We screened 596 cord-blood samples by nested short PCRs for the gag and pX regions of HTLV-1 which are capable of detecting a single copy. The samples were derived from 449 and 147 babies born to seropositive and seronegative mothers respectively. Of these, 20 samples were positive in at least one of the PCRs: 9 (45%) were positive in both PCRs, but 10 and 1 samples in either the pX or the gag PCR respectively. These samples were tested further in nested long PCRs directed for gag-pX, gag-pol and pol-pX regions capable of detecting 8, 1 and 2 copies respectively. Of 9 dually positive samples, 7 (77%) showed the predicted 6.2-kbp band in the gag-pX PCR; only 2 of them had the predicted band alone; 7 samples had discrete bands shorter than the predicted size. In the gag-pol PCR, all 9 samples showed the predicted 2.2-kbp band alone. In the pol-pX PCR, 819 samples showed the predicted 4.2-kbp band, including one with an additional 2.1-kbp band, and the last a 1.0-kbp band alone. Thus, all of the dually positive samples had proviruses harboring gag, pol and pX priming sites. In contrast, none of the 11 singly positive samples showed the predicted band in the gag-pX PCR: 5 had no visible band, and the other 6 had shorter bands only. None of these 11 samples showed any positive signal in either gag-pol or pol-pX PCR. Our results suggest that HTLV-1 proviruses in the cord blood are frequently defective.