Populations of the littorine gastropod Bembicium vittatum in 47 isolated tidal ponds were used to test effects of isolation and habitat on genetic divergence. The mean G(ST) at 13 polymorphic allozyme loci between pond populations and their adjacent shore sites, separated by surface barriers of 1-100 m, was 0.122, which is equivalent to subdivision over distances of 4-5 km among shore sites. At distances up to 15 km, subdivision among pond populations (mean G(ST)-0.181) was substantially greater than among shore sites (mean G(ST)-0.115); beyond 45 km mean G(ST) was 0.342 among ponds, compared with 0.279 among shore sites. Genetic divergence of pond populations from their adjacent shore counterparts was less in those ponds with greater interchange with the ocean. Greater divergence was associated with decreased allozyme variation within the pond populations, which was also lower in smaller ponds with smaller populations. With one exception, allelic frequencies did not differ consistently between ponds and shores or in association with differences in habitat among the ponds. The exception was the consistently higher frequency of Lap1(112) in pond populations. A mark-recapture experiment failed to detect differential survival or growth of Lnp1 genotypes, either within or among ponds and shore sites, leaving open the mechanisms of this apparently adaptive divergence. Overall, the results confirm a substantial increase in genetic divergence associated with the tidal ponds, and favour the interpretation that degree of isolation and population size, rather than localized selection, are the major determinants of the allozyme divergence.