Much is known about the genetic and developmental basis of mutant phenotypes in the laboratory. However, much less is known about the mechanisms that control naturally evolved phenotypes in the wild. Are specific types of genes or mutations preferred? How are developmental programs modified to generate altered phenotypes? Finally, in cases where the same phenotype has evolved in independent lineages (convergent evolution), are the same genes or mutations used?The threespine stickleback fish (Gasterosteus aculeatus) is an emerging model organism in which the approaches and tools of developmental biology, genetics, and genomics can be applied to understand evolutionary changes. Marine sticklebacks have repeatedly colonized and adapted to new freshwater environments throughout the Northern Hemisphere. Freshwater sticklebacks have evolved many changes in their skeleton relative to marine fish. For example, freshwater sticklebacks have evolved changes to their defensive armor in response to decreased predation and changes in bones used for feeding in response to differences in diet. Despite their phenotypic differences, marine and freshwater sticklebacks can be bred in the lab. Crosses between marine and freshwater sticklebacks can be used to decipher the genetic basis of their evolved phenotypic differences.In chapter 2 of this thesis, I present the results of a large quantitative genetic study of stickleback skeletal evolution. I mapped regions of the genome, known as quantitative trait loci (QTL), affecting a wide range of skeletal traits. Over 100 QTL were mapped, and a meta-analysis of their properties was performed. Using these QTL, several general questions in evolutionary genetics were addressed. First, I explored the dominance of the QTL. The dominance of a QTL is a genetic property that might affect likelihood of being used during adaptation. Surprisingly, a large proportion of skeletal QTL had an additive (intermediate) dominance. Second, I investigated the modularity of QTL. To do this, I examined serially repeating skeletal elements such as vertebrae. Most (76%) QTL that influenced such traits had specific, modular effects on just a subset of the possible domains of the phenotype. Finally, I examined whether the QTL clustered in the genome. Many large-effect skeletal QTL were clustered on chromosomes 4, 20, and 21. Inheritance of these clusters of linked adaptive alleles might allow multiple aspects of skeletal morphology to evolve rapidly and simultaneously. In summary, the large set of skeletal QTL was used to explore the genetic properties of evolved mutations. These QTL are a starting point for the identification of the molecular basis of evolved changes in the vertebrate skeleton.In chapter 3 of this thesis, I focus on one skeletal trait, gill raker number. Gill rakers are periodically patterned bones in a fish’s throat important for feeding. In response to a change in diet in freshwater populations, hundreds of freshwater stickleback populations have independently evolved a reduction in gill raker number. This phenomenon is known as gill raker reduction. I found that heritable marine/freshwater differences in gill raker number were present early in development. The gill raker number differences were due to a difference in the spacing between adjacent gill rakers. The marine/freshwater number and spacing differences were present at the earliest detectable point of gill raker specification. This result suggests that gill raker reduction is caused by an early-acting change to a lateral inhibition process controlling raker bud spacing. Next, I performed linkage mapping in F2 fish from crosses with three independently derived freshwater populations. In all three crosses, gill raker QTL mapped to chromosomes 4 and 20, suggesting a similar genetic basis. Finally, the chromosome 4 and 20 QTL affected the early spacing of gill raker buds. Collectively, these data demonstrated that parallel developmental genetic features underlie the convergent evolution of gill raker reduction in freshwater sticklebacks. These results suggest that even highly polygenic adaptive traits can have a predictable developmental genetic basis.In chapter 4 of this thesis, I present an economical method that uses next-generation sequencing to perform genome-wide genotyping of hundreds of multiplexed samples. This method was used to genotype two large marine by freshwater F2 crosses of over 350 fish each. The resulting maps significantly improved the stickleback genome assembly by making over 100 changes to the order and orientation of the scaffolds that compose the genome assembly. In the revised genome assembly, 95% of the assembly was anchored to a chromosome, compared to 87% in the original assembly. To assess linkage map quality, I mapped quantitative trait loci (QTL) controlling lateral plate number. Plate number mapped as expected to a 200 kilobase genomic region containing Ectodysplasin. In addition, plate number mapped to a chromosome 7 QTL overlapping a previously identified modifier QTL. Finally, I examined gill raker length. I found that the two freshwater populations convergently evolved shorter gill rakers relative to marine fish. I mapped eight QTL controlling gill raker length in the two crosses. However, none of the QTL were overlapping between the two crosses. Thus, the convergent evolution gill raker length reduction has a surprisingly non-parallel genetic basis.In chapter 5 of this thesis, I present additional experiments to determine the genetic basis of evolved reduction in gill raker number. Using the genotyping method developed in chapter 4, I mapped QTL controlling raker number and spacing in two large marine by freshwater F2 crosses of over 350 fish each. Overlapping QTL on chromosomes 4, 16, and 20 were found in both crosses. In addition, 14 unique modifier QTL were mapped. Together these results indicate that a combination of similar and different genetic basis underlies the convergent evolution of gill raker number. Using crosses with recombinant chromosomes, I fine-mapped the chromosome 4 and 20 QTL, substantially narrowing the interval sizes. Surprisingly, the chromosome 4 QTL in the two crosses mapped to non-overlapping regions. Therefore, even though the chromosome 4 QTL appeared to map to similar genomic regions in the two crosses, a different genetic basis was actually used. An excellent candidate gene, Fibroblast Growth Factor 20 (Fgf20) was located within one of the fine-mapped chromosome 4 QTL. I mutated the coding region of Fgf20 with a genome-editing technology called Transcription Activator-Like Effector Nucleases (TALENs). Induced loss of function mutations in Fgf20 resulted in a decrease in gill raker number. The Fgf20 mutants had a concomitant increase in gill raker spacing, similar to the evolved freshwater phenotype. Induced mutations in a second gene, Smad5, also resulted in a raker phenotype. However, Smad5 was excluded from the fine-mapped QTL intervals, so it likely does not underlie gill raker evolution. No coding changes in Fgf20 likely cause evolved gill raker reduction. These results suggest a model in which cis-regulatory mutations that affect Fgf20 expression levels or spatial patterns contribute to evolved gill raker reduction.Together, this work makes significant progress towards understanding the genetic and developmental basis of skeletal evolution in stickleback fish. The results have broad implications for understanding the process of how genetic variation contributes to adaptive evolution.