Experience-based priming of body parts: A study of action imitation

Abstract

Two important dimensions of action are the movement and the body part with which the movement is effected. Experiment 1 tested whether automatic imitation is sensitive to the body part dimension of action. We found that hand and foot movements were selectively primed by observation of a corresponding, task-irrelevant effector in motion. Experiment 2 used this body part priming effect to investigate the role of sensorimotor learning in the development of imitation. The results showed that incompatible training, in which observation of hand movements was paired with execution of foot movements and vice versa, led to a greater reduction in body part priming than compatible training, in which subjects experienced typical contingencies between observation and execution of hand and foot movements. These findings are consistent with the assumption that overt behavioral imitation is mediated by the mirror neuron system, which is somatotopically organized. Our results also support the hypothesis that the development of imitation and the mirror neuron system are driven by correlated sensorimotor learning.

Introduction

One of the most important findings to emerge from action perception research has been that the observation and the execution of body movements activate a common cortical network. This network, known as the ‘mirror neuron system’, is active when movements are executed without visual feedback, and when the same actions are passively observed (e.g. Di Pellegrino et al., 1992, Iacoboni et al., 1999, Raos et al., 2006; for a review see Rizzolatti and Craighero, 2004). Neurons exhibiting ‘mirror’ properties, that is, a close correspondence between the visual input (observed action) and the motor output (performed action) in the neural response, are known to exist in ventral premotor area F5 (e.g. Di Pellegrino et al., 1992, Gallese et al., 1996) and inferior parietal lobule area 7b of the macaque (e.g. Fogassi et al., 1998, Gallese et al., 2002), and areas with similar characteristics have been identified in homologous regions in human premotor cortex, often centered on Broca's area in the inferior frontal gyrus, and parietal areas (e.g. Iacoboni et al., 1999). The present study is concerned with two questions: To what extent does the activity of the mirror neuron system parallel overt, behavioral imitation, and how do we acquire the capacity to map observed onto executed actions?

Overt behavioral imitation occurs when one individual, an ‘observer’, copies the body movement of another individual, a ‘model’. More specifically, observation of some dimension of the model's body movement (e.g. its rate or topography) causes the observer's behavior to become more like that of the model on the observed dimension (Heyes, 2001, Box 1). For example, when two people, A and B, are in conversation, each tends to imitate the incidental foot-shaking and face-rubbing behavior of the other; the frequency of foot-shaking by A increases when B engages in foot-shaking, but not when B engages in face-rubbing, whereas face-rubbing by A increases when B engages in face-rubbing, but not when B engages in foot-shaking (Chartrand and Bargh, 1999).

Superficially, it is plausible that imitation is mediated by the mirror neuron system. Imitation requires a neural mechanism that can map observed onto executed actions, and the mirror neuron system appears to fulfill that function. Furthermore, the hypothesis that imitation is mediated by the mirror neuron system is supported by evidence that the mirror neuron system, and particularly the inferior frontal gyrus, is more active during imitation than during either observation or execution of actions (e.g. Aziz-Zadeh et al., 2006, Iacoboni et al., 1999, Koski et al., 2003, Nishitani and Hari, 2000, Williams et al., 2006), and is involved in imitation learning (Buccino et al., 2004, Vogt et al., 2007). However, a strong connection between imitation and the mirror neuron system remains to be established. Mirror neurons have been found in monkeys, animals which are apparently incapable of imitation (Rizzolatti, 2005, Visalberghi and Fragaszy, 2001), and greater activation of the mirror neuron system during imitation than during observation alone or execution alone does not show unambiguously that the mirror neuron system mediates imitation. An imitation task involves action observation, action execution, and, critically, matching or translation of observed into executed actions. Therefore, this result may indicate, not that the mirror neuron system translates observed into executed actions, a distinctively imitative function, but that the effects of observation and execution on the mirror neuron system are additive.

A strong connection between imitation and the mirror neuron system, indicating that the former is mediated by the latter, would be established if it could be shown that a range of action variables have parallel effects on imitation and on the mirror neuron system. Two important variables or dimensions of action are effector and movement (Chaminade et al., 2005, Meltzoff and Moore, 1997). The effector dimension relates to the limb or body part used to perform an action, whereas the movement dimension relates to the topography or trajectory of the effector. For example, to wave at another person, we typically use a hand (effector), and a lateral, parabolic trajectory (movement). Research to date has indicated that movement variables have parallel effects on imitation and the mirror neuron system (e.g. Di Pellegrino et al., 1992, Ferrari et al., 2003, Craighero et al., 2002, Puce et al., 2000, Stürmer et al., 2000), but it is not yet known whether the mirror neuron system's sensitivity to effector variables, as indicated by neurological measures, is reflected in overt imitative performance.

