We attempted to measure cardiopulmonary effects, CD4 counts, and perceived sense of well-being in 25 individuals moderately to severely immunocompromised from HIV infection (mean entry CD4 count = 144.microliters-1) before and after a 24-wk program of exercise training. Only six subjects completed the 24-wk program. All six showed evidence of a training effect. Statistically significant improvements were seen in maximal oxygen consumption (VO2max), oxygen pulse, and minute ventilation. Submaximal exercise performance improved significantly by 12 wk in the 10 individuals available for testing: decreases were seen in heart rate, rate pressure product, and rate of perceived exertion. White blood cell counts and T-lymphocyte subsets were stable at 12 and 24 wk in the subjects available for testing. High depression/anxiety scores on a mental health inventory (General Health Questionnaire) correlated with low CD4 counts. Scores did not correlate with compliance with the exercise program. There was a trend (P < 0.10) for scores to improve over time among those individuals who attended > or = 80% of scheduled exercise sessions. We conclude that exercise training is feasible and beneficial for some HIV-infected individuals.