The Exaggeration Engine: A Century of Turning the Cuteness Reflex Past Nature

Early in the twentieth century, teddy bears changed shape. The first ones looked like bears. The 1903 specimen in one Cambridge collection has a low forehead and a long snout, and is muzzled. The bears that followed carried shorter snouts and higher foreheads, faces pushed toward the proportions of a human infant. Nobody announced the change. It happened across competing manufacturers, through a long run of small decisions about what would sell.

Robert Hinde and Les Barden measured that drift in Animal Behaviour in 1985, comparing forehead height and snout length across the bears in that collection. They were careful about what it meant. Teddy bears do not reproduce, they wrote, but they are made for sale, so the process can be pictured as a form of selection determined by human preferences. Then they broke their own analogy twice: the variation is not random, because the makers know what they are selecting for, and the products themselves can create the preferences they later satisfy. They conceded that a correlation like theirs cannot on its own be convincing evidence.

The hedging is worth keeping, because the uncomfortable implication survives it. If a product’s shape moves steadily in one direction without anyone planning the whole arc, the thing doing the steering is not a designer’s intention. It is whatever is in the buyer that responds.

The charge

Daniel Harris looked at that response and did not like what he saw. His 1992 essay “Cuteness” is the most quotable attack on the subject anyone has written, and its claim is that cute is a species of damage rather than a species of beauty. He calls it “the aesthetic of deformity and dejection.” The mechanism, in his account, is pity: the grotesque reads as cute because the grotesque is pitiable, and the aesthetic works by manufacturing creatures deformed enough to feel sorry for.

He is very good on objects. Teddy bears are “stump-limbed and obese.” The doll So Shy Sherri he files as an “anatomical disaster.” And he puts the blame on the people drawing them: “Cuteness, in short, is not something we find in our children but something we do to them.” What the designer is doing, he says, is an unconscious attempt to maim, hobble and embarrass the thing being idolized.

I leaned on that argument in a previous note, where it did real work. It deserves a harder test than I gave it, because on the central question of what cuteness actually is, Harris turns out to be substantially wrong, and the way he is wrong is more interesting than the polemic.

The reflex is real, and it is fast

Harris is arguing against Konrad Lorenz without quite saying so. Lorenz proposed in 1943 that a specific cluster of infant features (large head, high forehead, big low-set eyes, round cheeks, small nose and mouth, short thick limbs) works as an innate releaser for caretaking behavior. He called it the Kindchenschema. Harris’s whole position depends on this being a story we were sold rather than a reflex we arrived with.

The timing evidence is the problem. Morten Kringelbach and colleagues, reviewing the neuroscience in Trends in Cognitive Sciences in 2016, report that infant faces produce activity in the orbitofrontal cortex at around 130 milliseconds, with rapid selective responses in parents and non-parents alike. That is faster than the machinery of considered emotional appraisal. Whatever is happening at a seventh of a second is not a preference being consulted. Melanie Glocker and colleagues had already shown that high baby-schema faces activate the nucleus accumbens, a reward structure, in women who have never had children. The mechanism is not confined to our own species either. Adapting to puppy faces measurably shifts how cute people then rate human infant faces, which points to a coding of cuteness shared across species and explains why a puppy works on you at all.

You cannot advertise your way to a brain response that fast in people who have never had children. Something was there before the toy industry found it.

And the feeling is not pity

Harris’s second claim is about which emotion is running. He says pity. The measurements say otherwise, and this is where his account comes apart most cleanly.

Hiroshi Nittono’s group had people look at cute images and then perform a delicate manual task. Performance improved by roughly 44 percent, against about 12 percent for images of adult animals. Attention narrowed. Pleasant images of food did nothing comparable. That is a profile of focused approach, of leaning in toward something you want to handle carefully. It is not the profile of sorrow.

Gary Sherman and Jonathan Haidt push further and challenge Lorenz from the other side: cuteness, they argue, is less a trigger for parental care than a trigger for sociality, a prompt to engage and to attribute a mind to the thing you are looking at. Cuteness humanizes. That is why cute objects get named and talked to, and why toys are built to invite play rather than nursing.

Even the aggressive-sounding part of the response has been measured and turns out to be something else. The urge to squeeze a cute thing, which Harris would read as the sadism surfacing, is what Oriana Aragón’s group called a dimorphous expression: an outward display that contradicts the inner state, regulating an excess of positive feeling. The researchers are explicit that no desire to cause harm is involved.

Joshua Paul Dale, who has done more than anyone to make cuteness a serious field, states the objection directly. He argues that reading the desire to master a cute object as sadistic reflects a linguistic bias in English, and that the pleasure of being overwhelmed by cuteness is better understood as masochistic, aimed at the self rather than the object. His positive definition is the useful one: cuteness is an appeal, an invitation to engage in companionship, cooperative play and emotional exchange.

That is a real rebuttal to the note I published before this one, and I would rather print it than route around it.

What is actually happening

The honest position is not that Harris was wrong about everything. He was wrong about the emotion and wrong about the origin, and right about the deformation.

The teddy bear did drift. Stephen Jay Gould documented the same drift in Mickey Mouse, writing for the character’s fiftieth birthday. He measured three signs of what he called creeping juvenility: eye size rose from 27 to 42 percent of head length, head length from 42.7 to 48.1 percent of body length, and the cranial vault expanded substantially. The limbs thickened and shortened too, though Gould did not quantify that one. A character he described as originally “a rambunctious, even slightly sadistic fellow” had been remade into something safe. Sanrio built an industry on the principle from 1971. What none of this requires is that the emotion be manufactured. It requires only that a real reflex exists and that the market can find its edges by trial.

