The objective was to assess the ability of the in situ mobile nylon bag method for predicting small intestinal and total tract starch digestibility. Starch disappearance was measured for 18 samples of different cereals and legumes subjected to different physical and chemical processing methods and compared with coherent in vivo digestibility. Starch disappearance was measured both with and without initial ruminal pre-incubation during 4 h. Bags were retrieved from either the ileal cannula or faeces. Two dry Danish Holstein cows fitted with rumen cannulas were used for rumen pre-incubations and two lactating Danish Holstein cows fitted with duodenal and ileal cannulas were used for intestinal incubations. Rumen pre-incubation had no significant effect on disappearance from bags recovered in faeces. The disappearance of legume starch was lower, both in the rumen and small intestine, compared with starch from barley, wheat, oats, ear maize and maize. Transit times of the mobile bags from duodenum to ileum were not significantly different between feeds. A weak positive correlation was found between in vivo small intestinal and total tract digestibility of starch and disappearance obtained using the mobile bag technique across a broad range of starch sources. Omitting two less conventional starch sources (NaOH wheat and xylose-treated barley) resulted in a high (0.87) correlation between total tract in vivo digestibility and mobile bag disappearance. The use of the mobile bag method for estimation of in vivo starch digestibility will therefore depend on the starch type.