C&C propose that cultural complexity arises through social tinkering across six stages, with linguistically formulated representations entering the picture only at Stages 5 and 6. The ordering misrepresents what enculturation does to agents because, according to cognitive integration (Menary, Reference Menary2007a, Reference Menary, Newen, de Bruin and Gallagher2018), learning the public representational practices that C&C reserve for the higher stages reorganises cognitive capacities at earlier stages. This means the higher stages cannot be reached without representationally mediated cognitive transformations at earlier stages. Once this point is recognised, the alignment story C&C offer for cultural learning in section 4.1 cannot do the work they ask of it, and the strict sequencing of stages gives way to a recursive picture in which representational practice and cognitive transformation co-construct each other.

C&C model enculturation as a coordination problem solved through Schelling-style focal points (section 4.1, p. 10). Cultural rules feel natural to new learners because previous generations of learners with the same cognitive apparatus shaped those rules into salient solutions. The argument requires that cognitive capacities stay fixed across generations, since otherwise the focal points selected by past learners would not align with the salience structure of present ones. Holding the apparatus fixed lets C&C treat enculturation as a tractable extrapolation problem within a stable representational space. By contrast, we emphasise enculturation as the learning of normative patterned practices that transform the cognitive capacities of individuals and groups, entailing novel ways of thinking and conceptualising which in turn generate modes of interaction that would not otherwise be possible (Menary, Reference Menary2007a, Reference Menary, Newen, de Bruin and Gallagher2018; Menary & Gillett, Reference Menary, Gillett and Kiverstein2017).

We are not convinced that empirical work on cognitive practices supports the assumption that cognitive capacities stay fixed. For example, reading reorganises the neural architecture of the reader through cortical recycling/retraining, recruiting visual word form areas previously dedicated to other functions (Dehaene, Reference Dehaene2009; Menary, Reference Menary2014). Similarly, mathematical training restructures the representational capacities of the practitioner so that quantitative relations become perceivable in ways unavailable to the untrained agent (Fabry, Reference Fabry2020; Menary, Reference Menary2015). But this restructuring does not only happen at Stages 5 and 6; it is not a purely linguistic representational shift. Cortical retraining is a shift in function due to a change in cognitive practices for manipulating representations. These cognitive practices and the representations shift over time as the practices and representations become more abstract, or general.

Let us take C&C’s example of writing. C&C treat writing at Stage 5 as a technology for codifying social rules that previously existed as informal norms, with literacy intensifying the codification process (p. 20). The framing follows Donald’s (Reference Donald1993) account of writing as an external memory system, where the inscription medium supplements internal memory and extends communication across time. However, cognitive integration recognises this function of writing whilst denying that it exhausts what writing does. Primarily because writing becomes a medium for representing abstract thinking, but also a medium in which that thinking is done (Menary, Reference Menary2007b).

The invention of writing from proto-writing systems has occurred at least three times: in Mesopotamia prior to 3000 BCE, Ancient China prior to 1000 BCE, and Mesoamerica around 500 BCE (Houston, Reference Houston2004). We focus on the Mesopotamian transition from arithmetic token to script (Overmann, Reference Overmann2021; Schmandt-Besserat, Reference Schmandt-Besserat1996). From around 7500 BCE, communities in the Near East tracked goods using small tokens of distinct shapes, each commodity having its own counters. These tokens are thought to have been an object-specific counting system. Around 3700 BCE, tokens were sealed inside clay envelopes with impressions of the tokens marked on the outside before sealing, which over the next century allowed the impressions to replace the enclosed tokens and the envelopes to flatten into tablets. By around 3100 BCE, pictographic script appeared on tablets in which the symbol for an item was paired with a separate numeral indicating quantity. The cognitive change here is not reducible to externalisation. Object-specific counting and abstract counting are different cognitive operations, and the second became available only when the representational medium had been reworked through millennia of social tinkering that took the token through enclosure, impression, and eventual replacement by the tablet. This is not to say that non-literate cultures lack advanced mathematics beyond counting (Overmann, 2021). But these abstractions did not exist as cognitive operations awaiting linguistic codification at Stage 5 in C&C’s framework. They came into being through the development of cognitive practices around evolving representational tools, which transformed what the practitioners could think. Therefore, the important cognitive transformations happened earlier and were built upon in each transition. One can also find similar historical accounts of how a capacity for abstraction plays out through the gradual historical refinement of cognitive tools in wayfinding technologies (Menary & Gillett, Reference Menary and Gillett2022; Uttal, Reference Uttal2000) and systems of measurement (Cooperrider & Gentner, Reference Cooperrider and Gentner2019).

C&C invoke a virtuous cycle between cultural complexity and brain size operating across hominin evolution (p. 30) without addressing how the cycle plays out within a single lifetime. Cognitive development supplies the mechanism that C&C leave implicit. The distinctiveness of human developmental plasticity within engineered learning environments is key to understanding the importance of cognitive evolution (Sterelny, Reference Sterelny2003; Stotz, Reference Stotz2010). Children acquire cognitive practices through extended apprenticeship in niches designed by previous generations to support that acquisition, with the plasticity making transformation possible and the niche channelling its direction.

The implication for the six-stage scheme is that Stages 5 and 6 cannot be reached by agents whose cognitive capacities have not already been reshaped by enculturated practice with the representational systems those stages deploy. Stage 6 ratcheting in mathematics presupposes mathematical cognitive practices that themselves required Stage 5 representational technologies to come into being, which in turn required the cognitive transformations that mastery of those technologies produces. The stages form loops rather than a strict ordering, with cognitive transformation depending on representational practice and representational practice depending on cognitive transformation. C&C are right that complex cultural systems emerge from local human action without overarching design. However, they underestimate how thoroughly that action has reworked the human minds that perform it.