We present a simple two-stage procedure to define haplotype blocks and construct a statistic to test whether a polymorphism belongs to a block. Applying this method to the data of Gabriel et al. [2002] yielded longer haplotype blocks than were originally reported with a similar average percentage of common haplotypes in blocks. Furthermore, across regions of the genome and among the four populations that were studied, we found that linkage disequilibrium between a given single nucleotide polymorphism (SNP) and the haplotype block was a monotonic function of distance. This correlation was essentially independent of the minor allele frequency of the putative causal SNP when it fell outside of the block, however it was strongly dependent on the minor allele frequency when the SNP was internal to the block. These results have direct application to the design of candidate gene or region-wide association studies.