The purpose of this study was to evaluate the efficacy and reliability of intraoperative spinal monitoring using spinal cord-evoked and compound muscle action potentials. We reviewed 716 cases of spinal monitoring performed over 15 years. The series contained 672 patients with spinal functions that could be monitored intraoperatively; 44 (6.1%) were impossible to record. Based on the 21 impossible-to-record patients, it is evident that missing a serious case such as a Frankel B type spinal cord injury indicates the limitations of the current monitoring methods for stimulating and recording. The monitoring outcomes were true-negative in 652 patients, true-positive in 12, false-negative in four, and false-positive in four. In two of the patients with false-negative results, postoperative myelomere motor paralysis was observed temporarily even though it was possible to record the muscle-evoked potential after electrical stimulation to the brain [Br(E)-MsEP] at the end of the operation. In cases in which the spinal parenchyma or spinal nerve root might be selectively damaged, Br(E)-MsEPs may not diagnose the disorder accurately. By employing multimodal monitoring, it should be possible to eliminate patients with false-negative results and to detect spinal disorders during the early stages, as well as to examine whether the abnormality that had been recorded by a single method is false-positive.