.

379–390
q 2005 by the Ecological Society of America

MAMMALS, EDGE EFFECTS, AND THE LOSS OF TROPICAL

FOREST DIVERSITY
NIGEL M. ASQUITH1,2,3 AND MÓNICA MEJı́A-CHANG2
1Department of Zoology, Duke University, Box 90325, Durham, North Carolina 27708-0325 USA
2Smithsonian Tropical Research Institute, Apartado 2072, Balboa, Panama

Abstract. Relative to the surrounding mainland forests, a subset of tree species dom-
inates wind-exposed, forested islands in Gatún Lake, Panama. We explored how tree di-
versity in these fragments has been affected by (1) impoverishment of the mammal com-
munity and (2) changes in abiotic conditions following island formation ca. 90 years ago.
To test effects of changes in the mammal community, we assessed seed and seedling survival
for nine tree species in five forests: small islands with no mammals; small islands with
spiny rats but no larger mammals; medium islands (intermediate mammal community);
Barro Colorado Island (intermittently present puma and jaguar); and mainland forests (intact
mammal community). To test effects of abiotic stress, we chose experimental sites at wind-
exposed, wind-protected, and interior forest sites. We predicted that fragments with less
diverse mammal communities would be characterized by (1) fewer seeds dispersed and
cached, (2) lower long-term seed survival, and (3) higher rates of seedling herbivory by
mammals. Where alteration of the environment has caused greater exposure to dry-season
winds, we predicted that (4) germination and seedling establishment and (5) dry-season
seedling survival would be low. Further, we expected that (6) dry-season seedling survival
would increase if soil moisture levels were raised, but that (7) wet-season seedling survival
is independent of wind exposure.
In the larger forests, seed and seedling survival were low, and differences in mammal
community composition had little effect. Small islands that supported only rats had the
same low seed survival as larger forests. Elimination of all mammals resulted in much
higher seed survival, and slightly higher seedling survival. Germination and seedling sur-
vival were lower at wind-exposed than at protected sites, but irrigation had no effect on
dry-season seedling survival. It appears that the dominant species on Gatún Lake fragments
have passed through a series of filters: seeds must first avoid being eaten by spiny rats,
survive the first dry season, and then avoid seedling herbivory by rats. Biotic and abiotic
factors determine seedling survival in tropical forests; synergistic changes in these factors
can thus lead to dramatic losses of diversity.
Key words: agouti; Barro Colorado Island, Panama; Dasyprocta spp.; edge effects; irrigation
experiments; island biogeography; Proechimys semispinosus; Protium panamense; seed dispersal;
seedling herbivory; spiny rat; tropical forest diversity.

INTRODUCTION It is well documented that ecosystem fragmentation

lowers species richness (MacArthur and Wilson 1967,
Why are intact tropical rain forests so species rich? Laurance 1997). Although ecologists have often noted
Although many mechanisms have been proposed, in- that particular species disappear following fragmen-
cluding niche differentiation (Ashton 1969), pest pres- tation, there are few mechanistic explanations of why
sure (Janzen 1970), intermediate disturbance (Connell (but see Leigh et al. 1993, Terborgh et al. 1997, 2001).
1978), chance (Hubbell and Foster 1986), and forest Apart from the simple answer that a fragment is too
productivity (Phillips et al. 1994), the answer to the small to support a viable population of a species, we
diversity question still eludes ecologists. In this paper have rarely been able even to guess the mechanisms
we attempt to shed light on the question by addressing underlying fragmentation-induced species loss. This
the complement: why are fragmented tropical forests information gap is especially acute for tropical forests:
so species poor? we do not know which of the processes that maintain
the diversity of intact forests are so altered by frag-
mentation that local extinctions result.
Manuscript received 2 September 2003; revised 27 June 2004; Some of the more obvious changes that follow trop-
accepted 2 July 2004. Corresponding Editor: R. S. Ostfeld. ical forest fragmentation are abiotic. Near fragment
3 Present address: Regional Conservation Strategies
Group, Conservation International, 1919 M. Street NW, edges, the forest becomes warmer, better lit, and less
Washington, D.C., 20036 USA. humid (Fetcher et al. 1985). Edge creation alters the
E-mail: [email protected] microclimate in ways that harm some species, increas-
379
380 NIGEL M. ASQUITH AND MÓNICA MEJÍA-CHANG Ecology, Vol. 86, No. 2

ing mortality rates and decreasing recruitment (Laur- This paper experimentally explores the mechanisms
ance et al. 1998). For example, seedling abundance in of species loss proposed by Leigh et al. (1993). Our
an Australian fragment varied depending on whether general hypothesis is that changes in fragments’ biotic
or not the edge was exposed to prevailing winds (Turton and abiotic environment (specifically, changes in the
and Freiburger 1997). Changes in a fragment’s abiotic abundance and diversity of herbivorous mammals, and
environment thus can directly affect seedling recruit- wind-induced changes in soil moisture) have had pro-
ment, and in the long term can affect tree species di- found effects on seedling recruitment and tree diversity
versity. on these islands. We assessed seed and seedling sur-
Indirect biotic changes also occur in a fragmented vival under a variety of treatments in a range of dif-
forest. Differences in herbivore foraging patterns ferent-sized fragments, and tested seven predictions.
caused by fragmentation-induced changes in mammal We predicted that in fragments with fewer mammal
community composition can lead to changes in seedling species (1) fewer seeds are dispersed and cached, (2)
recruitment (Dirzo and Miranda 1991, Wright and Dub- long-term seed survival is lower, and (3) seedling her-
er 2001). Putz et al. (1990) hypothesized that differ- bivory is higher than in fragments with more mammal
ences in mammal community composition have influ- species. Where a fragment’s abiotic environment has
enced tree diversity on islands in Lago Gatún, Panama. been altered by increased exposure to dry-season
Specifically, a lack of seed-eating mammals on small winds, we predicted that (4) germination and early
islands may have allowed increased recruitment of seedling survival is lower, and (5) dry-season survival
mammal-preferred tree species, resulting in changes in of established seedlings is lower than at more protected
forest composition (Terborgh 1992). Moreover, in in- sites. We also predicted that (6) during the severe dry
tact forests, mammals such as agouti (Dasyprocta spp.) season, seedling survival increases at exposed sites if
often disperse and bury seeds, some of which survive soil moisture levels are raised, and (7) wet-season seed-
to germinate (Smythe 1989). In fragments where mam- ling survival does not differ between exposed and in-
mals are absent, or where mammals have been lost to terior sites. Here, we test these predictions, show how
changes in biotic and abiotic conditions have affected
hunters, such dispersal does not occur. Mammal com-
seedling dynamics, and demonstrate how these changes
munity composition thus may play an important role
have reduced tree species diversity in Panamanian for-
in the maintenance or loss of tree species diversity.
est fragments.
Is there a synergy between abiotic and biotic changes
in fragments that further reduces tree diversity? The MATERIALS AND METHODS
numerous small islands that were created during the
Our study sites were located in and around Gatún
1912 construction of the Panama Canal provide a trac-
Lake, central Panama. Islands were formed in the lake
table study system for addressing this question. Most
during the construction of the Panama Canal, when
of these islands appear to have supported species-rich
rising waters of the Chagres river isolated forested hill-
tropical forest immediately post-isolation, but they rap-
tops into numerous islands ranging in size from ,1 ha
idly lost much of their species diversity. Many of the to the 1500-ha Barro Colorado Island (BCI). Annual
smaller islands were soon dominated by a restricted rainfall on BCI (98109 N, 798519 W) is 2600 mm, with
number of species: Protium panamense, Oenocarpus a four-month dry season beginning in December.
mapora, Attalea butyracea (formerly Scheelea zonen- Strong dry-season winds blow across the lake from the
sis), and Swartzia simplex, with Protium panamense north and northeast.
comprising ;60% of the stems on one island by 1980.
Leigh et al. (1993) hypothesized that three mechanisms Tree and sapling diversity
have contributed to this loss of diversity: exposure of On seven small (,2.2-ha) islands, we measured,
adult trees to damaging windstorms; wind-induced identified, and permanently marked every tree with a
changes in the islands’ abiotic environment; and the diameter at breast height (dbh) .10 cm. We included
alteration of the mammal community relative to intact all six islands on which we would conduct seed and
forest. seedling survival experiments, and the six small islands
Understanding the interplay of abiotic and biotic fac- studied by Leigh et al. 1993 (five islands were common
tors is likely to be critical for ensuring the maintenance to both groups). Because we were interested in natural
of diversity in tropical forests. Although changes in a mortality and recruitment, we chose islands that, as far
fragment’s abiotic environment can profoundly affect as could be determined, have been continuously for-
seedling survival at the forest edge, such edge effects ested since isolation in 1912. Stumps of drowned trees
may be of little long-term importance in larger frag- surround each island, and aerial photographs taken as
ments. In contrast, if changes in mammal community early as 1927 suggest that the islands were still forested
composition drive changes in seedling dynamics, forest following construction of the Panama Canal (Leigh et
diversity even in large fragments may be at risk if al. 1993). Five islands were ‘‘exposed’’ to dry-season
hunting or other disturbances persist (Wright et al. trade winds (i.e., .1 km of open water to the north
2000). and northeast), and two islands were ‘‘protected’’ from
February 2005 MAMMALS, EDGES, AND FOREST DIVERSITY 381

