The Face in the Wall Socket

Look at the electrical outlet nearest to you. Two holes and a slot, arranged with the dumb symmetry of industrial safety codes, and yet the thing is grinning at you. It has a face. Not a metaphorical face, not a face you have to squint to assemble. A face that arrives fully formed, unbidden, the instant your eyes land on it. You did not decide to see it. It saw you first.

The front end of a car does the same. Headlight eyes, a grille for a mouth, an expression somewhere between eager and smug depending on the model. Clouds do it, and burnt toast, and the water stain spreading across a ceiling tile, and the moon, which has watched over human nights with the same patient, lopsided face for as long as anyone has bothered to look up. This is pareidolia: the tendency to perceive meaningful faces in objects and patterns that contain no face at all.

It is tempting to file this under the heading of harmless quirks, the visual equivalent of a hiccup. It is not. Pareidolia is one of the most revealing errors the human brain makes, precisely because it is not really an error. It is a window into how perception actually works, which turns out to be almost nothing like the passive recording we assume it to be. The face in the socket is not a glitch in an otherwise faithful camera. It is the camera doing exactly what it was built to do, a few million years too eagerly.

A sandwich worth twenty-eight thousand dollars

The most famous face in the world may have been made of grilled cheese. In 1994, a woman named Diana Duyser in Hollywood, Florida, took a bite of a grilled cheese sandwich and noticed, in the pattern of the browned surface, what she believed to be the face of the Virgin Mary looking back at her. She stopped eating. She kept the sandwich in a plastic case for a decade, and it did not, she insisted, grow a spot of mould. In 2004 she sold it on eBay to an online casino for twenty-eight thousand dollars.

The sandwich is easy to laugh at, and plenty of people did. But the impulse behind it is not fringe. It runs through the whole of human history and every culture that has ever recorded its impressions. The man in the moon. The rabbit some cultures see there instead. The face on Mars, a rock formation photographed by the Viking 1 orbiter in 1976 that launched a thousand conspiracy theories before higher-resolution images revealed an ordinary eroded mesa. Jesus appears on toast, on tortillas, in the wood grain of church doors. Saints materialise in rust stains and pancake batter. What all of these share is not credulity. It is a nervous system that treats faces as the single most important thing it could possibly find, and would rather invent one than risk overlooking a real one.

The word itself is Greek in origin, stitched together from para, meaning beside or alongside, and eidolon, meaning image or shape. Beside the true image, a false one. The term was coined in the nineteenth century by the German psychiatrist Ludwig Kahlbaum, who used it in an 1866 paper describing the vivid, false perceptions that surfaced in his patients ¹. For a long stretch of the century that followed, science regarded pareidolia the way Kahlbaum had: as a symptom, a small failure of an otherwise reliable perceptual apparatus, an error to be noted and, ideally, corrected. Something had misfired. The machine was showing you a face where there was none, and that was a defect.

Then the scanners arrived, and the story turned inside out.

The patch of cortex that only wants faces

Tucked into the underside of the temporal lobe, roughly behind and below your ears, sits a small strip of cortex that appears to have almost no other purpose than to recognise faces. It is called the fusiform face area, and its discovery reframed everything.

In 1997, the neuroscientist Nancy Kanwisher and her colleagues at MIT published a landmark study in the Journal of Neuroscience mapping this region with functional magnetic resonance imaging ². They showed subjects faces, then objects, then scrambled images, and watched which parts of the brain responded. One region lit up dramatically for faces and stayed comparatively quiet for everything else. Show it a face and it fired. Show it a chair, a hand, a house, and it barely stirred. Kanwisher had found what looked like a dedicated module, a piece of neural hardware specialised for a single category of visual information among all the infinite things the eye can see.

This alone would be remarkable. But the truly strange finding came when researchers began feeding the fusiform face area things that were not faces at all. Show it the front of a car, an electrical outlet, a cloud arranged just so, and the region responds anyway, as if a genuine face had walked into the room. The specialisation is real, but it is not fussy. It is exquisitely tuned to a rough template, two things above a third, roughly symmetrical, roughly the proportions of eyes above a mouth, and it will accept a startling range of impostors that satisfy that template. The system errs, consistently and by design, on the side of yes.

