The ability to understand the actions of others is crucial for all social species. Recent electrophysiological and neuroimaging studies indicate that understanding the actions of others is mediated by the mirror neuron system (MNS), in which an observed action is mapped onto the observer's own motor representation of that action. Although there has been considerable progress in elucidating the mechanisms and functions of the direct-matching process, we still know little about its developmental aspects. In this article, we first provide a brief overview of the functions of the direct-matching process in the MNS. Next, we review the neurophysiological and behavioral evidence for the developmental aspects of the direct-matching process, indicating that it is already functional at least by the age of 6 months, and that perception and action are mutually influenced and directly linked in early infancy. Finally, we discuss the unresolved questions about the onset of the direct-matching process and suggest directions for future research.
Why and how do we understand the actions of others even though we are different individuals? Recently, research has identified the neurons that suggest the mechanism for action understanding (Gallese, Fadiga, Fogassi, & Rizzolatti, 1996; Rizzolatti, Fadiga, Gallase, & Fogassi, 1996). They are called “mirror neurons” (the mirror neuron system (MNS) in humans). The discovery of mirror neurons has provided an opportunity to gain insight into the neural basis of action understanding. Mirror neurons have the property of responding both when an individual executes an action and the individual observes the same action in others, indicating that the perception and execution of actions share a common neural representation. Thus, the MNS is thought to mediate action understanding.
In this article, we provide a brief overview of the properties of the MNS, derived from electrophysiological and neuroimaging studies, and explain the functions of the direct-matching process as a mechanism of action understanding. Next, we review the neurophysiological and behavioral evidence about the developmental aspect of the direct-matching process, suggesting the link between perception and action in early infancy. Finally, we discuss some unanswered questions about the onset of the direct-matching process and suggest future directions for research in this area.
The mirror neuron system and the direct-matching hypothesis
A number of studies using different methods have shown that a similar system appears to exist in humans (Grafton, Arbib, Fadiga, & Rizzolatti, 1996; Iacoboni et al., 1999; Nishitani & Hari, 2000). As a result, the phenomenon attributable to the MNS has lead to the direct-matching hypothesis, in which it is suggested that our understanding of others' actions derives from a direct-matching process, by which an observed action is mapped onto the observer's own motor representation3 of that action (Rizzolatti & Craighero, 2004; Rizzolatti, Fogassi, & Gallese, 2001). According to the direct-matching hypothesis, the action of another is understood when its observation causes the motor system of the observer to resonate (Rizzolatti et al., 2001). The mapping of the observed action onto the observer's own motor system is direct and automatic, and does not involve sophisticated perceptual analysis. This automatically induced motor representation of the observed action corresponds to what is spontaneously generated during active action, and the outcome of which is known to the acting individual. Therefore, the observer can understand another's action because they know the outcome of the action when they do it (Gallese et al., 2004). The schema of the direct-matching hypothesis is illustrated in Figure 1.
Investigating the development of the direct-matching process is potentially useful for understanding the functions of the direct-matching process, because, as Southgate, Johnson, Osborne, and Csibra (2009) argue, the limited but developing motor repertoire of infants has the potential to shed light on which capacities are modulated by motor-system recruitment. Thus, understanding how the direct-matching process develops is a challenging question, not only for developmental psychology, but also for neuroscience. Furthermore, if the direct-matching process plays an important role in social cognition (Gallese, 2003), understanding how the direct-matching process develops is also an important question for understanding the social development of children.
Neurophysiological evidence for the functioning MNS
The suppression of mu rhythm4 is considered to be an index of motor activation, and its presence during action observation is considered to be a probable index of MNS activity (Hari, Forss, Avikainen, Kirveskari, Salenius, & Rizzolatti, 1998). A recent electroencephalography (EEG) study of children aged between 52 and 133 months showed that mu rhythm suppression occurs during both the execution and the observation of hand actions (Lepage & Théoret, 2006).
