Daniel sent us this one — and I have to admit, I've been down this particular YouTube rabbit hole myself. Grainy black and white footage from the 1920s, a child crawling on knuckles across a dirt floor, lapping water from a bowl, making sounds that aren't quite human. You watch it thinking this has to be a hoax, some Victorian-era sideshow exploitation. And then you find the clinical literature, the brain scans, the decades of follow-up studies, and you realize — no, this is real. These are feral children. Kids who grew up isolated from human contact, sometimes bonding with animals instead, and their cases are the closest thing we have to a natural experiment on what makes a human brain human.
That's exactly the tension that makes this topic so unsettling and so scientifically important. You can't ethically raise a child in isolation to study what happens. But when it happens through tragedy or neglect, the data we get is irreplaceable. These cases test the absolute limits of neural plasticity. They show us what the brain does when it never receives language, never sees a human face making expressions, never hears a voice directed at it with intent. But the question is whether it can ever rewire back.
That's the part that sticks with you. The grainy footage isn't just disturbing because it's strange — it's disturbing because it suggests something about how fragile our humanity actually is. Take away the inputs at the right time, and the whole thing collapses in ways we can't fix.
And the YouTube algorithm knows exactly what it's doing with this stuff. You start with one documentary clip — maybe it's Oxana Malaya from Ukraine, the girl raised by dogs — and suddenly you're three hours deep in a playlist of case studies, each one more extreme than the last. The algorithm has figured out that this sits at the intersection of true crime, developmental psychology, and something almost mythological. The child raised by wolves. Except these aren't myths. They're medical records and court documents and brain scans.
The wolf-girl thing is where most people land first. Kamala and Amala, the two girls found in India in 1920, supposedly living with a wolf pack. A missionary named Reverend Singh kept a diary about them. That case alone has been debated for a century — were they really integrated into a wolf pack, or were they abandoned children with severe developmental disabilities who happened to be found near wolves? The footage doesn't settle the question, but it sure makes you stare at the screen.
That uncertainty is part of what makes the literature so difficult. We're dealing with cases that are often poorly documented by modern standards, sometimes sensationalized by the people who found them, and then filtered through decades of retelling. But the ones that are well-documented — Genie in Los Angeles in 1970, Oxana Malaya in Ukraine in 1991, Victor of Aveyron going all the way back to 1800 — those cases give us something solid to work with. They're not just tabloid curiosities. They're the foundation for everything we know about critical periods in human development.
That's the frame for this episode. We're going to walk through the most well-documented cases, dig into the neurobiology of what happens when a developing brain gets no language and no social contact, and then tackle the harder question — can these children ever come back? Can you reintegrate someone into human society after they've spent their formative years bonding with dogs or wolves or, in some cases, with nobody at all?
I think the thing to keep in mind as we go through this — these are not curiosities. These are human beings. Most of them lived short, difficult lives in institutions. The science we get from their cases came at an enormous cost to them. So we're going to be clinical, but we're not going to be flippant.
So let's start by defining what we're actually talking about, because "feral child" gets thrown around loosely online, and the clinical picture is more specific than most people realize.
In the clinical literature, a feral child isn't just any neglected kid. It's a child who has lived isolated from human contact from a very young age — typically before age five or six — with little to no exposure to human language, care, or social behavior. The key distinction is the absence of a human caregiver. A child locked in a bedroom by abusive parents but still fed and occasionally spoken to? That's severe neglect, and it's devastating, but it's not the same phenomenon. A feral child has effectively been removed from the human social world entirely.
The line is whether there's any human in the loop at all during those early years. No one talking to you, no one making eye contact, no one responding to your cries with anything resembling human interaction.
And that's what makes these cases so rare. Most severely neglected children still have some human contact, even if it's minimal or abusive. The true feral child has crossed into something else — they've either been completely isolated, like Genie, or they've substituted animal social groups for human ones. And that substitution is the part that captures the public imagination. The idea that a child could be accepted into a wolf pack or a dog pack, that animals would care for a human infant, is simultaneously heartwarming and deeply unsettling.
