R2.1. Condition 1: Probabilistic use of enduring media

The first requirement for reasoning from absence here is the probabilistic use of enduring materials. As discussed in section 10.2 of the target article, many artefacts which may contain enduring media need not do so. However, in Sterelny's hafting example, this condition is well satisfied: Hafted tools are typically weapons, or tools of manufacture/butchery which, by dint of their function, often incorporate hard, enduring media (target article, sect. 12). Thus, if hafted tool-heads were produced at scale, they should appear in any well-sampled assemblage (as perhaps they do, see Wilkins et al., Reference Wilkins, Schoville, Brown and Chazan2012) and their absence is informative. Modified ochre is another enduring material, so should also be archaeologically visible were it widely used. As Ben-Oren et al. highlight, its patchiness throughout the human record may indicate only situational importance. Here, again, negative evidence proves informative.

Note though, that this condition is met only for a small subset of total material repertoires even among contemporary African foragers. The data presented in the target article show that, excluding traded media, an estimated 92% of Mbuti, 83% of Hadza, and 76% of G//ana artefactual repertoires contained no enduring materials. Focusing only on that minority of artefacts where absence is informative may seriously blinker us to the potential complexity of both Pliocene and Pleistocene hominin technology (Pascual-Garrido & Almeida-Warren, Reference Pascual-Garrido and Almeida-Warren2021). This is central to commentaries from Falk, Tzafestas, and Gallup & Eldakar, who highlight that coordinated hunting strategies (also see Boyd & Richerson, Reference Boyd and Richerson2022; Lang & Kundt, Reference Lang and Kundt2023; Stibbard-Hawkes, Reference Stibbard-Hawkes2023), children's play objects (also see Lew-Levy, Andersen, Lavi, & Riede, Reference Lew-Levy, Andersen, Lavi and Riede2022a) and organised team-competition (Turnbull, Reference Turnbull, Leacock and Lee1982) would each be traceless. Falk makes the case strongly, arguing that during the period of hominin evolution proceeding the Lomekwian (i.e., the “botanic age”; Falk, Reference Falk2024), we should not assume technological nor cognitive stagnation despite the paucity of material evidence. Falk highlights the credible importance of infant-carrying slings among early bipedal hominins (see also Suddendorf, Kirkland, Bulley, Redshaw, & Langley, Reference Suddendorf, Kirkland, Bulley, Redshaw and Langley2020; Wall-Scheffler, Geiger, & Steudel-Numbers, Reference Wall-Scheffler, Geiger and Steudel-Numbers2007; and, for critical discussion, Sterelny, Reference Sterelny2021a). Slings are routinely manufactured from perishable media by both the Hadza and Mbuti for good reason – soft, pliable materials like leather and cloth are much better suited to carrying wriggly infants than are rigid bone or stone vessels. The absence of evidence for soft containers before 50,000 BP (reviewed Suddendorf et al., Reference Suddendorf, Kirkland, Bulley, Redshaw and Langley2020) is therefore inferentially meaningless.

Numerous other important categories of material evidence suffer similar issues. For instance, Sterelny also proposes “Late Pleistocene hominins [probably] did not live in towns and villages (or work metal); a secure conclusion based on lack of evidence for settled lives (or metal working).” This appears true of metalworking and stone buildings, and also of large, densely populated towns which may leave signature patterns of mortuary remains and refuse. Concluding that settled lives were absent during the Pliestocene seems less sure. Many once-temporary Hadza camps have today become permanently occupied, characterised by wooden dwellings that use similar construction methods but are larger than traditional Hadza huts. These buildings do not incorporate earthworks or storage pits, and it is unclear whether, at greater time-depths, such settlements would be discernibly different from large, seasonally occupied camps. Indeed, there are multiple instances of sturdy wooden dwellings from Late Pleistocene Eurasia suggestive of settled or, minimally, semi-settled lives (al-Nahar & Olszewski, Reference al-Nahar and Olszewski2016; Maher, Reference Maher2019; Maher & Conkey, Reference Maher and Conkey2019). The recent discovery of apparent joinery (Barham et al., Reference Barham, Duller, Candy, Scott, Cartwright, Peterson and Nkombwe2023) from 476,000 BP might suggest semi-permanent wooden structures in the Chibanian (and see Singh & Glowacki, Reference Singh and Glowacki2022). Determining settlement, occupation and mobility patterns from ancient remains is complex (Maher, Reference Maher2019; Padilla-Iglesias & Bischoff, Reference Padilla-Iglesias and Bischoff2024) and beyond the scope of this reply. Yet all structures in the target article dataset are botanic, and good ethnoarchaeology exploring mobility-patterns (Padilla-Iglesias & Bischoff, Reference Padilla-Iglesias and Bischoff2024) and the different signatures left by semi-sedentary and sedentary camps (e.g., Kent, Reference Kent, Kroll and Price1991; Yellen, Reference Yellen1977, Reference Yellen1991), may provide cause for caution in reasoning from absence.

R2.2. Condition 2: A probabilistic link between capacity and expression

It is Sterelny's second claim about hafting, however, which I take more serious issue with: That its absence provides evidence that H. heidelbergensis lacked “the capacity to haft tools.” Sterelny argues that if a capacity to make hafted tools were available, there would be a strong incentive to use them. “Capacity” can be construed in different ways and could simply mean “know-how.” However, following the target article and Haidle's (Reference Haidle, Haidle, Conard and Bolus2016) distinction between performance versus capacity (discussed by Will), I take it to mean intrinsic or underlying faculties; or some extension of cognitive faculties resulting from, for example, enmeshment in denser social networks (discussed in Sterelny, Reference Sterelny2011, Reference Sterelny2021b).

If so, with hindsight, it is easy to overestimate the likelihood that even “modern” minds would reinvent and retain particular technologies. Hafting not only requires knowing how to attach a stone to a stick, but knowing that it can be done and why it would be useful. The knowledge to create and efficiently use weapons technologies can be lost and, as Milks reminds us, hafting is not universal. Indeed, even “simpler” vital technologies have permanently left the repertoires of modern humans; fire-making requires two dry sticks and tinder (the hand-drill method), yet a recent ethnographic review revealed five cases where it was lost and not reinvented (McCauley, Collard, & Sandgathe, Reference McCauley, Collard and Sandgathe2020). It is no surety that living Homo sapiens with the demonstrable capacity to produce and use hafted tools would probabilistically reinvent them. It is difficult to assess what cumulative cultural mountain ranges (see Lombard, Reference Lombard, Haidle, Conard and Bolus2016) must be navigated to invent hafting, or how this probability is influenced by social contexts (Sterelny, Reference Sterelny2021b).

Weakening (perhaps breaking) the inferential link between capacity and production also weakens the probabilistic link between negative evidence and inference. This has broader implications. Will highlights the need for carefully constructed bridging theories, citing Coolidge, Haidle, Lombard, and Wynn (Reference Coolidge, Haidle, Lombard and Wynn2016) as a blueprint. I strongly endorse this suggestion, with the caveat that when addressing capacity differences, we must be explicit and cautious about inferences we are drawing from negative evidence. Indeed, Coolidge et al. (Reference Coolidge, Haidle, Lombard and Wynn2016) provide a case study. They investigate the distinct cognitive systems involved in the creation of hunting bows, linking them to episodic and working memory. Like Sterelny they also cast doubt on the capacity of H. heidelbergensis who, they contend, “probably could not have invented a bow and arrow, and almost certainly could not have conceived of, and organised, the entire system” (p. 224). However, beyond tentatively exploring H. sapiens comparative endocast data (Bruner, Reference Bruner2010; Schacter & Addis, Reference Schacter and Addis2007), they provide little evidence that H. heidelbergensis lacked these requisite capacities. The tacit implication (made explicit elsewhere; Fajardo, Kozowyk, & Langejans, Reference Fajardo, Kozowyk and Langejans2023; Wadley, Reference Wadley2021) appears to be that technological difference indicates capacity difference, and that H. heidelbergensis lacked cognitive prerequisites for bow-making because they lacked bows. In building future bridging theories, we must make any inferences from absence explicit – perhaps even constructing negative bridging theories. We must not over-rate the probabilistic link between absence and capacity. We must not underrate the intelligence of past humans where evidence (positive or negative) is lacking. To achieve these aims, one path forward is changing our null model.

