Pupil Size and Intelligence

By Neuroskeptic | November 20, 2016 12:54 pm

Are the eyes the windows to intelligence? In an interesting paper, Georgia psychologists Jason S. Tsukahara and colleagues report that there’s a positive correlation between pupil size and cognitive ability.

It’s well known that our pupil size varies over time due to changes in both emotional state and cognitive ‘effort’. As Tsukahara et al. put it

Starting in the 1960s it became apparent to psychologists that the size of the pupil is related to more than just the amount of light entering the eyes. Pupil size also reflects internal mental processes. For instance, in a simple memory span task, pupil size precisely tracks changes in memory load, dilating with each new item held in memory and constricting as each item is subsequently recalled.

However, Tsukahara et al. are saying that there’s relationship between pupil size and cognitive ability across individuals. Here’s the key figure showing the correlation between baseline (resting) pupil diameter and fluid intelligence (Gf) in n=331 volunteers. Fluid intelligence is thought to be a major component of IQ. As you can see, the association is present in each of the three ethnic groups studied:


In additional experiments, Tsukahara et al. show that baseline pupil size is a reliable measure, staying more or less constant within the same individual over repeated test sessions (r>0.7); and that the correlation with intelligence can’t be explained by some confounding variables such as age, gender, or nicotine and caffeine consumption.

Now, when I first saw this paper, I was sure there had to be something wrong with Tsukahara et al.’s methods. Indeed, you might say I rolled my eyes, because it seemed ‘too good to be true’ that a measure as simple as pupil size would correlate with something as nebulous as IQ. But after reading the paper, I actually haven’t been able to spot any flaws… although perhaps my readers will.

Assuming the association is real, what is the cause of it? Tsukahara et al. speculate that the neurotransmitter norepinephrine might be the link between the pupils and intelligence:

Although we did not obtain any direct neural measures of brain function, neuroscience research has shown a close association of pupil size with activity in the locus coeruleus-norepinephrine system. [In the brain, norepinephrine] modulates the gain of target neurons to be more sensitive to incoming signals (both excitatory and inhibitory)… this modulation of neural gain has an effect on the strength of functional connectivity throughout the brain.

In other words, maybe people with higher trait norepinephrine signalling have both larger pupils and more efficient neural information processing as a result. I must admit that this seems rather too simplistic to me, but again, I can’t really point to anything wrong with this logic specifically.

ResearchBlogging.orgTsukahara JS, Harrison TL, & Engle RW (2016). The relationship between baseline pupil size and intelligence. Cognitive Psychology, 91, 109-123 PMID: 27821254

CATEGORIZED UNDER: papers, select, Top Posts
  • https://plus.google.com/u/0/101046916407340625977/posts Rolf Degen

    If this should hold water, someone might create a nice little iPhone app for an instant intelligence test, something long sought for. Someone already suggested a catchy name, the “Eye Q”.

    • Rob Neff

      This is a bell curve, averaged over hundreds of people. You can see many examples in the chart of high intelligence but low pupil size. So it’s a neat idea, but hardly accurate.

  • http://sites.google.com/site/todorovicana/ Ana Todorovic

    Couple of random thoughts:

    1. Intelligence correlates with height. Maybe pupil size does too?
    2. Pupil size diminishes with age.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      They controlled for age. As regards height, I agree they should have looked at this, but I don’t think it could explain the whole of the pupil size-IQ correlation, because the height-IQ correlation is approximately only r=0.2… although then again it might be higher than this in the population studied here.

  • smut clyde

    This is why I wear dark glasses all the time, even indoors. My pupils dilate and make me smarter.

  • MP

    isn’t this something we’re able to interpret instinctively? you can often tell if a person is clever by looking at his or her eyes.

  • Pingback: Sunday Assorted Links | Marginal Counterrevolution()

  • Costa Vakalopoulos

    The plant belladonna was used in Rennaissance Italy to dilate pupils and appe ar more beautiful but it caused blurred vision and prolonged use even blindness. Not so smart!

  • Uri Cohen

    What about coffee or other socialy accepted stimulant? I can easily imagine how it correlates positively with both pupil size and intelligence.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      They included “caffeine use in the last 8 hours” as one of the confound variables, but this was just a binary variable, yes caffeine or no caffeine. It might have been better to quantify the dosage of caffeine. Habitual caffeine use might also be important over and above actual use that day.

  • http://EVIDENCEr.ORG Evie S.

