Cook et al. articulate two conclusions with which we wholeheartedly agree: (1) The functions of mirror neurons (MNs) cannot be determined based only on patterns of neural activation during action observation. Independent measures of the putative social-cognitive functions of MNs are needed and have, so far, not been sufficiently integrated with neural measures; and (2) understanding the developmental origins of MNs and the broader systems in which they are situated is essential for understanding their functional significance.

Given the centrality of developmental processes to Cook et al.'s arguments, we find it surprising that they do not engage the developmental literature more fully. They propose a relatively simple learning process – the formation of contingency-based associations between visual and motor experience – to account for the existence of MNs. For example, they propose that MNs reflect repeated experiences with reaching for objects and seeing the resulting hand movements. This kind of learning seems very likely to occur, but without a fuller consideration of motor and social-cognitive development, it is difficult to see how any important social-cognitive functions could arise from motor experience. In fact, several bodies of experimental work with human infants indicate that much richer connections exist between motor experience and social cognition.

Developmental research shows that infants' actions are prospectively goal-directed from very early in infancy (von Hofsten Reference von Hofsten1980; Reference von Hofsten2004), and during the first year, manual skills become increasingly well-organized (Thelen et al. Reference Thelen, Corbetta and Spencer1996; von Hofsten & Ronnqvist Reference von Hofsten and Ronnqvist1988). For example, Claxton et al. (Reference Claxton, Keen and McCarty2003) demonstrated that infants reach for objects differently depending on what they intend on doing next: They are faster to reach for a ball if they are going to throw it versus place it into a container. Further, over the course of the first year of life, infants begin to systematically anticipate the shape, size, and orientation of the objects that they grasp (von Hofsten & Ronnqvist Reference von Hofsten and Ronnqvist1988). This body of work shows that motor competence even in young infants involves abstract action plans, as it does in adults (Rosenbaum Reference Rosenbaum1991). This fact about infants' actions has implications for the role that action experience might play in infants' perception of others' actions as organized by goals.

In fact, converging research has shown that infants also view others' actions as structured by goals. Infants encode others' actions in terms of the relation between agent and goal (e.g., Brandone & Wellman Reference Brandone and Wellman2009; Luo & Johnson Reference Luo and Johnson2009; Sodian & Thoermer Reference Sodian and Thoermer2004; Sommerville & Woodward Reference Sommerville and Woodward2005; Woodward Reference Woodward1998), selectively imitate the goals of others' actions (Gerson & Woodward Reference Gerson and Woodward2012; Hamlin et al. Reference Hamlin, Hallinan and Woodward2008; Meltzoff Reference Meltzoff1995), and anticipate the outcomes of others' actions based on their goals (Cannon & Woodward Reference Cannon and Woodward2012; Gredebäck et al. Reference Gredebäck, Stasiewicz, Falck-Ytter, Rosander and von Hofsten2009; Kanakogi & Itakura Reference Kanakogi and Itakura2010; Krogh-Jespersen & Woodward, under review). Moreover, across these findings, matched comparison conditions and fine-grained analyses of infants' attention during the tasks have shown that infants' responses reflect more than simply attention to physical movements or low-level associations between hands and objects. Instead, this body of evidence shows that infants analyze others' behavior in terms of the abstract relational structure that organizes goal-directed actions.

Importantly, infants' action understanding is related to and shaped by their action experience. The emergence of goal-directed actions in infants' own motor repertoires correlates with their analysis of these actions in others (e.g., Brune & Woodward Reference Brune and Woodward2007; Cannon et al. Reference Cannon, Woodward, Gredebäck, von Hofsten and Turek2012; Kanakogi & Itakura Reference Kanakogi and Itakura2011; Loucks & Sommerville Reference Loucks and Sommerville2012; Sommerville & Woodward Reference Sommerville and Woodward2005). Critically, interventions that change infants' own actions render changes in their analysis of others' action goals. For example, 3-month-old infants are not yet efficient at reaching, but, given training to use Velcro “sticky” mittens to apprehend objects, they subsequently demonstrate an understanding of others' reaches as goal-directed (Sommerville et al. Reference Sommerville, Woodward and Needham2005). Matched training that involves passively observing others' reaches does not have this effect (Gerson & Woodward Reference Gerson and Woodward2014; for related findings, see Libertus & Needham Reference Libertus and Needham2010; Sommerville et al. Reference Sommerville, Hildebrand and Crane2008). Thus, this body of work shows that infants' own experience producing goal-directed actions informs their understanding of the goals that structure others' actions.

Nevertheless, studies of infants have made only preliminary progress in the domain in which the MN hypothesis originated – neural processes. We agree with Cook and colleagues that a systems approach is needed to evaluate the functional relations that may be signaled by the firing properties of MNs. In human infant research, we have neither the precision of single-cell recordings nor (yet) an analysis of connectivity among potential components of the mirror neuron system (MNS). Even so, infants evidence neural activity in the motor system during observation of others' actions (Marshall et al. Reference Marshall, Young and Meltzoff2011; Nyström et al. Reference Nyström, Ljunghammar, Rosander and von Hofsten2011; Southgate et al. Reference Southgate, Johnson, Osborne and Csibra2009). Critically, changes in infants' motor experience modulate this neural response to others' actions. Developments in infants' motor skill affect the motor system's response to others' actions (Cannon et al., under review; van Elk et al. Reference Van Elk, van Schie, Hunnius, Vesper and Bekkering2008), and short-term manipulations of motor experience in infants generate similar effects (Marshall et al. Reference Marshall, Saby and Meltzoff2013; Paulus et al. Reference Paulus, Hunnius, van Elk and Bekkering2012). Yet, as Cook et al. point out with regard to the adult work, the critical connection between the MNS and social understanding has not been made for infants. Establishing this connection requires integrating neural techniques with behavioral methods for investigating social cognition in infants.

Cook et al. use a developmental framework to argue against over-interpretation of MN findings. While we see merit in their argument that many open questions still exist concerning the MNS, we also advise against throwing out the baby with the bathwater. Rather than using developmental arguments to minimize the potential significance of MNs for social cognition, the field should be pushing forward to understand the links between neural systems, social cognition, and motor skill. Because each of these systems undergoes rapid and dramatic change during early ontogeny, a developmental approach is likely to shed the most light on the links between them.