We explored postmortem ethanol diffusion from the stomach using a human cadaver model with multiple blood site sampling. In all, 400 ml of alcohol solution (5%, 10%, 20%, or 40% methanol and ethanol weight/volume in water) was introduced into the stomach by oesophageal tube. Blood methanol concentrations correlated with ethanol concentrations (methanol range, 1–676 mg%; ethanol range, 1–531 mg%; r = 0.9973). The pattern of ethanol diffusion showed marked between-case variability. Typically, concentrations were highest in pericardial fluid and, in decreasing order, in left pulmonary vein, aorta, left heart, pulmonary artery, superior vena cava, inferior vena cava, right heart, right pulmonary vein, and femoral vein. Diffusional flux was broadly proportional to the concentration of ethanol used. It was time-dependent (as assessed by 24-h and 48-h sampling) and markedly inhibited by refrigeration at 4°C. After gastric instillation of 400 ml of 5% solution for 48 h at room temperature in paired cadavers, ethanol concentrations (mg%) were as follows: pericardial fluid 135, 222; aorta 50, 68; left heart 77, 26; right heart 41, 28; femoral vein 0. Using a 10% solution, ethanol concentrations (mg%) were as follows: pericardial fluid 401, 255; aorta 129, 134; left heart 61, 93; right heart 31, 41; femoral vein 5, 7. Introducing 50 ml of 10% alcohol solution into the oesophagus after oesophagogastric junction ligation produced similar aortic blood ethanol concentrations. This suggests that postmortem gastro-oesophageal reflux and diffusion from the oesophagus is the mechanism of artefactual elevation of aortic blood ethanol. Introducing 150 ml of 10% alcohol solution directly into the lesser sac resulted in marked reduction of diffusion into the pericardial sac. This finding suggests that close apposition of the gastric fundus against the diaphragm and a large gastric volume are significant factors that influence diffusional flux. Deflation and anchoring of the left lung at the apex had little effect, suggesting that diffusion into the left pulmonary vein is via the pericardial fluid rather than the lung substance. In nine fatalities with known alcohol consumption shortly before death, the highest observed stomach contents ethanol concentration was 8.7%. Two cases showed marked variations in blood ethanol concentrations in 10 samples with ranges (mg%) of 97–238 and 278–1395; pericardial fluid 1060 and 686; vitreous humour 34 and 225; and stomach contents 300 ml at 5.5% and 85 ml at 1.9%, respectively. We conclude that postmortem diffusion of ethanol from the stomach into the blood is a potentially significant artefact in a small minority of cases. Both aortic and cardiac chamber blood may be affected by this artefact. Between-case and within-case variability causes unpredictability. For postmortem ethanol analysis, the optimum sampling technique appears to be aspiration by needle puncture of the femoral vein (or external iliac vein) after cross-clamping proximal to the sampling site.