Notched sound therapy has not been shown to work better than simply listening to music: most comparisons of notched against ordinary music, including two of three meta-analyses, find the notch adds nothing, though the music itself may help.
This is the treatment with the most satisfying mechanism in the whole field, and the trials keep failing to find the ingredient the mechanism depends on. Both halves of that sentence are worth understanding, and this site has a particular obligation to get it right — it ships a free tool that generates notched sound.
The idea
The mechanism behind most tinnitus on this site is reduced input at some frequency, followed by the auditory system raising its own activity to compensate. Notched sound therapy takes that seriously and inverts it.
Remove a narrow band of frequencies — the band containing your own tinnitus pitch — from music you enjoy. Then listen for a couple of hours a day, for months. The neighbouring frequencies keep firing; the tinnitus frequency region gets nothing. The hope is that lateral inhibition from those active neighbours quiets the overactive region, and the map reorganises.
It is called tailor-made notched music training, TMNMT in the literature. It costs nothing, it is pleasant, and the reasoning is clean enough that you can see why it attracted attention.
The 2010 result that started it
A trial published in PNAS in 2010 exposed people with chronic tinnitus to self-chosen, enjoyable music notched around their individual tinnitus frequency. After twelve months of regular listening, the treatment group showed significantly reduced subjective tinnitus loudness — and, the part that made the paper land, reduced evoked activity in the auditory cortex areas corresponding to the tinnitus frequency, compared with a group given analogously processed placebo music.
That is a self-report and a physiological measurement moving together, which is rare in this field and is exactly what you would want if the mechanism were real.
Now the number that governs how much weight it carries: n = 8 in the treatment group, and 8 in the placebo group.
Sixteen people. That is a proof-of-concept, and the authors present it as one. It is not a reason to disbelieve the finding; it is a reason that everything after it matters more.
The replication attempt, and the detail nobody quotes
In 2016, BMC Neurology published what its authors describe as the first TMNMT study planned and conducted to CONSORT standards. 100 participants with chronic tonal tinnitus, randomised, double-blinded, listening to notched music two hours a day for three months, with follow-up a month after stopping.
The primary outcomes were the Tinnitus Handicap Questionnaire and visual analogue scales for loudness, awareness, distress and handicap.
No effect was found for the primary outcome measures.
Three things come after that sentence, and all three belong here:
- Tinnitus loudness at follow-up was significantly reduced in the treatment group compared with control. That is a real, positive finding on the most direct measure of the sound itself.
- A post-hoc analysis correcting for low reliability in the visual analogue scales found a significant reduction in the overall VAS mean score in the treatment group. Post-hoc, and stated as such.
- And the detail that essentially never gets quoted: on the Tinnitus Questionnaire, a secondary measure, the treatment group's distress scores went up while the placebo group's went down. It did not reach significance in post-hoc analysis and it had disappeared by follow-up — so it is not a finding that notched music makes distress worse. It is a reminder that in a trial where the headline is null, the noise goes in both directions.
The authors' own conclusion is precise: after three months the effect is observable in the most direct rating of tinnitus perception — loudness — while more global measures of distress do not show relevant changes.
Against another therapy, it wins
A 2023 randomised trial in Ear and Hearing did something different: instead of comparing notched music against placebo, it compared it against tinnitus retraining therapy, a widely used treatment.
120 people randomised, 112 completing, single-blinded, three months, with TRT's sound component adjusted to two hours a day for feasibility.
- Both groups improved. The paper's conclusion is that both TMNMT and TRT alleviated chronic subjective tinnitus effectively.
- The between-group difference on the Tinnitus Handicap Inventory favoured notched music: −6.90 points at one month (95% CI −13.53 to −0.27) and −6.17 at three months (95% CI −13.04 to 0.71).
- The authors describe these as closely reaching clinical significance for effective relief, and conclude TMNMT appeared more effective and potentially superior to TRT.
Read the second confidence interval, though. At three months it crosses zero. The one-month interval touches −0.27, barely clearing it. "Closely reached" is the authors' own hedge and it is doing real work.
The trial that tests the notch directly
Everything above compares notched music against placebo, or against another therapy. A 2021 trial in the Journal of International Advanced Otology asked the question that matters most, and its objective sentence says so: to compare the outcomes of tailor-made notched music therapy and ordinary music.
104 patients with chronic subjective tinnitus, double-blind, randomised to notched music or to ordinary music, assessed at one, three and six months.
The Tinnitus Handicap Inventory and the total visual analogue scale showed no significant differences between the groups.
The authors add one qualifier, and it should be kept: the treatment group showed a greater reduction in scores than the control group over the follow-up period — a trend in the right direction that did not reach significance. Their conclusion is appropriately flat. Notched music therapy is an optional treatment for bothersome subjective tinnitus, and more research is needed for firm conclusions about its benefits and cost-effectiveness.
Two groups both listening to music they chose. One had a notch in it. After six months there was no measurable difference between them.
Where two 2024 meta-analyses landed
Here is where the story resolves, and it resolves against the notch rather than against the therapy.
Two separate systematic reviews and meta-analyses, both published in 2024, by different groups, from different study sets.
The first, in the European Archives of Oto-Rhino-Laryngology, evaluated 234 papers and included seven clinical trials, five of them using the Tinnitus Handicap Inventory. It found no statistically significant effect of TMNMT on tinnitus handicap at three months or at six. Its heterogeneity was enormous — I² = 79.96% — meaning the included studies disagreed with each other profoundly. Its authors note that four of five studies using visual analogue scales showed positive effects, but that there were not enough of them to pool.
