To understand how allopatric speciation proceeds, we need information on barriers to gene flow, their antiquity, and their efficacy. For marine organisms with planktonic larvae, much of this information can only be obtained through the determination of divergence between populations. We evaluated the importance of ocean barriers by studying the mitochondrial DNA phylogeography of Tripneustes, a pantropical genus of shallow water sea urchin. A region of cytochrome oxidase I (COI) was sequenced in 187 individuals from locations around the globe. The COI phylogeny agreed with a previously published phylogeny of bindin that barriers important to the evolution of Tripneustes are: (1) the cold water upwelling close to the tip of South Africa, (2) the Isthmus of Panama, (3) the long stretch of deep water separating the eastern from the western Atlantic, and (4) the freshwater plume of the Orinoco and the Amazon rivers between the Caribbean and the coast of Brazil. These barriers have previously been shown to be important in at least a subset of the shallow water marine organisms in which phylogeography has been studied. In contrast, the Eastern Pacific Barrier, 5000 km of deep water between the central and the eastern Pacific that has caused the deepest splits in other genera of sea urchins, is remarkably unimportant as a cause of genetic subdivision in Tripneustes. There is also no discernible subdivision between the Pacific and Indian Ocean populations of this genus. The most common COI haplotype is found in the eastern, central, and western Pacific as well as the Indian Ocean. Morphology, COI, and bindin data agree that T. depressus from the eastern Pacific and T. gratilla from the western Pacific are, in fact, the same species. The distribution of haplotype differences in the Indo-Pacific exhibits characteristics expected from a sea urchin genus with ephemeral local populations, but with high fecundity, dispersal, and growth: there is little phylogenetic structure, and mismatch distributions conform to models of recent population expansion on a nearly global scale. Yet, comparisons between local populations produce large and significant F(ST) values, indicating nonrandom haplotype distribution. This apparent local differentiation is only weakly reflected in regional divergence, and there is no evidence of isolation by distance in correlations between F(ST) values and either geographical or current distance. Thus, Tripneustes in the Indo-Pacific (but not in the Atlantic) seems to be one large metapopulation spanning two oceans and containing chaotic, nonequilibrium local variation, produced by the haphazard arrival of larvae or by unpredictable local extinction.