a given number of species, (2) that each species is
found on a given number of islands, and (3) that
each species is permitted to colonize islands con-
stituting only a subset of island sizes’ (Connor and
Simberloff 1979, p. 1132). They based their claims
on data for New Hebridean (Vanuatu) birds, West
Indian birds, and West Indian bats, which they
analysed for evidence of Diamond’s assembly
rules. The statistical part of their analysis failed to
support the existence of assembly rules in their
data.
The quote given earlier referred to tautologies
and trivial features within the assembly rules.
These arguments are difficult to summarize but
amount to an attack on the line of reasoning used to
construct the assembly rules. In illustration, the
third rule ‘A combination that is stable on a large or
species-rich island may be unstable on a small or
species-poor island’ was argued to be reducible to
‘A combination which is found on species-rich
islands may not be found on species-poor islands’
because, first, species number is used operationally
by Diamond in lieu of island size and, secondly,
because stability is largely dependent on the com-
bination of species in question having been
observed. As large islands contain on average more
species, it would be expected by chance alone that
some combinations found on rich islands will not
occur on species-poor islands. By this line of rea-
soning, the third rule was deemed to be a trivial
outcome of the definitions used and otherwise a
probabilistic outcome.
Another important line of criticism concerned
the chequerboard distributions, which they
argued could represent ongoing cases of geo-
graphic speciation without re-invasion, i.e. they
could be cases of divergence between allopatric or
parapatric lineages, rather than being the outcome
of competitive exclusion. The position they adopt
at the end of their paper is clear. The assembly
rules were flawed because at no time was a simple
null hypothesis framed and tested, competitive
effects were invoked without proof and where
simpler explanations were available. Nonetheless,
Connor and Simberloff stressed that they were not
implying that island species are distributed
randomly and that interspecific competition does
not occur, just that neither non-random distributions
nor the primacy of competition in determining
distributions were convincingly demonstrated by
Diamond’s analyses.
Not surprisingly, these criticisms drew a
response, in the form of consecutive articles by
Gilpin and Diamond (1982) and by Diamond and
Gilpin (1983), and indeed generated a flurry of
articles and comments within the literature of the
late 1970s and early 1980s. Perhaps the core of
Diamond and Gilpin’s (1982) response is a chal-
lenge to the primacy of Connor and Simberloff’s
conception of a null hypothesis. Diamond and
Gilpin (1982) take the reasonable position that
biogeographical distributions may be influenced
by many factors, including competition, predation,
dispersal, habitat, climate and chance, and that the
mix is liable to vary with the group of organisms
considered and with spatial and temporal scale.
Further, it follows that no single theory, hypothe-
sis, or conceptual position has logical primacy, or
special claim to be the most parsimonious null
hypothesis (they signify this by placing ‘null’ in
quotation marks throughout). Each should be
regarded as a competing or contributory hypothe-
sis to be evaluated on its merits. This is a pragmatic
position adopted generally in the present book:
different answers may be found to similar ques-
tions posed in different island biogeographical
contexts. However, there is, of course, more to the
debate than this. Diamond and Gilpin in fact took
the opportunity to respond to a series of papers by
Simberloff, Connor, Strong, and others, employing
null hypotheses. The problem with null hypothe-
ses, i.e. hypotheses of ‘no effect’, is the assump-
tions that are made in their formulation (Colwell
and Winkler 1984; Shrader-Frechette and McCoy
1993). Diamond and Gilpin argue that in generat-
ing their null distributions Connor and Simberloff
(1979) took the following inappropriate steps:1 Whereas assembly rule theory posits competitive
effects within guilds, they diluted their data by
including data for the whole species pool instead of
only for the target guild.
2 They incorporated hidden effects of competition
into their constraints. The assumptions employedISLAND ASSEMBLY THEORY 115