The logic behind that bias is captured in a single line that many researchers in the field have variations of: the brain would rather see a false face than miss a real one. It is not being careless. It is making a bet, and the bet is deliberately lopsided.

Faces in pure noise

If the fusiform face area merely responded to face-like objects, you could still argue it was being fooled by the world. The deeper demonstration came from showing people nothing at all.

In 2014, a team led by the developmental psychologist Kang Lee at the University of Toronto ran an experiment that stripped the illusion down to its essentials ³. They placed subjects in a scanner and showed them images of pure visual noise, the grey static you might remember from an old untuned television, containing no faces, no objects, no structure whatsoever. Then they told some of the participants that faces were hidden inside the static, camouflaged, waiting to be found.

The participants saw faces. Not occasionally, but reliably: roughly a third of the time, they reported spotting a face in what was, objectively, random dots. And when they did, their fusiform face area lit up in the same way it would for a photograph of an actual human being. The expectation of a face was enough to manufacture one, and the brain treated its own invention as real perception. The illusion was not happening at the eyes. It was happening much further up, in the machinery of interpretation itself.

This is the finding that overturned Kahlbaum’s framing. Perception, it turns out, is not a faithful transcription of the light hitting the retina. The brain does not passively record the world and hand you the results. It actively predicts the world, generating a constant stream of best guesses about what is out there and then checking those guesses against incoming sensory data, correcting on the fly. This model, sometimes called predictive processing, has become one of the dominant frameworks in contemporary neuroscience. Under it, what you experience as seeing is really a controlled hallucination, an internally generated forecast that the senses are allowed to nudge but never fully author.

Seen this way, pareidolia stops being an error at all. It is the prediction engine running slightly ahead of the evidence. Two dots above a line satisfy the face forecast, and the forecast wins before slower, more deliberate processing can intervene. You see the face first and reason about the outlet second.

The pepper that looked cheerful

If illusory faces were merely detected and then dismissed, they would be little more than curiosities. What makes them uncanny is how much processing the brain lavishes on them after the fact. We do not just see a face in the car grille. We read its mood.

In 2020, David Alais and colleagues at the University of Sydney examined exactly this, publishing their work in the Proceedings of the Royal Society B ⁴. They showed participants a stream of illusory faces, the kind found in everyday objects: a startled-looking electrical socket, a stern building facade, a bell pepper that appeared to be grinning. People had no trouble assigning emotions to these non-faces. A pepper looked cheerful. A house looked disapproving. The expressions felt as legible as those on real human faces.

More striking still was what the researchers found next. The illusion did not stay contained. Exposure to a sequence of illusory faces expressing a particular emotion biased how participants subsequently judged the emotion of real human faces, an effect known as sensory adaptation. In other words, the brain was not filing fake faces in a separate mental folder marked not real. It was routing them through the same emotional-reading circuitry it uses for actual people, and that circuitry got tired and skewed in exactly the way it would after looking at a real crowd. To the machinery, the pepper was, for a moment, a person.

Why the brain chose to be wrong

All of this raises the obvious question. Why would evolution build a perceptual system so willing to be wrong, so quick to conjure company that is not there?

The answer lies in a piece of logic sometimes called error-management theory, and it turns on a simple asymmetry. Imagine an ancestor moving through tall grass at the edge of dusk, when the light is failing and shapes dissolve into ambiguity. A dark form ahead could be a shadow cast by a rock. It could also be a predator crouched and waiting. The brain has to decide, fast, and it will sometimes get it wrong. The crucial point is that the two possible errors do not cost the same.

Mistake a rock for a lion, and you lose a moment. Your heart races, you freeze, you feel foolish when the shadow resolves into stone. Mistake a lion for a rock, and you lose everything. Under those stakes, over enough generations, a nervous system that leans hard toward the first error will vastly outbreed one that risks the second. Better a thousand false alarms than a single fatal miss. The face-detection system inherited that same bias. In the ancestral world, the faces that mattered most were the ones you might otherwise fail to notice: the rival at the treeline, the ally in the dark, the infant whose gaze needed answering. A brain that saw faces everywhere would waste a little attention. A brain that missed them would not pass on its genes.

