We Need More Research on how CO2 Affects Cognition
A growing body of research indicates there are potentially large health benefits to improving indoor air quality. Indoor spaces are often poorly ventilated and expose occupants to high levels of contaminants, such as pathogens, particulates, and other pollutants. Because people spend nearly all their time indoors, improving indoor air quality would hugely benefit public health. (For more on the potential benefits of improving indoor air quality, see Juan Cambeiro and my recent IFP paper.)
One pollutant of indoor air is carbon dioxide (CO2). Outdoors, average CO2 levels are around 400 parts per million (ppm), or 0.04%. A building that meets ASHRAE indoor air quality standards will have fewer than 1000 ppm of CO2. However, because buildings are often poorly ventilated, in practice many buildings (particularly high-occupancy buildings such as schools) will have levels much higher than this:
Historically, CO2 was mostly used as an indicator of indoor air quality – since people exhale CO2, indoor CO2 concentrations will rise in the absence of good ventilation. But within the last several decades, evidence has emerged that CO2 itself might be a pollutant, and that even slightly elevated levels might have negative health effects, particularly on cognition.
If CO2 levels affect cognitive function, the amount they do matters a great deal. A cost-benefit analysis on increased ventilation done for the UK’s Royal Academy of Engineering estimated that improvements in productivity resulting from better ventilation (which air quality experts often attribute in part to reduced CO2) were about 1.4% per year, or about £1.5 billion per year (roughly $15 billion a year if applied to the US). In “Healthy Buildings," Allen and Macomber likewise project that if improved cognitive function from better ventilation translates to higher worker productivity, improvements to ventilation pay for themselves many times over, in the form of increased business revenue (per their cost-benefit analysis, productivity effects would massively outweigh every other potential benefit of increased ventilation).
But there’s a problem: the literature on CO2s effects on cognition is extremely inconsistent. In some studies, relatively small increases in CO2 levels cause large impacts on cognitive performance. Other studies don’t find any cognitive impacts at all, even when CO2 levels are ten to one hundred times higher than atmospheric levels. We’re left with an extremely muddy picture of the relationship between CO2 levels and cognition.
Studies on the effects of CO2 on cognitive performance
The best survey of evidence of the effects of CO2 on cognition is the review article Du et al 2020. The authors looked at 37 different studies of the effects of CO2 on cognitive performance, dating back to 1982. Studies fell into two broad categories: studies that manipulated CO2 directly (such as by releasing CO2 from stored tanks), and studies that manipulated it indirectly by altering the level of ventilation. Bear in mind that studies that manipulated ventilation would also affect the concentration of pollutants other than CO2, such as particulates or bioeffluents, which may also have effects on cognition.
The studies reviewed used a variety of different measures of cognitive performance. Some studies used very low-level measures of performance, such as ability to do arithmetic, or find typos in printed text. Others used more abstract measures. Many studies used the Strategic Management Simulation (SMS), which assesses “high-level decision-making performance” (Du et al). And some studies measured worker performance directly. Measures such as memory, learning ability, and reaction time were also looked at.
The review finds mixed results. On the one hand, many studies found that increasing CO2 levels resulted in worse cognitive performance. About half the studies that manipulated CO2 directly, and around 85% of the studies that manipulated ventilation, found worse performance when CO2 levels were higher. Negative impacts were found across a variety of different measures of performance. The authors note that “for all measurements of work performance, and for most measurements of brain function…there is at least one study showing a significant cognitive impact.”
For instance, in Allen et al 2016, 24 subjects performed the SMS test over a period of a week, while indoor environmental conditions (including CO2 and VOC level) varied. At higher CO2 levels, performance on the test showed significant declines - the authors note that “On average, a 400-ppm increase in CO2 was associated with a 21% decrease in a typical participant’s cognitive scores across all domains after adjusting for participant.”
