Cook et al. present a compelling case that mirror neurons (MNs) have a developmental origin in associative learning. Moreover, they legitimately argue that empirical testing is required to determine whether MNs and mirror systems have evolutionary origins as adaptive specializations, echoing criticism of adaptationist “just-so” stories in other fields (Pigliucci & Kaplan Reference Pigliucci and Kaplan2000). Here, I discuss whether work on mirror systems can be informed by, and inform, the fields of social information use and social learning. I leave aside discussion of communicative signals, by definition adaptive specializations.

Many animals use social information (information provided by other individuals) and social learning (learning from this information; Reader & Biro Reference Reader and Biro2010). Debate over mirror system origin and function can be viewed as part of a broader debate over the origins of a reliance on social cues and of the mechanisms underlying social information use, a debate Heyes (Reference Heyes1994; Reference Heyes2012a; Reference Heyes2012c) has also championed. Besides the fact that mirror systems utilize social information, there are numerous points of intersection between the two research fields. Mirror systems have been proposed to underlie various forms of social learning, including stimulus enhancement, emulation, and imitation learning (Byrne Reference Byrne2002; Keysers & Perrett Reference Keysers and Perrett2004), and such systems could potentially associate personal and conspecific location and thus also underlie local enhancement. Social learning propensities, biases, and processes have been proposed to be products of general learning processes, in a similar fashion to the Cook et al. proposal (Church Reference Church1957; Heyes Reference Heyes1994; Reference Heyes2012c; Keysers & Perrett Reference Keysers and Perrett2004; Laland & Bateson Reference Laland and Bateson2001; Leadbeater & Chittka Reference Leadbeater and Chittka2007; Miller & Dollard Reference Miller and Dollard1941). Furthermore, like mirror systems, the assumption that social learning is an adaptive specialization has been questioned, as has whether any such adaptive specialization would involve input systems rather than the learning mechanisms themselves (Caldwell & Whiten Reference Caldwell and Whiten2002; Heyes Reference Heyes2012c; Lefebvre & Giraldeau Reference Lefebvre, Giraldeau, Heyes and Galef1996; Reader et al. Reference Reader, Hager and Laland2011). These points of intersection suggest the two fields may usefully inform each other.

Experiential effects on the propensity to use and learn from social information have been demonstrated in several species (Kendal et al. Reference Kendal, Coolen, Laland, Dukas and Ratcliffe2009), supporting the idea that responses to social cues can be learned. However, flexibility alone is insufficient to demonstrate that the value and meaning of social cues are acquired by learning, since flexibility could be genetically encoded. For example, individuals could follow evolved unlearned rules-of-thumb of when, where, and how to employ social information (Rendell et al. Reference Rendell, Fogarty, Hoppitt, Morgan, Webster and Laland2011). Direct manipulation of the benefits of social information provides superior evidence for learned biases in social information use. For example, sparrows raised with an artificial parent that had reliably indicated food were more likely to approach feeding conspecifics than if the parent had not reliably indicated food (Katsnelson et al. Reference Katsnelson, Motro, Feldman and Lotem2008). Similarly, in finches manipulation of the net benefits of attending to others resulted in changes in individual tendencies to use social information, with lags that suggested the birds were learning the optimal response on the basis of received rewards (Mottley & Giraldeau Reference Mottley and Giraldeau2000). Perhaps the most compelling current evidence for associative learning shaping social information use involves the acquisition of matching and nonmatching responses during social learning. Dawson et al. (Reference Dawson, Avargues-Weber, Chittka and Leadbeater2013) trained bumblebees in a feeding array where conspecific “demonstrators” indicated either the presence of sweet sucrose or bitter quinine. Bees thus readily learned to approach or avoid conspecifics. Later, the bees observed demonstrators at one color of flower in a two-color array. Bees previously rewarded for approaching conspecifics were more likely to choose the same color as demonstrators, whereas the reverse was true in the quinine-trained bees. Such data strikingly parallel mirror and counter-mirror effects observed in budgerigars and dogs (Mui et al. Reference Mui, Haselgrove, Pearce and Heyes2008; Range et al. Reference Range, Huber and Heyes2011).

Although these examples provide evidence that experience can shape reliance on social cues, interpreting all individual and between-species variation in social information use as the result of prior learning would risk telling associationist “just-so” stories. Studies of the evolution of learning provide useful insights into when an evolutionary account may explain variation in social information use (e.g., Boyd & Richerson Reference Boyd and Richerson1985; Dunlap & Stephens Reference Dunlap and Stephens2009; Johnston Reference Johnston1982). When opportunities for learning are limited, learning or errors are costly, or the optimal response to a social cue is highly predictable, natural selection could shape genetically encoded predispositions to respond in a certain manner to particular social cues. Similarly, if experiences in early life predict later payoffs of social information, and there are costs to learning, early life experience may result in fixed social learning tendencies during adulthood (Lindeyer et al. Reference Lindeyer, Meaney and Reader2013).

There are several instances where responses to social cues appear fixed. Cases such as humans copying the most successful individual even when this is suboptimal (Offerman & Schotter Reference Offerman and Schotter2009), birds ignoring reliable asocial information to copy conspecifics (Rieucau & Giraldeau Reference Rieucau and Giraldeau2009), and the aforementioned counter-mirror effects taking longer to develop than mirror effects could all be the result of the extensive social experience individuals have prior to testing. However, other examples are more difficult to explain in terms of experiential effects. For example, several avian species use conspecific and heterospecific nesting or breeding success during their own habitat selection without clear opportunities to learn to use these cues (although experience can shape later choices; Morand-Ferron et al. Reference Morand-Ferron, Doligez, Dall, Reader, Breed and Moore2010). Restrictions on the stimuli monkeys and warblers socially learn about are also consistent with an adaptive specialization account (Davies & Welbergen Reference Davies and Welbergen2009; Mineka & Cook Reference Mineka, Cook, Zentall and Galef1988, but see Heyes Reference Heyes1994). Such predispositions would reduce errors during social learning, rather like predispositions to attend to conspecifics protect young birds from errors during filial imprinting (Horn Reference Horn2004).

Given that predispositions are expected for certain forms of social information use, the unconstrained flexibility of mirror systems that Cook et al. note raises two possibilities, assuming that mirror system efficiency is a determinant of fitness. Either (1) mirror system flexibility is vital to their adaptive function, suggesting that social cues have variable meanings that must be learned, or (2) evolved alternatives to associatively acquired mirror systems are constrained, perhaps by their cost. The broad affordances of associative learning may mean that beneficial mirror systems come virtually “for free,” reducing the likelihood of alternate evolved solutions.

If mirror systems and social learning tendencies are the products of general learning processes, the evolution of social and general intelligence may be closely entwined (Brown & Brüne Reference Brown and Brüne2012; Dunbar & Shultz Reference Dunbar and Shultz2007; Reader et al. Reference Reader, Hager and Laland2011). Furthermore, because associative learning and social information use are phylogenetically widespread, mirror systems could be studied in species such as insects, where evolutionary studies could examine the related but separate questions of adaptiveness and adaptive specialization. Particularly informative would be studies of species where deviating from group behavior carries strong costs, such as certain fish (Bates & Chappell Reference Bates and Chappell2002). Studies of links between mirror system efficacy and behavioral competence are essential, ideally by measuring costs and benefits for fitness in “real-world” group situations to establish when individuals out-compete or interact more effectively with others. Cook et al. present a parsimonious model that inspires broad application and testing of mirror system concepts.