A previously described endo-(1----4)-beta-D-xylanase produced by Aspergillus niger was allowed to react with linear unlabeled and labeled D-xylo-oligosaccharides ranging from D-xylotriose to D-xylo-octaose. No evidence of multiple attack or of condensation and trans-D-xylosylation reactions was found. Maximum rates and Michaelis constants were measured at 40 degrees and pH 4.85. The former increased with increasing chain-length from D-xylotriose through D-xylohexaose to approximately 70% of that on soluble larchwood D-xylan, and then decreased slightly for D-xyloheptaose and D-xylo-octaose. Michaelis constants decreased monotonically with increasing chain-length. Bond-cleavage frequencies were highest near the reducing end of short substrates, with the locus of highest frequencies moving towards the middle of larger substrates. These data indicated that the endo-D-xylanase has five main subsites, with the catalytic site located between the third and fourth subsites, counting from the nonreducing end of the bound substrate. The subsite to the nonreducing side of the catalytic site strongly repels its corresponding D-xylosyl residue, while the two subsites farther towards the nonreducing end of the substrate strongly attract their corresponding residues. The subsite to the reducing side of the catalytic site moderately attracts D-xylosyl residues, while the next one towards the reducing end has a high affinity for them. The residual error of the numerical estimation was allocated largely to the Michaelis constants of the different D-xylo-oligosaccharides, whose calculated values were appreciably smaller than measured values, especially for shorter substrates. This suggests that the subsite model cannot fully account for the experimental data. Estimated and measured values of maximum rates, bond-cleavage frequencies, and dissociation constant when the active site is fully occupied by substrate agreed more closely with each other.