Trends in Cognitive Sciences
OpinionCrossmodal and action-specific: neuroimaging the human mirror neuron system
Introduction
In the early 1990s a seminal paper [1] reported the existence of neurons in macaque frontal area F5 that showed remarkable tuning properties: these neurons not only fired when the monkey executed a specific action (such as grasping a pellet of food) but also when it observed an experimenter performing the same action. Soon, more reports of this type of visuomotor neurons, later termed ‘mirror neurons’, followed. Some of the key findings were that, first, neurons with similar properties were found in macaque parietal regions PF and PFG 2, 3, which together with F5 1, 4 were termed the frontoparietal ‘mirror neuron system’ (Figure 1a) 5, 6. Second, mirror neurons also respond when an object is initially viewed but the subsequent reach-to-grasp is obscured by a screen [7], showing an influence of contextual knowledge on mirror neuron activity. Third, some mirror neurons respond differentially to the observation of the same motor act (e.g., grasping) in the context of different actions (e.g., grasping to eat or placing an object), suggesting a mechanism by which the final goal of a series of actions could be understood 2, 3. Fourth, the class of mirror neurons is heterogeneous with respect to tuning properties of individual neurons on various dimensions including hand and direction preference [4], distance to the observed actor [8], and viewpoint of the observed action [9].
If humans are endowed with such neurons as well, many have argued that this would provide an explanation for how people solve the ‘correspondence problem’ 4, 10 of imitation and of learning and understanding actions performed by others. Given the anatomical location of F5 – in the premotor cortex – a popular interpretation was (and is) that this occurs through a simulation or direct matching mechanism, where observing someone else activates the same motor circuits as when executing that action ‘from within’, through a ‘motor resonance’ process 5, 6, 11. According to this interpretation, people can understand the actions of others by mapping them directly onto their own motor repertoire. More generally, the idea that visual and motor representations of actions share a common neural ‘code’ may also help explain findings showing that task-irrelevant spatial [12], symbolic [13], body-related [14], and affordance [15] aspects of stimuli can affect subsequent action responses. Similar effects are also found in more-complex situations, as in the ‘chameleon’ effect – the tendency of humans to mimic the actions of social partners [16]. Furthermore, such a mirror mechanism 5, 6 has also been proposed to underlie more general processes – beyond action representations – such as the automatic understanding of the feelings (i.e., empathy) [17] and thoughts (i.e., mentalising) [18] of others. It is also argued that mirror neurons play a role in language acquisition [19] given the close proximity of macaque F5 and its putative human homologue of Broca's area. Finally, it has been suggested that a dysfunction of mirror neurons is the underlying mechanism of autism [20] (but see [21]).
The putative explanatory power of mirror neurons for this wide range of human social phenomena has led to the prediction that ‘mirror neurons will do for psychology what DNA did for biology’ [22]. Although not without critics 21, 22, 23, 24, 25, 26, 27, these thoughts – all resting on the concept of a ‘human mirror neuron system’ (HMNS) 5, 28, 29 – make the effort to identify and characterise this system all the more important [30].
Section snippets
A human ‘mirror neuron system’?
Macaque and human brains differ significantly, therefore the findings from macaques do not necessarily extend to humans in a straight forward manner. Indeed, the last-known common ancestor of macaques and humans is estimated to have lived 30 million years ago, resulting in partial but imperfect homology between the species [31]. For example, although early visual areas seem to map well, significant differences have been found between macaques and humans in higher-level associative areas of the
Crossmodal responses: necessary but not sufficient
Many previous fMRI studies of the HMNS have followed the logic that the presence of mirror neurons in a region can be inferred when an increased response results from observing and executing actions compared with a baseline condition without a task. In other words: a crossmodal response across the visual and motor modalities. Ostensibly these studies provide overwhelming evidence in support of a HMNS by showing such a response for viewed and executed actions in frontal and parietal regions [6],
A crucial ingredient: action specificity
A further key property of mirror neurons is action specificity 30, 42. That is, according to a direct-matching account [6], different actions must elicit distinguishable neural signatures, regardless of whether they are seen or executed. Considered at the level of a neural population, one might expect that the response associated with a particular action should be similar whether that action is performed or observed but should be dissimilar for different actions (Figure 2).
The first attempts to
Multivariate pattern analysis
An alternative approach to study action-specificity of neural representations is multivariate pattern analysis (MVPA) 48, 49, 50, 51. This sensitive approach has been used successfully to decode subtle differences between stimulus-evoked activity patterns and even to reconstruct static and dynamic percepts from brain signals [52]. Such sensitivity is important if the proportion of mirror neurons in a region is small, as suggested by macaque (∼5% to ∼17% [4]) and human (∼8% [35]) studies.
Viewpoint invariance
Observing our own actions is typically associated with a particular ‘first-person’ visual perspective on those actions. By contrast, we typically see others’ actions from a variety of different ‘third-person’ perspectives. A key property of macaque mirror neurons is that they seem to relate the motor aspects of an action (which are inherently personal or ‘first person’) to the visual aspects of the same action even when that action is performed by others and, hence, seen from a different
Concluding remarks and future directions
Recent advances in fMRI methods have been applied to examine the HMNS by extrapolating some of the key properties of mirror neurons to the population level. Already, these first steps paint a picture that differs in some ways from the canonical frontoparietal model of the HMNS (Figure 1) that emerged from the first wave of fMRI studies in this area.
First, with MVPA, PMv shows stronger first-person than third-person view representations of actions, which could mean it is less involved in
Acknowledgements
We thank Marius Peelen, Emily Cross, and Nikolaus Kriegeskorte for helpful suggestions on an earlier version of this manuscript. We acknowledge the Economic and Social Research Council, the Leverhulme Trust, and the Boehringer Ingelheim Fonds for funding support.
References (102)
I know what you are doing. a neurophysiological study
Neuron
(2001)View-based encoding of actions in mirror neurons of area F5 in macaque premotor cortex
Curr. Biol.
(2011)- et al.
Imitation: is cognitive neuroscience solving the correspondence problem?
Trends Cogn. Sci.
(2005) - et al.
Mirror neurons and the simulation theory of mind-reading
Trends Cogn. Sci.
(1998) - et al.
Language within our grasp
Trends Neurosci.
(1998) Normal movement selectivity in autism
Neuron
(2010)Where do mirror neurons come from?
Neurosci. Biobehav. Rev.
(2010)More than one pathway to action understanding
Trends Cogn. Sci.
(2011)- et al.
A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content
J. Physiol. Paris
(2008) A unifying view of the basis of social cognition
Trends Cogn. Sci.
(2004)