TURNOVER 105Rosenzweig (1995) invokes both competition and
predation as reasons why extinction rates should
rise with increasing size of the assemblage. The
data he cites for a direct role of predation are not
entirely convincing of a general effect, as they come
(with one exception) from predators introduced by
humans to remote islands, the biota of which had
evolved without them. The exception was for data
for orb spiders on very small islands, where they
were preyed upon by lizards. The support in this
case is not entirely unequivocal (Toft and Schoener
1983) and, indeed, Spiller and Schoener (1995) have
shown that the effects of the lizards on spider den-
sities within these islands vary significantly over
time as a function of rainfall variations (data on
turnover were not given in the latter study).
Additional evidence for predators causing rising
extinction rates have come from studies by
Thornton (1996) and his colleagues of the coloniza-
tion of the newly formed island of Anak Krakatau
(the fourth Krakatau island, which emerged from
the sea in 1930). As the island began to support
appreciable areas of vegetation in the 1980s, frugiv-
orous and insectivorous birds were able to colonize
in greater numbers, but following the establish-
ment of raptors on the island, several species suf-
fered considerable reductions. In two cases at least,
their disappearance from the island was attributed
to predation. This is an intriguing finding, as in
other ecological contexts predators have often been
invoked as preventing competitive exclusion and
thereby enhancing diversity at lower trophic levels
(Caswell 1978; Begon et al. 1986). In the mid-1980s
Anak Krakatau was able to support only a single
pair of oriental hobbies, which in 1989 were
replaced by a pair of peregrine falcons, suggesting
the position of top predator to be rather marginal.
Anak Krakatau’s avifauna is set in the context of
three nearby islands, and individuals found on the
island may well be only partially dependent on it
for food supplies. For instance, Thornton (1996) has
noted that the home range of the other avian pred-
ator on Anak Krakatau, the barn owl, is sufficient to
take in all four Krakatau islands. In such cases, a
predator may be able to remove one element of its
in situfood supply while supplementing its diet on
prey from other nearby islands (equally, new
arrivals or stragglers may be killed by resident
predators on small islands, cf. Diamond 1974). In
this fashion, Anak Krakatau may be functioning not
as a closed but as an open system (cf. Caswell 1978).
The role of predators may also be scale
dependent, such that development of the island’s
ecosystems to a greater biomass might diminish the
ability of predators to cause the total loss of a prey
species. This suggestion is consistent with
Diamond’s (1975a) observations of contrasts
between Long Island and the more depauperate
Ritter, both being islands recovering from volcanic
disturbance. Caswell (1978) makes a similar sugges-
tion on the basis of his non-equilibrium modelling
of predator–prey relationships in open systems.
In order to demonstrate that the extinction curve
does indeed have a concave, rising form, produced
by species interactions, Rosenzweig (1995) had to
resort to a complex line of reasoning, based on
ISARS from two data sets, each of island archipela-
gos created at the end of the last ice age and
presumed to be undergoing relaxation. This consti-
tutes a rather indirect line of proof, reliant upon a
series of assumptions being taken as read. It is
intriguing that he had to work so hard to find sup-
port for such a basic element of the EMIB. The gen-
erality of the concave, rising trend of extinction does
not seem to be founded on much empirical evidence
yet. Data can also be found that, at face value, con-
tradict this expectation.
There is no reason to assume that there is a single
path to extinction. In the process of a species declin-
ing (perhaps in fluctuating fashion) to extinction,
its population must, for some period, be small. The
small size of the population may be a good predic-
tor of extinction, but why is the population small in
the first place? The explanation could lie in the
trophic status of the organism; in predation, dis-
ease, or disaster; in resource shortage; or in compe-
tition with other species of its trophic level. It is, as
Williamson (1981) notes, remarkable that the equi-
librium theory was widely accepted with almost no
evidence on the distribution of probabilities of
going extinct and how these vary (cf. Pimm et al.
1988). It is arguable that, in many cases, habitat
availability and stability, and food web connections
(and continuous availability of resources), are more