A two-stage in vitro system (IVDE) for estimating AME(n) in poultry feeds was investigated. For 71 diets ranging from 2.2 to 3.4 kcal/g, the average AME(n) was 2.889 kcal/g and the mean IVDE value was 3.005 kcal/g. From the 71 diets, 30 (42.20/o) showed differences between AME(n) and IVDE of less than .100 kcal/g and represented diets across the AME(n) range of values. The statistical analysis of the data showed a standard error of the estimate (SEE) of .152 kcal/g for the 71 diets assayed. No clear differences in accuracy of AME(n) among the diets, as related to the composition and proportion of ingredients, were observed. Thus, the IVDE method gave different AME(n) for diets of similar composition. The application of the IVDE system to selected ingredients showed that the AME(n) of com was underestimated by the method. However the AME(n) of roasted, extruded soybeans and oats was estimated accurately by the IVDE method. Other ingredients were greatly overestimated by the in vitro technique (soybean meal, corn gluten meal, and barley). The results of applying the IVDE method for estimating AME(n) showed the limitations of this technique with regard to the universality of its application. Although the method was successful in estimating AME(n) values of diets and ingredients, for many samples the IVDE technique did not give acceptable results.