range 7·7 to 3·7]) was reversed to net water absorption by pre-treatment with granisetron (1·3 [ 2·0 to 5·0]; p<0·004 Mann Whitney) (figure). Net electrolyte movement paralleled net water movement. This represents the prevention by granisetron of a median net shift in water movement of 6·7 mL cm h (about 3·3 L/h of secretion when extrapolated to the whole small intestine) and is similar to the degree of reversal of secretion seen when perfusing with a glucose-electrolyte oral rehydration solution. This study provides supporting evidence for the role of 5-HT in cholera and, since 5-HT3 receptors are neuronally located, confirms the significance of neuronal mechanisms in intestinal secretion. Eherer et al studied the effect of the 5-HT3 antagonists, tropisetron and ondansetron, and tropisetron in combination with the 5-HT2 antagonist, ketanserin, on cholera toxin-induced secretion in a similar human perfusion model but found no reduction in secretion; on the contrary, tropisetron with ketanserin and ondansetron alone appeared to enhance secretion. Differing receptor binding profiles and potencies of the various 5-HT3 receptor antagonists may account for these findings. We do not consider that changes in jejunal motility would account for our findings. Granisetron is not thought to alter jejunal motility and this is confirmed by our finding that the proximal and distal port [C]-PEG recoveries (mean of 80% [SD 15·1] and 59% [SD 23·6] respectively) were similar to those in previous studies. Furthermore, the 30 cm test segment and perfusion rate of 15 mL/min, set to mimic the pathophysiological secretory state, minimise the reliance upon peristalsis for transit. Oral rehydration solutions remain the cornerstone of treatment for cholera and other acute diarrhoeal states. Nonetheless, highly selective agents such as granisetron, targeted at the pathophysiological processes which drive secretion, are likely to emerge as potent anti-secretory treatments.