winds by other land masses (i.e., ,50 m of open water puma (Puma concolor) as frequent visitors
to the north and northeast). We also recorded tree spe- (Wright and Duber 2001). On each peninsula, we
cies diversity on an exposed and a protected edge of conducted experiments at an exposed, a protected,
each of two medium (3.5–17 ha) islands, three BCI and an interior site.
peninsulas, and on a mainland peninsula (we analyzed 5) Mainland forests surrounding Gatun Lake (n 5
these sites collectively as ‘‘large fragments’’). At each 3). Whereas BCI and the small and medium is-
large fragment, we measured, identified, and perma- lands have suffered minimal hunting, the other-
nently marked the first 100 trees (dbh . 10 cm) en- wise-intact mammal faunas of mainland penin-
countered within 40 m of the forest edge. sulas were unprotected and suffered severe poach-
At each site, we also randomly located and demar- ing until 1980 (Wright et al. 2000). On each pen-
cated five 5 3 5 m plots and measured, identified, and insula (Bohio, Gigante, and Palenquilla), we
tagged every sapling, shrub, and liana with dbh .1 cm. undertook experiments at an exposed, a protected,
We calculated abundance and species richness for each and an interior site (interior sites on BCI and in
site, and compared the Shannon diversity index for mainland forests were in protected valleys .100
each data set using factorial ANOVA with exposure m from the forest edge).
and forest type (i.e., small islands vs. large fragments)
main effects. Within each forest type, experimental site location
was random, except that each site was under a closed
Experiments forest canopy and had an area of level ground large
enough for two mammal-exclosure cages and paired
To assess how seed and seedling survival are affected
nearby control plots (same ground area, but uncaged).
by differences in mammal community composition, we
Cages were constructed of galvanized wire cloth (1.5
located 33 experimental sites in forests that supported
m high 3 1 3 1 m), were open at the top, and were
five different mammal communities. To assess the ef-
staked to the soil to prevent passage of burrowing mam-
fect of abiotic factors on survival, we located nine of
mals underneath. Although climbing mammals and
these sites in the forest interior, 12 within 10 m of an
large mammals such as deer had access through the
‘‘exposed’’ forest edge, and 12 within 10 m of a ‘‘pro-
open tops, Asquith et al. (1997) showed that disap-
tected’’ forest edge (definitions of exposed and pro-
pearance/mortality of Gustavia superba seeds and
tected as given previously). Specifically, sites were lo-
seedlings from within similar cages was ,10%, vs. 84–
cated as follows.
100% outside. Water, light, wind, and insects could
1) Small (,2.2 ha) islands with no mammals ( n 5 pass through the cages.
3 islands). These islands were exposed to dry- Each cage was located in a similar light environment,
season trade winds (islands 15, 19, and 37 in Adler as determined by the proportion of open canopy above
and Seamon [1991]). each cage measured with a Spherical Crown Densiom-
2) Small (,2.2 ha) islands with only spiny rats, eter (Forestry Suppliers, Jackson, Mississippi, USA).
Proechimys semispinosus (n 5 3 islands). One The proportion of open canopy for cages and controls
island was exposed to dry-season winds, and two at each site were averaged, and the arc sine transformed
islands were relatively protected by other land means for each site were compared by ANOVA with
masses (islands 7, 10, and 43 in Adler and Seamon main effects of forest type and location (edge vs. in-
[1991]). terior). As small islands lacked interior sites, we used
3) Medium-sized islands (3.5–17 ha) (n 5 3) that two separate ANOVA to check each main effect in-
lack deer, tapir, and large predators, but have the dividually. No difference was detected in light levels
expected complement of small mammals (e.g., in the cages between forest types (F3,26 5 0.57, P 5
agouti, spiny rats, and squirrels) (Appendix A). 0.64) or location (F2,27 5 0.27, P 5 0.76).
On each island (islands 3 and 13 in Adler and We verified the presence/absence of spiny rats on
Seamon [1991], and island 54 in Mangan and Ad- each small island through trapping, but did not further
ler [2002]) we undertook experiments at an ex- census other mammals. Not only are accurate abun-
posed site, a protected site, and an interior site. dances of mammals extremely difficult to obtain
The interior sites were as close to the center of (Wright et al. 1994), but also mammal populations vary
the island as possible, although island geometry over orders of magnitude in relatively short time pe-
required that these sites were 40–50 m from the riods (e.g., Proechimys on Gatún Lake islands, Adler
forest edge. This distance is greater than the 40 1996).
m invoked by Kapos (1989) as the zone into which For experiments investigating the effects of mammal
the forest is influenced by the edge (but cf. Laur- community composition, each small island was an in-
ance et al. 1998). dependent sample. Each medium island, BCI, and each
4) Peninsulas on Barro Colorado Island (1600 ha) (n mainland peninsula had an exposed, a protected, and
5 3) that support an intact mammal community an interior site. Although the three sites located on each
(Glanz 1991) with jaguar (Panthera onca) and larger island or peninsula were connected by contig-
382 NIGEL M. ASQUITH AND MÓNICA MEJÍA-CHANG Ecology, Vol. 86, No. 2

