Whether commercial, bean-derived alpha-amylase inhibitor preparations failed to decrease starch digestion in humans because of insufficient antiamylase activity, destruction by gastrointestinal secretions, or decreased activity in the presence of starch is unknown. We used a simple partial purification procedure to markedly concentrate the inhibitor (sixfold to eightfold by total protein content, and 30-40-fold by dry weight). Compared with a commercial preparation and crude bean extract, this partially purified inhibitor inactivated intraduodenal, intraileal, and salivary amylase in vitro faster and more completely (p less than 0.001); its specific activity was not affected by exposure to gastric juice and was only minimally reduced by duodenal juice. Whereas the rate of amylase inhibition by inhibitor was markedly slowed in the presence of nondietary liquid starch, dietary solid starch had only a minimal effect. Consequently, the partially purified inhibitor had no effect on liquid starch digestion, but decreased in vitro digestion of dietary starch in a dose-dependent manner (p less than 0.001). Perfusion of the partially purified inhibitor (2.0, 3.5, or 5.0 mg/ml at 5 ml/min) into the duodenum of humans rapidly inhibited greater than 94%, greater than 99%, or greater than 99.9% of intraluminal amylase activity. We conclude that commercial amylase inhibitors failed to decrease starch digestion in vivo mainly because they have insufficient antiamylase activity. However, a partially purified inhibitor with increased specific activity is stable in human gastrointestinal secretions, slows dietary starch digestion in vitro, rapidly inactivates amylase in the human intestinal lumen, and, at acceptable oral doses, may decrease intraluminal digestion of starch in humans. Such an inhibitor therefore deserves study.