The purpose of this study was to evaluate by CT the usefulness of the non-distension of the stomach in determining invasion of the gastric cancer into perigastric space and adjacent organs. Forty-eight patients with pathologically proved gastric cancer were studied by conventional CT. Patients were examined using two techniques: (a) non-distension of the stomach in the supine position and 34 patients additionally in prone position; and (b) distension of the stomach with water or air in the supine and/or prone position. CT findings by both techniques were separately analysed preoperatively and compared to surgical findings. Invasion of perigastric fat was better demonstrated by the non-distension technique in 15 of the 36 patients with pathologically proven fat infiltration. Non-distension technique was more accurate than distension in detecting: (a) involvement of gastric ligaments (80 versus 67% for gastrohepatic, 85 versus 73% for gastrocolic and 80% by both techniques for gastrosplenic ligament); (b) perigastric lymphadenopathy (86% by both techniques for lymph nodes sited at the gastrohepatic ligament, 85 versus 75% for gastrocolic and 85 versus 80% for gastrosplenic ligament lymph nodes, respectively); and (c) pancreatic invasion (86 versus 80%). The prone position with non-distended stomach was particularly helpful in excluding pancreatic invasion in five patients with carcinoma of the gastric body. The distension technique was more accurate in demonstrating perigastric extension in gastroesophageal junction tumors in two patients. In conclusion, additional CT of the non-distended stomach with the patient in prone position can provide further evidence about infiltration of the perigastric fat, ligaments, lymph nodes and pancreas, in patients with gastric carcinoma, with the exception of gastroesophageal junction tumors.