The article by Cook et al. addresses the question whether mirror neurons (MNs) are a genetic adaptation for action understanding. The authors argue that if this were the case, one might predict that their functioning would be protected against “environmental perturbations” (sect. 7.1, para. 1). They make the further claim that if it could be demonstrated that the functioning of MNs can be modulated by associative learning, such an outcome could be taken as evidence against the genetic adaptation hypothesis.
One may question the prediction that if mirror neurons were “designed by evolution” for action understanding, their response properties should likely be protected against experience-based modulations. There are numerous examples in the literature demonstrating innate neuronal machinery that is modulated by experience (e.g., experience-based modulations of ocular dominance columns in V1: Wiesel & Hubel Reference Wiesel and Hubel1965; activation of primary visual cortex during Braille reading in early blind participants: Buechel et al. Reference Buechel, Price, Frackowiak and Friston1998; enlargement of the cortical representation of neighboring digits after deafferentation of single digits: Merzenich et al. Reference Merzenich, Kaas, Wall, Sur, Nelson and Felleman1983). Such plasticity, while constrained by the innate connectivity pattern of cortical and subcortical areas, enables the brain to flexibly adjust to a dynamic environment and to both permanent and temporary changes of the input. It is far from obvious why one should assume that a function that is innate would be protected from such plasticity. Thus, although Cook et al. convincingly demonstrate that the properties of MNs can be modulated by experience, the studies discussed in their article are inconclusive regarding the question whether the capability to match visual and motor representations of actions is innate.
We have argued, in another context, that the observed object category-specific organization in the visual ventral stream is driven primarily by distinct long-range connections to downstream processes (Mahon & Caramazza Reference Mahon and Caramazza2011). Different domains of objects are associated with different types of processes. For example, animate but not inanimate objects involve computing affective/social responses. The different processes that characterize different object domains involve distinct, even distant, areas of the brain that must be connected to function effectively as domain-specific networks. On this view, then, visual cortical organization is determined in part by the need to satisfy innate connectivity constraints. The innateness of these constraints is revealed by the fact that the large-scale, domain-specific organization of visual cortex remains invariant in congenitally blind subjects, that is, in the absence of visual input (e.g., Mahon et al. Reference Mahon, Anzelotti, Schwarzbach, Zampini and Caramazza2009). However, in these subjects the properties of the neurons in “visual” areas have undergone extensive modification: they now respond to completely different sensory inputs even as they retain their domain-specificity. This shows that plasticity does not imply absence of innate neural organization. Likewise, it seems reasonable to assume that the capacity of MNs to match visual and motor representations is made possible by the innate connectivity between ventral premotor cortex/macaque F5 and parietal cortex (AIP [anterior intraparietal], PFG [parietal] areas), which receives visual input from areas IT (inferotemporal cortex), STS (superior temporal sulcus), and MTG (middle temporal gyrus) (Borra et al. Reference Borra, Belmalih, Calzavara, Gerbella, Murata, Rozzi and Luppino2008; Luppino et al. Reference Luppino, Murata, Govoni and Matelli1999; Matelli et al. Reference Matelli, Camarada, Glickstein and Rizzolatti1986; Muakkassa & Strick Reference Muakkassa and Strick1979; Petrides & Pandya Reference Petrides and Pandya1984; Webster et al. Reference Webster, Bachevalier and Ungerleider1994). This innate connectivity defines the types of processes MNs can participate in while allowing for extensive local plasticity.
In our view, the fundamental question that needs asking is not whether specific associations between visual and motor representations of actions are present at birth – which we take as a given – but whether the link between visual and motor representations takes the form proposed by mirror neuron theorists. A mechanism specialized for connecting visual and motor functions is fundamental for any cognitive function, or otherwise we would lack the ability to react appropriately to sensory input. It seems reasonable to assume that such a basic mechanism should be genetically determined. What remain to be worked out are the anatomical and functional structure of the innate connections between visual to motor representations and the precise nature of the representations involved in this process. In the context of the latter issue, figuring out the role played by MNs in action understanding is key, but as Cook et al. note, the role of these neurons remains poorly understood.
One source of difficulty is the fact that even granting that MNs are involved in matching visual and motor aspects of actions, this does not imply that they are required for action understanding. It is typically argued that (mirror) neurons that code the abstract “goal” of an action (e.g., “grasping food,” irrespective of the type of grip and the effector involved) should be involved in action understanding. It is less clear, however, what is “motor” about such abstract representations, and how these differ from conceptual representations outside the macaque mirror neuron system. What is needed is evidence that MNs are necessary for comprehension as opposed to being activated as a consequence of action understanding, for example, to react appropriately to the observed action, or to coordinate our actions with those of other people. Unequivocal evidence demonstrating a causal link between the functioning of MNs and the ability to understand actions has not been presented thus far. We believe that the answer to this question is unlikely to come from investigation of modulation of MNs by associative learning alone, important as this might be for a full characterization of these neurons, but through the investigation of the impact of temporary and permanent lesions to areas containing mirror neurons and their connections.
