cognitive science
and more
Is pupil mimicry real? And if so, what is it?

If you see someone yawn, you won't be able to suppress a yawn of your own. If you see someone take a sip from her coffee, you are likely to take a sip as well. If you see someone shift her gaze, your gaze is likely to follow. These are all examples of automatic imitation, or mimicry.

An adorable Rhesus monkey imitating tongue protrusion. (Source, License: CC-by 2.5)

Mimicry may serve a social function. If you see someone take a sip from her coffee, some of the same brain regions ("mirror neurons") become active that would also become active when you take a sip yourself. This neural overlap is, in a sense, a direct form of neural empathy: When I see you do something, it's like I'm doing it myself. And this overlap increases the likelihood of automatic imitation, because seeing someone else perform an action primes your brain to perform the same action yourself.

All of this may sound fanciful, and not all researchers buy into the mimicry-as-empathy notion to the same extent. But by and large this is accepted psychological theory. And personally I think that there's a lot to it.

Pupil mimicry is a specific type of automatic imitation, in which you automatically adjust the size of your eye's pupils based on the pupils of someone you're looking at. So if I would look at you, and if your pupils would dilate, then my own pupils would dilate as well. In the past years, several high-profile studies on pupil mimicry have been published, including a study by Prochazkova and colleagues that just appeared in PNAS, and which prompted me to write this post.

But is pupil mimicry real? And if so, what does it reflect? Let's apply some much-needed scrutiny.

Pupils of humans, cats, lizards, eagles, and octopuses. (Source)

The basic methodological problem when studying pupil mimicry is that pupils are dark, and that darkness causes pupils to dilate; that's the pupil light response. Therefore, if you look at someone whose pupils dilate, visual input becomes overall darker, and your own pupils dilate as a result. Therefore, the pupil light response could by itself be sufficient to explain pupil mimicry. If that's true, pupil mimicry would not be mimicry at all, at least not in the social sense in which the term is usually understood.

Surprisingly, the only pupil-mimicry study that I know of that controlled for brightness is a study by Derksen and colleagues, which just appeared in Journal of Cognition, and on which I'm a co-author. In a nutshell, we first replicated pupil mimicry using the same videos of constricting or dilating pupils that had been used in previous studies, including (it seems) the study by Prochazkova and colleagues. Although our results were mixed, overall we confirmed that something akin to pupil mimicry exists when brightness is not controlled for. However, when we equated the brightness of the videos by slightly increasing the brightness of large pupils, we no longer found any pupil mimicry.

To equate the brightness of the eye region, large pupils were made slightly brighter, resulting in slightly washed-out-looking pupils in the center panel. (Source)

What does this mean? It means that brightness of the eye region is a confound that may explain the results of previous studies on pupil mimicry. It does not prove that pupil mimicry is not a real thing. For example, proponents of pupil mimicry might reasonably counter that by increasing the brightness of the pupils in our videos, we created strange-looking faces (see above) that do not elicit mimicry. So we just don't know. Pupil mimicry may be a real thing, in the sense that it may be a genuine form of social imitation, or it may be a brightness artifact.

Now let's zoom in on the study by Prochazkova and colleagues. The authors make several claims, which I'll consider in turn. I'm not questioning the reliability of their data, nor am I suggesting that there's anything fishy with the results per se. But I'm skeptical that their results support all of their claims. And I fear that my criticisms apply to the pupil-mimicry literature as a whole.

Prochazkova and colleagues conducted a variation of a so-called trust game, in which participants decided on each trial how much money to give to a virtual partner, which was a video of an eye region with constricting, static, or dilating pupils. (In a regular trust game, the partner then receives triple the given amount, and returns some proportion of this to the giver. This creates an interesting dynamic between partners, in which trust and sense of fairness determine how much money is given and returned. However, it seems that Prochazkova and colleagues did not include any interaction.) Crucially, the authors found that participants gave more money to virtual partners with dilating pupils than to partners with constricting pupils, suggesting that participants trusted faces with dilating pupils more than faces with constricting pupils. This strikes me as a reasonable claim.

Prochazkova and colleagues also measured pupil mimicry, and replicated the typical finding that participants' pupils dilated when the pupils of their virtual partners dilated, relative to when they constricted. However, the claim that this reflects social mimicry is problematic because of the brightness confound that I explained above.

Next, Prochazkova and colleagues split the data based on whether or not participants indeed mimicked their partner's pupil size. Because mimicry occurs on average, but not on every trial. The authors then looked at how mimicry modulated the amount of money that participants gave to their virtual partners. Here's where things get interesting: When participants mimicked the pupil dilation of partners whose pupils dilated, they invested more money; in contrast, when participants mimicked the pupil constriction of partners whose pupils constricted, they invested less money. Based on this, the authors concluded that mimicry makes trustworthy partners even more trustworthy, and untrustworthy partners even less trustworthy.

But … let's look at this data differently, as a main effect of pupil size: Small pupils are associated with less money being given to virtual partners. That's all. And there are many reasons why this might be the case. For example, over time participants might grow bored with the task, which would certainly cause their pupils to constrict and possibly cause them to give less money to their virtual partners. This is just one possible explanation. But the important point is that the specific claim of the authors is not well supported by the data.

Finally, the authors compared brain activity, measured with fMRI, between trials in which participants mimicked or did not mimic their partner's pupil size. And they found that brain regions forming the so-called Theory of Mind (ToM) network were more active during mimicry than no-mimicry trials. The ToM network is associated with social behavior, so this finding fits well with the idea that pupil mimicry is a social phenomenon.

But … let's again look at the data differently. Say that on some trials participants are not engaged in the task. On those trials, participants wouldn't pay attention to their virtual partners, and their pupil size would therefore not be affected by the brightness of their partner's pupils (because the pupil light response is affected by attention); and possibly, their ToM network would also be less active when they don't engage in the trust game. In other words, the fact that there is a correlation between pupil mimicry and activity of the ToM network does not mean that pupil mimicry is a social phenomenon; it might reflect overall task engagement, or something else. (It's correlational data, so with some creativity you can fit any number of explanations onto it.)

In summary, the notion that pupil mimicry reflects social imitation is weakly supported, despite many studies on the topic. It might well be that pupil mimicry results from the fact that eye regions with large pupils are darker than eye regions with small pupils. Our failure to observe pupil mimicry when controlling for brightness suggests that this is the case, although our study should be a starting point, rather the final word on the matter. I encourage researchers on pupil mimicry to take a step back and firmly establish that the phenomenon is real, in the sense that it reflects social imitation. And only then use it as a tool to test more subtle social phenomena, such as trust, coorporation, and ingroup-outgroup differences. Because right now, they may be building a house of cards.


  • Prochazkova, E., Prochazkova, L., Giffin, M. R., Scholte, H. S., Dreu, C. K. W. D., & Kret, M. E. (2018). Pupil mimicry promotes trust through the theory-of-mind network. Proceedings of the National Academy of Sciences.
  • Derksen, M., Alphen, J. van, Schaap, S., Mathôt, S., & Naber, M. (2018). Pupil mimicry is the result of brightness perception of the iris and pupil. Journal of Cognition.