Following the birth of a baby girl confirmed to be homozygous normal for the delta F508 deletion causing cystic fibrosis (CF), many single-gene defects have been diagnosed by polymerase chain reaction (PCR) for preimplantation genetic diagnosis (PGD). A few misdiagnoses have been reported but no large-scale studies have been performed to assess the accuracy of diagnosis in a clinical setting. Here we focus on a series of 15 delta F508 PGD cycles performed at the Hammersmith hospital in an 18 month period. All the spare embryos that had not been selected for transfer after clinical diagnosis were disaggregated and the blastomeres were analysed individually to confirm the clinical results and assess the reliability of single blastomere analysis by the nested PCR method. A total of 484 blastomeres from 112 embryos of different delta F508 genotypes were analysed. The amplification rate for nucleated blastomeres was 95 per cent and the overall accuracy of diagnosis was 89 per cent. Using these figures, we calculate that the chance of selecting an affected embryo instead of a homozygous unaffected or heterozygous carrier is 1.3 per cent, and 0.3 per cent of selecting an affected embryo as unaffected when heterozygotes were not considered for transfer. Misdiagnoses risks were negligible when embryos were considered for transfer after obtaining two concordant results from the same embryo. This study highlights the fact that heterozygous carrier embryos are more often associated with misdiagnoses, due to the failure of amplification of one of the two alleles in heterozygous cells (allele dropout (ADO)) and undetected contamination. In a recessive condition such as CF, ADO cannot result in a serious error. Misdiagnoses due to contamination are potentially more dangerous, they, however, can be limited by only selecting homozygous unaffected embryos for transfer as the risks are quadrupled when heterozygotes are also considered for transfer. For diagnoses of dominant conditions we strongly recommend the systematic analysis of two blastomeres per embryo and the transfer of only embryos with two independent concordant results.