R3.1. Balancing probabilities: The null model is a comparator not a conjecture

Bedetti & Allen characterise my position as (A) rejecting the assumption that earlier humans had different capacities and (B) extending “modern” cognitive capabilities to all members of our genus. I agree that this conclusion is unwarranted, but clarify that this is not my position. Although I believe, like many commenters (e.g., Milks, Blessing), that the difference between us and other human species is often overestimated, I do not believe there is presently sufficient evidence to accept a derived null (i.e., all members of our genus had similar capacities), or to definitively reject a primitive null (i.e., earlier members of our genus were less sophisticated). I must therefore explain what adopting a derived null means in practice.

It is important to appreciate that null hypotheses – inferential heuristics developed by early twentieth century statisticians (Fisher, Reference Fisher1928; Neyman & Pearson, Reference Neyman and Pearson1928; Pernet, Reference Pernet2016) – are neither conjectures, conclusions, nor statements of prior belief. In frequentist p-value hypothesis testing, null hypotheses are used to calculate the probability of seeing the observed results given no genuine relevant relationship between variables of interest. Although p-values are becoming less prevalent in certain sciences (McElreath, Reference McElreath2016), even in Bayesian or informal inferential reasoning, null models still have utility as comparators. I provide an example from my own work.

I recently co-authored a study exploring eyesight (visual acuity) among the Hadza (Stibbard-Hawkes & Apicella, Reference Stibbard-Hawkes and Apicella2022). There is extensive evidence that increasing classroom education is associated, globally, with worsening distance vision (Ku et al., Reference Ku, Steptoe, Lai, Hu, Chu, Yen and Chen2019; Morgan, French, & Rose, Reference Morgan, French and Rose2018; Mountjoy et al., Reference Mountjoy, Davies, Plotnikov, Smith, Rodriguez, Williams and Atan2018). We had data on Hadza school participation, so decided to investigate this relationship. Here, our hypothesis (H1) was that school participation should worsen distance vision, and our null hypothesis (H0) was that schooling should have no (or a positive) effect. After running analyses, we found no strong statistical evidence of a negative association between school participation and visual acuity (estimates were wide and crossed zero), and were unable to reject the null hypothesis. Note, though, this does not mean the null hypothesis is true (see Pernet, Reference Pernet2016). In fact, given the wider literature (Morgan et al., Reference Morgan, French and Rose2018), we should still expect schooling to worsen eyesight, and the absence of clear statistical evidence was probably consequence of a small sample with low school attendance. It could alternatively be, however, that there is some peculiarity of Hadza school-participation which diminishes (or reverses) its impact on eyesight. Here, by setting a high inferential hurdle, null hypothesis testing forces us to rely on hard evidence to support our conclusions, rather than accepting (reasoned) conjecture.

Like both Bedetti & Allen and Irurtzun, I find it most parsimonious (see McGrew, Reference McGrew2010) that earlier members of both our lineage and our genus did not have brains, minds and cognitive capacities identical (Boeckx, Reference Boeckx2023; Meneganzin & Killin, Reference Meneganzin and Killin2024) to ours. Adopting a “cognitively modern” null is wholly compatible with this belief. The new null model simply changes the inferential bar we must clear. As Vasil highlights, it prevents the more parsimonious primitive/ancestral null from being tacitly accepted, and forces us to think clearly about how differences would manifest using a probabilistic “evidence-based approach” (see Bedetti & Allen; Liu & Stout). Compared to the primitive null, a “derived”/“cognitively modern” null raises that evidentiary burden of proof.

R3.2. Beyond reasonable doubt: Overturning the derived null requires extraordinary evidence

Irurtzun argues that adopting a derived null would raise the burden of proof so high it “may amount to a scientific dead-end.” Although it might not be currently possible to conclusively overturn the derived null, it is not a dead-end. Instead, it forces us to employ multiple categories of probabilistic evidence, and appreciate their limitations.

Irurtzun cautions that no artefact would definitively show any particular human species did not possess any particular cognitive trait or capacity. This may be true. Although, as discussed in section R.2, there are instances where we may update beliefs in light of the absence of certain artefacts (e.g., modified ochre, metals, hafted stones), most allow only limited probabilistic inferences concerning capacity difference (target article, sect. 10.1). Other authors may put more stock in the inferential utility of absent artefacts (Coolidge et al., Reference Coolidge, Haidle, Lombard and Wynn2016; Sterelny, Reference Sterelny2014), for instance interpreting correlations between encephalisation and toolkit complexity as causal (Bedetti & Allen) rather than parallel. Even so, in disproving the derived null hypothesis (H0), we must be explicit about the probabilistic (Irurtzun; Liu & Stout; Will) inferential utility of the evidence, both positive and negative. If our aim is to get closer to the truth, properly reckoning the weaknesses of artefactual data is no bad thing.

Irurtzun also cautions that to overturn a derived null we could only rely on negative, not positive evidence. However, there are several types of enduring positive evidence we can employ. Bedetti & Allen highlight one of them: Brain size. The relationships between (relative) brain size, capacity and behaviour are complicated. Much research including, notably, certain formulations of the Upper Palaeolithic transition model, assume that fundamental reorganisations of brain structures could occur without large shifts in endocranial volume (e.g., Gabora & Smith, Reference Gabora, Smith, Henley, Rossano and Kardas2019; Klein, Reference Klein1995). Recent work, focusing on Homo naledi, has reached similar conclusions from the opposite vantage, suggesting that small brains need not prohibit “symbolic” behaviour (Berger et al., Reference Berger, Hawks, Fuentes, van Rooyen, Tsikoane, Ramalepa and Molopyane2023; Fuentes et al., Reference Fuentes, Kissel, Spikins, Molopyane, Hawks and Berger2023; though see Foecke, Queffelec, & Pickering, Reference Foecke, Queffelec and Pickering2024). I am sympathetic to Bedetti & Allen's view that, as brain tissue is costly, inter-species size differences must serve some adaptive function. The fact that different brain tissues may be asymmetrically costly (Castrillon et al., Reference Castrillon, Epp, Bose, Fraticelli, Hechler, Belenya and Riedl2023) complicates but does not obviate this supposition. Indeed, skeletal evidence of this type can reasonably be taken as positive probabilistic evidence against the derived null. The question is how and to what extent? What particular capacities do the approximately 500–700 cubic centimetres of cranial expansion that differentiate us from early members of our genus enable?

