Double contrast barium enema (DCBE) and colonoscopy were prospectively compared with rigid sigmoidoscopy, rectal biopsy and microbiological examination in the analysis of stool specimens in 58 HIV-1-infected patients with diarrhoea (more than three liquid motions/day for > 1 month). In 26 patients no cause for the diarrhoea was found. In 17 patients the cause of diarrhoea was microbiological, and in 19 rectal histology provided a specific diagnosis. In all these patients sigmoidoscopic appearances were abnormal except in those with Cryptosporidium alone. Colonoscopy provided additional information in only one individual, with cytomegalovirus ulcers of the transverse colon. DCBE was abnormal in only seven cases (cytomegalovirus in three, Kaposi''s sarcoma in two, Giardia lamblia in two) and in no case provided additional information. A combination of stool microbiology and rectal histology gave a sensitivity of 97% with a positive predictive value of 100%. The sensitivities of DCBE and colonoscopy with were low (16 and 62%, respectively) although the specificity for each test was high, with high positive predictive values. We conclude that neither barium enema nor colonoscopy add usefully to rigid sigmoidoscopic biopsy and stool microscopy in HIV-positive patients with diarrhoea.