Yes, open-office noise measurably affects people, but not always in the way headlines say. Laboratory studies show that intelligible background speech impairs some kinds of work, most consistently verbal short-term memory, and a simulated-office experiment found that eight minutes of office noise raised physiological stress and negative mood without lowering task performance. Large occupant surveys find that the lack of speech privacy, more than loudness, is what open-plan workers complain about most. The fixes with the most direct evidence are the ones that make nearby speech less intelligible: sound absorption, sound masking, screens and distance, zoning, and enclosed rooms for focused or private work.
This article reads the studies that ranking summaries cite, says what each can and cannot show, and links each finding to a type of remedy. Where absorption goes on an open floor is covered in open office acoustics, and the choice between absorption and masking in sound masking or acoustic panels.
The short version: what the evidence supports
Different kinds of studies answer different questions, and most confusion comes from quoting one kind as if it were another. A laboratory experiment can show cause and effect under controlled conditions, usually over minutes or hours. A field study shows what happens in a real office, with less control over what caused what. A survey shows what large numbers of people report, but not why.
| Finding | Study type | Scale | What it cannot show |
|---|---|---|---|
| More intelligible background speech impairs verbal short-term memory and some other tasks | Systematic review of lab experiments (Haapakangas and colleagues, 2020) | 14 studies, 34 tests | Effects on whole workdays, or on complex real tasks, where the evidence is limited |
| Speech at an excellent open-office intelligibility level did not impair tasks that a poor one did | Lab experiment (Haka and colleagues, 2009) | 37 students, 4 hours | Long-term effects; real offices |
| Office noise raised stress markers and negative mood; task performance did not drop | Lab experiment (Sander and colleagues, 2021) | 40 participants, 8 minutes per condition | Effects of a full workday, or of real coworkers |
| Zoned and team layouts rated better than plain open plan; open plan had more loud peaks | Field experiment in one company (Pitchforth and colleagues, 2020) | 288 employees, 2 weeks per design | Which of several simultaneous changes caused the difference |
| Sound privacy is the most common source of dissatisfaction in open plans | Survey database (Kim and de Dear, 2013) | 42,764 responses, 303 buildings | Cause and effect; performance |
| Lack of privacy predicts acoustic dissatisfaction more than noise disturbance | Field measurements plus surveys (Yadav and colleagues, 2025) | 349 occupants, 28 offices | Performance outcomes |
Two terms recur. Speech intelligibility is how much of a nearby conversation you can understand, often measured as the speech transmission index described below. Speech privacy is the other side of the same coin: how well your own conversation stays with the people you are talking to. Noise disturbance is a broader feeling of being bothered by sound, whatever its source. The studies treat them separately, and so should anyone building a case.
Read together, the studies agree on a pattern more than on a number. Speech that can be understood is the problem; the feeling of being overheard is part of it; and layouts that separate quiet work from talk tend to rate better. What the evidence does not support is a single productivity figure for “open-plan noise” that applies to every office.
Why speech, not loudness, is the main problem
The noise that bothers people in open offices is mostly other people talking. An ISO news article introducing the open-plan office measurement standard, written by Jens Holger Rindel, puts it plainly: “The foremost source of noise in open-plan offices comes from telephone calls and discussions between co-workers. Conversations are particularly disturbing because of the information content.” The same ISO article notes that many people talking at once at a distance can no longer be understood and can even mask speech nearer by.
Researchers capture “how understandable” with the speech transmission index (STI), a number from 0 to 1 that rises as speech becomes more intelligible. In a 2009 laboratory experiment, Haka and colleagues played speech at the same level, 48 dBA, at three intelligibility settings: STI 0.10 for a private office, 0.35 for “an acoustically excellent open office” and 0.65 for “an acoustically poor open office.” Thirty-seven students worked through five tasks in each condition over 4 hours. The researchers changed intelligibility by changing the balance between the speech and a masking sound, which kept the overall level constant.
Three of the tasks, including serial recall and a working-memory task, were worse only at STI 0.65. Questionnaires, by contrast, showed disturbance rising at every step, which led the authors to conclude that “subjective comfort was disturbed more easily than performance.” Their practical implication is directly about remedies: “Proper acoustic design requires both the use of high acoustic absorption and an appropriate masking sound.”
