Double contrast barium radiology and upper gastrointestinal endoscopy were compared prospectively on 45 occasions in patients infected with HIV who presented with upper gastrointestinal symptoms. In 40 cases, a definite pathological diagnosis was reached and in four cases no organic basis for symptoms could be found. A correct and complete diagnosis was made on visual endoscopic criteria in 43 cases (95.5%) but in only 14 cases (31.1%) from barium studies alone. Radiology showed no abnormality in 22 cases where pathological changes were documented (oesophageal candidiasis in 21 cases). Where pathological confirmation of diagnosis existed (40 cases), endoscopy (without pathological support) had a sensitivity of 97.5% and a specificity of 100% compared with the sensitivity and specificity of 25 and 100% for barium studies. The difference between the sensitivities of the two methods was highly significant (P less than 0.005). The combination of oral candidiasis and upper gastrointestinal symptoms without dysphagia or weight loss was so strongly associated with uncomplicated oesophageal candidiasis (negative predictive value 93%; P less than 0.025), that this is supported as a basis for therapy without the need for further investigation, although if upper gastrointestinal investigation is required, endoscopy should be the method of choice.