That there is discontinuity as well as continuity in the (nonlinear) evolution of cognitive systems is hardly news. In the case of contemporary humans, it is also self-evident that a runaway culture of cognition has led to an exponential growth in knowledge way outside the time frame of classical Darwinian evolution. Externalised as in the third world of books (Popper Reference Popper1972), these knowledge-gaining systems are the product of a society of minds, public, open to scrutiny, repair, refutation, and (collective) redevelopment. These demand in turn extended periods of top-down instruction to secure for individuals what Piaget once dubbed the status of an “epistemic agent” capable of participation in the rational world of the sciences. Within this domain there is no possibility by definition of a nonhuman entry. And if this is the central thesis of Penn et al., then we have no disagreement with them.

However, despite the near impossible challenge of separating human cultural products from individual competences (Wundt Reference Wundt1898), the authors seem bent on assessing the “evolutionary” and cognitive status of the individual. Motivated by an adult end-state PSS (physical symbol system) -type model, they seek to explain discontinuity in terms of the individual's possession of a “supermodule.” Installed in a space between more primitive nonhuman forms of cognition (and even human natural language), such a module is used to claim a special biological uniqueness for humans (but not Neanderthals), allegedly with a representational capability that forever separates us even from our closest simian relatives.

One difficulty in following this thesis is that when espousing their case for human structural superiority, the authors veer between task criteria which are adult end-state, context free, and formal, such as “systematicity,” omnidirectionality, and “analogy” considered in isolation from content – and those which are embedded in “world knowledge” – such as functional analogy, theory of mind (ToM), and higher-order structural apprehension of perceived relations of similarity and difference. These heterogeneous domains are bridged with words such as “cognising,” “abstraction,” and “re-representation,” giving rise to the impression these all share the property of an in-the-head compliance between formal structures and individual agents.

This not only conflates private with cultural constructions as templates for the individual mind, it also ditches in the process those elements of human cognition regarded by many as core and normative, namely, commonsense reasoning, bounded rationality, choice transitivity, and subjective scales of judgement (based on adaptive value rather than truth) as well as other sources of knowledge derived from perception and action – all of which are subject to principled influences of learning and development. In contrast, Penn et al.'s own characterisation of human cognition both diminishes the role of development and eliminates completely the role of learning. This is despite the fact that many of the authors they cite are at pains to point out that the human competences they describe are often the product of many years of human development (Halford Reference Halford1993; Piaget Reference Piaget1970) and/or considerable explicit tuition (Kotovsky & Gentner 1996; Siegal et al. Reference Siegal, Varley and Want2001) within a physical and social environment. To take just two examples, cross-dimensional analogy even at the concrete level and involving only three levels along a relational dimension (greatest, middle, least) are not “made reliably” until 8 years of age unless specific training is provided (Kotovsky & Gentner 1996). Similarly, symbolic transitive inference (not transitive choice) is slow to develop in the human child, despite claims to the contrary by the authors (Chalmers & McGonigle Reference Chalmers, McGonigle, Smith, Dockrell and Tomlinson1997; Clark Reference Clark1969; Hunter Reference Hunter1957). The significance of the gradual acquisition of these core cognitive skills in humans for comparable studies in nonhumans is ignored by the authors in their review of equivalent comparative data. Where are the ground rules as suggested by, for example, Gentner, for training subjects to discount surface structure variation in favour of the underlying relational one?

And it is in this direction, that is, of “growing cognition” either in a model (Smith & Breazeal Reference Smith and Breazeal2007) or assessing it in long-term learning-based studies with children and nonhumans (McGonigle & Chalmers Reference McGonigle and Chalmers2006), that many now see the best prospect for providing an account of how we came to be what we now are (Griffiths & Stotz Reference Griffiths and Stotz2000). Clustered under the banner of “embodied cognition,” furthermore, the scope of what constitutes a cognitive system within such programs reaches far beyond the formal horizons limiting that of Penn et al. Here action plays a crucial role in cognitive regulation, pointing up some of their most serious areas of neglect. In particular, new programs of research on serial ordering mechanisms and executive control with primates (McGonigle & Chalmers Reference McGonigle and Chalmers2006; Terrace Reference Terrace2005b) open a window on the emergence of complex behaviors from simple relational primitives. These are already showing the growth of systematicity, compositionality, and the deployment of economic strategies with experience, which reduce cognitive costs (McGonigle et al. Reference McGonigle, Chalmers and Dickinson2003); as well as an upward bound trajectory with a momentum that has not so far been limited by any glass ceiling on simian achievements. It is the study of these action grammars in the nonhuman, in fact, that may well reveal the genesis of the way we plan and think (Bloom Reference Bloom2000), providing in turn a platform for the human linguistic system (Hinzen, in press). As for evolutionary discontinuities, promising candidates from this research are not supermodules, but rather, the instruments for cognitive externalization through the superior fine motor control in the human which enable complex gesture (Corballis Reference Corballis2002), articulate speech (Pinker & Jackendoff Reference Pinker and Jackendoff2005), and the ability to both manipulate objects and construct collections (Tallis Reference Tallis2003). In this way cognitive products are made available to agents as a result of their own activity and could prove crucial to the sharing of knowledge and cultural achievements (Donald Reference Donald1991).

Already with a high prescriptive value, we need to maintain the momentum of systematic developmental and comparative programs of this sort. In an exciting area still largely in a vacuum created more by experimental neglect than animal failures, this rush to judgement by Penn et al. will put this fragile yet exciting new comparative agenda at risk.