64% of listeners in HARMAN’s headphone research preferred a headphone tuned close to the Harman target curve, the frequency response recipe that has become the default reference for headphone tuning since it was first presented in 2013 (GRAS Sound and Vibration, Harman case study 2021). The same research program built a statistical model that predicts in-ear headphone ratings from measurements at r = 0.91 (Harman patent US11197095B2, 2021), and its multi-country test drew 283 listeners from Canada, the United States, Germany and China (HARMAN Professional, AKG research overview 2023). Since 2022, newer studies have tied or challenged it, from a 2024 in-ear test where it shared first place to a 2026 Stanford CCRMA capstone that reported a rival target preferred in 82.3% of weighted comparisons. We aggregated data from Acoustics Today, the Journal of the Audio Engineering Society, the Journal of the Acoustical Society of America, HARMAN’s own patent and research pages, SenseLab (FORCE Technology), Knowles, RTINGS.com and Stanford CCRMA. For hardware context on in-ear designs, see our in-ear monitor statistics.
A quick definition used throughout: a target curve is the frequency response a headphone should measure on a standard test fixture (an artificial ear or head and torso simulator) to sound right. The older reference, the diffuse-field (DF) target, assumes sound arriving equally from all directions; the Harman target instead mimics loudspeakers in a typical listening room, with more bass.
TL;DR
- 64% of listeners formed the Harman Target Lovers segment, 15% wanted more bass and 21% less bass (GRAS Sound and Vibration, Harman case study 2021).
- HARMAN’s 2018 validation pitted the target against 31 headphone models from 18 manufacturers, priced $60 to $4,000 (HARMAN Professional, AKG research overview 2023).
- HARMAN’s in-ear preference model predicts ratings at r = 0.91, with an RMSE of 5.505 points on a 100-point scale (Harman patent US11197095B2, 2021).
- The AKG K371, released in July 2018, was the first AKG Professional headphone tuned to the Harman target (Copper Magazine, Olive interview 2023).
- 283 headphones measured: neither response nor estimated quality was related to price (Breebaart, JASA 2017).
- Young listeners with normal hearing preferred 12 dB more treble near 16 kHz than the Harman in-ear curve; the oldest group wanted 21 dB (Knowles white paper 2022).
- The Harman IE 2019 target and the SoundGuys curve tied at 65.69 and 65.97 out of 100 in a 2024 test (Headphonesty, reporting Olive AES 2024).
- 4 of 5 target curves were statistically tied among 35 listeners; only a low-bass curve clearly lost (RTINGS.com listening study 2024).
- A 2026 capstone reported PEQdB Diamond beta preferred over Harman IE 2019 in 82.3% (in-ear) and 84.4% (over-ear) of weighted comparisons (Stanford CCRMA, Chala 2026).
- For spatial audio, a flat response beat the Harman target, which still won for stereo music (Engel et al., JAES 2022).
1. How the Harman Target Was Built: Seven Years of Blind Tests
The Harman target is unusual among audio standards because it came from preference data, not from a physical model of the ear. The sample grew from a 10-person pilot to a 283-person test in four countries, and in the 2018 validation listeners on average preferred the target-equalized version in 28 of the models tested. Commercial models that already measured close to the target were rated about equally, which is what a valid target should produce.
The most consequential design choice is the bass shelf: the Harman target carries 5 to 6 dB more bass than the diffuse-field and free-field targets that headphone standards recommended as of 2022. That single decision is why many reviewers describe Harman-tuned headphones as warmer than older studio references. Most recent available data: Acoustics Today, Olive 2022 and HARMAN Professional 2023, summarizing AES papers from 2013 to 2019.
