Manhattan restaurants average 78 dBA and bars 81 dBA, and 71% of the restaurants measured sit at levels not conducive to conversation (Acoustical Society of America, SoundPrint lay-language paper 2017). A nationally weighted survey analysed by CDC and NIOSH scientists found that 68.0% of US adults call a loud dining experience annoying and 64.8% are likely to avoid a noisy restaurant (CDC authors, The Hearing Journal 2022). The physics explains why rooms spiral upward: in day-long monitoring, talkers added an average 0.54 dB of voice for every 1 dB of background noise (Whittico et al., JASA 2020), and every raised voice becomes someone else’s background. We aggregated data from the Acoustical Society of America, CDC and NIOSH, Zagat, Action on Hearing Loss, the WHO, and peer-reviewed work in JASA, Scientific Reports, Ear and Hearing and Acta Acustica, plus the crowdsourced SoundPrint database. For the venue-side fixes that help listeners with hearing aids, see our hearing loop and assistive listening statistics.
TL;DR
- Manhattan restaurants average 78 dBA and bars 81 dBA in crowdsourced readings (ASA / SoundPrint, 2017).
- 50 restaurants in 10 US food cities averaged around 79 dB at peak Saturday dining hours (Oticon via Hearing Review, 2018).
- Noise was the most bothersome dining issue for 24% of diners, ahead of service at 23% (Zagat 2018 Dining Trends Survey).
- 64.8% of US adults are likely to avoid a noisy restaurant (CDC authors, SummerStyles 2021, Hearing Journal 2022).
- Lombard slope rises from 0.2 dB/dB at a 50 dB background to 0.5 dB/dB at 80 dB (Nijs et al., JASA 2008).
- Older adults’ voice rise steepens to 0.51 dB/dB above 58.3 dB(A) of restaurant noise (Bottalico et al., Scientific Reports 2022).
- A student canteen’s acoustic capacity was estimated at 189 people, about 72% table occupancy (Zelem et al., Acta Acustica 2023).
- 79% of UK Speak Easy respondents have left a restaurant, cafe or pub early because of noise (Action on Hearing Loss, 2016).
- In 59 loud NYC venues, the median and mean level was 92 dBA (Spira-Cohen et al., J Occup Environ Hyg 2017).
- 25% of Apple Hearing Study participants exceed the WHO daily environmental sound limit (Apple / University of Michigan, 2021).
1. Measured Decibel Levels in Restaurants and Bars
The striking pattern is not one loud outlier, it is the clustering. Averages from phone meters in Manhattan, mystery diners in ten US food cities and precision meters in Hong Kong all land between about 74 and 81 dBA for a busy service. 71% of Manhattan restaurants measured sat at levels not conducive to conversation (Acoustical Society of America, SoundPrint lay-language paper 2017), which means loudness is the default operating condition of a successful room, not a design accident.
Bars are a step louder again, and the difference matters more than the 3 dB gap suggests: in the same dataset the share of venues judged potentially dangerous to hearing roughly doubles from restaurants to bars.
| Metric | Value | Source |
|---|---|---|
| Average sound level, Manhattan restaurants / bars (Jul 2015 to Jun 2017) | 78 dBA / 81 dBA | ASA / SoundPrint lay-language paper 2017 |
| Bars at levels not conducive to conversation | 90% | ASA / SoundPrint lay-language paper 2017 |
| Restaurants / bars potentially dangerous to hearing health | approximately 31% / 60% | ASA / SoundPrint lay-language paper 2017 |
| Peak Saturday dining noise, 50 restaurants in 10 US cities | around 79 dB | Oticon study via Hearing Review 2018 |
| Noisiest venues in the Oticon sample | 81.92 and 82.19 dB | Oticon study via Hearing Review 2018 |
| Hong Kong restaurant noise, measured range | 66.7 to 82.6 dBA | To and Chung, Noise and Vibration Worldwide 2015 |
| Hong Kong restaurant noise, mean | 73.9 dBA (SD = 3.81 dBA) | To and Chung, Noise and Vibration Worldwide 2015 |
| Occupied noise levels across ten eating establishments | 45 to 82 dB(A) | Hodgson et al., JASA 2007 |
Context note: the Manhattan venues were each measured at least three times, Wednesday to Saturday between about 7 and 9 PM, and some downtown neighbourhoods (Flatiron, Gramercy, East Village, Lower East Side) averaged as high as 82 dBA. The Oticon mystery-diner study and the To and Chung Hong Kong measurements used different instruments and sampling windows, so treat the cross-city comparison as directional. Most recent available data: SoundPrint 2017, Oticon 2018, To and Chung 2015.
