One or two research groups have published work on carbon dioxide and cognition. The state of the published literature is confusing.

Here is one paper on the topic. The authors investigate a proprietary cognitive benchmark, and experimentally manipulate carbon dioxide levels (without affecting other measures of air quality). They find implausibly large effects from increased carbon dioxide concentrations.

If the reported effects are real and the suggested interpretation is correct, I think it would be a big deal. To put this in perspective, carbon dioxide concentrations in my room vary between 500 and 1500 ppm depending on whether I open the windows. The experiment reports cognitive effects for moving from 600 and 1000 ppm, and they are relatively large compared to interindividual differences.

I haven't spent much time looking into this (maybe 30 minutes, and another 30 minutes to write this post). I expect that if we spent some time looking into indoor CO2 we could have a much better sense of what was going on, by some combination of better literature review, discussion with experts, looking into the benchmark they used, and just generally thinking about it.

So, here's a proposal:

  • If someone looks into this and writes a post that improves our collective understanding of the issue, I will be willing to buy part of an associated certificate of impact, at a price of around $100*N, where N is my own totally made up estimate of how many hours of my own time it would take to produce a similarly useful writeup. I'd buy up to 50% of the certificate at that price.
  • Whether or not they want to sell me some of the certificate, on May 1 I'll give a $500 prize to the author of the best publicly-available analysis of the issue. If the best analysis draws heavily on someone else's work, I'll use my discretion: I may split the prize arbitrarily, and may give it to the earlier post even if it is not quite as excellent.

Some clarifications:

  • I reserve the right to call off the whole thing at any time during the next week (April 5 - April 11), if there is a simple reason why indoor CO2 is not practically relevant / these results are grossly misleading and someone brings it up in the comments.
  • The metric for quality is "how useful it is to Paul." I hope that's a useful proxy for how useful it is in general, but no guarantees. I am generally a pretty skeptical person. I would care a lot about even a modest but well-established effect on performance. 
  • These don't need to be new analyses, either for the prize or the purchase.
  • I reserve the right to resolve all ambiguities arbitrarily, and in the end to do whatever I feel like. But I promise I am generally a nice guy. 
(Thanks to Andrew Critch for mentioning these results to me and Jessica Taylor for lending me a CO2 monitor so that I could see variability in indoor CO2 levels. I apologize for deliberately not doing my homework on this post.)





More posts like this

Sorted by Click to highlight new comments since: Today at 6:20 AM

CO2 has been used a marker for acceptable ventilation since the 1920s. Commercial building ventilation systems are often linked to CO2 sensors in the return air, and increase or decrease outdoor ventilation air to meet a set target. Typically 800-1000ppm. This is known as "demand control ventilation", and is pretty common, especially in newer buildings. A tricky thing with this is that the sensors are notorious for drifting out of calibration over time, so many system have minimum damper positions built in to make sure enough fresh air is getting to the space.

Residential and commercial ventilation is regulated under ASHRAE standard 62.1 and 62.2 The standards are set by committee, and are a big battle/tradeoff between (1) energy cost from heating, cooling, and moving all that ventilation air (2) odor, (3) air quality risks (like NOx, particulates, VOCs), and (4) impairment from CO2.

In the most recent update, ASHRAE finally removed the infiltration allowance for residential buildings. Before, residential buildings could be built with just exhaust fans, with the fresh air being made up through infiltration, meaning air coming through cracks around walls, windows, doors, etc. But there was disagreement as to whether this air is really fresh. Insulation tends to "filter" the air and remove particulates, but it can also pick up bad stuff like mold spores, VOCs, etc. The decision was to no longer count infiltration, largely because buildings are getting tighter and shouldn't be leaking that much. Plus infiltration can pull in warm, moist air which can condense in walls if the building is air conditioned, causing mold and water damage. Now, residential must match commercial buildings in supplying fresh, filtered dedicated supply air. That will be BIG change in residential construction practice if/when states adopt it. (1/3 of states don't follow it, 1/3 are keyed to old 2004,2007,2010 standards, 1/3 are recent). For now I recommend cracking open windows, running bath fans occasionally, and ESPECIALLY installing and using a cooking range hood exhaust fan while cooking. Can't emphasize the kitchen range hood enough - thats probably #1 most important thing to do to improve air quality. Well, I'm assuming you're not smoking indoors and not having campfires in your backyard. Those are worse.

There is some question as to whether the ventilation recommendations should be stricter. CO2 is a good proxy for a lot of other air pollutants that are harder to measure. So performance or other benefits from reducing CO2 levels may be coming from getting rid of other air pollutants in the space (VOCs, NOx in particular). However, some recent work out of LBNL has found some slight differences in performance between 600 and 1000ppm. (study) It would be very tough to stay at 600ppm for most buildings, because that would require a LOT of ventilation air. Doable with natural ventilation, but in places with cold winters? That gets really expensive.

