Loudness Just-Noticeable Difference (JND) Statistics (2026): 50+ Data Points on Weber's Law, Hearing Loss, and Cochlear Implants

Loudness just-noticeable difference statistics 2026: infants need 3 to 12 dB vs 1 to 2 dB in adults, implant users get 6.6 to 45.2 steps, and the 1 dB rule bends.

A healthy adult ear needs only about 1 to 2 dB to notice a tone getting louder, while infants aged 7 to 9 months need 3 to 12 dB, and none of them reliably notice a drop in level (Sinnott and Aslin, Journal of the Acoustical Society of America 1985). The step is not fixed for adults either. The classic fit by Jesteadt, Wier and Green (JASA 1977) implies a step of roughly 1.2 dB at 20 dB above threshold, shrinking to about 0.5 dB at 80 dB above threshold (our calculation, shown in section 1). Cochlear implant users can have as few as 6.6 discriminable steps across their entire electric range (Nelson et al., JASA 1996). For very brief high-frequency tones, the Weber fraction can reach 14 dB at moderate levels (Carlyon and Moore, JASA 1984). We aggregated data from the Journal of the Acoustical Society of America, Perception and Psychophysics, Trends in Hearing, Otology and Neurotology, the EBU, the ISO 226 revision team, and other primary sources. This roundup is about what the ear can tell apart; for how platforms set target levels, see our companion piece on streaming loudness normalization statistics.

TL;DR

  • Weber fraction for pulsed tones: 0.463 (I/I0)^-0.072 from 200 to 8,000 Hz (Jesteadt, Wier and Green, JASA 1977).
  • That fit implies about 1.2 dB at 20 dB above threshold and about 0.5 dB at 80 dB above threshold (our calculation from Jesteadt et al. 1977).
  • Near-miss exponents are usually on the order of 0.9 instead of Weber’s 1.0 (Parker and Schneider, Perception and Psychophysics 1980).
  • Brief 30 ms high-frequency tones in bandstop noise: Weber fraction up to 14 dB at moderate levels (Carlyon and Moore, JASA 1984).
  • Infants 3 to 12 dB versus adults 1 to 2 dB at 1 kHz (Sinnott and Aslin, JASA 1985).
  • 20 older professional musicians were no better than 20 nonmusicians at intensity discrimination, p = .96 (Grassi et al., PLoS ONE 2017).
  • 82 of 85 tests (96%) in cochlear implant candidates detected a 5 dB step (Khandalavala et al., Otology and Neurotology 2026).
  • Cochlear implant users: 6.6 to 45.2 discriminable steps across the electric range (Nelson et al., JASA 1996).
  • Suggested noticeable hearing-aid steps: 3 dB for fine-tuning, 5 dB for self-fitting (Caswell-Midwinter and Whitmer, Trends in Hearing 2019).
  • ISO 226:2023 moved the equal-loudness contours by 0.6 dB at most, below a typical loudness JND (Suzuki, Takeshima and Kurakata, Acoustical Science and Technology 2024).
  • Loudness meters must read EBU test signals within plus or minus 0.1 LU (EBU Tech 3341 v4, 2023).

1. Weber’s Law and the Near Miss: How the Step Shrinks With Level

Weber’s law says the just-noticeable change is a fixed fraction of the starting intensity, so the step in decibels should be the same at every level. Hearing breaks that rule in a consistent way, and that is the single most important fact for anyone quoting “1 dB” as the loudness JND. The fit by Jesteadt, Wier and Green across 200 to 8,000 Hz gives ΔI/I = 0.463 (I/I0)^-0.072, where I0 is the listener’s threshold intensity. The negative exponent means the ear gets sharper as the sound gets louder.

That small exponent is the famous “near miss” to Weber’s law. It held at every frequency Schacknow and Raab tested, and later work by Florentine and colleagues showed that the step is roughly independent of frequency up to about 4 kHz before growing at higher frequencies. Here “sensation level” (SL) means the level above that listener’s own hearing threshold, not the absolute sound level. Read the original Jesteadt et al. 1977 paper and the Florentine, Buus and Mason 1987 level-discrimination study for the full curves.