A 2020 systematic review revised the model that links STI to performance. The earlier model, published by Hongisto in 2005, had shaped office acoustic design guidelines, but the review’s authors note it “was based on scarce empirical data.” Haapakangas, Hongisto and Liebl pooled 14 laboratory studies reporting 34 tests. Across all tests, “performance begins to decrease approximately above STI = 0.21 while the maximum decrease is reached at STI = 0.44.”
The review also separated task types, which most summaries skip. Verbal short-term memory tasks “were most strongly and very consistently affected,” with performance deteriorating between STI 0.12 and 0.51. The authors found “some evidence” for verbal working memory tasks and “limited evidence in complex verbal tasks,” and called for more research on task-specific effects.
Two practical points follow. The harm comes from intelligibility, which depends on how much speech reaches you relative to the background sound, not on loudness alone. And because the modeled decline levels off near STI 0.44, a floor where nearby speech is very clear has to bring it below roughly that point before the model predicts any gain in performance; smaller improvements may still be felt as less disturbance.
Which kinds of work suffer most
The task-type findings are the most useful part of this research for a manager, and the part summaries most often drop. Verbal short-term memory means holding words, numbers or sequences in mind for a few seconds while working with them: keeping a figure in your head while you type it, following a long reference number or a contract clause, rehearsing what you are about to write. Those are the tasks the review found “most strongly and very consistently affected” by intelligible speech.
The 2009 experiment points the same way. The tasks that suffered at STI 0.65 were the operation span task, a test of working memory; serial recall, remembering items in order; and the activation of prior knowledge from long-term memory. By contrast, the 2021 stress experiment used proofreading and found no performance drop over 8 minutes.
The evidence is thinner for complex, real-world verbal work such as writing a report or reasoning through a problem; the review calls it “limited.” That is not evidence that such work is unaffected. It means the effect has been measured less often, and a manager should not promise a number for it.
Two practical readings follow. On that evidence, people doing detail-heavy verbal work, such as editing, data entry, accounting or legal review, are the likeliest to benefit from lower speech intelligibility. And people’s sense of disturbance is a more sensitive signal than their output: in the 2009 study, ratings worsened at every step while performance changed only at the worst setting.
Why the decibel numbers in studies are hard to compare
Office noise studies report decibels, but they rarely measure the same thing, and putting their numbers side by side misleads. Four kinds of figure appear in the studies here.
| Figure | Example | What it describes |
|---|---|---|
| Level of a test sound at the listener | 48 dBA in the 2009 experiment; 59.1 dB(A) against 36.3 dB(A) in the 2021 experiment | The sound played to participants in a lab |
| Peaks above a chosen threshold | Readings above 70 dB in the 2020 field experiment | How often sensors, read every 15 minutes, recorded loud moments in a working office |
| Background noise limit | 50 dBA average for open areas for concentration, first tier of WELL v2 S02 | Noise from building systems and outside, not from people talking |
| Masking level limit | No more than 48 dBA in open areas under WELL v2 S06 | Artificial background sound added on purpose |
The 2009 experiment shows why level alone is the wrong yardstick: all three conditions were played at the same 48 dBA, yet only the most intelligible one hurt performance. Masking makes the same point from the other side. Added background sound raises the measured level while lowering intelligibility, and it is one of the remedies with laboratory support.
So when a study or a vendor quotes a decibel figure, ask three things: whose sound it is, whether it is an average or a peak, and whether the office was occupied. The 2021 authors, for instance, chose their 59.1 dB(A) soundscape on the basis of earlier literature reporting typical open-office background levels of 52 to 58 dB(A), a figure they cite rather than one they measured.
The stress experiment: what eight minutes of office noise did
The study behind “open-plan office noise is stressful” headlines is a careful laboratory experiment with a short exposure. Sander and colleagues, publishing in the Journal of Management & Organization in 2021, compared a simulated open-plan soundscape with a quiet private-office condition.
The method, from the full text:
- Participants. Forty healthy volunteers aged 17 to 44, mean age 22.65, mostly undergraduate and postgraduate students (34 women, 6 men). The authors’ power analysis called for at least 34.
