Cook et al., in the section on mirror neuron basics (sect. 2), provide an incomplete description of the mirror system in the macaque, ignoring some recent developments. Using monkey fMRI, Nelissen et al. (Reference Nelissen, Luppino, Vanduffel, Rizzolatti and Orban2005) visualized the frontal regions involved in action observation. This study is extremely important as it makes a point which has escaped most researchers in the human imaging field: In order to activate F5c, where mirror neurons (MNs) are housed, in fMRI, the actor has to be in view in the videos. Simply showing an isolated hand preforming an action is insufficient. Cook et al. also ignore the subsequent study (Nelissen et al. Reference Nelissen, Borra, Gerbella, Rozzi, Luppino, Vanduffel, Rizzolatti and Orban2011), combining fMRI experiments with anatomical tracer studies. This later study revealed that the visual signals conveying action observation travel from the superior temporal sulcus (STS) to F5c over two parietal way stations: cytoarchitectonic PFG and the anterior intraparietal (AIP) area. Thus, this report indicates that, contrary to the assertion in section 3.2 (para. 2), STS neurons need not be linked to premotor neurons for the MNs to be generated, but rather, that visual and motor signals have to be associated in only a few parietal areas: PFG and AIP. This latter area, while not a classical mirror area, also houses MNs, according to fMRI (Nelissen et al. Reference Nelissen, Borra, Gerbella, Rozzi, Luppino, Vanduffel, Rizzolatti and Orban2011) and unpublished single-cell results from studies conducted in Parma and Japan. Although this restricted association in parietal cortex is compatible with a genetic/evolutionary as well as an associative learning origin, it suggests that a domain-general mechanism such as associative learning may not be needed and that a more specialized hybrid mechanism, such as exaptation (sect. 8.2), might be more relevant than Cook et al. indicate. The sentence “there is no evidence that the sensorimotor learning involved in MN development is modified or constrained relative to the associative learning that occurs in standard conditioning experiments” (sect. 8.2, para. 1) may thus require serious revision.

The other development regarding monkeys concern two recent studies from the Lemon group showing how the mirror signal is “extinguished” by the addition of suppressed MNs among cortico-spinal neurons in F5 or M1 (Kraskov et al. Reference Kraskov, Dancause, Quallo, Shepherd and Lemon2009; Vigneswaran et al. Reference Vigneswaran, Philipp, Lemon and Kraskov2013). Any account of how mirror neurons acquire their intriguing properties should take into account this transformation from purely excitatory to mixed excitatory/suppressed populations of MNs along the motor hierarchy. This point is again unaddressed, and it may well be that a genetic or hybrid mechanism can account for this range of responses more easily than an associative learning process. The data from monkeys concerning suppressed MNs also bear upon the interpretation of the Mukamel et al. (Reference Mukamel, Ekstrom, Kaplan, Iacoboni and Fried2010) data, suggesting that these human recordings may have been made in areas situated at a level of the motor hierarchy other than the planning level of the classical mirror areas (PFG/AIP and F5c).

In the section discussing mirror neuron basics, the authors also claim that the human mirror system is known. Ironically, the only mirror area for which the homology is known is the least documented in monkeys: AIP. There is indeed excellent evidence for the existence of this homologue, a combination of dorsal intraparietal sulcus anterior (DIPSA) and the so-called putative human AIP (phAIP) in anterior IPS (Durand et al. Reference Durand, Peeters, Norman, Todd and Orban2009). In contrast, most evidence for human areas cited by Cook et al. is weak or inexistent. Overlapping activations for action observation and execution are by no means proof for the existence of MNs, as the voxels contain thousands of neurons (Dinstein et al. Reference Dinstein, Gardner, Jazayeri and Heeger2008). Similarly, the repetition suppression studies have yielded contradictory results, unsurprisingly so, given that repetition suppression overestimates selectivity (Sawamura et al. Reference Sawamura, Orban and Vogels2006) and visual responses of premotor neurons do not adapt (Caggiano et al. Reference Caggiano, Pomper, Fleischer, Fogassi, Giese and Thier2013). That imitation constitutes an argument supporting the existence of a mirror system in humans is debatable, and that the inferior frontal gyrus (IFG) comprises a classical mirror area is unlikely, since it is not activated by action observation in well-controlled studies (Abdollahi et al. Reference Abdollahi, Jastorff and Orban2013; Jastorff et al. Reference Jastorff, Begliomini, Fabbri-Destro, Rizzolatti and Orban2010). Finally, the direct evidence of Mukamel et al. (Reference Mukamel, Ekstrom, Kaplan, Iacoboni and Fried2010) is also very disputable, because for technical reasons the so-called mirror neurons were, unlike the monkey MNs, recorded outside the areas involved in the planning of grasping. Indeed, so far it has been possible to record single neurons only in the medial regions of the hemispheres of epileptic patients undergoing pre-surgical physiological evaluation, preventing any exploration of anterior IPS or ventral premotor cortex. Given the recent evidence that medial fronto-parietal areas are involved in the observation of climbing and locomotion (Abdollahi et al. Reference Abdollahi, Jastorff and Orban2013), it is likely that Mukamel et al. (Reference Mukamel, Ekstrom, Kaplan, Iacoboni and Fried2010) tested neurons with far from optimal actions. The absence of clear evidence for human MNs and areas reduces the value of all the sensorimotor learning experiments that are cited in support of the authors' viewpoint (sects. 7.1 & 7.2), as they have been performed only in humans.

Finally, Cook et al. appear relatively uninformed about the value of rodent experiments, which they propose as an alternative to monkey studies (sect. 9.1, experimentation, para. 3). Even if rats do grasp with their forepaws, this behavior differs markedly from human grasping, reducing the relevance of proposed experiments for understanding the origin of human MNs. In general, given the immense difference in cortical surface area between the two species (for mice, the ratio is 1/1,000 [Rakic Reference Rakic2009], compared to 1/9.2 for macaque [Van Essen et al. Reference Van Essen, Glasser, Dierker, Harwell and Coalson2012]), the value of rodents as a model for the human brain may be overrated. Given the importance of the the posterior parietal cortex (PPC) for an extended mirror system, suggested by recent studies (Abdollahi et al. Reference Abdollahi, Jastorff and Orban2013; Filimon et al. Reference Filimon, Nelson, Hagler and Sereno2007), its limited development in rodents might further devaluate the rodent model in the present context. In fact, it may be the case that the most dorsal parts of the human superior parietal lobule (SPL) are evolutionarily the oldest parts of PPC (which we may share with rodents), whereas the more ventral regions of the inferior parietal lobule (IPL) are the most evolutionarily novel (Mantini et al. Reference Mantini, Corbetta, Romani, Orban and Vanduffel2013). Given the limitations of the rodent model, lesion or reversible inactivation experiments in monkeys would be more relevant than asserted by Cook et al.