Single base substitutions in DNA mismatch repair genes which are predicted to lead either to missense or silent mutations, or to intronic variants outside the highly conserved splicing region are often found in hereditary nonpolyposis colorectal cancer (HNPCC) families. In order to use the variants for predictive testing in persons at risk, their pathogenicity has to be evaluated. There is growing evidence that some substitutions have a detrimental influence on splicing. We examined 19 unclassified variants (UVs) detected in MSH2 or MLH1 genes in patients suspected of HNPCC for expression at RNA level. We demonstrate that 10 of the 19 UVs analyzed affect splicing. For example, the substitution MLH1,c.2103G > C in the last position of exon 18 does not result in a missense mutation as theoretically predicted (p.Gln701His), but leads to a complete loss of exon 18. The substitution MLH1,c.1038G > C (predicted effect p.Gln346His) leads to complete inactivation of the mutant allele by skipping of exons 10 and 11, and by activation of a cryptic intronic splice site. Similarly, the intronic variant MLH1,c.306+2dupT results in loss of exon 3 and a frameshift mutation due to a new splice donor site 5 bp upstream. Furthermore, we confirmed complete exon skipping for the mutations MLH1,c.1731G > A and MLH1,c.677G > A. Partial exon skipping was demonstrated for the mutations MSH2,c.1275A > G, MLH1,c.588+5G > A, MLH1,c.790+4A > G and MLH1,c.1984A > C. In contrast, five missense mutations (MSH2,c.4G > A, MSH2,c.2123T > A, MLH1,c.464T > G, MLH1,c.875T > C and MLH1,c.2210A > T) were found in similar proportions in the mRNA as in the genomic DNA. We conclude that the mRNA examination should precede functional tests at protein level.