Ecosystem Decay in Closed Forest Fragments 309Further levels of complexity and ambiguity are
inherent in the common use of “open” fragments
(here defined as habitat patches in a land-use
matrix) in the vast majority of fragmentation
studies. The character of the matrix surround-
ing open patches is rarely homogeneous in terms
of vegetation and distances among patches, and
in many cases will be perceived in different ways
by the various species inhabiting the landscape.
For some species, the matrix may be freely porous,
whereas for others, it may serve as an imperme-
able barrier (Malcolm 1994, Leveyet al. 2005).
Since the influence of matrix character has sel-
dom been considered in studies of fragmentation
(but see Gasconet al. 1999), the use of open frag-
ments entails serious ambiguities, as it is rarely
known which species are using different patches
or moving amongst them.
The clearest interpretation of results can be
obtained from “closed” fragments that are severely
isolated from the surrounding matrix. Coloniza-
tion by plants and animals of islands as remote
as the Hawaiian archipelago demonstrates that
no fragment on earth is absolutely closed, so the
permeability of a matrix is a relative concept (Rose
and Polis 2000). Nevertheless, water constitutes
a less heterogeneous and permeable barrier for
non-volant terrestrial species than a mainland
habitat matrix (Gasconet al. 2000).
Since closed fragments can be thoroughly
inventoried (Terborghet al. 1997a,b), one can
accurately determine which species are present
and which are not. Knowing which species are
absent can be as important as knowing which
species are present. For example, many preda-
tors are wide-ranging and may routinely tra-
verse unsuitable habitat to hunt in high quality
patches. Because predation is difficult to detect,
it is likely to go unnoticed in large complex
habitats. With closed fragments, one has better
knowledge of which predators are present, and
consequently a greater ability to determine their
influence on community dynamics.
Another advantage of closed fragments is that
density-dependent emigration rarely influences
the dynamics of resident animal populations.
When possible, many species will emigrate in
response to high densities. Animal populations
of open fragments thus rarely exhibit increased
density, here termed “hyperabundance,” a com-
mon feature of animal populations on islands
or closed fragments (Crowell 1962, Grant 1965,
Krebset al.1969, MacArthuret al. 1972, Morse
1977, Caseet al. 1979, Emlen 1979, Wright
1980, George 1987, Blondelet al. 1988, Polis and
Hurd 1995, Adler 1996).
Our purpose in this chapter is to synthesize the
results of a 10-year investigation of closed forest
fragments isolated as islands in Lago Guri. Lago
Guri is a vast hydroelectric impoundment created
in 1986 with the completion of the Raul Leoní
dam on the lower Caroní River in the state of
Bolívar, Venezuela. Lago Guri contains hundreds
of land-bridge islands, all formerly interconnected
as parts of a continuous forested landscape. The
islands range in size from less than 0.1 ha to more
than 1000 ha (Morales and Gorzula 1986), con-
stituting a giant experiment on the role of spatial
scale in ecology (Diamond 2001).
A crucial feature of the islands we studied is
that they were essentially devoid of predators
of vertebrates during the period of our research
(1993–2003; Terborghet al. 1997a, 2001). In
1960, Hairston, Smith, and Slobodkin (HSS) pro-
posed a simple trophic cascade model which could
be used to predict the consequences of predator
removal. According to the HSS model, predators
regulate consumers to low population densities,
thereby allowing plants to escape damaging levels
of herbivory. Therefore, predator loss or removal
should allow herbivores to increase to levels at
which damage to vegetation would be manifest.
The HSS model was subsequently generalized by
Oksanenet al. (1981) who proposed that a per-
turbation at one trophic level would propagate
downward through a food web with alternating
positive and negative effects at lower levels (Paine
1980, Carpenter and Kitchell 1993, Terborgh
2005). Based on these theoretical models, our
expectation for predator-free Guri islands was to
find higher-than-normal densities of consumers
and evidence of damage to vegetation.
Both these expectations were abundantly
affirmed. However, at least initially, we did not
appreciate the complexity of the trophic cascade
on predator-free Guri islands. The combination
of restricted area, isolation, and the absence of
predators of vertebrates (and some invertebrates)