The second, in the American Journal of Otolaryngology, pooled three studies: 99 people on TMNMT against 109 on sound therapy. It found no significant difference in tinnitus distress between notched music and simply listening to music — and unlike the other review, its studies agreed almost perfectly with each other, I² = 0%.
That second comparison is the one that matters. Not "does notched music help" but "does the notch help". Three studies, low heterogeneity, no difference.
A third meta-analysis, in 2025, disagrees
Honesty requires this section, because it cuts against the two above.
A 2025 systematic review and meta-analysis in the European Archives of Oto-Rhino-Laryngology — a different group from the 2024 review in the same journal — pooled 14 randomised controlled trials and 793 patients, 395 on notched music therapy against 398 on conventional music therapy. It found notched music better at three months:
- Tinnitus Handicap Inventory — MD −8.62 (95% CI −17.01 to −0.23), P = 0.044
- Tinnitus VAS — MD −1.13 (95% CI −2.49 to −0.11), P = 0.003
Its sensitivity analysis was stable, it found no significant publication bias, and notch frequency range, publication year and patient age did not explain the results. Its conclusion is that notched music therapy is worthy of clinical promotion.
Now read the interval on the headline number. −17.01 to −0.23. The upper bound is effectively zero, and P = 0.044 barely clears the conventional line. An estimate running from "comfortably larger than the 7-point responder threshold" to "no change any person could detect" is not a measurement of an effect. It is a statement that the effect has not been pinned down.
So the honest position is now: three meta-analyses, two null and one marginal, with the marginal one unable to exclude approximately nothing. Add the 2026 hearing-aid trial below, which found no between-group difference at all, and the pattern is of a treatment that keeps failing to separate from its own control.
It is also a reason to correct this site's earlier framing. The section above was originally headed "the comparison that decides it", written when two reviews agreed and nothing contradicted them. A third has since disagreed. The interval undercuts it heavily, but the heading claimed more certainty than the evidence now supports, and it has been changed.
A 2026 trial puts the notch inside a hearing aid
The place most people will meet notch therapy is not a music player — it is a setting in a hearing aid. A randomised controlled multicentre study in the International Journal of Audiology tested it exactly there, against two alternatives: conventional amplification, and increased gain at the tinnitus pitch.
Tinnitus distress fell significantly in all three groups after hearing aid fitting. Between the groups, no significant difference was found on the outcome measures — with one exception, and it does not favour the notch.
Speech intelligibility benefit was greater with conventional amplification than with notch therapy (p = .009) or with increased gain at the tinnitus pitch (p = .022).
That is more informative than a plain null result, because it prices the notch. Notching a hearing aid means withholding amplification at frequencies the wearer needs in order to understand speech. This trial measured what that costs and what it buys: you give up some hearing benefit, and you get no additional tinnitus benefit in return.
The authors' recommendation is explicit — conventional, objectively verified hearing aid amplification for tinnitus patients, on the grounds that it improves tinnitus symptoms and speech intelligibility simultaneously rather than trading one against the other.
What that most likely means
The honest reading is not that notched sound therapy does nothing. It is that the active ingredient may be the music, not the notch.
Listening to two hours a day of music you have chosen because you enjoy it, deliberately, for months, is a real intervention. It is pleasant, it fills silence, and quiet is when tinnitus is loudest — why sound therapy works at all rests substantially on that. Against no treatment, or against a different therapy, that shows up as benefit. Against the same music without a notch in it, it does not.
There is a second reading worth keeping open. The strongest surviving signal across the trials is on loudness rather than on distress or handicap. Loudness is what the mechanism predicts should change, and the instruments that measure distress are not designed to detect it. A treatment that genuinely reduces perceived loudness without reducing distress would look exactly like this in the literature — and would still be worth less to most people than one that works the other way round, because distress is the thing that damages a life.
Where the guidelines land
The 2024 VA/DoD tinnitus guideline found insufficient evidence to recommend for or against sound therapy with altered music, naming notched music therapy and spectrally altered music. That sits beside a weak recommendation in favour of the therapeutic use of sound in general — the guideline supports sound, and reserves judgement on the notch, which is where the trials above leave it too. The older AAO-HNS guideline lists sound therapy in general as an option.
Why this site still offers the tool
Because a free, pleasant, harmless thing with a plausible mechanism and an unconfirmed active ingredient is a reasonable thing to be able to try, provided nobody is misled about what it is.
The tinnitus frequency test matches your tinnitus frequency by ear and can generate optional notched audio in the browser, with no account and no download. Several apps charge a subscription for the same thing. What none of them can honestly promise, and what this page will not, is that the notch is what makes the difference.
If you have money or attention to spend on tinnitus rather than time, the treatment comparison shows where the evidence is: CBT and its internet-delivered form, and hearing aids where there is hearing loss to correct.
What to do with this
- If you want to try it, it is free here and no source describes harm. Give it the schedule the trials used — around two hours a day, for at least three months — because anything less has not been tested.
- Expect loudness, not relief. That is the outcome the surviving evidence points at, and it is not the outcome most people are hoping for.
- If the notched version is a hassle, the meta-analysis says plain music performed the same. Listening to music you love is the part with support behind it.
- Do not let it displace the treatments with better evidence. Three months of daily listening is a real commitment, and it is not a substitute for a course of CBT.
- If someone is selling it to you, the two 2024 meta-analyses are the papers to ask them about, and the second one compared it against ordinary music.