Genomic instability in the form of repeat number variation at microsatellite loci has been reported in several human cancers. To resolve current confusion regarding frequency of microsatellite instability (MSI), standardised protocols have proposed use of a consensus set of informative loci; it is claimed that analysis of 2 apparently quasi-homozygous, quasi-monomorphic, mononucleotide repeats (BAT-25 and BAT-26) is sufficiently accurate to define MSI (obviating need for corresponding constitutive DNA). We examined these loci in 163 breast cancers, 108 of which had previously been analysed at 11-24 other loci and found to have MSI in 38% of them. For BAT-26 only 1/153 (homozygous) tumours showed a contracted allele, with minor size variations (1-6 bp) between individuals. For BAT-25 repeat contractions were unambiguously observed in 12 (7.4%) tumours; only 4 of these were previously designated MSI+. DNA from normal individuals showed significant allelic variation in 8/159 (5%) cases for BAT-25; no instance of heterozygosity was seen for BAT-26. Subsequently, we analysed normal DNA from the 12 individuals whose tumours had MSI at BAT-25; in 2 cases there was germline heterozygosity. We conclude that analysis with BAT-26 (in contrast to other loci) was not a useful detector of MSI. With BAT-25, a low frequency of MSI not much greater than germline polymorphism, also limits the utility of this marker for determining MSI in breast cancer.