Hyperpolymorphic short tandem repetitive DNA sequences, STRs or microsatellites, have become widely used in human identification, particularly in criminal cases and in mass disasters. In such cases the substrates for the analyses may be decomposed biological material, a fact that has to be taken into account when choosing the appropriate casework methods. In this paper we report the evaluation of five different DNA extraction methods, namely the phenol-chloroform, the silica based, the InstaGene Matrix (BioTest), the glass fiber filter, and the Chelex based methods. The substrates for the analyses are decomposed human liver tissue specimens from forensic autopsy cases. Extracted DNA was quantified and DNA profiled by a set of seven STRs. We have compared laboratory time consumption and costs of the five methods, showing that the Chelex method is the more rapid and less expensive of the methods, the phenol-chlorophorm and silica extractions being the most time consuming and resource demanding ones. A full profile was obtained by the silica method in nine out of ten cases and this method failed to give a reliable type in four out of 70 STR analyses. The phenol-chlorophorm and the glass fiber filter methods failed in 16 analyses, the InstaGene Matrix (BioTest) in 25 and the Chelex extracts in 56 of the 70 STR analyses. By multiple logistic regression we show that the difference between the silica procedure and the other methods are statistically significant. In our hands, the silica gel extraction procedure is an obvious choice when the biological material available is decomposed human tissue--even if this procedure is one of the more laborious ones.