| Metric | Value | Source |
|---|---|---|
| Trained listeners in the first Harman target pilot test | 10 | HARMAN Professional, AKG research overview 2023 |
| Listeners in the multi-country follow-up test | 283, from Canada, the United States, Germany and China | HARMAN Professional, AKG research overview 2023 |
| Headphones in the 2018 validation study | 31 models from 18 manufacturers, priced $60 to $4,000, rated by 130 listeners | HARMAN Professional, AKG research overview 2023 |
| Commercial models where listeners preferred the Harman-equalized version over the original | 28 of 31 (see Data watch) | HARMAN Professional, AKG research overview 2023 |
| Extra bass in the Harman target compared with diffuse-field and free-field targets | 5 to 6 dB | Acoustics Today, Olive 2022 |
| Extra bass in the in-ear target compared with the around-ear target | 4 dB | Harman patent US11197095B2, 2021 |
| Duration of HARMAN’s headphone sound-quality research program | 7 years, 2012 to 2019 | Copper Magazine, Olive interview 2023 |
Context: the program was run by HARMAN, which also sells AKG and JBL headphones, and much of the listener pool was trained HARMAN staff. The independent studies in section 6 matter partly because they test the curve outside that setting.
2. Listener Segments: Who Prefers the Target and Who Does Not
The headline 64% hides a clear pattern. Preferred bass level is the main feature that sorted listeners into classes, and the minority classes are not random: 50% of listeners aged 50 and over fell into the Less Bass Is Better class, compared with 13% of listeners in their 30s. Women were about three times as likely as men to land there.
Training also mattered. Not one trained listener fell into the Less Bass class, while 30% of trained listeners wanted more bass. The practical takeaway HARMAN drew is that a single target plus a bass control or app setting covers most people better than separate products for each group. Most recent available data: Acoustics Today, Olive 2022, Table 1, a cluster analysis of HARMAN’s earlier listening tests.
| Metric | Value | Source |
|---|---|---|
| Listeners in each segment: Harman Target Lovers, More Bass Is Better, Less Bass Is Better | 64%, 15%, 21% | GRAS Sound and Vibration, Harman case study 2021 |
| Bass and treble preferred by the Less Bass Is Better segment, relative to the Harman curve | 2 to 3 dB less bass and 1 dB more output above 1 kHz | GRAS Sound and Vibration, Harman case study 2021 |
| Men vs women in the Less Bass Is Better class, and in the More Bass Is Better class | Less Bass: 13% vs 40%; More Bass: 18% vs 4% | Acoustics Today, Olive 2022 |
| Trained vs untrained listeners in the Less Bass Is Better class | 0% vs 25% | Acoustics Today, Olive 2022 |
| Harman Target Lovers among listeners in their 30s vs listeners aged 50 and over | 74% vs 30% | Acoustics Today, Olive 2022 |
| Listeners aged 50 and over in the Less Bass Is Better class | 50% | Acoustics Today, Olive 2022 |
Outlier note: the article itself names the small number of female listeners as a legitimate criticism, so the 40% figure rests on a thin subgroup.
3. Predicting Preference from Measurements: How Accurate the Models Are
If a measurement can stand in for a listening panel, a headphone maker can check a design in seconds instead of weeks. HARMAN’s models do that by scoring how far a headphone’s response deviates from the target, using the spread (standard deviation) and tilt (absolute slope) of that error curve. The in-ear model correlates with real ratings at r = 0.91, and it held up when tested against six separate listening tests.
An r of 0.91 means the model explains most, but not all, of what listeners hear: its typical error (RMSE, root mean square error) is about 5.5 points on a 100-point scale, enough to separate a good headphone from a poor one but not two close competitors. These are the model’s fit on HARMAN’s own data rather than an independent replication. Most recent available data: Harman patent US11197095B2, granted 2021, and the AES 143rd Convention best paper announcement, 2017.
| Metric | Value | Source |
|---|---|---|
| In-ear model correlation with measured ratings, and typical prediction error | r = 0.91 (R² = 0.819), RMSE 5.505 points on a 100-point scale | Harman patent US11197095B2, 2021 |
| Around-ear and on-ear model correlation and error | r = 0.86, RMSE 6.705 | Harman patent US11197095B2, 2021 |
| In-ear model correlation across six validation listening tests | r = 0.89 to 0.98 | Harman patent US11197095B2, 2021 |
| In-ear model with 10 of 29 headphones removed (10 repeats) | mean r = 0.90, range 0.83 to 0.95 | Harman patent US11197095B2, 2021 |
| Listeners and in-ear models behind the in-ear model (AES best paper 2017) | 71 trained and untrained listeners, 30 in-ear models | AES 143rd Convention best paper announcement 2017 |
Context: the around-ear and on-ear model uses only two predictors (error spread and slope); a third, the mean error, was dropped because it added little.
