During 1989, 316 members of a cohort of homosexual men were tested for HIV-specific DNA by the polymerase chain reaction (PCR) using a pair of gag-region primers. Of 125 HIV-seronegative subjects, 123 (98.4%) were PCR-negative while 158 (82.7%) of 191 HIV-seropositive subjects were PCR-positive. Fewer of the 33 subjects who were seropositive and PCR-negative were at Centers for Disease Control (CDC) stage IV than the seropositive, PCR-positive subjects (6 versus 25%;P= 0.030). The seropositive, PCR-negative group had higher median CD4 counts (640 versus 490 ± 106cells/I;P= 0.006), higher CD4: CD8 ratios (0.92 versus 0.64;P= 0.004), lower immunoglobulin (Ig) G levels (1290 versus 1645 mg/dl; P = 0.002), lower IgA levels (168 versus 251 mg/dl;P< 0.001), and lower C1q binding activity (8 versus 14%; P = 0.010) than the seropositive, PCR-positive subjects. The median rate of CD4 cell decline in the 3 years preceding the PCR sample was less marked in the seropositive, PCR-negative group than the seropositive, PCR-positive group (—58 versus — 77 ± 106cells/I per year; P = 0.028). To control for duration of infection, we restricted the analysis to the subgroups of 11 seropositive, PCR-negative subjects and 34 seropositive, PCR-positive subjects who had seroconverted earlier in the cohort study. Both subgroups had similar durations of infection, yet the same pattern of differences persisted. Moreover, despite a median of 51 months since seroconversion, the seropositive, PCR-negative subgroup had CD4 counts and antecedent rates of CD4 cell decline that were similar to those of the 123 seronegative, PCR-negative homosexual controls who were not infected with HIV. We conclude that PCR can identify a group of HIV-infected people with very low levels of proviral DNA who demonstrate slower progression of the effects of HIV.