On Trinidad (4520 km2), 12 bird species in foothills of the Northern Range (11 hummingbirds and Coereba flaveola, a passerine) often use floral nectar for carbohydrate food. A subset of 6 (5 hummingbirds and Coereba) occur on Tobago, a smaller (295 km2), more isolated island. For 13 mo we compared these assemblages at study sites having similar climates and similar sets of bird—visited flowers (34 species at Trinidad sites, 31 on Tobago). Density of nectar resources for birds varied seasonally>100—fold on each island. During peaks in total flower density, much nectar available to birds went unused. In contrast, during a flower—poor 3—mo period (September—November 1977), nectar available in flowers was held near zero because supplies were consumed as fast as they were secreted. Competition for this limited supply of nectar appeared to be intense on both islands. Diets of coexisting bird populations diverged, especially at the Tobago site, where only 3 nectar—feeding bird species, with disparate morphologies, persisted: 1 short—billed hummingbird, 1 long—billed hummingbird, and Coereba. At other seasons, though, a variety, of bird species consumed nectar on both islands, and often diets of different species were nearly identical. Despite the twofold difference in species richness, the two bird assemblages used nectar resources similarly, contrary to the theory that species—rich and species—poor guilds should differ in patterns of resource use. There were no statistical differences between the islands in the ratio of demand for nectar to nectar supply (expected by theory to be lower on the average on species—poor Tobago), monthly variation in that ratio (expected to be greater on Tobago), or breadth of the entire guild's diet (expected to be lesser on Tobago). One reason may have been that supplies of nectar at both sites fluctuated so widely that demand for nectar in neither bird assemblage could change quickly enough to keep pace. Second, foraging by the hummingbird Chrysolampis mosquitus, a migrant common on Tobago for the 9 mo of the year outside the period of food shortage, made up in part for nectar use by the 6 hummingbird species found on Trinidad but not Tobago. C. mosquitus on Trinidad rarely entered our primary study site, instead shifting to a type of habitat not available on Tobago and relatively little used by other hummingbird species. We found few significant differences in the intensity of competition on the two islands. One hummingbird species, Amazilia tobaci, won most interspecific aggressive encounters over nectar on both islands. There were no significant differences in the intensity of exploitative competition (measured as mean interspecific crowding) experienced by most species on different islands. C. mosquitus, though, may have left Tobago during the 3—mo nectar shortage due to a combination of interference competition (from A. tobaci) and exploitative competition (from all 3 resident species), whereas at that time the Trinidad C. mosquitus population apparently left the island due to a shortage of appropriate flowers in its preferred habitat. Species composition on Tobago can partly be explained by examining traits of species present on Trinidad, the principal source pool. Although individuals in all 5 Tobago hummingbird populations were significantly larger in one or more morphological dimensions (bill length, wing length, mass, or wing disc loading) than conspecifics on Trinidad, the adaptive significance of these size increases is obscure in all but one case (A. tobaci). In addition to Chrysolampis mosquitus and Coereba flaveola, a passerine that consumes fruits and insects as well as nectar, the other birds that used nectar at the Trinidad study sites were: 2 large—bodied, stout—billed hummingbird species that consume many arthropods and that, like C. mosquitus, may sometimes migrate to the mainland; and two distinct 4—species sets of resident hummingbirds, one set with short bills (28 mm), often curved, bills. Only one species from each of these 4—species sets was found on Tobago. We examined Trinidad's 4—species sets closely to see if the Tobago representative from each, short—billed A. tobaci or long—billed Glaucis hirsuta, was a random draw with respect to population traits thought to enhance persistence on small islands. Diets of A. tobaci and G. hirsuta on Trinidad were not particularly broad, nor was either Trinidad population particularly free of exploitative competition, relative to all other species there. At the Trinidad study site, though, A. tobaci maintained the highest and least variable demand for nectar (hence, presumably, the highest and least variable population density) of any short—billed species, and G. hirsuta maintained a higher and less variable demand than 2 of 3 other long—billed species. We argue that the relatively low seasonal variation in demand for nectar (and, presumably, in population density) within the second—growth habitat is the most straightforward explanation for persistence on Tobago of A. tobaci and G. hirsuta, rather than other species from their respective groups. The result implies that the other 6 Trinidadian populations wander relatively more than A. tobaci or G. hirsuta, respectively, among different habitat types and thus depend more heavily on a regional habitat mosaic not likely to occur on a small island such as Tobago.