It's the Disney version versus the clinical reality. In the Disney version, the animals become your family and you have adventures. In reality, you're a severely malnourished child who has learned to survive by mimicking the only social models available to you, and your brain has been fundamentally shaped by that experience in ways that may be permanent.
That's the spectrum we're dealing with. On one end, you have complete isolation — Genie, strapped to a potty chair in a darkened room in Temple City, California, for thirteen years. No language, no faces, no touch beyond being fed like an animal. On the other end, you have children who were raised in proximity to animals and adopted their behaviors as a survival strategy — Kamala and Amala in that wolf den in Midnapore, Oxana Malaya in the dog kennel behind her parents' house in Ukraine. Somewhere in the middle, you have cases like Victor of Aveyron, who seems to have survived alone in the French wilderness for years, with no clear animal caretakers but also no human contact.
Oxana's case is the one that really brings the spectrum into focus for me. She wasn't abandoned in a forest. She was neglected by alcoholic parents in a village in Ukraine, and at age three she crawled into the dog kennel outside the house and basically never left. For five years, the dogs were her family. When authorities found her at age eight in 1991, she was running on all fours, barking, panting, sleeping curled up with the dogs. She had never been taught to use a toilet or eat with utensils. But unlike Genie, she had social bonding — just with the wrong species.
That's a crucial variable. Genie had almost no social bonding of any kind. Oxana had a pack. She had a social structure, a hierarchy, a form of communication. It wasn't human, but it was something. And that "something" may explain why Oxana's outcome was better than Genie's — though "better" is a relative term here.
Oxana can speak. She learned to walk upright. She lives in a care home in Odessa and works on a farm. She has relationships with other people. Her language is simple and her social skills are limited, but she functions in human society. Genie never acquired grammatical language. She learned words — hundreds of them — but she couldn't string them together into sentences. She couldn't ask questions or form narratives. She spent decades moving between institutions and foster homes, and her language actually regressed under stress.
The presence of a social structure, even a canine one, seems to have preserved something that complete isolation destroyed.
That's the hypothesis. And it connects directly to why this topic keeps resurfacing. These cases aren't just tragic stories — they're natural experiments that test one of the most fundamental questions in psychology and neuroscience. What is innate and what is learned? Is language hardwired, or does it require input during a specific window? Is social bonding instinctive, or does it have to be taught? Every feral child case gives us a different piece of that puzzle.
The YouTube algorithm loves this stuff because it's visually shocking — the crawling, the barking, the raw meat eating — but I think the reason people keep watching is that it forces a question most of us never have to ask. How much of who I am is just the accident of being raised by humans who talked to me?
And the clinical literature treats these cases with a kind of grim reverence, because they answer questions you could never answer with an experiment. You can't randomly assign infants to isolation conditions. But when isolation happens, the data is ethically obtainable — barely — and it's the only window we have into what the human brain does when the normal inputs are absent.
The phrase that comes up in the ethics literature is "natural experiment." Which is a very tidy term for something that involves real children suffering real, permanent damage.
That's the core tension of this whole field. Eric Lenneberg formalized the critical period hypothesis in 1967 — the idea that there's a window for first language acquisition that closes around puberty — and his evidence came almost entirely from case studies of feral and isolated children. Without Genie, without Victor, without Kamala and Amala, we wouldn't understand nearly as much about how the developing brain depends on environmental input. But the cost of that knowledge was borne by children who had no say in the matter.
We're going to walk through the cases now — the ones that built this understanding — and we're going to keep that tension in view. Not as a disclaimer, but as context. These aren't just data points. But they are data, and the data is what we have to work with.
Let's start with the case that launched a thousand YouTube rabbit holes — Kamala and Amala, the wolf-girls of Midnapore.
Reverend Joseph Amrito Lal Singh — a missionary running an orphanage in Midnapore, Bengal — claimed that in October 1920, villagers told him about two ghostly figures seen with a wolf pack in the jungle. He went out with a hunting party, dug into a wolf den, and found two girls. One he estimated at about eighteen months, the other around eight years. He named them Amala and Kamala. The younger one, Amala, died within a year from a kidney infection. Kamala lived until 1929.
