Objectives. Following lesions in somatosensory pathways, deafferentation pain often occurs. Patients report that the pain is qualitatively complex, and its treatment can be difficult. Mirror visual feedback (MVF) treatment can improve deafferentation pain. We sought to classify the qualities of the pain in order to examine whether the potential analgesic effect of MVF depends on these qualities. Methods. Twenty-two patients with phantom limb pain, or pain related to spinal cord or nerve injury, performed a single MVF procedure. Before and after the MVF procedure, we evaluated phantom limb awareness, movement representation of the phantom or affected/paralysed limb, pain intensity on an 11-point numerical rating scale (010) and the qualities of the pain [skin surface-mediated (superficial pain) vs deep tissue-mediated (deep pain)] using lists of pain descriptors for each of the two categories. Results. Fifteen of the patients perceived the willed visuomotor imagery of the phantom or affected/paralysed limb after the MVF procedure. In most of the patients, a reduction in pain intensity and a decrease in the reporting of deep-pain descriptors were linked to the emergence of willed visuomotor imagery. Conclusions. In this pilot study, we roughly classified the pain descriptor items into two types for evaluating the qualities of deafferentation pain. We found that visually induced motor imagery by MVF was more effective for reducing deep pain than superficial pain. This suggests that the analgesic effect of MVF treatment does depend on the qualities of the pain. Further research will be required to confirm that this effect is a specific consequence of MVF.