did not restrict analyses to particular ecological
guilds, this ‘intrafamily’ result is both unsurpris-
ing and by no means contradictory to Diamond’s
(1975a) arguments.
For the record, Sanderson et al. (1998), Grant
(1998), and Gotelli and Entsminger (2001) have con-
tinued the debate on how to construct null model
simulations, again using the New Hebridean bird
data as the subject matter. Although differing in the
details, the overall picture is that some degree of
non-randomness can indeed be detected.
Such debates are rarely settled by the triumph of
one viewpoint to the complete exclusion of the
other, but, as in this case, can sometimes result in
the re-examination and refinement of the original
models and positions. Connor and Simberloff’s
(1979) intervention led to increased attention to the
problems of statistical analyses in island assembly
structure, which was their stated intent. Even now,
it would be a bold move to declare the debate over,
but we offer the following conclusions. The original
(Diamond 1975a) statements of island assembly
theory laid too much emphasis on competition at
the expense of the many other factors (Table 5.1)
that could be involved in structuring species distri-
butions across islands. Although there is room for
debate about details of the assembly rules Diamond
identified in his original analyses, the evidence is
convincing that some non-randomness is involved.
Moreover, within the avifaunal data (especially that
from the Bismarcks), it is possible to find what
appear to be compelling cases for competitive
effects within guildsin particular cases.
At this point, we have to introduce a final note of
caution: the patterns in distribution of the avifauna,
and especially pigeons and doves may have been
altered by humans. Steadman (1997b) provides a
comprehensive review of prehistoric bone data for
the extinction of columbids (pigeons and doves)
from Polynesian islands. He notes that losses
include the extinction of at least nine species in six
genera, as well as the extirpation of numerous
island populations of still extant species. If it were
not for humans, a typically West Polynesian island
would support at least six or seven species in four
or five genera, instead of the observed one to six
species in one to five genera. He comments that one
of Diamond’s supertramp species, Ducula pacifica,
which is found in both West and East Polynesia,
appears to have expanded its range eastward after
the arrival of people, which would be consistent
with a response to human induced extirpation of
populations of more competitive species. Having
reviewed Polynesian evidence, Steadman notes
that there is by contrast a paucity of prehistoric
bone data from Melanesia. Some losses have, how-
ever, been documented, and Steadman considers it
likely that many more await discovery. In time, it
may turn out that some of the chequerboards and
other assembly rule patterns of columbids at the
heart of these debates have been structured not just
by the ‘natural’ forces of competition, habitat con-
trols, volcanic eruptions, and other historical
processes, but also by human interference.
There has been something of a resurgence of
interest in assembly rules in recent years (e.g.
Weiher and Keddy 1995, Fox and Fox 2000). Gotelli
and McCabe’s (2002) meta-analysis of 96 previously
published matrices found them to be highly non-
random. As predicted by island assembly theory,
there were fewer species combinations, more che-
querboard species pairs, and less co-occurrence in
real matrices than expected by chance. Although
many of their data sets are not concerned with or
are not limited to insular data and thus lie slightly
outside the scope of this review, these findings are
nonetheless significant in the present context.Exploring incidence functions
One of the problems of evaluating island assembly
theory as detailed above is that the biogeographical
context of the island avifaunas under discussion is
in each case quite complex. We now review some
simpler systems, involving non-volant mammals
on small, near-shore islands.
Hanski (1986) discusses occurrence and turnover
of three species of shrews on islands in lakes in
Finland, a system in which individual species pop-
ulations could be studied in some detail. The
largest species, Sorex araneus, was found to have a
low extinction rate except on islands of less than
2 ha, on which demographic stochasticity becomes
important. The smaller S.caecutiensandS. minutus,118 COMMUNITY ASSEMBLY AND DYNAMICS