Clean Air, Cloudy Evidence: What Do We Really Know About Ventilation and Learning?

Keywords: IAQ, educational buildings, cognitive performance, learning

We have a duty to protect the vulnerable, and that includes the children who spend over a thousand hours each year in classrooms. For those of us working in ventilation, that responsibility often translates into a push for cleaner air.

During the COVID-19 pandemic, some ventilation engineers became household names in their own countries doing amazing jobs of making people aware of the importance of IAQ. CO2 sensors and HEPA filters were installed in classrooms across the globe. Since the pandemic, there has been renewed interest in improving IAQ in schools, fuelled by concerns about infection, but also by wider pre-existing claims about the health risks from contaminant exposure, and the effects of IAQ on attendance, cognition, and academic performance.

Advocates now campaign for ever higher airflow rates, often invoking the language of safety and science. But not all evidence is equal. And beyond a certain point, more air does not always mean more benefit. The concentration of a contaminant is inversely proportional to the clean airflow rate, and so there is a law of diminishing returns where the benefits gained from increasing the airflow rate represent a proportionally smaller gain as more is added. This is illustrated in Figure 1, although it should be noted that the space conditioning load can be eliminated if secondary air systems are used, such as portable air cleaners. When the concentration of a contaminant of concern is in the left half of the plot, where the concentration gradient is steepest, it may be desirable to reduce its concentration. But when it is in the right side where the gradient is small, it may not be worth it. There are often many contaminants of concern present in indoor air and, for a particular airflow rate, they may be in different parts of the plot. During the COVID-19 pandemic, this plot was used explain to policy makers why it is important to treat under-ventilated spaces first, and why a blanket approach to all spaces everywhere is sub-optimal.

There is a strong health-based justification for ensuring acceptable IAQ in schools. This means meeting existing standards and guidelines. But going beyond that requires major investments in system upgrades, or the widespread deployment of air cleaners, and so it would demand strong evidence.

Figure 1. The law of diminishing returns: the relationships between contaminant concentration, energy, and the clean airflow rate.

What is Indoor Air Quality?

A General Definition

A definition of indoor air quality is often a subject of tension. Most would agree that the phrase ‘indoor air’ considers air in a building and immediately around it, which may be brought inside. It is the word ‘quality’ that is contentious. Quality is a measure of excellence. It is a relative parameter. A horse was considered an excellent mode of transport until the car was invented. So, when we think of IAQ, it is relative to something. But considering the quality of indoor air relative to uncontaminated air may not be an efficient approach because it may cause resources to be expended on the unnecessary design, manufacture, and running of systems. So how should we think about IAQ? Should we strive for the best, the most excellent quality? Or, should it be good enough? Should it be acceptable? What would make IAQ acceptable?

ASHRAE defines acceptable indoor air quality as “air in which there are no known contaminants at harmful concentrations, as determined by cognizant authorities, and with which a substantial majority (80% or more) of the people exposed do not express dissatisfaction” (ASHRAE 2022). Acceptability is not a categorical outcome (bad, average, good, excellent), as some experts often mistakenly say it is. It is firmly binary: the quality is either acceptable, or it is not. It is the criteria of acceptability that might change as new evidence emergences, improving IAQ. EN16798-1:2019 also talks about acceptable IAQ but does not define it. IAQ is implicitly considered acceptable when the IAQ criteria for a space are defined and then achieved. This is our current paradigm.

Dose and Response

Any person is exposed to airborne contaminants when they occupy the same space as contaminated air and their exposure is a function of the concentration and the time spent in the space. The dose received is a function of breathing rate, breath volume, and uptake by the lungs (Jones et al., 2021). These factors are affected by metabolic rate and physiology that may be a function of age and sex (Persily 2017). None of these parameters are constants and so the dose received in different spaces and for different occupancy scenarios will also be different. Small children respond differently to young adults. Therefore, an air quality metric should identify when the quality of indoor air is unacceptable and should be based on its effects on human health and comfort.

A person’s responses to a dose may be immediate and are known as acute responses. We have tended to categorise these responses as a building related illness when we know the cause, or as sick building syndrome when we don’t. We know some contaminants cause acute effects and the world health organization (WHO) regulates PM2.5 (particles with a diameter of less than 2.5 microns), PM10 (particles with a diameter of less than 10 microns), nitrogen dioxide (NO2), ozone (O3), carbon monoxide (CO), sulphur dioxide (SO2) and formaldehyde (HCHO) over a 24-hour period. This is because there are established correlations between the presence of these contaminants and negative health effects. However, we cannot yet say, and may never be able to say, that a specific health outcome is likely at a given contaminant concentration, such as a child suffering an asthma attack at some PM2.5 concentration.

