The corticospinal tract (CST) is one of the descending tracts that carry the forelimb volitional information. Using flexible multi-electrode arrays (MEAs), the CST signals were recorded in rats during active and resting states. The power spectral density (PSD) of CST signals during the active state were notably higher than those observed during resting or anesthesia. Average inter-channel coherences below 30 Hz were considerably higher for reach-to-pull task compared to face grooming and the quiet state, suggesting the presence of volitional information in the recorded CST signals. A novel feature extraction technique (spectrum of cross-spectrum) was applied to obtain feature vectors and used in a classification algorithm. The extension and flexion phases of the forelimb movements were classified with 95% accuracy. The results warrant further investigations for extraction of more detailed volitional motor control information from the CST signals.