RELATION AND TURNOVER—THE EVIDENCE 273relationship between the percentage of forest loss
on a regional basis and regional species losses. In
short, the use of species–area regressions in this
way appears to be currently without a theoretical
justification, and thus represents merely an
exercise in curve-fitting.
Taken as such, what do we learn from such
studies? The Grelle et al. (2005) study illustrates
that, in practice, there is no reason to favour a z
value of 0.25 over other slope parameters in
making what amounts to a back-of-the-envelope
sketch of how regional habitat loss may relate to
loss of species regionally. In practice, we would
be better to use species-level assessments
(whether underpinned by rigorous survey data or
expert assessments) than to rely on such
species–area fits. Such assessments enableattention to be given to species judged most at
risk, and necessarily involve an assessment of what
factors threaten particular species. This sort of
information would seem to be of inherent value to
conservation agencies. Moreover, the argument
that we should use species–area modelling in the
absence of adequate data for other regions or
other taxa, arguably breaks down with the
realization that the success of the study in fitting
species–area models to independent data on
species threats varied between taxa, with
extinction risks of reptiles and amphibians not
successfully predicted in any of their models. This
is one area within conservation biogeography
where current theory appears inadequate and in
urgent need of attention (as Whittaker et al.
2005).threatened species (Birdlife EBA factsheet 75,
www.birdlife.org/datazone/, visited July2004).
Given these varying forms of human intervention, it
may be impossible to resolve the question of whether
the Atlantic forests system provides support for, or a
refutation of, the theory of species relaxation.
The assumption that fragmentation will necessarily
lead to ongoing species losses through relaxation
has been challenged by Kellman (1996). He argues
that reference to past environmental change and
species responses to it shows that many species are
more flexible in their ecological requirements than
generally recognized. Kellman contends that many
plant communities are actually undersaturated and
that systems of fragments may be able to sustain
increased species densities for significant periods of
time (although not necessarily indefinitely). He
bases his case in large part on studies from
Neotropical gallery forests, which constitute nar-
row peninsulars of forest extending through a
matrix of savanna ecosystems. According to equi-
librium theory, the tips of the peninsulas should be
impoverished because of their relative isolation.
This extension of island theory is termed the penin-
sula effect. It has been tested at a coarse scale for
Florida and for Baja California, and has generally
been found to be a fairly weak effect (Brown 1987;
Means and Simberloff 1987; Brown and Opler
1990). In the Neotropical gallery forests, Kellman
(1996) reports that there is no evidence for floral
impoverishment with distance from the peninsu-
lar base, possibly because the frugivores that
disperse the seeds of the plants are capable of
adjusting their ranging behaviour to accommo-
date the long, thin peninsular configuration, or
indeed to move through the sparsely wooded
savanna matrix surrounding the peninsulas. These
peninsulas have a long history, and thus appear to
provide evidence for adequate gene flow of plant
populations to the peninsular tips over long peri-
ods, i.e. there has been plenty of time for ‘lag
effects’ to unfold, but at least for plants, there is no
evidence of species impoverishment. However, it is
unclear how good an analogue these peninsulas
provide for the rapidconversion of forested areas to
fragmented systems.
There is a long-term fragmentation project which
sheds some light on these questions, and in which
data were collected before fragmentation: the
Minimum Critical Size of Ecosystem project. It is
located near Manaus, in the Brazilian Amazon, and
sites were turned into habitat islands, isolated from
intact forests by just 70 to 650 m of pasture, between
1980 and 1984. While some patches remained