Responses that act over a lifetime are known as chronic responses. The harm caused by years of exposure may be illness, disability, and/or premature death. The WHO regulates PM2.5, PM10, and NO2 over an annual period.

How Might we be Wrong?

Rethinking the Regulation of Contaminants

An exposure limit value (ELV) is the highest amount of a substance considered safe to breathe over a certain period, based on current health evidence. They are used in occupational environments to prevent or reduce risks to health from hazards, such as vibrations, by setting a maximum quantity experienced over an exposure time. A problem with them is that it isn’t clear how a change in an ELV, say by 10%, would affect occupant health. This can only be done with knowledge of the dose-response relationship.

ELVs are given by regulatory authorities for criteria contaminants that are known to have a direct effect on human health, but they don’t agree with each other. For example, the WHO and the US environmental protection agency give wildly different ELVs for the same contaminants, such as PM2.5 (USEPA 2024; WHO 2021). If both organizations consider the same risk of harm, they should agree. There is also harm inequity between contaminants, because exposure to different contaminants at their respective ELVs should result in the same level of harm from each. The selection of ELVs is generally undocumented and subjective. A further problem is that there are often many criteria contaminants in standards and guidelines; for example, the UK’s school IAQ guideline document, BB101, gives ELVs for 16 contaminants (DFE 2018). Prescribing lists of ELVs is unwise because a diagnostic procedure is required for each of them, and time and cost constraints make enforcing the list impossible. It makes more sense to identify contaminants based on the dual conditions of being harmful and commonly present in indoor air. Then contaminants can be ranked by the harm they cause and the most harmful targeted for mitigation. The econometric of harm used, the disability adjusted life year (DALY), allows the harm from each contaminant to summed to estimate the total harm caused by all airborne contaminants (Morantes 2024). The total harm in DALYs can then be compared against DALY values for other common hazards, such as transport injuries or smoking.

This type of harm analysis has now been done for homes and offices (Morantes et al., 2024). In homes, of the 45 evaluated, six contaminants were found to cause 99% of the total harm: PM2.5 (~66% of all harm), PM10-2.5 (~13%), HCHO (~9), and NO2 (~8%), radon (~2%), and O3 (~1%). These are the most harmful contaminants by around an order of magnitude. In offices, five contaminants cause 99% of the total harm: PM2.5 (~93%), HCHO (~5%), and O3 (~1%). The population harm from exposure to airborne contaminants in homes is over 7 times higher than offices. Would we expect to find the same schools? Airborne contaminants in school classrooms generally originate from outside, so it is very likely that the contaminants identified for offices will also be important, particularly PM, although the proportions each contributes to the total harm will be different. The air in homes is likely much more harmful than in schools because there are prominent contaminant sources in homes that do not exist in school classrooms, such as cooking, and because children spend more time at home.

The Mirage of Causal Claims

Ventilation and air quality are increasingly promoted as levers for improving academic and behavioural outcomes in educational buildings (Allen 2021). But much of the supporting evidence comes from observational studies that are not designed to establish causality. There are many observational studies, where researchers watch what happens in real life without trying to change or control anything, which demonstrate some effect of ventilation on some measure of productivity or cognitive ability. However, they are often conducted over relatively short periods, and it is not possible to extrapolate the effects over longer periods or identify what metrics of productivity improvement might be expected. These studies often poorly control for confounders like age, sex, sleep, nutrition, noise, and teacher quality, or attempt to account for everything in models built on small samples whose statistical power is low so that the chance it detects a real effect when there actually is one, is also low. The results are inevitably noisy. In some cases, the observed effect of IAQ on test performance is smaller than the natural variability in the test scores themselves.

The literature is littered with examples of what were originally perceived to be correlations to improved productivity, which over time are discovered to not be causal at all. The most famous is the Hawthorne Effect, where research suggested that productivity at an Electric Works was increased by improving the lighting levels in the factory, a conclusion that was later found to be false because the workers had modified their behaviour when they are aware of being observed (Levitt 2021). Without robust data and studies there is a likelihood that reported correlations may not have any causative features related to the ventilation rates.

