Modern speciation theory is heavily influenced by Mayr's postulate that prolonged geographical isolation is necessary for differentiated populations to evolve reproductive isolation. Present-day distributions of sister species are consistent with allopatric or peripatric speciation in many terrestrial and freshwater animal groups. However, the oceans present few obstacles to dispersal for marine taxa with planktonic larvae, and sister species are not often split at biogeographical breakpoints in the sea. Theory predicts that disruptive selection on habitat choice or resource use can split a population into divergent ecotypes without physical separation, yet sympatric speciation is still often viewed as improbable. Here, I review phylogenetic evidence from diverse marine gastropods to test Mayr's prediction that recently diverged sister species should not be sympatric over most of their ranges. In contrast to expectations, young sister species are often broadly sympatric in many gastropod groups, suggesting that classical models of allopatric divergence are insufficient to explain marine speciation. I discuss four mechanisms that may contribute to this deviation from predicted biogeographical patterns: transient allopatry along continuous coastlines, rapid evolution of gamete recognition proteins, shifts to non-planktonic development, and ecological divergence. The available evidence argues that patterns of marine speciation depend on complex interactions between geography, life history, and ecology, often resulting in local radiations within a basin or endemic to an island group. Whether selection acts on ecotypes in sympatry or on populations during secondary contact, ecological factors may promote speciation in the sea at smaller spatial scales than expected. I highlight areas for future study to improve our understanding of the forces generating marine biodiversity, and why the geography of speciation may be fundamentally different for shallow-water animals.