2. Diner Complaints and Avoidance
Noise and service keep trading first place as the thing diners dislike most, and the evidence now spans restaurant guides, consumer testing and a probability-sampled national panel. 64.8% of US adults say they are very or somewhat likely to avoid a noisy restaurant (CDC authors, SummerStyles 2021 analysis in The Hearing Journal 2022). The commercial reading is blunt: a loud room may turn tables faster, but it also filters out a large share of the market before anyone books.
The age gradient is where operators should look hardest. Adults aged 63 and older had 2.7 times the adjusted odds (aOR = 2.7) of saying they would avoid a noisy restaurant compared with adults aged 18-32, adults aged 49-62 had an aOR of 2.1, and women were also more likely than men to avoid one (aOR = 1.2), in the same analysis.
| Metric | Value | Source |
|---|---|---|
| Noise as most bothersome dining issue, US (nearly 13,000 diners) | 24% (service 23%, crowds 15%, high prices 12%, parking 10%) | Zagat 2018 Dining Trends Survey |
| Noise as chief complaint, Portland OR diners | 33% | Zagat State of American Dining 2016 (9,248 diners, 26 cities) |
| Noise complaints outweighing service, New York / Boston / San Francisco | 32% / 30% / 26% | Zagat State of American Dining 2016 |
| US adults who find a loud dining experience somewhat or very annoying | 68.0% | CDC authors, SummerStyles 2021 (Hearing Journal 2022) |
| US adults who feel restaurants are too loud half the time to always | 23.3% (never or seldom: 22.4%) | CDC authors, SummerStyles 2021 (Hearing Journal 2022) |
| Average share of chain dining visits flagged for excessive noise | 16% | Consumer Reports restaurant survey 2016 |
| UK public who left a venue because it was too noisy / would spend more for a quieter one | 49% / 75% | Action on Hearing Loss general public survey 2016 (n = 1,200) |
Context note: the SummerStyles panel had 4,085 completed responses (71.2% of 5,741 sampled) and was weighted to the US adult population. Zagat’s 2016 national ranking put service first and noise second, while its 2016 survey archive shows noise winning in several big cities. Most recent available data: Zagat 2018, Consumer Reports 2016, SummerStyles 2021.
3. The Lombard Effect: How Much Louder Each Diner Gets
The Lombard effect is the feedback loop at the centre of every loud restaurant: noise makes people talk louder, and louder talk is the noise. The key number is the slope, the decibels of voice added per decibel of background. The slope grows from about 0.2 dB/dB at a 50 dB background to 0.5 dB/dB at 80 dB (Nijs, Saher and den Ouden, JASA 2008), which explains why a half-full room feels fine and a full one suddenly does not.
The restaurant-specific study of older adults adds a useful threshold and a surprise: hearing loss did not statistically change the Lombard effect. People with and without hearing loss push their voices the same way; what differs is how much of the result they can understand (see section 5).
| Metric | Value | Source |
|---|---|---|
| Lombard slope at 50 dB / 80 dB background | 0.2 / 0.5 dB/dB | Nijs et al., JASA 2008 |
| Voice-level breakpoint, adults over 60 in restaurant noise; slope below / above it | 58.3 dB(A) (95% CI 53.4 to 63.3); 0.27 / 0.51 dB/dB | Bottalico, Piper and Legner, Scientific Reports 2022 |
| Average slope in ambulatory monitoring (vocally healthy women) | 0.54 dB/dB | Whittico et al., JASA 2020 |
| Predicted voice level at 50 / 65 / 80 dBA background | 72.8 / about 79.9 / 87.1 dB SPL | Whittico et al., JASA 2020 |
| Median background noise across participants’ monitored days | 58.3 dBA | Whittico et al., JASA 2020 |
| Noise level at which the Lombard effect starts | around 45 dB (speech level 55 dB) | Rindel, Acoustics in Practice 2019 |
Context note: Whittico et al. report that some of the loudest environments their participants logged were restaurants or bars, and that earlier studies put typical slopes in the range of 0.3 to 1 dB/dB. Lab and field slopes differ because real rooms add music, dishes and laughter on top of speech. Most recent available data: Bottalico 2022, Whittico 2020, Nijs 2008.