Presently, secure answers to this question may be beyond reach. The link between large brains and “symbolism,” at least, is not secure. Although the H. naledi data may not be sound (see Foecke et al., Reference Foecke, Queffelec and Pickering2024; Martinón-Torres, Garate, Herries, & Petraglia, Reference Martinón-Torres, Garate, Herries and Petraglia2023), Verpooten & De Tiège and Tennie & Planer show us that many material features often associated with complex symbolic cognition in archaeological contexts – acoustic technologies, non-utilitarian decoration/marking, and manuports – are produced by species less encephalised than any ancient human. From the other direction, Ben-Oren et al. speculate that cognitively modern humans need not create any complex or symbolic artefacts at all (contra Barceló-Coblijn). The connection between large brains and language is also tentative and complicated (Albessard-Ball & Balzeau, Reference Albessard-Ball and Balzeau2018, Kuleshova et al.).

As Kuleshova et al. highlight, there are other categories of positive evidence, though none are definitive. Cranial anatomy (Kochiyama et al., Reference Kochiyama, Ogihara, Tanabe, Kondo, Amano, Hasegawa and Akazawa2018) and endocasts (Bruner, Reference Bruner2010; de León, Bienvenu, Akazawa, & Zollikofer, Reference de León, Bienvenu, Akazawa and Zollikofer2016; Labra et al., Reference Labra, Mounier, Leprince, Rivière, Didier, Bardinet and Balzeau2024; Mounier, Noûs, & Balzeau, Reference Mounier, Noûs and Balzeau2020; Schacter & Addis, Reference Schacter and Addis2007) may provide scant clues about the relative expansion of different cortical structures, although their utility is limited (Neubauer, Reference Neubauer2014) and open to multiple interpretation (Labra et al., Reference Labra, Mounier, Leprince, Rivière, Didier, Bardinet and Balzeau2024; Mounier et al., Reference Mounier, Noûs and Balzeau2020). Other anatomical features – laryngeal or ear morphology – may provide clues about language, though most important features are soft tissues; evidence remains inconclusive (discussed in Albessard-Ball & Balzeau, Reference Albessard-Ball and Balzeau2018) and often suggests continuity (Conde-Valverde et al., Reference Conde-Valverde, Martínez, Quam, Rosa, Velez, Lorenzo and Arsuaga2021). For younger hominins, ancient DNA and comparative genomics show more promise. Present studies offer only associational hints (Kuhlwilm & Boeckx, Reference Kuhlwilm and Boeckx2019; Pauly, Johnson, Feltus, & Casanova, Reference Pauly, Johnson, Feltus and Casanova2024; Skov et al., Reference Skov, Peyrégne, Popli, Iasi, Devièse, Slon and Peter2022) and, as Blessing warns us, many genotype–phenotype association studies are badly confounded. Yet, as the field advances, and as comparative genomics (Jorstad et al., Reference Jorstad, Song, Exposito-Alonso, Suresh, Castro-Pacheco, Krienen and Bakken2023; Suresh et al., Reference Suresh, Crow, Jorstad, Hodge, Lein, Dobin and Gillis2023) and our knowledge of the causal pathways between genotype, ontogeny and phenotype improve over the next century, positive genetic evidence may one day allow us to overturn elements of the derived null. Again, setting a higher hurdle forces us to be precise about the limitations of existing evidence.

This does not represent an inferential “dead-end” – instead we must adopt as many categories of evidence, positive and negative, as possible. If we are unable to prove beyond reasonable doubt that other members of our genus were not similar in capacity to ourselves, it may temper beliefs about species-level capacity differences.

R3.3. Until proven guilty: Failing to overturn the derived null will temper our biases

As Irurtzun tells us “a certain homogeneity of traits is a core assumption for intra-species groupings.” This is correct, and one of several a priori reasons why we should hesitate to ascribe material culture differences within our species (e.g., Gabora & Smith, Reference Gabora, Smith, Henley, Rossano and Kardas2019; Kelly, Mackie, & Kandel, Reference Kelly, Mackie and Kandel2023; Klein, Reference Klein1995) to differences in capacity (see McBrearty & Brooks, Reference McBrearty and Brooks2000). Irurtzun also warns that we cannot extend the same logic beyond intra-taxon groupings. Bedetti & Allen ask “why place the boundary of Homo-sapiens-level cognition at [the genus level] rather than another?” Here, the reply to both is the same: Our genus is an intra-taxon grouping. Although less similar than members of the same species, a certain homogeneity of traits is a core assumption for genus-level groupings also. We should not expect similarities to cease above the neck. It is easy to forget this, and the derived null provides a heuristic check on our biases.

Here, Bedetti & Allen draw useful parallels between the present discussion and debates from comparative animal cognition. They highlight Morgan's canon, a parsimony heuristic which cautions no animal activity should be interpreted in terms of higher-order cognitive processes when it may be explained by less complex ones. Morgan's canon is a useful check on the “romantic” (sensu Dennett, Reference Dennett1983) tendency to anthropomorphise non-human animals. It is seen at work in commentaries by Tennie & Planer and Verpooten & De Tiège (also Tennie, Bandini, van Schaik, & Hopper, Reference Tennie, Bandini, van Schaik and Hopper2020). The opposite to Morgan's canon is perhaps Macphail's null, which posits we should, by default, assume no differences in intelligence between species (Macphail, Reference Macphail1987; Macphail, Barlow, & Weiskrantz, Reference Macphail, Barlow and Weiskrantz1985). Macphail argued that different species often employ similar general processes of associative learning (Macphail & Bolhuis, Reference Macphail and Bolhuis2001; Macphail et al., Reference Macphail, Barlow and Weiskrantz1985) and that between-species differences in task performance often result from contextual factors (Macphail et al., Reference Macphail, Barlow and Weiskrantz1985). For instance, in early vocal language acquisition studies, apes failed to master vocabularies of over a few utterances, but when researchers switched to gestural communication or lexigrams, representatives of those same species were able to learn large (100–1000) vocabularies (Gold & Watson, Reference Gold and Watson2018; Patterson & Cohn, Reference Patterson and Cohn1990; Savage-Rumbaugh, Shanker, & Taylor, Reference Savage-Rumbaugh, Shanker and Taylor1998) alongside elements of syntax (Savage-Rumbaugh et al., Reference Savage-Rumbaugh, Shanker and Taylor1998). The difference was context, not capacity.

There are many instances where Macphail's null has been rejected (Bastos & Taylor, Reference Bastos and Taylor2020). Macphail himself was hesitant to ascribe human-like intellect to non-human animals (Macphail et al., Reference Macphail, Barlow and Weiskrantz1985). When Macphail's ideas were published in a target article (Macphail, Reference Macphail1987), commenters were sceptical, yet almost 40 years later, the heuristic continues to provide a check on confirmatory biases in designing and interpreting behavioural experiments (Bastos & Taylor, Reference Bastos and Taylor2020; Pepperberg, Reference Pepperberg2020). Where Morgan's canon provides a check on anthropomorphism, Macphail's null provides a check on anthropocentrism. The question is, when comparing our capacities with those of other humans, where does the prevailing bias lie?

Early Western anthropologists at least were characterised by the tendency to view themselves as exceptional. The target article highlighted endemic early twentieth century discrimination against hunter-gatherer populations (Woodburn, Reference Woodburn1997), which had its basis in technological differences (Bagshawe, Reference Bagshawe1925; Blurton Jones & Konner, Reference Blurton Jones, Konner and Johannes1989) and was central in shaping policy (McDowell, Reference McDowell1984). Milks instantiates similar biases in discussions of the Tasmanian toolkit, including Noetling's (Reference Noetling1911) suggestion that their “failure” to add stone points to spears indicated the lack of a modern mind. This juxtaposes uncomfortably with present discussions of H. heidelbergensis hafting (sect. R.2). Although progress has been made, and work by Sahlins (Reference Sahlins1972) and Woodburn (Reference Woodburn1982) has helped ameliorate such biases, they continue colouring perceptions (Lavi, Rudge, & Warren, Reference Lavi, Rudge and Warren2024). Indeed, one commentary (Sergiou & Gabora) questions whether contemporary hunter-gatherers should be considered just as cognitively sophisticated as other humans (contra Blurton Jones).