uous forest, sites were considered independent because in young forests and on steep ravines. Its fruits are
the mammals of interest (Proechimys, Dasyprocta, orange, 4–4.5 cm, and are held in a massive, pendant
Agouti, and Sciurus) have home ranges much smaller cluster (Croat 1978). Virola nobilis A. C. Smith (Myr-
than the minimum distances between sites (Glanz 1991, isticaceae) is a canopy tree. A fibrous capsule dehisces
Adler 2000). to expose a gray 2 cm diameter seed enclosed by a red
aril 1 mm thick (Howe 1993). Dipteryx panamensis
Abiotic environment (Pitt.) Rec. & Mell (Leguminosae) is a common, large
During the dry season (December–April), we mea- (40–50 m) canopy tree. Its drupes have a thin exocarp
sured wind velocity and soil moisture along two ex- and a hard, thick-walled endocarp that encloses a sin-
posure gradients. Each gradient comprised two small gle, 4 cm diameter seed (DeSteven and Putz 1984).
exposed islands, one site on the exposed edge of a BCI Prioria copaifera Griseb. (Fabaceae) is a common can-
peninsula, one site 40 m in from an exposed edge, one opy tree in seasonally inundated forest, but also occurs
site on a protected edge of a BCI peninsula, one site on BCI on well-drained upland soils. Its single-seeded
40 m in from the edge, and one interior forest site. woody legumes are .8 cm in diameter (Dalling et al.
Modal wind velocity (meters per second) was measured 1997).
at 1 m height using an Omnisensor Air Velocity Meter Seeds of all species are consumed by mammals, and
(Model 1640, TSI, Saint Paul, Minnesota, USA) at each may be dispersed and buried by agouti, Dasyprocta
site for 10 min between 11:30 and 13:30 hours on at punctata (e.g., Forget 1992, 1993). Prior to experi-
least four dry-season days in 1997. Wind velocity could ments, the mesocarp, fruits, and arils were removed
not be measured at all of the sites on the same day, so from P. panamense, S. simplex, G. superba, A. stan-
sites measured each day were chosen randomly. Modal dleyanum, and V. nobilis.
wind velocity values for each 10-min period were com-
pared by ANOVA. We measured soil moisture content Mammalian seed dispersal and predation
by collecting ;300 g of soil from the upper 15 cm of Prediction 1: The ratio of seeds dispersed and
soil at each of these sites. Five samples for each site cached by mammals per seed eaten is lower in frag-
were pooled and weighed, dried for three days at 708F ments with less diverse mammal communities.—We as-
(21.18C), and then reweighed. sessed seed dispersal rates by attaching threads to seeds
of eight species (not the small-seeded P. panamense)
Tree species and following their fate. Eight seeds of each species
We used nine tree species in seed and seedling sur- were placed outside the exclosures on the forest floor
vival experiments. Three of these species are common at 33 sites, within a 2-week period for each species
on small islands in Gatún Lake (Leigh et al. 1993) and (May 1996–June 1997). Each seed was threaded with
also occur in intact forest. Oenocarpus mapora H. a 50 cm length of monofilament thread, to which was
Karst (Arecaceae) (Oenocarpus panamanus Bailey) is attached a 2 cm length of flagging tape to facilitate
a slender 8–20 m monoecious palm, common on BCI. relocation (Asquith et al. 1999).
Its 2.0–2.5 cm fruits are blackish-green (Croat 1978). After 14 days, we searched within a 10 m radius of
Protium panamense Rose (Burseraceae) is usually ,12 the site to relocate threads and tapes. If the thread was
m tall and is abundant on BCI. Its fruits are 2–3 cm relocated with the seed still attached and it had been
drupaceous capsules, that turn red at maturity and con- moved and/or buried, secondary dispersal was indi-
tain a white, fleshy mesocarp surrounding one or two cated. If the thread alone was recovered, we assumed
0.5 cm diameter green seeds (Croat 1978). Swartzia that the seed had been consumed. We tested this as-
simplex (SW.) Spreng. (Caesalpiniaceae) is a 15-m tree sumption by attaching similar threads to A. standley-
that is relatively common on BCI. Its fruits are orange anum seeds, which have an inedible endocarp (Asquith
pods, 2.5–5 cm long, that enclose a white aril and a et al. 1997). Almost 80% (23 of 29) of the threads from
single smooth 2–3 cm diameter black seed. which the seed had been removed were within a few
The other six study species are relatively uncommon centimeters of fragments of an opened endocarp, sug-
on the small Gatún Lake islands studied by Leigh et gesting consumption by granivorous mammals.
al. (1993). Calophyllum longifolium Willd. (Clusi- For analysis, we excluded seeds that were not re-
aceae) is a 35 m tall canopy tree, frequent in the BCI moved, seeds that were not recovered, and seeds whose
forest (and actually very common on a small, unstudied thread was not recovered. We found no evidence that
island). Its fruits are round, green, ;3 cm in diameter, the small islands support scatter-hoarding mammals
with a thin fleshy exocarp surrounding a single large (but see Adler and Kestell 1998), so small islands were
seed (Croat 1978). Gustavia superba (H.B.K.) Berg. excluded from analysis. We also excluded S. simplex
(Lecythidaceae) is a small (10–15 cm dbh) pioneer tree because none of its seeds were dispersed, and D. pan-
common in the secondary forest understory on BCI. amensis because only two of its seeds were dispersed
Its large fruits (150–600 g) contain 5–50 seeds, each in any forest type. Proportions of seeds dispersed at
3–15 g (Sork 1987). Astrocaryum standleyanum Bailey each site were ranked because some groups showed no
(Arecaceae) is a monoecious 15 m tall palm, common variance (Conover and Iman 1981). The rankings were
February 2005 MAMMALS, EDGES, AND FOREST DIVERSITY 383

