Heyes (Reference Heyes2018) proposes that cultural evolutionary psychology (henceforth CEP), and with it, cultural evolution, underlies many human-specific cognitive mechanisms. To mark their cultural source, Heyes calls these mechanisms “cognitive gadgets.” A cultural source is certainly likely for some human phenomena (such as Heyes’ example of reading). It may also be correct for other mechanisms traditionally regarded as cognitive instincts. I am not completely convinced of all the aspects of CEP (yet?), but to foster readability, my comment will read as if I were already a full CEP convert.

Heyes discusses four cognitive gadgets that form the “mechanisms of cultural learning”: selective social learning, mindreading, language, and imitation. Here, I will focus on imitation (the copying of the form of an actionFootnote ¹). I fully agree with Heyes that imitation is logically required for (large) parts of human culture – specifically for culture based on actions (Heyes Reference Heyes2018; Tennie et al. Reference Tennie, Call and Tomasello2012).

Any claim for human-specific cognitive abilities benefits from a “control” comparison with humans’ closest living relative – that is, for nonhuman great apes (henceforth apes). Heyes (Reference Heyes2018) herself frequently mentions apes, but does not clearly say whether, in her view, apes spontaneously imitate or notFootnote ² and whether ape imitation would (have to) be due to an “imitation gadget.”

Finding spontaneous ape imitation – that is, without any human interference – would mean one of two things: (a) apes may then have a cognitive instinct to imitateFootnote ³ or (b) they, too, may have evolved their own variant of CEP – including an imitation gadget. Empirically, if apes spontaneously imitate in either of these ways, we should see at least two types of evidence: (1) Wild ape behaviour should show “smoking gun” signs of underlying imitation, and (2) captive apesFootnote ⁴ should not require human interference to show imitation. Does the current empirical data demonstrate these two patterns?

Imitation transmits the form of actions, automatically creating path-dependent differences over time (e.g., due to unavoidable copying error; Eerkens & Lipo Reference Eerkens and Lipo2005). This allows the detection of “smoking gun” signs of imitation: If wild ape cultures were based on imitation, we should see action form differences across time and between populations – for example, as different gesture sets/dialects. However, empirically, we find instead overwhelming similarity in gestural form across populations – and this extends even to captive populations (see analysis in Byrne Reference Byrne2016). The picture for ape material culture is more complicated but essentially the same: Although these behaviours are more likely to show differential frequencies across populations, the forms of also these behaviours neither require nor indicate imitation (e.g., Tennie et al. Reference Tennie, Call and Tomasello2009; Reference Tennie, Premo, Braun and McPherron2017).

What about captive apes? Unenculturated, apes consistently fail to imitate in controlled settings – where imitation would be the sole key to success (Clay & Tennie Reference Clay and Tennie2018; Tennie et al. Reference Tennie, Call and Tomasello2012; Tomasello et al. Reference Tomasello, Call, Warren, Frost, Carpenter and Nagell1997). After human training/enculturation (henceforth enculturation) apes can show clear (albeit often limited) evidence for imitation (Custance et al. Reference Custance, Whiten and Bard1995; Tomasello et al. Reference Tomasello, Savage-Rumbaugh and Kruger1993b). What explains this pattern is that unenculturated apes lack important brain structures for imitation but that human enculturation produces these structures in their brains (Pope et al. Reference Pope, Taglialatela, Skiba and Hopkins2018).

In sum, the best matched “control” for humans – apes – lack imitation as either an “ape cognitive gadget” and/or as a cognitive instinct. Whenever apes imitate, they do so because of human enculturation ape imitation is a gadget lent to apes by humans. This supports Heyes’ “wealth of the stimulus” argument (p. 46) regarding the cultural transmission of imitation.

Given this somewhat successful cross-species cultural transmission of a single (imitation) gadget, could apes also become a “model species” for the cultural evolution of whole CEP? For this, high levels of human enculturation would be necessary. In the past, this was sometimes achieved when studying ape capacity for language. However, high enculturation tends to have negative consequences for apes (Freeman & Ross Reference Freeman and Ross2014). And so, although we should therefore not repeat these studies, we can re-examine old data.

After a human “kickstart” enculturation process (the human transmission of one or more gadgets towards apes), the argument whether or not apes could ever sustain the continuous transmission of cognitive gadgets across pure ape generations (as a “stable ape CEP”Footnote ⁵) should depend in part on the fidelity with which cognitive gadgets can be passed from ape to ape after human influence has been removed. To examine this fidelity, an experiment would require an initial phase, where a “seed” ape population is enculturated so as to acquire – to the highest possible degree – one or more human cognitive gadgets. Then, after adding unenculturated apes (subjects) to the seed population, human enculturation efforts must cease both to seeds and subjects. Next, subjects need to be observed over extended time periods – even across ape generations, in case of initial success. Would the affected subjects “catch” cognitive gadgets from the seed population – and if so, how many and to what degree? Would these subjects later be able to enculturate additional subjects? There is exactly one long-term study with systematic data collection that came close to this hypothetical setup. This study introduced a single, unenculturated male chimpanzee – 10-month-old LoulisFootnote ⁶ – to a seed population of four sign-language trained conspecifics. During the 63 months that Loulis spent in this experimental setup,Footnote ⁷ he reportedly used 51 different signs. In addition, Loulis was claimed to have learned these signs (mostly) via ape-to-ape imitation (Fouts et al. Reference Fouts, Fouts, Van Cantfort, Gardner, Gardner and Van Cantfort1989). There are many reasons to doubt these and related claims (Rivas Reference Rivas2003; Reference Rivas2005), but even when taken at face value, a signing usage of 51 different signs after human-to-ape-to-ape transmission would already be substantially smaller than the claimed repertoire of the seed group after human-to-ape transmission (e.g., already at 36 months of age, one of the “seeds” (Washoe) was claimed to have mastered 85 different signs; Gardner & Gardner Reference Gardner, Gardner, Schrier and Stollnitz1971).

