Anderson's (Reference Anderson2014) After Phrenology makes an authoritative and timely contribution to the literature by describing two key developmental mechanisms (i.e., neural reuse and, more speculatively, interactive differentiation and search) that have important implications for our understanding of the brain. In particular, we applaud his provision of a rigorous, empirically informed treatment of the functional diversity and overlap of different brain regions that also explains how higher-order, domain-general cognitive capacities emerge from evolved neural circuitry. Arguably, these contributions call massive modularity into serious question. Like many others, we agree with Anderson that in light of the evidence, this view can no longer be reasonably sustained. As such, we think that one of the chief virtues of Anderson's work is that it brings us one step closer to the abandonment of massive modularity as a model of the brain. Naturally, any appeal for greater collaboration across the cognitive sciences should also be lauded (particularly one accompanied by such clearly specified and promising research directions), and Anderson's broader treatment of the functional properties of the brain is comprehensive, scholarly, and elucidating.

That being said, Anderson also seems to tell us little new. Buller (Reference Buller2005), for example, forwarded neural plasticity as an important developmental mechanism that accounts for domain-general capacities in his critique of massive modularity a decade ago, and as Anderson readily admits, he is not the first to focus on neural reuse to explain how evolved neural mechanisms come to perform new cognitive functions. Similarly, cognitive scientists have been emphasizing the complexity of the brain and its development, the functional integration of neural regions, the need to use sophisticated dynamical methods to analyze such complexities, and the pitfalls of modularity for many years now (e.g., Fotopoulou Reference Fotopoulou2014; Karmiloff-Smith Reference Karmiloff-Smith1992; Kelso Reference Kelso1995; McIntosh Reference McIntosh2000; Mesulam Reference Mesulam1990; Meunier et al. Reference Meunier, Lambiotte and Bullmore2010; Pfeifer & Allen Reference Pfeifer and Allen2012; Price & Friston Reference Price and Friston2002). As a consolidating synopsis of disparate research programs, After Phrenology is undoubtedly impressive, but it arguably falls short of constituting the inception of a new paradigm – Anderson suggests that we use his treatise as an impetus to “get to work,” but cognitive scientists have already been undertaking such work for quite some time.

More critically, although we certainly agree that the brain can be seen as an action control system, we are nonetheless skeptical of Anderson's claim that perception is not reconstructive. In particular, highly influential predictive coding approaches in neuroscience call his action-oriented framework into question by placing perception center stage (or at least on an equal footing with action). Take, for example, Friston's (Reference Friston2005; Reference Friston2010) free-energy principle, which asserts that the brain instantiates a biological imperative to model the world. According to this perspective, perception and action operate synergistically to minimize prediction errors and optimize an individual's internal representations of the environment. A key corollary of this hypothesis is active inference: the idea that all behavior can be understood in terms of the selective sampling of sensory data so that we experience what we expect to (in order to avoid surprises). That is, we act upon the world to ensure that our predictions are self-fulfilling (Friston et al. Reference Friston, Daunizeau and Kiebel2009; Reference Friston, Daunizeau, Kilner and Kiebel2010). Of course, one way to interpret this view is to suppose that action subserves perception, but at the very least, it cautions against definitively attributing causal primacy to it.

Notably, there are times when Anderson also appeals to more of a synergistic relationship between action and perception, and he discusses ideas that are highly reminiscent of predictive coding such as perceptual hypothesis generation (e.g., Gregory Reference Gregory1980). As such, we imagine that Anderson has a clear idea about how predictive coding accounts align with his own view, or – considering his professed aversion both to global brain theories and the notion of reconstructive perception – he might be inclined to reject them outright. In any case, given the widely recognized explanatory power of these models (e.g., Clark Reference Clark2013b; Hohwy Reference Hohwy2013) – and their capacity to accommodate the sort of phenomena that Anderson calls upon to substantiate his own view (e.g., the size–weight illusion; see Clark Reference Clark2013a; Lupyan Reference Lupyan2015) – his failure to explicitly address them renders his conclusion that we should privilege hypotheses of action control somewhat questionable and premature. For our part, we wonder why Anderson feels the need to advocate such a polarizing view – the way we see it, given insufficient evidence to the contrary, one should assume more of a dialectical, reciprocally causal relationship between action and perception.

This brings us to our greatest reservation about Anderson's work – its cursory and misleading construal of evolutionary psychology. Of particular note, his assertion that species-typical cognitive patterns emerge from the repeated assembly of reliably recurrent developmental resources was proposed by an evolutionary psychologist at the turn of the century (Caporael Reference Caporael1997; Reference Caporael2001) and has been echoed by many others since (e.g., Badcock Reference Badcock2012; Frankenhuis et al. Reference Frankenhuis, Panchanathan and Barrett2013; Geary & Bjorklund Reference Geary and Bjorklund2000; Kenrick Reference Kenrick2001; Lickliter & Honeycutt Reference Lickliter and Honeycutt2003). Given his critique of massive modularity – which he appears to erroneously equate with the broader paradigm of evolutionary psychology itself – we think it would only be fair to acknowledge the theoretical heterogeneity of the field, not to mention its provision of a similar view of the evolved brain (and the development of its functional properties) long before him.