analyzed by factorial ANOVA with main effects of mination and early seedling survival of C. longifolium,
species, forest type, and exposure (exposed edge, pro- G. superba, and O. mapora were recorded in October/
tected edge, forest interior). November 1996, P. panamense in May 1997, and A.
Prediction 2: Long-term seed survival is lower in standleyanum, D. panamensis, S. simplex, and P. co-
fragments with less diverse mammal communities.— paifera in January 1998. By the end of the experiment
After 2–4 months, we recorded the numbers of seeds (November 1998), too few seeds of V. nobilis had ger-
at each site that were eaten, dispersed, or not moved. minated to allow analysis. Proportions of seedlings of
Proportions of seeds that were eaten were ranked before each species surviving at each site were arcsine-trans-
analysis by ANOVA (Conover and Iman 1981), with formed before analysis. Because small islands lack the
main effects of species, forest type, and exposure. interior exposure level, they were excluded from the
Seeds for which neither seed nor thread was recovered ANOVA, which compared germination in the three
were excluded from the analysis, because we did not larger forest types, analyzing the main effects of spe-
know their fate. Small islands with and without spiny cies, forest type, and exposure.
rats were included in the analysis. However, because Prediction 5: Survival of established seedlings dur-
small islands lack the three wind/drought exposure lev- ing the season of abiotic (moisture) stress is lower at
els, the ANOVA design was not fully factorial. There- exposed than at interior sites.—Three, two, and five
fore, the three-way interaction term, and one two-way seedlings, respectively, of O. mapora, G. superba, and
interaction term (forest type and exposure level) were S. simplex were transplanted from the growing house
excluded from analysis. on BCI into 1-m2 wire cages at nine pairs of exposed
Prediction 3: Seedling herbivory is higher in frag- and interior sites (on medium islands, BCI, and the
ments with less diverse mammal communities.—Seed- mainland) in the late wet season (November 1996).
lings were transplanted from a growing house at BCI These seedlings had established and were independent
to exposed small islands and to exposed and interior of their seeds. The number of seedlings at each site
sites in each of the three larger forest types. Two groups that were alive in January 1997 and that survived the
of three, two, and five seedlings, respectively, of O. dry season was recorded in May 1997. Proportions of
mapora, G. superba, and S. simplex were transplanted surviving seedlings were arcsine-transformed and an-
to each of 22 sites in November 1996, and two groups alyzed by ANOVA with exposure and species main
of four D. panamensis seedlings were transplanted in effects.
the early wet season in June 1997. Prediction 6: Seedling survival during the season of
Each seedling group was planted at least 3 m apart, high moisture stress increases at exposed sites if soil
and was protected from mammals by a wire exclosure moisture levels are raised.—Two 1-m2 wire cages were
cage. The number of established seedlings of each spe- placed at each of seven sites along the two exposure
cies was recorded in July 1997. Cages around one group gradients. Into each cage, we planted 10 O. mapora, 8
of each species were then removed, such that those G. superba, 12 S. simplex, and 7 C. longifolium seed-
seedlings were then exposed to mammalian herbivores. lings in early November 1996. Numbers of established
The total number of surviving seedlings in each treat- seedlings were recorded at the start of the dry season
ment was recorded at the beginning of the dry season (January 1997). Seedlings in one randomly chosen cage
(six months later). Proportions of seedlings surviving at each site were irrigated with 36 L of lake water twice
were ranked and analyzed by ANOVA (Conover and per week throughout the dry season. Lake water has
Iman 1981) with main effects of forest type, treatment, extremely low total nitrogen and phosphate content (R.
and exposure (exposed and interior). There was no dif- Stallard, personal communication), so the irrigation
ference in survival rates among species, so the totals treatment did not add many nutrients to experimental
for all species were pooled to increase sample size. plots. Gravimetric soil moisture in experimental cages
Because the larger forest types (medium islands, BCI, was measured at 0–15 cm depth every 2 weeks during
and mainland peninsulas) had similar levels of seedling the dry season, and compared with ambient soil mois-
survival (84–88% in the cages, 81–90% in the open), ture levels in non-irrigated cages. Soil samples were
these sites were also pooled before analysis. The three- taken for moisture analysis at least 3 days after the
way interaction term and one two-way interaction term most recent watering. Dry season seedling survival was
(forest type and exposure level) were excluded from recorded in April 1997, and proportions of seedlings
the analysis. surviving on islands, at exposed sites, and at protected/
interior sites were arcsine transformed and analyzed by
Seed germination and seedling survival ANOVA with species, exposure, and treatment main
Prediction 4: Germination and early seedling sur- effects.
vival are lower at exposed sites than at protected Prediction 7: Seedling survival during the season of
sites.—Between May 1996 and June 1997, we placed low moisture stress (wet season) does not differ be-
eight seeds of each of nine species in the 1-m2 wire tween exposed and interior sites.—As part of the mam-
cages at each of 33 sites. All seeds of a given species malian herbivory experiment (test of prediction 3),
were placed at all sites within a 2-week period. Ger- seedlings of four species, O. mapora, G. superba, S.
384 NIGEL M. ASQUITH AND MÓNICA MEJÍA-CHANG Ecology, Vol. 86, No. 2

TABLE 1. Fragment tree and sapling diversity at sites in trees on two islands, and O. mapora representing 60%
Gatún Lake as measured by the Shannon Index, H.
of the saplings on one island. However, in comparison
Trees, H Saplings, H compared with six other forests around Gatún Lake,
dominance on small islands is not as spectacular as it
Sites n Mean 1 SE Mean 1 SE
first appeared (Leigh et al. [1993] compared the islands
Exposed islands 5 1.77 0.17 1.88 0.30 with only one mainland peninsula). For example, Gus-
Protected islands 2 2.91 0.13 1.71 0.30
Exposed mainland 6 2.64 0.19 2.69 0.15 tavia superba comprised 52% of the trees at a protected
Protected mainland 6 2.68 0.18 2.69 0.13 peninsula site and Macrocnemum glabrescens com-
prised 39% of the trees at a protected BCI site.
In addition, the P. panamense of Leigh et al. (1993)
actually appears to be two species, P. panamense and
simplex, and D. panamensis, were transplanted to ex-
P. tenuifolium: this marginally increases the number of
posed and interior sites in each of the five forest types.
species in the subset dominating small islands, al-
Seedling survivorship was measured at the start (July
though these two congeners appear ecologically sim-
1997) and at end of the wet season (January 1998), and
ilar. Despite these caveats, exposed Gatún Lake islands
proportions of seedlings that survived were ranked and
exhibit lower species diversity and higher relative dom-
analyzed by ANOVA with treatment, exposure, and
inance of both trees and saplings than sites in larger
forest type main effects. This analysis tested for an
forest fragments. In contrast, protected small islands,
effect of wet-season exposure on seedling survival.
despite exhibiting very low sapling diversity, still
RESULTS maintain high tree species diversity.
Tree and sapling diversity Abiotic environment
Protium panamense dominated the tree and sapling Modal wind velocity differed significantly among
community on three of seven small islands, and at three sites (F5,68 5 17.4, P , 0.001; Appendix C): it was
of six exposed large fragments. P. panamense also lowest at interior sites (0.53 m/s) and 40 m in from
dominated the sapling community at one protected site, protected edges (0.55 m/s), and highest on small islands
whereas Protium tenuifolium dominated two small is- (1.89 m/s). Post hoc comparisons with an adjusted a
lands (Appendix B). Oenocarpus mapora dominated 5 0.025 showed that sites 40 m from exposed edges
the tree and sapling community on one protected small were significantly less windy than exposed edges (F1,68
island. Attalea butyracea dominated the tree commu- 5 6.31, P 5 0.01) and windier than protected edges
nity on one island, was second most dominant tree (F1,68 5 5.63, P 5 0.02). Throughout most of the dry
species on two small islands, and third most dominant season, soil moisture levels at the two exposed sites
species on two islands. Swartzia simplex was second and on the small islands were lower than at the pro-
most dominant species in two island tree communities, tected and interior sites (Fig. 1, controls).
and fourth most dominant on one island.
In species-rich tropical forests, apparent diversity Mammalian seed dispersal and predation
will tend to increase purely as a function of sample The ratios of seeds dispersed and cached per seed
size when sample sizes are small. However, no positive eaten by mammals did not differ among medium is-
relationship was detected between the number of stems lands, BCI, and the mainland (F2, 103 5 0.1, P 5 0.9;
and the number of species in the sapling plots: indeed, Appendix D). Mammal community diversity was ap-
the sites with lowest diversity had some of the largest parently unrelated to patterns of seed dispersal in these
sample sizes. Tree diversity was higher at protected larger fragments (cf. Asquith et al. 1997). There were
than exposed sites (F1,14 5 10.2, P 5 0.006; Table 1). no differences between edges and interior sites (F2, 103
Islands had lower diversity than large fragments ( F1,14 5 1, P 5 0.4), but there were differences among species
5 10.0, P 5 0.007), but the forest type 3 exposure (F5, 103 5 4.7, P , 0.001). Mammals buried at least
effect was also significant (F1,14 5 6.4, P 5 0.02). Ex- 59% of Astrocaryum standleyanum seeds on each of
posure had a significant effect on diversity on small the medium islands, BCI, and the mainland. In contrast,
islands, but not in large fragments. Sapling diversity no S. simplex seeds, and virtually no Dipteryx pana-
was significantly higher in large fragments than on is- mensis seeds, were buried.
lands (F1,14 5 9.0, P 5 0.01), but no significant ex- The proportion of seeds eaten after 2–4 months did
posure (F1,14 5 0.22, P 5 0.65) or interaction (F1,14 5 not differ between edge and interior sites (F2, 195 5 1.5,
0.15, P 5 0.71) effects were detected. P 5 0.23), but there were differences between species
Small, exposed islands in Gatún Lake are dominated (F7, 195 5 9.3, P , 0.001; Table 2). Most species suf-
by a subset of mainland tree species, primarily P. pan- fered high (.60%) predation where mammals were
amense, P. tenuifolium, A. butyracea, O. mapora, and present, whereas ,25% of A. standleyanum seeds were
S. simplex (for dominance on other small islands, see eaten across all sites. Forest types also differed ( F4, 195
Adler [2000]) On some islands, this dominance is pro- 5 12.4, P , 0.001): post hoc Bonferroni comparisons
found, with P. panamense comprising .60% of the showed that small islands with no mammals had sig-
February 2005 MAMMALS, EDGES, AND FOREST DIVERSITY 385

