The M129V mutation in the human prion protein gene (PRNP) is the primary site linked to susceptibility to prion diseases in humans. The heterozygous state of this allele has been proven to be more resistant to prion-related diseases such as Creutzfeldt-Jakob disease. Based on a study of the common genetic variations, it has been proposed that balancing selection has played a critical role in shaping the distribution of 129V in worldwide populations. Kreitman and Di-Rienzo [Kreitman and Di-Rienzo (2004)Trend Genet 20:300-304] challenged this hypothesis by pointing out that the exclusion of polymorphisms with low frequency may introduce an ascertainment bias and, in turn, lead to a wrong conclusion. By studying sequence variations in samples from the Human Genome Diversity Project, Soldevila et al. [Soldevila et al. (2005) Trends Genet 21:389-391] showed that this ascertainment bias does exist. We argue that the evidence presented by Soldevila et al. may be compromised by the samples tested since they were selected from a population that may be substructured. In this study, we re-evaluated the hypothesis of balancing selection in a natural Chinese population using a much longer segment encompassing the entire genomic region of the PRNP gene (15 kb). We showed that the pattern of genetic variation in PRNP is not consistent with the presence of balancing selection in this gene.