The associative account described in the target article provides a viable explanation for the origin of mirror neurons. We argue here that if mirror neurons develop purely by associative learning, then they cannot by themselves explain intentional action understanding. Higher-level processes seem to be involved in the formation of associations as well as in their application during action understanding.

Mirror neuron–based associative learning may be understood according to associative learning theories, in addition to sensorimotor learning theories. This is important for a comprehensive understanding of the role of mirror neurons and related hormone modulators, such as oxytocin, in complex social interactions such as among parent–infant dyads and in examples of mirror neuron function that involve abnormal motor systems such as depression.

Speech is commonly claimed to relate to mirror neurons because of the alluring surface analogy of mirror neurons to the Motor Theory of speech perception, which posits that perception and production draw upon common motor-articulatory representations. We argue that the analogy fails and highlight examples of systems-level developmental approaches that have been more fruitful in revealing perception–production associations.

Hebbian Learning should not be reduced to contiguity, as it detects contingency and causality. Hebbian Learning accounts of mirror neurons make predictions that differ from associative learning: Through Hebbian Learning, mirror neurons become dynamic networks that calculate predictions and prediction errors and relate to ideomotor theories. The social force of imitation is important for mirror neuron emergence and suggests canalization.

The rational-decision approach is superior to the associative-learning approach of Cook et al. at explaining why mirror neurons fire or do not fire – even when the stimulus is the same. The rational-decision approach is superior because it starts with the analysis of the intention of the organism, that is, with the identification of the specific objective or goal that the organism is trying to maximize.

Ever since their discovery, mirror neurons have generated much interest and debate. A commonly held view of mirror neuron function is that they transform “visual information into knowledge,” thus enabling action understanding and non-verbal social communication between con-specifics (Rizzolatti & Craighero 2004). This functionality is thought to be so important that it has been argued that mirror neurons must be a result of selective pressure.

The target article argues that developmental processes are key to understanding the mirror neuron system, yet neglects several bodies of developmental research that are informative for doing so. Infants' actions and action understanding are structured by goals, and the former lends structure to the latter. Evaluating the origins and functions of mirror neurons depends on integrating investigations of neural, social-cognitive and motor development.

We argue, by analogy to the neural organization of the object recognition system, that demonstration of modulation of mirror neurons by associative learning does not imply absence of genetic adaptation. Innate connectivity defines the types of processes mirror neurons can participate in while allowing for extensive local plasticity. However, the proper function of these neurons remains to be worked out.

An associative learning account of mirror neurons should not preclude genetic evolution of its underlying mechanisms. On the contrary, an associative learning framework for cognitive development should seek heritable variation in the learning rules and in the data-acquisition mechanisms that construct associative networks, demonstrating how small genetic modifications of associative elements can give rise to the evolution of complex cognition.

Cook et al. propose that mirror neurons emerge developmentally through a domain-general associative mechanism. We argue that experience-sensitivity does not rule out an adaptive or genetic argument for mirror neuron function, and that current evidence suggests that mirror neurons are more specialized than the authors' account would predict. We propose that future work integrate behavioral and neurophysiological techniques used with primates to examine the proposed functions of mirror neurons in action understanding.

Some birds and mammals have vocal communication systems in which coordination between individuals is important. Examples would include duetting or antiphonal calling in some birds and mammals, rapid exchanges of the same vocalization, and vocal exchanges between paired individuals and other nearby pairs. Mirror neurons may play a role in such systems but become functional only after experience.

Cook et al. argue that mirror neurons originate from associative learning processes, without evolutionary influence from social-cognitive mechanisms. We disagree with this claim and present arguments based upon cross-species comparisons, EEG findings, and developmental neuroscience that the evolution of mirror neurons is most likely driven simultaneously and interactively by evolutionarily adaptive psychological mechanisms and lower-level biological mechanisms that support them.

Cook et al. overstate the evidence supporting their associative account of mirror neurons in humans: most studies do not address a key property, action-specificity that generalizes across the visual and motor domains. Multivariate pattern analysis (MVPA) of neuroimaging data can address this concern, and we illustrate how MVPA can be used to test key predictions of their account.

The description of the mirror neuron system provided by Cook et al. is incomplete for the macaque, and incorrect for humans. This is relevant to exaptation versus associative learning as the underlying mechanism generating mirror neurons, and to the sensorimotor learning as evidence for the authors' viewpoint. The proposed additional testing of the mirror system in rodents is unrealistic.

Contagious behaviors such as yawning and itching/scratching have mirror-like properties and clearly defined stimulus and motor parameters; they are also relatively easy to study and should be part of the debate about mirror neurons and the neurological mechanisms of social behavior. The broadly tuned, multimodal stimuli of contagious behavior challenge present accounts of mirror mechanisms that focus on specific, mirrored acts.

Investigations of biases and experiential effects on social learning, social information use, and mirror systems can usefully inform one another. Unconstrained learning is predicted to shape mirror systems when the optimal response to an observed act varies, but constraints may emerge when immediate error-free responses are required and evolutionary or developmental history reliably predicts the optimal response. Given the power of associative learning, such constraints may be rare.

Cook et al. argue that mirror neurons originate in sensorimotor associative learning and that their function is determined by their origin. Both these claims are hard to accept. It is here suggested that a major role in the origin of the mirror mechanism is played by top-down connections rather than by associative learning.

Like Cook et al., we suggest that mirror neurons are a fascinating product of cross-modal learning. As predicted by an associative account, responses in motor regions are observed for novel and/or abstract visual stimuli such as point-light and android movements. Domain-specific mirror responses also emerge as a function of audiomotor expertise that is slowly acquired over years of intensive training.

Neonatal imitation should not exclusively be considered at the population-level; instead, we propose that inconsistent findings regarding its occurrence result from important individual differences in imitative responses. We also highlight what we consider to be a false dichotomy of genetic versus learning accounts of the development of mirror neurons, and instead suggest a more parsimonious epigenetic perspective.

Commentators have tended to focus on the conceptual framework of our article, the contrast between genetic and associative accounts of mirror neurons, and to challenge it with additional possibilities rather than empirical data. This makes the empirically focused comments especially valuable. The mirror neuron debate is replete with ideas; what it needs now are system-level theories and careful experiments – tests and testability.