Home > Uncategorized > Why Do You See That Face that’s NOT There?

Why Do You See That Face that’s NOT There?

Have you ever thought about seeing things that are not there? At first glance, this might sound like a bizarre suggestion, unless, of course, you are a philosophically-minded person (if that’s the case, read Descartes!); it is, undoubtedly, a logical possibility, but it is a possibility that seems only able to realize itself in a movie, or in the cases of unfortunate people who suffer from mental disorders. But interestingly, this kind of phenomenon does exist in our life, and it is actually very prevalent, at least for seeing one particular object: faces.


I guess you are now suspicious, but recall the last time that you or someone else saw a face in the cloud; or this image showing a cute, smiling “face” on the back of a chair. But obviously, there is no face. Still, we recognize them, with considerable ease. This tendency for us to see faces where there aren’t any is called face pareidolia. And this tendency can be very useful, and sometimes even profitable. Artists, such as Giuseppe Arcimboldo, have long exploited this tendency to create some of the most imaginative paintings (to read more about Giuseppe Arcimboldo’s work, click here); moreover, the ability to see Virgin Mary in toast is obviously worth 28,000 dollars on eBay (to see the news, click here). Given its prevalence and potential value, it’s natural to wonder how exactly do we recognize those non-existent faces? Saddly, I don’t know the answer for sure, but perhaps I could offer some possible explanation.

To understand face-pareidolia, I believe it’s actually important to first look at the way we recognize actual faces, for these two phenomena might share similar underlying cognitive process. It turns out that face recognition, though not strictly unique in the realm of pattern recognition which is the process by which we recognize the input sensory information, is still quite special; it features what cognitive psychologists refer to as holistic processing. In other words, not only does basic features, such as the color of your eyes, and the shape of your month, matter, but the holistic spatial relationship between these basic components also play an arguably more important role in face recognition. This holistic processing is supported perhaps first by Robert Yin’s famous experiment (1969), in which he demonstrated that faces presented up-side down were much more difficult to identify than faces presented upright. Moreover, this effect, called inversion effect, was disproportionately manifested in face recognition compared with recognitions of other objects (e.g. house, airplanes, and even men in motion), where smaller inversion effects are found. If you are not convinced yet, then consider another study done by James W. Tanaka and Martha J. Farah (1993). In their study, participants were told to study a bunch of faces, and then were given a recognition test, in which they were presented with a part of a face either within a whole face, or isolated out, to recognize. Consistent with the holistic processing theory, they found that participants could recognize a part of a face more easily when it’s presented within a whole face than when it’s isolated out. Moreover, this effect also manifested itself more profoundly in face recognition compared with other object recognition.

Thus, even though it is impossible for us to see actual eyes, or nose in the cloud or on the back of a chair, we still can, theoretically speaking, perceive something as a face, as long as the holistic pattern of all the basic features themselves resembles that of a face.

However, the importance of holistic processing relies entirely on the assumption that normal face recognition and face pareidolia share similar cognitive processes. If this assumption is not true, then the previous paragraphs can be readily considered as severe digressions committed by some crazy cognitive psychologists who just can’t stop talking about their work. Fortunately, it is not an assumption that is purely imaginary. In a study published in 1997, Nancy Kanwisher, Josh McDermott, and Marvin Chun discovered a particular area, called Fusiform Face area, in human brain that is very active when people are looking at faces but not other objects. A more recent experiment (Josef Parvizi, et. al., 2012) also demonstrated that sending a small electric current to the Fusiform Face area could induce distortions in people’s face recognition. Hence, it is very clear that Fusiform Face area is very involved in normal face recognition process. If research could demonstrate that this same area is also very active in the process of face pareidolia, then it will be reasonable to think that normal face recognition and face pareidolia at least share some similar underlying cognitive processes. And indeed such study (Liu et. al., 2014) exists, in which right Fusiform Face area was shown to be particularly active when participants believed that they “see” faces.

Thus, I hope you are now convinced that the holistic processing actually plays an important role in face pareidolia. However, you might still not be satisfied, since this holistic processing does not seem to explain the considerable ease we experienced when we perceive those faces. Hence, it seems that holistic processing is not the entire story.

Scrambled face from Johnson et. al., 1991.

In order to account for this extraordinary ease, I think perhaps we could first think of the reasons why certain tasks might be very easy and natural to perform. One possible reason is practice. If we’ve been practicing certain task for a very long time, we would expect to perform that task with ease. Hence, if holistic processing is developed in a very early stage, then it will be possible that, through extensive, though not conscious, practices, the holistic processing becomes really natural, and eventually becomes the dominant processing in face recognition. So is it? The short answer is yes. First, it is well-known that newborn babies—babies within one hour of birth—are more willing to track normal faces than scrambled faces in which, though basic features such as eyes, and nose exist, they exist in a spatially abnormal fashion (Johnson et. al., 1991). Second, the aforementioned inversion effect, which is considered as a hallmark for holistic processing, was also observed in 1-3-day-old infants (Turati et. al., 2006). Thus, it is reasonable to conclude that holistic process is indeed developed very early in life. In fact, face pareidolia has also been recorded in 10-month-olds and 12-month-olds (Kato & Mugitani, 2015). Hence, it seems that we could argue that the ease we experienced in the process of face pareidolia is a result of a prolonged practice of  holistic processing in face recognition and face pareidolia that started almost in the first year of life.


