OBJECTIVE To investigate whether high levels of depressive symptomatology at baseline predict more rapid decline of CD4 lymphocyte counts and progression of clinical disease in persons infected with the human immunodeficiency virus (HIV). METHODS Prospective cohort study with semiannual data collection waves and up to 66 months of follow-up. METHODS Population-based probability sample of single men in areas of San Francisco with high case rates of the acquired immunodeficiency syndrome (AIDS). METHODS All 330 homosexual or bisexual men who by January 1985 had serological evidence of HIV infection but had not had an AIDS diagnosis. Analysis of CD4 lymphocyte change was performed for 277 subjects (83.9%) who had three or more CD4 lymphocyte counts recorded during the study period January 1985 through July 1990. METHODS Depressive symptoms were assessed using the Center for Epidemiologic Studies-Depression scale (CES-D). All subjects were classified according to two indicators of depression: (1) as overall depressed using a cut point of 16 or higher on the complete CES-D, and (2) as affectively depressed using a cut point of more than 1 SD above the mean on a subscale of the CES-D measuring affective depression. Laboratory and symptom measures, antiretroviral use, demographics, and behavioral measures were also used. The primary outcome measure was the rate of change of the CD4 lymphocyte count. Secondary outcomes were AIDS-free survival and mortality. RESULTS At baseline 65 subjects (19.7%) were classified as depressed on the overall scale and 53 (16.1%) were classified as depressed on the affective scale. The unadjusted mean rate of CD4 change was 38% greater for overall depressed subjects than for the overall nondepressed (-0.0812 vs -0.0588 x 10(9)/L [-81.2 vs -58.8/microL per year; P = .07) and 34% greater for affectively depressed subjects than for the affectively nondepressed (-0.0804 vs -0.0598 x 10(9)/L per year; P = .06). In hierarchical multivariate analysis controlling for antiretroviral use, symptoms, and other predictors, baseline overall depression was associated with an excess decline in CD4 count of -0.0285 x 10(9)/L per year (95% confidence interval, -0.0496 to -0.0073), and baseline affective depression was associated with an excess decline in CD4 count of -0.0236 x 10(9)/L per year (95% confidence interval, -0.0464 to -0.0008). Neither overall depression nor affective depression was significantly associated with earlier AIDS diagnosis or earlier mortality. CONCLUSIONS Overall depression and affective depression predicted a more rapid decline in CD4 lymphocyte counts; this association was not attributable to baseline physiological differences. While the mechanism of the association remains unknown and cannot be addressed directly by this study, the data suggest that it can be explained neither as simply a reflection of perceived somatic symptoms nor as the result of differences in recreational drug and alcohol use. Further study is necessary to determine whether treating depression can alter the course of HIV infection.