MetricValueSource
Fitted Weber fraction for pulsed tones, 200 to 8,000 Hz, 5 to 80 dB sensation levelΔI/I = 0.463 (I/I0)^-0.072Jesteadt, Wier and Green, JASA 1977
Level step implied by that fit at 20 dB sensation level (our calculation, see note)about 1.2 dBDerived from Jesteadt et al. 1977
Level step implied by that fit at 80 dB sensation level (our calculation, see note)about 0.5 dBDerived from Jesteadt et al. 1977
Typical near-miss exponent in loudness discrimination studies (1.0 would be exact Weber’s law)on the order of 0.9Parker and Schneider, Perception and Psychophysics 1980
Tone-burst frequencies at which a near miss to Weber’s law appeared (250, 1,000, 4,000 and 7,000 Hz tested)all fourSchacknow and Raab, Perception and Psychophysics 1973
Level discrimination at high levels vs low levels, tones from 0.25 to 16 kHz, six normal listenerssteps generally smaller at high levels, at all but the highest frequenciesFlorentine, Buus and Mason, JASA 1987
Frequency dependence of the step at equal sound pressure levelslargely independent of frequency up to about 4 kHz, rising above 4 kHzFlorentine, Buus and Mason, JASA 1987
Deviation from Weber’s law at 14 kHz compared with 1 kHzless deviation at 14 kHzFlorentine, JASA 1983

Note on the derived rows: the level step is ΔL = 10 log10(1 + ΔI/I). At 20 dB SL, I/I0 = 100, so 100^-0.072 = 0.718 and ΔI/I = 0.463 x 0.718 = 0.332, giving ΔL = 10 log10(1.332) = 1.25 dB. At 80 dB SL, (10^8)^-0.072 = 0.266, ΔI/I = 0.123 and ΔL = 10 log10(1.123) = 0.50 dB. Most recent available data: the canonical pure-tone datasets in this section date from 1973 to 1987.

2. Duration, Frequency and Sound Type: Why One Number Never Fits

The near miss describes long, steady tones. Shorten the sound or move it high in frequency and the picture changes dramatically. With 30 ms tones above 5 kHz, Carlyon and Moore found that when bandstop noise restricts the spread of excitation along the cochlea, the Weber fraction 10 log(ΔI/I) at moderate levels can be as large as 14 dB, an extreme departure from Weber’s law that Nizami later found to be even more severe at 6.5 kHz.

Real-world sounds add another layer. Broadband noise is among the easiest signals to judge, with level JNDs below 1 dB, but speech is harder because its level fluctuates from moment to moment. In the study by Whitmer and Akeroyd, the stimuli were single words from the FAAF corpus and sentences from the ASL corpus (the study stimuli, kept under their corpus names). Their finding is that the larger JND for words and sentences than for steady noise held for both hearing-impaired and normal-hearing participants. The Whitmer and Akeroyd 2011 paper and the Florentine 1986 duration study carry the details.

MetricValueSource
Average slope of level-discrimination improvement with duration (tones at 250, 500 and 8,000 Hz, 2 ms to 2 s)-0.28, compared with -0.5 predicted for an optimum detectorFlorentine, JASA 1986
Longest duration up to which level discrimination keeps improvingat least 2 s, except at 250 Hz (between 0.5 and 1 s)Florentine, JASA 1986
Weber fraction 10 log(ΔI/I) at moderate levels for 30 ms high-frequency tones with bandstop noisecan be as large as 14 dBCarlyon and Moore, JASA 1984
Same high-frequency task with longer 225 ms tonesonly a small effect, compared with a large deterioration at 30 msCarlyon and Moore, JASA 1984
Intensity-difference limen for 6.5 kHz tones from 30 to 90 dB SPLan even more severe departure from Weber’s lawNizami, Perception and Psychophysics 2006
Level JND for broadband noise, normal-hearing listeners in the 2011 study (8 tested)0.8 dBWhitmer and Akeroyd, Ear and Hearing 2011
Level JND for speech, hearing-impaired listeners wearing hearing aids: single words vs sentences2.9 dB vs 2.2 dB (p < .05)Whitmer and Akeroyd, Ear and Hearing 2011

Context: the Carlyon and Moore figure is expressed as 10 log(ΔI/I), not as a level difference in dB SPL; the two scales converge only for large steps.

