Considerable advances are being made in our understanding of divergent selection and ecological speciation. However, it remains unclear to what extent diversifying selection, responsible for divergence of distinct populations, also contributes to earlier stages of population diversification such as resource polymorphisms that persist despite gene flow. A general expectation of resource polymorphism is that phenotypic trade-offs influence fitness via trophic performance, resulting in disruptive selection. Despite numerous polymorphisms exhibiting morphological variation consistent with expectations of trophic trade-offs, there have been few attempts to examine these in the context of natural selection. As a result, we do not know how phenotype affects fitness and whether morphologically-based performance trade-offs contribute to the earliest stages of adaptive diversification. In this thesis, I evaluated the hypothesis that morphological trade-offs influence fitness via trophic performance, thus contributing to disruptive selection. First, I characterised the shape-dependence of resource use in polymorphic Arctic charr (Salvelinus alpinus) in two contrasting Icelandic lakes: Thingvallavatn and Vatnshlíðarvatn. In both lakes, resource use was shape-dependent consistent with biomechanical expectations differentiating morphs. However, shape-resource associations within morphs did not always correspond to patterns distinguishing morphs. I then estimated effects of shape on growth via trophic resource use, using a path analytical approach and calculating performance gradients. Though observations were consistent with disruptive selection of shape in Thingvallavatn, there were no effects in either lake consistent with the hypothesised function of trophic performance. Instead, I observed shape-dependent parasitism and non-trophic effects of shape influencing growth, and contrasting patterns of size-dependent mortality between coexisting morphs. Finally, I demonstrated the conditions under which performance gradients can be considered valid estimates of selection, and clarified what inferences can be made where direct estimates of fitness in nature are unobtainable. This thesis contributes to our understanding of early diversification by demonstrating inconsistencies between general expectations of adaptive mechanisms and processes evident in nature. Intuitive mechanisms such as morphological trade-offs in trophic performance might not affect fitness in the wild and expected phenotype-fitness relationships may arise from a complex of alternative mechanisms.