We have investigated HTLV-I and HTLV-II infection in children born to HTLV-I-seropositive or indeterminate Western blot mothers in Martinique by using the polymerase chain reaction (PCR). Only HTLV-I and no HTLV-II-positive samples were found in this study. All the samples from HTLV-I-seropositive children and adults were PCR positive, whereas the four HIV-I-seropositive and Western blot HTLV-I-negative mothers and their eight children were all PCR negative. Therefore, PCR and serology were in complete agreement in these patients. However, two of the six mothers who were first indeterminate by Western blot, and who later became seronegative, were found positive by PCR. Of the 27 children (ages 2-12 years), born to HTLV-I-seropositive and PCR-positive mothers, 2 were seropositive and PCR positive, 5 were seronegative and PCR positive with 2 primer pairs in gag and pol, and 4 were seronegative and PCR positive with only 1 of the primer pairs. In contrast to an initial rate of transmission of 7% estimated by serology we found a rate of transmission of 28 to 41% (whether or not children who were positive with only one of the primer pairs were included). Thus, our study confirms that PCR is useful in detecting HTLV-I infection in children before seroconversion and underlines the potential lack of sensitivity of serology to detect contaminating HTLV-I blood units in endemic areas.