FIG. 1. Dry-season gravimetric soil water content for six different sites in Gatún Lake, Panama. For each site, percentage
soil water content is shown for irrigated (solid circles) and control (open circles) plots. The vertical line represents the start
of the irrigation experiment (9 January).

nificantly lower disappearance rates (,35%) than each thermore, rates of seed destruction after 2–4 months
of the other forest types (each comparison P , 0.01), were virtually indistinguishable between the larger for-
but that no other forest types differed from each other. est types and islands with rats only, even though the
Thus, no differences were found between rates of seed mammal communities in these forests are radically dif-
survival after 2–4 months on medium islands, BCI, and ferent.
the mainland. This suggests that even if components In contrast, on small islands with no resident mam-
of the mammal community are absent or at reduced mals, seed survival was high for all eight tree species.
densities, other seed consumers will compensate. Fur- Seeds that would be consumed quickly by mammalian

TABLE 2. Proportion of unprotected seeds eaten during 2–4 months at sites in Panamá.

Small islands Small islands

Mainland BCI Medium islands with rats without rats
Species Mean 1 SE Mean 1 SE Mean 1 SE Mean 1 SE Mean 1 SE

O. mapora 0.99 0.01 0.91 0.06 0.72 0.14 1.00 0.00 0.33 0.33
A. butyraceae 0.89 0.05 0.82 0.08 0.85 0.12 1.00 0.00 0.33 0.33
G. superba 1.00 0.00 0.92 0.04 0.81 0.10 1.00 0.00 0.33 0.33
C. longifolium 0.68 0.15 0.37 0.13 0.71 0.09 0.90 0.10 0.33 0.33
V. nobilis 0.97 0.02 0.94 0.05 0.69 0.16 0.38 0.31 0.33 0.33
A. standleyanum 0.25 0.10 0.15 0.07 0.23 0.08 0.19 0.19 0.11 0.11
D. panamensis 0.60 0.12 0.42 0.14 0.63 0.16 0.09 0.04 0.09 0.09
P. copaifera 0.83 0.10 0.87 0.07 0.80 0.10 0.67 0.33 0.00 0.00
386 NIGEL M. ASQUITH AND MÓNICA MEJÍA-CHANG Ecology, Vol. 86, No. 2

TABLE 3. Mean number (and 1 SE ) of protected seeds (n 5 8 seeds per species) germinating and surviving at exposed,
protected, and interior sites.

Species code
Site type Om Pp Ss Gs Cl As Dp Pc
Small islands
Exposed 0.8 (1.5) 3.0 (1.4) 2.3 (2.2) 0.0 (0.0) 1.2 (2.5) 1.0 (1.4) 2.5 (1.7) 1.5 (2.5)
Protected 0.0 (0.0) 1.0 (0.0) 3.5 (2.1) 3.0 (4.2) 2.5 (3.5) 1.0 (1.4) 1.0 (1.4) 0.5 (0.7)
Medium islands
Exposed 3.3 (2.9) 1.3 (1.2) 4.3 (1.5) 4.3 (3.1) 2.0 (3.4) 1.3 (1.5) 2.0 (1.7) 1.7 (1.2)
Protected 5.3 (2.1) 2.3 (0.6) 6.0 (1.0) 4.7 (3.2) 2.6 (2.3) 1.3 (0.6) 3.3 (1.2) 4.3 (2.1)
Interior 4.3 (3.2) 1.3 (1.2) 5.0 (2.6) 5.6 (1.5) 0.6 (1.2) 2.0 (2.0) 1.7 (0.6) 3.7 (3.5)
BCI
Exposed 5.0 (3.4) 0.7 (0.6) 4.3 (1.2) 7.3 (1.2) 1.7 (2.9) 0.7 (1.2) 1.3 (1.2) 3.0 (1.0)
Protected 3.7 (2.1) 3.3 (0.6) 4.7 (0.6) 7.3 (1.2) 2.3 (2.5) 1.0 (1.0) 1.3 (1.2) 5.7 (1.2)
Interior 3.0 (3.6) 0.3 (0.6) 5.7 (2.1) 6.3 (2.1) 0.0 (0.0) 2.3 (2.1) 1.0 (1.0) 5.3 (1.2)
Mainland
Exposed 2.3 (4.0) 1.0 (0.0) 4.5 (0.7) 6.0 (0.0) 3.5 (0.7) 1.0 (1.4) 1.5 (2.1) 3.5 (0.7)
Protected 4.0 (4.0) 1.3 (1.5) 6.0 (1.0) 3.7 (4.0) 5.0 (1.0) 1.7 (2.1) 2.0 (2.0) 6.7 (1.2)
Interior 2.3 (3.2) 0.7 (0.6) 3.3 (3.5) 4.0 (2.8) 5.5 (2.1) 2.5 (0.7) 2.3 (1.5) 5.7 (0.6)
Note: Species codes are: Om, Oenocarpus mapora; Pp, Protium panamense; Ss, Swartzia simplex; Gs, Gustavia superba;
Cl, Calophyllum longifolium; As, Astrocaryum standleyanum; Dp, Dipteryx panamensis; Pc, Prioria copaifera.

