a random community, without reference to particular
patterns of turnover. Connor and McCoy (1979) term
thispassive sampling, and advocate its use as a null
hypothesis against all alternatives (debated in
Sugihara 1981; Connor et al. 1983).
●Theequilibrium model (EMIB)postulates the
number of species on an island as a dynamic
equilibrium between immigration and extinction,
dependent on island isolation and area. Only this
hypothesis involves a constant turnover of species.
●Habitat diversity. The number of species may be
a function of the number of habitats: the larger the
island, the larger the number of habitats.
●Incidence functions. Some species can occur
only on large islands because they need large terri-
tories, others only on small islands where they can
escape from competition (cf. Diamond 1974).
●Species-energy theory. According to this theory,
the capacity for richness is considered a function of
the resource base of the island, where the latter may
be estimated, for example, using total primary pro-
ductivity multiplied by area (Wright 1983). This
mechanism may account for variation in species
richness but is essentially neutral on the issue of
turnover.
●Small-island effect. According to this theory, cer-
tain species cannot occur on islands below a certain
size (e.g. Whitehead and Jones 1969). This effect
may be more apparent in marine islands than non-
marine isolates.
●Small-island habitat effect. In contrast to the
previous idea, it has been suggested that in some
systems, small islands may be different in character
because of their smallness, so that they actually
possess habitats not possessed by larger islands
and thus sample an extra little ‘pool’ of species.
Alternatively, they may just have a greater diversity
of habitats than anticipated from their area, e.g. via
telescoping of altitudinal zones (Chapter 2).
●The disturbance hypothesis postulates that
small islands or ‘habitat islands’ suffer greater dis-
turbance, and disturbance removes species or
makes sites less suitable for a portion of the species
pool (e.g. McGuinness 1984, and for a contrary
example Wardle et al. 1997). Disturbance might also
open up sites to invasion by new members.
The above list illustrates that simply finding a
significant relationship between species and area
is not conclusive in evaluating the EMIB. Kelly
et al. (1989) made use of a rather underexploited
feature of the EMIB: that given knowledge of the
species pool, it should be possible to predict the
proportion of species in common between
islands, or between equal-sized sample areason the
islands. Kelly et al. (1989) have used the latter
approach to evaluate the first three of the above
hypotheses for 23 islands within Lake Manapouri,
New Zealand. It had previously been established
that there was a statistically significant ISAR for
the whole flora, with some indication that habitat
diversity might be the cause. In their study, they
sampled for richness of vascular plant species in
two vegetation types (beech forest and manuka
scrub) with a fixed quadrat size. This should elim-
inate the effect of island area on observed species
richness if either the random placement or the
habitat diversity hypothesis is correct, but should
leave an effect of island area if the equilibrium
hypothesis is correct. The equilibrium hypothesis
failed this particular test. As their sampling
design was also intended to eliminate the random
placement and habitat diversity hypotheses, they
suggested that other ideas, such as incidence
functions or the small-island effect, might have
explanatory value.
Rosenzweig (1995) has expressed some doubts
as to the meaning of this refutation, arguing that
the samples were too small, that the restriction to
single habitat types was inappropriate, and that
the islands were insufficiently isolated. The plots
used by Kelly et al. (1989) were of 100 m^2 , which is
not only a fairly standard size for use in New
Zealand forests of this character (J. B. Wilson, per-
sonal communication), but is also respectable in
relation to islands varying from 242 m^2 to
2.675 km^2 in total area. The restriction of sampling
to particular forest types also seems a reasonable
procedure for distinguishing between the three
hypotheses as stated above. The point about isola-
tion, although it does not invalidate the test, is a
fair one. Unfortunately for advocates of the EMIB,
similar problems apply to many other island eco-
logical studies. Arguably, this study exemplifies88 SPECIES NUMBERS GAMES: THE MACROECOLOGY OF ISLAND BIOTAS