How Tonaleve builds your sound profile
The psychoacoustic protocol the app uses to calibrate your audio: 2AFC comparisons with logarithmic bisection, octave check, threshold via Levitt staircase, level in dB SL, and consistency verification across repeats. This is an internal calibration, not a clinical measurement.
Last reviewed: August 2026
What this is. The sound profile is the internal reference Tonaleve uses to calibrate the audio you’ll listen to: a tone in Hz and a comfortable level in dB SL. It is not a clinical measurement, it does not produce a diagnostic result, and it does not replace an audiological evaluation. We publish the full protocol so you know exactly where every number comes from.
Why finding the tone is so hard
Finding the tone that resembles what you hear is not trivial, and the literature documents this bluntly:
- Classic pitch matching (adjusting a tone with a slider until it “sounds the same”) has very poor test–retest reliability. In comparative studies, the standard clinical method achieved reproducibility of only 23% for pitch and 40% for loudness, while similarity-rating methods reached 84% and 73% respectively (Standard Clinical Method vs Tinnitus Likeness Rating).
- Two-week stability of the pitch match shows an ICC ≈ 0.73, and for the loudness match just ICC ≈ 0.48 (PLoS ONE 2014).
- Octave confusion — mistaking the true tone for its upper or lower octave — is a classic and frequent error in people without musical training. Recursive repetition of the match reduces this error and improves test–retest reliability (Repetitive Recursive Matching, Audiology & Neurotology 2020).
- Recent work proposes treating pitch not as a single number but as a distribution with a confidence interval (McMillan & Cannon, A Bayesian Perspective on Tinnitus Pitch Matching, Ear & Hearing 2014).
Tonaleve is designed around these findings: every step in the protocol targets an error documented in the literature.
The protocol, step by step
Calibration is a guided seven-step wizard. None of the steps require musical or audiological knowledge.
1. Sound character
Before searching for the frequency, you identify the timbre of your tinnitus by comparing four candidates: pure tone, narrowband noise, pink noise, and white noise. The samples are presented in the high-frequency range (2–10 kHz), where most tonal tinnitus is concentrated, and are loudness-compensated with an approximation of the ISO 226 equal-loudness contour (~40 phon) so no candidate “wins” simply by sounding louder.
2. Audibility map
A quick pre-check sweeps 10 fixed frequencies between 250 Hz and 12 kHz and builds a map of which regions you can hear and at what level. This map combines an ISO 226-based prior with your actual responses, and serves a critical function: avoiding the number-one methodological error in matching apps — voting “sounds like it” for a candidate you can’t actually hear because it falls in a region with hearing loss. Every subsequent stimulus is presented at a target level of ~18 dB above your estimated threshold at that frequency; if you can’t hear it, the system raises the level in a controlled way and, if it’s still inaudible, excludes that region rather than letting it contaminate the result.
3. Frequency search via A/B comparisons (2AFC)
Instead of a free slider — the method with the worst documented reliability — Tonaleve uses two-alternative forced choice (2AFC) with logarithmic bisection:
- You’re presented with two tones, A and B, within the candidate range.
- You choose which one is closer to your tinnitus (or “they sound the same” / “I can’t hear one”).
- Each response narrows the range and feeds a continuous similarity curve (Gaussian kernel over 80 log-spaced bins), so the result doesn’t depend on a single lucky choice but on the accumulation of evidence.
- The search ends when the range is narrow enough or when the comparison cap is reached (11 by default).
The final estimate is the geometric mean of the final range — mathematically the center of the interval on a logarithmic scale, which is how the ear perceives pitch. Along with the frequency, the system computes a confidence value that explicitly discounts regions excluded for inaudibility, so as not to inflate certainty with exclusions that aren’t evidence.
4. Octave check
With the candidate frequency f, the protocol explicitly compares f/2, f, and 2f — the
direct control against octave confusion recommended by the recursive-matching literature —
and then allows a fine adjustment of ±1 and ±10 Hz.
5. Hearing threshold at your frequency
To express the level of your tinnitus in a meaningful unit, your hearing threshold is first measured exactly at the matched frequency, using an adaptive 2-down/1-up staircase (Levitt transform): the level drops after two correct responses and rises after one miss, with an initial step of 8 dB that halves at each reversal (minimum 2 dB). The measurement ends after 6 reversals, and the threshold is estimated as the mean of the last 4 — the standard adaptive-psychophysics procedure.
6. Loudness match in dB SL
You adjust the loudness of a tone at your frequency until it matches your tinnitus, on a scale of 0 to 45 dB SL — decibels above your own threshold measured in the previous step.
Why dB SL and not “volume %”: dB SL (Sensation Level) is the methodological backbone of Tonaleve. Because the threshold and the match are measured with the same headphones and the same audio chain, the equipment’s coloration cancels out: the result is comparable across sessions even without an absolute hardware calibration. It’s the same reason clinical tinnitometry reports the loudness match in dB SL.
7. Consistency verification (test–retest)
The protocol repeats the match 2–3 times and calculates the spread of your results in semitones. A spread under 0.5 semitones is classified as excellent consistency; under 1.2, good; above that, low — and in that case the result is presented with the corresponding caveat. This repetition is exactly what the evidence points to as the best predictor of a reliable match.
What you get at the end
| Metric | What it is | How it’s obtained |
|---|---|---|
| PM (pitch match) | The tone closest to what you hear, in Hz | 2AFC + log bisection + octave check |
| Confidence | How narrow the range ended up | Final bracket width, discounting inaudible regions |
| Threshold | Your minimum audible level at that frequency | Levitt 2-down/1-up staircase |
| LM (loudness match) | The level at which you hear it, in dB SL | Matching relative to your own threshold |
| Consistency | Test–retest spread, in semitones | 2–3 repeats of the match |
That profile directly feeds the sound program: the notch of the notched stimulus is centered on your PM, session levels are set in dB SL relative to your LM, and progress is charted over time with each recalibration.
Limits of the calibration — what Tonaleve is not
Methodological honesty requires stating what this calibration cannot do:
- It is not a clinical audiometry. It does not replace tinnitometry performed with a calibrated audiometer according to ISO 8253 / IEC 60645, in a booth, by a professional.
- It does not work in absolute dB HL. Without lab calibration of the headphones, levels are relative to your threshold (dB SL), not absolute. This is a limitation of every consumer app, and Tonaleve states it rather than hiding it.
- Bluetooth headphones are not “flat.” Lossy codecs, DSP, and noise cancellation introduce coloration and latency. The dB SL design minimizes the impact, but it’s recommended to always use the same headphones, with the same settings.
- The profile is global, not split by ear in the current version.
- It does not diagnose or treat anything. Tonaleve is a general wellness product and not a medical device; tinnitus is a symptom and its cause is evaluated by a healthcare professional.
The full references for this page are in Sources and bibliography.
