The mirror neuron mechanism
The most important aspect of MNs is their capacity to match the visual/acoustic cortical representation of a biological stimulus with its corresponding somatomotor or visceromotor representation. Thus, MNs constitute a mechanism that not only explains the automatic decoding of the motor acts of others, but also many other types of processes involved in social cognitive functions (e.g., emotion recognition, imitation, oro-facial communication). Recent studies have shown the existence of MNs active during both listening and production of species-typical song in singing birds (Prather et al. Reference Prather, Peters, Nowicki and Mooney2008), suggesting that mirror mechanisms are probably very primitive solutions evolved in different vertebrate classes to elaborate sensory information for social cognition. Furthermore, behavioural evidence suggests that other vertebrates such as rats or dogs (Range et al. Reference Range, Viranyi and Huber2007; Zentall & Levine Reference Zentall and Levine1972) or even invertebrates such as the octopus (Fiorito & Scotto Reference Fiorito and Scotto1992), could be endowed with some form of mirror mechanism.
By using the same matching mechanism, mirroring may occur, even within the same species, at different levels: in the comprehension of goals (Cattaneo et al. Reference Cattaneo, Caruana, Jezzini and Rizzolatti2009; Fogassi et al. Reference Fogassi, Ferrari, Gesierich, Rozzi, Chersi and Rizzolatti2005; Rizzolatti et al Reference Rizzolatti, Fogassi, Gallese and Gazzaniga2004) or of meaningful communicative or symbolic gestures (Ferrari et al. Reference Ferrari, Gallese, Rizzolatti and Fogassi2003; Lui et al. Reference Lui, Buccino, Duzzi, Benuzzi, Crisi, Baraldi, Nichelli, Porro and Rizzolatti2008) (high level), and in the decoding of observed movements rather than of motor acts (Catmur et al. Reference Catmur, Walsh and Heyes2007; Fadiga et al. Reference Fadiga, Fogassi, Pavesi and Rizzolatti1995) (low level).
Mirror neurons and sensorimotor associative learning
Cook et al. claim that MNs are the result of associative learning rather than an adaptation selected by evolution for action understanding genetically coded in humans and ancestors. Following the authors' reasoning, if we knew the experience of every monkey since birth, we could predict the formation, in monkeys living in different developmental environments, of different types of MNs. Furthermore, many of the typical primate behavioural functions of daily life would be the result of associative learning, and therefore we should observe a large inter-individual behavioural variability. However, it is well known that, for example, object- or space-related sensorimotor transformations for reaching-grasping actions are grounded on dorsal cortical circuits (Rizzolatti & Matelli Reference Rizzolatti and Matelli2003) that are phylogenetically very old and, in the primates' lineage, very similar among different species. Just as there are these dedicated circuits, linking specific parietal and premotor areas (Rizzolatti & Luppino Reference Rizzolatti and Luppino2001), so there is also a dedicated mirror circuit for hand actions observation, linking anterior superior temporal sulcus (aSTS)↔inferior parietal cortex (PFG)↔ventral premotor cortex (area F5c) (Nelissen et al. Reference Nelissen, Borra, Gerbella, Rozzi, Luppino, Vanduffel, Rizzolatti and Orban2011, p. 3754). It is evident that such selected circuits cannot re-build every time. They provide, rather, the neuroanatomical scaffold for both hard-wired and newly-learned sensorimotor transformations. In these circuits, pre-existing and new motor representations are matched with their corresponding sensory representations.
As an example of an anatomo-functional circuit in which the mirror matching mechanism operates, a series of recent works (Bonini et al. Reference Bonini, Rozzi, Serventi, Simone, Ferrari and Fogassi2010; Fogassi et al. Reference Fogassi, Ferrari, Gesierich, Rozzi, Chersi and Rizzolatti2005) has shown that the discharge of purely motor and MNs of monkey premotor (area F5c) and parietal (area PFG) cortex is modulated, during grasping observation/execution, depending on the behavioural goals of specific executed or observed action sequences (grasp-to-eat or grasp-to-place). Furthermore, the percentage of MNs tuned for the hard-wired action (grasp-to-eat) is much higher than that of MNs tuned for the learned action (grasp-to-place). Overall, these and other data suggest that the mirror mechanism, deeply rooted in primate evolution, also plays a strong role in the extension of action understanding capacity to new actions, in motor skill consolidation (Cross et al. Reference Cross, de Hamilton and Grafton2006) and in observation-based rehabilitation (Ertelt et al. Reference Ertelt, Small, Solodkin, Dettmers, McNamara, Binkofski and Buccino2007). These processes could also benefit from associative learning.
