Of the four classroom conditions, air has the strongest evidence

Evidence · May 2026 · 7 min read · 20 sources

Of the four classroom conditions I analysed, air has the strongest evidence: studies of three different designs link fresher air with higher test scores. Sound has one London study behind it. The best-known daylight result, from 1999, did not recur when the same group studied Fresno in 2003.

Written by

Christian Huser
neuecon partners

A district can measure air and sound in the rooms it already has.

Pupils in rows seen from behind in a brick hall, daylight through tall arched windows.

Boston's 4,000 sensors

Boston Public Schools measures the air in its classrooms before it decides what to build. In May 2026, it was the one district that did so among the 20 sources I analysed for a report for The Built Review on four classroom conditions (air, sound, light and green space) and on the design of the room as a whole. Denver and Montgomery County have announced comparable programmes.

Since 2021, the district has installed more than 4,000 air sensors in more than 100 schools. Each sensor reads temperature, humidity, carbon dioxide, carbon monoxide and two size classes of airborne particles, PM2.5 and PM10, and parents can follow the readings on a public dashboard in real time. The environmental health researcher Patricia Fabian and her group at the Boston University School of Public Health analyse the data. The readings showed faults in the ventilation systems before the equipment failed. When I analysed the sources, Fabian's group was still building the models that link the readings to how pupils learn.

The programme cost about USD 7 million from federal COVID relief and close to USD 200,000 from a city clean-air grant, for the sensors, their installation, the dashboard and several years of analysis. Sensor hardware costs a few hundred US dollars per room.

Mendell's figure is the one for a school board

Pupils breathe out carbon dioxide, so a high level in a classroom is a sign of little fresh air. In 2020, the indoor-environment researcher Pawel Wargocki and colleagues pooled published studies of classroom carbon dioxide and school performance, and their model puts the gain from a fall from 2,300 to 900 parts per million at about 5% on tests of learning progress. No single study measured that 5%.

A comparison across 70 schools found the same direction in 2015. The public health researcher Ulla Haverinen-Shaughnessy and the indoor-air researcher Richard Shaughnessy measured the ventilation in 140 fifth-grade classes in these schools, with 3,109 pupils. Each additional litre of fresh air per second per person went with about 7 more points on a standardised maths test, with a 95% confidence interval of 1 to 12 points, so chance alone is unlikely to explain the link. Without outliers, the figure was 11 points, with an interval of 2 to 20. The classes had on average 3.6 litres of fresh air per second per person.

I would give a school board the figure of the indoor-air researcher Mark Mendell and colleagues, because they recorded carbon dioxide in each classroom every day. In a study published in 2016, they did so in about 150 classrooms in 28 California schools and estimated the ventilation from the readings. Each 10% more ventilation went with 0.6 more points on the English test. If ventilation had no link with the scores, chance would produce a result this large in about one study in a hundred. For maths, the figure was similar, and chance could explain it. Mendell and colleagues concluded that most of their models showed small positive associations and that few of them ruled out chance.

For sound, I found one study. In 2008, the acoustics researcher Bridget Shield and the psychologist Julie Dockrell measured the noise outside 142 London primary schools and inside 16 of them and compared it with the national test results of pupils aged 11. Schools with more noise outside had lower results, with correlations around minus 0.4, where 0 would mean no link. This link remained after they controlled for the shares of pupils with free school meals, a measure of poverty, with English as an additional language and with special educational needs. The noise inside the occupied classroom went with lower English results at a correlation around minus 0.7. The psychologist Jacob Cohen set the convention that a correlation of 0.5 in either direction counts as large.

A district that measures sound should measure it with pupils in the room, because Shield and Dockrell found their stronger link there. The US standard ANSI/ASA S12.60 of the Acoustical Society of America and the American National Standards Institute sets its targets for the empty room: a reverberation time below 0.6 seconds, the time a sound takes to fade after it stops, and background noise below about 35 decibels, weighted to human hearing. I found no study that tested classroom sound with a second kind of design.

Fresno's results went with the view out

The case for daylight comes from one group, and its second large study did not repeat its first. In 1999, the architect Lisa Heschong and her firm, the Heschong Mahone Group, analysed the records of more than 21,000 pupils in more than 2,000 classrooms in three districts for the utility Pacific Gas and Electric. Only the Capistrano district in California had data on progress over a year. There, pupils in the most daylit rooms progressed about 20% faster in maths and 26% faster in reading than pupils in the least daylit rooms. Heschong and colleagues analysed the data again in 2002 with further controls, among them a check on whether more experienced teachers had the brighter rooms, and the Capistrano result remained.

In 2003, Heschong's group studied more than 8,000 pupils in 450 classrooms in Fresno for the California Energy Commission, and there the amount of daylight did not predict progress. A better view out of the window went with higher results, and glare and direct sun with lower ones.

A 2007 committee questioned figures like Barrett's 16%

Barrett's 16% comes from one model of one school year. In 2015, the built-environment researcher Peter Barrett and colleagues at the University of Salford studied 3,766 pupils in 153 classrooms in 27 schools. Their multilevel model separates the effect of the room from the effect of the pupil, and in it the physical design of the classroom explained about 16% of the differences in pupils' progress over one school year. I found no study that rebutted the figure.

The objection to such figures is eight years older. In 2007, a committee of the US National Research Council reviewed the evidence on green schools and wrote in its Finding 8 that the methods that correlate the condition of a building with pupil achievement cannot show whether any relationship exists. Pupils are not assigned to schools at random, and poorer and minority children are over-represented in older and worse buildings, so an apparent effect of the building can be in part an effect of poverty. Finding 8 applies to Barrett's model. The London noise study met it with its control for free school meals.

Designers work with targets for the whole room, such as 20 to 50% of a wall left clear, or a minimum share of green space in the playground. Barrett and colleagues found that walls of moderate visual complexity went with more progress than cluttered or bare walls, and they tested no exact share.

Barcelona measured green space from satellites

Green space around a school went with better working memory in one large study. In 2015, the environmental epidemiologist Payam Dadvand and colleagues tested about 2,600 children aged seven to ten in Barcelona four times over twelve months. Children with more greenness at and around their school developed their working memory more and were less inattentive, and lower air pollution explained part of that link. Dadvand and colleagues measured the greenness on satellite images and tested no change to a schoolyard.

Measure air and sound first

The evidence puts air ahead of sound, and sound ahead of daylight and whole-room design. A district can measure air and sound in the rooms it already has. For daylight, I would spend the budget on the view out and on glare control, the two features that went with results in Fresno.

Sources

Related

Evidence · Apr 2026

Daylight on a ward went with less pain medication after surgery →