We studied the use of gas curve subtraction to distinguish between two fractions soluble in neutral detergent solution. Samples of unfractionated (whole) forage, residue insoluble in 90% ethanol, and isolated NDF were fermented in vitro, and gas production was monitored. The gas volume associated with the ethanol solubles (A fraction) was determined as the difference between the gas from the whole forage and from the ethanol residue. The gas yield associated with the fraction insoluble in 90% ethanol but soluble in neutral detergent solution (B fraction) was determined by subtracting the isolated NDF gas curve from the corresponding ethanol residue curve. This experiment included three forages (alfalfa, bromegrass, and orchardgrass) harvested at two maturities and preserved by freeze-drying or ensiling to form a 3 x 2 x 2 factorial arrangement. Ensiling reduced the rate of gas formation from the A fraction by approximately 30% (P < .01). Ensiling increased (P < .05) the size of the A fraction (2 to 10% of DM) but did not change the volume of gas produced (P > .05). The gas yield from the B1 fraction was reduced 40% (P < .05) by ensiling with no significant change in rate. Curve subtraction of gas production profiles may be used to obtain rate estimates for multiple neutral detergent soluble pools. The separation of the neutral detergent solubles into two pools clarified the effects caused by ensiling. Changes due to ensiling on the rate of gas produced were associated with the A fraction, and the effects on final gas volume were associated with the B1 fraction.