OBJECTIVE To assess the diagnostic usefulness of motor evoked potentials (MEPs) and to identify the optimal method for calculating the central conduction time. The test results were evaluated in a prospective study of 1023 neurological patients. METHODS We evaluated the correlation between clinical and electrophysiological findings, the accuracy, the sensitivity, the percentage of subclinical abnormalities and the false negative rates of MEPs in different neurological disorders. In patients with lower motor neuron involvement, we compared the central conduction time calculated as the difference between the latency of the cortical and magnetic root stimulation responses with that calculated using the F-wave method. RESULTS The agreement index between electrophysiological and clinical findings was 87%. The overall accuracy of the test was 0.97. The higher sensitivity values were demonstrated in spinal cord disorders (0.85), hereditary spastic paraplegia (0.80) and motor neuron diseases (0.74). The higher percentages of subclinical abnormalities were found in motor neuron disorders (26%) muscular diseases (24%), multiple sclerosis (13.5%) and spinal cord diseases (12.5%). The higher false negative rates were found in sylvian stroke (0.36) and hereditary spastic paraplegia (0.16). Central conduction study using magnetic paravertebral stimulation but not using the F-wave method, resulted in 12% and 10% of false positive values in lower limb multiradiculopathies and in neuropathies, respectively. CONCLUSIONS MEPs represent a highly accurate diagnostic test. MEP clinical value is maximum in motor neuron, muscle and spinal cord diseases. In patients with lower motor neuron involvement, the gold standard for central conduction determination is the F-wave method.