Examination of the "null" model of connor and simberloff for species co-occurrences on Islands.

Among birds of the Bismarck Archipelago, only certain combinations of the species in a guild coexist on islands, and some species that are very similar ecologically have mutually exclusive distributions. Diamond (1975) interpreted these patterns as biologically significant, involving effects such as competition. Connor and Simberloff (1979) claimed such patterns to be not recognizably different from random, because they were scarcely distinguishable from those generated by a "null" distribution supposedly not incorporating competition.On examining the analysis by Connor and Simberloff, we find that it actually yields the opposite conclusion: three faunas tested have grossly non-random structures, while their test is unworkable for the fourth fauna. The method of Connor and Simberloff for generating a null distribution has many fatal weaknesses: dilution of relevant data from guilds with irrelevant data from the whole species pool; hidden incorporation of effects of competition into the constraints; inability to recognize a checkerboard, the extreme example of a distribution produced by competition; reliance on inefficient Monte Carlo simulations; severe constraints that sometimes make generation of a null distribution impossible; failure to weight species combinations; and failure to identify the direction of non-randomness or the species combinations most responsible.Finally, we use other recent studies by Simberloff and coleagues to examine the value of constructing null hypotheses in community ecology.