Six laboratories participated in a study to compare the sensitivity and specificity of RT-PCR tests for the detection of classical swine fever virus (CSFV). Sets of coded samples were prepared by serial dilution of positive samples and then distributed to each of the laboratories. One set comprised 25 samples of random primed cDNA, synthesised from viral RNA representative of different pestiviruses. The other set comprised samples of blood and serum obtained from virus-free or CSFV-infected pigs. Each laboratory tested the samples using PCR/RT-PCR according to a set of standardised protocols that specified the exact conditions and requirements for inclusion of control samples. Two types of test were evaluated. One amplified a part of the 5'-non coding region of the pestivirus genome by means of a closed, one-tube RT-nested PCR. The other amplified a part of the NS5B gene using non-nested RT-PCR. The results of the laboratories were compared with one another, and with those obtained earlier when similar samples were tested by the same laboratories using non-standardised methods [Paton et al., Classical swine fever virus: a ring test to evaluate RT-PCR detection methods, Vet. Microbiol., in press]. Standardisation of the protocols resulted in a more consistent test sensitivity. Three laboratories avoided significant false positive results. Others that did not, could nevertheless recognise that test specificity was inadequate from the results obtained with the control samples. Minimum requirements for the inclusion of adequate controls and periodic proficiency testing are proposed.