HYPOTHESES ON SEED SIZE - TESTS USING THE SEMIARID FLORA OF WESTERN NEW-SOUTH-WALES, AUSTRALIA

Seed size varies over several orders of magnitude in any one community. We outline a number of hypotheses that could account for this variation, briefly discuss the reasoning and evidence underlying each of these hypotheses, and then test each hypothesis with a database of 248 species from the semiarid woodlands of western New South Wales, Australia. Information on seed weight, growth form, plant longevity, height, dispersal mode, dormancy, and germination season was used. We considered not only pairwise relationships between seed weight and each other variable, but also alternative hypotheses whereby relationships arose as a result of indirect correlations through other variables. The strongest associations of seed size were with plant height and growth form. The seed-size variation accounted for by growth form largely overlapped with that accounted for by plant height, but each also accounted for some further variation independently of the other. Of the five hypotheses tested, the correlative patterns were inconsistent for two. Two others showed the predicted pattern, but these patterns could alternatively be interpreted as arising from secondary correlation via the combination of plant height and growth form. Only plant height, growth form, and dispersal mode had significant associations with seed size independent of the other attributes measured.