Anderson's underestimation of evolutionary psychology is particularly apparent in his treatment of massive modularity. Like most critics of this view, he fails to distinguish between massive modularity as an explanatory account of the properties of the brain and its methodological value as a research heuristic. Regardless of the veracity of massive modularity, evolutionary computational theories continue to guide research in a systematic and highly productive way, providing a wealth of insights into the adaptive significance of mental processes and behavior (e.g., Buss Reference Buss2012; Crawford & Krebs Reference Crawford and Krebs2008; Dunbar & Barrett Reference Dunbar and Barrett2007; Ellis & Bjorklund Reference Ellis and Bjorklund2005; Pinker Reference Pinker1997). Indeed, the convenience of the phenotypic gambit is that it can produce substantive, testable hypotheses of behavior without demanding recourse to a mechanistic explanation of how it occurs. In other words, evolutionary psychologists chiefly concentrate on ultimate or functional analyses of cognitive and behavioral patterns, not their neurobiological instantiation (Klasios Reference Klasios2014). With this in mind, the pitfalls of massive modularity by no means vitiate evolutionary psychology.

Consider the illustrative case of attentional biases toward threatening stimuli. For example, research on snake detection shows that both humans (Masataka et al. Reference Masataka, Hayakawa and Kawai2010) and other primates (Shibasaki & Kawai Reference Shibasaki and Kawai2009) display faster reaction times when detecting a snake in an array of neutral pictures, relative to a neutral picture in an array of snakes. Such findings resonate with a wealth of studies suggesting that humans exhibit a perceptual bias toward evolutionarily recurrent threats; biases that occur early in infancy – well before the acquisition of threat-relevant fears – and are likely to play a causal role in privileging fear learning for certain stimuli (LoBue & Rakison Reference LoBue and Rakison2013). Note, too, that this need not denote a dedicated “threat-detection” or “snake-detection” module. Although some researchers have certainly suggested as much, others have left the mechanisms responsible for this bias open to question – this has not, however, prevented them from emphasizing its adaptive significance (see LeBou & Rakison 2013). Indeed, to an evolutionary psychologist, such phenomena make clear sense. If, however, we were to follow Anderson's suggestion by abandoning this approach, we would be hard pressed to conjure an equally parsimonious explanation for them.

By contrast, Anderson's own suggestions for research seem overly neurocentric – advocating a rather nonsubstantive, descriptive approach to cognitive science. He tells us what cognitive scientists should look for – and provides excellent advice on how to go about it – but says very little about what, exactly, we should expect to find. The idea that distinct patterns of functional neural activity load on to “neuroscientifically relevant psychological (NRP) factors” (sect. 6, para. 4) is self-evidently true, but it also provides few insights into what these factors actually are. We therefore question the capacity of his approach to provide substantive hypotheses about overt behavior – precisely what one would expect from an action-oriented psychology. For this reason, although Anderson's work will no doubt prove useful to cognitive neuroscientists, we strongly suspect that most psychological scientists will find evolutionary psychology far more fruitful than his own scheme.

Ultimately, although Anderson's contribution is clearly important, some of his central claims warrant skepticism, and it is far from revolutionary. In particular, he recapitulates a perspective of the evolved brain that originated from the very paradigm he tries to refute, suggesting that evolutionary psychology and his own view might be a lot more compatible than he thinks. By our estimation, once we set a massively modular view of functional neuroanatomy aside, evolutionary psychology and Anderson's approach have the potential to be highly complementary. The former addresses the ultimate “why” questions of psychological science by elucidating the adaptive significance of cognition and behavior; the latter, mechanistic account can help us explain “how” these adaptive patterns are neurobiologically realized. As evolutionary psychologists have long been telling us (e.g., Badcock Reference Badcock2012; Barrett Reference Barrett2008; Dewsbury Reference Dewsbury2009; Frankenhuis et al. Reference Frankenhuis, Panchanathan and Barrett2013; Kenrick et al. Reference Kenrick, Maner, Butner, Li, Becker and Schaller2002; Ploeger et al. Reference Ploeger, Van Der Maas and Raijmakers2008; Scott-Phillips et al. Reference Scott-Phillips, Dickins and West2011), these distinct levels of explanation can be mutually informative and reinforcing. In short, although we entirely agree that cognitive neuroscientists should hunt for NRP factors, we think that evolutionary psychology is one of the most promising places for them to look.