This study aimed at investigating and quantifying the time course of abnormal synergies in the affected upper limbs of chronic stroke patients treated with a neuro-rehabilitation robot. Eight age-matched normal and seven chronic stroke patients were recruited to perform rectilinear tracking movements in four directions for four months. Kinematic, kinetic and electromyographic (EMG) data were used to develop two biomechanical indices and one EMG assessment index from principal component analysis (PCA) to characterize abnormal synergies. Initially, compared with those in normal subjects, the correlation between the elbow angle and forearm pro-supination torque was smaller, the variation of the forearm torque was larger and the abnormal co-contraction of the elbow and shoulder muscles, shown by the coactivation ratio derived from EMGs, was similar for the stroke patients. The correlation increased, the variation decreased and the coactivation ratio decreased with the time course of treatment. In summary, the proposed assessment indices indicated significant improvement of abnormal synergies after four months of treatment with the robot.