An immunocapture polymerase chain reaction (IC-PCR) protocol and ELISA were compared for their effectiveness in detecting plum pox virus (PPV) in dormant plum material. Although the IC-PCR was about one thousand times more sensitive than ELISA, PPV was detected by ELISA in 71-80% of bark samples collected in December, January and March 1996/97 from pot-grown rootstock trees inoculated with PPV the previous March, compared with 85-86% detection in the same samples by IC-PCR. In similar samples from one-year-old shoots taken from infected branches of orchard trees, 66-81% were positive by ELISA compared with 81-87% by IC-PCR. With bulked samples taken from the fibrous roots of the pot-grown trees, PPV was detected in 92-100% of samples by IC-PCR in winter compared with only 38-65% by ELISA. These results were confirmed in samples from the roots and shoots of the same trees in 1997/98. Three samples per shoot would have been sufficient to detect PPV by ELISA in 87 of the 88 infected shoots tested during the two winters. However, infected shoots are irregularly distributed in diseased trees and PCR assays of root samples offer the potential for improving the reliability of identifying trees infected with PPV.