Mirror neurons, action understanding and social interaction: implications for educational neuroscience

• ¹ Department of Psychology, Institute of Psychiatry, Psychology and Neuroscience, King's College London, United Kingdom
• ² University of Oxford, United Kingdom

Mirror neurons are cells which fire to both the observation and execution of actions (di Pellegrino et al., 1992). Originally found in premotor area F5 of the macaque monkey, they have subsequently been found in additional brain areas including the inferior parietal lobule and dorsal premotor cortex (Fogassi et al., 2005; Tkach et al., 2007). In humans, converging neuroscientific evidence for the presence of mirror neurons (Fadiga et al., 1995; Iacoboni et al., 1999; Kilner et al., 2009) has led some researchers to claim that mirror neurons are involved in a process known as ‘action understanding’ (Binkofski & Buccino, 2006). However, what is meant by the term ‘action understanding’ is disputed (Cook et al., 2014; Hickok, 2009), and this has left considerable confusion in the literature, regarding both what ‘action understanding’ refers to, and whether mirror neurons are involved in such a process. Below, we outline our concerns regarding these two points; and emphasise the importance of clarifying not only the evidence for mirror neuron involvement in ‘action understanding’, but also the definition of ’action understanding’, before proposing that either mirror neurons or ‘action understanding’ should be targets for educational neuroscience interventions (Zhu et al., 2011). ‘Action understanding’ has been defined as the ability to identify others’ actions, infer the goal of their actions, and extract the intention underlying their actions (Gallese et al., 1996). This definition suggests that ‘action understanding’ can be separated into three distinct processes: identifying the configuration of body parts involved in an action; identifying its goal; and identifying the actor’s intention. However, these three processes differ in terms of the amount of generalization required across action features. More specifically, identifying an action requires discrimination of the perceptual features (such as the type of grip used: e.g. precision or whole-hand), based on the configural relationship between different body parts. Identifying goals and intentions, however, requires generalization across the perceptual features of observed actions. This is because a goal (e.g. “to grasp”) or intention (e.g. “in order to drink”) can be achieved using a range of different grip types; and, crucially, the same grip type can be used to achieve a number of different goals and intentions. Since there is not a one-to-one correspondence between body part configurations, goals, and intentions (e.g. Jacob & Jeannerod, 2005), the same pattern of mirror neuronal firing cannot simultaneously represent the other’s action, their goal, and their intention. Although this has been acknowledged in the literature (Kilner, 2011), researchers investigating the neurocognitive mechanisms underlying ‘action understanding’ typically focus on only one of the above processes (e.g. Fogassi et al., 2005; Spunt et al., 2011; Umiltà et al., 2001); but problematically, all three processes are referred to as ‘action understanding’. The non-specific use of the term ‘action understanding’ has led to contradictory results regarding whether mirror neurons (or, in human studies, brain areas thought to contain mirror neurons) are involved in ‘action understanding’. For example, some researchers have claimed that brain regions containing mirror neurons are involved in identifying others’ intentions (Michael et al., 2014). Michael et al. found that disruption to the premotor cortex by continuous theta burst stimulation impaired accuracy on a task involving the analysis of the actor’s intention. In this task, participants had to combine information on the configuration of a hand with contextual information in order to deduce the actor’s intention. However, closer inspection of the results showed that accuracy on a purely perceptual task, in which the participant matched the configuration of a hand depicted in a video to a hand depicted in an image, was impaired to the same extent as the intention task. If mirror neurons are involved in identifying intentions, then stimulation should have reduced accuracy on the task involving the analysis of intentions to a greater extent than the purely perceptual task. Therefore, the results only provide convincing evidence for the role of mirror neurons in the identification of the perceptual features of actions (see Catmur, 2014). Similarly, other researchers have claimed to have found evidence that mirror neurons are able to distinguish between different goals (Hafri et al., 2017). However, other evidence has demonstrated that mirror neuron brain areas encode different exemplars of actions based on their perceptual features (Nicholson et al., 2017), suggesting that mirror neuron areas may only appear to encode the goals of observed actions when those goals are perceptually distinct. Moreover, generalization across the perceptual features of observed actions appears to occur in other, non-motor, brain areas, such as the lateral occipitotemporal cortex (Wurm et al., 2016; Wurm & Lingnau, 2015) or the dorsomedial prefrontal cortex (Spunt & Adolphs, 2014; Spunt & Lieberman, 2013), which are not thought to contain mirror neurons. This summary of the literature suggests that there is some evidence to support mirror neuron involvement in identification of the configuration of body parts involved in an observed action, but not in identifying the goal or the intention of the action. However, since the current definition of ‘action understanding’ does not distinguish between these processes, whether this evidence supports mirror neuron involvement in ‘action understanding’ clearly depends on which particular sub-definition of this term is adopted. The possible involvement of mirror neurons in ‘action understanding’ is particularly important for educational neuroscience. Both the psychological process implied by the term ‘action understanding’, and its potential neural underpinnings in terms of mirror neurons, have the potential to be targeted in educational interventions: most notably, for those with autism spectrum disorder (ASD). ASD is a neurodevelopmental disorder involving deficits in social communication and social interaction, and restricted, repetitive patterns of behavior (DSM-5). Perhaps the most striking social difficulty in ASD, a deficit in theory of mind (Baron-Cohen et al., 1985), has been associated – at least theoretically – with the concept of ‘action understanding’ (Koster-Hale & Saxe, 2013). Certain researchers have therefore proposed that mirror neurons are a possible neural substrate for these social abilities (Gallese, 2003), with impairments in mirror neuron functioning playing a role in the social deficits seen in ASD (e.g. Khalil et al., 2018; Oberman et al., 2005). However, the link between the term 'action understanding' and social abilities such as theory of mind is theoretical rather than empirical. Moreover, as discussed above, the term itself refers to three distinct processes, all of which are likely to have different associations with specific social abilities. For example, ASD has been associated with difficulties in identifying the intentions underlying observed actions but not with difficulties in identifying the actions themselves (Boria et al., 2009). If, as proposed above, mirror neuron involvement in ‘action understanding’ is limited to the identification of the configuration of body parts in an observed action, and not in the identification of the goal or intention underlying that action, then these data are not consistent with the proposed association between mirror neuron dysfunction and the social symptoms of ASD. Despite the unsubstantiated nature of the links between mirror neurons, ‘action understanding’, and social ability, educational interventions targeted at altering mirror neuron firing have been designed in an attempt to improve the social abilities of individuals with ASD. One such intervention is neurofeedback therapy, which uses operant conditioning to alter brain responses. Neurofeedback interventions targeting mirror neuron function have focused on the mu rhythm, an 8–12 Hz electroencephalographic frequency band recorded from sensorimotor cortex that is proposed to index mirror neuron activity (e.g. Hari, 2006). While some researchers have claimed success with this intervention (Pineda et al., 2014), at the meta-analytical level there is no support for its effectiveness (Holtmann et al., 2011). Indeed, debate is ongoing regarding whether neurofeedback therapy does target mu activity (Hobson & Bishop, 2017); whether mu activity is associated with mirror neurons (Hobson & Bishop, 2016); and whether mirror neurons are indeed linked to deficits in social abilities (Enticott et al., 2013). In this context, therefore, we argue that the first step in establishing whether educational neuroscience interventions should target mirror neuron function should be to determine the role of mirror neurons in each of the three processes currently subsumed under the term ‘action understanding’. The next step is to determine the association between each of those processes and social abilities, including theory of mind. Unless mirror neurons are involved in a process that contributes to such social abilities, educational interventions targeting mirror neuron activation, such as neurofeedback therapy, are unlikely to be productive. Therefore, until a clearer definition of ‘action understanding’ is agreed, and until mirror neuron involvement in such a process is clarified, researchers and clinicians should be wary of educational interventions targeting such factors.