In short
Pareidolia is the perception of a meaningful pattern, most often a face, in something that does not contain one, such as a cloud, the front of a car, an electrical socket, the grain of wood or a slice of burnt toast. It shows up in the Man in the Moon, the 1976 photograph of a face on Mars, the religious images reported in food and water stains, the smiley in two dots and a curve, and in the faces that noise-detection experiments find people reporting in pure static.
How pareidolia works
The visual system is tuned to find faces fast, and the tuning is set loose enough to fire on anything with roughly the right layout: two marks side by side above a third. A socket, the grille of a car, a pepper with two dents and a crease all meet that description, and all of them read as faces before you can decide whether they are. The detector is early and cheap, and its errors are cheap too: missing a real face costs more, for a social animal, than seeing a face in a bush.
Brain imaging shows that the error is a real face response rather than an afterthought. Wardle and colleagues recorded magnetoencephalography in 2020 while people viewed real faces, objects that happen to look like faces, and matched objects that do not, and found that the face-like objects engaged the brain's face response within about a sixth of a second, at the same stage as real faces, while over the next few hundred milliseconds the pattern of activity came to resemble that for ordinary objects. The authors read this as the brain representing the thing as both face and object, with the object information building after the face response. Liu and colleagues found in 2014 that when people reported a face in pure noise, the right fusiform face area, the region most associated with face perception, was active, as it is for real faces.
The tendency is shared with other primates and runs deeper than culture. Taubert and colleagues showed in 2017 that rhesus monkeys looked at face-like objects the way they look at real monkey faces, fixating the illusory eyes and mouth. People readily assign illusory faces an expression, an age and a gender, and a 2022 study by Wardle and colleagues found that illusory faces are judged male far more often than female, which suggests the face system fills in a default rather than reading one off the object.
Expectation turns the gain up. Liu's participants saw faces in noise because they were told faces were there, and the same goes for religious images in food, the Face on Mars, and voices heard in recordings. Pareidolia is not a disorder and, by itself, not a sign of one; it is the ordinary cost of a face detector set to be sensitive, and the demo above produces it with three gray shapes.
The original study
Two studies anchor this page: one on whether people will see faces where there are none, one on what the brain does when they do. Jiangang Liu and colleagues published "Seeing Jesus in toast: Neural and behavioral correlates of face pareidolia" in Cortex in 2014. Participants viewed images of visual noise while lying in a scanner and were told that half of the images contained a face, hidden in the noise, which they should detect. In fact, after a practice phase with faces faintly present, every image in the test phase was pure noise. Participants reported a face in about 34 percent of the noise images anyway. Comparing trials where a face was reported with trials where it was not, the right fusiform face area was more active for the reported faces. The same design with letters produced letter reports in noise and activity in regions associated with letters, so the effect tracked what the person expected to see.
Susan Wardle and colleagues published "Rapid and dynamic processing of face pareidolia in the human brain" in Nature Communications in 2020. Using magnetoencephalography, which follows brain activity on a millisecond scale, they showed people photographs of human faces, of everyday objects that resemble faces, such as a pepper or a handbag with the right configuration, and of similar objects that do not. The face-like objects were represented like faces within about a sixth of a second of appearing, the stage at which real faces are detected, and over the following few hundred milliseconds their representation moved toward the other objects. The face response came first and the object information followed; the face need not disappear for the object to be recognized. Taken together, the two studies show that pareidolia is produced by the ordinary face machinery, that it is fast, and that expectation can trigger it in noise that contains nothing at all.
Everyday examples
Faces are the commonest case, but the same tendency finds figures, animals and messages in sparse or random input.
| Situation | What the bias does | What helps |
|---|---|---|
| A slice of toast, a tortilla or a water stain is reported to show a religious figure | Two marks and a line below them become a face, and the expectation of a particular face fills in the rest | Cover one of the marks and look again, to see how much of the face that single mark was carrying |
| The front of a car looks friendly or aggressive | Headlights and grille are read as eyes and mouth, and the expression affects how the car is judged | Notice that the judgment arrives before any reason for it, and treat it as styling |
| A 1976 orbiter photograph shows a face on Mars | Low resolution, a low sun and a mesa with two shadows produce a face that later images at higher resolution dissolved | Look for the same feature under different lighting or resolution before concluding it is there |
| A voice or a message is heard in a recording played backward or in static | The auditory system finds words in noise, and the words it finds are the ones suggested in advance | Listen without being told what to hear, and compare what different listeners report |
| A chart of random numbers looks like it has a trend or a cycle | The pattern-finding that produces faces in clouds produces trends in noise | Ask what the same chart would look like if the numbers were shuffled |
How to counter it
Pareidolia is not a fault to fix, but there are ways to keep it from making decisions for you.
