Abstract
(1) Floristic lists have been compiled for twenty-three islands in Lake Manapouri, New Zealand. (2) Wilson''s modification of Gilpin and Diamond''s test was used to examine deviation from a null model in which species are distributed independently of each other. There was very clear deviation from the null model of species co-occurrences for native species, especially for woody species and for records inland of the lake shore. The deviations were explicable as habitat differences. The exotic guild showed no departure from the null model. (3) Incidence functions were examined. There was very little evidence for the existence of ''supertramps'', i.e. species tending to occur on small islands. However, the few species showing such a tendency were species that might be expected, from their general ecology, to be among any supertramps. (4) Determinacy is considered, and especially nesting, i.e. the extent to which the flora of a species-poor island is a subset of the flora of species-richer islands. An index of nesting is introduced. Significant nesting existed amongst the native flora, especially for the woody species and especially inland, though it was far from complete. Nesting amongst the exotic species was lower, but still significant. (5) The results are compatible with a community structure determined primarily by the response of individual species to habitat, similar to Gleason''s model. The exotic guild is clearly differently structured from the native guild, and has probably not yet reached equilibrium. Inland occurrences show more structure than shore ones, probably because of the greater variety in shore habitats between islands. Woody species consistently showed more structure than herbaceous ones. This points to community structure with dominant woody ''climax'' species occupying the sites ecologically suitable for them, with a second tier of species dependent on them, or in marginal or disturbed sites.