One common pitfall is the overuse of CO2 as a proxy for indoor air quality (ASHRAE 2025). While it can serve as an indicator of the ventilation rate in some circumstances, CO2 is not harmful in a typical classroom. It is not an indicator of IAQ because its concentrations are uncorrelated with other contaminants. This is because it is solely diluted by outdoor air, whereas PM, HCHO, O3, NO2 and airborne pathogens have multiple removal mechanisms. Yet, it dominates many analyses. This has skewed both academic discourse and public messaging. Claims that cognitive performance improves at lower CO2 concentrations are often built on weak data and unvalidated tests, sometimes conducted over short periods or with repetitive online formats that risk disengagement.

One example is a recent study that measured CO2 in university lecture rooms and tested students' cognitive performance after class using a Stroop task (Dedesko 2025). The researchers found that higher CO2 was linked to slightly worse test results. But an examination of the method shows the evidence is weak. The study was underpowered, with only 54 students completing enough tests, which is about half the number the researchers had planned. The rooms were also fairly well ventilated, with CO2 mostly below 1,200 ppm. Nevertheless, the analysis focused on the upper end of CO2 concentrations and used a flexible model to link them to performance. But small and noisy effects in observational data can be misleading, especially without clear controls for confounding factors like fatigue or time of day. The supposed impact—one fewer correct answer on a multi-trial task—is tiny, and likely meaningless in the real-world. So, while the research question might be valid, the study doesn’t offer strong evidence that CO2 harms cognition.

A final example of an erroneous causal claims is the use of total volatile organic compound (TVOC) as a health metric. It is uncorrelated with any negative health effect in people. This could be explained by the harm analyses in homes and offices (Morantes et al., 2024), which show that HCHO is the most important VOC in buildings, and others can largely be ignored by standards.

When Correlation Becomes Causation

Built environment research frequently slips into causal language when associations are observed. The problem is not just with motives or funding sources, but with the methods themselves. Selection bias, publication bias, and weak statistical controls distort the narrative. Without the pre-registration of studies or formal quality assessment tools like GRADE or ROBINS-I, it becomes difficult to separate real effects from artefacts (Guyatt 2008; Sterne 2016).

Evidence of long-term effects on academic attainment is especially thin. Most studies cover short time frames and fail to track outcomes beyond immediate test scores. The lack of randomised trials (where people are randomly assigned to different groups to fairly test the effects of a treatment or intervention) or high-quality quasi-experiments (the testing of an intervention without random assignment but with a careful design to make the comparison groups as similar as possible) means the field is still guessing. As a result, it is possible to cherry-pick findings that support almost any conclusion. Cherry picking creates a false sense of certainty and leads to overconfident policy recommendations.

The most egregious example of this was a study of Italian-schools that claimed to show that increasing ventilation through mechanical ventilation reduces SARS-CoV-2 airborne transmission by over 80%, but the ventilation rates for naturally ventilated schools were all assumed to be 0.5 ACH and not measured, a likely huge underestimation of actual flow rates (Buonanno 2022). Many other confounders such as geographical location were not considered (Madhusudanan 2023).

Reducing the transmission of pathogens that cause acute respiratory infections (ARI), such as SARS-CoV-2, in schools is challenging. There are several reasons for this. Dr Mildred Weeks, in early studies on upper-room UV, warned that these interventions were unlikely to reduce ARI transmission at the community level. This was, she argued, “because of the short incubation period believed to be characteristic of such illnesses, the multiplicity of exposure in and out of the school, and the confusion of the patterns of spread by the inclusion of adults in the population at risk.” Despite these reservations, the Millbank Memorial Fund funded trials of upper-room UV in schools. As predicted, the studies found no reduction in the incidence of ARI (Downes, 1950).

Other factors also limit the effectiveness of ventilation and air cleaning. Close-range transmission dominates in many settings, where air cleaning and ventilation do little to reduce dose. Over time, population immunity also shrinks the susceptible pool. Moreover, emission rates of pathogens vary dramatically between individuals — by as much as six orders of magnitude. When emission rates are low, the concentration of pathogen in room air is already very low, even in unventilated spaces (Jones 2024).

However, in naïve populations, equivalent ventilation may have some role in slowing the spread of infection. Perkins et al. (1947) observed slower spread of measles in classrooms with upper-room UV, though the total number of cases across the school remained similar to the control.

Reviewing literature

A gold-standard literature review, like those published by the Cochrane Collaboration, is a systematic review that follows strict, transparent methods to gather and assess all the best available research on a specific question. It starts with a clearly defined question and uses comprehensive search strategies to find every relevant study, not just the most convenient or well-known ones. Reviewers assess the quality of each study using standard criteria, often focusing on randomised trials. They then summarise the results, sometimes using meta-analysis to combine findings statistically. Every step is carefully documented to reduce bias, so the review is repeatable and the conclusions as reliable as possible.

