Plant pathology developed as a field after the nineteenth century potato famine that led to more than one million deaths across Europe. One important component of plant defense is the plant resistance (R) gene family, which includes an R-gene that recognizes the causal pathogen of the potato famine. R-genes are a vital component of the plant’s secondary defense response, and function by recognizing specific pathogen-released Avirulence genes or cellular changes effected by these genes, then inducing a strong defense response. R-gene evolution is thought to be driven both by selection on pathogen recognition functions and fitness costs of resistance. As a result, R-genes offer some of the most spectacular examples of polymorphism in plant genomes, suggesting selection may drive multiple aspects of R-gene variation. As my dissertation research, I investigated three facets of R-gene polymorphism in Arabidopsis thaliana. First, I tested for evidence of adaptive variation in R-gene expression (MacQueen and Bergelson, in press). R-genes exhibit unusually extensive variation in basal expression levels when compared across accessions collected from different environments. I characterized R-gene expression variation to explore whether R-gene expression was up-regulated in environments favoring pathogen proliferation and down-regulated when risks of infection are low; down-regulation would follow if costs of R-gene expression negatively impact plant fitness in the absence of disease. Surprisingly, almost every change in the environment led to an increase in R-gene expression, a response that was distinct from the average transcriptome response and from that of other stress response genes. These changes in expression were functional in that environmental change prior to infection affected levels of specific disease resistance to isolates of Pseudomonas syringae. In addition, there were strong latitudinal clines in basal R-gene expression and clines in R-gene expression plasticity with drought and high temperatures. These results suggest that variation in R-gene expression across environments may be shaped by natural selection to reduce fitness costs of R-gene expression in permissive or predictable environments. Second, I measured fitness costs of resistance for the R-gene Rps2, which is under balancing selection for expressed resistant and susceptible clades of alleles. To do this, I conducted a large field fitness trial using 22 isogenic lines differing only in the allele of the R-gene Rps2 that they carried. I found that resistant alleles of Rps2 do not carry high fitness costs relative to susceptible alleles in the absence of disease. Instead, all alleles confer fitness benefits relative to an artificially constructed null, due to the additional role of RPS2 as a negative regulator of defense. Variation in the presence or absence of costs suggests an interplay between costs and the genetic architecture of resistance. Third, I investigated the effects of an ancient duplication of the R-gene Rpp8 on polymorphism at that locus. I examined patterns of synonymous and nonsynonymous sequence polymorphism, haplotype diversity, and linkage disequilibrium within and between homologs of Rpp8 in 28 accessions of A. thaliana, and modeled the effects of duplicate distance, intergenic gene conversion rate, and number of duplicates on the patterns of polymorphism expected at this locus. The data were consistent with selection on Rpp8 copy number and genetic architecture to generate and share an excess of novel genotypes. In combination, the three projects above helped elucidate the forces that select on R-gene genomic architecture and levels of expression.