It is difficult not to see parallels (Shea, Reference Shea2011; Zilhão, Reference Zilhão, Akazawa, Nishiaki and Aoki2014) between shifting twentieth century conceptions of hunter-gatherer minds and more recent discourses surrounding the “sapience” of ancient humans. Slimak (Reference Slimak2024) gives an evocative example of a leading Russian academy of science member declaring that Neanderthals “have no soul,” a question that has occupied academic theologians (Gaine, Reference Gaine2021; Moritz, Reference Moritz2012, Reference Moritz2015). This is an extreme example that yet throws light on “unspoken, unconscious assumptions which underlie great swathes of our understanding of this humanity” (Slimak, Reference Slimak2024, p. 12). Meneganzin and Killin (Reference Meneganzin and Killin2024) explore how “aesthetic capacity” was long considered unique to Homo sapiens (and see Slimak, Reference Slimak2024, Ch. 5). Blessing highlights the related fact that categories of artefact which were once accepted as evidence of H. sapiens' behavioural modernity are routinely dismissed in Neanderthal contexts (also see target article, sect. 4 & 16).

Given the outcomes of Western exceptionalism during the nineteenth and twentieth centuries (Elkins, Reference Elkins2022), such comparison inevitably takes on a moral aspect. Equally pertinent here though, is that anthropocentrism (perhaps sapiens-centrism) has routinely and demonstrably diminished the predictive accuracy of our theories: In the 1990s and 2000s, research consensus held that modern capacities appeared after the origins of our species (Klein, Reference Klein1995), an idea that receives continuing attention (Gabora & Smith, Reference Gabora, Smith, Henley, Rossano and Kardas2019; Kelly et al., Reference Kelly, Mackie and Kandel2023; Klein, Reference Klein2017). Earlier African evidence (McBrearty & Brooks, Reference McBrearty and Brooks2000) paved the way towards overturning this consensus. In the 2000s and 2010s, research consensus held that Neanderthals were less cognitively capable than H. sapiens. Increasing material evidence (Fajardo et al., Reference Fajardo, Kozowyk and Langejans2023; Hardy et al., Reference Hardy, Moncel, Kerfant, Lebon, Bellot-Gurlet and Mélard2020; Hoffmann et al., Reference Hoffmann, Standish, García-Diez, Pettitt, Milton, Zilhão and Pike2020; Mazza et al., Reference Mazza, Martini, Sala, Magi, Colombini, Giachi and Ribechini2006; Meneganzin & Killin, Reference Meneganzin and Killin2024) has led to a reassessment and the newly emerging consensus that Neanderthals, if not indistinguishable, were not “inferior” (Hardy et al., Reference Hardy, Moncel, Kerfant, Lebon, Bellot-Gurlet and Mélard2020; Mazza et al., Reference Mazza, Martini, Sala, Magi, Colombini, Giachi and Ribechini2006; Meneganzin & Killin, Reference Meneganzin and Killin2024; Blessing). Today, we still assume capacity differences between ourselves and middle Pleistocene hominins (Coolidge et al., Reference Coolidge, Haidle, Lombard and Wynn2016; Metcalfe, Reference Metcalfe2023; Sterelny), which emerging evidence is only beginning to throw into contention (Barham et al., Reference Barham, Duller, Candy, Scott, Cartwright, Peterson and Nkombwe2023). Although, as Liu & Stout emphasise, discussions are greatly more sophisticated today than the turn of the millennium, there is a pattern, and we may not yet have escaped the “modern human superiority complex” (Villa & Roebroeks, Reference Villa and Roebroeks2014).

Recognising these biases is particularly important in cases where evidence is unlikely to preserve. The temporal isolation of the various wooden artefacts (Barham et al., Reference Barham, Duller, Candy, Scott, Cartwright, Peterson and Nkombwe2023; Belitzky, Goren-Inbar, & Werker, Reference Belitzky, Goren-Inbar and Werker1991) probably reflects preservation not expression. Falk reminds us that we should not underrate cognitive sophistication during the first half of hominin evolutionary history either; even though preservation biases mean we may never find secure evidence of most technologies. Here the derived null may be profitably extended beyond our genus to our whole lineage – although as we approach the base of the clade, arguments of homology become more relevant (discussed in McGrew, Reference McGrew2010; Sayers & Lovejoy, Reference Sayers and Lovejoy2008; Sayers, Raghanti, & Lovejoy, Reference Sayers, Raghanti and Lovejoy2012).

The derived null, as above, is not conjecture, nor a statement of prior beliefs. We may still improve inference by strategically adopting weakly (Lemoine, Reference Lemoine2019) or more strongly (Lee & Vanpaemel, Reference Lee and Vanpaemel2018) informative priors which depart from our null. Instead, the derived null offers both a perspective shift, and a challenge. In failing to overturn it, and in reflecting upon that failure, we might moderate our beliefs somewhat. If we are ever to overturn it beyond reasonable doubt, we must leverage all available evidence.

R4.1. Comparative hunter-gatherers anthropology reveals broader trends in behaviour

Although the thesis of the target article requires no particular similarity between the toolkits of contemporary and Palaeolithic hunter-gatherers, similarities go beyond shared nomenclature: All hunter-gatherers hunted and gathered. Contemporary hunter-gatherers are not atavisms, nor relics, but modern, living populations, situated within and increasingly integrated into (Pollom, Cross, Herlosky, Ford, & Crittenden, Reference Pollom, Cross, Herlosky, Ford and Crittenden2021; Stagnaro, Stibbard-Hawkes, & Apicella, Reference Stagnaro, Stibbard-Hawkes and Apicella2022) global market systems. Some of my Hadza collaborators, who continue to forage, today have phones, and post videos on Facebook. My subject area, Human Behavioural Ecology, should better recognise this, but is not predicated on contemporary foragers being identical to past ones. Instead, it relies on a probabilistic gambit similar to that used by Sterelny; that rational people, faced with similar economic or ecological challenges, will converge on similar solutions (Blurton Jones, Reference Blurton Jones1982; Borgerhoff Mulder & Schacht, Reference Borgerhoff Mulder and Schacht2012; Cronk, Reference Cronk1991; Cunningham, Reference Cunningham2003; Kelly, Reference Kelly2013; Layton, Reference Layton, Panter-Brick, Rowley-Conwy and Layton2001; Mace, Reference Mace2014; Marlowe, Reference Marlowe2004; O'Connell, Reference O'Connell1995; Wiessner, Reference Wiessner2022). This is supported by evidence that many forager groups globally converge on similar patterns of behaviour (Bird & Codding, Reference Bird, Codding, Scott and Kosslyn2015; Grove, Reference Grove2009; Hoffman, Farquharson, & Venkataraman, Reference Hoffman, Farquharson and Venkataraman2023; Kelly, Reference Kelly2013; Smith et al., Reference Smith, Bettinger, Bishop, Blundell, Cashdan, Casimir and Stini1983; Woodburn, Reference Woodburn1982) despite having no recent shared history.

