A total of 240 male broilers were used in a 21-d experiment to study the effect of reference diet type on nitrogen corrected-apparent metabolizable energy (AMEn) of soybean meal (SBM) and canola meal (CM). The two reference diets were corn-SBM or corn-CM. To prepare experimental diets, graded levels of SBM or CM at 300 and 450 g/kg were added to each of the two basal diets to make a total of 10 diets. The experimental diets were fed to birds from d 14-21. Excreta were collected on d 20 and 21. The AMEn values were calculated using the difference or the regression methods for SBM using SBM- and CM reference diets, and for CM using SBM- and CM-reference diets. The data for AMEn determined using the difference method were analyzed as 2×2×2 factorial to determine the effect of reference diet (corn-SBM and corn-CM), protein feedstuff (SBM, and CM), and substitution level (300 and 450g/kg) of assay feedstuff in the reference diets. There was significant (P < 0.05) three-way interaction between reference diet, feedstuff, and substitution level. For AMEn of SBM determined using CM reference diet, AMEn increased (P < 0.05) as the inclusion level of SBM in the CM reference diet increased. The AMEn of CM decreased (P < 0.05) with increasing inclusion level of CM in the SBM-reference diet. Data on AMEn determined by the regression method were analyzed as a 2×2 factorial to determine the effect of the reference diet (SBM- or CM-based) or feedstuff (SBM or CM). There was no significant reference diet × feedstuff interaction. AMEn was greater (P < 0.01) when determined in a corn-CM reference diet compared with corn-SBM reference diet but there was no significant difference between AMEn of SBM and CM. For comparison of AMEn calculated by the two methods, the data were analyzed as a 2×2×2 factorial to incorporate the effect of reference diet type (SBM- or CM-based), protein feedstuff (SBM or CM) and method (difference or regression). There were no significant 3- or 2-way interactions. AMEn calculated using the regression method gave greater (P < 0.01) values compared to the difference method. In addition, AMEn was greater (P < 0.01) when determined using a corn-CM reference diet compared to corn-SBM.