Evidence that the mirror neuron system is sensitive to movement type has been provided by studies on both monkeys and humans (e.g. Di Pellegrino et al., 1992, Ferrari et al., 2003, Puce et al., 2000). Parallel evidence of behavioral sensitivity to movement type comes from many studies of imitation (Bertenthal et al., 2006, Brass et al., 2000, Brass et al., 2001, Castiello et al., 2002, Chartrand and Bargh, 1999, Craighero et al., 2002, Dimberg et al., 2000, Heyes et al., 2005, Kerzel and Bekkering, 2000, Kilner et al., 2003, Press et al., 2005, Stanley et al., 2007, Stürmer et al., 2000, Vogt et al., 2003), but is particularly clear in research on ‘automatic imitation’ using stimulus–response compatibility paradigms. For example, when participants have been instructed to make a pre-specified response (e.g. opening their hand) as soon as an observed hand begins to move, they initiate their response movement faster if the observed hand is performing a compatible movement (opening) than if it is performing an incompatible (closing) movement (Heyes et al., 2005, Press et al., 2005, Stürmer et al., 2000). Effects such as this show movement sensitivity in visuomotor priming: an observed action can prime the execution of the same action, but not of a different action, performed with the same body part. The matching of observed and executed movement types that occurs in automatic imitation is not reducible to the effects of either simple or complex spatial compatibility (Press et al., in press; see also Brass et al., 2001, Bertenthal et al., 2006).

Recent studies using neurological measures have shown that the human mirror neuron system is sensitive, not only to movement type, but to effector type: it responds differentially to the observation of different body parts in motion (Buccino et al., 2001, Sakreida et al., 2005, Wheaton et al., 2001, Wheaton et al., 2004). Buccino et al. (2001) and Wheaton et al. (2004) used neuroimaging to demonstrate that hand, foot and mouth actions selectively activate distinct regions of human ventral premotor and parietal cortex. Importantly, Wheaton et al. (2004) showed this somatotopic pattern of activation even when movements were held constant across effectors (opening and closing movements of a hand and a mouth, respectively). These findings indicate that the mirror neuron system codes the body parts involved in action, but there is no directly corresponding evidence that imitative behavior is sensitive to effector type.

A recent study has shown priming of index finger lifting movements by the observation of a lifting index finger, relative to observation of a lifting middle finger, and vice versa for middle finger lifting movements, even when fingers were in incongruent spatial locations (Bertenthal et al., 2006, Experiment 3b). While this provides some evidence that body part priming in imitative behavior is possible, it does not establish a parallel between imitation and the mirror neuron system because there is, at present, no corresponding evidence that body part coding in the mirror neuron system is selective for individual finger movements. In order to demonstrate that the body part dimension of an action affects both imitation and the mirror neuron system, it must be shown that, in parallel with the findings of Buccino et al. (2001) and Wheaton et al. (2004), observation of a hand in motion selectively primes hand movement, rather than foot movement, and vice versa for observation of a foot in motion.

Several studies suggest the occurrence of effector priming for various combinations of hand, foot and mouth movements in human and nonhuman subjects (e.g. Akins and Zentall, 1996, Bach and Tipper, 2007, Berger and Hadley, 1975, Dawson and Foss, 1965, Voelkl and Huber, 2000), but, on closer examination, it becomes clear that none of these studies isolated the effects of effector observation from those of movement observation. For example, Bach and Tipper (2007) found that observation of a model kicking a ball facilitated foot responses relative to hand responses, whereas observation of a model typing on a computer keyboard facilitated hand responses relative to foot responses. Although interesting in its own right, this result does not demonstrate effector priming because it could have been observation of hand use (effector priming), or observation of a movement – repetitive tapping – typically performed with the hand (movement priming) that facilitated hand responses.

To find out whether, like the mirror neuron system, imitation is sensitive to the effector dimension of action, Experiment 1 used an automatic imitation procedure in which we held constant the movement trajectory that was observed and executed, and varied only the effector that was selected for the performance of this movement type. In this choice-RT study, participants were required to lift their hand or their foot in response to a task-relevant letter (H or F), while a task-irrelevant image of a hand or foot lifting movement was simultaneously presented. If body parts are matched in automatic imitation, observed effectors should prime response effectors. That is, hand lifting responses should be initiated faster in the presence of a hand than a foot lifting action, while foot lifting responses should be faster when a foot rather than a hand action is observed.