Action recognition and action understanding
Cook et al. claim that there is no consensus on the concept of action understanding and on its distinction from action perception and recognition. If the objective of the nervous system were to simply ensure action recognition, probably the visual system would be enough: some sectors of aSTS would be the best candidate for recognizing the actions of others. However, if we assume that the motor system is crucial for cognition because it provides information about motor goals, aSTS areas alone are not able to support action understanding because they do not show motor responses (Perrett et al. Reference Perrett, Harries, Bevan, Thomas, Benson, Mistlin, Chitty, Hietanen and Ortega1989). Thus, we can hypothesize that the reciprocal neuroanatomical links between high-order visual areas and motor areas endow individuals with two main abilities: (1) to interpret the vision of a hand approaching an object in terms of goal; (2) to better perceive the details of the observed actions.
Two recent studies support these two functions. The first (Caggiano et al. Reference Caggiano, Fogassi, Rizzolatti, Thier and Casile2009) shows that in half of recorded MNs the intensity of the visual response is different depending on whether the observed grasping is performed within or outside the peri-personal space of the monkey, and this effect can be further modulated by the possibility for the monkey to act or not in its peri-personal space. The second (Caggiano et al. Reference Caggiano, Fogassi, Rizzolatti, Pomper, Thier, Giese and Casile2011) shows that 75% of the recorded MNs discharge more strongly when the monkey sees the action either from an egocentric or a third-person perspective. These results can be interpreted as the demonstration that MNs, although always encoding, in their output, the goal of a motor act (they are basically motor neurons), can, at the very same time, in conjunction (through feedback projections) with high-order visual areas sensitive to biological stimuli, provide information on specific details of the observed action, by enhancing the activity of the sensory neurons that are more selective for those details.
The core issue of the target article is whether mirror neurons (MNs) arise by associative learning. My comments, based on neuroanatomical and electrophysiological expertise, focus on this and other related issues (mirror mechanism and action understanding) addressed by the article.
The mirror neuron mechanism
The most important aspect of MNs is their capacity to match the visual/acoustic cortical representation of a biological stimulus with its corresponding somatomotor or visceromotor representation. Thus, MNs constitute a mechanism that not only explains the automatic decoding of the motor acts of others, but also many other types of processes involved in social cognitive functions (e.g., emotion recognition, imitation, oro-facial communication). Recent studies have shown the existence of MNs active during both listening and production of species-typical song in singing birds (Prather et al. Reference Prather, Peters, Nowicki and Mooney2008), suggesting that mirror mechanisms are probably very primitive solutions evolved in different vertebrate classes to elaborate sensory information for social cognition. Furthermore, behavioural evidence suggests that other vertebrates such as rats or dogs (Range et al. Reference Range, Viranyi and Huber2007; Zentall & Levine Reference Zentall and Levine1972) or even invertebrates such as the octopus (Fiorito & Scotto Reference Fiorito and Scotto1992), could be endowed with some form of mirror mechanism.
By using the same matching mechanism, mirroring may occur, even within the same species, at different levels: in the comprehension of goals (Cattaneo et al. Reference Cattaneo, Caruana, Jezzini and Rizzolatti2009; Fogassi et al. Reference Fogassi, Ferrari, Gesierich, Rozzi, Chersi and Rizzolatti2005; Rizzolatti et al Reference Rizzolatti, Fogassi, Gallese and Gazzaniga2004) or of meaningful communicative or symbolic gestures (Ferrari et al. Reference Ferrari, Gallese, Rizzolatti and Fogassi2003; Lui et al. Reference Lui, Buccino, Duzzi, Benuzzi, Crisi, Baraldi, Nichelli, Porro and Rizzolatti2008) (high level), and in the decoding of observed movements rather than of motor acts (Catmur et al. Reference Catmur, Walsh and Heyes2007; Fadiga et al. Reference Fadiga, Fogassi, Pavesi and Rizzolatti1995) (low level).