3. Infants, Children, Musicians and Other Species

Development is where the loudness JND moves most. Infants aged 7 to 9 months needed 3 to 12 dB to hear a 1 kHz tone get louder, against 1 to 2 dB for young adults, and not one of the 26 infants could discriminate a decrement, meaning a step down in level. School-age children are much closer to adults but still noisier: their psychometric functions, the curves linking step size to percent correct, were shallower than adults’.

Training does not seem to sharpen this particular skill. In a sample of 20 professional musicians and 20 nonmusicians, all aged 65 to 84, musicians beat nonmusicians on four of six auditory tasks, yet intensity discrimination was not one of them. Across species, monkeys, chinchillas and even goldfish land in the same few-decibel range as humans. See the Sinnott and Aslin infant study and the Grassi et al. musician study.

MetricValueSource
Intensity difference limen at 1 kHz: infants aged 7 to 9 months (n = 26) vs young adults (n = 6)3 to 12 dB vs 1 to 2 dBSinnott and Aslin, JASA 1985
Infants who successfully discriminated intensity decrementsnoneSinnott and Aslin, JASA 1985
Intensity discrimination threshold at 500 Hz, 65 dB SPL, in 10 log(ΔI/I): children aged 4.9 to 9.4 years (n = 16) vs adults (n = 11)0.20 dB vs -4.96 dBBuss, Hall and Grose, JASA 2009
Difference in psychometric function slope, adults vs children (children shallower)0.72 on average (t24 = 3.45, p <0.01)Buss, Hall and Grose, JASA 2009
Intensity discrimination at 1 kHz, 70 dB SPL: 20 professional musicians vs 20 nonmusicians aged 65 to 84no difference, F(1,38) < .01, p = .96Grassi et al., PLoS ONE 2017
Intensity difference limens at 1 kHz: monkeys vs humans1.5 to 2.0 dB vs 1.0 to 1.8 dBSinnott, Petersen and Hopp, JASA 1985
Chinchilla intensity difference limen near threshold vs at the highest level testedapproximately 8 dB vs approximately 3.5 dBSaunders, Shivapuja and Salvi, JASA 1987
Goldfish intensity thresholds at 55 dB sensation level, 100 to 1,600 Hz1.44 to 2.2 dB (0.7 dB higher on average at 35 dB sensation level)Fay, JASA 1989

Note on the Buss et al. units: converting with ΔL = 10 log10(1 + 10^(x/10)), the child mean of 0.20 dB corresponds to ΔI/I = 1.047 and a level step of about 3.1 dB, and the adult mean of -4.96 dB corresponds to ΔI/I = 0.319 and a level step of about 1.2 dB (our calculation). Most recent available data: the infant and primate figures are from 1985.

4. Hearing Loss and the Clinic: From the SISI Test to Implant Surgery

Clinicians have used the loudness JND as a diagnostic tool for decades. The Short Increment Sensitivity Index (SISI) rests on the observation that ears with cochlear damage often hear small level changes unusually well, because of recruitment (loudness growing abnormally fast above an elevated threshold). A SISI score of 70 to 100% on 1 dB increments is read as consistent with cochlear pathology, though the Interacoustics guide notes that low scores do not rule it out and that the test should never be the only site-of-lesion assessment.

A 2026 study pushes the same logic into the operating room. Among cochlear implant candidates with a median pure-tone average of 68 dB, a 5 dB difference was detected in 82 of 85 tests (96%), which supports using a 5 dB loudness cue to monitor residual hearing during awake implantation. Hidden damage can also show up here: Epp and colleagues (JASA 2012) reported increased intensity discrimination thresholds in tinnitus patients whose audiograms were normal, a pattern that overlaps with cochlear synaptopathy and hidden hearing loss statistics. Sources: the Interacoustics SISI guide and Khandalavala et al. 2026.