So the twist, when you sit with it, is not that pareidolia is a bug in an otherwise well-built system. Pareidolia is the system. The very same machinery that misfires on burnt toast is the machinery that lets you glance across a crowded room and instantly recognise a friend, read the flicker of doubt on their face, register that something is wrong before a word is spoken. You could not have the second without the first. Turn off the tendency to over-detect, and you would not gain a more accurate view of the world. You would lose the ability to see faces at all.

The fingerprints on the guess

There is a final, stranger detail that reveals just how much of ourselves we pour into these split-second inventions. When researchers, including the Sydney group and others, asked people to assign a gender to illusory faces, the answers were lopsided. Studies published around 2020 found that people overwhelmingly perceive these accidental faces as male, in some samples on the order of four to one, even when the underlying object contains nothing that could remotely signal a gender ⁵. A face conjured out of a power outlet arrives, more often than not, read as a man.

The object cannot be male or female. It is a piece of moulded plastic. The bias lives entirely in the observer, which means that even the fastest, most automatic layer of perception, the part that operates in a tenth of a second and long before conscious thought, carries the fingerprints of our assumptions. We do not see the world as it is. We see it pre-annotated with the defaults our brains have absorbed.

And this searching begins astonishingly early. Newborns only days old, with almost no visual experience of anything, will already orient toward simple stimuli arranged in the pattern of a face, preferring them over the same elements scrambled ⁶. You did not learn to see faces the way you learned to read or to ride a bicycle. You arrived already looking, the template installed before your first breath, the guessing machine switched on and hunting from the start.

There is something almost tender in that. The same instinct that keeps a species alive in the dark also refuses to let us be alone. It finds company in the grain of a wooden door, warmth in the grille of a passing car, a watchful presence in the moon. To be human, in this sense, is to be incapable of standing in an empty room without sensing that something, somewhere, is looking back. So the next time a wall outlet grins at you from across the room, there is no need to look away or feel foolish. That grin is not a mistake. It is the oldest part of you, ancient and alert and still, after all this time, keeping watch.

Sources

1. Kanwisher, N., McDermott, J., Chun, M. M., ‘The Fusiform Face Area: A Module in Human Extrastriate Cortex Specialized for Face Perception,’ Journal of Neuroscience, 1997. — https://www.jneurosci.org/content/17/11/4302
2. Liu, J., Li, J., Feng, L., Li, L., Tian, J., Lee, K., ‘Seeing Jesus in toast: Neural and behavioral correlates of face pareidolia,’ Cortex, 2014. — https://www.sciencedirect.com/science/article/abs/pii/S0010945214000288
3. Alais, D., Xu, Y., Wardle, S. G., Taubert, J., ‘A shared mechanism for facial expression in human faces and face pareidolia,’ Proceedings of the Royal Society B, 2021. — https://royalsocietypublishing.org/doi/10.1098/rspb.2021.0966
4. Wardle, S. G., Taubert, J., Teichmann, L., Baker, C. I., ‘Rapid and dynamic processing of face pareidolia in the human brain,’ Nature Communications, 2020. — https://www.nature.com/articles/s41467-020-18325-8
5. Johnson, M. H., Dziurawiec, S., Ellis, H., Morton, J., ‘Newborns’ preferential tracking of face-like stimuli and its subsequent decline,’ Cognition, 1991. — https://www.sciencedirect.com/science/article/abs/pii/001002779190045Q
6. Haselton, M. G., Nettle, D., ‘The Paranoid Optimist: An Integrative Evolutionary Model of Cognitive Biases,’ Personality and Social Psychology Review, 2006. — https://journals.sagepub.com/doi/10.1207/s15327957pspr1001_3

Related reading

• The Private Color Inside Your Head
• The Clock in Your Head Is a Storyteller

More from the Mind edition →