- Conditions. Each person did a proofreading task under both conditions, in counterbalanced order: 10 minutes per condition, of which 8 minutes were the task.
- Sound. A loudspeaker played people speaking, walking, printing, ringing telephones and typing, at an average of 59.1 dB(A) at the head. The quiet condition averaged 36.3 dB(A), from air conditioning and computer fans.
- Measures. Heart rate and heart rate variability, skin conductance, facial emotion recognition software and a self-report mood scale.
The results: open-plan noise “did not reduce immediate cognitive task performance,” but it did reduce wellbeing on self-reported mood, facial expressions and physiological stress markers. In a piece republished by Bond University, the lead author summarized the effects as negative mood up 25% and sweat response up 34%.
What it can and cannot show
The design supports a causal claim: the same people, in the same room, reacted differently to two soundscapes. The authors are also clear about its limits. They call the short exposure “the most significant limitation” and suggest it may explain why performance did not drop. They expect effects on stress and mood to be greater over a full day, which is a reasonable expectation rather than a measured result.
Three more limits matter when quoting the study. The sample was young and mostly female students rather than office workers. The noise came from a loudspeaker, not from colleagues whose conversations carry meaning for the listener. And the noisy condition was louder overall as well as more speech-filled, so the study does not separate the effect of speech from the effect of level.
The authors also offer an explanation for the gap between unchanged performance and raised stress: “Continuing to focus on a task under distraction requires compensatory effort, indicated here by physiological changes, and eventually may deplete self-control resources.” Put simply, people kept up their output by working harder at it, and the study measured the cost of that effort rather than a drop in results. That fits the 2009 finding that comfort suffers before performance does.
For remedies, the study does not test one. The lead author’s article mentions acoustic treatments and sound masking as options, which is advice rather than a finding of this experiment.
What the field experiment at a technology company found
The most practical evidence comes from a randomized field experiment in a working office. Pitchforth and colleagues, reporting in PLoS ONE in 2020, moved 288 employees from 22 teams at a technology company through four office designs, two weeks in each, in an order that differed for every group.
The four designs were a plain open plan (the control), a zoned open plan, an activity-based layout and team offices. The zoned open plan kept open desks but added closed doors between working and collaboration areas, capped each room at 40 people against a maximum of 72 in the control, and gave every room at least two phone booths. The team offices held four or six desks each, with each space “delineated by walls of sound absorbing panels.”
The headline results, all compared with the plain open plan:
| Measure | Zoned open plan | Team offices | Activity based |
|---|---|---|---|
| Satisfaction with the workplace | 12% higher | 8% higher | 14% lower |
| Self-reported productivity | 17% higher | 10% higher | 14% lower |
| Git commits (an objective output measure) | No significant effect of design | No significant effect | No significant effect |
| Sound peaks outside the study’s healthy range (above 70 dB) | Open plan had them 29% more often | Open plan had them 20% more often | Open plan had them 27% more often |
The authors write that “noise was the main environmental variable to differentiate office designs,” and that part of employees’ preference for the other designs may be that they were less distracting. The abstract’s phrase “higher levels of unsafe noise” refers to readings above the study’s own 70 dB threshold, set for the experiment, not to a hearing-safety limit.
What it can and cannot show
The study’s strengths are real: randomized order, real teams doing real work, and objective sensor data alongside surveys. Its limits are as important for anyone quoting it.
- Several changes at once. The zoned design changed doors, room size, phone booths and plants together. The study shows that the package worked better; it cannot say how much of that came from acoustics.
- Mostly self-reported outcomes. Productivity was the employees’ own rating. The one objective output measure, code commits, did not differ by design.
- One company, a young workforce. The authors note the sample was younger than the general population and recommend repeating the experiment elsewhere.
- Two weeks per design. Enough time to settle in, but short enough that novelty may play a part.
The method is worth borrowing even where the findings are not. The company ran its comparison inside an ordinary working floor: teams rotated through the designs in different orders, each design got two weeks, and the same survey and sensors were used throughout. A smaller organization can run a modest version, such as treating one zone, swapping two teams after two weeks and asking the same questions in both places, and learn more about its own floor than any published percentage can tell it.
