To evaluate the potential use of time-frequency analysis and its reliability in intraoperative somatosensory evoked potential (SEP) monitoring. One hundred ninety-one patients undergoing thoracic and/or lumbar spinal surgery were studied retrospectively. The SEP signals were recorded during different stages of surgery. Averaged SEP was analyzed by short-time Fourier transform. The main peak in the time-frequency interpretation of SEP was measured in peak power, peak time, and peak frequency. The variability of these parameters was compared with that of amplitude and latency during different stages of surgery. The reliability of these parameters was also compared in true-positive and false-positive cases. During different surgical stages for the posterior tibial nerve SEP, the intrasubject variability of peak power was found to be more stable than that of amplitude, while the intrasubject variability of peak time did not show any difference compared with that of latency. The peak frequency presented stable during surgery. Moreover, the true-positive SEP case showed that peak power may detect the potential injury earlier than amplitude does. The false-positive outcomes could be reduced by the proposed method. The SEP peak component was found stable and reliable during the different stages of surgery. For clinical application purpose, time-frequency analysis was suggested to be an additional monitoring method besides the conventional amplitude/latency measurement since it provided a more reproducible and prompt response to the potential injury in intraoperative SEP monitoring.