Variation in depth of radionuclide within the stomach may result in a significant error in the measurement of gastric emptying if no attempt is made to correct for gamma-ray attenuation by the patient's tissues. In this study a method of attenuation correction, which uses a single posteriorly located scintillation camera and correction factors derived from a lateral image of the stomach, was compared with a two-camera geometric mean method, both in phantom studies and in five volunteer subjects. A meal of 100 g of ground beef containing 99Tcm-chicken liver, and 150 ml of water, was used in the in vivo studies. In all subjects the geometric mean data showed that solid food emptied in two phases: an initial lag period, followed by a linear emptying phase. Using the geometric mean data as a standard, the anterior camera overestimated the 50% emptying time (T50) by an average of 15% (range 5-18) and the posterior camera underestimated this parameter by 15% (4-22). The posterior data, corrected for attenuation using the lateral image method, underestimated the T50 by 2% (-7 to +7). The difference in the distances of the proximal and distal stomach from the posterior detector was large in all subjects (mean 5.7 cm, range 3.9-7.4). We conclude that attenuation effects may account for large errors in the measurement of gastric emptying with radionuclide methods and that the application of correction factors derived from a lateral image of the stomach reduces these errors.