Mirror neurons and sensorimotor associative learning
Cook et al. claim that MNs are the result of associative learning rather than an adaptation selected by evolution for action understanding genetically coded in humans and ancestors. Following the authors' reasoning, if we knew the experience of every monkey since birth, we could predict the formation, in monkeys living in different developmental environments, of different types of MNs. Furthermore, many of the typical primate behavioural functions of daily life would be the result of associative learning, and therefore we should observe a large inter-individual behavioural variability. However, it is well known that, for example, object- or space-related sensorimotor transformations for reaching-grasping actions are grounded on dorsal cortical circuits (Rizzolatti & Matelli Reference Rizzolatti and Matelli2003) that are phylogenetically very old and, in the primates' lineage, very similar among different species. Just as there are these dedicated circuits, linking specific parietal and premotor areas (Rizzolatti & Luppino Reference Rizzolatti and Luppino2001), so there is also a dedicated mirror circuit for hand actions observation, linking anterior superior temporal sulcus (aSTS)↔inferior parietal cortex (PFG)↔ventral premotor cortex (area F5c) (Nelissen et al. Reference Nelissen, Borra, Gerbella, Rozzi, Luppino, Vanduffel, Rizzolatti and Orban2011, p. 3754). It is evident that such selected circuits cannot re-build every time. They provide, rather, the neuroanatomical scaffold for both hard-wired and newly-learned sensorimotor transformations. In these circuits, pre-existing and new motor representations are matched with their corresponding sensory representations.
As an example of an anatomo-functional circuit in which the mirror matching mechanism operates, a series of recent works (Bonini et al. Reference Bonini, Rozzi, Serventi, Simone, Ferrari and Fogassi2010; Fogassi et al. Reference Fogassi, Ferrari, Gesierich, Rozzi, Chersi and Rizzolatti2005) has shown that the discharge of purely motor and MNs of monkey premotor (area F5c) and parietal (area PFG) cortex is modulated, during grasping observation/execution, depending on the behavioural goals of specific executed or observed action sequences (grasp-to-eat or grasp-to-place). Furthermore, the percentage of MNs tuned for the hard-wired action (grasp-to-eat) is much higher than that of MNs tuned for the learned action (grasp-to-place). Overall, these and other data suggest that the mirror mechanism, deeply rooted in primate evolution, also plays a strong role in the extension of action understanding capacity to new actions, in motor skill consolidation (Cross et al. Reference Cross, de Hamilton and Grafton2006) and in observation-based rehabilitation (Ertelt et al. Reference Ertelt, Small, Solodkin, Dettmers, McNamara, Binkofski and Buccino2007). These processes could also benefit from associative learning.
Action recognition and action understanding
Cook et al. claim that there is no consensus on the concept of action understanding and on its distinction from action perception and recognition. If the objective of the nervous system were to simply ensure action recognition, probably the visual system would be enough: some sectors of aSTS would be the best candidate for recognizing the actions of others. However, if we assume that the motor system is crucial for cognition because it provides information about motor goals, aSTS areas alone are not able to support action understanding because they do not show motor responses (Perrett et al. Reference Perrett, Harries, Bevan, Thomas, Benson, Mistlin, Chitty, Hietanen and Ortega1989). Thus, we can hypothesize that the reciprocal neuroanatomical links between high-order visual areas and motor areas endow individuals with two main abilities: (1) to interpret the vision of a hand approaching an object in terms of goal; (2) to better perceive the details of the observed actions.
Two recent studies support these two functions. The first (Caggiano et al. Reference Caggiano, Fogassi, Rizzolatti, Thier and Casile2009) shows that in half of recorded MNs the intensity of the visual response is different depending on whether the observed grasping is performed within or outside the peri-personal space of the monkey, and this effect can be further modulated by the possibility for the monkey to act or not in its peri-personal space. The second (Caggiano et al. Reference Caggiano, Fogassi, Rizzolatti, Pomper, Thier, Giese and Casile2011) shows that 75% of the recorded MNs discharge more strongly when the monkey sees the action either from an egocentric or a third-person perspective. These results can be interpreted as the demonstration that MNs, although always encoding, in their output, the goal of a motor act (they are basically motor neurons), can, at the very same time, in conjunction (through feedback projections) with high-order visual areas sensitive to biological stimuli, provide information on specific details of the observed action, by enhancing the activity of the sensory neurons that are more selective for those details.