For CO2, OSHA permissible exposure limit for workers is 5000ppm. And NIOSH's short term exposure limit is 30,000ppm, though that would be really bad to be in. There will certainly be cognitive impairment at these levels, but it probably won't cause damage.

If your concern is productivity, work in a ventilated space. Or just open windows / turn on the bath fan / range hood just before you start until after you finish working at home.

If you really want to see some elevated CO2 levels, get in a car with a few other people and go on a road trip without rolling down the windows. And then understand why everyone gets groggy and tired.
ASHRAE also sets ventilation standards for moving transport (airlines, trains, buses), but people are in charge of ventilation in their personal cars or trucks and don't ventilate as much. Or big trucks like semis, where the drivers are already impaired from driving long hours. eek.

If you want more information on air quality and ventilation, check out LBNL's indoor air group: Or NIST's group:

Happy to answer more questions about ventilation / air quality. Interesting field. the elevator "why this is important" (1) air pollution is a big risk factor a lot of health problems (2)humans spend most of their time indoors (90% + in developed world) (3)a lot of that time is spent sleeping. (4)lots of the world still cooks in their homes with biomass, unventilated. And this pollutes the air in the whole neighborhood. Big problem. Make sure your kitchen, bedroom, and workplace are well ventilated.

Thanks for the information about ventilation.

I linked to the LBNL study in my post. You call the differences "slight," but I'm not yet convinced. Naively, the effects look big enough to justify the expense of maintaining very low co2 levels (e.g. opening windows + running the heat in winter, changing where I work). What I want to better understand is whether those effects are real, what the actual effect size is for tasks I care about, and so on.

I looked at the study again, the reported effects from 2500ppm also just seem totally implausible. On many of the scales they report drops from e.g. 80th percentile to 30th percentile.

10,000ppm causes heavy breathing and confusion, from 50,000ppm upwards you can die in a couple of hours. I don't think going from 80th to 30th on complex cognitive tasks is totally implausible with an increase from 600 to 2500ppm.

I don't understand why you find it so surprising given that (it seems) you had no previous knowledge of the area. Where did your prior come from? I can't be surprised by the fact the Planck length is 1.61x10^-35 and not 2.5x10^-30 given that I had not idea what it was.

  • After using a CO2 monitor in my house, I know that my room goes from 400 to 1300 depending on ventilation and is not noticeably stuffy (I can't reliably tell whether the window is open without looking). Other rooms in the house go above 2000 with windows closed. A difference between 80th and 30th percentile looks like a big difference that would be easily noticed, either introspectively or by someone else.

ETA: for reference, that's a 500/2400 point drop on the SAT.

  • Given the range of indoor CO2 levels in buildings, this effect looks like it is extremely economically relevant, so I'd expect to have encountered it before (and would expect people to take it more seriously). So that leads me to expect that my interpretation is wrong.

Lots of good points here. A couple comments: 1) I don't think it is so critical to ventilate an electric oven/range. But it is important for natural gas and critical for propane or anything else. 2) it is most important to crack open windows and use bathroom fans when the outside temperature is close to the inside temperature because that is when you get the least infiltration from buoyancy (and the increase in heating/air conditioning energy use is small). 3) the increased heating/cooling energy can be mitigated with a heat recovery ventilator

CO2 in the bloodstream is sedating and many people are averse to being in "stuffy" rooms so this does have an air of plausibility to it. Great initiative!

Yeah, I really hope someone breathes some life into this issue.

It will never work, you have your head in the clouds.

I agree this is good to run to ground. But one thing to keep in mind is that I believe the oxygen that one exhales is about four percentage points lower than inhaled. That means on average in the lung the oxygen is about two percentage points lower, meaning the average carbon dioxide level is about 20,000 ppm. So I would be surprised to see big impacts based only on CO2 changing around 500 ppm because you can get bigger changes in the lung CO2 just by changing breathing rate slightly.

But I promise I am generally a nice guy.

I'm very happy to vouch for this ;-)

Thanks Paul, nice idea. Look forward to reading what people come up with!

Another recent study that shows just controlling CO2 doesn't yield the same performance losses. Maybe it is something else in the air?

Air Pollution, Temperature and Cognitive Performance in the Short Run: Evidence from Women’s Ability to Recall Dates

Abstract: Cognitive performance is important to productivity across many fields and potentially correlated to air pollution and extreme temperatures. We study the effects of daily ambient air pollution and monthly temperature on women’s ability of recalling dates across 42 developing countries from 1997 to 2009. We use an estimated natural air pollution data, and calculate the AQI to get an aggregate effect of air pollution. We find that one standard deviation increase in the AQI leads to a statistical decrease in women’s probability to recall dates such as birthdays, marriage date or children’s birthdays by 0.44 percentage point. Furthermore, there is a nonlinear effect of air pollution with a suggesting AQI threshold 150. We also find each degree day above 30°C increase the probability of women fail to recall children’s birthdays by 0.17 percentage point. Moreover, by doing a sub-sample estimation, we find that air pollution and temperature particularly affect uneducated women.

