A total of 72 male broilers were used in the study to investigate the effect of adaptation length and substitution levels on AME of soybean meal (SBM) determined by both the difference and regression methods. At 14 days old, the broilers were allocated to 4 treatments with 6 replicates of 3 birds per replicate cage. The experimental diets were fed for either 4 days or 7 days as adaptation lengths, and excreta were collected on days 20 and 21. Within each adaptation period, 4 diets comprising one basal and 3 test diets in which SBM substituted the energy yielding components of the basal diets at 150, 300, or 450 g/kg were fed, corresponding to levels 1, 2 or 3, respectively. The AME of SBM was calculated using the difference method at each substitution levels. In addition, AME of SBM was calculated by the regression method using four different combinations of the substitution levels (i.e. 0,1 and 2;0,1 and 3;0,2 and 3; or 0, 1, 2 and 3). The data for AME determined by the regression method were analyzed as 2×4 factorial for the influence of adaptation length (2) and combination of substitution levels (4). There was no adaptation × substitution level interaction for AME determined by both regression and difference methods. For AME determined by the difference method at different substitution levels, 150 g/kg gave lower AME values (P < 0.01) than that determined at 300 and 450 g/kg substitution levels but adaptation length had no effect. For AME determined by the regression method, substitution levels had no effect on AME of SBM but AME was greater (P < 0.05) in birds with 4-day adaptation compared with 7-day adaptation length. AME values determined by the two methods were also statistically compared. There was significant method × adaptation length interaction (P < 0.05) for SBM metabolizable energy. The AME determined by regression method after 4-day adaptation length was greater (P < 0.10) than that determined after 7-day adaptation length, but there was no difference in AME for SBM determined by the difference method at different adaptation lengths.