It is possible to overextend hunter-gatherer referential models beyond their utility, and the approach has numerous detractors (Gosselain, Reference Gosselain2016; Graeber & Wengrow, Reference Graeber and Wengrow2021; Pargeter et al., Reference Pargeter, MacKay, Mitchell, Shea and Stewart2016; Schrire, Reference Schrire1980). Moreover, Human Behavioural Ecology underrates the role of cultural processes (discussed in Mace, Reference Mace2014) and, like analogous utility maximisation models from behavioural economics (Hodgson, Reference Hodgson2012), over-rates human rationality more generally. People do not inevitably find optimal solutions or technologies (sect. R.2). Yet, where hard evidence is lacking, if employed with appropriate caution, forager studies may still help us reason about the unknown.

This discussion has explored what we can infer from the presence and absence of different materials. Ethnoarchaeology must have a continuing role here (Lane, Reference Lane, Cummings, Jordan and Zvelebil2014; Yellen, Reference Yellen1977). Where the present dataset only analyses artefact repertoires, studies of abandoned camps may provide information on the frequency and footprint of different artefactual traces (Yellen, Reference Yellen1977; O'Connell, Hawkes, & Jones, Reference O'Connell, Hawkes, Jones and Price1991). Ethnoarchaeology can also tease apart the subtle difference in both site organisations, and refuse production that characterises mobile and more sedentary camps (Kent, Reference Kent, Kroll and Price1991, Reference Kent and Kent1996; Yellen, Reference Yellen1991). Computational models, validated by ethnographic data, can explore the role of population structures in shaping and maintaining toolkit diversity (Ben-Oren, Kolodny, & Creanza, Reference Ben-Oren, Kolodny and Creanza2023a; Premo, Reference Premo, Mesoudi and Aoki2015; Zonker, Padilla-Iglesias, & Djurdjevac Conrad, Reference Zonker, Padilla-Iglesias and Djurdjevac Conrad2023), and how mobility shapes and limits the available evidence (Padilla-Iglesias & Bischoff, Reference Padilla-Iglesias and Bischoff2024). Ethnoarchaeology is also key to exploring how mobility, environment, network structures and functional constraints predictably influence the design, variation and expression of different technologies (Collard, Kemery, & Banks, Reference Collard, Kemery and Banks2005; Padilla-Iglesias et al., Reference Padilla-Iglesias, Blanco-Portillo, Pricop, Ioannidis, Bickel, Manica and Migliano2024; Shott, Reference Shott1986; Wiessner, Reference Wiessner1983).

Hunter-gatherer ethnography may also help us reason about patterns of behaviour and material expression which leave no traces. Several commentaries provide examples (Falk; Gallup & Eldakar; Tzafestas; Vasil). As Gallup & Eldakar and Tzafestas show us, the ethnographic record is replete with accounts of team competition (Sutton-Smith & Roberts, Reference Sutton-Smith and Roberts1971) and play that leave no enduring signature. For instance, the Mbuti often participated in organised, ritualised tugs-of-war (Turnbull, Reference Turnbull, Leacock and Lee1982). Although Hadza ethnographers have documented gambling games (“Lukuchuko,” see Woodburn, Reference Woodburn1970), there are few accounts of physical competition beyond those organised by researchers (Blurton-Jones & Marlowe, Reference Blurton-Jones and Marlowe2002; Stibbard-Hawkes, Amir, & Apicella, Reference Stibbard-Hawkes, Amir and Apicella2023; Stibbard-Hawkes, Attenborough, & Marlowe, Reference Stibbard-Hawkes, Attenborough and Marlowe2018). Yet, anecdotally, I have witnessed three impromptu Hadza physical skills contests: An archery contest between two men; a baobab climbing race; and an impromptu pull-up competition. We should assume these are ethnographically under-reported. Research into the functional role (Cohen, Davis, & Taylor, Reference Cohen, Davis and Taylor2023; Cohen, Ejsmond-Frey, Knight, & Dunbar, Reference Cohen, Ejsmond-Frey, Knight and Dunbar2009; Gray, Reference Gray2009; Lew-Levy, Reckin, Lavi, Cristóbal-Azkarate, & Ellis-Davies, Reference Lew-Levy, Reckin, Lavi, Cristóbal-Azkarate and Ellis-Davies2017; e.g., in cooperation, pedagogy and group cohesion), and ethnographic contexts of competition and play among both children (Lew-Levy et al., Reference Lew-Levy, Bombjaková, Milks, Kiabiya Ntamboudila, Kline and Broesch2022b; Lew-Levy, Boyette, Crittenden, Hewlett, & Lamb, Reference Lew-Levy, Boyette, Crittenden, Hewlett and Lamb2020) and adults (Gray, Reference Gray2009), allows us to build provisional models of past behaviour.

Vasil also discusses the centrality of ethnography in exploring the cognitive demands of group hunting, especially persistent hunting (also see Lieberman, Bramble, Raichlen, & Shea, Reference Lieberman, Bramble, Raichlen and Shea2007b; Morin & Winterhalder, Reference Morin and Winterhalder2024). Shared intentionality and group hunting have also featured in models of ritual evolution (Lang & Kundt, Reference Lang and Kundt2023). Here, remembering Morgan's canon (Stibbard-Hawkes, Reference Stibbard-Hawkes2023) cautioned that cooperative hunting is found in non-human species (Boesch, Reference Boesch2002; Stander, Reference Stander1992), and emphasised that not all contemporary foragers engage in extensive cooperative hunting (Berbesque, Wood, Crittenden, Mabulla, & Marlowe, Reference Berbesque, Wood, Crittenden, Mabulla and Marlowe2016; Wood et al., Reference Wood, Harris, Raichlen, Pontzer, Sayre, Sancilio and Jones2021). Moreover, there is debate over the past prevalence of persistence hunting (Morin & Winterhalder, Reference Morin and Winterhalder2024; Pickering & Bunn, Reference Pickering and Bunn2007). Nonetheless, cooperative hunting, especially drive hunting, is common globally (Boyd & Richerson, Reference Boyd and Richerson2022), including among the Mbuti (Wilkie & Curran, Reference Wilkie and Curran1991). It probably has deep roots. Ethnographic analogy is vital for exploring the forms hunting might have taken in the past and the contexts in which different practices were adaptive (Codding, Bird, & Bird, Reference Codding, Bird and Bird2011; Hoffman et al., Reference Hoffman, Farquharson and Venkataraman2023; Stibbard-Hawkes, Attenborough, Mabulla, & Marlowe, Reference Stibbard-Hawkes, Attenborough, Mabulla and Marlowe2020; Winterhalder, Reference Winterhalder1986).

Last, Grüning & Grüning propose a model for comparing the material complexity of different societies, focusing on assemblage diversity, artefact design complexity and the functional necessity of artefacts. Here, as Falk argues, we should expect that much complexity will be expressed only in perishable media (target article, p. 12). Material datasets like those in the SI file (or see Buckley, Reference Buckley2023) may allow us to calibrate such models, by comparing estimates of material complexity derived from complete repertoires to those derived only from enduring artefacts. These two estimates might differ. Indeed, as Sergiou & Gabora conclude “our ancestors may have been much more creative than what we can surmise from the existing archaeological record.” Although we must remember that the ethnographic record represents only a recent and limited sampling of the full sweep of modern human diversity, it remains the primary source for setting expectations and mapping the latent shapes of past creativity.

R4.2. Hunter-gatherers, materiality and cognitive ecologies

Several commenters highlight both that artefacts are imbedded in social contexts (e.g., Ben-Oren et al.; Sterelny; Sergiou & Gabora) and that objects and materials may influence and extend our social and cognitive worlds (Di Paolo et al.). This has been called “materiality” (Hussain & Will, Reference Hussain and Will2021) or, more holistically, “cognitive ecology” and the “extended mind” (Sterelny, Reference Sterelny2019; Tribble & Keene, Reference Tribble, Keene, Tribble and Keene2011). Several commentaries provide a chance to further explore the prospects and potential pitfalls of these approaches with reference to contemporary hunter-gatherers.

