You're probably wrong about daylight.

Issue 005: In the Dark about Daylighting

It’s been a long summer, ya’ll. Lots of traveling, lots of experiences with place attachment (the topic of last month's issue). Visiting new places with my preschooler had me wondering about place attachment in kids. Lo and behold, there was a brand-new article published in Journal of Environmental Psychology on exactly this topic. It’s so new that the full-text isn’t even available yet, but I’m keeping my eye on it. Kismet.

This week, we’re shining light on a landmark daylighting study. I’ve seen it referenced so many times. I’ve been in multiple project interviews where well-meaning designers present it. It’s been cited in 200+ other scientific articles. And it’s bullshit. Let’s *light it up* (shout-out to my SJM readers).

And finally, an offer just for Read the Room subscribers: If you’re hooked on evidence-based design and you want to integrate science into your own design practice, I just opened enrollment for the first two cohorts of my new course - Evidence-Based Design: Practical Foundations. 12 weeks, six modules, two opportunities for 1:1 coaching with me, and a final portfolio-ready EBD project. The first lesson is available for free here. Get $100 off the registration fee with promo code READTHEROOM.

Warmest welcome,
Dr. Kati Peditto

In the late ‘90s, the Pacific Gas & Electric and the California Energy Commission combined forces to contract a study. The Heschong Mahone Group was a well-known building science firm and they agreed to investigate daylight on behalf of the utility.

This was a big study. The HMG team pulled test scores and demographic data on more than 21,000 students across three districts in California, Washington, and Colorado. They went into schools in each district and assigned classrooms a daylight rating from 0 to 5, based on window size, skylights, and glazing.

After controlling for around 50 variables (sounds impressive!), HMG found students in the most daylit classrooms progressed 20% faster in math and 26% faster in reading over a year than students in the least daylit ones.

I’ve seen this result - 20% improvement in math, 26% improvement in reading - in so many places, including two separate LEED and WELL project verification narratives.

Built environment research often has a “random assignment” problem. It’s very hard to randomly sort people into different spaces, especially schools and houses that may be lower-quality. We talked about this in the first issue.

In this case, nobody randomly assigned kids to daylit and non-daylit classrooms. Peer reviewers found a really obvious issue with that: what if more experienced (or simply better) teachers ended up in the nicer rooms with more windows?

And also… they only visited a sample of the classrooms to take measurements. Otherwise, they estimated daylight scores based on architectural drawings, photos, and maintenance records. According to the study itself: “They were not verified with on-site measurements since we could not visit enough classrooms under similar daylighting conditions.”

To their credit, the HMG team went back and checked for bias. The original article came out in 1999, and they did a re-analysis in 2001. This time, they surveyed teachers directly and they turned electric lighting on/off. They didn't find evidence for any bias, at least not on the variables they measured. This is actually a cool example of peer review working as intended!

But…. it's also still an observational study, and they were searching for issues after the results came in rather than ruling them out beforehand. Not great research technique.

Perhaps I sound like a conspiracy theorist, but there is a big red flag that researchers are trained to identify. When evaluating the strength of a study, readers should always investigate the funding source. In this case, a utility paid for research meant to build the case for daylighting incentives and energy code changes. That doesn't necessarily make the findings “wrong” or false or totally inadmissible. But it's very important context.

In addition to measuring schools again, HMG ran a better-controlled (IMO) study in 2003 with office workers. Workers with a view of the outdoors, compared to none, scored 10% better on a cognitive acuity test and 16% better on a memory test. Are you a better employee if you have 16% better memory? Questionable.

Especially when they measured and isolated the effect of daylight specifically. A 10% increase in measured daylight predicted only a 0.4% improvement in cognitive accuracy, and no measurable effect on memory at all. Not 20%. Not 26%. Basically nothing. This is far less impressive than the spurious relationship between the divorce rate in Maine and per capita consumption of margarine.

A research review from Lawrence Berkeley National Lab's indoor air quality program looked at the daylighting and performance literature as a whole. Systematic reviews are at the top of the evidence hierarchy, so they give us more information than one single study alone. They concluded the following: too few studies, with results too inconsistent, to draw a firm conclusion either way.

Anecdotally, I think we all know that windows have substantial benefit. Perhaps it’s the view of nature, perhaps it’s the opportunity for positive distraction, perhaps it’s the light itself. Regardless, the point of this newsletter is not to argue about windows, just the evidence about windows. And there’s far better evidence out there now, so stop citing the HMG study, please, I beg.

Now for the better evidence.

AKA: I downloaded these papers and vetted them, so you don’t have to.

Daylighting is indeed A GOOD THING for so many reasons. Circadian rhythm, views of nature, improved energy use (mostly). We should all be helping our clients invest in better daylight!

But none of that depends on a poorly-designed study from 25 years ago.

If you recommend a particular glazing strategy or window-to-floor ratio, and you back it up with the HMG study, you’re going to be on unsteady ground. Instead of citing a correlational study from 1999 funded by a utility, try one of the studies above.

Good design doesn't need inflated data to be justified. Good data is out there for good design!

If you want to learn more about identifying strong evidence, using science to inform design, and conducting design research, check out a preview from my new course below. Now enrolling for fall and winter cohorts - learn more here and use code READTHEROOM for $100 off.

If you found this valuable, forward it to one person in your firm who should be reading it. To work with me on an upcoming project or speaking engagement, reply to this email or visit katipeditto.com.

Until next week,

Dr. Kati Peditto