    Aren’t eyes technically part of the brain, anyway? Did they look at iris diameter, and does that correlate with anything?

    If we just accept the finding, is it because the more light on the retina, the better one can think?

    Or is it a biological correlate of something else having to do with smarts?

    Or, (this is it!) because large pupils are considered attractive, those who have them attract the best mates, who happen to be not just well off and good looking, but smart?

    • Opio Faegansen

      The biological correlate seems most likely. After all, it’s right there in the article- neurotransmitters like norepinephrine. We’ve got illicit drugs that people take for their ‘mind-expanding’ effects which also cause pupil dilation. We’ve got psychological disorders that are often correlated with high levels of intelligence for which pupil dilation is a symptom. When the brain is working overtime, the pupils grow to match. Intelligent folks almost certainly work their brains harder, so it’s reasonable to expect their brains are drowning in neurotransmitters related to brain interconnectivity, physical reward response, mood, etc…

      I know it sounds boring compared to sexual selection or the magical power of light entering the pupil, but I’m gonna stick with the notion of a simple bio-mechanical response to increased brain activity.

      • pauljason p VanHorn

        Very well said…as for some of the other comments i grow so weary of idiotic remarks about race etc… Science is neither racist nor political…chalk it up as interesting and move on…

        • pauljason p VanHorn

          What i would love to see is this study performed on the blind…

          • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

            Very good idea!

    • Paul O

      To see the eyes as being part of the brain is brilliant.

  • OWilson

    I just checked.

    My pupils are extraordinarily large.

    (so are my hands!) :)

    • Jiveazz Turkey

      Same here!

      Unfortunately, the same can’t be said about the brain between my legs *sigh*

  • davidpearce

    Alternatively, IQ tests measure just the autistic component of general intelligence (ASD folk have larger resting pupil size)

  • Rock guy

    The statistical basis for this study is weak at best, i.e., r is less than 0.4 for all three populations. I am amazed that it was published.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      That’s not that small of an r, there are many smaller r’s published in psychology.

      • Rock guy

        I did not know this, having not followed the psychology literature. Maybe this is one reason that psychology is considered a ‘pseudoscience’ by some researchers in other fields. Long past time for psychology studies to adopt the statistical rigor demanded in biology, geology, chemistry, etc.

        • Opio Faegansen

          Or, maybe it’s sometimes considered ‘pseudoscience’ because reactionaries like you blurt whatever random thought is running through your head on public forums before even bothering to type 5 words into google that might help you understand why they interpret statistics differently. Long past time for armchair scientists like you to educate themselves before lecturing everyone else about statistical rigor. Here, you can start with this: http://rpsychologist.com/d3/cohend/

          • jayb1

            So. How do youuuuu feeeeel about that. Hmmmm…. Uhhhhh…. Hmmmmmm……

        • Franck Ramus

          For one thing, statistical rigor has nothing to do with effect size. It is entirely expected that sciences of complex systems involving countless factors have to deal with small effect sizes for each factor, more so than sciences focused on elementary particles in vacuum.
          And biology is in the same ballpark as psychology, for that matter.

  • http://www.mazepath.com/uncleal/qz4.htm Uncle Al

    I demand retraction of the racist component. It smacks of The Bell Curve whose politically defective statistics summed 80 years of intelligence testing across the nation in education, industry, and the military.

    Who are we to say that a person who can barely score 800 on the traditional SAT will not be beautiful as a Federal administrator allocating tens of $millions of taxpayer money to her friends each year?

    • Martin

      Larger pupil sizes are found at higher latitudes where incident light levels are lower because the Sun is at a steeper angle so more light gets absorbed in the atmosphere and less reflections occur due to the steepness.

      So larger pupils —> to see better in areas with less light.

      There is also a correlation between latitude and intelligence. Colder environments provided a selective pressure towards higher intelligence because of increased hardship in survival.

      If science is racist then its still science. Just because it doesn’t fit with your political prejudice doesn’t mean its not the truth.


      • http://www.mazepath.com/uncleal/qz4.htm Uncle Al

        Good man. Believe what you falsifiably see, not what you are told should be. For instance, presidential elections have fulcrums not cushions.

        • Martin

          Provide some science that back up your claims, otherwise let the science speak.

      • Tony Lyda

        Just speaking of science, there is only one race. The human one. Genetically, there is too much overlap to define race. Race is just a social construct.

        • Martin

          If race is just a social construct then I self identify as black. O wait, I can’t cause race is guided by genetics you fool.

