Establishing metrics of diversification can calibrate the observed scope of diversity within a lineage and the potential for further phenotypic diversification. There are two potential ways to calibrate differences between phenotypes. The first metric is based on the structure of the network of direct and indirect connections between elements, such as the genes, proteins, enzymes and metabolites that underlie a phenotype. The second metric characterizes the dynamic properties that determine the strength of the interactions among elements, and influence which elements are the most likely to interact. Determining how the connectivity and strength of interactions between elements lead to specific phenotypic variations provides insight into the tempo and mode of observed evolutionary changes. In this dissertation, I proposed and tested hypotheses for how the structure and metabolic flux of a biochemical network delineate patterns of phenotypic variation. I first examined the role of structural properties in shaping observed patterns of carotenoid diversification in avian plumage. I found that the diversification of species-specific carotenoid networks was predictable from the connectivity of the underlying metabolic network. The compounds with the most enzymatic reactions, that were part of the greatest number of distinct pathways, were more conserved across species’ networks than compounds associated with the fewest enzymatic reactions. These results established that compounds with the greatest connectivity act as hotspots for the diversification of pathways between species. Next, I investigated how dynamic properties of biochemical networks influence patterns of phenotypic variation in the concentration and occurrence of compounds. Specifically, I examined if the rate of compound production, known as metabolic flux, is coordinated among compounds in relation to their structural properties. I developed predictions for how different distributions of flux could cause distinct diversification patterns in the concentrations and presence of compounds in a biochemical network. I then tested the effect of metabolic network structure on the concentrations of carotenoids in the plumage of male house finches (Haemorhous mexicanus) from the same population. I assessed whether the structure of a network corresponds to a specific distribution of flux among compounds, or if flux is independent of network structure. I found that flux coevolves with network structure; concentrations of metabolically derived compounds depended on the number of reactions per compound. There were strong correlations between compound concentrations within a network structure, and the strengths of these correlations varied among structures. These findings suggest that changes in network structure, and not independent changes in flux, influence local adaptations in the concentrations of compounds. Lastly, the influence of carotenoid network structure in the evolutionary diversification of compounds across species of birds depends on how the structure of the network itself evolves. To test whether the carotenoid metabolic network structure evolves in birds, I examined the patterns of carotenoid co-occurrence across ancestral and extant species. I found that the same groups of compounds are always gained or lost together even as lineages diverge further from each other. These findings establish that the diversification of carotenoids in birds is constrained by the structure of an ancestral network, and does not evolve independently within a lineage. Taken together, the results of this dissertation establish that local adaptations and the evolutionary diversification of carotenoid metabolism are qualitatively predictable from the structure of an ancestral enzymatic network, and this suggests there is significant structural determinism in phenotypic evolution.