Satish et al 2012, also using the SMS in a study of 24 subjects, found lowered performance at 1000 ppm and 2500 ppm CO2 levels, compared to 600 ppm:
A third study, also by Allen, looked at the effect of CO2 levels on pilots. 30 pilots flew several flights in a flight simulator at different levels of CO2 (700 ppm, 1500 ppm, and 2500 ppm). Pilot performance, as evaluated by an FAA Pilot Examiner, was worse at 2500 ppm, and there was “suggestive evidence” of lower performance at 1500 ppm.
The above studies suggest that CO2 has a negative effect on performance. But other studies found no effects, even when looking at the same cognitive measures of performance. Du et al note that “within each brain function or work performance category…often only half or fewer of the tests see significant changes in cognitive function. Furthermore, even the same cognitive test can have contrary findings across different studies or between measurements of speed and accuracy.”
For instance, two studies found logical reasoning speed declined with elevated CO2 levels: Wargocki and Wyon 2007a, which studied two classrooms of 10 to 12 year old children exposed to different levels of ventilation (resulting in 900 ppm and 1300 ppm CO2 levels respectively), and Sayers et al 1987, which studied 10 undergraduates performance at CO2 levels up to 75,000 ppm. However, three other studies – Sheehy et al 1982 (n = 4-6), Bloch-Salisbury et al 2000 (n = 9), and Wargocki and Wyon 2007b (n=32-45) – found no such effects.
Du et al also note that “generally no dose-response relationship between CO2 concentrations and performance changes was found.” That is, higher CO2 levels didn’t necessarily result in worse cognitive performance:
On the contrary, a number of changes in performance at relatively low CO2 concentrations (approximately 1000 ppm) are contradicted at higher CO2 concentrations (approximately 3000 ppm). For example, adding CO2 and reducing ventilation rates caused impaired or improved attention performance (measured by d2 test accuracy), respectively, at CO2 concentrations of approximately 1100 ppm, but the effects were reversed when CO2 concentrations reached 3000 ppm.
However, many of these studies had very small numbers of participants, or were otherwise underpowered. Of the studies that manipulated CO2 directly, 6 of them had fewer than 10 participants.
In other cases, the studies were performed on physiologically distinct populations. Many of the studies were done on people in specialized professions such as divers, astronauts, and naval submarine operators. Not only will members of these professions be much more physically fit than average (a naval submariner must pass the Navy Physical Readiness Test several times a year), but they’ll often have experience operating in environments of very high CO2. In nuclear submarines, CO2 levels may be at up to 5000 ppm for 90-day durations, and for short periods are allowed to be as high as 40,000 ppm, 100 times atmospheric levels. In spacecraft, CO2 levels are typically between 2000 to 5000 ppm, and may reach 10,000 ppm or above for short periods of time. This makes it hard to generalize findings on these sorts of subjects.
To address this, Du et al also looked at a subset of studies that both met certain criteria for strength, and that didn’t involve members of specialized professions, such as astronauts, submariners, and pilots. This reduced the 37 studies down to nine (including Allen et al 2016 and Satish et al 2012 mentioned above): four which manipulated CO2 directly, four which manipulated ventilation, and one which manipulated both. For each study, they calculated the relative effect size found.
For the studies that looked at CO2 directly, the only statistically significant finding was decreased performance on the SMS at elevated CO2 levels.
For studies that manipulated ventilation levels, Du et al found that “many cognitive functions have a relatively consistent decline in speed measurements with lower ventilation (attention, executive functioning, reasoning, calculation and text processing).” Though speed was found to significantly decrease, accuracy wasn’t. In fact, the authors note that in many cases accuracy increased at higher levels of CO2.
However, even here the authors note that the results are inconsistent:
For example, addition speed and error rates were influenced but not subtraction, and significant changes in reasoning, visual perception, and text processing performance were only observed in one of the 3 or 4 studies for speed or accuracy measures.
As we noted earlier, some professions, such as astronauts and submariners, are routinely exposed to extremely high levels of CO2 – submarine CO2 levels can be 10 to 100 times as high as atmospheric levels. Because of this, the military has been examining the effects of CO2 on cognitive performance for decades. Many of these studies aren’t included in the Du et al review, but are nonetheless interesting to consider.
