A repeated flexure apparatus which has been used for much of the recent research on fatigue behavior of asphalt paving was modified to accommodate 15-in.-long beam specimens with 3-in. instead of 1½-in. widths and depths. Controlled stress flexural fatigue studies were conducted for 10 different paving mixes using the larger cross section specimens in an effort to reduce test variability, and in so doing reduce time and cost for determining the fatigue behavior of a paving mixture. Based on results of the studies and on published work of others, it was found that variability of flexural fatigue tests was reduced when 3-in. instead of 1½-in. width and depth specimens were used. Correlation coefficients ranging from 0.91 to 0.99 were obtained for stress-fracture life and strain-fracture life regression lines when from 6 to 9 specimens with 3-in. widths and depths were tested at 70 F. Flexural fatigue behavior determined for 8 or 9 specimens with 3-in. widths and depths appears to be the same as determined for 2 or more times as many specimens with 1½-in. widths and depths.