Twenty eight patients with unilateral cervical radiculopathy were studied by somatosensory evoked potentials (SEPs) from nerve stimulation at the wrist and from skin stimulation at the first, third or fifth finger depending on the root involved. In order to evaluate the reliability of various "radicular SEP patterns" as described in the literature, absolute latencies and side-to-side differences of the brachial plexus component from the supraclavicular fossa (N9), the medullary component (N13) from the cervical vertebra Cv7, and the primary cortical component (N20, P25) were assessed. Side-to-side differences of the amplitudes of N20/P25 and of the conduction times across the intervertebral fossa (interval N9-N13) were analysed. After nerve stimulation, 68% of the patients had false negative findings on the symptomatic, while 36% had positive findings on the asymptomatic side. After segmental stimulation, 72% of the patients had false negative findings on the symptomatic, while 22% had positive findings on the asymptomatic side. It is concluded that SEPs following nerve and segmental stimulation do not reliably confirm clear-cut already established diagnoses of unilateral radiculopathy with sensory and motor deficit. Therefore, they will not be helpful in the electrophysiological investigation of cervicobrachialgias of unknown origin.