In a much younger population, Nyström (2008) reported that 6-month-old infants showed significantly higher event-related potential (ERP) activation, similar to that seen in adults, when they viewed others' goal-directed actions (e.g. reaching for an object), although they failed to find mu rhythm suppression in 6-month-old infants. Also, Southgate et al. (2009) showed that 9-month-old infants exhibited mu rhythm suppression during observation of others' grasping actions that directly matched the neural signal that occurred during their own actions. In addition, there is a neurophysiological study of infants using near-infrared spectroscopy (NIRS). Shimada and Hiraki (2006) demonstrated that, as in adults, the motor area of 6- to 7-month-old infants was activated both when they observed a live person manipulating a toy and when the infants themselves performed an action. Although they also found activation in the same area during action observation in the TV setting, there is a significant difference in activity between the observation of another's action and object movement only in the live setting.
Furthermore, a study of 14- to 16-month-old infants demonstrated that suppression of the mu rhythm in infants during action observation was directly related to the amount of their own action experience (van Elk, van Schie, Hunnius, Vesper, & Bekkering, 2008). In this study, the infants, who had more experience with crawling than with walking, showed a stronger suppression during observation of crawling than of walking, and the degree of suppression was correlated to the their own crawling experience.
Taken together, these studies suggest that the MNS in infants is functional in early infancy, indicating that the direct-matching process is already functional at least as early as 6 months old. Thus, these results suggest that there is some possibility that perception and execution of action are directly linked in early infancy.
The mutual influence between perception and action in infancy
The existence of the MNS leads to speculation that the perception and execution of an action may be mutually influenced. Indeed, a number of adult behavioral studies indicate that the execution of actions affects action perception (Casile & Giese, 2006). Conversely, observing the actions of others influences the execution of one's own actions (Kilner, Paulignan, & Blakemore, 2003). In addition, it has been discovered that even neonates spontaneously imitate the actions of others (Meltzoff & Moore, 1977). These findings raise the possibility that action perception and execution might be tightly interconnected from very early in development.
To date, several behavioral studies have showed that action perception and execution are mutually influenced in early infancy. With regard to studies investigating the effect of action execution on action perception in infancy, Hauf, Aschersleben, and Prinz (2007) demonstrated that action experience influenced the perception of actions performed by others. In their experiments, after having played with an object, infants watched two movies presenting adults who acted on either the same object or a different object. The 9- and 11-month-old infants looked longer at the same-object movie than the different-object movie, but not in watching objects or persons per se. Furthermore, Sommerville, Woodward, and Needham (2005) demonstrated that action execution has causal effects on 3-month-old infants' perception of others' actions. In their study, these pre-reaching infants, who had experienced training with Velcro mittens,5 enabling them to experience the reaching action directly, showed sensitivity to the actor's goal. By contrast, infants who observed the actions of others before the mittens task showed no such effect.
In another study demonstrating the effect of action perception on action execution, Longo and Bertenthal (2006) showed that 9-month-old infants' actions were influenced by the observation of others' actions. The infants made perseverative reaching (A-not-B) errors6 after they observed active search by another person, as well as after they performed active search at one location.
Taken together, these studies have shown that action perception and action execution have mutual influence, indicating a bidirectional relation. These results suggest not only that perception and execution of action are directly linked in infancy, but also that there is a bidirectional causal effect of perception and action in infancy.
The link between perception and action in infancy
The neurophysiological and behavioral studies mentioned above indicate the possibility that perception and execution of action are directly linked in early infancy. Indeed, recent studies have shown that there is a developmental link between perception and action in early infancy (Reid, Belsky, & Johnson, 2005; Sommerville & Woodward, 2005). For example, Reid et al. (2005) demonstrated that 8-month-old infants who had a higher general motor ability discriminated others' anatomically possible and impossible grasping actions better than infants who had a lower general motor ability. Also, a study of 10-month-old infants (Sommerville & Woodward, 2005) reported that individual difference in the ability to attribute the goal to the means-end behavior and the frequency of action production were strongly correlated.
In addition, Sanefuji, Ohgami, and Hashiya (2008) reported that when infants observed another's actions in the form of point-light information, their preference was related to the type of locomotion that was relevant to them. Whereas 8-month-old infants who could crawl but not walk looked longer at the crawling actions than the walking actions, 12-month-old infants who could walk looked longer at the walking actions than the crawling actions.
