Detection on automated DNA sequencers of polymerase chain reaction (PCR) products of tetra- and penta-nucleotide short tandem repeat (STR) loci frequently reveals one or more extra peaks along with the true, major allele peak. The most frequent extra peak pattern is a single smaller peak which is one repeat unit shorter than the true allele peak. The existence of such artificial peaks is of special importance when the methods are used for forensic investigations because the artificial extra peaks may simulate true alleles when samples containing mixtures of DNA from different individuals are analyzed. We have investigated the relative levels of formation of extra peaks in 14 STR marker systems. We found that not only the parameters of the PCR but also factors determining the stringency during the post-PCR and pre-electrophoresis handling of samples were of importance for the formation of extra peaks. In our hands, the amounts of extra peaks were reduced (i) if the samples were effectively denatured immediately before loading, (ii) if they contained substantial amounts of formamide (i.e. > or = 50%), and (iii) if the temperature of the electrophoresis gel was above a certain level (i.e. > or = 43 degrees C). The results suggest that extra peaks may in part be due to re-annealing of the PCR product under suboptimal conditions. When efforts had been made to reduce the post-PCR formation of extra peaks, the relative peak areas of the extra peaks ranged from 1% to 17% of those of the true alleles. Similar results were obtained when the PCR products were analyzed under native conditions. Low-copy genome analysis excluded that somatic heterogeneity of the STR regions caused the extra peaks. The systems HumVWA31A, HumFibra/FGA, and D21S11 were especially affected by low-stringency conditions, while Hum-TH01, HumCD4, and D12S391 were virtually unaffected by low-stringency conditions. Replacement of the Taq DNA polymerase with DNA polymerases with lower processivity resulted in higher levels of extra peaks. Our results support the hypothesis that extra peaks are produced due to slipped-strand mispairing.