Massive parallel sequencing enables identification of numerous genetic variants in mutant organisms, but determining pathogenicity of any one mutation can be daunting. The most commonly studied preclinical model of retinitis pigmentosa called the "rodless" (rd1) mouse is homozygous for two mutations: a nonsense point mutation (Y347X) and an intronic insertion of a leukemia virus (Xmv-28). Distinguishing which mutation causes retinal degeneration is still under debate nearly a century after the discovery of this model organism. Here, we performed gene editing using the CRISPR/Cas9 system and demonstrated that the Y347X mutation is the causative variant of disease. Genome editing in the first generation produced animals that were mosaic for the corrected allele but still showed neurofunction preservation despite low repair frequencies. Furthermore, second-generation CRISPR-repaired mice showed an even more robust rescue and amelioration of the disease. This predicts excellent outcomes for gene editing in diseased human tissue, as Pde6b, the mutated gene in rd1 mice, has an orthologous intron-exon relationship comparable with the human PDE6B gene. Not only do these findings resolve the debate surrounding the source of neurodegeneration in the rd1 model, but they also provide the first example of homology-directed recombination-mediated gene correction in the visual system.