Experimentally assessing the relative importance of predation and competition as agents of selection.

Field experiments that measure natural selection in response to manipulations of the selective regime are extremely rare 1 , even in systems where the ecological basis of adaptation has been studied extensively. The adaptive radiation of Caribbean Anolis lizards has been studied for decades Natural selection is often measured in wild animal populations, but only rarely are changes in selection quantified in response to experimental manipulations of the putative ecological agents of selection Lizards in the genus Anolis comprise one of the most well-studied vertebrate radiations on the planet, with nearly 400 species distributed throughout the Caribbean and the mainland Americas We tested the relative significance of predation and competition by manipulating replicate island populations of the brown anole, A. sagrei, in The Bahamas. We selected small, offshore study islands based on their similarity in vegetation structure, such that variation in island area (range 800-2,300 m 2 ) would generate a natural range of population densities (0.09-0.30 lizards per square metre) 22 when seeded with lizards (n 5 40-80 males, 150 females per island). During spring (May), we captured and uniquely marked lizards from a single population on the nearby main island of Great Exuma and then released them to our experimental islands. Females were included in estimates of total lizard density on each island, but our analyses of selection focused on adult males (see Methods). For each male, we measured body size (snout-vent length to the nearest 1.0 mm), relative hindlimb length (to the nearest 0.5 mm) and stamina (running endurance on a treadmill, in seconds) following standard methods 19 . Four months later, at the end of the breeding season (September), we conducted population censuses to recapture survivors and measure the strength and form of viability selection on these traits We conducted three predator treatments (no predators, bird predators only, bird and snake predators) with two island replicates of each treatment. Islands in the 'no predators' treatment were enshrouded with No-Tangle bird-proof netting to exclude avian predators. Islands in the 'bird predators' treatment received an equal amount of netting around their perimeters to control for the structural addition of nets, but were left open to avian predators. Islands in the 'bird and snake predators' treatment were seeded with predatory snakes (Alsophis vudii, Bahamian racer), such that lizards were exposed to both avian and terrestrial predators. Racers are naturall