The 2023 Royal Society’s systematic review of the effectiveness of non-pharmaceutical interventions against the transmission mission of SARS-CoV-2 found that the evidence suggests that ventilation, air cleaning, and reduced room occupancy may help reduce transmission in some settings, but the supporting studies were generally of low quality (they were assessed to have a critical risk of bias in at least one area), so confidence in this conclusion is also low (Madhusudanan 2023). Whereas the 2022 Lancet COVID Commission was not a systematic review, but an expert consensus report informed by a targeted literature review (Sachs 2022). This means that it cannot claim to be an unbiased or comprehensive synthesis of the evidence and should be considered as expert guidance informed by selected studies, not as a definitive evidence summary. This is probably why it based its recommended airflow rates on the confounded Italian Schools study mentioned earlier.

Better Evidence Please

In medicine and public health, strong claims require strong evidence. Randomised controlled trials (RCTs) sit at the top of evidence hierarchies, followed by cluster trials, quasi-experiments, cohort studies, and cross-sectional surveys. Each has a place, but they differ in how confidently they can support cause-and-effect claims. In building science, these hierarchies are rarely applied. Instead, we rely heavily on observational data and post-hoc analysis, sometimes with minimal transparency about methods or assumptions.

Tools like GRADE and ROBINS-I help rate the quality of evidence and assess the risk of bias. They are standard in clinical research but seldom used when evaluating environmental interventions in schools. Applying them here would expose the uncertainty behind many IAQ–performance studies and help guide decisions toward more credible findings.

Until better trials are conducted, we should be cautious about promoting increased ventilation and air cleaning as mechanisms to improve academic performance and well-being. Meeting statutory ventilation standards is a sensible baseline. But going further — particularly when it involves capital investment, energy use, and disruption — should be justified by more than suggestive data.

What Should Be Done?

Beyond meeting current standards, the case for further improvement on academic grounds is weak. The pollutants most likely to affect health and learning, such as formaldehyde, PM2.5, HCHO, NO2, and O₃, are well known from studies in homes and offices. These should be the focus of our attention. Here, ASHRAE is leading the way (Jones, 2023), its committee on Ventilation and Acceptable Indoor Air Quality in Residential Buildings (62.2) has produced an addendum that would add a harm-based IAQ procedure as an alternative compliance method (Jones 2023). The procedure is harm-based rather than based on individual contaminant limits. The 62.2 committee has used the harm analysis in homes to determine that its IAQ Procedure needs to consider only 3 contaminants, and only the sum of the harm from those three contaminants needs to be limited. The concept has broad implications and should be considered for school.

But the assumption that marginal improvements in ventilation will meaningfully boost cognition or attainment is not supported by strong evidence. We should study the effects of ventilation on academic attainment and behaviour properly. The cost-benefits should be compared against other interventions, like giving children breakfast, improving teacher quality, reducing class sizes, and increasing physical activity.

Conclusion: Better Trials, Not Bigger Fans

The drive to improve IAQ in schools is well intentioned. It reflects a desire to protect children and promote learning, which are goals we share. But we must distinguish between what feels intuitively right and what is scientifically supported. The belief that ventilation boosts academic performance is not backed by strong, consistent, or causal evidence. Where studies do exist, they are often thin on rigour and thick with noise.

Schools should meet existing ventilation standards. That is a reasonable and defensible position. But claims that more air will improve cognition, behaviour, or test scores need better support. Overstating these effects risks misdirecting public funds and eroding trust in our industry when expected gains do not materialise.

In a world of limited budgets, we must be honest about trade-offs. Every intervention comes with capital, operational, and environmental costs. When those costs are weighed against uncertain benefits, we owe it to schools and students to demand better evidence. That means fewer headlines and more high-quality trials. Until then, we should focus on what we know works.

References

Allen, J. G., Eitland, E., Klingensmith, L., MacNaughton, P., Cedeno Laurent, J., Spengler, J., & Bernstein, A. (2021). Schools for Health: Foundations for Student Success. Harvard T.H. Chan School of Public Health. https://healthybuildings.hsph.harvard.edu/wp-content/uploads/2024/10/Schools_ForHealth_UpdatedJan21.pdf.

American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE). (2022). ANSI/ASHRAE Standard 62.1-2022: Ventilation for acceptable indoor air quality. ASHRAE.

ASHRAE. (2025). ASHRAE position document on indoor carbon dioxide. American Society of Heating, Refrigerating and Air-Conditioning Engineers. https://www.ashrae.org/file%20library/about/position%20documents/pd_indoorcarbondioxide_2025.pdf.