MetricValueSource
SISI protocol: increment size, duration and base tone1 dB increments of 200 ms on a tone at 20 dB sensation level, 20 presentationsInteracoustics clinical guide (vendor documentation)
SISI score range read as consistent with cochlear pathology and recruitment70 to 100%Interacoustics clinical guide (vendor documentation)
SISI score range read as normal hearing, conductive or retrocochlear loss0 to 20% (20 to 70% inconclusive)Interacoustics clinical guide (vendor documentation)
Hearing-impaired adults (n = 38, mean age 67.2 years, average losses 26 to 94 dB HL): level JND with vs without hearing aidsno significant differenceWhitmer and Akeroyd, Ear and Hearing 2011
Cochlear implant candidates (20 ears): tests in which a 5 dB difference was detected at 10 dB above threshold82 of 85 (96%)Khandalavala et al., Otology and Neurotology 2026
Same candidates: loudness adaptation reported at 10 dB vs 20 dB above threshold55% (n = 11) vs 40% (n = 8)Khandalavala et al., Otology and Neurotology 2026
Hearing-aid acclimatization: four listeners with average aid gain of 20 dB at 3 kHza change in intensity discrimination appeared only for the 3 kHz band, at levels above 85 dB SPLRobinson and Gatehouse, JASA 1995
Adults with deafness in one ear (n = 11): intensity discrimination in the hearing earenhancedMaslin, Taylor, Plack and Munro, JASA 2015

Context: SISI cut-offs differ slightly between manuals; the Interacoustics guide itself says interpretation guidance varies between sources.

5. Cochlear Implants: When Loudness Has Only a Few Steps

Electric hearing compresses the whole loudness range into a narrow window of current, so the JND matters far more than in acoustic hearing. The landmark measurement in eight implanted subjects found that the cumulative number of discriminable intensity steps across the electric dynamic range ranged from as few as 6.6 to as many as 45.2. For a listener at the low end, the entire range from barely audible to uncomfortably loud holds only a handful of distinguishable loudness levels.

Processing choices move that number. Raising the pulse rate from 500 to 3,000 pulses per second can increase the loudness JND by 60% near the top of the range, the equivalent of losing 38% of the discriminable steps. Simulations with normal-hearing listeners suggest a floor of roughly 5 dB of dynamic range and eight steps for good speech identification. For how implants pair with a hearing aid in the other ear, see bimodal cochlear implant statistics. Primary sources: Nelson et al. 1996 and Azadpour, McKay and Svirsky 2018.

MetricValueSource
Sensitivity to intensity change in dB (more negative = finer), electric hearing in 8 implanted subjects vs acoustic hearing-0.42 to -13.5 dB vs +0.60 to -3.34 dBNelson et al., JASA 1996
Cumulative discriminable intensity steps across the electric dynamic range6.6 to 45.2Nelson et al., JASA 1996
Average improvement in Weber fractions across the electric dynamic range8 dBNelson et al., JASA 1996
Increase in loudness JND when pulse rate rises from 500 to 3,000 pulses per second (upper dynamic range)can be 60%Azadpour, McKay and Svirsky, JARO 2018
Equivalent reduction in the number of discriminable steps38%Azadpour, McKay and Svirsky, JARO 2018
Difference limens for a level change spread across two electrodes vs within one electrode, 9 users (11 ears)nearly 3 times largerMcKay, Rickard and Henshall, JARO 2018
Correlation of those across-electrode limens with speech recognitionr = 0.67, P = 0.019McKay, Rickard and Henshall, JARO 2018
Vocoder simulation with 7 normal-hearing listeners: conditions that kept performance highat least 5 dB dynamic range and eight or more discriminable stepsStafford et al., Ear and Hearing 2014

Outliers: children with cochlear implants performed significantly worse than typically developing peers for a 4,000 Hz warble tone (Tak and Yathiraj, Journal of International Advanced Otology 2019, n = 30). In implanted monkeys, Pfingst, Burnett and Sutton (JASA 1983) measured electric difference limens as low as 0.5 dB near the upper limit, a reminder that animal electric data may not transfer directly to human users. Most recent available data for the step counts: Nelson et al. 1996.

