Imitation is a cognitive skill that human beings seem to master at remarkable ease. The common notion is that imitation is an important mechanism in social learning. The Oxford dictionary names ‘a child learns to speak by imitation’ as an example. We use imitation for learning motor skills and for understanding the actions and mental states of others. Heyes defines imitation in the context of psychology and neuroscience in the following way: “a topographic resemblance between the behavior of the copier (or ‘observer’ or ‘self’) and the agent who is copied (‘model’, ‘other’); where the parts of the observer’s body move in the same way relative to one another as the parts of the model’s body.”
The imitative behavior of newborn human infants is proposed to play a crucial role in understanding that others are ‘like me’, the development of the theory of mind and empathy. Imitation learning is also necessary for the “development of culture-specific skills, such as communicative gestures or ritualistic body movements, and promotes cooperative social interaction.”
But how does imitation work exactly? When copying a certain body movement of another person you’re faced with the following challenge, called the ‘self-other correspondence problem’: You can see how the other person moves, but you cannot see yourself moving from this outside perspective. However, you can feel your own body moving, but not the body of the other person. How does the imitator know what pattern of motor activation will make their action look like that one of the model?
Considering the cognitive challenges of successful imitation and the fact that non-human species rarely imitate this poses the question of the development of this skill. There are two main directions in developmental psychology: Either (1) humans are highly skilled imitators, because of a special and inborn mechanism or (2) imitative behavior of infants is learned and emerges together with the development of motor, cognitive and social skills in a given sociocultural environment.
Facial Imitation in Newborns
In 1977, Meltzoff and Moore proposed that facial imitation of human infants might be inborn. They showed that infants between 12 and 21 days of age matched three facial gestures of the experimenter: tongue protrusion, mouth opening and lip protrusion. The authors used these findings to argue for an inborn mechanism and against a learned response, because at this age the actions of the infant could not have been reinforced by others yet.
They describe this mechanism in the Active Intermodal Matching Model. This model claims that imitation is a matching-to-target process, where the perceived and produced action are “coded within a common supramodal framework”. This means, that when an action is observed with the intention to imitate, it’s visual representation is converted into a supramodal representation, which contains information about how the movement is executed. The active nature of this model is captured by a proprioceptive feedback loop.
While this model doesn’t specify how the information is coded, it implies, however, that human infants “inherit considerable knowledge about their own bodies and action capabilities and how those map onto the bodies and actions of others”. Meltzoff (1988) speaks of the ‘Homo imitans’, being equipped with this genetically inherited intermodal matching mechanism that makes us highly skilled imitators and sets us apart from other animals. This ability is also often believed to be the basis for other cognitive abilities, for example empathy or language.
There seems to be widespread agreement that infants match adult’s behavior. Nevertheless, critics of the Active Intermodal Matching Model pose the question whether the observed behavior is really imitative or has another explanation. For example, the matching of tongue protruding, a common behavior in newborns, may just be a by-product of arousal instead of indented imitation.
The idea of similar neural processing between perception and action was fueled by the discovery of mirror neurons in rhesus macaque monkeys by Rizzolatti and colleagues (1996). In single-cell recordings the authors showed that neurons in the F5 premotor area where active during the observation of a goal-oriented action of others. Mirror neurons seem to have sensory and motor properties, which makes them a strong candidate for supporting the claim that imitation is an automatic and innate mechanism. Admittedly thus far, they have only been observed in adult monkeys and for meaningful, goal-directed actions, such as reaching for food. Also, it’s been shown that activity of mirror neurons correlates with action understanding.
Even if these mirror neurons haven’t been directly observed in humans, this led to the idea of a human mirror system, which captures the belief that we comprehend actions of others by activating the same neural structures that would be engaged if we were to produce the action ourselves. The main brain areas involved in this human mirror system are the premotor cortex, the supplementary motor areas, the primary somatosensory cortex and inferior parietal cortex. This mirror system has also been argued to not only be involved in imitation, but also in learning new skills, understanding emotions of others, empathy, theory of mind and language development. However, there’s weak experimental evidence for the abovementioned functions.
Moreover, the extent of activation in the human mirror system has been shown to reflect the motor repertoire of the individual itself. The more expertise we have in performing an action, the stronger the brain activation in the mirror system when observing this action. It seems like sensorimotor associations of related stimulus-response pairs get stronger as they repeatedly occur together. That’s why Campbell and Cunnington (2017) propose a dynamical top-down modulation of the human mirror system, via frontal control networks, that is meeting changing task-demands.
In conclusion, if that’s how the human mirror system operates, then it is highly unlikely to be involved in behavioral matching in newborns, unless some human mirror neurons would be inherited and pre-programmed without the benefit of experience. Infants have no pre-natal opportunity to observe actions of others, and their visual system shortly after birth is highly immature. Additionally, there’s lacking evidence of action understanding in human newborns so far. Note, however, that this doesn’t mean that imitation isn’t an important learning mechanism for human infants, it’s just unlikely to be an inborn ability present in newborns.