Buonanno, G., Ricolfi, L., Morawska, L., & Stabile, L. (2022). Increasing ventilation reduces SARS-CoV-2 airborne transmission in schools: A retrospective cohort study in Italy’s Marche region. Frontiers in Public Health, 10. https://doi.org/10.3389/fpubh.2022.1087087.

EN 16798-1:2019, Energy performance of buildings – Ventilation for buildings – Part 1: Indoor environmental input parameters for design and assessment of energy performance of buildings addressing indoor air quality, thermal environment, lighting and acoustics – Module M1-6, European Committee for Standardization, Brussels, 2019.

Department for Education. (2018). Building Bulletin 101: Guidelines on ventilation, thermal comfort and indoor air quality in schools. https://www.gov.uk/government/publications/building-bulletin-101-ventilation-for-school-buildingsGOV.UK [accessed 12th May 2025].

Dedesko, S., Pendleton, J., Young, A. S., Coull, B. A., Spengler, J. D., & Allen, J. G. (2025). Associations between indoor air exposures and cognitive test scores among university students in classrooms with increased ventilation rates for COVID-19 risk management. Journal of Exposure Science and Environmental Epidemiology. https://doi.org/10.1038/s41370-025-00770-6.

Downes, J. (1950). Control of Acute Respiratory Illness by Ultra-Violet Light. American Journal of Public Health and the Nations Health, 40(12), 1512–1520. https://doi.org/10.2105/AJPH.40.12.1512.

Guyatt, G. H., Oxman, A. D., Vist, G. E., Kunz, R., Falck-Ytter, Y., Alonso-Coello, P., & Schünemann, H. J. (2008). GRADE: An emerging consensus on rating quality of evidence and strength of recommendations. BMJ, 336(7650), 924–926. https://doi.org/10.1136/bmj.39489.470347.AD.

Jones, B., Sharpe, P., Iddon, C., Hathway, E. A., Noakes, C. J., & Fitzgerald, S. (2021). Modelling uncertainty in the relative risk of exposure to the SARS-CoV-2 virus by airborne aerosol transmission in well mixed indoor air. Building and Environment, 191. https://doi.org/10.1016/j.buildenv.2021.107617.

Jones BM. Dallying with DALYs: A Proposed Harm-Based IAQ Procedure for Standard 62.2. ASHRAE Journal. 2023:24-7.

Jones, B., Iddon, C., & Sherman, M. (2024). Quantifying quanta: Determining emission rates from clinical data. Indoor Environments, 1(3), 100025. https://doi.org/10.1016/j.indenv.2024.100025.

Levitt, Steven D., and John A. List. 2011. "Was There Really a Hawthorne Effect at the Hawthorne Plant? An Analysis of the Original Illumination Experiments." American Economic Journal: Applied Economics 3 (1): 224–38.

Madhusudanan, A., Iddon, C., Cevik, M., Naismith, J. H., & Fitzgerald, S. (2023). Non-pharmaceutical interventions for COVID-19: a systematic review on environmental control measures. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 381(2257). https://doi.org/10.1098/rsta.2023.0130.

Morantes G, Jones B, Molina C, Sherman MH. Harm from Residential Indoor Air Contaminants. Environmental Science & Technology. 2024;58(1):242-57.

Perkins, J. E., Bahlke, A. M., & Silverman, H. F. (1947). Effect of Ultra-violet Irradiation of Classrooms on Spread of Measles in Large Rural Central Schools Preliminary Report. American Journal of Public Health and the Nation’s Health, 37(5), 529–537. http://www.ncbi.nlm.nih.gov/pubmed/18016521.

Persily, A., & de Jonge, L. (2017). Carbon dioxide generation rates for building occupants. Indoor Air, 27(5), 868–879. https://doi.org/10.1111/ina.12383.

Sachs, J. D. et al. (2022). The Lancet Commission on lessons for the future from the COVID-19 pandemic. The Lancet, 400(10359), 1224–1280.

Sterne, J. A. C. et al. (2016). ROBINS-I: A tool for assessing risk of bias in non-randomised studies of interventions. BMJ, 355, i4919. https://doi.org/10.1136/bmj.i4919.

U.S. Environmental Protection Agency (USEPA). (2024, April 10). National ambient air quality standards (NAAQS) table. https://www.epa.gov/criteria-air-pollutants/naaqs-table.

World Health Organization. (2021). WHO global air quality guidelines: Particulate matter (PM₂.₅ and PM₁₀), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide. https://www.who.int/publications/i/item/9789240034228.