          • Lexy Rednax

            There is only one race here: The race to the bottom. Keep in mind there are other species and we are only beginning to understand their intelligence let alone try to ascribe meta-characteristics or quantify their intelligence. We do not fully understand our own, either. What I see from the abstract is an association between one activity and one poorly characterized meta-characteristic. That’s alchemy, not science.

          • Mankoi

            No, it’s a construct because it’s an arbitrary way of categorizing people. People’s skin tones vary based on genetics; this is true. However, people’s nose size also varies based on genetics. We divide people into race based on color, and not nose size, but that’s completely arbitrary. Neither measure speaks to any underlying difference, but we use one to categorize people, and generally not the other.

            In summary, skin tone is guided by genetics. The concept of race is a social construct.

        • Tol

          Maybe it ironical that in Post-Socialistic contry Racelogy is a valid fild of study.Leading biology researcher in Russia laught of western genetic that denied the existence of race. They think that evil capitalist destroyed race concept to masked their own racism

      • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

        However, the correlation of intelligence with cognitive ability survived controlling for ethnicity…

      • Small_Businessman

        Uncle Al has pin-head sized pupils.

      • facefault

        >Colder environments provided a selective pressure towards higher intelligence because of increased hardship in survival.

        Used to be racists argued that Europe was more hospitable than Africa, and that that allowed the development of tool use. Now they argue the reverse. There is no actual evidence of either, just wishful thinking.

        • Martin

          Try numerous scientific studies you pseudoscientist

      • http://larryetkin.efoliomn.com/ Larry Etkin

        Unless I’m missing something, as I read his synopsis, this study isn’t talking about race at all. It’s addressing deviation from the baseline across all individuals or subgroups.

    • Apollo Grace

      Uncle Al, what are you talking about with the “racist component”? They broke it down over three ethnic groups, and got the same correlation across those groups. There’s zero assertion here that any ethnic group is any smarter than any other.

    • Susan Rand

      Beautiful? What has beauty to do with it?

    • Rob Neff

      What “racist component”? You mean you want retraction of the data?
      Keep in mind also, N of 331 is a pretty small sample size.

  • http://cherishthescientist.net mareserinitatis

    Are people with higher IQs also more likely to have myopic vision?

    • Rob Neff

      Good point. People who grew up doing a lot of reading tend to need glasses. That needs to be part of the data set at least.

      • Maia

        Intelligence has to do with a lot more variables than how much reading a person has done. (I am a heavy reader, so this objection is not self-serving).

  • gatorallin

    if true, can it have anything to do with reading, so that pupils that read or gather more information move side to side and thus like a muscle grow in size?

  • slocklin

    I would have suggested something like alertness, but I was interrupted by that ridiculous plot of pupil size data. Psychologists fail at linear regression, again.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      What’s ridiculous about it? I mean, it is not very aesthetically pleasing, but the data look sensible?

  • Pingback: Your intelligence might depend on your pupil size – Mother Nature Network | SelfHelpEA()

  • Patrice Boivin

    could there be a correlation with this as well?

  • Stephanie knepper

    Geeze, I have an MA in Constitutional Law, and my pupils are not very large. I guess I are a idiot.

  • Apollo Grace

    OK, this does cast an interesting light on hallucinogens…

  • Lorie Franceschi

    If we go by this so called “study” and expand it to the rest of the mammalia class, there are mammals that have a very much higher intelligence than humans because of their pupil size.

  • Elaine Dolan

    Yes, please retract the racial implication. Here’s how: begin the enormous list of ALL(beyond one’s capacity to quickly recall) that INTELLIGENCE entails.

  • cld

    Don’t women have larger pupils generally?

    Is this to say the average woman will have a higher average IQ than men?

    • Bronwyn (デイ)

      Perhaps it’s a relative-to-the-mean thing where sex and/or gender need to be taken into account as well. There could be a stronger trend visible if we consider fluctuation around the sex/gender’s mean.


  • travois

    I have beady little eyes but I can’t figure out if that’s good or bad….help!

  • Juno_Moneta

    It explains how, when I’m higher than a kite, and my pupils are as huge as saucers, I’m so damn smart.

  • Tony Reno

    I can think of a far simpler cause. Near Sightedness is a common correlation with lots of reading. Whether being nearsighted makes reading more comfortable, or reading a lot makes eyes nearsighted, or it is just coincidental is beside the point.

