The subjects in the study will be 144 men and women with HIV infection who report experiencing 3 or more episodes of diarrhea (non-pathogen related) per 24 hour period for 3 weeks or more. Subjects will be randomized to one of four experimental intervention conditions: Condition 1 subjects receive true acupuncture and true moxibustion; Condition 2 subjects receive true acupuncture and placebo moxibustion; Condition 3, subjects receive true moxibustion and sham acupuncture; Condition 4( Control Group), subjects receive sham acupuncture and placebo moxibustion. Subjects in Conditions 1,2,3,& 4 will attend 20 scheduled sessions over 24 weeks. Week 1 is a baseline session followed by two sessions per week for weeks 2-8 (sessions 2-15), one session per week for weeks 9, 10, 11 and 12 (sessions 16, 17, 18 & 19) and a final follow-up session at week 24. All subjects will complete daily bowel movement and medication data collection diaries for the duration of the study. Measurement of quality of life and level of functioning will be taken pre-intervention (session 1), week 6 (session 10), week 12 (session 19) and week 24 (session 20). All interventions will be implemented by licensed acupuncturists trained in traditional Chinese medicine. This study is designed to assess the efficacy of two alternative medicine treatments for chronic diarrhea associated with HIV in a prospective, randomized, controlled, blinded, parallel groups study under the intent-to-treat principle. True acupuncture, moxibustion, and combination therapy, in which specific meridian points are stimulated according to protocol, will be compared to each other and with the control group, with one-way ANOVA models for pre-treatment minus post-treatment difference scores for diarrhea frequency and stool consistency as the dependent measures and treatment group assignment (Conditions 1 - 4) as the independent variable. Average pretreatment diarrhea frequency and stool consistency scores will be entered as covariates into these models. Sample size determination for the above analysis, based on preliminary data, with 80% power and a two-tailed type I error rate of .05% by the method of Holm (1979) and a 20% attrition factor indicates the need for 36 subjects assigned to each condition to detect a 0.95-standardized difference between the most extreme experimental conditions.