First the pitfalls. The target article warns against assuming strong boundaries between cognitive ecologies, especially where doing so creates “a hierarchy of cultural forms” or classifies “some as more cognitively or behaviourally modern [or] complex… than others” (p. 17). Sergiou & Gabora's commentary provides opportunities to revisit this. They question whether there is sufficient evidence that living hunter-gatherers are just as cognitively sophisticated as other societies and highlight that “cognitive sophistication is not uniform even amongst contemporary human populations.” They emphasise the concept of “nous,” meaning “intellect, understanding, or reason,” and ask whether “contemporary foragers differ from other contemporary human populations with respect to either nurture or nous, and even differences due to nature.” Here I implore caution.

It is true that reason is mediated by cultural context (Henrich, Heine, & Norenzayan, Reference Henrich, Heine and Norenzayan2010; Reuning, Reference Reuning, Cronbach and Drenth1972). As Henrich et al. (Reference Henrich, Heine and Norenzayan2010) relate, Kalahari foragers are much less susceptible to Müller-Lyer “arrow-and-stick” illusions than north Americans. Likewise, the Hadza language does not have an extensive native counting system so, to ease comprehension, researchers have often translated numbered Likert-scales into vernacular language (e.g., Smith & Apicella, Reference Smith and Apicella2022; Stibbard-Hawkes et al., Reference Stibbard-Hawkes, Abarbanell, Mabulla, Endeko, Legare and Apicella2024). However, despite patterning by subsistence type, differences in reasoning are task-specific. Blurton Jones highlights the difficulties that Western anthropologists face navigating the nuanced web of expectations that characterise hunter-gatherer sharing and personal property relations (Bahuchet, Reference Bahuchet1990; Crittenden & Zes, Reference Crittenden and Zes2015; Lee, Reference Lee, Boyd and Lassiter2011; McCall, Reference McCall2000; Peterson, Reference Peterson1993; Stibbard-Hawkes, Smith, & Apicella, Reference Stibbard-Hawkes, Smith and Apicella2022; Woodburn, Reference Woodburn and Hann1998). Several studies describe the complex deductive logic involved in interpreting animal spoor, or the deep factual knowledge hunters have about animal behaviour (Blurton Jones & Konner, Reference Blurton Jones, Konner and Johannes1989; Stibbard-Hawkes et al., Reference Stibbard-Hawkes, Attenborough and Marlowe2018). Given the innumerable domains of life where we bring reason (“nous”) to bear, while we may compare task-specific performance, satisfactorily operationalising and comparing the “cognitive sophistication” of whole societies is unachievable, perhaps meaningless (discussed in Buckner, Reference Buckner2022; Reuning, Reference Reuning, Cronbach and Drenth1972).

Moreover, culturally mediated differences in “nous” are neither fixed nor bounded. Presently, I am working over e-mail with a Hadza collaborator – who grew up in the bush – on the cognitively demanding task of research methods, design and translation. Just as he has, in adulthood, acquired the technological and linguistic skills to do this complex work, so might I, with practise, learn the nuances of spoor identification (Blurton Jones & Konner, Reference Blurton Jones, Konner and Johannes1989). The lesson, again, is that we should not overemphasise cultural boundaries, which are permeable. As for cognitive differences between hunter-gatherers and others “due to nature” (Sergiou & Gabora), such inquiry is impossibly confounded by context and ontogeny. As Blessing emphasises, the topic has been the purview of junk scientists (see Sear, Reference Sear2022), and prone to malappropriation (Panofsky, Dasgupta, & Iturriaga, Reference Panofsky, Dasgupta and Iturriaga2021). It has no place in this discussion.

If these pitfalls are avoided, we gain much by investigating inter-contextual patterning in the ways minds are extended by the materials we use, and material repertoires are shaped by context. Di Paolo et al. explore how certain materials may influence reasoning. They propose that non-perishable objects might last longer as design referents (e.g., via retrieval from abandoned camps, see Padilla-Iglesias & Bischoff, Reference Padilla-Iglesias and Bischoff2024), and that branch-weaving might foster skills such as tool disassembly, reassembly and component replacement. Ben-Oren et al. and Sterelny consider how social context might influence toolsets. Sterelny explores how denser social networks might increase the utility of social identity symbols (Kuhn & Stiner, Reference Kuhn, Stiner, Mellars, Bar-Yosef and Stringer2007a). He suggests that shell beads make particularly useful identity markers (Kuhn & Stiner, Reference Kuhn, Stiner, Mellars, Bar-Yosef and Stringer2007a; Miller & Wang, Reference Miller and Wang2022). Ben-Oren et al. discuss the related idea that material symbols play some functional role in intra-group social cohesion (e.g., McElreath, Boyd, & Richerson, Reference McElreath, Boyd and Richerson2003), and so may be more prevalent among denser or more interconnected populations (Bar-Yosef, Reference Bar-Yosef, Conkey, Soffer, Stratmann and Jablonski1997; Hovers & Belfer-Cohen, Reference Hovers, Belfer-Cohen, Overmann and Coolidge2024; Miller & Wang, Reference Miller and Wang2022).

These are each topics where forager ethnography can provide insights. It can contextualise how material objects are situated in particular ecologies or social contexts (Riede, Lew-Levy, Johannsen, Lavi, & Andersen, Reference Riede, Lew-Levy, Johannsen, Lavi and Andersen2023; Skaanes, Reference Skaanes2017; Sterelny, Reference Sterelny2021c; Wiessner, Reference Wiessner1983, Reference Wiessner1984). Quantitative cross-cultural research (e.g., Collard, Buchanan, Morin, & Costopoulos, Reference Collard, Buchanan, Morin and Costopoulos2011; Riede et al., Reference Riede, Lew-Levy, Johannsen, Lavi and Andersen2023), may test hypotheses statistically – for instance, investigating the relationships between demography, population structure and symbol-use. Here while the present analysis did not include demographic variables (Sergiou & Gabora), other studies have explored the relationship between demography and “toolkit complexity” (see Ben-Oren, Strassberg, Hovers, Kolodny, & Creanza, Reference Ben-Oren, Strassberg, Hovers, Kolodny and Creanza2023b; Kline & Boyd, Reference Kline and Boyd2010). It would be fruitful to explore the relationship between population size and structure and the intensity and design of semiotic artefacts across cultures. Here, we must not assume living foragers are facsimiles of the past, and appreciate that similar artefacts can take on different semiotic functions (Barker, Power, Heap, Puurtinen, & Sosis, Reference Barker, Power, Heap, Puurtinen and Sosis2019; Ozaita, Baronchelli, & Sánchez, Reference Ozaita, Baronchelli and Sánchez2022; Wiessner, Reference Wiessner1983). Yet, where we prove that humans probabilistically respond to environmental and demographic challenges in predictable ways, we may go beyond jingling.