4. How Far Commercial Headphones Drift from the Target, and Why Price Does Not Fix It
The largest independent measurement set is still Jeroen Breebaart’s 283 headphones. Price had no meaningful relationship with how close a headphone measured to the target: in the analysis of retail price, deviation from the target was not significant (p > 0.10), while headphone type was. Spending more buys slightly more consistent responses between models, not a more accurate one.
Type matters more than budget. Around-ear models land closest to the Harman target and on-ear models furthest, a ranking that HARMAN’s own model reached independently on 158 headphones. Most recent available data: Breebaart, Journal of the Acoustical Society of America 2017.
| Metric | Value | Source |
|---|---|---|
| Headphones measured (in-ear, supra-aural and circumaural) | 283 | Breebaart, JASA 2017 |
| Effect on retail price in a two-way ANOVA: headphone type vs deviation from the target curve (RMSE) | type: F(279, 2) = 21.05, p < 0.0001; deviation: F(279, 1) = 2.59, p > 0.10 | Breebaart, JASA 2017 |
| Spread of responses between headphones (RMSD across frequency), cheapest price quartile vs most expensive quartile | 4.3 dB vs 3.5 dB | Breebaart, JASA 2017 |
| Extra bass in in-ear headphones at 100 Hz compared with other types | about 4 dB | Breebaart, JASA 2017 |
| Headphones scored by HARMAN’s two models, with the type ranking | 158; around-ear closest to target, on-ear worst, in-ear in between | Acoustics Today, Olive 2022 |
| Extra bass in JBL-branded headphones compared with the target | about 2 dB | Copper Magazine, Olive interview 2023 |
Context: Breebaart also found that six basis functions explain 90% of the variation between all 283 responses, the first resembling published target curves. In other words, most headphones are variations on a few shapes. HARMAN’s own JBL line sits about 2 dB above the target in bass, closer to the taste of the More Bass segment from section 2.
5. Bass and Treble Preference Shifts by Training, Gender and Hearing Age
Small decibel numbers carry real weight here: a 2 to 3 dB tilt is clearly audible. Untrained listeners pushed both ends up, and women pulled both ends down. The biggest shift is in the top octave: Knowles found young listeners preferred 12 dB more treble near 16 kHz than the Harman in-ear curve, a region the original HARMAN tests could not reliably cover above 10 kHz.
Hearing age drives the treble result. Knowles measured each listener’s hearing and found preference climbed about 3 dB for every 10 years of equivalent hearing age, which links this topic to age-related hearing decline (see our presbycusis progression statistics). Knowles makes the balanced armature tweeters its paper recommends, so treat the size of the boost as a vendor finding, though the work was later peer reviewed in the Journal of the Audio Engineering Society (Miller and Downey 2023). Most recent available data: Knowles white paper 2022 and Harman findings reported in 2016.
