Because roughage in feedlot diets is one of the most expensive ingredients on an energy basis, regimens that minimize roughage usage are of interest. Crossbred steers of British breeds (n = 112, initial BW = 405 kg) were used to compare the feeding of diets containing 2% roughage from d 22 through 84 and 10% roughage from d 85 to finish (d 133; 2/10%) to the feeding of 10% roughage throughout the finishing period (10/10%); all diets were based on steam-flaked sorghum grain and contained monensin and tylosin. When the 2% roughage diet was fed, steers consumed less feed (6.8 vs 7.8 kg/d, P less than .01), tended to gain less (1.11 vs 1.20 kg/d, P = .13), and were numerically more efficient (16.5 vs 15.5 kg of gain/100 kg of DMI, P greater than .2) than steers fed the 10% roughage diet (10/10%). After the roughage content was increased from 2 to 10% on d 85 (all steers fed 10% roughage), steers fed the 2/10% regimen had greater DMI (8.4 vs 8.0 kg/d, P = .08) and ADG (1.29 vs 1.09 kg, P = .06), and tended to be more efficient (15.4 vs 13.6 kg of gain/100 kg of DMI, P = .10) than steers fed the 10/10% regimen. Steers fed the two regimens had similar (P greater than .2) overall gain performance. The 2/10% regimen tended to have a greater percentage of Choice carcasses (58 vs 42%, P = .14) and numerically more liver abscesses (24 vs 15%, P greater than .2) than the 10/10% regimen.(ABSTRACT TRUNCATED AT 250 WORDS)