Imitating is one of the most powerful and widespread strategies in biology, shaping everything from how children learn language to how orchids trick bees into pollinating them. It sounds simple, but the science behind copying another’s actions, appearance, or sounds turns out to be surprisingly layered. Whether newborn humans actually imitate facial expressions remains hotly debated, the brain circuitry involved is more complex than early theories suggested, and the reach of imitation extends far beyond primates into reptiles, insects, and even artificial intelligence systems that learn by watching demonstrations.
The Newborn Imitation Debate
For decades, one of the most cited claims in developmental psychology was that newborns can imitate facial expressions within hours of birth. A landmark 1982 study reported that babies averaging just 36 hours old appeared to match happy, sad, and surprised faces posed by a live model, with observers able to guess the model’s expression simply by watching the baby’s face.1PubMed. Discrimination and imitation of facial expression by neonates This finding was enormously influential. It became a cornerstone of the idea that humans are born with an innate system for understanding other people’s actions.
That cornerstone has cracked. A comprehensive longitudinal study tested whether infants could imitate any of nine different social gestures, including mouth opening, tongue protrusion, happy and sad facial expressions, and various sounds. The results were striking: the researchers found no evidence for imitation of any of the nine gestures. For some, babies were equally likely to produce the gesture regardless of what the model was doing. For others, the babies’ behavior changed over time but wasn’t predicted by what they were watching.2Current Biology. Comprehensive Longitudinal Study Challenges Loci of Neonatal Imitation Even tongue protrusion, the single gesture most associated with newborn imitation in earlier studies, didn’t hold up when tested against proper controls.
The debate isn’t fully settled, but the weight of evidence has shifted. A 2021 review in a leading cognitive science journal noted that neonatal imitation is still widely accepted as fact and cited as evidence of an inborn mirror neuron system, even though powerful longitudinal data and new analyses have reinvigorated the possibility that newborns do not actually imitate at all.3PubMed. Do Newborns Have the Ability to Imitate? What earlier researchers may have been observing was something more reflexive or exploratory, not true matching of another person’s actions.
How Children Learn by Copying
Even if newborn imitation is questionable, there’s no doubt that older infants and young children become extraordinary imitators. By the end of the first year, babies reliably copy actions they see adults perform. By age three or four, something even more curious emerges: children begin to overimitate. They faithfully reproduce steps in a task that are clearly unnecessary, copying pointless actions an adult performs even when those steps have nothing to do with achieving a goal.
This behavior initially puzzled researchers. It seemed irrational for a child to copy a useless step when a shortcut was available. The assumption was that overimitation was just a social nicety, kids going along with what an adult did to be polite or to maintain a social connection. But research has shown it runs deeper than that. Children overimitate irrelevant actions that even chimpanzees, who are excellent tool users, simply ignore.4PubMed Central. The hidden structure of overimitation Chimpanzees skip the unnecessary steps and go straight for the reward. Children don’t, and they continue to overimitate even when explicitly told the extra steps aren’t needed.
The leading interpretation is that overimitation serves cultural learning. Children aren’t just trying to get a treat out of a box; they’re absorbing how things are done. In a world where much of human technology and practice is opaque to a naive observer, copying everything an expert does, including parts you don’t yet understand, is a reasonable default strategy. It prioritizes fidelity over efficiency, which turns out to be critical for passing complex knowledge from one generation to the next.
High-Fidelity Copying and Cumulative Culture
This preference for faithful reproduction connects to a broader point about why humans have culture in a way that other species largely don’t. Modeling work has shown that even small improvements in how accurately information gets passed between individuals can be the difference between cultural stagnation and the “ratchet effect,” where each generation builds on what came before rather than starting from scratch.5PubMed Central. Transmission fidelity is the key to the build-up of cumulative culture The model didn’t require dramatic changes in brain architecture. It suggested that high-fidelity transmission, the kind that overimitating children are especially good at, is the key driver of cumulative culture.
