Week 1: The pattern-seeking animal

Why a brain built to spot predators in leaves now sees faces in toast, streaks in coin flips, and messages in static.

In 1976, NASA’s Viking 1 orbiter photographed a mesa in Mars’s Cydonia region, lit from a low sun angle, with two symmetrical shadows sitting where eyes would go and a ridge where a mouth would go. NASA’s own caption called it “a huge rock formation… which resembles a human head,” shadows and all, and moved on. The picture didn’t move on with them — it anchored two decades of speculation about an artificial monument on Mars. When the Mars Global Surveyor re-photographed the same mesa in 1998 and 2001, at roughly ten times the resolution and a different sun angle, the face was gone: an ordinary eroded hill, no eyes, no mouth. Nothing on Mars had changed. What changed was the angle of the light and the resolution of the instrument doing the seeing.

The Face on Mars is a useful place to start a course about meaningful coincidence, because it’s a small, checkable version of a much bigger claim this week makes: the meaning was never in the rock. It was supplied, cheaply and automatically, by a visual system built to find faces even where there are none to find — and it’s the same system that will be doing the work in every other case this course examines, right through to a recommendation feed in week 11.

Two costly mistakes are available to an animal deciding whether a rustle in the grass is a predator: treat noise as signal (waste energy, look foolish), or treat signal as noise (get eaten). Evolution did not need pattern detection to be accurate. It needed it to be cheap and to fail in the survivable direction. That asymmetry is the starting point for the whole course: humans inherited a detector tuned to over-fire, and it has never been recalibrated for a world made of headlines, horoscopes and recommendation algorithms instead of grass.

Psychologists have names for the two halves of this. Pareidolia is seeing a specific, recognisable shape in ambiguous sensory input — a face in a power socket, a word in static. Apophenia is the broader tendency to see connection or meaning in unrelated events — noticing that the phone rang right as you thought of the caller, and feeling that this means something. Neither is a malfunction. Both are the ordinary output of a system doing exactly what it evolved to do, at a rate tuned for a much noisier and more dangerous environment than a lecture theatre.

The detector has a favourite hypothesis#

Pattern-seeking has a narrower, sharper cousin worth naming on its own: agency detection. It isn’t enough for a brain to notice a pattern — the pattern that matters most, evolutionarily, is someone did this on purpose. Researchers in the cognitive science of religion describe something close to a hyperactive agency detector: faced with an ambiguous cause, the mind defaults to assuming an intentional agent (a predator, a rival, a watching mind) rather than an impersonal one (wind, erosion, chance), because wrongly assuming agency costs almost nothing, while wrongly ruling it out can cost everything. A shape that merely resembles a rock formation gets noticed and forgotten. A shape that resembles a face gets treated, for a fraction of a second, as something looking back — which is very likely why Cydonia unsettled people in a way an equally improbable rock-that-looks-like-a-shoe never would have. Eyes read as attention, and attention reads as agency.

The same effect, three explanations#

None of the following actually compete with each other — they describe the same phenomenon at three different levels, and holding all three at once is more accurate than picking a favourite:

• Evolutionary: the asymmetric-cost account above — a detector tuned to over-fire because a false alarm and a missed threat are not equally costly.
• Neurological: the visual system has a region, the fusiform face area, that appears to specialise in face detection and can be triggered by very little structural information — two dots and a line will often do it. Pareidolia may be this specific circuit firing on ambiguous input, not a general-purpose “meaning detector” switching on.
• Cultural: what an ambiguous shape gets read as — a face, a religious apparition, an omen of disaster — depends heavily on what a viewer has been taught to look for. The circuitry may be near-universal; the vocabulary it reaches for is learned.

The meaning was never in the rock.

Try it yourself, before the digital version#

Spend one minute on the oldest version of this experiment. Look at some source of visual noise you have nearby — a textured ceiling, static on a detuned screen, the marbling in a countertop — and count how many recognisable shapes (faces, animals, objects) you can find. Most people find several within the first thirty seconds. Nothing was placed there for you to find; you supplied all of it.

What this changes#

This week’s answer to the course’s opening question is deliberately unglamorous: a coincidence starts becoming meaningful the instant a detector like this one fires, and it fires constantly, on nothing, by design rather than malfunction. Every case this course examines — an omen, a dream, a birth chart, a card, a coin toss, a hexagram, a number, a recommendation feed — starts at exactly this step. What changes from week to week is only what happens after the noticing: who links it to what, and whether that link survives being checked.

This week is deliberately the least dramatic in the course. There is no omen, no oracle, no algorithm yet — just the raw machinery, so that every stranger case later in the semester can be traced back to it rather than treated as its own mystery.

Further reading#

• Stewart Elliott Guthrie, Faces in the Clouds: A New Theory of Religion (1993) — the fullest development of pareidolia and agency detection as a single underlying mechanism.

Test the detector#

The activity below is the same coin-sequence problem researchers have used for decades to study this effect. Guess first, then read what was actually going on — the point isn’t to score a right answer, it’s to notice which sequence felt more random, and ask why.