0198566123.pdf

in their null models build in occurrence frequen-
cies and incidences of species which can be shown
to be influenced by competition by reference
to occurrence frequencies of members of the
same guild in different archipelagos (Bismarck,
Solomon, and New Hebrides). These data show
that the fewer competing species of the same
guild that share an archipelago, the higher the
occurrence frequency of particular species, often
by a factor of 10 or more.
3 They produced a statistical procedure incapable
of recognizing a chequerboard distribution.
4 Simulation procedures used were inadequate to
provide realistic simulations of the matrices.

This is an incomplete listing of the objections
raised by Diamond and Gilpin (1982), but the gen-
eral position they adopt is summed up as follows:
‘how can one pretend that one’s “null model” is
everything-significant-except-X, when it was con-
structed by rearranging an observed database that
may have been organized by X?’ (Diamond and
Gilpin 1982, p. 73). The problem thus afflicts the
proponents on both side of the argument, and is
one that has since reappeared in other applications
of null models (papers in Strong et al. 1984; Weiher
and Keddy 1995). All null models involve assump-
tions; the trick is to recognize what they are, the dis-
agreements come over which are the most realistic
and appropriate set to use (Box 5.1).
In their second paper, Gilpin and Diamond
(1982) developed a test that they regarded as an
improvement over the Monte Carlo modelling
used by Connor and Simberloff, and affirmed their
earlier conclusion that the distribution patterns
were indeed non-random. Comparing results for
the Bismarck Islands and the New Hebrides, they
found non-random patterns in each system, but
that the richer Bismarck Islands had clearer evid-
ence of competitive effects. The New Hebrides
have a smaller avifauna (56 compared with 151
species) and thus negative associations (as would
be produced by competition) did not exceed a
random expectation: neither were chequerboard
distributions evident for the New Hebrides. In
contrast, they re-affirmed the existence of negative
associations and chequerboard distributions

amongst ecologically similar pairs of Bismarck

species. They found that some pairs of species have

more exclusive distributions than expected by

chance, and invoked as likely controlling factors,

competition, differing distributional strategies,

and different geographical origins. Examples of the

first two have been given above. In illustration of

the third category, the hawk Falco berigorahas

spread from the west, whereas the parrot

Chalopsitta cardinalishas spread from the east, but

in each case to a limited degree. They are not mem-

bers of the same guild, obviously, and their cur-

rently exclusive distributions relate to their

differing geographical origins rather than competi-

tion. Other species pairs were found to have more

coincident distributions than would be expected

by chance alone, and this they interpreted in rela-

tion to shared habitat, single-island endemisms,

shared distributional strategies, or shared geo-

graphical origins (Table 5.1).

As Gilpin and Diamond (1982) point out, much of

the information about non-random co-occurrences

is actually contained in the incidence functions––the

first element in the whole analysis––which are

much more intuitively accessible sources of infor-

mation than are the more sophisticated modelling

exercises. Equally, the chequerboard distributions

provided unambiguous patterns, strongly sugges-

tive of competitive effects: in such cases restricting

analyses to members within a single guild is clearly

necessary (Diamond 1975a). Thus, where observed

within guilds of pigeons for which there is excellent

evidence of colonizing ability across moderate

116 COMMUNITY ASSEMBLY AND DYNAMICS

Table 5.1Factors invoked in explanation of non-random

co-occurrence of island birds by Diamond and Gilpin in their

articles on assembly rules

Negative Positive

Competition Shared habitat

Differing distributional strategies Shared distributional

strategies

Differing distributional origins Shared geographical

origins

Single-island endemics