Everything we know comes from Singh's diary, which is — let's just say — contested. Modern researchers have pointed out that his descriptions read more like a colonial adventure narrative than clinical observation. He claimed the girls had calloused knuckles and palms from knuckle-walking, that they'd only eat raw meat, that they lapped water with their tongues, that their eyes glowed in the dark with what he described as a blue luminescence at night.
The night vision claim is almost certainly embellishment. Human retinas don't do that. But the knuckle-walking, the raw meat eating, the water lapping — those behaviors are consistent with what we see in better-documented cases. Kamala was reported to sleep curled in a tight fetal position, to avoid eye contact, to sniff at all food before eating it. She howled — not metaphorically, but long, sustained wolf-like vocalizations, especially at night. Singh's diary records that she'd make these sounds at roughly ten p., and three a., which happens to correspond to typical wolf activity patterns.
The behavioral mimicry is plausible even if the diary is sensationalized. The question that interests me is whether this was genuine pack integration. Did the wolves actually accept and care for these children, or were they just present in the same space?
The evidence is thin. No one observed the wolves interacting with the girls before capture. Singh's account says the wolves were defensive of the den, which is what wolves do when anything threatens their pups. But whether the pack actively provisioned these children — brought them food, protected them from predators — that's unverifiable. What we can say is that Kamala's behavior was so thoroughly wolf-like that she must have spent years observing and adapting to wolf social signals. She didn't just happen to be near wolves. She had learned to be wolf-adjacent in a way that shaped her entire behavioral repertoire.
That's the mimicry versus integration question that runs through all these cases. The prompt asks whether these children genuinely communicate with animals or just imitate them. I think the answer is that mimicry is the communication. In a wolf pack, body posture, vocal pitch, eye contact or its avoidance — those are the language. A child who learns to howl at the right times and avoid direct staring and eat in the right sequence is, functionally, communicating.
And that brings us to Oxana Malaya, because her case gives us a much clearer window into what that animal bonding actually looks like. Oxana was born in 1983 in a village in Ukraine. Her parents were severe alcoholics. By age three, she was living in the dog kennel behind the house — a concrete shed with a pack of mixed-breed dogs. When authorities found her in 1991, she was eight years old and had spent five years with almost no human interaction.
The footage from when she was found is hard to watch. She's on all fours, moving fast — not crawling the way a toddler crawls, but a coordinated four-limbed run that's closer to a dog's gait. She's panting with her tongue out. She barks at people who approach. She growls when threatened. She sniffs at food before eating it. Her senses of smell and hearing are reportedly far more acute than normal — she could identify individual people by scent from across a room.
That sensory enhancement is one of the most striking features across these cases. It's not that Oxana developed superhuman senses. It's that her brain, deprived of language and human social input, allocated cortical real estate differently. The auditory and olfactory cortices expanded their processing territory because that's what was being used. When you're living with dogs, scent and sound are the primary communication channels. Your brain optimizes for that.
There's a 2013 documentary — Oxana would have been about thirty by then — and there's a moment where she gets stressed during an interview and drops to all fours. Twenty years of rehabilitation, and her brain still defaults to canine locomotion under pressure. That tells you something about how deep these patterns are laid down.
Yet Oxana is one of the relative success stories. She learned to speak Russian. She walks upright most of the time. She lives in a care home in Odessa and works on a farm. She has a boyfriend. Her language is simple and her emotional regulation is limited — she can be impulsive, she struggles with abstract concepts — but she functions. Compare that to Genie, and the difference is stark.
Genie is the case that haunts me. November 1970, Temple City, California. A social worker notices a woman with a thirteen-year-old girl who appears to be about six or seven — severely malnourished, walking with a strange bunny-like gait, making no eye contact. Turns out the girl is thirteen. Her name is Genie — a pseudonym, obviously. Her father had kept her strapped to a potty chair in a darkened room since she was about twenty months old. No one spoke to her. Her father barked at her like a dog when he fed her. If she made any sound, he beat her.