6. Hearing-Aid Fine-Tuning: The JND Behind Every Gain Click

Every fitting-software click assumes the patient can hear the change. One research team measured that assumption directly in hearing-impaired listeners and found that the ear is far less sensitive to a gain change in one frequency band than to a change in overall level. Their results suggest noticeable steps of 3 dB for clinician fine-tuning and 5 dB for self-fitting, so smaller adjustments risk going unheard.

Speech makes the problem harder than noise. With real sentences, single octave-band changes had to reach 6.3 to 9.6 dB before they were noticed, and when 32 experienced hearing-aid users were asked whether a change made sound ‘better’ or ‘worse’, the median threshold for any preference was 4 to 12 dB. Fine adjustments a clinician makes in response to a vague complaint may simply be inaudible. Consumer devices face the same limit, which is one reason to read OTC hearing aid return rate statistics alongside this section. Sources: gain increments in speech and preference thresholds, 2021.

MetricValueSource
Gain-increment JND in speech-shaped noise, 38 hearing-impaired listeners: single frequency bands (excluding 0.25 kHz) vs the 0.25 kHz low-pass band2.8 dB vs 4.5 dBCaswell-Midwinter and Whitmer, Trends in Hearing 2019 (noise study)
Suggested noticeable step: clinician fine-tuning vs self-fitting3 dB vs 5 dBCaswell-Midwinter and Whitmer, Trends in Hearing 2019 (noise study)
Gain-increment JND in sentences, 41 hearing-impaired listeners: octave bands at 0.25, 1 and 4 kHz6.3, 6.7 and 9.6 dBCaswell-Midwinter and Whitmer, Trends in Hearing 2019 (speech study)
Same study: broadband gain-increment JND (interquartile range)2.0 dB (1.5 to 2.7)Caswell-Midwinter and Whitmer, Trends in Hearing 2019 (speech study)
Median preference thresholds for ‘better’ or ‘worse’ judgments, 32 adults aged 51 to 75 with hearing-aid experience4 to 12 dB, increasing with frequencyCaswell-Midwinter and Whitmer, International Journal of Audiology 2021

Context: these JNDs were measured at d’ of 1, a minimal discriminability criterion; a stricter criterion gives larger steps.

7. Standards, Meters and the “1 dB Rule” Examined

Rules of thumb put the loudness JND near 1 dB. The Praat phonetics manual summarizes Jesteadt’s data as “humans can detect an intensity difference of 1 phon”, and loudness meters treat 1 LU as exactly 1 dB. Against that yardstick, the 2023 revision of ISO 226 moved the equal-loudness contours by 0.6 dB at most, a change within the 0.5 to 1.2 dB range of steps implied by the Jesteadt fit in section 1.

Loudness models are a different story. When three standardized methods were checked against the new ISO 226:2023 contours, their root-mean-square (RMS) errors ranged from 1.57 to 5.55 dB, so the best model sits near one JND and the worst sits several JNDs away. Meter accuracy, by contrast, is held to a tenth of that. The practical lesson: a 0.5 LU difference between two masters is real for the meter but borderline for the ear. Game teams face the same tolerance question, covered in game loudness LUFS compliance statistics. Sources: Suzuki, Takeshima and Kurakata 2024, Sottek et al. 2023, EBU Tech 3341 and the Praat manual.

MetricValueSource
First-approximation intensity JND, in loudness-level unitsabout 1 phonPraat manual (Paul Boersma, University of Amsterdam), summarizing Jesteadt et al. 1977
Maximum difference between the ISO 226:2023 and ISO 226:2003 equal-loudness contours0.6 dB at mostSuzuki, Takeshima and Kurakata, Acoustical Science and Technology 2024
Lowering of the 20 Hz hearing threshold taken from ISO 389-7:2019 into the revision0.4 dBSuzuki, Takeshima and Kurakata, Acoustical Science and Technology 2024
ISO 226:2023 publication date, length and predecessorMarch 2023, 20 pages, replaces ISO 226:2003DIN Media listing for ISO 226:2023
RMS deviation from ISO 226:2023 contours, 100 Hz to 10 kHz at 20 to 80 phon: ISO 532-1 / ISO 532-3 / ECMA-418-25.55 / 3.04 / 1.57 dBSottek et al., Forum Acusticum 2023
RMS deviation of ISO 532-1 from the older ISO 226:1987 contours2.84 dBSottek et al., Forum Acusticum 2023
Tolerance for loudness meter readings on EBU Mode test signalsplus or minus 0.1 LUEBU Tech 3341 v4, November 2023
Loudness unit vs decibel; momentary and short-term meter windows1 LU = 1 dB; 0.4 s and 3 sEBU Tech 3341 v4, November 2023

Context: ECMA-418-2 is the Sottek Hearing Model, named after the first author of the comparison paper, so this ranking is not an independent test. “Phon” is the unit of loudness level, equal to the dB SPL of a 1 kHz tone judged equally loud.

