Tropical Forest Community Ecology

312 Jo hn Terborg hand Kennet hFeeley

Medium islands ranged in area from 4 to
25 ha. In addition to the species present on small
islands, M islands supported a scatterhoarding
rodent (agouti,Dasyproctaspp.), one or some-
times two predators of social insects (nine-banded
armadillo,Dasypus novemcinctus – all medium
islands; tamandua,Tamandua tetradactylus– one
island), and an additional generalist herbivore,
the tortoise (Geochelone carbonaria). A mesopreda-
tor(capuchinmonkey,Cebus olivaceus)waspresent
on three M islands.
Large landmasses served as controls. So that
we could benefit from multiple large landmass
sites while limiting pseudoreplication, we used
two large islands (88 and 190 ha) and two sites
2 km apart on the nearby mainland. The large
landmasses supported complete or nearly com-
plete faunas, including all primates known to
occur in the region, mesopredators (opossum,
Didelphis marsupialis;tayra,Eyra barbara; coati,
Nasua nasua), predators of vertebrates (felids,
raptors, snakes), and large terrestrial mammals
such as deer (Mazama americana), collared pec-
cary (Tayassu tajacu), and tapir (Tapirus terrestris).
Jaguar, puma, and harpy eagle were all confirmed
for the mainland and the 190 ha island, but were
not observed on the 88 ha island or any smaller
island.

Vegetation dynamics

We investigated the prediction that the vegeta-
tion of predator-free S and M islands would suffer
herbivore-induced damage by censusing seedlings
andmonitoringwoodystemsof severalsizeclasses
on islands and the mainland. Trees≥10 cm diam-
eter at breast height (dbh) were monitored for sur-
vival, recruitment, mortality, and growth in plots
on 24 islands and at two mainland sites. Within
14 of these plots (located at 5 S, 4 M, 5 L sites)
we subsampled small (≥1 m tall,<1 cm dbh) and
large (≥1 cm dbh and<10 cm dbh) saplings in
15 m×15 m (225 m^2 ) subplots, two in each tree
plot.Inaddition,wecensusedseedlings(<1mtall)
in a total of 136, 2 m×2 m plots located on 17 S
islands. Tree plots were recensused after either 1
or 5 years, and sapling plots after 5 years (seedling
plots were not recensused). Overall, we tagged,

mapped, measured, and identified 5609 adult

trees, 7027 saplings, and 4086 seedlings repre-

senting more than 320 woody species (exclusive

of lianas).

RESULTS

Hyperdensity of persistent populations

and functional imbalances

Faunal surveys revealed that most animal

populations persisting on S and M islands

displayed increased density (hyperabundance)

compared with mainland/large landmass pop-

ulations. Hyperabundance was documented for

some species of birds (Terborghet al. 1997b,

Feeley and Terborgh 2006), various mammals,

reptiles, amphibians, and spiders (Terborghet al.

1997a), leaf-cutter ants (Rao 2000), rodents

(Lambertet al. 2003), tortoises (Aponteet al.

2003), and howler monkeys (Terborgh et al.

2001; Table 18.1). In addition, some islands pre-

sumably supported high numbers of arthropods

(other than leaf-cutter ants) judging from the

hyperabundance of their predators (lizards, anu-

rans, birds, and spiders; Terborghet al. 1997a,

Feeley and Terborgh 2006). Common iguanas

were present on S and M islands at roughly

10 times mainland density; howler monkeys and

rodents at 20–30 times, and leaf-cutter ants at up

to 100 times. Importantly, species loss/persistence

was strongly non-random with regard to guild

membershi psuch that some functional grou ps

were over-represented (predators of invertebrates,

rodents), whereas others were under-represented

(pollinators, seed predators) or altogether absent

(predators of vertebrates; Terborghet al. 1997a,

Feeleyet al.2007).

Causes of hyperabundance

It is tempting to attribute the hyperabundance

of birds, lizards, amphibians, rodents, and her-

bivores to the absence of predators. Although

this causal link can only be inferred, the circum-

stances are highly suggestive. We found primary