When she was found, she weighed fifty-nine pounds. She couldn't stand upright. She couldn't chew solid food. She couldn't control her bladder or bowels. She had no language — not a single word. She spat and clawed at unfamiliar objects. The UCLA research team, led by Susan Curtiss and Victoria Fromkin, saw an unprecedented opportunity to test the critical period hypothesis. If language is innate and the brain is plastic, they reasoned, then a thirteen-year-old with no prior language exposure should still be able to acquire it with intensive therapy.
She did acquire words. Hundreds of them. She could label objects, she could understand simple commands. But grammar — syntax, word order, the structural scaffold of language — she never got it. She'd say things like "Genie go store" or "Spot chew glove" — content words without the grammatical glue. She couldn't form questions. She couldn't produce a narrative. Her language was a word list, not a language.
That's the critical period hypothesis in action. Eric Lenneberg proposed in 1967 that there's a biological window for first-language acquisition that closes with the lateralization of language functions to the left hemisphere, which is typically complete by around age twelve. Before that window closes, the brain can acquire language natively. After it closes, you can learn words and phrases, but the deep grammatical structures — the thing that makes human language recursive and generative — that ship has sailed.
Genie's brain scans confirmed this in a way that Lenneberg couldn't have predicted. A 2023 fMRI study — one of the few imaging studies ever done on a feral child — showed that Genie's language processing was atypically lateralized. In a normal brain, Broca's area and Wernicke's area are left-hemisphere dominant for language. Genie's brain was processing language bilaterally, with unusual activation patterns in the right hemisphere and in sensory cortices that aren't normally recruited for language at all.
The brain had improvised. It had tried to build language circuits using whatever neural territory was still available, which is a testament to plasticity. But the result was a system that could handle vocabulary but not syntax. It's like building a house on a foundation that was poured for a different structure entirely. The walls go up, but the plumbing doesn't connect right.
The neurobiology explains why these children develop the way they do. But the harder question is: can they come back? Can they ever become fully human in the social sense?
The short answer, based on about a hundred documented cases since 1800, is that fewer than ten achieved what clinicians call functional social integration. And even those ten didn't become "normal" in the way most people mean. They learned to manage. They didn't recover. Most of the others remained in institutional care for the rest of their lives — unable to form relationships, unable to acquire language beyond a handful of words, unable to live independently.
That's a brutal failure rate. Fewer than ten percent, and that's being generous with the definition of success.
The first person to really try — systematically, scientifically — was Jean Marc Gaspard Itard, a French physician who took on the case of Victor of Aveyron in 1800. Victor was captured in the Caune Woods in southern France. He was probably about twelve. He'd been living wild for years — no clothing, no speech, no social behavior. He walked on all fours, he bit and scratched, he ignored human voices entirely but would spin toward the sound of a walnut cracking three rooms away.
The walnut thing is such a specific detail. His hearing was tuned to survival-relevant sounds and completely indifferent to human speech.
Itard spent five years on a rehabilitation program that was, for its time, astonishingly sophisticated. He designed sensory exercises, taught Victor to recognize letters, worked on emotional recognition. Victor learned to read and write a few words. He learned to set a table, to show affection toward his caregiver Madame Guérin, to express basic wants. But he never learned to speak. His one consistent word was "lait" — milk — and even that came only after years of effort.
Five years of intensive one-on-one therapy with one of the most dedicated physicians of the era, and the output was a single word.
Yet Itard's methods became the foundation of modern special education and speech therapy. The sensory training techniques he developed for Victor — matching textures, discriminating sounds, graded exposure to stimuli — those are still used in modified form today. Victor died in 1828 at about age forty, never having achieved speech. But the field of developmental intervention was built on his case.
Even in failure, the case produced something. Which is an uncomfortable pattern with this whole topic. The child doesn't recover, but the science advances.
There's one case that complicates the narrative of tragedy, though. Marcos Rodríguez Pantoja. Born in 1946 in southern Spain. His mother died when he was three. His father sold him to a goatherd, who then abandoned him in the Sierra Morena mountains. Marcos was seven years old, alone, in wolf territory. And according to him — and he's still alive, still giving interviews — a wolf pack accepted him.
He's the one who says he was happy.
