1. Center for Tropical Research, Institute of the Environment, University of California, Los Angeles, California 90065, USA
2. Organismic Biology, Ecology and Evolution, University of California, Los Angeles, California 90065, USA

Correspondence to: Ryan Calsbeek¹ Email: calsbeek@ucla.edu

Islands are considered to be natural laboratories in which to examine evolution because of the implicit assumption that limited gene flow allows tests of evolutionary processes in isolated replicates¹. Here we show that this well-accepted idea requires re-examination. Island inundation during hurricanes can have devastating effects on lizard populations in the Bahamas^{2, 3}. After severe storms, islands may be recolonized by over-water dispersal of lizards from neighbouring islands³. High levels of gene flow may homogenize genes responsible for divergence, and are widely viewed as a constraining force on evolution^{4, 5}. Ultimately, the magnitude of gene flow determines the extent to which populations diverge from one another, and whether or not they eventually form new species^{6, 7}. We show that patterns of gene flow among island populations of Anolis lizards are best explained by prevailing ocean currents, and that over-water dispersal has evolutionary consequences. Across islands, divergence in fitness-related morphology decreases with increasing gene flow⁵. Results suggest that over-water dispersal after hurricanes constrains adaptive diversification in Anolis lizards, and that it may have an important but previously undocumented role in this classical example of adaptive radiation.