In imitation, observed actions are related to the same executed actions. For example, when we imitate an observed foot action, we are more likely to perform this action with our foot than with our hand. Similarly, the observation of foot actions activates areas of premotor and parietal cortex involved in the execution of foot actions more than areas involved in the execution of hand actions (e.g. Buccino et al., 2001). It may seem obvious that observed foot actions are more similar to executed foot than hand actions, but, on reflection, it is evident that this intriguing and distinctive feature of imitation, and of the mirror neuron system, needs to be explained (Brass and Heyes, 2005). Given that the actions of the self and those of others are not seen from a third party perspective, it is not clear how the processes underlying imitation ‘know’ that observed foot actions are equivalent to executed foot actions rather than to executed hand actions, or how the mirror neuron system acquires its ‘mirror’ properties.

One possibility is that the matching of observed to executed actions is innate and experience-independent (Meltzoff and Moore, 1997, Gallese and Goldman, 1998). However, this is an unlikely hypothesis for the mirror neuron system in the light of several recent studies demonstrating its responsiveness to ‘unnatural stimuli’, such as the observation of tool use (Ferrari et al., 2005, Järveläinen et al., 2004, Obayashi et al., 2001) and the sound of paper ripping (Kohler et al., 2002), and of studies showing that the responsivity of the mirror neuron system varies with expertise in the observed action domain (e.g. Calvo-Merino et al., 2005, Lahav et al., 2007). If the development of imitation, and the mirror neuron system, is instead experience-dependent, then it may draw on three kinds of experience: unimodal sensory (Ferrari et al., 2005), unimodal motor (Calvo-Merino et al., 2006) or sensorimotor experience (Heyes, 2001, Heyes et al., 2005, Keysers and Perrett, 2004, Lahav et al., 2007, Obayashi et al., 2001). Unimodal sensory experience is provided by passive observation of an action, whereas unimodal motor experience arises from repeated execution of an action. In contrast, sensorimotor experience derives from correlated observation and execution of the same action. Thus, development of the capacity to, for example, match observed with executed foot movements, could result from observation of foot movements (unimodal sensory), execution of foot movements (unimodal motor), or from correlated experience of observing foot movements while performing foot movements (sensorimotor).

Building on the results of Experiment 1, Experiment 2 examined the role of sensorimotor experience in the development of imitation when unimodal sensory and unimodal motor experience were controlled. According to one sensorimotor account, the Associative Sequence Learning (ASL) model of imitation and the mirror neuron system (e.g. Heyes, 2001, Brass and Heyes, 2005), unimodal sensory and unimodal motor experience contribute to the establishment of sensory and motor representations, respectively, as a pre-requisite for sensorimotor learning, but they are not sufficient to establish the direct matching between an observed action and the same executed action that is necessary for imitation. ASL proposes that, instead, the formation of links between sensory and motor representations of the same action, and therefore of mirror neurons, depends on the correlated experience of observing and executing the same action. In the course of normal development, correlated experience of this kind is obtained through self-observation (e.g. watching one's own hand while it is moving), and through social interactions in which the individual is imitated by others, or engages in synchronous action with another agent. The model implies that when observation of one action is repeatedly paired with execution of another, nonmatching links will be formed between a sensory representation of one action (e.g. foot lifting) and a motor representation of a different action (e.g. hand lifting). On the basis of this assumption, Experiment 2 used body part priming to measure the strength of automatic imitation before and after a period in which subjects received ‘counter-mirror’ sensorimotor training. ‘Counter-mirror’ sensorimotor training required a group of participants (the incompatible group) to execute hand movements while observing foot movements and vice versa over the course of several training sessions. Another group (the compatible group), which served as a control, executed hand movements while observing hand movements, and foot movements while observing foot movements. The actions were familiar – they had been observed and executed by the subjects repeatedly before the experiment – and the groups received equal amounts of unimodal sensory and unimodal motor experience of the hand and foot actions in the course of training. Therefore, if the development of imitation depends on unimodal visual and/or motor experience alone, training should not result in an appreciable difference between groups; the compatible and incompatible groups should show equivalent priming effects before and after training. However, if the sensorimotor hypothesis is correct, and the development of imitation depends on the contingency experienced between action observation and action execution, then the incompatible group, but not the compatible group, should show less effector priming after training than before training.