This framing helps explain why human imitation feels different from what we see in other animals. Many species learn socially, but the precision and scope of human copying, especially our willingness to reproduce actions we don’t understand, creates a ratchet that allows technologies and practices to grow more complex over time. A child imitating a cooking technique doesn’t need to understand why each step matters. The faithful copy preserves the information until the child is old enough to figure that out.
The Brain Circuitry Behind Imitation
Understanding how imitation works in the brain has been a major focus of neuroscience since the 1990s, when mirror neurons were first discovered in macaque monkeys. These are neurons that fire both when an animal performs an action and when it watches someone else perform the same action. In humans, the brain network involved in imitation includes regions in the frontal and parietal lobes, along with the superior temporal sulcus, a strip of cortex involved in perceiving biological motion.6PubMed. Neural mechanisms of imitation
But where does this system come from? Two competing models have offered starkly different answers. One, the active intermodal-mapping (AIM) model, proposed that the ability to match observed actions to one’s own body is innate, wired in from birth. The rival view, known as associative sequence learning (ASL), argues that the connections between seeing and doing are learned through experience, particularly through the rich social environments human infants grow up in.7PubMed Central. Homo imitans? Seven reasons why imitation couldn’t possibly be associative Experimental work using automated measurement of how well people can copy facial expressions without visual feedback has supported the learning-based ASL model. People don’t get better at imitating facial expressions without being able to see themselves doing it, which is what you’d expect if the skill depends on learned visual-motor associations rather than an innate matching system.8PubMed. Facial self-imitation: objective measurement reveals no improvement without visual feedback
This matters because the “innate mirror system” narrative has been enormously influential, not just in basic science but in clinical theories about conditions like autism. If the mirror system is largely learned rather than hardwired, the explanations for why some people struggle with imitation may need to be rethought entirely.
Autism and the “Broken Mirror” Idea
One of the most widely publicized clinical applications of mirror neuron theory was the “broken mirror hypothesis” of autism. The idea was straightforward: if mirror neurons drive our ability to understand and imitate others, then a dysfunction in this system could explain the social and communicative difficulties associated with autism. It was a tidy explanation, and it attracted enormous media attention.
The evidence, however, has not been kind to the hypothesis. A systematic review of all published neuroscience studies examining mirror system integrity in autism, using methods ranging from brain imaging to eye tracking, found that the data were very mixed. Studies using emotional stimuli sometimes found group differences, but studies using non-emotional hand actions did not. Overall, there was little evidence for a global dysfunction of the mirror system in autism.9PubMed Central. Reflecting on the mirror neuron system in autism: a systematic review of current theories A later review confirmed that research has produced insufficient evidence to support the broken mirror hypothesis in its original form.10PubMed Central. Continuing to look in the mirror: A review of neuroscientific evidence for the broken mirror hypothesis, EP-M model and STORM model of autism spectrum conditions
Alternative models now suggest that what looks like an imitation problem in autistic individuals may actually reflect differences in how social context shapes responses, not a fundamental inability to mirror actions. The distinction matters for how therapies are designed and for how we understand what is actually going on.
Vocal Imitation and the Roots of Language
One specialized form of imitation, the ability to reproduce sounds heard from others, sits at the foundation of spoken language. Vocal learning is surprisingly rare in the animal kingdom. Among mammals, it appears in humans, some cetaceans, and certain bats. Among birds, only three groups have it: songbirds, parrots, and hummingbirds.11PubMed Central. The constitutive differential transcriptome of a brain circuit for vocal learning Research into the song system of birds has identified thousands of genes that are differentially active in the brain regions responsible for vocal learning, offering clues about the molecular machinery that makes this kind of imitation possible.