The article by Cook et al. addresses the question whether mirror neurons (MNs) are a genetic adaptation for action understanding. The authors argue that if this were the case, one might predict that their functioning would be protected against “environmental perturbations” (sect. 7.1, para. 1). They make the further claim that if it could be demonstrated that the functioning of MNs can be modulated by associative learning, such an outcome could be taken as evidence against the genetic adaptation hypothesis.
One may question the prediction that if mirror neurons were “designed by evolution” for action understanding, their response properties should likely be protected against experience-based modulations. There are numerous examples in the literature demonstrating innate neuronal machinery that is modulated by experience (e.g., experience-based modulations of ocular dominance columns in V1: Wiesel & Hubel Reference Wiesel and Hubel1965; activation of primary visual cortex during Braille reading in early blind participants: Buechel et al. Reference Buechel, Price, Frackowiak and Friston1998; enlargement of the cortical representation of neighboring digits after deafferentation of single digits: Merzenich et al. Reference Merzenich, Kaas, Wall, Sur, Nelson and Felleman1983). Such plasticity, while constrained by the innate connectivity pattern of cortical and subcortical areas, enables the brain to flexibly adjust to a dynamic environment and to both permanent and temporary changes of the input. It is far from obvious why one should assume that a function that is innate would be protected from such plasticity. Thus, although Cook et al. convincingly demonstrate that the properties of MNs can be modulated by experience, the studies discussed in their article are inconclusive regarding the question whether the capability to match visual and motor representations of actions is innate.
We have argued, in another context, that the observed object category-specific organization in the visual ventral stream is driven primarily by distinct long-range connections to downstream processes (Mahon & Caramazza Reference Mahon and Caramazza2011). Different domains of objects are associated with different types of processes. For example, animate but not inanimate objects involve computing affective/social responses. The different processes that characterize different object domains involve distinct, even distant, areas of the brain that must be connected to function effectively as domain-specific networks. On this view, then, visual cortical organization is determined in part by the need to satisfy innate connectivity constraints. The innateness of these constraints is revealed by the fact that the large-scale, domain-specific organization of visual cortex remains invariant in congenitally blind subjects, that is, in the absence of visual input (e.g., Mahon et al. Reference Mahon, Anzelotti, Schwarzbach, Zampini and Caramazza2009). However, in these subjects the properties of the neurons in “visual” areas have undergone extensive modification: they now respond to completely different sensory inputs even as they retain their domain-specificity. This shows that plasticity does not imply absence of innate neural organization. Likewise, it seems reasonable to assume that the capacity of MNs to match visual and motor representations is made possible by the innate connectivity between ventral premotor cortex/macaque F5 and parietal cortex (AIP [anterior intraparietal], PFG [parietal] areas), which receives visual input from areas IT (inferotemporal cortex), STS (superior temporal sulcus), and MTG (middle temporal gyrus) (Borra et al. Reference Borra, Belmalih, Calzavara, Gerbella, Murata, Rozzi and Luppino2008; Luppino et al. Reference Luppino, Murata, Govoni and Matelli1999; Matelli et al. Reference Matelli, Camarada, Glickstein and Rizzolatti1986; Muakkassa & Strick Reference Muakkassa and Strick1979; Petrides & Pandya Reference Petrides and Pandya1984; Webster et al. Reference Webster, Bachevalier and Ungerleider1994). This innate connectivity defines the types of processes MNs can participate in while allowing for extensive local plasticity.
In our view, the fundamental question that needs asking is not whether specific associations between visual and motor representations of actions are present at birth – which we take as a given – but whether the link between visual and motor representations takes the form proposed by mirror neuron theorists. A mechanism specialized for connecting visual and motor functions is fundamental for any cognitive function, or otherwise we would lack the ability to react appropriately to sensory input. It seems reasonable to assume that such a basic mechanism should be genetically determined. What remain to be worked out are the anatomical and functional structure of the innate connections between visual to motor representations and the precise nature of the representations involved in this process. In the context of the latter issue, figuring out the role played by MNs in action understanding is key, but as Cook et al. note, the role of these neurons remains poorly understood.
One source of difficulty is the fact that even granting that MNs are involved in matching visual and motor aspects of actions, this does not imply that they are required for action understanding. It is typically argued that (mirror) neurons that code the abstract “goal” of an action (e.g., “grasping food,” irrespective of the type of grip and the effector involved) should be involved in action understanding. It is less clear, however, what is “motor” about such abstract representations, and how these differ from conceptual representations outside the macaque mirror neuron system. What is needed is evidence that MNs are necessary for comprehension as opposed to being activated as a consequence of action understanding, for example, to react appropriately to the observed action, or to coordinate our actions with those of other people. Unequivocal evidence demonstrating a causal link between the functioning of MNs and the ability to understand actions has not been presented thus far. We believe that the answer to this question is unlikely to come from investigation of modulation of MNs by associative learning alone, important as this might be for a full characterization of these neurons, but through the investigation of the impact of temporary and permanent lesions to areas containing mirror neurons and their connections.