| Metric | Value | Source |
|---|---|---|
| Untrained vs trained listeners | 1 to 2 dB more bass and 2 to 3 dB more treble | Peter Mapp, Sound and Communications 2016, reporting HARMAN research |
| Women vs men | about 1 dB less bass and 2 dB less treble | Peter Mapp, Sound and Communications 2016, reporting HARMAN research |
| Headphones vs in-room loudspeakers, same listeners | about 2 dB less bass and 3 dB less treble on headphones | Peter Mapp, Sound and Communications 2016, reporting HARMAN research |
| Bass boost preferred by the More Bass Is Better segment | 4 to 6 dB above the target | Copper Magazine, Olive interview 2023 |
| Second listener class in a 2024 in-ear test vs the first class | 2 dB less bass and 2 dB more treble | Olive, AES Show 2024 abstract |
| Preferred boost near 16 kHz vs the Harman in-ear curve: young normal-hearing listeners vs oldest hearing-age group | 12 dB vs 21 dB (nearly 70 listeners) | Knowles white paper 2022 (vendor research) |
| Added treble boost per 10 years of equivalent hearing age | about 3 dB | Knowles white paper 2022 (vendor research) |
| Listeners whose preferred boost fell within 5 dB of their group average | 95% | Knowles white paper 2022 (vendor research) |
Context: the Knowles result does not contradict HARMAN below 10 kHz; it kept the Harman curve there and only tested the 10 to 20 kHz region.
6. Newer Targets and Independent Tests, 2022 to 2026
The newer studies change the question from “is Harman right” to “how many curves are equally good”. Moving to the more accurate B&K 5128 head and torso simulator, SenseLab asked 56 naive listeners in Denmark and Japan to rate 32 curves, and in a 2024 HARMAN test the Harman IE 2019 target and the SoundGuys curve finished 0.28 points apart out of 100, a statistical tie. RTINGS.com found 4 of 5 curves tied.
The strongest challenge comes from a 2026 Stanford CCRMA capstone, which reported its PEQdB Diamond beta target was preferred over Harman IE 2019 in more than 8 of 10 weighted comparisons. Its author founded PEQdB and its listeners took an online test identified only by browser cookies, so it needs independent replication. Meanwhile, spatial audio may need a different target entirely: Meta Reality Labs (Engel et al., JAES 2022) found a flat response preferred for spatial content. The SenseLab ratings are public as an open Zenodo dataset.
| Metric | Value | Source |
|---|---|---|
| Naive listeners and curves in the SenseLab B&K 5128 study | 56 listeners (30 Danish, 26 Japanese) rating 32 curves from 8 closed headphones | Ravizza et al., SenseLab Zenodo dataset 2023 |
| Mean preference, SoundGuys curve vs Harman IE 2019 (32 listeners) | 65.97 vs 65.69 out of 100, statistically tied | Headphonesty 2024, reporting Olive AES 2024 |
| Listeners in the class preferring the top two in-ear targets | 72% (Class 2: 28%) | Olive, AES Show 2024 abstract |
| Curves statistically tied in a 35-listener double-blind test | 4 of 5; the low-bass curve lost (p = 0.00001 vs the tilted diffuse-field curve) | RTINGS.com listening study 2024 |
| Participants who picked the same favorite curve on both music clips | 63% | RTINGS.com listening study 2024 |
| Weighted comparisons won by PEQdB Diamond beta over Harman IE 2019, in-ear vs over-ear | 82.3% vs 84.4% (704 in-ear and 489 over-ear listeners) | Stanford CCRMA, Chala 2026 |
| Adaptive target search: odds the best-ranked curve beat the starting curve | 1.29 times (z = 2.72, p = .018) | Ravizza et al., arXiv 2025 |
| Spatial vs stereo content study design | 3 tests, 7 target responses, more than 20 listeners per test | Engel et al., JAES 2022 |
Outlier note: the RTINGS and 2024 HARMAN panels were company employees (6 of 35 RTINGS participants were women; 25 of 32 HARMAN listeners were male, per SoundGuys 2024). The Engel et al. and SenseLab rows are older than three years. Most recent available data: Engel et al., JAES 2022, and Ravizza et al., SenseLab dataset 2023. Sources: Stanford CCRMA capstone, RTINGS.com study, AES Show 2024 abstract, arXiv 2512.08313.
7. Industry Adoption and the Measurement Tools Behind It
Adoption moved from HARMAN’s own catalog to the review ecosystem. The AKG K371 in July 2018 was the first AKG Professional headphone built to the target, and by August 2025 RTINGS.com let readers compare headphones against 6 reference curves, with the Harman target now one option among several rather than the default.
