The aim of this study was to investigate the accuracy of 17 forensic odontologists identifying individuals from two sets of radiographs, one regarded as ante- and the other as postmortem. Each case was observed twice and only one pair out of 31 did not match. The observers were asked to comment about each case, classifying it as easy, moderate or difficult. The results show that one observer was totally correct in the first analysis while four observers made no errors the second time. In the first evaluation 14 observers made between one and seven errors and two observers made 11 errors each. In the second evaluation 12 observers made between one and seven errors and one observer made 13 errors. At the first evaluation, the observers judged 18 of the cases as easy, eight as medium and five as difficult. At the second evaluation, the observers pronounced 13 of the cases as easy, 13 as medium and five as difficult. The corresponding values for the authors were 6, 12 and 13. Most of the mistakes were made on the cases with no restorations and the incorrect answers were found mostly among the difficult cases. In practical forensic work however additional dental chart information is usually available to the forensic odontologist.