Now I hope I have already provided a convincing argument for holistic processing’s central role in this phenomenon—face pareidolia. But that’s still not the end of the story. You now might be curious that if face pareidolia is indeed rooted in such fundamental and universal cognitive process, why are some people better at perceiving these illusory faces than others? The answer might be that, though face pareidolia shares a lot of similar underlying cognitive process with normal face recognition, it is still possible to differentiate them. More specifically, face pareidolia might requires more top-down processing, which is the process that utilizes our prior knowledge and experience to guide our cognitive behaviors. This seems to be a logical speculation, since the perceptual information sent by those alleged “faces” does not contain information indicating eyes and month; nevertheless, we still recognize them as eyes and month, suggesting that information must come from an avenue that’s different from the bottom-up and perceptual pathway. And it could only be the top-down processing. This theory is supported by a study (Riekki et. al., 2013) comparing face pareidolia between paranormal and/or religious believers and skeptics and/or non-believers. In that study, it is shown that paranormal and/or religious believers, whose experiences and beliefs differ from those of skeptics and/or non-believers, seem to have a more lenient criterion for accepting something as faces, and are more prone to claim to see faces where there aren’t any.

In conclusion, though I don’t claim to have a definite explanation for face pareidolia, I do illustrate some possible factors—holistic processing, being developed at a very early stage, and top-down processing—that influence this phenomenon. So what’s the take-home point? If you want to better capitalize your own face pareidolia, practice more and pay very close attention to your toaster.


Johnson, M. H., Dziurawiec, S., Ellis, H., & Morton, J. (1991). Newborns’ preferential tracking of face-like stimuli and its subsequent decline. Cognition, 40, 1-19. doi: http://doi.org/10.1016/0010-0277(91)90045-6

Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face area: A module in human extrastriate cortex specialized for face perception. Journal of Neuroscience, 17 (11), 4302-4311. Retrieved from http://www.jneurosci.org/

Kato, M., & Mugitani, R. (2015). Pareidolia in Infants. PLOS ONE. doi: 10.1371/journal.pone.0118539

Liu, J., Li, J., Feng, L., Li, L. Tian, J. & Lee, K. (2014). Seeing Jesus in toast: Neural and behavioral correlates of face pareidolia. Cortex, 53, 60-77. doi: http://doi.org/10.1016/j.cortex.2014.01.013

Parvizi, J., Jacques, C., Foster, B. F., Withoft, N., Rangarajan, V., Weiner, K. S., & Grill-Spector, K. (2012). Electrical stimulation of human fusiform face-selective regions distorts face perception. Journal of Neuroscience, 32 (43), 14915-14920. doi: https://doi.org/10.1523/JNEUROSCI.2609-12.2012

Riekki, T., Lindeman, M., Aleneff, M., Halme, A., & Nuortimo, A. (2013). Paranormal and religious believers are more prone to illusory face perception than skeptics and non-believers. Applied Cognitive Psychology, 27, 150-155. doi: 10.1002/acp.2874

Tanaka, J. W., & Farah, M. F. (1993). Parts and wholes in face recognition. The Quarterly Journal of Experimental Psychology, 46A (2), 225-245. doi: http://dx.doi.org/10.1080/14640749308401045

Turati, C., Cassia, V. M., Simion, F., & Leo, I. (2006). Newborns’ face recognition: Role of inner and outer facial features. Child Development, 77 (2), 297-311. doi: 10.1111/j.1467-8624.2006.00871.x

Yin, R. K. (1969). Looking at upside-down faces. Journal of Experimental Psychology, 81, 141-145. doi: http://dx.doi.org/10.1037/h0027474

Categories: Uncategorized Tags: , ,
  1. May 3rd, 2017 at 10:44 | #1

    Very interesting, well written post! Something that I thought of while reading this is also the influence of frequency and memory. For example, something like an outlet is a very common one to see faces in as we probably look at them on daily basis. This would make our pattern recognition of it as looking like a face very quick. However, if something is more complex or unfamiliar it may take longer. For instance, you are laying on the ground with friends looking at the clouds and one says she sees a person’s angry face, but it takes you 5-10s to recognize what she is talking about. I also think attention would play a role in this, for the second example you would need to expend resources to deliberately be trying to pattern recognize something, whereas the outlet example would be much more quick and automatic.
    I also stumbled across this interesting article that looks at how viewing things as faces enhances our object recognition, which is a nice positive! https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5016824/

  2. sxu20
    May 3rd, 2017 at 13:46 | #2

    @Alexandria Lucas
    Thank you for your comment! I agree that attention might also play a role in face pareidolia, which does not seem possible without people spending at least some amount of attentional resources to the target. However, I am rather suspicious of the influence of frequency. It is true that we see outlets on a daily basis, but I think this high frequency should enhance our ability to recognize outlets instead of faces. I am more inclined to explain your comparison between seeing “faces” in outlets and seeing “faces” in the clouds in terms of the configurations these two different kinds of stimuli present. I think we are able to recognize “faces” in outlets faster than in clouds because the configuration of the outlets resembles faces more than that of the clouds does. Finally, thank you for pointing out that article!

  3. April 24th, 2018 at 23:20 | #3

    Very interesting blog post ( I also like your pictures and examples of pareidolia)! Your description of the face inversion effect and holistic processing and how this ties into pareidolia is quite informative and makes a lot of sense. I would have been interested to hear more about the role of top-down processing in pattern recognition and how this relates to the phenomenon of pareidolia. I feel like this is an important factor and could be expanded upon more as an underlying explanation for pareidolia.

You must be logged in to post a comment.