What is especially illuminating, of course, are the long-term effects of human-to-ape-to-ape transmission – did enculturation effects/gadgets persist in Loulis? When we look at the entire group's (seed plus Loulis) later performances (across four corpora)Footnote ⁸ the difference in persistence between subject and seed becomes highly apparent. In this dataset, the seed chimpanzees “imitated” (= responded to like with like) known signs much more frequently than Loulis did (around four times as often; Rivas Reference Rivas2003). That is, the human-installed ape imitation gadget seems to have already lost most of its power within the first generation of ape-to-ape transmission.Footnote ⁹ Equally important, whereas the seed chimpanzees still showed evidence of using between 38 and 55 different signs, Loulis now merely showed evidence for four different signs (Rivas Reference Rivas2003).Footnote ¹⁰ However, two of these “signs” (GIMME and HURRY) simply resembled species-typical behaviour (Rivas Reference Rivas2003). The third sign (THAT/THERE/YOU) involved “pointing” to various entities by way of extending the entire hand – which non–language-trained captive chimpanzees also do (Leavens & Hopkins Reference Leavens and Hopkins1999).

Finally, Loulis's fourth sign (CHASE) involved wrist-hitting actions using both armsFootnote ¹¹ – but a related gesture (Rivas Reference Rivas2005) also develops (and in a similar play function) in non–language-trained chimpanzees (wrist hitting another chimpanzee with one arm; Tomasello et al. Reference Tomasello, Gust and Forst1989). Note also that, at the time, Loulis's use of these four signs was studied by Rivas, Loulis had already additionally been exposed to several years of human-to-ape transmission (which included signing CHASE to him). And so, none of the “signs” Loulis persisted to use need be attributed to ape-to-ape transmission – and most should probably not even count as signs (except maybe one single sign; and even this sign perhaps only in part).

In sum, in terms of long-term effect of human-to-ape-to-ape transmission, Loulis demonstrated a grand total of zero signs that he clearly learned from the seed population. In addition, he also showed a relative lack of motivation to sign compared to the seed (human-trained) chimpanzees. Even his level of prompted “imitation” of seen gestures/signs was heavily reduced compared to the seed chimpanzees.

Overall then, the “Loulis experiment” – using a suitable methodological design – uncovered that even a short-chained human-to-ape-to-ape transmission of cognitive gadgets did not survive well. Any potential small gadget-residue in Loulis (~one sign plus weak imitation?) would likely fail to pass down to later ape-to-ape transmission steps. Thus, I must disagree with Gardner and Gardner (Reference Gardner, Gardner, Gardner, Gardner and Van Cantfort1989) who concluded that ape “sign language is robust and self-supporting” (p. 25). At least when chimpanzee seeds are human-trained to these levels in only these two cognitive gadget domains (communication and imitationFootnote ¹²) and when using a small seed/subject population, apes on their own seem unable to stabilise an ape CEP. The leakage of this system proved too large – apes still fall back into their cognitive baseline (i.e., to evolutionary psychology, rather than to CEP). The analogy here is filling an unclosed bathtub with water – that can be done, but the bathtub will empty itself as soon as the water supply stops or even shrinks.

The (theoretical) question remains as to whether humans could ever kickstart a stable ape CEP – and if so, how? In general, we should not forget that the only available “style” in which humans can enculturate themselves and other species could have evolved culturally to fit our and only our human biology (as a co-evolved bathtub-plug system). If we had the knowledge to devise enculturation ways specifically suited to ape biology, then perhaps an “artificial ape enculturation bathtub plug” could be transmitted to them, and then a stable ape CEP might succeed.

But, perhaps, we could get apes to evolve ape-specific enculturation styles on their own – via cultural evolution. For this, more and deeper cognitive gadgets important for cultural learning would have to be human-transferred to an ape seed population. It is not altogether clear how this could be done, but if it could, then at least some successful survival of gadgets across at least two ape generations could be induced. Once such multiple-generation ape-to-ape transmission happens, ape-specific enculturation styles could theoretically evolve culturally – perhaps even to a level able to stabilise ape CEP.

All the above is water on Heyes’ (Reference Heyes2018) mills. I also agree with Heyes that we need to engage in historical theory of human CEP. Clearly, in our own lineage, we must have evolved CEP ourselves entirely from scratch (likely in a feed-forward process). But when and why did our lineage evolve CEP? Knowing when not to look saves from misattributing factors. Given that imitation plays a crucial role, time periods with an absence of “smoking gun” (see above) evidence for imitation can therefore be dismissed. Using this logic, we recently found that the imitation gadget was likely absent from our lineage prior to ~500,000 years ago (compare Tennie et al. Reference Tennie, Braun, Premo and McPherron2016; Reference Tennie, Premo, Braun and McPherron2017). In accordance with Heyes (Reference Heyes2018, p. 212), we therefore state that imitation evolved late (roughly within the last 500,000 years).