Many genomic methodologies rely on the presence and extent of linkage disequilibrium (LD) between markers and genetic variants underlying traits of interest, but the extent of LD in the horse has yet to be comprehensively characterized. In this study, we evaluate the extent and decay of LD in a sample of 817 Thoroughbreds. Horses were genotyped for over 50 000 single nucleotide polymorphism (SNP) markers across the genome, with 34 848 autosomal SNPs used in the final analysis. Linkage disequilibrium, as measured by the squared correlation coefficient (r2), was found to be relatively high between closely linked markers (>0.6 at 5 kb) and to extend over long distances, with average r2 maintained above non-syntenic levels for single nucleotide polymorphisms (SNPs) up to 20 Mb apart. Using formulae which relate expected LD to effective population size (Ne), and assuming a constant actual population size, Ne was estimated to be 100 in our population. Values of historical Ne, calculated assuming linear population growth, suggested a decrease in Ne since the distant past, reaching a minimum twenty generations ago, followed by a subsequent increase until the present time. The qualitative trends observed in Ne can be rationalized by current knowledge of the history of the Thoroughbred breed, and inbreeding statistics obtained from published pedigree analyses are in agreement with observed values of Ne. Given the high LD observed and the small estimated Ne, genomic methodologies such as genomic selection could feasibly be applied to this population using the existing SNP marker set.