of the equilibrium thinking at the heart of
MacArthur and Wilson’s (1967) monograph. Thus,
for instance, he writes,
The fact that the larger land-bridge islands around New
Guinea still have double their equilibrium species number
[of birds], more than 10 000 years after the land bridges
were severed, emphasizes for what long times many
species can escape their inevitable doom on a large island’
(Diamond 1975a, p. 370).
If it takes in excess of 10 000 years to reach equi-
librium, the effect would appear to be a relatively
weak one, potentially or actually to be overridden
by other factors. The inevitability of their doom, the
fate ultimately of all living things, as the fossil
record reveals, is not necessarily in practice the
consequence of their island systems moving closer
to a biotically regulatedequilibrium. Simberloff
(1978) made a similar point when he argued that
the statistical analyses of distribution were not
uniquely interpretable as a function of diffuse
competition.
A second critical paper was more strident. It
claimed to show that ‘every assembly rule is either
tautological, trivial or a pattern expected were
species distributed at random’ (Connor and
Simberloff 1979, p. 1132). The rules derived in
Diamond’s analyses were cast in terms of ‘permit-
ted’ and ‘forbidden’ combinations (as rule 2, above)
and ‘resistance’ to invasion. Thus, although they
were derived empirically mostly from distribu-
tional data, competition was invoked within their
formulation. One element of the critique concerned
the great difficulty of demonstrating a particular
role over time for competition in determining
present-day distributional patterns (Law and
Watkinson 1989; Schoener 1989). Simberloff (1978)
argued that the role of competition had been over-
played by several authors, citing Diamond’s stud-
ies in illustration. His argument focused largely on
the methods by which non-random distributions
are calculated. He argued that confidence limits
were not given on the combinational rules, which
were therefore difficult to evaluate, and that there
were too few islands involved to allow rigorous
statistical assessments. He argued that the null
hypothesis should be formulated in terms of
species being considered essentially as similar
units, i.e. without differing dispersal abilities and
ecologies, and that these species should be mod-
elled as bombarding islands entirely at random.
The different islands should be accepted as having
different properties, however. While accepting that
the first of these assumptions is biologically unreal-
istic, Simberloff’s argument was that it provided a
baseline by which to examine biogeographical dis-
tributions. When the data are not consistent with
this null model, ecological phenomena, such as
competition or dispersal differences, can then be
examined for explanations. Simberloff also argued
that the biological characteristics of each species
should be examined for explanatory value as being
a more parsimonious explanation than competi-
tion. Thus one species might be a superior colonist
with respect to another because it has better disper-
sal ability and/or better persistence once immigra-
tion occurs (as a function of per capita birth and
death rates).
Thus Simberloff, whose own formative experi-
ence was the mangrove islet study (a fast develop-
ing, near-source, dynamic system), adopted a
position which is arguably closer to the stochastic,
dynamic equilibrium model of island biogeogra-
phy (EMIB) than did Diamond in his seminal
study of New Guinea island avifaunas (a slow,
more isolated, and less dynamic system with a
strong evolutionary signal). This difference in
view was recognized by Simberloff, who noted
that botanists and invertebrate zoologists have
most often sought distributional explanations in
responses of individual species to physical factors,
whereas vertebrate zoologists commonly invoke
competition. The generality of competitive effects
was in Simberloff’s view largely unproven, and
the reason was that appropriate statistical tests
had yet to be applied to sufficient data sets—
indeed such data sets were, by the nature of the
statistical requirements, bound to be difficult to
obtain. This line of argument was extended by
Connor and Simberloff (1979) who contended that
the patterns deduced by Diamond as the outcome
of competition could be produced by random
processes providing that the following three
constraints were accepted: ‘(1) that each island has114 COMMUNITY ASSEMBLY AND DYNAMICS