A proportion of gastric adenocarcinomas exhibit replication errors manifested as microsatellite instability. The clinicopathological and prognostic significance of this abnormality remains uncertain. This study aimed to determine the importance of microsatellite instability by analysing a large series of gastric carcinomas from an English population. Using a novel fluorescent polymerase chain reaction technique, we amplified 11 microsatellite sequences from paired normal and carcinoma DNA from 101 patients who underwent a potentially curative resection for gastric carcinoma. Overall, 21% of cases demonstrated microsatellite instability in at least one locus. At least four loci were examined in each case. A replication error positive phenotype (minimum of 29% of loci affected) was detected in 9% of cases. There was no statistically significant association between the presence of microsatellite instability or replication error positive phenotype and the patient's age, sex, tumour site, stage, node status, histological subtype or grade. Carcinomas confined to the mucosa or submucosa (T1) showed a significantly higher frequency of instability and replication error positive phenotypes than T3 lesions (P = 0.03 and P = 0.05, respectively). A larger proportion of patients who were microsatellite instability or replication error positive were alive at 5 years compared with those who were negative but this did not reach statistical significance (P = 0.15 and P = 0.16, respectively). We identified a subset of gastric carcinomas from a relatively low-risk population which showed evidence of microsatellite instability. There were no statistically significant 5-year survival advantages in cases demonstrating microsatellite instability or replication error positive phenotypes. The detection of microsatellite instability is of limited prognostic value in gastric carcinoma.