After a maximum of 11 years (median 8·3 years) from the time of HIV seroconversion. 25 out of 59 (42%) of CMV-seropositive haemophiliacs had progressed to AIDS, as opposed to eight out of 50 (16%) CMV scroncgatives. The age-adjusted relative risk for AIDS among CMV scropositives was 2·4 (P= 003). In order to determine how this adverse effect is mediated, the mean rate of decline in serial CD4+ lymphocyte counts was studied. CD4+ lymphocyte counts tended to decline more rapidly in CMV seropositives than in seronegatives (−0·087 ± 109/l per annum versus-0·082 ± 109/l per annum), but this difference did not reach statistical significance. The average CD4+ lymphocyte count at the lime of HIV seroconversion was estimated to be similar in CMV seropositives and negatives, because in HIV-l-ncgative haemophiliacs the CD4+ counts were virtually identical, after adjustment for age (0·94 ± 10+/l and 0·97 ± 109/l. respectively). The median CD4+ cell count at which AIDS developed was higher in the CMV-seropositive group (0·07 ± 109/l) than in the seronegative group (0·04 ± 109/l). but this difference did not reach statistical significance. We conclude from these findings that the adverse effect of CMV is not wholly mediated via a more rapid loss of CD4+ cells. We discuss other processes that may be mediated by CMV, such as a functional deficiency of residual CD4+ cells, or dissemination of HIV in other organs, which may be important in determining the earlier onset of AIDS among CMV-seropositive subjects.