Up until now, the exact mechanism by which neuromodulation using sacral nerve stimulation works still remains unknown. Recent studies of pelvic floor contraction during peripheral nerve evaluation (PNE) have shown that several muscle responses are reflexly mediated. However, whether these reflexes originate from a segmental level within the sacral spinal cord or from supraspinal neuronal centers involving spino-bulbospinal pathways remains to be determined. Therefore, recordings of external anal sphincter reflexes during PNE in complete spinal cord injury (SCI) patients was performed. Three patients with complete SCI (two female and one male) suffering from neurogenic incontinence underwent the first stage of a stimulator implant (peripheral nerve evaluation). Stimulation was obtained through an electrode placed in the S3 foramen. Concomitantly, the latency of the response of the striated anal sphincter was measured using a concentric needle electrode placed in the striated part of the anal sphincter. Repeated measurements within an interval of 30 min were performed to confirm the reliability of the responses. In all subjects, reflex responses of early and late latencies could be recorded. The early response showed a mean latency of 41.2 ms (range 33.3-62 ms), which corresponds to a segmental reflex, similar to the pudendo-anal reflex. The late response appeared with a mean latency of 189.4 ms (range: 183.3-197.8 ms) and with high variability and fatigability, suggesting a reflex response organized at polysegmental spinal levels. Despite successful foramina lead placement, none of our complete SCI patients showed any improvement of neurogenic incontinence after 5 days of treatment by sacral nerve stimulation. The findings confirm that the anal contractions observed during peripheral nerve evaluation are reflex responses mediated by afferent pathways. Both the early and late reflex responses are of spinal origin, since they were obtained in complete SCI patients in whom all spino-bulbo spinal loops are supposed to be interrupted. The finding that neuromodulation is working in non-neurogenic patients but is less successful in complete SCI patients could give evidence that preserved spino-bulbo spinal loops contribute to the positive effects of neuromodulation. The role of an eventual spino-bulbo spinal loop acting during neuromodulation in non-neurogenic patients has to be assessed in further studies.