Hmm, variables correlated with air pollution across nations would seem very difficult to control for.

Was the winner decided?

Here's another study which found the same thing:

Both studies were published in a top journal in the area (13% acceptance rate) and have respectable university professors in the area among their authors. The reason other studies did not found the same thing is because they weren't looking at that range. The common assumption was that anything below 800ppm is equally perfect. You will find some blogs in the area giving this same explanation. The effect size really is suspiciously huge, but it is not entirely implausible either given that we already know CO2 does have a brutal effect on cognition at higher ppm and understand the mechanism fairly well.

Anders made the following calculations. A human consumes about 45 liters of oxygen per hour, producing about 10 liters of carbon dioxide per hour. To maintain a good CO2 level you want 8-10 l/s/person in an office. Now, if the airspeed is 0.15 m/s (about the limit of drafty) you need 0.01/ 0.15 = 6.7 cm^2 window area.

My credence on the finding being true is high. Even if it is false, the cost would be leaving the window very slightly opened (6.7cm2 is pretty small), or opening in every other hour. I've been doing so since the last study came out and I would add the room also feels more pleasant like that.

Anders made the following calculations...

The quoted calculation seems to assume that the indoor air is 100% CO2, when in fact it is about 0.1% CO2? So your conclusions seem to be off by a factor of 1000. Actually a factor of 5000 if you are trying to maintain 600ppm, since the outdoor air also has 400ppm and presumably the net flux is 0.

ETA: Actually maybe that was how you moved from l/hour to l/second, your figures seem about right for keeping levels at 700ppm assuming your airspeed.

Also a cracked window just doesn't seem to do it empirically.

My credence on the finding being true is high.

I had also seen the replication, and I believe that the paper correctly reports the result of an experiment. (And its certainly not publication bias with p < 0.0001 or whatever.) The question is whether a particular interpretation of the results is correct. At a minimum it depends on just what the test is measured.

Why not test this? Probably only suggesting this because I'm reluctant to trust one or two papers on this alone. Would be cheap to do. eg: -Write two similar tests of the key dimensions of performance you care about. -Recruit a number of participants -Put each test in an envelope marked 1 or 2 for first or second, then put two envelopes in a bigger envelope, making sure that the smaller envelopes marked 1 and 2 don't contain the same test. -Assign people to two rooms. In one, a friend has raised CO2 to 1200ppm, in another, its 600ppm. You don't know which and you don't tell them which. They do test one first in one of the rooms, and test two second in the other. -Look at results

Here is the kind of claim Id' want to understand better:

High levels of predictive validity for the SMS (r > 0.60 with real-world success as judged by peers and as demonstrated by income, job level, promotions, and level in organizations), as well as high levels of test–retest reliability across the four simulation scenarios (r = 0.72–0.94) have repeatedly been demonstrated (Breuer and Streufert 1995; Streufert et al. 1988). Additional validity is demonstrated by the deterioration of various performance indicators with 0.05% blood alcohol intoxication and seriously diminished functioning with intoxication at the 0.10 level (Satish and Streufert 2002). Baseline scores at 600 ppm CO2 for the participants in this study, mostly current science and engineering students from a top U.S. university, were all average or above.

I generally appreciate your posts and comments, but I wonder whether the EA forum is the right forum for this particular one. It's quite specific. Compare to this post about a similar specific issue and the top comment:

Edit: What I originally thought this post would be about is dangers to cognition from climate change, which I think would be a slightly better fit.

The paper also isn't accessible.

Thanks for pointing that out.

Compare to this post about a similar specific issue

One difference is that the current post is about a concrete project that has a utilitarian justification and is being philanthropically funded. In some abstract sense it seems like this should be as on-topic as other object-level discussion.

Edit: What I originally thought this post would be about is dangers to cognition from climate change, which I think would be a slightly better fit.

I don't disagree that people might be more interested in that, but it's psychologically/sociologically interesting that it would be a better fit. It's not clear what the relevant difference is.

The paper also isn't accessible.

Thanks, linked to non-paywalled version.

Well, the upvotes are speaking for themselves! :-)

My point was that it's not clear to me how it relates to doing 'the most good'. If you had mentioned you want to have this research to influence public policy or to highlight it as an underappreciated risk of global warming, the connection would be less removed than very specific productivity advice (compared to very specific finance advice).

Fwiw, I really enjoy the more specific posts on the forum. I find them more valuable that the broader comments on the movement posts, and I think the usefulness of the forum would increase if more posts were like this one.