4. Crowd Density, Absorption, and Acoustic Capacity
Because voices respond to noise, room acoustics work harder in restaurants than in rooms with fixed sound sources. Classic theory predicts roughly 3 to 4 dB less noise when absorbing area doubles; with talkers in the loop, doubling the absorbing area cut noise by 5.5 dB in the Delft measurements (Nijs et al., JASA 2008). Absorption is a double win: it soaks up sound and it calms the voices producing it.
The more useful design concept is acoustic capacity, the maximum headcount a room can hold before conversation at 1 m becomes insufficient. It turns a decibel target into a seat count an owner can act on, and measured canteens show the ceiling is real (Zelem et al., Acta Acustica 2023).
| Metric | Value | Source |
|---|---|---|
| Noise reduction per doubling of absorbing area (constant number of talkers) | 5.5 dB | Nijs et al., JASA 2008 |
| Sound level rise per doubling of talkers | 3 dB (low density) to 6 dB (dense crowds) | Nijs et al., JASA 2008 |
| Maximum ambient noise for sufficient conversation at 1 m (SNR of -3 dB) | 71 dB | Rindel, Acoustics in Practice 2019 |
| Acoustic capacity rule of thumb | N max = V / (20 T), V in m3, T in seconds | Rindel, Acoustics in Practice 2019 |
| Typical fully occupied restaurant vs its acoustic capacity | exceeded by a factor of 2 or more | Rindel, Acoustics in Practice 2019 |
| Restaurant with 2.4 m2 or more absorbing surface per person | more than 3 dB rise per doubling of people | Zelem et al., Acta Acustica 2023 |
| Student canteen estimated acoustic capacity | 189 people (about 72% table occupancy) | Zelem et al., Acta Acustica 2023 |
| Canteen background noise (L_A,95) at peak occupancy | as high as 69 dB | Zelem et al., Acta Acustica 2023 |
Context note: in the Zelem canteen, people began communicating less once absorbing surface fell under 1.5 m2 per person (80 people), and the level stopped rising from 150 people onward, at 0.8 m2 per person. Rindel recommends keeping background music around 60 dB to 65 dB at most, since music feeds the same Lombard loop. Most recent available data: Zelem 2023, Rindel 2019.
5. Conversation Difficulty and Hearing Loss
For listeners with hearing loss, the restaurant problem is not volume alone but signal-to-noise ratio. A room that is merely tiring for a normal-hearing diner can be unworkable for someone who needs a few extra decibels of separation between the voice across the table and the babble behind it. Adults over 60 with moderate-to-severe hearing loss fell to 50% word recognition at about 50 dB(A) of background noise, versus about 60 dB(A) for peers with normal hearing (Bottalico et al., Scientific Reports 2022). That 10 dB gap means the same table can be workable for one diner and unworkable for the person sitting next to them.
The UK Speak Easy survey, run by a hearing-loss charity, shows what that means in behaviour: people do not complain, they leave and do not come back (Action on Hearing Loss, Access to cafes, pubs and restaurants policy statement 2018). The neighbouring problem of trouble hearing in noise despite a normal audiogram is covered in our cochlear synaptopathy and hidden hearing loss statistics.
| Metric | Value | Source |
|---|---|---|
| Respondents with difficulty holding a conversation due to background music / environmental noise | 79% / 81% | Action on Hearing Loss, Speak Easy 2016 |
| Respondents who left a venue early because of noise / would not return if noise is too high | 79% / 91% | Action on Hearing Loss, Speak Easy 2016 |
| Norwegians with impaired hearing reporting often/always difficulty conversing in canteens, restaurants and cafes | 51% (88% including sometimes) | NIBR 2011 survey, as reported in Rindel, Acoustics in Practice 2019 |
| Extra SNR needed by people with a slight hearing disorder (and non-native listeners) | approximately 3 dB | ISO 9921, as cited in Rindel, Acoustics in Practice 2019 |
| 50% word recognition point: normal / mild / moderate-to-severe hearing loss | about 60 / 55 / 50 dB(A) | Bottalico et al., Scientific Reports 2022 |
| Background level above which average conversational SNR can turn negative | approximately 69 dB SPL at 1 m (75 dB SPL at 0.5 m) | Weisser and Buchholz, JASA 2019 |
| Virtual-restaurant intelligibility at reverberation time 0.8 s / 1.8 s | 76.6% / 71.9% correct | Ellis, Crukley and Souza, Ear and Hearing 2024 |
Context note: the Ellis experiment compared 20 older listeners with impaired hearing against 20 listeners with clinically normal hearing, at SNRs from -8.5 to 11.5 dB; the normal-hearing group did better at low SNRs and the groups converged as SNR improved. Bottalico et al. conclude that noise below 50 dB(A) has minimal impact on intelligibility and vocal effort for older adults. Most recent available data: Speak Easy 2016, Bottalico 2022, Ellis 2024.