The tools are changing too. The B&K 5128 head and torso simulator reproduces what a headphone sees below 100 Hz and above 8 to 10 kHz more accurately than older ear simulators, which is why targets are being re-derived for it (fixture design is covered in our dummy head binaural microphone statistics). Sources: RTINGS.com Test Bench 2.0 and audioXpress 2025.
| Metric | Value | Source |
|---|---|---|
| First AKG Professional headphone tuned to the Harman target | AKG K371, released July 2018 | Copper Magazine, Olive interview 2023 |
| HARMAN-brand AKG and JBL headphones designed to or close to the curve | nearly all | GRAS Sound and Vibration, Harman case study 2021 (vendor claim) |
| Selectable bass levels on the Mark Levinson No. 5909 headphone | 3 | Copper Magazine, Olive interview 2023 |
| Curves in the RTINGS.com comparison tool (Test Bench 2.0) | 6: RTINGS, Harman in-ear, two SoundGuys curves, SenseLab/Aizu, plus the B&K 5128 diffuse field as a reference | RTINGS.com Test Bench 2.0, 2025 |
| Sound signature categories added for sorting headphones | 7 | RTINGS.com Test Bench 2.0, 2025 |
| Registered participants, AES Headphone Technology conference, 2025 vs 2016 | about 200 vs 160 | audioXpress 2025 |
| New artificial ear for the B&K 5128 announced at the 2025 conference | HBK Type 4621, capable to 40 kHz | audioXpress 2025 |
Context: the SenseLab/Aizu curve in RTINGS’s tool is the average of the five highest-rated curves from the 56-listener study in section 6, a direct path from published research to a consumer review site. The HARMAN product rows are older than three years. Most recent available data: Copper Magazine, Olive interview 2023, and GRAS Sound and Vibration, Harman case study 2021.
Summary: Headphone Harman Target Curve Preference by the Numbers
| Metric | Value | Source |
|---|---|---|
| Listeners in the Harman Target Lovers segment | 64% | GRAS Sound and Vibration, Harman case study 2021 |
| Listeners in the multi-country Harman test | 283 in 4 countries | HARMAN Professional, AKG research overview 2023 |
| Preference model accuracy, in-ear model vs around-ear and on-ear model | r = 0.91 vs r = 0.86 | Harman patent US11197095B2, 2021 |
| Listeners aged 50 and over in the Less Bass Is Better class | 50% | Acoustics Today, Olive 2022 |
| Bass boost preferred by the More Bass Is Better segment | 4 to 6 dB | Copper Magazine, Olive interview 2023 |
| Untrained vs trained listeners, treble | 2 to 3 dB more treble | Peter Mapp, Sound and Communications 2016 |
| Headphones measured with no price effect | 283 | Breebaart, JASA 2017 |
| Preferred 16 kHz boost over the Harman in-ear curve, young vs oldest group | 12 dB vs 21 dB | Knowles white paper 2022 |
| Treble boost per 10 years of hearing age | about 3 dB | Knowles white paper 2022 |
| SenseLab B&K 5128 study size | 56 listeners, 32 curves | Ravizza et al., SenseLab Zenodo dataset 2023 |
| Harman IE 2019 vs SoundGuys curve | 65.69 vs 65.97, tied | Headphonesty 2024, reporting Olive AES 2024 |
| Class preferring the top two in-ear targets | 72% | Olive, AES Show 2024 abstract |
| Curves statistically tied among 35 listeners | 4 of 5 | RTINGS.com listening study 2024 |
| PEQdB Diamond beta preferred over Harman IE 2019 (in-ear) | 82.3% of weighted comparisons | Stanford CCRMA, Chala 2026 |
| Reference curves in RTINGS.com’s comparison tool | 6 | RTINGS.com Test Bench 2.0, 2025 |
Methodology and Sources
- Acoustics Today, Sean E. Olive, The Perception and Measurement of Headphone Sound Quality: What Do Listeners Prefer? (Spring 2022, vol. 18, no. 1), including Table 1 on listener segments and the 158-headphone model results
- HARMAN Professional, Defining the Standard: The Science Behind AKG Reference Response Studio Headphones (2023)