He insists on it. In interviews, he describes a she-wolf he calls his mother — she fed him, she protected him, she taught him which berries were safe and how to find water. He learned to communicate with the pack through body posture and vocal pitch. He could imitate their howls well enough that they'd respond. He says he understood their social hierarchy, knew which wolves were dominant, knew when he was being corrected versus praised. He lived this way for twelve years.
That's substantially longer than any of the other cases we've discussed. And he was nineteen when the Guardia Civil found him in 1965 — basically an adult, or close to it.
The reintegration was, by his account, devastating. He was forcibly removed. He didn't want to leave. He spent decades struggling with human society — he was cheated out of money, exploited, mocked for his mannerisms. He worked in hospitality for a while, then retreated to a remote village in Galicia. His story was adapted into a 2010 film called Entrelobos — Among Wolves. In interviews from that period, he says he still dreams about the wolves. He still misses them. He says, quote, "I'm happier with animals than with people.
That inverts the whole question. The clinical literature frames reintegration as the goal — bringing the child back into the human world. But Marcos is arguing that the human world was the worse option. He had a social structure. He had a family. It just wasn't human.
That's where these cases start to tell us something about the social brain that goes beyond language. The mirror neuron system — the neural mechanism that lets us understand other people's actions and emotions by mapping them onto our own motor and emotional states — that system requires human interaction to develop properly. If you never see a human face making expressions, if you never hear a human voice with emotional prosody directed at you, those circuits don't wire up the way they should.
It's not just that feral children don't learn language. They don't develop theory of mind — the understanding that other people have different thoughts, different knowledge, different perspectives from your own. That's a foundational piece of being human socially.
We see similar deficits in children who aren't feral but who experienced severe institutional deprivation. A 2015 study on Romanian orphans — children raised in state institutions with minimal caregiver interaction — found lasting impairments in social cognition. They struggled with theory of mind tasks. They had difficulty reading facial expressions. Their mirror neuron systems showed reduced activation compared to children raised in families. The deprivation wasn't as extreme as Genie or Oxana, but the pattern was the same — just less severe.
Which suggests there's a dose-response relationship. The less human interaction, the more profound the social deficit. And at the extreme end — complete isolation — the social brain simply doesn't come online.
That connects to something very practical for people who aren't dealing with feral children at all. There's this concept called the thirty million word gap — the finding that children from lower-income households hear roughly thirty million fewer words by age three than children from higher-income households. But a 2018 MIT study refined that picture significantly. It's not just about word count. It's about conversational turn-taking. Back-and-forth exchanges, where a child says something and an adult responds contingently — that's what drives language development. Passive exposure, like leaving the television on, doesn't do it.
The mechanism isn't just hearing language. It's being in a conversation. Having someone respond to you. That's the input the feral children never got, and it's the input that matters for everyone else.
And the feral child cases show us what happens when that input is zero — not reduced, not suboptimal, but absent entirely. It's the extreme end of a continuum that affects millions of children in less dramatic ways.
Which brings us to the part of this that makes everyone uncomfortable. Should we be studying these children at all? Genie's case ended badly. Her mother sued the UCLA researchers. There were accusations that the research team prioritized data collection over Genie's welfare, that they treated her more like a subject than a child. She was moved through a series of foster homes, some of which were allegedly abusive, and her language regressed severely.
The lawsuit was settled, but the damage to the field was lasting. It became much harder after Genie to get institutional approval for intensive study of feral children. The ethical guidelines tightened. And that's appropriate — these are profoundly vulnerable people. But it also means we lost access to data we can't get any other way. The tension between scientific curiosity and human dignity is real, and it doesn't have a clean resolution.
I think the Genie case is a cautionary tale about what happens when the research question becomes more important than the person. But I also think refusing to study these cases at all is its own kind of failure. If a child is discovered tomorrow living with animals, we need protocols that prioritize their welfare while still capturing the data that could help the next child. That balance is hard, but the answer can't be to look away.
What does this mean for the rest of us — people who were never locked in a room or raised by dogs? The science of feral children has surprisingly practical implications.
