To gain insight into the evolution of xanthophyll synthesis in Arabidopsis thaliana, we analyzed two pairs of duplicated carotenoid hydroxylase enzymes in Arabidopsis thaliana: the cytochrome P450 enzymes CYP97A3 and CYP97C1, and non-heme di-iron enzymes, BCH1 and BCH2. Hydroxylated carotenes did not accumulate in a quadruple mutant for these four genes, demonstrating that they encode the full complement of carotenoid hydroxylases in A. thaliana. We were thus able to infer definitively the activity of each enzyme in vivo based on the phenotypes of selected double and triple mutant genotypes. The CYP97 and BCH gene pairs are primarily responsible for hydroxylation of alpha- and beta-carotenes, respectively, but exhibit some overlapping activities, most notably in hydroxylation of the beta-ring of alpha-carotene. Surprisingly, triple mutants containing only CYP97C1 or CYP97A3 activity produced 74 and 6% of the wild-type lutein level, indicating that CYP97C1 can efficiently hydroxylate both the beta- and epsilon-rings of alpha-carotene and that CYP97A3 also has low activity toward the epsilon-ring of alpha-carotene. The modes of functional divergence for the gene pairs appear distinct, with the CYP97 duplicates being strongly co-expressed but encoding enzymes with different in vivo substrates, while the BCH duplicates encode isozymes that show significant expression divergence in reproductive organs. By integrating the evolutionary history and substrate specificities of each extant enzyme with the phenotypic responses of various mutant genotypes to high light stress, we propose two likely scenarios for the evolution of alpha-xanthophyll biosynthesis in plants from ancestral organisms.