Cook et al. have done a great service by marshaling the mounting evidence that experience influences the operation of mirror neurons (MNs). This work reinforces other recent calls (e.g., Henrich et al. Reference Henrich, Heine and Norenzayan2010) to study the effects of variable environments, both in humans (across cultures) and in nonhuman primates (across rearing and field settings), on the development of neural and psychological systems. Cook et al. advance this issue with the hypothesis that specific aspects of developmental environments influence the ontogeny and functions of MN systems. Encouragingly, anthropologists have documented relevant population variation, including in adult interactions with infants (e.g., Ochs & Schieffelin Reference Ochs, Schieffelin, Shweder and Levine1984), and in interactive practices across cultural domains (Fiske Reference Fiske1992). Primatologists and comparative psychologists have also increasingly catalogued variation in social behaviors across natural populations (Whiten et al. Reference Whiten, Goodall, McGrew, Nishida, Reynolds, Sugiyama, Tutin, Wrangham and Boesch2001) and captive environments (Call & Tomasello Reference Call, Tomasello, Russon, Bard and Parker1996; Russell et al. Reference Russell, Lyn, Schaeffer and Hopkins2011). While the painstaking comparative groundwork is just beginning, it is hard to dispute Cook et al. on the value of the enterprise.

However, Cook et al. rely on a curiously impoverished evolutionary framework that likely hinders the empirical project they advocate. The authors do not clearly distinguish and relate types of biological causation (Laland et al. Reference Laland, Sterelny, Odling-Smee, Hoppitt and Uller2011). To be sure, they do clearly propose both a proximate/ontogenetic and an ultimate/functional explanation of MNs – MN development is driven by action-perception contingency learning, and is a by-product of associative learning mechanisms that bear no evidence of adaptive specialization for MNs. However, Cook et al. develop alternative approaches to this “associative” account that are arbitrary conjunctions of ultimate and proximate explanations. Their “genetic” account weds a nativist ontogenetic explanation to a specific ultimate function, “action understanding.” That this account is said to “combine” questions of origin and function, while the associative account “dissociates” them, is an artificial consequence of the “genetic” hypothesis as presented. The authors are mistaken when they state that, “If MNs were a genetic adaptation, it is likely that their properties would be relatively invariant across developmental environments” (sect. 9.1.1). Natural selection operates on developmental systems through the phenotypes they produce, and MN development could be largely experience-dependent while having a specific evolved function (see Barrett Reference Barrett2012). Likewise, MNs could be highly canalized and reliably developing, yet could have evolved for functions other than “action understanding,” including empathy (Gallese Reference Gallese2003; Preston & de Waal Reference Preston and de Waal2002) – a widely discussed and investigated ultimate hypothesis that the target article relegates to a footnote. The “genetic” hypothesis is at best an arbitrary hypothesis, and at worst a straw man.

In addition, contrasting “canalization” and “exaptation” as Cook et al. do has the unfortunate effect of obscuring interesting phylogenetic questions. While the authors argue that “the motivation for invoking exaptation is not compelling” (sect. 8.2, para. 1), they advocate the purest possible exaptation hypothesis – an ancient adaptation was coopted for producing MNs without any “secondary adaptation” (Gould & Vrba Reference Gould and Vrba1982). The authors go on to argue that any species capable of associative learning should be capable of developing MNs, and they predict controlled training regiments will produce MNs in lab rats. Intriguing, but this raises a question: Why do humans and macaques (and likely other species: Mancini et al. Reference Mancini, Ferrari and Palagi2013; Mui et al. Reference Mui, Haselgrove, Pearce and Heyes2008; Range et al. Reference Range, Huber and Heyes2011) “naturally” develop mirror neurons, while rats apparently do not? Is this a happy accident of variation in early developmental environments? Or has there been varying selection pressures across species for secondary adaptations that facilitate MN development (such as mother–infant face-to-face interaction; Ferrari et al. Reference Ferrari, Paukner, Ionica and Suomi2009b)?

In general, the target article is overly restrictive in discussing such canalizing mechanisms. Cook et al. discuss only one attention bias, infant self-observation of reaching. They overlook other biases, both in infants and in adults, that could facilitate MN development. For example, in humans the properties of infant-directed speech (Bryant & Barrett Reference Bryant and Barrett2007) and adult encouragement and imitation (Bornstein et al. Reference Bornstein, Tamis-LeMonda, Tal, Ludemann, Toda, Rahn, Pêcheux, Azuma and Vardi1992) show evidence of invariance across cultures, while infants show perceptual preferences for faces (Valenza et al. Reference Valenza, Simion, Cassia and Umiltà1996) and infant-directed motion (Brand & Shallcross Reference Brand and Shallcross2008). In addition, no mention is made of mounting evidence that associative learning mechanisms often evince domain-specific design features, such as preparedness to learn particular fitness-relevant contingencies (Barrett & Broesch Reference Barrett and Broesch2012; Garcia & Koelling Reference Garcia and Koelling1966).

Cook et al. also fail to appreciate the likely evolutionary consequences of MNs reliably developing in a species. Although ultimate and proximate explanations are classically considered orthogonal, they can be reciprocally related, such as when proximate mechanisms become part of the selective environment (Laland et al. Reference Laland, Sterelny, Odling-Smee, Hoppitt and Uller2011). Selection for behaviors and capacities that exploit MNs would be a clear case of this. Expressions of emotion may be a case in point. There is strong evidence that emotional contagion and mimicry of expressions are mediated by MNs (Carr et al. Reference Carr, Iacoboni, Dubeau, Mazziotta and Lenzi2003; Kircher et al. Reference Kircher, Pohl, Krach, Thimm, Schulte-Rüther, Anders and Mathiak2013; Wicker et al. Reference Wicker, Keysers, Plailly, Royet, Gallese, Rizzolatti and Aiguier2003) and have significant consequences for social competencies (Pfeifer et al. Reference Pfeifer, Iacoboni, Mazziotta and Dapretto2008) and relationships (Lakin et al. Reference Lakin, Jefferis, Cheng and Chartrand2003). Given both mirror neurons and selection for empathy and/or emotional coordination, evolution could well craft emotional expressions for facilitating shared emotion. For example, we have argued that laughter evolved as a medium for playful emotional contagion that taps MNs to facilitate mutually beneficial social play (Gervais & Wilson Reference Gervais and Wilson2005; see also Davila-Ross et al. Reference Davila-Ross, Allcock, Thomas and Bard2011). To the extent that human expressions of emotion are elaborated homologues of ancestral ape expressions (Parr et al. Reference Parr, Waller and Vick2007), one might implicate uniquely human selection pressures that attended ratcheting interdependence (Tomasello et al. Reference Tomasello, Melis, Tennie, Wyman and Herrmann2012). Further, the emergence of cultural evolutionary processes in humans (Boyd et al. Reference Boyd, Richerson and Henrich2011) may have created an additional consequence of MNs – selection for cultural practices (e.g., religious rituals; Atkinson & Whitehouse Reference Atkinson and Whitehouse2011; Fiske Reference Fiske1992) that facilitate MN development and build social bonds through them.

Even if mirror neurons develop through associative learning, they may be reliably developing adaptations. Evidence of secondary adaptation, including biological and cultural traits designed to exploit MNs, would be evidence of a history of such reliable development. The investigation of such derived traits should thus be integral to the mirror neuron enterprise.