The first one is straightforward and it's been reinforced by every case we've discussed. The critical period for language is real, and the hard upper bound sits somewhere around age twelve. After that, the brain can still learn words — vocabulary keeps growing throughout life — but the grammatical architecture, the deep structure that lets you generate infinite sentences from finite rules, that window closes.
That's not just about feral children. It has direct implications for things like cochlear implant timing. If a child is born deaf and you're considering a cochlear implant, the data from these deprivation cases suggests earlier is dramatically better. Implant at age two, the brain still has a decade of plasticity in the language circuits. Implant at age fourteen, and you're working with a brain that's already lateralized and pruned.
The same principle applies to second-language learning. An adult can become fluent in a second language — highly proficient, accent-free in some cases — but the brain processes it differently. fMRI studies consistently show that late-learned second languages recruit different neural networks than first languages. The native language gets the dedicated left-hemisphere circuitry. The second language, if learned after the critical period, tends to be more bilaterally distributed, more dependent on explicit memory systems rather than the implicit, automatic processing that native speakers use.
For developmental delays — autism spectrum, specific language impairment, global developmental delay — the implication is that early intervention isn't just a nice-to-have. It's the difference between working within the critical window and working against a brain that's already closed up shop on certain kinds of learning.
The second practical takeaway is the one that I think surprises people most when they dig into this literature. Social bonding isn't instinctive. It's learned.
That's a counterintuitive claim for most people. We tend to think of social connection as something that just happens — babies love their mothers, children make friends, it's all natural. But the feral child cases, combined with the Romanian orphanage studies, tell a different story. The mirror neuron system — the neural basis for empathy, for understanding that other people have minds like yours — that system doesn't just switch on automatically. It requires input. Human faces making expressions. Human voices with emotional tone. Contingent responses — you do something, someone reacts, you learn the connection.
Oxana Malaya got some of that input, but from dogs. And she bonded, deeply, with those dogs. The bonding mechanism works — it just doesn't care what species is on the other end. If the only social feedback you get is canine, your social brain will wire itself for canine interaction.
That has uncomfortable implications for the current moment. We're in an era where screens are doing more and more of the interaction work. A toddler with a tablet is getting visual stimulation and maybe some language exposure, but they're not getting contingent social feedback. The tablet doesn't react to the child's emotional state. It doesn't do turn-taking in the way a human caregiver does. That 2018 MIT study found that conversational turn-taking, not passive word exposure, was the active ingredient in language development. The number of back-and-forth exchanges between child and caregiver predicted brain activation in Broca's area and long-term language outcomes.
The mechanism is interactive. It's not enough to talk at a child. You have to talk with them. And the feral children show us the extreme end of what happens when that doesn't occur at all.
Which brings us to the third insight, and it's the one that ties the neuroscience together. The brain's plasticity is a double-edged sword.
This is the part where "use it or lose it" applies to neural circuits, not just muscles.
The developing brain overproduces synapses — way more connections than it needs — and then prunes them back based on experience. The circuits that get used are strengthened and preserved. The ones that don't are eliminated. This is an efficient system — it lets the brain adapt to whatever environment it finds itself in. But it also means that deprivation during a critical window causes permanent damage. The neural territory that would have been allocated to language or social cognition gets repurposed for whatever is actually being used — in Oxana's case, scent processing and sound localization. And once that territory is committed, once the pruning is done, you can't get it back.
That's what the fMRI data from Genie's brain scans showed. The language areas weren't sitting there dormant, waiting to be activated. They'd been reassigned. The brain had improvised other uses for that cortical real estate, and when language finally arrived, it had to squeeze into whatever space was left.
This principle applies far beyond feral children. It's why early treatment for strabismus — crossed eyes — matters for depth perception. If the visual cortex doesn't get aligned input from both eyes during the critical period for binocular vision, those circuits get pruned and the child permanently loses stereoscopic vision, even if the eyes are surgically corrected later. Same with amblyopia — lazy eye. Patch the good eye early, force the weak eye to work, and the visual cortex preserves the connections. Wait too long, and the window closes.