The evolutionary story of how imitation led to language in humans extends beyond just copying sounds. One framework proposes that a capacity for practiced, rehearsed skill imitation provided early hominins with survival benefits, and the same abilities later gave rise to “mimetic” expression: pantomime, ritual, and eventually proto-words. Spoken language, in this view, evolved out of vocal mimicry but only long after more basic imitative abilities had already transformed how humans coordinated group activities.12PubMed. Key cognitive preconditions for the evolution of language Language, then, is a late product of a much older imitative capacity.
Social Learning in Animals Far Beyond Primates
One common misconception about imitation is that it’s largely a primate affair. The reality is that social learning through imitation appears across a remarkable range of species, including some that share very little recent evolutionary history with us.
Bearded dragons, a common pet reptile, demonstrated genuine imitation in a controlled experiment. When one lizard watched another open a trap door by pushing it to a specific side, the observer reliably opened the door to the same side. A control group that saw the door move on its own, without a demonstrating animal, couldn’t solve the task. The distinction rules out simpler explanations like just being attracted to movement near the door.13PubMed. Social learning by imitation in a reptile (Pogona vitticeps) The researchers argued that reptiles possess cognitive abilities comparable to those observed in mammals and birds, suggesting imitation relies on ancient neural mechanisms rather than recent evolutionary innovations.
Bumblebees have pushed the boundary even further. In experiments with puzzle boxes that could be opened in two different ways, bees that observed a trained demonstrator acquired the demonstrated technique and stuck with it, even after discovering the alternative method on their own.14PubMed Central. Bumblebees acquire alternative puzzle-box solutions via social learning In a separate study, about 60 percent of bees that watched a skilled demonstrator pull a string to access a hidden flower managed to solve the task on their first attempt. Bees that didn’t get to watch a demonstration mostly failed, trying instead to reach the reward from above the apparatus.15PLOS Biology. Associative Mechanisms Allow for Social Learning and Cultural Transmission of String Pulling in an Insect These findings don’t mean bees have the same internal experience of learning that humans do, but they challenge the assumption that copying complex, non-natural behaviors requires a large brain.
Mimicry in the Wild
Outside the laboratory, some of the most spectacular examples of imitation are found in the natural world, where mimicry has evolved as a survival strategy. The term is used differently here than in cognitive science. When biologists talk about mimicry, they usually mean evolved resemblance between one organism and something else in its environment. The organism isn’t consciously choosing to copy; natural selection has shaped its body, behavior, or chemistry to resemble another species.
Octopuses are a fascinating exception because they appear to make real-time decisions about what to imitate. The Indo-Malayan mimic octopus displays a repertoire of postures and body patterns that clearly impersonate venomous animals sharing its habitat, and observations suggest it selects the most appropriate form of mimicry based on the nature of the perceived threat.16PubMed Central. Dynamic mimicry in an Indo-Malayan octopus In the Caribbean, a different species, Macrotritopus defilippi, was filmed mimicking the swimming posture, speed, and coloration of a common flounder. It deployed this disguise only while swimming across open sand, when its movement would have given away its usual camouflage.17PubMed. A “Mimic Octopus” in the Atlantic: Flatfish mimicry and camouflage by Macrotritopus defilippi Studies of related species off North Sulawesi, Indonesia, confirmed the same pattern: octopuses switched from camouflage to active flounder mimicry only when motion made hiding impossible, using it as a primary defense against visual predators.18Biological Journal of the Linnean Society. Mimicry and foraging behaviour of two tropical sand-flat octopus species off North Sulawesi, Indonesia
Brood parasites offer yet another angle. The screaming cowbird, which lays its eggs in the nests of baywing birds, has evolved fledgling-stage mimicry so precise that its young are visually and vocally indistinguishable from baywing chicks. This matters because baywing parents actively reject foreign young after they leave the nest. Shiny cowbirds, a related but less specialized parasite, suffered post-fledging mortality rates of about 83 percent due to host rejection, while screaming cowbirds experienced virtually none.19PubMed Central. Host-parasite coevolution beyond the nestling stage? Mimicry of host fledglings by the specialist screaming cowbird The imitation of the host’s appearance and begging calls is entirely genetic, shaped over evolutionary time rather than learned.