So cuteness is a dial. Underneath sits a genuine, fast, pre-cultural signal. On top sits a century of commercial pressure turning that signal up past anything nature produces. Real infants do not have eyes that size. The industry is amplifying a channel that was already open, neither inventing the response nor leaving it untouched.

The evidence for the drift is real and softer than the story usually told, and the honest version says so. Hinde and Barden’s snout finding appears to have been complete by the 1930s, which makes it a sharp early shift rather than a century-long creep. Paul Morris and colleagues found in 1995 that four-year-olds did not prefer the baby-featured bears at all, while six- and eight-year-olds did. Their study tested only children, up to age eight, so the safe reading stays narrow: the drift is not driven by the youngest child the toy is ostensibly bought for, and a taste for baby features arrives later than the toy does. A preregistered replication published in 2025, working from 250 bears across eight manufacturers between 1900 and 1980, confirmed that forehead height increased but did not confirm the rounder heads or the bigger eyes. That replication is also where the power critique comes from: the work it was checking rested on fifteen bears and three Mickey drawings. The direction of travel survives. The tidy version of it does not.

Past sweetness

The dial has a far end, and the far end is where the interesting work is now.

Simon May’s argument is that the force of cute has never really been about helplessness inviting protection. He puts it down to what he calls a teasing indeterminacy, an instability of power, gender, age, ethnicity and morality that leaves you unsure how to place the thing in front of you. His reading of Harris’s own example turns it around. So Shy Sherri, swollen ankles and pigeon-toed feet included, “is endearing precisely as an ‘anatomical disaster.’” Cute compels us, May writes, “not because of such vulnerabilities, but because we stumble across them in what we expect to be a Garden of Eden.” What he is describing sits closer to the uncanny than to pity.

Hello Kitty makes the point commercially. She has no mouth. Christine Yano reads that absence as the product: the mouthless face is a sounding board that synchronizes with the mood of her viewer, a blank mirror that never fails. Sanrio’s own account runs alongside it, that Kitty has no mouth so she may better reflect the feelings of those who look upon her. Cuteness here works by leaving a space rather than filling one.

Japanese looks at first like it hands Harris a win. Kawaii and kawaisō, the word for pitiable, descend from a common ancestor, kawahayushi, assembled from the words for face and flushing and used to name a sense of shame. Cute and pitiful really are related at the root. The relation is historical only, though. Shiri Lieber-Milo and Hiroshi Nittono record that the connotations of pity and shame had almost entirely faded by the end of the Edo period, and that modern kawaii carries almost no negative charge. The etymology preserves a trace of where the concept started. It does not describe what the word does now.

And then Labubu, which is the cleanest evidence that sweetness stopped being the point. Kasing Lung’s creature has a mouthful of pointed teeth, oversized ears and a grin that reads as mischief rather than innocence. Sold by Pop Mart in blind boxes, the property behind it became the company’s largest by a wide margin: RMB 3.04 billion in 2024, a rise of 726.6 percent on the year before, and 23.3 percent of all company revenue. None of Harris’s vocabulary fits: it is awake, slightly feral, and it sold enormously.

The boom did not hold, and the correction is worth stating rather than skipping. Resale prices peaked around the middle of 2025 and fell through the third quarter, with mini Labubu resale down roughly 30 percent by early September, and Reuters reported the drop had more to do with Pop Mart flooding supply than with demand disappearing. The durable claim here was never about price. It is that the shape which won this much attention was the ugly one. I would file that under what Japanese calls kimo-kawaii, creepy-cute, though that is my lens applied to a Chinese product rather than a category its makers claimed.

Why this matters off the toy shelf

Every engineered signal works the way cute works. You locate a response that is already running in the person, and then you turn it up. Trust signals do this, which is the desk’s actual subject. A security badge, a green checkmark, an audit seal, a founder’s face on an about page: none of these create confidence from nothing. They borrow a reflex and amplify it.

Which means the design question is never only whether a signal works. It is how far up the dial you are, and whether the thing you are amplifying is still connected to something true underneath. Turn a real signal up moderately and it reads as reassuring. Turn it up past what the evidence supports and it starts reading as strange, then as suspicious. That is the same failure I described in color as evidence: a confident bright applied to a shaky number does not make the number stronger, it makes the reader’s trust worse calibrated than before they looked.

The toy industry found the edge of that curve by running the experiment for a hundred years across millions of units. Most people designing a trust surface get one attempt and no feedback.

What the evidence licenses

The innate substrate is solid: the fast orbitofrontal response, the reward activation in non-parents, the cross-species transfer. The reframing from pity to approach and sociality is well supported by the behavioral work. The commercial amplification is documented, though the specific neoteny-drift measurements are weaker than their popularity suggests, and I have tried to say so where it matters.

What stays genuinely contested is the moral reading. Harris and Ngai hear aggression in the way we handle cute things; Dale hears an invitation and calls the aggression reading an artifact of one language. I do not think the evidence settles that, and my own previous note came down harder on one side of it than the record strictly supports. What survives from Harris after all the corrections is narrower than his essay claims and still worth keeping: the shapes were pushed, deliberately, past nature, by people who were paid to find out how far the pushing could go.

FIGURE · IN PRODUCTION

The exaggeration curve: an innate baseline (real infant proportions) on the left, a century of commercial drift moving right, a sweetness peak, and an uncanny edge beyond it. Teddy bear, Mickey, Hello Kitty and Labubu plotted along the curve. Legend: what is amplified (head ratio, eye size, limb reduction) against how the result reads (ignorable, appealing, strange).