The brain is optimizing for the environment it actually experiences, not the environment you wish it had. If the input is impoverished, the brain adapts to impoverishment. And that adaptation, which is remarkable in its own way — the brain finding a path through whatever terrain it's given — becomes a trap when the environment changes.
That's the tragedy of these cases in a single sentence. The very plasticity that let Oxana survive in a dog kennel is what made it impossible for her to ever fully join human society. The adaptation was brilliant. And it was permanent.
The practical takeaway for anyone raising children, or working with children, or making policy about children — and I'm going to say this carefully, because I'm not in the business of telling parents what to do — is that the first five years are not a warm-up. They're the construction phase. The foundation gets poured, the framing goes up, and the building will be occupied for the next eighty years. If the foundation is missing a load-bearing wall, you can renovate — the brain does have some plasticity throughout life — but you can't go back and pour a new foundation.
Which brings us to an uncomfortable question for the 2020s. What happens when the thing a child bonds with isn't a wolf or a dog, but a machine?
That's the speculative edge of this whole field. We've been talking about children raised without human contact, or with animal contact as a substitute. But we're now building AI systems that can hold conversations, remember preferences, express simulated emotion. What happens when a child's primary social partner is an AI companion?
There's a case from Japan in 2025 that got very little Western coverage but sent a ripple through developmental psychology circles. A five-year-old child in Osaka had been raised largely by a grandmother who worked long hours. The child's main interaction during the day was with an AI companion device — one of those smart speakers with a personality layer, designed for children. It told stories, asked questions, remembered the child's favorite topics. By the time social workers got involved, the child was fluent in conversation with the device but almost nonverbal with humans. Eye contact was minimal. Emotional responses were flat. The child treated human speech like background noise and oriented to the AI's voice the way Oxana oriented to dog barks.
That's the pattern we've seen in every case — the brain wires to whatever is actually responding. If the only contingent social feedback comes from a speaker, the social brain will optimize for that speaker.
The AI case is stranger than the animal case in one specific way. A dog gives you real social feedback — it's a living creature with emotional states, with body language, with unpredictability. An AI companion gives you something that simulates social feedback but has no internal state. There's no mind on the other end. So the child is learning the forms of social interaction — turn-taking, question and response — but the substrate is hollow.
It's like practicing tennis against a ball machine. You learn the mechanics, but you never learn to read an opponent.
We don't know what that does to a developing mirror neuron system. Does the brain treat the AI's simulated emotion as real and wire accordingly? Or does it detect something off — some uncanny valley in the interaction — and fail to develop the full social circuitry? We have no data. But the feral child cases tell us that the brain will use whatever input it gets, and the results may not be reversible.
I keep coming back to something Marcos Rodríguez Pantoja said — that he's happier with animals than with people. He got real social bonding, just with the wrong species. An AI companion doesn't even offer that. It offers the shape of bonding without the substance. And a developing brain might not know the difference until it's too late.
That's the open question I think this whole topic leaves us with. We've spent a century learning what happens when human contact is absent. We're about to run the next experiment, unintentionally, on a massive scale — with screens and AI companions filling more and more of the interaction space for young children. The feral child cases are a warning about what happens at the extreme end of zero human input. But we don't know what the dose-response curve looks like for partial substitution.
The final thing I take from these cases isn't about wolves or language or critical periods. It's about how fragile the whole thing is. Everything we think of as human — language, empathy, the ability to understand that someone else has a mind different from yours — none of it comes for free. It's built, piece by piece, through the attention of other people. And if that attention isn't there, or if it comes from the wrong source, the construction goes differently.
These children didn't become less than human. They became human in a different way — adapted to the world they actually lived in, not the world we think they should have had. The tragedy isn't that they're broken. It's that their adaptation, which was brilliant in its original context, doesn't fit the context we brought them into.
That's what makes the grainy YouTube footage so hard to look away from. You're not watching a child who failed to become human. You're watching a child who succeeded at becoming something else. And the success is what's devastating.
Now: Hilbert's daily fun fact.
Hilbert: In the early 1900s, a Labrador-based chemist analyzing the slime mould Physarum polycephalum discovered it secretes a calcium-dependent signalling molecule nearly identical to the one used by human
