Divergent Natural selection for efficient exploitation of alternative resources is thought to be a primary cause of adaptive radiation. I transplanted two recently diverged stickleback species (Gasterosteus sp.) and their Hybrids between habitats to test three predictions of this hypothesis: (1) the form of natural selection on feeding traits should differ between resource environments; (2) fitness trade—offs should occur between morphs specialized for different resources; (3) feeding efficiency should be the basis of fitness differences between morphs. Growth rate was used as a surrogate component of fitness. All three predictions were upheld. The larger of the two species (the Benthic), which has a larger mouth, deeper body, and fewer, shorter gill rakers than the second species (the Limnetic), had a twofold growth advantage in the littoral zone of the lake. This pattern was reversed in the open—water habitat, where the Limnetic species grew at double the rate of the Benthic. Hybrids, which are morphologically intermediate, tended to be intermediate in growth in both habitats. Growth rates were closely predicted by earlier laboratory measurements of feeding efficiency of the three forms in the same habitats, thus linking growth (and probably fitness) with feeding performance and morphology. I compare the strengths of the present approach with other methods that have been used to test the role of resources in adaptive radiation. I also argue that steep trade—offs in fitness along the littoral—open—water habitat gradient have two further consequences. First, they may cause Hybrids to have low fitness, allowing Benthics and Limnetics to coexist in the absence of perfect premating isolation between them. Second, they may promote ecological character displacement. Steep trade—offs may thus help explain why divergence along the littoral—open—water gradient is so common in fish that have colonized low—diversity lakes.