granivores in an intact forest have a chance to ger- at exposed sites was lower than at protected sites (P
minate on small islands. Although some seeds disap- , 0.01), but that neither exposed nor protected sites
peared on the small islands where we did not record differed from interior sites. Germination was thus low-
resident mammals, we discovered some of these seeds est at exposed sites, highest at protected sites, and in-
decomposing, and invertebrates or mammalian visitors termediate at interior sites. Both exposed and protected
may have consumed the others. small islands exhibited very low germination rates.
Seedling herbivory on D. panamensis, O. mapora, Germination of Prioria copaifera was noticeably lower
G. superba, and S. simplex was minimal on medium on small islands and at exposed sites than at protected
islands, BCI, and the mainland forests, and on small sites, perhaps contributing to the rarity of this species
islands without rats. Indeed, herbivory was so low in on exposed islands. However, most of the other species,
large fragments and on the small islands without rats even those that are common on some small islands,
(and, for S. simplex, on islands with rats) that there also fared worse at exposed sites than at protected sites.
was no overall difference in seedling survival between Although germination of island-dominating species
caged and uncaged seedlings (treatment effect F1, 110 5 was low on small islands in comparison to larger for-
0.06, P 5 0.81; Appendix E). However, on small is- ests, of the three species that fared the best at exposed
lands with rats, 60% of D. panamensis, and 100% of sites, two, S. simplex and O. mapora, are common trees
O. mapora and G. superba seedlings were removed on small islands.
from outside the cages (forest type effect F2, 110 5 4.65, Dry-season survival of seedlings of S. simplex, O.
P 5 0.01). Swartzia simplex was the exception, suf- mapora, and G. superba was lower at exposed (Ss 5
fering virtually no herbivory on small islands with rats. 69%, Om 5 21%, Gs 5 68%) than at interior (Ss 5
The extremely high levels of herbivory may be a func- 91%, Om 5 64%, Gs 5 72%) sites (F1,55 5 4.4, P 5
tion of the abnormally high densities of rats on Gatun 0.04; Appendix F). O. mapora had significantly lower
Lake islands (Adler 1996). Post hoc comparisons con- survival than G. superba and S. simplex (species effect
firmed that fewer seedlings survived on small islands F2,55 5 12.1, P , 0.001, post hoc Bonferroni compar-
with rats than in large fragments (P 5 0.025), but that ison P , 0.01).
small islands without rats did not differ from large Irrigation increased soil moisture content at exposed,
fragments. protected, and interior sites (Fig. 1). Ambient soil mois-
ture ranged between 23% and 31% at the end of the
Seed germination and seedling survival dry season, whereas irrigation increased moisture to
There was no difference in germination rates among 27–36%. For G. superba, O. mapora, and S. simplex,
medium islands, BCI, and the mainland (F2, 136 5 0.08, neither irrigation nor exposure had a significant impact
P 5 0.9) (Table 3). The species effect was significant on seedling survival (Table 4). Survival of nonwatered
(F7, 136 5 13.1, P , 0.001), as was the forest 3 species S. simplex was virtually identical on islands and at
interaction (F14, 136 5 2.21, P 5 0.01. The exposure exposed and protected sites, and survival of non-irri-
effect was also significant (F2, 136 5 3.7, P 5 0.03): post gated Calophyllum longifolium was actually highest on
hoc Bonferroni comparisons showed that germination the small islands. The relatively high survival rates of
February 2005 MAMMALS, EDGES, AND FOREST DIVERSITY 387

TABLE 4. Mean proportion of caged O. mapora, S. simplex, G. superba, and C. longifolium

seedlings that survived the dry season under irrigated and non-irrigated conditions.

Protected/interior
Species Islands Exposed sites sites
and
condition Mean 1 SE n Mean 1 SE n Mean 1 SE n
O. mapora
Wet 1.00 0.00 36 0.89 0.09 36 0.90 0.09 52
Dry 0.73 0.38 35 0.81 0.38 34 0.63 0.11 50
S. simplex
Wet 1.00 0.00 47 0.83 0.07 47 0.98 0.04 59
Dry 0.96 0.04 47 0.96 0.05 49 0.95 0.07 60
G. superba
Wet 0.91 0.12 32 0.83 0.05 29 0.92 0.13 46
Dry 0.84 0.31 30 0.96 0.07 30 0.96 0.07 46
C. longifolium
Wet 0.93 0.10 15 0.64 0.47 15 0.40 0.35 20
Dry 0.29 0.49 14 0.25 0.05 18 0.22 0.23 27

non-irrigated seedlings in this experiment (as opposed is caused by both an island effect and an exposure
to much lower seedling survival in the previous ex- effect. Protected and exposed small islands have lower
periment) may reflect the fact that the irrigation ex- sapling diversity than larger forests, whereas all ex-
periment was established earlier, so the transplanted posed sites have lower tree diversity than protected
seedlings had longer to establish before the unusually sites.
harsh 1997 dry season started. The species 3 treatment A small number of species (Protium panamense,
interaction was significant (F3,80 5 4.6, P 5 0.005). Oenocarpus mapora, Attalea butyracea, and Swartzia
Post hoc contrasts (with an adjusted significance level simplex) may dominate exposed sites primarily because
of P 5 0.0125) within the species 3 treatment inter- they are better able to adapt to the unfavorable micro-
action showed that survival of non-irrigated O. mapora climate. Although germination of O. mapora and S.
did not differ from that of S. simplex and G. superba, simplex was lower on the small islands/exposed sites
but that C. longifolium survival was higher when seed- than at protected sites, these species had higher ger-
lings were irrigated (P , 0.001); survival of C. lon- mination rates than any other species except Gustavia
gifolium seedlings was lower than that of the other superba. Further, reduced dry-season soil moisture, ex-
species in both the irrigated (P , 0.001) and the non- acerbated by high winds on exposed small islands, had
irrigated (P , 0.001) treatments. Although seedlings only a minor effect on established S. simplex, G. su-
of C. longifolium, and to a lesser extent, O. mapora, perba, and O. mapora seedling survival. Species such
suffer from dry-season soil moisture stress, this effect as S. simplex appear to thrive in the dry, exposed, small-
occurs in all forest types, not just on exposed edges island environment.
and small islands. Calophyllum longifolium appears es- Large-seeded species may have an advantage over
pecially susceptible to water stress, so that even when small-seeded species because they have more reserves
soil moisture was increased through irrigation, dry- to survive their first dry season and develop a deep
season seedling survival remained low. root structure. However, this effect is a function of
To assess whether exposure to winds in the wet season species’ intrinsic susceptibility to water stress, and not
reduced the survival of seedlings that were protected just a function of seed size. Drought susceptibility at
from mammals (Appendix G), we reviewed the ANOVA exposed sites persists at the seedling stage, even for
table that tested prediction 3 for significance of the ex- apparently well-adapted species. Even after a two-
posure main effect and the exposure 3 treatment inter- month establishment phase, survival of Calophyllum
action. Neither the exposure effect (F1, 110 5 0.51, P 5 longifolium was drastically reduced at all sites by a
0.48) nor the exposure 3 treatment interaction (F1, 110 5 lack of available soil moisture. Dry-season winds on
0.24, P 5 0.62) was significant. Thus, there was no effect exposed Gatún Lake fragments probably exclude many
of exposure on wet-season survival of any species, sug- small-seeded species, and perhaps the large-seeded C.
gesting that exposure to the dry-season winds is the longifolium and Prioria copaifera.
critical exposure factor for seedling survival. A second important factor that determines which tree
species can survive on small islands is mammal com-
DISCUSSION munity composition. A lack of mammals favors some
A restricted subset of tree species dominates small, tree species over others. More than 85% of O. mapora,
wind-exposed Gatún Lake islands. As hypothesized by S. simplex, G. superba, and Virola nobilis seeds had
Leigh et al. (1993), the low diversity of these islands been destroyed after 2–4 months at mainland sites,
388 NIGEL M. ASQUITH AND MÓNICA MEJÍA-CHANG Ecology, Vol. 86, No. 2

