The pattern of X-chromosome inactivation in females is currently evaluated by assays of differential methylation in the genes between the active and the inactive X chromosomes, with methylation-sensitive enzymes. We report a new assay in the human androgen receptor (HUMARA) locus involving a methylation-specific polymerase chain reaction (M-PCR) technique, independent of the use of restriction enzymes. The assay involves the chemical modification of DNA with sodium bisulfite and subsequent PCR. By using the assay with specific primers for the methylated allele, we obtained an X-inactivation pattern based on the ratio of the maternal inactive X to the paternal inactive X. These patterns were consistent with those obtained by conventional PCR assay at the same locus in 48 female cases. We also obtained another X-inactivation pattern based on the ratio of the maternal active X to the paternal active X by using specific primers for the unmethylated allele. The latter pattern was complementary to the former pattern, and a combination of these patterns produced a reliable X-inactivation pattern. The assay revealed that 12 (11%) of the 105 normal females had non-random inactivation patterns (>80:20 or <20:80). Four patients with an X; autosome translocation showed extremely non-random patterns, and these results were consistent with those obtained by previous molecular/cytogenetic studies. We conclude that M-PCR provides an accurate assay for X-inactivation and that it can be performed on various DNA samples unsuitable for restriction digestion.