A new method for measuring the rate of in-vitro starch digestion in products with a structure 'as eaten' is introduced. An equivalent amount of potentially available starch from each product was chewed by subjects, expectorated into a beaker and incubated with pepsin. The incubate was thereafter transferred to a dialysis tubing and incubated with pancreatic alpha-amylase for 3 h. Samples were removed from the dialysate at time intervals and the degree of hydrolysis was calculated as the proportion of the potentially available starch degraded to maltose. A hydrolysis index (HI) was calculated as the area under the hydrolysis curve with the product as a percentage of the corresponding area with white wheat bread. The method was applied to 21 cereal and legume products, chosen to cover as wide a range as possible with respect to metabolic response, and to include several of the proposed mechanisms to differences in metabolic behaviour of starch. The accuracy of the in-vitro method was evaluated versus the metabolic responses obtained with the same products in healthy subjects. A significant correlation between HI and glycaemic index (GI) was obtained in cereal as well as in legume products. A significant correlation was also obtained between HI and insulin index (II) with pooled data from all products. However, in the case of II no correlation was obtained with the legume products only. It is concluded that the presently described in-vitro procedure offers a good potential to predict the metabolic behaviour of starchy foods.