Moreover, similar developmental links were found not only in the studies concerning action perception, but also in a study concerning action anticipation. Based on a study demonstrating that adults predict the goal of others' goal-directed actions (Flanagan & Johansson, 2003), Falck-Ytter, Gredebäck, and von Hofsten (2006) demonstrated that 12-month-old infants who had a motor representation of observed actions, made predictive eye movements to the goal when they observed manipulating hand actions, but that 6-month-olds, who did not have a motor representation of the observed actions, did not show any predictive eye movements.
In sum, these studies have shown that there is a developmental link between action perception and action execution in early infancy. However, these studies did not examine the direct link between perception of others' actions and one's own motor ability for the same action in infancy. For instance, Sommerville and Woodward (2005) showed a developmental link between the ability of action perception and the frequency of action production for the same action, and Reid et al. (2005) demonstrated the link between action perception and general motor ability. According to the direct-matching hypothesis (Rizzolatti et al., 2001; Rizzolatti & Craighero, 2004), the relationship between observed actions and the actions within the observer's motor representation has a one-to-one correspondence. Thus, to prove the direct-matching process in infancy, it is necessary to demonstrate the correlation between the development of action perception and action execution itself.
In the following section, we present evidence of the development of the direct-matching process by showing the direct link between the development of predictive eye movements for grasping action and the motor ability itself of the same action.
The direct link between action anticipation and motor ability in infancy
We investigated the direct link between action anticipation of others' actions and the motor ability itself for the same action in infancy (Kanakogi & Itakura, 2009). A previous study had demonstrated that 12-month-old infants, but not 6-month-old infants, showed proactive goal-directed eye movements when they observed manipulating hand actions (Falck-Ytter et al., 2006). The authors suggested that the 6-month-olds' lack of predictive eye movements might reflect their lack of a motor representation for the action. However, first, there is some possibility that 6-month-old infants do show such eye movements because the MNS functions in 6-month-old infants (Nyström, 2008; Shimada & Hiraki, 2006), and second, the previous study did not specifically test whether predictive eye movements were related to the infants' own motor ability for the same action. If the MNS mediates action anticipation (Kilner, Vargas, Duval, Blakemore, & Sirigu, 2004), then predictive eye movements must be found in younger infants when they observed anothers' actions for which they have a motor representation (such as grasping, which emerges at approximately 4–6 months of life), and the development of the predictive gaze must be related to the infants' motor ability itself for the same actions.
In our experiment, we presented movies of agents who reached toward one of two toys to 4- (n = 12), 6- (n = 12), 8- (n = 12), and 10-month-old infants (n = 12). The agents in the movies were grasp hand (GH condition), back of hand (BH condition) and mechanical claw (MC condition) (see Figure 2). The actions in each condition were goal-directed actions (GH condition), non-goal-directed actions (BH condition), and inanimate actions (MC condition). The GH condition was the experimental condition and the other conditions were control conditions because the MNS is activated when someone observes another's goal-directed actions, but not when someone observes another's non-goal-directed actions (Johnson-Frey et al., 2003; Muthukumaraswamy, Johnson, & McNair, 2004) or inanimate actions (Kilner et al., 2003; Tai, Scherfler, Brooks, Sawamoto, & Castiello, 2004).
After recording the eye movement using a gaze-recording technique while participants viewed the movies, participants performed a modified version of Rochat's (1992) grasping task, which measures the development of one-handed grasping, similar to the perceptual stimuli. We calculated the timing of gaze arrival at the goal relative to the arrival of agents' actions during viewing of the movies and the development of the one-handed grasping action.
A summary of our main results is illustrated in Figure 3. These results indicate that the development of predictive eye movements for the grasping action is strongly related to that of the infants' own motor ability for the same action. In each condition for 4-month-olds who could not grasp, we found no difference between the timing of gaze arrival to the goal. In contrast, in each condition for 6-, 8-, and 10-month-olds who could grasp, we found that the eye movements in the GH condition were significantly more predictive than in the other conditions, and that such eye movements of the GH condition were significantly more predictive than the time when the agents arrived at the goal. Furthermore, correlational analysis revealed that the development of predictive eye movements and the motor ability of the one-handed grasping action were significantly related only in the GH condition, even when controlling for age.