For remedies, the study points to separation more than to any single product: fewer people per room, doors between talk and focus, booths for calls, and enclosed team spaces lined with absorption. The activity-based design, which also had focus rooms and booths, rated worst, a reminder that offering quiet rooms is not the same as people using them.
Behavior told the same story as the surveys. The authors report that once employees no longer had to work in the open-plan design, “they spent more time at their desks,” measured by occupancy sensors rather than by anyone’s opinion. People voted with their feet, in a modest way, for the designs that separated talk from work.
What large surveys say: privacy is the bigger complaint
Surveys cannot prove causes, but at scale they show which problems are common. The largest in this area is the Center for the Built Environment database analyzed by Kim and de Dear in 2013: 42,764 responses from 303 office buildings.
The survey asked separately about “the noise level in your workspace” and “the sound privacy in your workspace (ability to have conversations without your neighbours overhearing and vice versa).” Sound privacy drew the highest dissatisfaction of any indoor environment factor: 59% in high-partition cubicles, 58% in low-partition cubicles and 49% in open plans with no or limited partitions, against 18% in private offices.
Two findings cut against common assumptions. Satisfaction with “ease of interaction” was no higher in open-plan offices than in private offices, and the benefits of easier interaction were smaller than the penalties of more noise and less privacy. And high cubicle partitions did not rescue sound privacy; high-partition cubicles reported the lowest satisfaction on 13 of 15 factors.
The database has the usual limits of a survey. It records what occupants reported in the buildings surveyed, so it shows which problems go together, not what causes what, and it says nothing directly about output. Its value is scale: with more than 40,000 answers, the ranking of complaints is unlikely to be a fluke of one building.
The US General Services Administration reached similar conclusions from its own buildings, and its guide is blunt about the source: “The real noise problems are people having conversations with colleagues on the phone, in their workstations or in adjacent circulation areas.” In Sound Matters, GSA reports that after seven federal offices were redesigned for collaboration, almost every environmental factor improved “except noise and voice privacy.” In a baseline survey of about 3,700 respondents, “60% said they could get more done if it were quieter,” 56% said the ability to insulate themselves from distractions was very important, and 50% said noise kept them from being as productive as they could be. Describing a Veterans Administration service center in Reno, the guide reports “much greater dissatisfaction with speech privacy than with noise intrusion,” the same order the large database shows.
A 2025 field study added measurements to the surveys. Yadav and colleagues surveyed 349 occupants in 28 offices and measured each room. As predictors of acoustic dissatisfaction, “the contribution of lack of privacy” was 25% higher than that of “noise disturbance.” The study’s findings on which measurements predicted dissatisfaction are covered in detail in sound masking or acoustic panels, and are not repeated here.
Two of its other findings speak to layout decisions. “Medium-sized offices exhibited higher dissatisfaction than larger (≥50 occupants) offices,” and dissatisfaction did not differ between offices with fixed seating and “more flexible and activity-based working configurations.” It did vary with ceiling height, the number of workstations and how long people had worked there. The authors close by noting “the complexities in characterizing occupants’ perceptions using instrumental acoustic measurements,” a fair warning against expecting any single measured number to predict how a floor will feel.
Do people get used to it?
A common hope is that complaints fade once people settle in. The literature review in Kim and de Dear’s paper reports the opposite: in earlier longitudinal studies of moves from enclosed to open-plan offices, “the occupants in these studies didn’t adapt or habituate to the change in spatial layout.” The International WELL Building Institute summarizes the evidence the same way in its WELL v2 text on background noise, stating that “employees are unable to habituate to noise in office environments over time.” Neither source is a new experiment, but both reflect the weight of studies that followed people over time.
Open offices change behavior, not only sound
Some of the most quoted open-office research is not about noise at all, and it is worth knowing which. Bernstein and Turban ran two intervention-based field studies of company headquarters moving to more open layouts, measuring face-to-face contact with wearable sensors and electronic contact from email and messaging servers, before and after the move. “Contrary to common belief, the volume of face-to-face interaction decreased significantly (approx. 70%) in both cases,” with more email and instant messaging instead.
That study measured interaction, not sound. Its relevance here is indirect: the usual argument for open plans is that they increase conversation, and these two cases found the opposite. A plausible reading, consistent with the privacy findings above, is that people talk less when everyone can hear them; the study itself does not test that explanation.