whereas only 33% of seeds had disappeared on islands successful recruitment will probably occur only if rats
free of small mammals. In contrast, a relatively con- are temporarily absent or at low density. If rats are
stant 10–25% of Astrocaryum standleyanum seeds dis- present, seedling herbivory can be as high as 100% for
appeared from all sites (many were dispersed and bur- some species. Although most species on small islands
ied). For most species, if an island supports spiny rats, can only avoid high rat herbivory when rat populations
then seed mortality is high. However, rat populations disappear, species such as S. simplex do not appear to
on Gatún Lake islands suffer local extinctions rela- be susceptible.
tively frequently (Adler and Seamon 1991). If a small There are undoubtedly other important filters, in-
island is temporarily without rats, seeds of fruiting trees cluding invertebrate seed attack (Hammond and Brown
may have a window of opportunity for establishment. 1998) and herbivory (Arnold and Asquith 2002). How-
We found no evidence to support the hypothesis that ever, these three filters are likely to play the most crit-
other differences in mammal community composition, ical role in determining a fragment’s tree species com-
specifically the absence of top predators (Terborgh position. Given the apparent susceptibility of all spe-
1992), can have an impact on seed and seedling dy- cies’ seedlings to water stress, the species that even-
namics and, thus, on forest diversity. Early survival of tually come to dominate a particular small island are
most tree species was unaffected by differences in probably those that had a bumper seed crop and then
mammal community composition among the larger faced a relatively wet dry season in the first year after
fragments (cf. Terborgh 1992, Dirzo and Miranda 1991, the island’s spiny rat population disappeared.
Asquith et al. 1997, Terborgh et al. 2001). What would have happened to tree species diversity
P. panamense had very low germination success at on Gatún Lake islands if exposure had not been such
exposed sites in our experiments, so its dominance on an important factor? On the two small, protected is-
small islands cannot be explained simply by a natural lands, adult tree diversity is still high. This belies an
propensity for survival in dry conditions. These Pro- important pattern in the understory: sapling diversity
tium spp. may be particularly resistant to attack by on protected small islands is significantly lower than
mammals; P. panamense seeds are ,5 mm, and al- at mainland sites, and similar to the low sapling and
though they are eaten by rats (Adler 1995), they are tree diversity on exposed small islands. Thus, although
not preferred. Perhaps P. panamense only dominates the loss will take longer, protected small islands may
some islands because intense seed predation and seed- ultimately lose as much tree species diversity as ex-
ling herbivory by rats excluded other species that are posed islands.
physiologically better able to withstand the small-is- The significant forest type 3 exposure interaction
land microclimate. The two Protium species are also effect in the tree diversity ANOVA indicates that the
dominant whereas A. butyraceae is rare in the wind- effects of island size (a proxy for mammal community
exposed mainland forests with diverse mammal com- composition) and exposure (indicator of a changed abi-
munities. Small-seeded Protium may thus have taken otic environment) have acted synergistically. On ex-
over exposed Gatún Lake islands primarily because posed small islands in Gatún Lake, it thus appears that
spiny rats ate the seeds of A. butyracea and the other the presence or absence of spiny rats and exposure to
large-seeded species that also could dominate. dry-season winds are synergistic filters through which
It appears that the dominant species on small Gatún tree species must pass to persist.
Lake fragments must pass through a series of selective If one of these filters is removed, as occurs on non-
filters or bottlenecks. exposed islands, more tree species will pass through
Filter 1: Avoid seed consumption by rats.—This fil- this bottleneck and diversity will remain higher for
ter will allow passage of many small-seeded species longer periods. However, the strong selective filter of
(Adler 1995), but large-seeded species can also pass 100% seedling herbivory for some species when rats
through, as long as they fruit when island rat popula- are present suggests that even in fragments that have
tions are low or absent. not experienced major changes in their abiotic envi-
Filter 2: Survive the first dry season.—Although the ronment, species recruitment across the community
roots of adult trees on small islands can reach lake will eventually be affected. Diversity will be lost, but
water, the seedlings of many species cannot tolerate the at a slower rate than if the exposure filter were also
physiological stress on islands with soil moisture lower operating simultaneously.
than 25%. The species that can tolerate such water It is clear that biotic and abiotic factors are important
stress tend to be large seeded (e.g., S. simplex). The determinants of seed and seedling survival in central
few large-seeded species that passed propagules Panamanian forests, and that changes in either the bi-
through the first filter are probably at a distinct advan- otic or abiotic environment can profoundly alter forest
tage when it comes to the second. However, if the first composition. Synergistic interactions between changes
dry season is relatively wet, many more species will in mammal community composition and exposure to
survive this filter. winds, as has occurred on small Gatún Lake islands,
Filter 3: Avoid seedling herbivory by rats.—This is can lead to an even more dramatic loss of tree species
again dependent on external factors for many species: diversity.
February 2005 MAMMALS, EDGES, AND FOREST DIVERSITY 389