- Test by removing. Cover the mark that seems to carry the face and look at what is left. This proves nothing on its own, but it shows how much of the face one mark was doing.
- Change the conditions. Look again under different light, from a different distance or at a different resolution. The Face on Mars went away when a sharper camera looked.
- Separate detection from interpretation. The face response arrives within about a sixth of a second, and the sense of what the thing is builds afterward. The face may stay; the second look is where you decide what you are looking at.
- Notice the expectation. Liu's participants saw faces because they were told to expect them. Ask what you were primed to find before you looked.
- Keep it for the cases where it matters. Seeing a face in a socket is harmless. Seeing a trend in random numbers, or a message in static, is the same mechanism with a cost.
What it is not
Pareidolia is the perceptual case of a broader family of pattern-finding.
- It is not the clustering illusion, which finds streaks and clusters in random sequences of events. Both come from the same drive to find structure, but the clustering illusion works on data over time rather than on a single image or sound.
- It is not anthropomorphism, the attribution of human intentions and feelings to animals, objects or weather. Pareidolia supplies the face; anthropomorphism supplies the mind behind it, and the two often arrive together.
- It is not illusory correlation, the perception of a link between two things that are not linked. That is a pattern across cases rather than a figure in one.
- It is not apophenia, the broader term coined by Klaus Conrad for finding meaning in unrelated things, of which pareidolia is the perceptual special case. Nor is it a symptom by itself; seeing faces in objects is common, and is found in other primates too.
Tests and tools that use this
Related biases
Pareidolia FAQ
What is pareidolia?
It is the perception of a meaningful pattern, usually a face, in something that does not contain one: clouds, sockets, car fronts, wood grain, toast. It happens because the visual system detects faces with a fast, loose rule, two marks above a third, that fires on many objects. Brain imaging shows that face-like objects engage the brain's face response at the same early stage as real faces, before the correction arrives.
Why do people see faces in things?
Because detecting faces quickly matters more than detecting them precisely. A detector tuned to fire on any face-like layout rarely misses a real face, at the cost of firing on sockets and peppers too, and missing a real face has historically cost more than a false alarm. Wardle and colleagues showed in 2020 that illusory faces trigger the brain's face response within about a sixth of a second. Expectation raises the rate: people told to expect faces in noise report them in about a third of the images.
How does the demo on this page work?
You rate eight images made of three gray shapes on a light rectangle. In four of them the shapes are arranged as two side by side above a third, the layout the face detector responds to; in the other four the same shapes are scattered. The page compares your average rating for the two arrangements. Rating the arranged shapes as more face-like than the scattered ones counts as pareidolia, which is the usual result. Only the outcome label leaves your browser.
Is pareidolia a sign of a mental health condition?
Not by itself. Seeing faces in objects is common and shared with other primates; Taubert and colleagues found in 2017 that rhesus monkeys look at face-like objects as they look at real faces. It is the ordinary cost of a sensitive face detector. The broader term apophenia, coined by Klaus Conrad, originally described a stage of psychosis, but the perceptual pareidolia this page covers is an ordinary feature of healthy vision and is not diagnostic of anything on its own.
What is the difference between pareidolia and apophenia?
Apophenia is the general tendency to perceive meaningful connections between unrelated things, including patterns in random data, coincidences read as signs, and hidden messages. Pareidolia is its perceptual special case: a figure, most often a face, seen in a specific image or heard in a specific sound. The clustering illusion, which finds streaks in random sequences, is another member of the same family on this site.
What are famous examples of pareidolia?
The Man in the Moon, which is the lunar maria read as a face. The Face on Mars, a mesa photographed by a Viking orbiter in 1976 under a low sun, which later higher-resolution images showed to be an ordinary landform. Religious figures reported in toast, tortillas and water stains. The faces in car fronts and in electrical sockets, and the smiley made of two dots and a curve, which works because the face detector needs no more than that.
References
- 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.org
- Wardle, S. G., Taubert, J., Teichmann, L., & Baker, C. I. (2020). Rapid and dynamic processing of face pareidolia in the human brain. Nature Communications, 11, 4518. doi.org
- Taubert, J., Wardle, S. G., Flessert, M., Leopold, D. A., & Ungerleider, L. G. (2017). Face pareidolia in the rhesus monkey. Current Biology, 27(16), 2505-2509. doi.org
- Wardle, S. G., Paranjape, S., Taubert, J., & Baker, C. I. (2022). Illusory faces are more likely to be perceived as male than female. Proceedings of the National Academy of Sciences, 119(5), e2117413119. doi.org
Cite this source
Psychology.com. (2026, October 20). Pareidolia. Psychology.com. https://psychology.com/biases/pareidolia