GSA’s guide observed the other side of the same coin. In one renovated GSA building, “the number of interactions in workstations doubled in the new workspace due to the increased visibility of work,” and GSA notes that this generates “more distractions for others doing quiet work.” Whether an open floor produces more talk or less seems to depend on the floor and the people. Either way, the talk that does happen is what neighbors hear.
How to read an office-noise study before you quote it
Most errors in summaries come from stretching a study beyond its design. Six questions catch most of them.
| Question | Why it matters | Example from the studies above |
|---|---|---|
| How long was the exposure? | Minutes in a lab are not a workday | Eight minutes per condition in the stress experiment |
| Who took part? | Students, office workers and one company’s staff differ | Mostly female students; one young corporate workforce |
| Was it a lab, a field site or a survey? | Labs show causes; fields show real conditions; surveys show prevalence | All three appear above |
| What was measured: feelings, physiology or output? | Comfort is disturbed more easily than performance | Haka’s questionnaires moved before task scores did |
| Was the noise speech, and was it meaningful? | Intelligible speech is the main driver | Loudspeaker soundscapes versus real colleagues |
| How many things changed at once? | Packages cannot be split after the fact | Doors, room size, booths and plants in the zoned design |
What the percentages are percentages of
Three different kinds of percentage appear in the studies above, and they are easy to mix up. The field experiment’s “12% higher satisfaction” compares average scores on a survey scale between two designs. The stress experiment’s 25% is a relative change in mood scores between two lab conditions. The survey database’s 59% is a share of respondents who chose the two lowest answers on a seven-point scale. None of them is a measure of output, and none can be added to another.
Apply the questions to a headline such as “office noise raises negative mood by 25%.” The figure is real, but it comes from an 8-minute laboratory session with students and a loudspeaker. Quoted with that context, it is a useful sign that noise has a physiological cost; quoted without it, it sounds like a measured effect on every office worker’s day.
Which fixes have evidence, and how strong it is
The studies above point to remedies that reduce how much intelligible speech reaches people who are not part of the conversation. The ISO article on open-plan measurement names the most important acoustic parameters as “the amount of sound absorption material (usually in the ceiling), the use of screens and background noise.” Each remedy changes something different, and the evidence behind each differs.
| Remedy | What it changes | Evidence behind it | Caveat |
|---|---|---|---|
| Ceiling absorption | Less speech carries across the floor by reflection | Haka’s design implication; ISO names it first | Works with screens and background sound, not alone |
| Screens between desks | Blocks the direct path at seated height | ISO names screens; WELL v2 counts rated screens | GSA warns high partitions can encourage louder talk |
| Sound masking | Raises background sound so speech is less intelligible | Haka’s lab conditions adjusted STI with masking | Needs careful levels; see the masking article |
| Zoning and distance | Keeps talk away from focused work | Zoned design rated best in the field experiment; ISO recommends grouping people who share information | Several changes were bundled together |
| Enclosed rooms and booths | Removes calls and meetings from the open floor | Private offices had the lowest privacy dissatisfaction; booths were part of the zoned design | Activity-based design with rooms still rated worst |
| Wall panels in meeting rooms | Shorter reverberation inside the room | Published reverberation limits for conference rooms, such as WELL and LEED | Does not stop speech leaking out through walls and doors |
Two remedies deserve an honest note about panels. Wall panels on an open floor do some good near walls, but the ceiling usually carries most of the absorption that limits how far speech travels; open office acoustics explains why. And no panel stops sound passing between rooms, so privacy for enclosed meetings is a wall, door and ceiling-plenum question.
Rating systems have started to encode the same logic. WELL v2 rewards high-performing ceiling absorption and rated desk screens in open workspaces, and gives an extra point for combining them with masking; WELL’s sound requirements walks through the features.
Why one measure rarely works alone
The studies agree that remedies work as a set. Haka and colleagues concluded that proper design “requires both the use of high acoustic absorption and an appropriate masking sound.” The ISO article lists absorption, screens and background sound together. GSA frames office noise control as the “ABC’s” of noise control: Absorb, Block, Cover. The same guide warns, under “Myth,” that higher cubicle partitions “provide small amounts of additional acoustical shielding” and “may encourage people to talk louder because they think they have more privacy.”