Summary: Loudness Just-Noticeable Difference by the Numbers

MetricValueSource
Fitted Weber fraction for pulsed tones, 200 to 8,000 HzΔI/I = 0.463 (I/I0)^-0.072Jesteadt, Wier and Green, JASA 1977
Implied level step at 20 vs 80 dB sensation level (our calculation)about 1.2 dB vs about 0.5 dBDerived from Jesteadt et al. 1977
Typical near-miss exponent (1.0 = Weber’s law)on the order of 0.9Parker and Schneider, Perception and Psychophysics 1980
Frequency dependence of the level steplargely flat up to about 4 kHz, rising aboveFlorentine, Buus and Mason, JASA 1987
Weber fraction for 30 ms high-frequency tones in bandstop noisecan be as large as 14 dBCarlyon and Moore, JASA 1984
Average slope of level-discrimination improvement with duration-0.28 (optimum detector: -0.5)Florentine, JASA 1986
Intensity difference limen at 1 kHz, infants vs adults3 to 12 dB vs 1 to 2 dBSinnott and Aslin, JASA 1985
Psychometric slope difference, adults vs children aged 4.9 to 9.40.72 on averageBuss, Hall and Grose, JASA 2009
Older musicians vs nonmusicians, intensity discriminationno difference, p = .96Grassi et al., PLoS ONE 2017
Intensity difference limens at 1 kHz, monkeys vs humans1.5 to 2.0 dB vs 1.0 to 1.8 dBSinnott, Petersen and Hopp, JASA 1985
SISI score range consistent with cochlear pathology70 to 100%Interacoustics clinical guide (vendor documentation)
Cochlear implant candidates detecting a 5 dB step82 of 85 tests (96%)Khandalavala et al., Otology and Neurotology 2026
Discriminable intensity steps across the electric dynamic range6.6 to 45.2Nelson et al., JASA 1996
Loudness JND increase from 500 to 3,000 pulses per secondcan be 60%Azadpour, McKay and Svirsky, JARO 2018
Across-electrode limens vs speech recognitionr = 0.67, P = 0.019McKay, Rickard and Henshall, JARO 2018
Octave-band gain JNDs in sentences, 0.25 / 1 / 4 kHz6.3 / 6.7 / 9.6 dBCaswell-Midwinter and Whitmer, Trends in Hearing 2019
Maximum ISO 226:2023 vs 2003 contour change0.6 dBSuzuki, Takeshima and Kurakata, Acoustical Science and Technology 2024
RMS model deviation from ISO 226:2023, ISO 532-1 / ISO 532-3 / ECMA-418-25.55 / 3.04 / 1.57 dBSottek et al., Forum Acusticum 2023
Loudness meter tolerance on EBU Mode test signalsplus or minus 0.1 LUEBU Tech 3341 v4, 2023

Methodology and Sources

We searched journal databases (Europe PMC, OpenAlex), publisher pages and standards bodies for measured loudness JNDs, intensity difference limens and level-discrimination thresholds, and kept only figures we could read in an abstract, a full text or an official document. Values are reported in the units each source uses (dB level step, 10 log(ΔI/I), percent, or LU); where we converted units, the math is shown in the note under the table.

Last updated: November 8, 2026. We update this roundup quarterly, and the next refresh is expected when the next edition of the ISO 532 loudness calculation series is published.

Try VoxBooster — 3-day free trial.

Real-time voice cloning, soundboard, and effects — wherever you already talk.

  • No credit card
  • ~30ms latency
  • Discord · Teams · OBS
Try free for 3 days