ACKNOWLEDGMENTS Fetcher, N., S. F. Oberdauer, and B. R. Strain. 1985. Vege-

tation effects on microclimate in lowland tropical forest in
We are extremely grateful to M. Riveros, N. Daguerre, A.
Costa Rica. International Journal of Biometerology 29:
E. Arnold, B. Shapiro, O. Calderón, and A. Hernandez for
145–155.
assistance in the field. The Smithsonian Tropical Research
Forget, P.-M. 1992. Seed removal and seed fate in Gustavia
Institute provided superb logistical support: many thanks to
superba (Lecythidiaceae). Biotropica 24:408–414.
D. Millan and O. Acevedo, and to the Islanders and Gam-
Forget, P.-M. 1993. Post-dispersal predation and scatter-
bodians for life on Mango Street. J. Terborgh and J. Wright
hoarding of Dipteryx panamensis (Papilionaceae) seeds by
directed the research, E. Leigh, A. Herre, P. Coley, and T.
rodents in Panama. Oecologia 94:255–261.
Kursar contributed their expertise in Panama, and J. Clark,
Glanz, W. E. 1991. Mammalian densities at protected versus
C. van Schaik, W. Morris, and P. Klopfer improved the pre-
hunted sites in Central Panama. Pages 163–173 in J. G.
sentation and analysis. Thanks to A. E. Arnold, J. Wright, E.
Robinson and K. H. Redford, editors. Neotropical wildlife
Leigh, J. Terborgh, P.-M. Forget, A. Herre, G. Adler, J. Putz,
use and conservation. University of Chicago Press, Chi-
L. Curran, R. Ostfeld, and an anonymous reviewer for their
cago, Illinois, USA.
comments on the manuscript. Funding was provided by the
Hammond, D. S., and V. K. Brown. 1998. Disturbance, phe-
Smithsonian Institution, the Forest History Society, the Con-
nology and life history characteristcs: factors influencing
servation, Food & Health Foundation, Inc., the Lincoln Park
distance/density dependent attack on tropical seeds and
Zoological Society, and the National Science Foundation
seedlings. Pages 51–78 in D. M Newbery, H. T. T. Prins,
(Grant DEB 9801320 to John Terborgh and N. M. Asquith).
and N. D. Brown, editors. Dynamics of tropical commu-
nities. Blackwell, Cambridge, UK.
LITERATURE CITED
Howe, H. F. 1993. Aspects of variation in a neotropical dis-
Adler, G. H. 1995. Fruit and seed exploitation by Central persal system. Vegetatio 107:149–162.
American spiny rats Proechimys semispinosus. Studies on Hubbell, S. P., and R. B. Foster. 1986. Biology, chance and
Neotropical Fauna and Environment 30:237–244. history and the structure of tropical rain forest tree com-
Adler, G. H. 1996. The island syndrome in isolated popu- munities. Pages 314–329 in J. Diamond and T. J. Case,
lations of a tropical forest rodent. Oecologia 108:694–700. editors. Community ecology. Harper and Row, New York,
Adler, G. H. 2000. Tropical forest diversity, forest structure New York, USA.
and the demography of a frugivorous rodent (Proechimys Janzen, D. H. 1970. Herbivores and the number of tree spe-
semispinosus). Journal of Zoology 250:57–74. cies in tropical forests. American Naturalist 104:501–528.
Adler, G. H., and D. W. Kestell. 1998. Seed predation and Kapos, V. 1989. Effects of isolation on the water status of
scatterhoarding by spiny rats (Proechimys semispinosus). forest patches in the Brazilian Amazon. Journal of Tropical
Biotropica 30:677–681. Ecology 5:173–185.
Adler, G. H., and J. O. Seamon. 1991. Distribution and abun- Laurance, W. F. 1997. Hyper-disturbed parks: edge effects
dance of a tropical rodent, the spiny rat, on islands in Pan- and the ecology of isolated rainforest reserves in tropical
ama. Journal of Tropical Ecology 7:349–360. Australia. Pages 71–83 in W. F. Laurance and R. O Bier-
Arnold, A. E. and N. M. Asquith. 2002. Herbivory in a frag- regaard, Jr., editors. Tropical forest fragments: ecology,
mented tropical forest: patterns from islands at Lago Gatún, management and conservation of fragmented communities.
Panama. Biodiversity and Conservation 11:1663–1680. University of Chicago Press, Chicago, Illinois, USA.
Ashton, P. S. 1969. Speciation among tropical forest trees: Laurance, W. F., L. V. Ferreira, J. M. Rankin-de Merona, and
some deduction in the light of recent research. Biological S. G. Laurance. 1998. Rain forest fragmentation and the
Journal of the Linnaean Society 1:155–196. dynamics of Amazonian tree communities. Ecology 79:
Asquith, N. M., J. Terborgh, A. E. Arnold, and M. Riveros 2032–2040.
C. 1999. Fruits the agouti ate: Hymenaea courbaril seed Leigh, E. G., Jr., S. J. Wright, E. A. Herre, and F. E. Putz.
fate when its disperser is absent from Lago Guri, Venezuela. 1993. The decline of tree diversity on newly isolated trop-
Journal of Tropical Ecology 15:229–235. ical islands: a test of a null hypothesis and some impli-
Asquith, N. M., S. J. Wright, and M. J. Clauss. 1997. Does cations. Evolutionary Ecology 7:76–102.
mammal community composition control recruitment in MacArthur, R. H., and E. O. Wilson. 1967. The theory of
neotropical forests? Evidence from Panama. Ecology 78: island biogeography. Princeton University Press, Princeton,
941–946. New Jersey, USA.
Connell, J. H. 1978. Diversity in tropical rain forests and Mangan, S. A., and G. H. Adler. 2002. Seasonal dispersal of
coral reefs. Science 199:1302–1309. arbiscular mycorrhizal fungi by spiny rats in a neotropical
Conover, W. J., and R. L. Iman. 1981. Rank transformations forest. Oecologia 131:587–597.
as a bridge between parametric and nonparametric statis- Phillips, O. L., P. Hall, A. H. Gentry, S. A. Sawyer, and R.
tics. American Statistician 35:124–129. Vasquez. 1994. Dynamics and species richness of tropical
Croat, T. B. 1978. The flora of Barro Colorado Island. Stan- rain forests. Proceedings of the National Academy of Sci-
ford University Press, Stanford, California, USA. ences (USA) 91:2805–2809.
Dalling, J. W., K. E. Harms, and R. Aizprua. 1997. Seed Putz, F. E., E. G. Leigh, and S. J. Wright. 1990. The arboreal
damage tolerance and seedling resprouting ability of Prior- vegetation on 70-year old islands in the Panama Canal.
ia copaifera in Panama. Journal of Tropical Ecology 13: Garden 14:18–23.
481–490. Smythe, N. 1989. Seed survival in the palm Astrocaryum
DeSteven, D., and F. E. Putz. 1984. Impact of mammals on standleyanum: evidence for dependence upon its seed dis-
early recruitment of a tropical canopy tree Dipteryx pan- persers. Biotropica 21:50–56.
amensis, in Panama. Oikos 43:207–216. Sork, V. L. 1987. Effects of predation and light on seedling
Dirzo, R., and A. Miranda. 1991. Altered patterns of her- establishment in Gustavia superba. Ecology 68:1341–1350.
bivory and diversity in the forest understory: a case study Terborgh, J. 1992. Maintenance of diversity in tropical for-
of the possible consequences of contemporary defaunation. ests. Biotropica 24(2b):283–292.
Pages 273–287 in P. W. Price, T. M. Lewinsohn, G. W. Terborgh, J., L. Lopez, P. Nuñez, M. Rao, G. Shahabuddin,
Fernandes, and W. W. Benson, editors. Plant–animal inter- G. Orihuela, M. Riveros, R. Ascanio, G. H. Adler, T. D.
actions: evolutionary ecology in tropical and temperate re- Lambert, and L. Balbas. 2001. Ecological meltdown in
gions. John Wiley, New York, New York, USA. predator-free forest fragments. Science 294:1923–1926.
390 NIGEL M. ASQUITH AND MÓNICA MEJÍA-CHANG Ecology, Vol. 86, No. 2

Terborgh, J., L. Lopez, J. Tello, D. Yu, and A. R. Bruni. 1997. Wright, S. J., and H. C. Duber. 2001. Poachers and forest
Transitory states in relaxing ecosystems of land bridge is- fragmentation alter seed dispersal, seed survival, and seed-
lands. Pages 256–274 in W. F. Laurance and R. O. Bier- ling recruitment in the palm Attalea butyraceae, with im-
regaard, Jr., editors. Tropical forest remnants: ecology, plications for tropical tree diversity. Biotropica 33:583–
management and conservation of fragmented communities. 595.
University of Chicago Press, Chicago, Illinois, USA. Wright, S. J., M. E. Gompper, and B. DeLeon. 1994. Are
Turton, S. M., and H. J. Freiburger. 1997. Edge and aspect large predators keystone species in Neotropical forests?
effects on the microclimate of a small tropical forest rem- The evidence from Barro Colorado Island. Oikos 71:279–
nant on the Atherton tableland, northeastern Australia. Pag- 294.
es 45–54 in W. F. Laurance and R. O. Bierregaard, Jr., Wright, S. J., H. Zeballos, I. Domı́nguez, M. M. Gallardo,
editors. Tropical forest fragments: ecology, management M. C. Moreno, and R. Ibáñez. 2000. Poachers alter mam-
and conservation of fragmented communities. University mal abundance, seed dispersal, and seed predation in a
of Chicago Press, Chicago, Illinois, USA. neotropical forest. Conservation Biology 14:227–239.

APPENDIX A
A table of the terrestrial mammal community at sites around Gatún Lake, Panama is available in ESA’s Electronic Data
Archive: Ecological Archives E086-020-A1.

APPENDIX B
A list of the dominant tree species on islands at exposed and protected sites around Gatún Lake is available in ESA’s
Electronic Data Archive: Ecological Archives E086-020-A2.

APPENDIX C
A table of modal midday wind velocity at 1 m above ground at Gatún Lake sites in the 1997 dry season is available in
ESA’s Electronic Data Archive: Ecological Archives E086-020-A3.

APPENDIX D
A table showing the proportion of unprotected seeds dispersed after two weeks is available in ESA’s Electronic Data
Archive: Ecological Archives E086-020-A4.

APPENDIX E
A table showing the proportion of protected Oenocarpus mapora, Swartzia simplex, Gustavia superba, and Dipteryx
panamensis seedlings surviving after six months is available in ESA’s Electronic Data Archive: Ecological Archives E086-
020-A5.

APPENDIX F
A table showing the proportion of protected Oenocarpus mapora, Swartzia simplex, and Gustavia superba seedlings that
survived the 1997 dry season at exposed and interior forest sites is available in ESA’s Electronic Data Archive: Ecological
Archives E086-020-A6.

APPENDIX G
A table showing the mean proportion (SE) and original n of mammal-protected seedlings that survived the wet season at
exposed and interior sites is available in ESA’s Electronic Data Archive: Ecological Archives E086-020-A7.