We investigated habitat selection in a trophically dimorphic population of pumpkinseed sunfish (Lepomis gibbosus) to determine whether littoral and limnetic ecotypes exhibit habitat or site fidelity. A transplant experiment was conducted, in which 998 pumpkinseeds captured from littoral and limnetic sites were marked and released in either the site of capture, the nearest site of the same habitat type, or the nearest site of the opposite habitat type. Daily recapture attempts over the course of the reproductive and growing season provided a 25% recapture rate, 40% of which were recaptured on multiple occasions at the same site. Site fidelity was very high in both ecotypes. Results estimated with a multi-state transition model indicated that the probability of a transplanted pumpkinseed returning to its site of origin ranged from 74% for limnetic pumpkinseeds released into a different limnetic or littoral site, to 93% for littoral pumpkinseeds released into a limnetic site. Furthermore, the probability of a pumpkinseed being recaptured at its site of origin if not transplanted was estimated at 97 and 98% for limnetic individuals and littoral individuals, respectively. Discriminant Function Analysis of helminth parasite loads sampled from littoral and limnetic individuals could classify site of origin with 96-100% accuracy, suggesting that the habitat and site fidelity patterns observed with mark-recapture are indicative of longterm habitat segegation of the two forms. The results of our experiment provide compelling evidence of correlated habitat selection as a function of home range fidelity within both ecotypes of a subtly dimorphic species. Such behaviour could have a significant effect on present or future gene flow.