While perusing the news last night, I was horrified to come across a set of articles telling me that less sleep is better for academic performance.  If that’s the case, I should be in tip-top mental shape right now, because the reporting on that study was bad enough to keep me tossing and turning in scientific consternation.  So, in solidarity with sleep-deprived teens everywhere, here’s why those headlines are wrong.

First, what does the study in question actually do and find?  The researchers used a large national survey that included students’ standardized test scores, self-reported sleep hours, and demographic data, and they looked at the correlations between test scores and hours of sleep.  Contrary to current recommendations for 8-9 hours of sleep, they found that teens with ~7 hours of sleep had the best test scores.  Getting either more and less sleep was associated with lower scores.

Ok, so what’s wrong with that?  Simply this: correlation ≠ causation.  Let’s consider how the study was done.  This was an observational survey, not a controlled experiment, so there could be all kinds of unknown factors that affect both hours of sleep and test scores.  Some of them, if you’ve ever been a teen in high school, are obvious.

For example, if students are getting a whole lot of sleep, they’re probably not staying up late to study or finish homework.  So maybe what we’re actually seeing here is that kids who study less get lower test scores, and they also have more time to sleep.  It’s not the increased sleep that hurts grades, but it looks like that since you’re not considering study time.  Similarly, exercise improves brainpower, but sports practices can cut away from available sleep time.  Another example: maybe strong social networks help academic performance, but also keep kids out late.  So there are three unobserved variables that could increase test scores but decrease sleep at the same time.  If we increased sleep hours while holding study hours, exercise, and social support constant, it’s entirely possible that we’d find 9 or 10 hours of sleep to be optimal, if you can find time for it.

The paper itself acknowledges that its non-experimental approach has limitations – there’s an entire section so titled that the reporters must have missed.  The authors do control for what variables they could, including age, gender, race/ethnicity, geographic region, family income, and parents’ education level.  However, they reject the idea that another unmeasured variable (like the three we suggested above) could be having opposite effects on sleep and test scores.  Their argument is that this would cause test scores to keep getting better as sleep decreases, instead of there being some “optimum” amount of sleep like they find in their data.  Again, anyone who’s in or remembers high school could easily refute that argument.  More studying gives you better grades up to a point, but after that it starts to cut out of your sleep time so much that it does more harm than good.

So high school teens: before you let your parents say the internets told them less sleep would be good for you, make sure they understand the difference between correlation and causation, and the virtues of controlled randomized trials.  Just don’t spend so long studying for your argument that it cuts out of your sleep time.