Despite growing interest in species' range shifts, little is known about the ecological and evolutionary factors that control geographic range boundaries. We investigated the processes that maintain the northern range limit of the mud fiddler crab (Uca pugnax) at North Scituate, Massachusetts, USA (42 degrees 14' N), located approximately 60 km north of Cape Cod. Larvae from five populations in Massachusetts were reared under controlled temperatures to test whether cooler water near the edge of this species' range inhibits planktonic development. Few larvae completed development at temperatures < 18 degrees C, a threshold that larvae would regularly encounter north of Cape Cod. Extensive salt marshes are present north of the current range boundary, and a transplant experiment using field enclosures confirmed that benthic fiddler crabs can survive severe winter conditions in this northern habitat. Taken with oceanographic data, these results suggest that the range boundary of fiddler crabs is likely maintained by the influence of cooler water temperatures on the larval phase. Analyses of mitochondrial DNA sequences from a neutral marker (COI) indicate high gene flow among U. pugnax populations in Massachusetts with little differentiation across Cape Cod. Consistent with predictions regarding the homogenizing influence of gene flow, larvae from source populations north and south of Cape Cod shared a common lower threshold for development. However, larvae produced near the range edge had faster growth rates than those from the south side of Cape Cod (typically reaching the final megalopal stage 1.0-5.5 d faster at 18 degrees C). Additional studies are needed to determine the mechanism underlying this counter-gradient variation in development time. We hypothesize that dispersal into cooler water on the north side of Cape Cod may act as a selection filter that sieves out slow developers from the larval pool by increasing planktonic duration and exposure to associated sources of mortality. Thus while high gene flow may prevent the evolution of greater cold tolerance in northern populations, recurrent selection on existing variation may lead to an unexpected concentration of favorable adaptations at the edge of the range. Such a pattern could permit edge populations to play a dominant and unrecognized role in future range extensions.