Theory suggests that speciation is possible without physical isolation of populations (hereafter, nonallopatric speciation), but recent nonallopatric models need the support of irrefutable empirical examples. We collected snails (Littorina saxatilis) from three areas on the NW coast of Spain to investigate the population genetic structure of two ecotypes. Earlier studies suggest that these ecotypes may represent incipient species: a large, thick-shelled ‘RB’ ecotype living among the barnacles in the upper intertidal zone and a small, thin-shelled ‘SU’ ecotype living among the mussels in the lower intertidal zone only 10–30 m away. The two ecotypes overlap and hybridize in a midshore zone only 1–3 m wide. Three different types of molecular markers [allozymes, mitochondrial DNA (mtDNA) and microsatellites] consistently indicated partial reproductive isolation between the RB and the SU ecotypes at a particular site. However, each ecotype was related more closely to the other ecotype from the same site than to the same ecotype from another site further along the Galician coast (25–77 km away). These findings supported earlier results based solely on allozyme variation and we could now reject the possibility that selection produced these patterns. The patterns of genetic variation supported a nonallopatric model in which the ecotypes are formed independently at each site by parallel evolution and where the reproductive barriers are a byproduct of divergent selection for body size. We argue that neither our laboratory hybridization experiments nor our molecular data are compatible with a model based on allopatric ecotype formation, secondary overlap and introgression.