In Manzey and Lorenz 1998, four subjects were confined for 26 days at CO2 levels up to 12000 ppm, and were given a test battery “including grammatical reasoning, memory search, unstable tracking, and dual tasks." No effects were found at 7000 ppm, and while tracking task performance declined at 12,000 ppm, the effect was small enough that it did “not appear to be of operational relevance."
The National Research Council’s guidance levels for submarine contaminants references several studies showing no impacts of high levels of CO2:
A number of studies suggest that CO2 exposures in the range of 15,000-40,000 ppm do not impair neurobehavioral performance. Schaefer (1961) reported that 23 crewmen exposed to CO2 at 15,000 ppm for 42 days in a submarine showed no psychomotor testing effects but showed moderate increases in anxiety, apathy, uncooperativeness, desire to leave, and sexual desire. In a 5-day exposure of seven subjects at a CO2 concentration of 30,000 ppm, Glatte et al. (1967) reported no effects on hand steadiness, vigilance, auditory monitoring, memory, or arithmetic and problem solving performance. Storm and Giannetta (1974) studied the effects of 2 weeks of exposure to CO2 at 40,000 ppm on psychomotor performance in a 6-week protocol that included a 2-week pre-exposure baseline period and 2 weeks of recovery. Twenty-four volunteers, ages 18-23, were selected for their motivation and their excellent health…
CO2 exposure did not affect performance on the tracking task or any of the six RPM subtests (Storm and Giannetta 1974). There was a learning effect for the tracking task during both pre-exposure and recovery, but the authors still thought it was appropriate to conclude the absence of a performance impairment. The authors considered it especially likely because previous papers had suggested that impairment is easier to detect during skill reacquisition, which occurred following the 2 weeks of exposure without practice, rather than at an asymptotic skill level (Storm and Giannetta 1974). Thus, CO2 at 40,000 ppm for 2 weeks did not affect performance on multiple tests of cognitive function in physically fit young airmen, a population probably not unlike submariners.
Finally, a 1941 study on exposure up to 60,000 ppm of CO2 found that “3-4% of carbon dioxide at atmospheric pressure caused no deterioration in manual or arithmetical skill, and in the two subjects tested, 6% of carbon dioxide caused no deterioration." And a 1947 study of exposure to CO2 up to 50,000 ppm for up to 72 hours found that “there was no consistent significant effect upon any of the auditory or visual functions measured, nor upon any of the paper-and-pencil test scores, after the subjects practiced.”
These submarine studies must be taken with a grain of salt. For one, many of them have extremely small sample sizes (the 1941 study tested just two participants!) For another, as we mentioned earlier, submariners are a physiologically distinct population, and findings may not generalize (though it would be interesting if “resist the negative effects of high CO2” was another benefit of regular exercise and physical fitness). But it’s nevertheless interesting that studies of very, very high levels of CO2 don’t seem to find negative cognitive impacts.
If high levels of CO2 had negative cognitive effects, it would be a very large public health issue. Because people spend so much time indoors, and because so many buildings have poor ventilation, the collective impact would be large even if the individual effects were small.
However, the evidence so far seems mixed. Many studies have found negative cognitive impacts on elevated levels of CO2, but many have not. There does not seem to be a clear dose-response relationship between CO2 levels and cognitive performance, or an aspect of cognition that is reliably affected outside of one measure of abstract thinking (the SMS). Outside of the SMS, the strongest studies do not seem to have found significant effects when CO2 levels are manipulated directly.
Improving ventilation seems to more reliably improve cognitive performance, but this also changes other environmental factors other than CO2 levels. It’s quite possible that those other environmental factors hamper cognition, while CO2 is less of a factor. What’s more, the Navy and NASA have long operated submarines and spacecraft at levels of CO2 that are 10-100 times atmospheric levels without (apparently) ill-effects, and several military studies have failed to find effects on cognitive performance on very high levels of CO2, though the distinctiveness of submariners and astronauts make generalizing this finding fraught.
Given the potential size of the harms and the fuzziness of existing findings, more research here is needed.
Thanks to Juan Cambeiro, Willy Chertman and Davin Pavlas for reading drafts of this. All errors are my own.