Chemical Imposters in the Plant Kingdom
Plants can’t move or vocalize, but some have evolved chemical mimicry so refined that it manipulates animal behavior at the molecular level. Sexually deceptive orchids in the genus Ophrys produce no nectar or other reward for pollinators. Instead, they synthesize volatile chemicals that mimic the sex pheromones of virgin female insects, luring males into attempting to mate with the flower. During these mating attempts, pollen gets transferred.20PubMed Central. Pollinator attraction in a sexually deceptive orchid by means of unconventional chemicals The deception goes beyond smell. Some sexually deceptive orchids combine specialized chemical volatiles with tactile structures that feel enough like an insect body to sustain the male’s attention.21Current Biology. Pollination by Sexual Deception in Caladenia plicata Relies on a Dual Volatile System Synthesized via Biosynthetically Distinct Pathways The specificity is extreme: each orchid species tends to target a single pollinator species, and the chemical mimicry is precise enough that closely related insect species are not deceived. Even the flowers’ color may play a role, potentially functioning as a visual lure that reinforces the chemical signal.22PubMed Central. Why sexually deceptive orchids have colored flowers
Machines That Learn by Watching
The same principle of learning through imitation has been adapted for artificial intelligence and robotics, though the underlying mechanisms are entirely different. In the AI world, “behavior cloning” is a supervised learning approach where a robot or software agent learns a skill by directly imitating expert demonstrations.23arXiv. Constrained Behavior Cloning for Robotic Learning You show the system how a task is done, and it tries to reproduce the observed behavior. This approach can lead to proficient performance with relatively limited training data, making it practical for tasks where programming explicit rules would be impossibly complex.24arXiv. Integrating Behavior Cloning and Reinforcement Learning for Improved Performance in Dense and Sparse Reward Environments
A related approach called inverse reinforcement learning takes the idea a step further: rather than just copying the actions an expert takes, the system tries to infer the underlying goals or reward signals driving the expert’s behavior.25Artificial Intelligence. A survey of inverse reinforcement learning: Challenges, methods and progress Where behavior cloning asks “what did the expert do?”, inverse reinforcement learning asks “what was the expert trying to achieve?” This parallels, at least loosely, the distinction between a child who copies every step of a task (overimitation) and one who understands the goal well enough to improvise. Recent work combining behavior cloning with other reinforcement learning methods has improved robot success rates by roughly 30 to 40 percent in both simulated and real-world tasks compared to standard imitation-only approaches, particularly for complex, multi-step operations.26arXiv. Constrained Behavior Cloning for Robotic Learning
The parallel between biological and machine imitation is real but limited. A child copying a parent’s cooking technique and a robot copying a human demonstrator’s arm movements share a surface-level structure: observe, encode, reproduce. But the child is embedded in a social world that gives the copied behavior meaning, while the robot is optimizing a mathematical function. The convergence on imitation as a learning strategy, across domains that have nothing else in common, says something about how useful copying turns out to be when you need to acquire complex behavior efficiently.
When Imitation Gets Weaponized
The same capacity for faithful reproduction that drives cultural learning and social bonding has a darker side when applied to digital media. AI systems trained on human faces and voices can now generate synthetic video and audio that convincingly imitates real people. Research testing how well humans detect these fakes found that AI detection models consistently outperformed human judges. In one study, 110 participants evaluated a mix of 20 real and 20 fake videos, and every AI model tested beat human accuracy on the same set of clips.27arXiv. Unmasking Illusions: Understanding Human Perception of Audiovisual Deepfakes Our perceptual systems evolved to interpret faces and voices in real time, not to catch pixel-level artifacts or subtle audio inconsistencies. The irony is that the same human sensitivity to social cues that makes imitation so powerful for learning also makes us vulnerable to convincing fakes, because we tend to trust faces and voices that look and sound familiar.