- GRAS Sound and Vibration, The Harman Curve Verified by GRAS, case study (2021)
- Copper Magazine issue 189, Talking With Dr. Sean Olive of Harman International, Part One (2023)
- HARMAN International, patent US11197095B2, One-click measurement for headphones (granted 2021)
- AES 143rd Convention best peer-reviewed paper announcement, Olive, Welti and Khonsaripour (Sports Video Group, 2017)
- Breebaart, No correlation between headphone frequency response and retail price, Journal of the Acoustical Society of America 141(6) EL526 (2017)
- Peter Mapp, Sex and Audio, Sound and Communications reprint in Enjoy the Music (2016)
- Knowles, An Improved Target Response Curve for Insert Headphones, white paper (2022) and Miller and Downey, Listener Preferences for High-Frequency Response of Insert Headphones, JAES 71(10) (2023)
- Engel, Alon, Scheumann, Crukley and Mehra, On the Differences in Preferred Headphone Response for Spatial and Stereo Content, JAES 70(4) (2022)
- Ravizza, Villegas, Volk and Stegenborg-Andersen, Preference ratings and 32 magnitude frequency response curves, Zenodo dataset (2023)
- Olive, A comparison of in-ear headphone target curves for the B&K HATS Type 5128, AES Show 2024 abstract, with mean ratings as reported by Headphonesty (2024) and listener demographics as reported by SoundGuys (2024)
- RTINGS.com, One Curve Doesn’t Fit All, listening study (2024) and RTINGS.com Headphones Test Bench 2.0 changelog (2025)
- Ravizza et al., An Adaptive Method for Target Curve Selection, arXiv 2512.08313 (2025)
- Chala, An Optimized Headphone Target Response Curve, Stanford CCRMA MA/MST capstone (2026)
- audioXpress, A Report from the 2025 AES International Conference on Headphone Technology, Finland (2025)
- Data watch: HARMAN Professional (2023) and Acoustics Today (2022) both say that, in the 31-model study, listeners preferred the Harman-equalized version in 28 of the models and that four models close to the target were equally preferred; 28 plus 4 is 32, so the groups likely overlap or the count includes the target itself, and we do not use the 28 as a headline figure. Sources also disagree on segment sizes in decibels. GRAS (2021) puts the More Bass segment at 3 to 6 dB and the Less Bass segment at 2 to 3 dB, while Acoustics Today (2022) gives 4 to 6 dB and 2 dB; we quote each source as written. The patent text gives r = 0.86 for the around-ear and on-ear model, but its Table 2 prints R = 0.741 and R² = 0.861 (the two values look swapped), and it describes five validation tests while its table lists six. For the 2024 in-ear test, the AES abstract says the second class preferred 2 dB more treble, while SoundGuys and Headphonesty report 1 to 2 dB. The Knowles white paper analyzed 200 songs from the last 10 years of the Billboard chart, while a 2022 press summary said 20 years. The 2025 adaptive-method preprint reports two different standard-deviation figures (1.72 to 0.98 dB and 1.85 to 0.99 dB), so we left them out. The 2026 Stanford capstone was written by the founder of PEQdB, whose target it tests, with online listeners identified by browser cookies; a Zenodo record says a related manuscript was submitted to JAES. Olive’s 2025 AES paper on preferred in-ear targets by method of adjustment was not used because only its paywalled abstract was available. The AES e-library pages for the 2017 to 2019 HARMAN papers could not be opened, so their figures come from the patent, Acoustics Today and HARMAN’s own pages.
Last updated: October 24, 2026. We update this roundup quarterly, and the next refresh is expected when the Journal of the Audio Engineering Society publishes the Chala and Wong target-curve manuscript now under submission.