Although we have known for over four decades, that hominin stone tool technology extends into deep antiquity (Leakey, 1971), the temporal extent of tool-assisted hominin behavior is still unknown.

The earliest visible evidence of frequent and systematic stone flake production is represented by Oldowan technology (Semaw et al., 1997; Semaw, 2000, 2006). The oldest archaeological evidence of the Oldowan at Bokol Dora (Braun et al., 2019) and Gona in Ethiopia (Semaw et al., 2003), dated to around 2.6 Ma, shows that this technology involved the reduction of cobbles to produce sharp-edged flakes (Stout et al., 2010). The diversity of reduction strategies along with the lack of knapping accidents suggested that the hominins behind this record already had a good understanding of the mechanics of stone fracture (Stout et al., 2010). The apparent sophistication of the early Oldowan has led some to suggest a more gradual developmental stage of stone tool use and production before 2.6 Ma (Panger et al., 2002; Carbonell et al., 2007). The discovery of flaked stone artifacts from Lomekwi 3 (West Turkana, Kenya), dated to 3.3 Ma (Harmand et al., 2015), provides a potential snapshot of an earlier stage of flake production (but see Dominguez-Rodrigo and Alcalá, 2019; Archer et al., 2020). At Lomekwi 3, a combination of passive hammer and bipolar knapping was used to produce flakes (Harmand et al., 2015; Lewis and Harmand, 2016). Although technologically different compared to the Oldowan (Braun et al., 2019), it is argued that the Lomekwian still shows an understanding of the mechanics of stone flaking, including conchoidal fracture, raw material quality, and the application of repeated exploitation strategies. In addition, this assemblage retains the earliest evidence of percussive behaviors in the archaeological record (Harmand et al., 2015; Lewis and Harmand, 2016).

Fossil evidence also supports an earlier origin for tool use by hominins. Many derived anatomical aspects of hominin hands are argued to have been adaptations for tool use (Marzke and Shackley, 1986; Marzke, 1992; Marzke and Marzke, 2000; Kivell et al., 2011; Skinner et al., 2015; Key et al., 2020) and appear in the fossil record before the earliest archaeological evidence of stone tool production. In addition, cut marked animal bones from Dikika (Ethiopia), dating to 3.39 Ma, highlight the possibility that hominins may have used naturally occurring sharp-edged stones to facilitate foraging (McPherron, 2010; Dominguez-Rodrigo et al., 2011). Thus, current advances in Pliocene archaeological research are pushing back the chronological range of tool use and toolmaking into a period before the emergence of the genus Homo (Villmoare et al., 2015). This suggests that the evolution of technology may not have been a linear development, but may have occurred independently in different hominin species (Rolian and Carvalho, 2017).

In the quest to interpret the processes that led to the emergence of stone flaking, chimpanzees have played a dominant role as referential models, given their phylogenetic proximity to hominins and their wide range of tool-aided behaviors (Davidson and McGrew, 2005). Although rare within the primate order, three other nonhuman primates habitually use stone tools in the wild. These include long-tailed macaques in Thailand and Myanmar (Macaca fascicularis: Malaivijitnond et al., 2007; Gumert et al., 2009), capuchin monkeys in Brazil (Sapajus libidinosus: Fragaszy et al., 2004; Falótico and Ottoni, 2016; Sapajus xanthosternos: Canale et al., 2009); and white-faced capuchin monkeys in Panama (Cebus capucinus: Barrett et al., 2018). Although each species uses stone hammers and anvils for a range of different percussive behaviors (Whiten et al., 1999; Gumert and Malaivijitnond, 2013; Falótico and Ottoni, 2016), all of these primates universally use stone tools to crack open nuts (Boesch and Boesch, 1990; Falótico and Ottoni, 2016; Luncz et al., 2017).

Within this context, it is suggested that the evolutionary origins of stone flaking may have derived from a ‘culture of pounding’ dominated by the use of stones for extractive foraging such as fracturing bones to access marrow or processing nuts (Kortlandt, 1986; McGrew, 1992; de Beaune, 2004; Davidson and McGrew, 2005; Marchant and McGrew, 2005; Whiten et al., 2009; Rolian and Carvalho, 2017; Thompson et al., 2019; Gürbüz and Lycett, 2021). Such perspectives are often influenced by the observed percussive behaviors of extant nonhuman primates. A prevailing hypothesis to explain the emergence of flake production from a pounding behavior to intentional flaking revolves around the unintentional detachment of sharp-edged flakes from nut-cracking anvils due to hammerstone mis-hits (McGrew, 1992; Marchant and McGrew, 2005; Rolian and Carvalho, 2017). It is suggested that the repeated re-use of the same anvils over time would lead to an accumulation of sharp-edged flakes within the vicinity of food processing, representing a potential opportunity for the reverse engineering of intentional flake production (McGrew, 1992). This hypothesis has been frequently invoked in discussions on the emergence of stone tool technology (Wynn and McGrew, 1989; Panger et al., 2002; Davidson and McGrew, 2005; Marchant and McGrew, 2005; Carvalho et al., 2009; Wynn et al., 2011) and is referred to here as the ‘by-product hypothesis.’