6. Hearing-Health Thresholds, Staff Exposure, and Crowdsourced Decibel Data
Most restaurant noise is a communication problem first and a hearing-damage problem second, but the loudest end of the market crosses into occupational-health territory. In a pilot study of 59 loud New York City restaurants, bars and clubs, the median and mean level were both 92 dBA (Spira-Cohen, Caffarelli and Fung, J Occup Environ Hyg 2017). For staff, who stay for a full shift rather than a two-hour dinner, the dose arithmetic is less forgiving.
Crowdsourced meters have turned these readings into a consumer dataset. SoundPrint now lists measurements from more than 150 countries (SoundPrint), and the Apple Hearing Study uses wearables to log ambient exposure at population scale. Musicians and venue workers face the same exposure curve, covered in our musician hearing loss and noise exposure statistics.
| Metric | Value | Source |
|---|---|---|
| Loud NYC venues with LAeq above 85 dBA during the visit | 80% (N = 47) | Spira-Cohen et al., J Occup Environ Hyg 2017 |
| Loud NYC venues where one visit exceeded the NIOSH daily noise dose | 49% (N = 29) | Spira-Cohen et al., J Occup Environ Hyg 2017 |
| Restaurant employee full-shift TWA exposures (180 measurements, 6 restaurants) | 69 to 90 dBA, mean 80 dBA | Green and Anthony, J Occup Environ Hyg 2015 |
| Cooks exceeding the NIOSH 85 dBA REL | 10.5% (mean 81 dBA) | Green and Anthony, J Occup Environ Hyg 2015 |
| NIOSH Recommended Exposure Limit | 85 dBA averaged over an eight-hour workday | CDC NIOSH, Noise and Occupational Hearing Loss |
| US workers exposed to hazardous noise at work each year | about 27 million | CDC NIOSH |
| Apple Hearing Study participants above the WHO daily environmental sound limit | 25% | Apple / University of Michigan Hearing Study update 2021 |
| SoundPrint crowdsourced coverage | 150+ countries, 1,600+ cities | SoundPrint website, accessed October 2026 |
Context note: clubs and lounges were louder than restaurants and bars in the NYC pilot (97 vs 91 dBA, p < 0.05). Green and Anthony measured day shifts only and note that evening exposures may be higher. Globally, the WHO projects nearly 2.5 billion people will have some degree of hearing loss by 2050 and estimates over 1 billion young adults are at risk from unsafe listening (WHO Deafness and hearing loss fact sheet 2026). Most recent available data: Spira-Cohen 2017, Green and Anthony 2015, Apple 2021.
Summary: Restaurant Noise and the Lombard Effect by the Numbers
| Metric | Value | Source |
|---|---|---|
| Average Manhattan restaurant / bar sound level | 78 dBA / 81 dBA | ASA / SoundPrint 2017 |
| Peak Saturday noise, 50 restaurants in 10 US cities | around 79 dB | Oticon via Hearing Review 2018 |
| Hong Kong restaurant mean noise | 73.9 dBA | To and Chung 2015 |
| Noise as most bothersome dining issue | 24% | Zagat 2018 |
| US adults who find loud dining annoying / likely to avoid a noisy restaurant | 68.0% / 64.8% | CDC authors, SummerStyles 2021 |
| Chain dining visits flagged for excessive noise | 16% | Consumer Reports 2016 |
| Lombard slope at 50 dB / 80 dB background | 0.2 / 0.5 dB/dB | Nijs et al., JASA 2008 |
| Voice-level breakpoint, adults over 60 | 58.3 dB(A) | Bottalico et al. 2022 |
| Ambulatory Lombard slope | 0.54 dB/dB | Whittico et al., JASA 2020 |
| Maximum noise for sufficient conversation at 1 m | 71 dB | Rindel 2019 |
| Student canteen acoustic capacity | 189 people | Zelem et al., Acta Acustica 2023 |
| Background level where conversational SNR can turn negative at 1 m | approximately 69 dB SPL | Weisser and Buchholz, JASA 2019 |
| UK respondents who left a venue early / would not return to a noisy venue | 79% / 91% | Action on Hearing Loss 2016 |
| 50% word recognition with moderate-to-severe hearing loss | about 50 dB(A) | Bottalico et al. 2022 |
| Loud NYC venues above 85 dBA | 80% | Spira-Cohen et al. 2017 |
| Restaurant employee mean full-shift exposure | 80 dBA | Green and Anthony 2015 |
| US workers exposed to hazardous noise each year | about 27 million | CDC NIOSH |
| Apple Hearing Study participants above WHO environmental limit | 25% | Apple / University of Michigan 2021 |
Methodology and Sources
- Acoustical Society of America, “Are restaurants and bars in New York City too loud?” lay-language paper (Gregory Scott, SoundPrint), 174th ASA meeting, 2017, companion to Farber and Wang, Analyses of crowd-sourced sound levels of restaurants and bars in New York City, Proceedings of Meetings on Acoustics 2017 (DOI 10.1121/2.0000674).
