The determination of the length of polymerase chain reaction (PCR)-amplified short tandem repeats (STRs) by denaturing capillary electrophoresis (CE) is a standard procedure for purposes of genotyping. We show that dye-specific mobility anomalies exist for 5'-fluorophor-labelled single-stranded DNA (ssDNA) fragments in CE using the performance-optimized polymer 4 (POP4) buffer sieving matrix, containing the entangled poly(N,N-dimethylacrylamide) polymer, urea, and 2-pyrrolidinone. The dye-specific retardation effects relative to coseparated GeneScan-500 [TAMRA] standard fragments can lead to wrong genotyping, even for allele-specific fragments of pentanucleotide STRs, when comparing the relative calculated sizes of identical fragments, labelled with rhodamine (ROX, TAMRA) or fluorescein dyes (FAM, 6-FAM, HEX, JOE, NED, TET): The size of fluorescein dye-labelled fragments of appr. 100 b in length appears to be smaller by up to 6.5 b. This effect becomes more dramatic with decreasing size: a 6-FAM-labelled 24-mer oligonucleotide appeared to be smaller by 11 b. In contrast, in classical urea/polyacrylamide slab-gel electrophoresis only a small dye-specific retardation of identical fragments is observed. The dye-specific effects are superimposed by weaker size and sequence-dependent anomalies of fragment mobility. Therefore, in denaturing CE the coseparation of a defined allele ladder labelled with the same dye as the unknown sample fragments remains the method of choice for accurate genotyping.