Acoustic deprivation, i.e. hearing loss, is responsible for a cascade of processes resulting in reorganisation of the cortex. Tinnitus mechanisms are explained by synchronization of the neural spontaneous activity and might be related to cortical re-mapping. Auditory discrimination training (ADT) has demonstrated in both animals and humans to induce tonotopical changes in the auditory pathways through neural plasticity. We hypothesize that ADT could have some effect on tinnitus perception. The objective of this study is to compare the effect on tinnitus following two paradigms of ADT. Only patients from 20 to 60 years of age were recruited. Inclusion criteria were pure tone tinnitus of mild or moderate handicap according to the Tinnitus Handicap Inventory score (<56). ADT patients were randomized in two groups: SAME (ADT in the same frequency of tinnitus pitch, 20 patients) and NONSAME (ADT in the frequency one-octave below tinnitus pitch, 21 patients). Groups of pair of tones (70% standard tones ST, 30% deviant tones ST + 0.1-0.5 kHz) were randomly mixed for 20 min/day during 1 month. Patient had to mark when the two sounds of the pair were similar or different. Control group included 26 patients from the waiting list (WLG). Patients were also divided according to the trained frequency and the deepest hearing-impaired frequency. Outcome parameters were set up according to the answer to the question "is your tinnitus better, same, or worse with the treatment?" (RESP), the tinnitus handicap inventory (THI) and the visual analogue scale from 1 to 10 on tinnitus intensity (VAS). Tinnitus improved in 42.2% of the patients (RESP). VAS and THI scores were reduced but only THI differences were statistically significant (P = 0.003). ADT patients improved significantly compared with WLG in RESP and THI scores (P < 0.01). Training frequencies one-octave below the tinnitus pitch (NONSAME) decreased significantly THI scores compared with patients trained frequencies similar to tinnitus pitch (SAME, P = 0.035). RESP and VAS scores decreased more in NONSAME group though differences were not significant. We did not find any differences when comparing the group training the deepest hearing-impaired frequency and the group who trained other frequencies. Auditory discrimination training significantly improved tinnitus handicap compared to a waiting list group. Those patients who trained frequencies one octave below the tinnitus pitch had better outcome than those who performed the ADT with frequencies similar to the tinnitus pitch (P = 0.035).