- Oticon restaurant noise study, reported by Hearing Review, 2018 (manufacturer-commissioned study).
- To and Chung, Restaurant noise: Levels and temporal characteristics, Noise and Vibration Worldwide 2015.
- Hodgson, Steininger and Razavi, Measurement and prediction of speech and noise levels and the Lombard effect in eating establishments, JASA 2007 (DOI 10.1121/1.2535571), abstract via Europe PMC.
- Zagat 2018 Dining Trends Survey and Zagat State of American Dining 2016 (blog archive).
- Eichwald, Murphy and Scinicariello (CDC / NIOSH / ATSDR), Study Shows Noisy Restaurants Pose Health Risks, The Hearing Journal, January 2022, based on Porter Novelli SummerStyles 2021.
- Consumer Reports, Top restaurant complaints and worst offenders, 2016.
- Action on Hearing Loss (RNID), Access to cafes, pubs and restaurants policy statement, March 2018, citing the Speak Easy survey (2016) and a general public survey (n = 1,200, 2016).
- Nijs, Saher and den Ouden, Effect of room absorption on human vocal output in multitalker situations, JASA 2008.
- Bottalico, Piper and Legner, Lombard effect, intelligibility, ambient noise, and willingness to spend time and money in a restaurant amongst older adults, Scientific Reports 2022.
- Whittico et al., Ambulatory monitoring of Lombard-related vocal characteristics in vocally healthy female speakers, JASA 2020.
- Rindel, Restaurant acoustics: Verbal communication in eating establishments, Acoustics in Practice (European Acoustics Association) 2019.
- Zelem, Chmelik, Glorieux and Rychtarikova, Correlation of room acoustic parameters and noise level in eating establishments, Acta Acustica 2023.
- Weisser and Buchholz, Conversational speech levels and signal-to-noise ratios in realistic acoustic conditions, JASA 2019 (PMID 30710956), abstract via Europe PMC.
- Ellis, Crukley and Souza, The effects of SNR, T60, WDRC speed, and DNR in a virtual restaurant setting, Ear and Hearing 2024.
- Spira-Cohen, Caffarelli and Fung, Pilot study of patron sound level exposure in loud restaurants, bars, and clubs in New York city, J Occup Environ Hyg 2017.
- Green and Anthony, Occupational Noise Exposure of Employees at Locally-Owned Restaurants in a College Town, J Occup Environ Hyg 2015.
- CDC NIOSH, Noise and Occupational Hearing Loss.
- Apple Hearing Study (University of Michigan) update, March 2021.
- SoundPrint crowdsourced decibel database and WHO Deafness and hearing loss fact sheet, March 2026.
- Data watch: most measured restaurant data predates 2023, so treat it as the most recent available rather than a 2026 snapshot. SoundPrint sample sizes differ inside the same ASA lay paper (more than 1,800 Manhattan venues in the body, more than 2,250 New York venues in the title), and the proceedings version is not fully public. Lombard slopes vary with method: controlled multitalker rooms converge near 0.5 dB/dB at high noise, while studies that include music and other sources report wider ranges. Zagat noise ranked second nationally in 2016 and first in 2018, and we found no newer national Zagat figure. The Oticon data is manufacturer-commissioned and the SoundPrint figures are crowdsourced from phone microphones, so neither is a calibrated survey. The Speak Easy percentages come from a charity’s own survey and are quoted from its 2018 policy statement because the original 2016 report is no longer online.
Last updated: October 18, 2026. We update this roundup quarterly, and the next refresh is expected when the CDC and NIOSH publish further SummerStyles noise analyses or SoundPrint releases new crowdsourced venue data.