The mean of individual coefficients of variation of acid output after modified sham feeding was 39% in 22 patients operated on with parietal cell vagotomy. The reproducibility of the interpretation of the sham feeding test as 'positive' or 'negative' was good. An intragastric infusion of a marker to correct for pyloric loss did not increase the accuracy of the test. The prognostic value of the qualitative estimation of the sham feeding test 2 months after operation to predict recurrent ulcer after parietal cell vagotomy was poor in 39 patients studied prospectively over 3 years. With the criterion sham feeding minus basal acid output over 1.0 mmol/30 min as a positive test, 63% of patients with a positive and 24% with a negative test later had recurrent ulcers. The consistency of the interpretation as either positive or negative was low in annual tests during the 3 years of follow-up study. After parietal cell vagotomy the sham-feeding-stimulated acid output was higher in patients with duodenal than in those with prepyloric recurrent ulcers and also in those without recurrences. This indicates that the amount of vagal innervation left after parietal cell vagotomy is of special importance in the occurrence of duodenal ulcer relapse.