AIMS--To determine whether neural invasion in advanced gastric cancer is of clinicopathological significance. METHODS--The study population comprised 121 cases of primary advanced gastric carcinoma. Two paraffin wax embedded blocks taken from the central tissue slice in each primary tumour were used. For definitive recognition of neural invasion, immunostaining for S-100 protein was applied to one slide; the other slide was stained with haematoxylin and eosin. RESULTS--Neural invasion was recognised in 34 of 121 (28%) primary gastric carcinomas. There were significant differences in tumour size, depth of tumour invasion, stage, and curability between patients with and without neural invasion. The five year survival rates of patients with and without neural invasion were 10 and 50%, respectively. Multivariate analysis, however, demonstrated that neural invasion was not an independent prognostic factor. CONCLUSIONS--Neural invasion could be an additional useful factor for providing information about the malignant potential of gastric carcinoma. This may be analogous to vessel permeation which is thought to be important, but is not an independent prognostic factor.