    The types of corrective lenses used for near sightedness reduce the apparent size of incoming objects (whenever I take my glasses of, things look larger). Perhaps the correlation is related to the response to corrective lenses for near sightedness.

    So my suggestion is that the researcher include an axis for nearsightedness.

  • heather horton

    Unlike humans, parrots control their pupil size. It is important to watch so one knows if the parrot wants to play vs. bite. Smart birds.

    • Rob Neff

      Humans do that too, it’s just unconscious so you’re not aware of it. Cats also do this, it’s very pronounced because of their pupil shape.

  • jimoppenheimer

    Take away those three highly hypothetical lines, and you see a nice circular scatter pattern. And while it can be said that it seems ever so slightly elevated on the righthand side, one must think, So What? The variation within the groups is far more than the change. This finding of who-cares results reminds me of the parapsychology results reported for so long, again and again. People predicted shapes correctly at a rate better than chance. This finding did not gain traction outside the, uh, interested community simply because it was a So-What finding. You might say, yes something’s happening, but it’s a minor thing, and we have no clue why it is happening.
    Is this worth following up? Well, in science I suppose everything is worth following up, even if it involves such obviously unrelated measurements, but I wouldn’t expect anyone to invest a lot of time, effort or money in it.

    • Kirby Mahard

      I thought so too…Until I read about research on Extrasensory Electroencephalographic Induction between Identical Twins.

      ‘Alpha rhythms have been elicited in one of a pair of identical twins as a result of evoking these rhythms in a conventional manner solely in the other.’

      Then there’s the strong telepathy that dogs have of their owners returning home.

      Even some skeptics (Richard Wiseman) have now begun to accept ESP.

      “I agree that by the standards of any other area of science that remote viewing is proven…”

  • Ajit

    Women used to put belladonna in their eyes to dilate the pupils and increase their attraction quotient. Apparently they knew this correlation a long time ago.

  • JoeM

    I’m not impressed! If you take away the lines drawn in, all I see is a bunch of dots pretty much scattered at random. Oh…but you’re claiming a statistical analysis was done to correlate the lines. Statistical analysis was done on a heck of a lot more than 330 people at election time, and they all told us Hiliary was a sure thing. Nice try…. but I don’t buy it.

  • A.Alexander

    The reality around as is so complicated, that looks. chaotic. We can impose the differen nets on it,thus catching some order(with the small correlations ). This different correlations may look like the surrogates of the connections or dependances, though false ones.

  • bwana

    Sounds a bit like this study is right up there with phrenology!?

  • Mike

    What several responders in this thread conclude, without being sure why, is that there isn’t much to Tsukahara’s study. They are correct. Tsukahara’s study “proves” that there is virtually no relationship between fluid intelligence and pupil size. Here’s why. The first clue is the r values (correlations) are low. But more important than the correlations is r-squared (variance) *100. This gives you an intuitively understandable way of understanding the study results in terms of prediction. r-squared tells you how many times you would make a correct prediction (y) by knowing x or vice versa. Thus, for Causcasians, r-squared is .1122. Multiply that by 100=11.22%. You would be correct 11 times out of 100 in predicting fluid intelligence (y) by knowing pupil size (x) and vice-versa. For the Other category, you would be correct only 14 times out of 100 times. You would do better, statistically, by just flipping a coin to make your prediction; you would make 50 correct predictions out of every 100. A coin flip prediction is roughly equivalent to a correlation of .7 with an r-squared of .49. So, as a general interpretive rule of thumb, any correlation below , say, .725 means you will make more incorrect predictions between x and y than correct ones. As correlations increase beyond .725, correct predictions become more frequent than incorrect ones.

    • Haven Monahan

      I don’t think your interpretations of r and R2 make any sense. What are you even trying to predict? Both IQ and pupil size are continuous variables, while a phrase like “correct 11 times out of 100” suggests a binary variable. The correlation between pupil size and IQ is not large, but it appears to be reliable/repeatable.

      If you have no information on the correlation between IQ and pupil size, your best bet, if you know a given individual’s value on one of the variables, is to predict that their value on the other variable is equal to the population mean value. If, on the other hand, you have information on an individual’s IQ or pupil size, you will be able to the predict to what direction and how much their value on the other variable deviates from the population mean. Given the low correlation, your prediction probably won’t be precise, but it’ll be better, on average, than simply assuming that the individual’s value is equal to the population mean.

      • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

        That’s right

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      “You would do better, statistically, by just flipping a coin to make your prediction; you would make 50 correct predictions out of every 100. A coin flip prediction is roughly equivalent to a correlation of .7 with an r-squared of .49”

      By definition, a correlation between two variables means that you can predict one of them at better than chance levels if you know the other one.

      • Mike

        No, I don’t think you are correct because you aren’t considering the size/strength of correlations. A correlation can be any number greater than 0 and no more than 1.0. If .001 is the correlation (r) between two variables and .4 is a correlation (r) between two other variables, you can make more correct predictions of one variable from the other at a .4 level than than you can when the correlation level is .001. But in neither case will you be able to make predictions greater than chance. They will be much less than chance. You can only make greater than chance predictions when the correlation is .71 or greater. We generally say that correlations in the .1, .2, .3 range are weak correlations, .4, .5,.6 are middling correlations, and .7 and beyond are strong correlations. When you calculate r-squared (variance or explained variance), you get a number that tells you how often your predictions will be correct when you estimate/predict/”guess” the value of one variable when you know the value of another.

    • D Samuel Schwarzkopf

      As others have pointed out already, this is incorrect. R-squared doesn’t give you a prediction accuracy. All this means is that 11% of the variability in IQ can be modeled as a function of pupil size. The other 89% are due to other, unknown factors. Unless the authors used cross-validation (which I doubt but I may be wrong as I admit I didn’t read this paper) the 11% is an overestimate as it only quantifies the goodness of the model fit for these specific data.

  • Anthony Kreinbrink

    I’d love to see the correlation coefficient for this data. I estimate it to be about 0.15, just eyeballing it. Surely looks like a plot of random noise to me.

    • Anthony Kreinbrink

      Oh, duh, r is on the right. No self-respecting scientist would stake results on such low r values. Scatter plot.

      • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

        I’m not sure the authors are staking their results on that plot. The plot is their results, and that’s what the data show. Would you suggest that they not publish these data at all because the r is too low?

        • Jon

          And the r isn’t low. The average effect size in psychology for main effects is r=.20. Human behaviors and traits are complex, multiply determined things. Nothing is caused by one (or even a small number) of factors. A correlation of r=.337 is huge.

          • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

            I wouldn’t say “huge”, but it is large in psychology terms. Even in absolute terms it’s by no means trivial.

  • Pingback: Your intelligence may depend on your pupil size()

  • Cee Betterchoice

    Looks like a pretty mediocre study. I don’t see a real trend at all. The study uses r, the correlation coefficient and the highest value is 0.337? That’s less than 50-50 a coin flip. Usually, you check the r-squared as well. That had a value of 0.113. This study is so bad, it isn’t even worth arguing over.

    • Jon

      Umm, the average effect size in psychology is a correlation of r=.20. Human behavior (and things like intelligence) are complex and multiply determined–nothing is caused by a single thing. A correlation of r=.337 between two different things is huge.

    • Grant

      In addition to what Jon said about 0.337 being quite significant as far as rule-of-thumb assessments of correlation goes, you appear to be confusing r-values with the likelihood of given outcomes. In particular, I think you’re incorrectly interpreting the r-value is the probability that one variable will have an impact on the other (at all) and, furthermore, erroneously comparing this value to the probability that a certain face of a coin will come up when flipped, as though a variable’s influence on another can be thought of as the outcome of a weighted coin. This is not correct. The r-value, known as the correlation coefficient, is essentially a normalized measure of covariance – it’s the degree to which changes in one variable are commensurate with changes in another, not a probability. The r-values, or even r^2, are not appropriate to interpret as probabilities because there is no enforced requirement that they sum to 1, as the axioms of probability require.

      When speaking of correlations, causation, and probabilities in the future, please make…


      ( •_•)>⌐■-■


      Better choices.

  • Jonathan Democritus Lettvin

    A great teaching neurologist taught me that pupils narrow for detail work (soldering discrete components) and widen for fight-or-flight (try pinching the back of someone’s arm while watching their pupils).

    When confronted by a xenophobe, my pupils might widen. Then again, they might narrow as I consider what can be done to fight the hate her attitude engenders.

    I wonder whether I would score low or high IQ by pupil size? I wonder whether I scored differently the day my cat died… or the day I worked out the mathematics of visual hyperacuity which depends on pupil size. Maybe I am particularly dumb when my pupils constrict as I write this. I know I am at my smartest as my pupils dilate in delight when my wife enters the room.

    Lies. Damn lies. And statistics.



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar