Scorpions are an excellent system for understanding biogeographical patterns. Most major scorpion lineages predate modern landforms, making them suitable for testing hypotheses of vicariance and dispersal. The Caribbean islands are endowed with a rich and largely endemic scorpion fauna, the origins of which have not been previously investigated with modern biogeographical methods. Three sets of hypotheses have been proposed to explain present patterns of diversity in the Caribbean: (1) connections via land bridges, (2) vicariance events, and (3) overwater dispersal from continents and among islands. The present study investigates the biogeographical diversification of the New World buthid scorpion subfamily Centruroidinae Kraus, 1955, a clade of seven genera and more than 110 species; infers the ancestral distributions of these scorpions; and tests the relative roles of vicariance and dispersal in the formation of their present distributions. A fossil-calibrated molecular phylogeny was estimated with a Bayesian criterion to infer the dates of diversification events from which ancestral distributions were reconstructed, and the relative likelihood of models of vicariance vs. dispersal, calculated. Although both the timing of diversification and the ancestral distributions were congruent with the GAARlandia land-bridge hypothesis, there was no significant difference between distance-dependent models with or without the land-bridge. Heteroctenus Pocock, 1893, the Caribbean-endemic sister taxon of Centruroides Marx, 1890 provides evidence for a Caribbean ancestor, which subsequently colonized Central America and North America, and eventually re-colonized the Greater Antilles. This 'reverse colonization' event of a continent from an island demonstrates the importance of islands as a potential source of biodiversity.