While the question whether mirror neurons are present at birth remains unsolved, there is, however, empirical evidence that mirror neurons for tool use can develop during ontogeny. This suggests that learning plays a crucial role. Generalist theories suggest that mirror neurons acquire their properties in the course of ontogeny and that therefore the capacity to imitate is a product of general associative learning processes.
The Associative Sequence Learning model states that each action has two representations: One encoding visual information and one encoding somatosensory and motor information. These two become linked as the result of being frequently experienced close together in time (contiguity) and are in a predictive relationship with one another (contingency). On a neural level, synaptic associations are formed through Hebbian learning. These direct links between visual and motor information make imitation possible. The brain areas that are active during imitation also show activation for passive action observation. This speaks for a close neural coupling between the two and against a specialized module for imitation.
Reid and colleagues (2017) showed in a recent study that the link between action perception and action production is already established in 10-weeks old infants. Using ERP recording, they showed brain activity in parietal regions in response to the perception of biological motion, but only for the group of infants that experienced ‘reflex walking’ shortly before. Importantly, this suggests that experience refines perception and that action perception and action production are already closely linked at this age.
Heyes (2015) argues strongly for this model and published a review article answering to seven common objections to the Associative Sequence Learning model. She presents arguments that while newborns do not imitate, young infants receive plentiful experience to learn to imitate, making imitative behavior be observable from a very young age on. For example, individual differences in associative learning ability at 1 month predict imitative performance at 9 months. Further, she argues that neither infants nor adults can imitate novel actions very well, imitation is goal-directed and improvement in imitation depends on visual feedback. All evidence for association learning, rather than for an innate mechanism.
Finally, a recently published comprehensive longitudinal study by Oostenbroek and collegues (2016) failed to provide evidence for newborn imitation. They presented a total number of 106 infants at 1, 3, 6 and 9 weeks of age with social and non-social models. The results indicated that infants at this age where just as likely to produce gestures in response to the control model than to matching model. Consequently, this speaks against an innate module for imitation and for associative learning.
So where do we stand?
Since humans are “hyper-social animals”, the prevalent view since the 1970s in the field was that an inborn ability to imitate forms “the foundation for the development of cognitive skills needed for cooperation, such as empathy, mind reading and language” on a psychological and neurobiological level. The Active Intermodal Matching Model was proposed by Meltzoff and Moore (1997), who observed behavioral matching of facial gestures in newborns. While many studies have failed to replicate Meltzoff and Moore’s findings, one has to keep in mind that infant research is very difficult, sample sizes are small and the interpretation of the results often challenging. A prominent example is the behavioral matching of tongue protrusion in newborns, which has been used to argue for an inborn imitation mechanism. However, it has been shown that newborns protrude their tongue to a number of arousing stimuli. This means, when you stick out your tongue in front of a newborn, it is very likely, that the newborn is also sticking out it’s tongue, but only because of arousal and not because it is imitating your action.
When mirror neurons were discovered in 1996 in monkeys, this encouraged researchers in cognitive neuroscience to focus on this perception-action interaction and imitation in humans. Together with the observation of behavioral matching in newborns this led to the idea of an innate mirror system in humans. What was often overlook however, is that the human mirror system seems to rely on experience, and is probably the by-product of learned associations. The Associative Sequence Learning Model captures this idea of learned associations, proposing vertical associations between perceptual and motor information that develop over ontogeny. In this model the ‘self-other correspondence problem’ of imitation is solved by general mechanisms of associative learning and action control, rather than an inborn specialist mechanism for imitation. Recent behavioral and neurophysiological evidence supports this idea.
Heyes (2016) poses the question why both views could coexist for such a long time. One reason is probably that a lot of the existing data is either inconsistent or open to interpretation. Furthermore, when there is presumable evidence that a neurocognitive mechanism is inborn, research on the development of this skill is often neglected. Therefore, it’s crucial to investigate the development of the human mirror system and imitation skills over ontogeny, in order to get a better understanding how learning to imitate forms the foundation of so many fundamental cognitive skills like social learning, language or socialization.
What is also extremely interesting is the way that Heyes (2016) writes about the progress in the field, to an understanding that imitation is not an innate skill. She explains that humans are often believed to be “‘special’ because we genetically inherit a set of complex cognitive mechanisms, dedicated to functions such as language, mental time travel, cheater detection, face recognition, and theory of mind “. Now that it was shown that imitation, which is believed to be the foundation of many of these skills, is not inborn, this poses the question why we develop so differently from other animals. Maybe, when we are born, our minds are not so different from other animals, but we’re really sensitive to social stimuli and have a high capacity for learning. And as our minds get more mature, we develop those more complex cognitive processes through social interaction.