The reason is that each measure covers a different path. Absorption reduces speech that arrives by reflection, mostly from the ceiling. Screens reduce speech that travels directly between desks. Masking lowers the intelligibility of whatever speech still arrives. Distance and zoning reduce how much speech is produced near people doing focused work. Leave one out and the others carry more of the load than they can.
What the research does not yet tell you
Honest gaps matter as much as findings when making a case, because they tell you which promises not to make. The authors of these studies name several themselves.
- Long exposures in real offices. The clearest cause-and-effect results come from sessions of 8 minutes to 4 hours. The stress experiment’s authors call short exposure their most significant limitation.
- Complex work. The 2020 review found limited evidence for complex verbal tasks and called for “further research” on task-specific effects.
- Single-remedy field trials. The field experiment changed several things at once, so it cannot price acoustics separately from doors, booths and room size.
- Other workforces. The field experiment drew on one company’s younger-than-average staff, and its authors recommend repeating it elsewhere.
- Measurement to perception. The 2025 study found that instrumental measures predict dissatisfaction only partly.
What the research does support is enough to act on: intelligible speech is the main problem, privacy matters to people as much as noise, comfort is disturbed before output is, and remedies work best together. What it does not support is a reliable productivity gain, in percent, for any particular office.
A fictional example: building the case for one floor
Example (fictional). Harbor Lane Software (fictional) has a 60-person open floor with a hard ceiling, and the office manager wants to know whether acoustic work is worth proposing. Rather than quoting headline percentages, she runs a short survey using the two questions from the large survey database: satisfaction with noise level, and with sound privacy, “the ability to have conversations without your neighbours overhearing and vice versa.” Of 40 respondents, 24 (60%) are dissatisfied with sound privacy and 14 (35%) with noise level, a pattern that matches the research: privacy first.
She then asks what kind of work is suffering. Developers report trouble with tasks that need holding details in mind, the kind of verbal short-term memory work most affected in the lab studies. Sales staff say they take calls at their desks because the two phone rooms are always booked.
Her proposal follows the evidence in order: a room acoustic assessment of the floor, ceiling absorption over the desk areas, screens between facing desks, two more phone booths, and a quiet zone at the far end away from the kitchen. She asks a masking supplier for a separate proposal and plans to repeat the same two survey questions three months after the work, at the same time of year.
Three months after the work is finished, the same questions go to the same team. Suppose 15 of 40 respondents (38%) are now dissatisfied with sound privacy, against 24 before, and 10 (25%) with noise level, against 14. She reports those as what they are, changes in one team’s answers, not as a productivity gain, and adds what the floor now has that it did not: fewer calls at desks and a quiet zone that people use.
The example uses invented numbers, but its method is the point. A survey with the same questions before and after, a clear link between complaints and remedies, and modest claims about productivity make a stronger internal case than any single statistic from a study done elsewhere.
What to do next
Start with your own floor rather than the headlines. Five steps give a decision maker something concrete.
- Ask the two questions. Survey satisfaction with noise level and with sound privacy separately, before any change. The open office acoustics checklist lists what else to record about the floor.
- Get the floor measured or estimated. Reverberation, background sound and how fast speech fades with distance are what the remedies above change. A room acoustic assessment estimates the starting point from the floor’s size and finishes.
- Match each complaint to a remedy. Privacy complaints point to absorption, screens, masking and enclosed rooms working together; complaints about calls point to booths and zoning; echo in meeting rooms points to panels inside those rooms, covered in why your meeting room echoes on video calls.
- Decide what you will claim. Promise fewer complaints about privacy and noise and fewer calls at desks, which you can check with the same survey. Do not promise a productivity percentage, which the research cannot back for your office.
- Check again after the change, with the same questions, the same people where possible and the same time of year, and share the result with the team that answered.
Fibson designs and makes acoustic panels to order and plans each room before anything is ordered; it does not sell masking systems or partitions. Its offices page describes how it approaches workplaces. If you want a written estimate and panel plan for a floor, send us the room with its plan, photos and the survey results you have.
