We studied the effects of robotic-assisted locomotor (LOKOMAT) training on neuromuscular abnormality associated with spasticity in persons with incomplete Spinal Cord Injury (SCI). LOKOMAT training was performed 3 days/week for 4 weeks, with up to 45 minutes of training per session. Subjects were evaluated before and after 1, 2, and 4 weeks of training, and the effects of training on the intrinsic (muscular) and reflexive components of the neuromuscular properties were quantified over the ankle range-of-motion. A linear (slope&intercept) regression was fit to the stiffness-angle curve. "Growth mixture" modeling was used to identify recovery classes for these parameters over the training period. Two distinct classes were observed. Class 1 subjects had initially higher reflex stiffness parameters (i.e., intercept and slope vs. ankle position) and reduced significantly over the training period. Class 2 subjects initially had lower reflex stiffness parameters and experienced non-significant reductions. Similar results were observed for the intrinsic stiffness intercept; however, intrinsic slope showed no significant improvement over training for either class. These findings demonstrate that LOKOMAT training is effective in reducing reflex and intrinsic stiffness (which abnormally increase in SCI) and improving the abnormal modulation of reflexes over the ankle range-of-motion.