R5.1. Clarifications

First, Liu & Stout caution that the target article “downplays widespread critiques of the ‘behavioral modernity’ construct over the past two decades”; but the target article provides extensive discussion of these important critiques (e.g., see sect. 3, 4, 13, 14, 15 and 16) and highlights multiple (40 + ) sources questioning either the soundness of the behavioural modernity concept or elements of its evidentiary basis, or related critical discussions of the evidence beyond our species (e.g., Albessard-Ball & Balzeau, Reference Albessard-Ball and Balzeau2018; Ames, Riel-Salvatore, & Collins, Reference Ames, Riel-Salvatore and Collins2013; Baquedano et al., Reference Baquedano, Arsuaga, Pérez-González, Laplana, Márquez, Huguet and Higham2023; Botha, Reference Botha2008, Reference Botha2010; Breyl, Reference Breyl2021; Brumm & Moore, Reference Brumm and Moore2005; Conard, Reference Conard, Henke and Tattersall2015; d'Errico & Stringer, Reference d'Errico and Stringer2011; d'Errico, Reference d'Errico2003; Dibble et al., Reference Dibble, Holdaway, Lin, Braun, Douglass, Iovita and Sandgathe2017; Haidle, Reference Haidle, Haidle, Conard and Bolus2016; Henrich, Kline, Muthukrishna, Shennan, & Thomas, Reference Henrich, Kline, Muthukrishna, Shennan and Thomas2016; Henshilwood & Dubreuil, Reference Henshilwood and Dubreuil2011; Hoffmann et al., Reference Hoffmann, Standish, García-Diez, Pettitt, Milton, Zilhão and Pike2020; Hopkinson, Reference Hopkinson, Roberts and Vander Linden2011; Kuhn & Stiner, Reference Kuhn and Stiner2007b; Langley, Clarkson, & Ulm, Reference Langley, Clarkson and Ulm2011; McBearty, Reference McBrearty2013; McBrearty & Brooks, Reference McBrearty and Brooks2000; McBrearty & Stringer, Reference McBrearty and Stringer2007; Milks, Reference Milks2020; Mounier et al., Reference Mounier, Noûs and Balzeau2020; Nowell, Reference Nowell, Hatfield and Pittman2013; Pascual-Garrido & Almeida-Warren, Reference Pascual-Garrido and Almeida-Warren2021; Powell, Shennan, & Thomas, Reference Powell, Shennan and Thomas2009; Scerri & Will, Reference Scerri and Will2023; Scerri et al., Reference Scerri, Thomas, Manica, Gunz, Stock, Stringer and Chikhi2018; Shea, Reference Shea2011; Speth, Reference Speth2004; Sterelny, Reference Sterelny2014, Reference Sterelny2011, Reference Sterelny2016, Reference Sterelny2021b, Reference Sterelny2019; Stringer, Reference Stringer2002; Villa & Roebroeks, Reference Villa and Roebroeks2014; Zilhão et al., Reference Zilhão, Angelucci, Badal-García, D'Errico, Daniel, Dayet and Zapata2010; Zilhão, Reference Zilhão2007).

Second, Liu & Stout warn that the target article “amplifies unilinear and teleological views of cognitive evolution and erroneously portrays them as consensus.” This is a misreading. The target article does caution against teleological reasoning (orthogenetic evolution), but does not portray such views as consensus. Instead, it states that while popular accounts (e.g., Harari, Vandermuelen, & Casanave, Reference Harari, Vandermuelen and Casanave2020) falsely present recent origins models “as a resolved consensus theory… Within the academy, recent origins and ‘revolution’ theories have been vigorously debated… research consensus has leaned towards gradualistic (McBrearty, Reference McBrearty2013) and mosaic (Conard, Reference Conard, Henke and Tattersall2015; Scerri et al., Reference Scerri, Thomas, Manica, Gunz, Stock, Stringer and Chikhi2018) theories of evolutionary change. Pure cultural evolutionary accounts, which assume no difference in intrinsic capacity… have become more widely accepted” (target article, sect. 3).

Third, Sergiou & Gabora and Liu & Stout both highlight the role of McBrearty and Brooks' seminal paper “The revolution that wasn't.” As Will; Blessing and Blurton Jones also note, this paper was critical in advancing the study of behavioural modernity beyond eurocentric “revolution” models. Its significance should not be understated. It is important also to clarify that the target article makes a different contribution. McBrearty and Brooks (Reference McBrearty and Brooks2000) reviewed the African archaeological record and demonstrated, decisively, that material complexity had deep roots. They promoted a gradualistic framework for understanding cognitive evolution. The target article, instead, contends that because contemporary humans need not leave extensive evidence of material sophistication, we should be more cautious in using material culture as a barometer for cognition.

As Blessing tells us, present findings do have relevance to “rubicon” definitions of behavioural modernity: “Any generalized threshold for what counts as modern, or… what it takes to be human, will [probably] exclude some populations.” “The assumption that we will eventually find the one difference between us and not us (e.g., Meneganzin & Currie, Reference Meneganzin and Currie2022), will always lead to… moving of the goal posts” (Blessing, Reference Blessing2023) so, as Liu & Stout emphasise, we must continue developing “approaches not organized around sterile… modern/pre-modern binaries.” However, the target article is not primarily a critique of the “behavioural modernity” concept, but aims to inspire reconsideration of the link between material culture and complex cognition more broadly.

This is no straw-man.

Indeed, Killin and Pain (Reference Killin and Pain2021) highlight that using stone artefacts to infer aspects of human cognitive and social evolution is “the core business of evolutionary cognitive archaeology” (p. 269). A central assumption is that “complex technologies over the last 3.3 million years provide a mirror to the cognitive developments that underpin behavioral changes” (Fajardo et al., Reference Fajardo, Kozowyk and Langejans2023, p. 1). Wynn, Overmann, and Malafouris (Reference Wynn, Overmann and Malafouris2021), in reviewing 50 years of research, characterise most as united in assuming that “bigger brains made better tools” (p. 99). Although models are often multifaceted and nuanced, as Will puts it, many continue to draw “straightforward connections between material culture and specific measures of cognitive capacities or behavioral complexity.” Sections 3 and 4 of the target article give numerous recent examples, with quotations and page numbers. Although many were caveated (e.g., Kelly et al., Reference Kelly, Mackie and Kandel2023, is exemplary), they yet employed those broader modes of inference under consideration. Here I highlight three further recent examples:

Backwell et al. (Reference Backwell, Bradfield, Carlson, Jashashvili, Wadley and d'Errico2018) tell us “Complex cognition emerged approximately 100kya” (p. 290). They link this to hunting technology, stating “the bow and arrow is thought to signal higher-level cognitive functioning and is considered a hallmark of complex modern human behaviour” (p. 290). But not all humans use bows (Davidson, Reference Davidson1936; Stibbard-Hawkes, Reference Stibbard-Hawkes and Moreau2020).

Mithen (Reference Mithen2024a) argues that Neanderthals frequently suffered injuries “from close-encounter hunting using thrusting spears” which “could have been avoided by using bows and arrows or thrown spears.” He takes this as evidence that they “did not design hunting weapons for killing specific types of animals in specific circumstances” and so “struggled to connect the different semantic clusters of words within their brains, such as those relating to animals, people and tools.” But there is evidence of contemporary humans – fully capable of connecting semantic clusters – hunting large, dangerous animals with hand-spears (Ichikawa, Reference Ichikawa, Konidaris, Barkai, Tourloukis and Harvati2021; Macphail, Reference Macphail1930). The Mbuti, for instance, use hand-spears in elephant hunts (Ichikawa, Reference Ichikawa, Konidaris, Barkai, Tourloukis and Harvati2021), which are better suited to immobilising an elephant's knees, and slashing its vulnerable lower abdomen, than are projectiles. Moreover, as Milks reminds us, it is easy to underrate the complexity of ostensibly “simple” technologies; creating non-hafted weapons, such as wooden long-spears, requires multi-step advanced woodworking techniques (Haidle, Reference Haidle, De Beaune, Coolidge and Wynn2009; Leder et al., Reference Leder, Lehmann, Milks, Koddenberg, Sietz, Vogel and Terberger2024; Milks et al., Reference Milks, Lehmann, Leder, Sietz, Koddenberg, Böhner and Terberger2023), alongside years of learning (and teaching) to utilise effectively (Lew-Levy et al., Reference Lew-Levy, Bombjaková, Milks, Kiabiya Ntamboudila, Kline and Broesch2022b). We should not dismiss these as “little more than a pointed stick” (Mithen, Reference Mithen2024b).

