Osiurak and Reynaud (O&R) present a novel approach to the study of cumulative technological culture (CTC) by focusing on the role of technical reasoning skills rather than the traditional high-fidelity social-learning approach prevalent in the literature. The authors discuss several candidate cognitive abilities (e.g., imitation, theory of mind, and working memory (WM)) that are widely viewed as prerequisites for the emergence of CTC and have also been argued to differ between humans and other animals. The authors maintain that none of these abilities are necessary for the occurrence of CTC but instead are merely boosting elements, with technological reasoning skills being the real drivers of CTC. Here, we focus on two of these cognitive capacities: imitation and WM.

To strengthen their claim for technological reasoning being the pivotal ability supporting CTC rather than imitation and other mechanisms of the social dimension, the authors argue that chimpanzees possess imitative abilities but do not show evidence of CTC. Therefore, according to their argumentation, imitation cannot be a driving force of CTC. However, the authors have neglected to discuss crucial literature on imitation in chimpanzees that contradicts their arguments and would invalidate their conclusion about this ability in great apes. The authors base their argument that chimpanzees can imitate on the so-called “two-action tests,” which originally claimed to test whether chimpanzees would imitate the mechanical actions of a demonstrator who was trained in one of two techniques to open a puzzle box (Horner & Whiten Reference Horner and Whiten2005; Horner et al. Reference Horner, Whiten, Flynn and de Waal2006; Whiten et al. Reference Whiten, Horner and de Waal2005). The finding that subjects will generally converge on the technique of the demonstrator has often been taken as evidence that chimpanzees can imitate actions. However, these tests are methodologically flawed, mainly because the two different actions often lead to two different results, making it impossible to disentangle result from action copying. Therefore, these tasks would be better named “two-action, two-results tests,” as they do not allow differentiating between social-learning mechanisms. Finally, in all of these studies, subjects individually reinnovated the non-seeded action, demonstrating that these solutions are not beyond the spontaneous capabilities of each individual, and do not strictly require imitative skills to be acquired.

In fact, studies aimed at directly testing for imitation in chimpanzees have failed to find this ability (Henrich & Tennie Reference Henrich, Tennie, Muller, Wrangham and Pilbeam2017). For instance, Tennie et al. (Reference Tennie, Call and Tomasello2012) conducted an experiment with the goal to elicit imitation of motor actions in captive chimpanzees using conspecific demonstrators. The authors found that chimpanzees showed no evidence of action imitation despite ample opportunities for observational learning (see also Tennie et al. Reference Tennie, Call and Tomasello2010, for a different task). Similar negative results have been found across great ape species and experimental tasks (bonobos: Clay and Tennie Reference Clay and Tennie2017; all great apes: Tennie et al. Reference Tennie, Call and Tomasello2009). Therefore, we disagree with the authors' interpretation of the great ape imitation literature, and encourage the authors to reassess their line of reasoning in this regard.

Regarding WM, the authors argue against the possibility that an increase in WM capacity in human evolution could have been a key driver of CTC. They claim that an increase in WM capacity could not explain why humans – but not great apes – are capable of “combined mechanical actions.” The authors rightly argue that “maintain[ing] multiple mechanical actions in memory for a period of time is cognitively different from being able to combine them.” However, they might be wrong in assuming that WM only allows for storing mechanical actions, but not for combining them. Although no single, agreed-upon definition of WM exists, most researchers today refer to the mere temporary maintenance of information in mind as short-term memory (Cowan Reference Cowan2017). In contrast, WM is defined by many – including Wynn and Coolidge (Reference Wynn, Coolidge, Mellars, Boyle, Bar-Yosef and Stringer2007), which are cited by the authors – as the ability to maintain and manipulate mental representations, going beyond simple storage and allowing for information processing and interrelating representations (Cowan Reference Cowan2017; Diamond Reference Diamond2013). Therefore, the authors seem to conflate the concepts of WM and short-term memory. If taken as a storage and processing capacity, WM increases could allow for combinations of mechanical actions – there might be no need to assume separate technical reasoning skills.

The authors further argue that WM could not generate content and thus technical reasoning skills would be necessary to do so. However, note that WM is thought to play a crucial role in the creation of novel procedures or tools: greater WM capacity allows one to “see novel combinations of tools and materials or look farther ahead in the procedural sequence to see novel consequences” (Wynn & Coolidge Reference Wynn and Coolidge2014).

One might even argue that technical reasoning itself requires a considerable amount of WM to evolve. For example, the authors suggest that the “outcome of technical reasoning is a mental simulation of the mechanical action […] and this simulation needs to be temporarily maintained.” WM – by definition – is precisely the platform needed for storing and simulating such mental content. Second, the authors describe technical reasoning as being “analogical.” Analogical reasoning itself (potentially also in the technical domain) relies critically on WM, as it requires one to hold multiple representations and their relational information in mind (Halford et al. Reference Halford, Wilson and Phillips1998; Simms et al. Reference Simms, Frausel and Richland2018). Thus, the authors' arguments against the role of WM for CTC seem weak. WM as a potential key factor for CTC should not be prematurely ruled out.

In conclusion, we agree with the authors that the role of technological reasoning skills in CTC has been underestimated, and the role of imitation perhaps overestimated. However, by neglecting to engage with the most recent developments in the field of great ape social learning, and by basing some of their conclusions about WM on weak arguments, the authors' current formulation of the technical reasoning hypothesis loses some credibility. Despite this, we applaud their efforts toward questioning the status quo of the literature on the cognitive mechanisms behind CTC, and look forwards to reading about experimental applications of their hypothesis.