Various studies have shown that the production of flakes is within the capacity of nonhuman primates (Wynn and McGrew, 1989; Wynn et al., 2011; Proffitt et al., 2016). Experiments with captive bonobos (Pan paniscus) have demonstrated their ability to produce flakes, which they then use to access food following training by human experimenters (Toth et al., 1993; Schick et al., 1999; Savage-Rumbaugh et al., 2007). The resulting core and flake assemblages differed significantly to both Oldowan and human-replicated assemblages (Pelegrin, 2005; Toth et al., 2006; Eren et al., 2020). In wild West African chimpanzees (Pan troglodytes verus), it is known that, occasionally during nut cracking, either the hammer or anvil are fractured in a way that produces sharp-edged flakes (Hannah and McGrew, 1987; Marchant and McGrew, 2005; Carvalho et al., 2008). The excavation of a modern chimpanzee nut-cracking site in the Taï Forest in Cote d'Ivoire (Panda 100), originally bolstered the suggestion that significant unintentional flake production was associated with chimpanzee nut cracking and furthermore that these flakes were comparable to hominin Oldowan flakes from Omo (dated to ∼2.3 Ma; Mercader, 2002). However, a recent reanalysis of the Panda 100 lithic assemblage has shown that flake detachments are in fact rare (Proffitt et al., 2018a). This is mainly due to the poor homogeneity of the available raw material (Proffitt et al., 2018a). Recent experimental studies, where captive chimpanzees have been provided with quartzite anvils, have also shown that accidental flake production is rare. Here, the lack of flakes was mainly caused by the lack of hammer mis-hits during nut cracking (where the anvil had direct contact with the hammerstone; Arroyo et al., 2016).

Captive robust capuchin monkeys (Sapajus sp.; see Alfaro et al., 2012 for revised taxonomic classifications) have also been shown to produce sharp-edged fragments when provided with adequate stone material (Westergaard, 1995; Westergaard and Suomi, 1995). The lithic material produced in those studies, however, only possess a passing resemblance to knapped flakes, and were produced primarily by throwing a core against a hard surface (Westergaard, 1995). Importantly, these studies do not show a natural behavior by a capuchin monkey. More recently, it was reported that wild capuchin monkeys (S. libidinosus) in Serra da Capivara National Park in Brazil (Mannu and Ottoni, 2009; Falótico and Ottoni, 2016) unintentionally detached conchoidal flakes from quartzite hammerstones during stone-on-stone percussive behavior (Proffitt et al., 2016). These flakes have the same technological attributes used to identify flakes produced by hominins and occur in greater numbers (Proffitt et al., 2016) compared to the known chimpanzee archaeological record (Proffitt et al., 2018a). Further, it has been reported that capuchin monkeys can significantly damage anvils during nut cracking, producing identifiable assemblages of debris (Visalberghi et al., 2013). However, the technological features of these unintentionally produced assemblages have not been reported (Visalberghi et al., 2013). As such, this behavior is an ideal candidate to explore the ‘by-product hypothesis.’

Until now, there has been no formal investigation of the circumstances under which the ‘by-product hypothesis’ could be possible. Furthermore, there has been no attempt to quantify the archaeological signature of nut cracking across raw material qualities/types. For the ‘by-product hypothesis’ to be feasible, a number of conditions must logically be met. First, both the anvil and the hammer must be made of stone. Second, the raw material properties of the hammer and anvil must be generally fine grained and isotropic, such that a successful conchoidal or wedging fracture can be initiated (Cotterell et al., 1985; Cotterell and Kamminga, 1987; Tsirk, 2014). Third, the potential for flakes to be detached unintentionally from the margins of anvils requires that the anvil itself has at least one angle between the active plane and the adjoining horizontal plane less than or approaching 90° (Cotterell and Kamminga, 1987; Dibble and Pelcin, 1995; Rezek et al., 2011; Magnani et al., 2014). Finally, even if all these conditions are met, accidental flakes will not occur if the percussive behavior is performed with a high degree of accuracy. The percussive activity needs to include mis-hits such that the hammerstone strikes the anvil directly with enough regularity to create substantial removals.

This study addresses whether the ‘by-product hypothesis’ for the emergence of stone flake technology could be a possibility, and if so, how it might be identifiable in the archaeological record. We target three specific questions: 1) Can natural foraging of nuts with stone hammers and anvils frequently produce sharp-edged flakes? 2) How does raw material quality affect the archaeological signature associated with nut cracking? 3) Are flakes associated with the ‘by-product hypothesis’ distinguishable from intentionally produced, freehand conchoidal flakes? For our study, we selected wild capuchins (Sapajus sp.), a nonhuman primate that uses a bimanual action to crack nuts. Wild bearded capuchin monkeys in Brazil adopt a bipedal stance, raise the hammerstone above their head, and use their prehensile tail to add postural stability (Liu et al., 2009). The bimanual handling of hammerstones generally decreases the precision of hits. Interestingly, a component of the oldest evidence of tool production, the Lomekwian technology, was likely produced through the passive hammer technique, a bimanual method of handling stone cores (Harmand et al., 2015; Macchi et al., 2021).

To address our research questions, we provided a group of capuchins with anvils of three different raw materials with varied densities and internal homogeneity, representing low, medium, and high quality. Quality here represents the ability of the stone to fracture predictably (Braun et al., 2009). These anvils were used to crack a standard number of oil palm nuts (Syagrus romanzoffiana). The resulting lithic assemblages associated with each anvil were subjected to a full use-wear, technological, and refit analysis to characterize their archaeological signature. The detached percussive products from the high-quality raw material were subsequently compared to an experimental sample of flakes, created through free-hand knapping by an experienced knapper using the same raw material type. This allowed the identification of potential diagnostic differences between flakes produced during a pounding behavior and those produced by a knapping activity.