Although these two examples concern single artefact categories, Sergiou & Gabora highlight that much contemporary research is “multi-pronged” and “focuses on identifying adaptive packages of features.” For example, Gabora and Smith (Reference Gabora, Smith, Henley, Rossano and Kardas2019) identify a major transition at 100 kya during what they consider “was likely a time of major neural reorganization” (p. 226). They highlight seven features that identify this transition: (a) Proliferation of task-specialised tools; (b) “elaborate burial sites indicating ritual”; (c) “personal symbolic ornamentation”; (d) representational cave-art; (e) complex hearths and structured living spaces; (f) calorie-gathering intensification; and (g) “bone and antler tools, sometimes with engraved designs.” They propose these changes resulted from “a fine-tuning of the biochemical mechanisms underlying the capacity to spontaneously shift between different modes of thought” (p. 227). Yet, as here, even where models are multi-pronged, we may still reconsider the inferential utility of one or other category of evidence. That contemporary humans – well capable of shifting between modes of thought – do not universally engage in elaborate ritual burial (target article, sect. 10, para. 3), or produce representational cave-art (target article, sect. 10, para. 2), or leave enduring evidence of personal symbolic ornamentation (target article, sect. 10.1) is surely relevant.

It is best to view this, as Liu & Stout remind us, not as a binary discussion over truths and falsehoods, but “as a disagreement about probabilities.” We must, however, recognise this disagreement as genuine.

R5.2. Prescriptions

I am nonetheless sympathetic to Liu & Stout's call to “move on from critique.” Scepticism from beyond the field is nothing new (Lewontin, Reference Lewontin, Osherson, Sternberg and Scarborough1998). Although the dataset is novel, as Will reminds us, many of the inferential problems identified here are well-known to Cognitive Archaeologists (reviewed in Killin & Pain, Reference Killin and Pain2021); and will appear trivial to some readers. Despite this, data-driven critical discussion remains important.

First, while certain issues are well-known, contextualising abstract lessons with hard data from living humans makes them tangible. It is easy, when thinking hypothetically about long-dead humans, to speculate that the hunting bow is “a hallmark of complex modern human behaviour”; but reflecting on the many complex, modern humans who do not produce them (Stibbard-Hawkes, Reference Stibbard-Hawkes and Moreau2020) may inspire researchers to moderate these and related claims. Theoretical diversity is valuable (Killin & Pain, Reference Killin and Pain2021) and inter-disciplinarity allows the synthesis of insights from different fields. Even where the target article provides little novel discussion, the data might still lead to a modest reassessment of the probabilistic inferential utility ascribed to certain types of material evidence.

Second, as Will identifies, Behavioural and Brain Sciences has a broad readership. Questions concerning the origins of the human mind have occupied linguists, psychologists, neuroscientists, primatologists and hunter-gatherer anthropologists (target article, abstract) – not forgetting cultural evolutionists, geneticists and theologians. Not all disciplines are equally well-versed in the inferential pitfalls of archaeological data – taphonomy, underdetermination, equifinality, absent evidence (Will). Kuleshova et al. remind us that language evolution research is fundamentally interdisciplinary, and that evidence from contemporary foragers can help in calibrating our baseline assumptions about how language might (or might not) manifest in the material record. Present discussion may be useful to readers of other specialisations also.

Last, and relatedly, as Will highlights, perspectives from beyond the field (e.g., Harari, Reference Harari2014) have given “distorted portrayals of human origins.” I know, first-hand, how difficult it is to prevent such distortions; a recent co-authored paper explored hunting exclusively among contemporary humans (Venkataraman et al., Reference Venkataraman, Hoffman, Farquharson, Davis, Hagen, Hames and Stibbard-Hawkes2024), yet a UK broadsheet newspaper called these “prehistoric societies” and ran the headline “Ancient man did most of the hunting after all” (Rhys, Reference Rhys2024, now corrected). Fully inoculating against misinterpretation is unfeasible, but there are steps that may limit it:

1. 1) Many papers give caveated yet precise chronological boundaries or date-ranges for particular capacities (e.g., Backwell et al., Reference Backwell, Bradfield, Carlson, Jashashvili, Wadley and d'Errico2018; Kelly et al., Reference Kelly, Mackie and Kandel2023; Paige & Perreault, Reference Paige and Perreault2024). These should be presented carefully, as they imply surety and are easy for non-experts to latch onto, leading to simplistic portrayals from science communicators. For instance, in a recent exploration of the origins of cumulative culture (Paige & Perreault, Reference Paige and Perreault2024), the authors clearly explain the inferential issues of taphonomy, but most press summaries emphasised only the date boundary (Putol, Reference Putol2024; Russ, Reference Russ2024; Timmer, Reference Timmer2024).

2. 2) Works linking particular artefacts to particular capacities (Coolidge et al., Reference Coolidge, Haidle, Lombard and Wynn2016) or neurocognitive mechanisms (Stout, Chaminade, Apel, Shafti, & Faisal, Reference Stout, Chaminade, Apel, Shafti and Faisal2021) are easy for popularisers, funders and even disciplinary outsiders to interpret as presence/absence acid tests for these capacities – even where this is not the intended interpretation. Researchers should explicitly address the role of negative evidence (Wallach, Reference Wallach2019) – what we could conclude were these artefacts absent. Where researchers make no inferences from absence, they should say so clearly. Where they do, they should ensure theories would be sound if applied to the ethnographic record.

3. 3) We should be more hesitant to comment on cognitive capacity shifts (Coolidge et al., Reference Coolidge, Haidle, Lombard and Wynn2016; Gabora & Smith, Reference Gabora, Smith, Henley, Rossano and Kardas2019; Kelly et al., Reference Kelly, Mackie and Kandel2023; Paige & Perreault, Reference Paige and Perreault2024). These may often simply be inferentially inaccessible and, whatever their link to the material record, must always precede expression (Haidle, Reference Haidle, Haidle, Conard and Bolus2016; Hovers & Belfer-Cohen, Reference Hovers, Belfer-Cohen, Hovers and Kuhn2006) by an undetermined interval. Indeed, questions concerning the social, environmental and ecological contexts of behavioural and technological change (Sterelny; Ben-Oren et al.; Di Paolo et al.) are often more interesting and accessible to inquiry (e.g., Blessing, Conard, & Bader, Reference Blessing, Conard and Bader2023; Hovers & Belfer-Cohen, Reference Hovers, Belfer-Cohen, Hovers and Kuhn2006, Reference Hovers, Belfer-Cohen, Overmann and Coolidge2024; Hussain & Will, Reference Hussain and Will2021; Miller & Wang, Reference Miller and Wang2022; Shea, Reference Shea2011). Many studies which do explore capacity shifts present findings (Backwell et al., Reference Backwell, Bradfield, Carlson, Jashashvili, Wadley and d'Errico2018; Kelly et al., Reference Kelly, Mackie and Kandel2023; Paige & Perreault, Reference Paige and Perreault2024) that would be similarly exciting were these discussions omitted.