Our findings suggest that: (a) predictive eye movements are found at a younger age than had been demonstrated by previous research; (b) the development of action anticipation is synchronized with the onset of infants' own motor ability; and (c) there is a direct link between the anticipation of others' actions and infants' own motor abilities. These findings indicate the direct link between perception and action in early infancy, supporting the mirror neuron account of action understanding through the direct-matching process (Rizzolatti & Craighero, 2004).
Discussion and future research directions
We have explored empirical data investigating the development of the direct-matching process. Although there is not enough evidence to explain the development of the direct-matching process conclusively, researchers have been elucidating the developmental aspects of the direct-matching process gradually. From the findings of the neurophysiological and behavioral studies mentioned above, we conclude that: (a) the direct-matching process is already functional at least by the age of 6 months; and (b) perception and action are mutually influenced and directly linked in early infancy. These conclusions suggest additional questions about when and how the matching process develops.
First, although we concluded that the direct-matching process is already functional at least by the age of 6 months, our conclusion does not contradict the possibility that the direct-matching process might be functional in infants aged less than 6 months. As shown in the studies mentioned above, whether we could reveal the direct-matching process depended heavily on the limited motor repertoire of infants at a given age. Thus, by using an index of behavior that infants can perform at a given age, we might find the direct-matching process in infants aged less than 6 months.
Second, if perception and action are mutually influenced and directly linked in early infancy, the causal effect between perception and action is bidirectional. Thus, both perceiving others' actions and executing actions by oneself are thought to be factors that facilitate the development of the direct-matching process.
Our suggestions are speculative, but they shed light on two unresolved questions about the onset of the direct-matching process. One question is whether the matching process is innate or is a learning process. The other question is what mechanism allows the direct-matching process to become functional. As follows, we explain the two unresolved questions about the onset of the direct-matching process and suggest directions for future research.
One of the important ways to answer the question is to investigate whether or not the direct-matching process is functional in newborns by using both neurophysiological and behavioral indexes. However, no one has investigated these. Until evidence is provided by measuring the brain activation and behavior of newborns during neonate imitation, this controversial discussion will remain unresolved and quite speculative.
With regard to the second question, there has been only one developmental study about what makes the direct-matching process functional. Sommerville et al. (2005) showed that action experience facilitated perception of others' actions, but not vice-versa, in 3-month-old infants. This result indicates the impact of action production on action perception, but does not provide evidence for the impact of action perception on action production. However, considering the bidirectional influence and direct link between perception and action in early infancy, as we discussed above, the perception experience should be a factor that makes the direct-matching process functional. Thus, it is possible that the authors simply could not discover the influence of the perception experience on action in their particular experiment.
By using a device to measure the behavioral change in detail (e.g. an eye tracking system) or employing another behavioral index to detect its change easily (e.g. eye movements) as a dependent variable, we might find out the impact of the perception experience on action production and demonstrate the bidirectional causal effects between perception and action.
Empirical evidence about the development of the direct-matching process is now gradually accumulating and additional research is about to begin. In the near future, if the questions mentioned above were answered, that would shed new light on the mechanisms and functions of the direct-matching process. Further research is particularly needed to investigate the onset of the direct-matching process.
This research was supported by grants to Shoji Itakura from Japan Society for the Promotion of Science (JSPS) (No: 01500004) and Nissan Science Foundation.
We thank Yuko Okumura and Tomoyo Morita for helpful comments on a previous version of this manuscript.
In this article, motor representation is defined as an internal description of a motor response.
The suppression of mu rhythm means attenuation of the resting-state sensorimotor alpha rhythm.
Velcro mittens are sticky and allow the infants to artificially grasp small objects.
A-not-B error means the formation of a prepotent response created by repetitive searching at one location.