1 Introduction

Studying non-human tool use helps us to understand adaptive contexts for this behavior, contributing to the understanding of the evolution of human cognition (Laland and Janik 2006). Non-human primates in particular have shown flexible and variable tool use in terms of functions and types of materials used, especially in chimpanzees (Sanz and Morgan 2009). Within Platyrrhini, the genus Sapajus has proved ideal for this type of research, with tool use well characterized in the species Sapajus libidinosus, particularly in the context of foraging (Mendes et al. 2015). The genus is marked for its highly extractive and manipulative foraging (Visalberghi and McGrew 1997), and in the last 15 years, many studies have described tool use in wild populations (Fragaszy et al. 2004; Ottoni and Izar 2008; Mendes et al. 2015). Studying stone tool use in platyrrhines provides the opportunity to elucidate characteristics of tool use evolution in an experiment outside of the human and ape lineage. Platyrrhines are an independent primate lineage that diverged from Catarrhini at about 40 mya (Springer et al. 2012 (mean time to a most recent common ancestor [TMRCA] molecular 40.1 mya); Schrago et al. 2013 (mean estimate TMRCA molecular: 36.1 mya, morphology-based: 43.1 mya)). Within this lineage, ancestral capuchins (crown Cebinae), as well as ancestral robust capuchins (stem Sapajus), may have faced convergent selection pressures to hominins during the late Miocene and throughout the Pliocene favoring the emergence of tool use and related cognitive traits, as suggested from genomic analyses pointing to sustained positive selection on cognition-related genes in the Cebidae, Cebinae, and Sapajus lineages during that time interval (Byrne et al. 2022). While habitual tool use has been described in several studies of different wild populations of Sapajus libidinosus, tool use occurs at least in captivity in all species of robust capuchins (Steinberg et al. 2022), and direct observations of tool use have been documented in other robust capuchin species in the wild, such as S. flavius (Lima et al. 2024), S. nigritus cucullatus (Rocha et al. 1998; dos Santos Gutierres et al. 2025), and S. apella (Fernandes 1991). A comparative study of tool use characteristics across Sapajus species allows us insight into what characteristics of stone tool use are stable versus flexible across this widespread group; here we focus on variation in stone tool use for nutcracking, the most widespread type of tool use in capuchin monkeys.

Many studies of tool use in Sapajus libidinosus report the use of stones to break open-encased fruits, such as palm fruits, Manihot seeds, and cashew nuts (Fragaszy et al. 2004; Visalberghi and Fragaszy 2013; Spagnoletti et al. 2011; Visalberghi et al. 2016; Moura and Lee 2004; Mannu and Ottoni 2009; Falótico and Ottoni 2016; Falótico et al. 2022). Visalberghi et al. (2007) took advantage of the natural habitual tool use by S. libidinosus in Fazenda Boa Vista, Piauí state, Brazil to provide a complete characterization of stone tool use and test various hypotheses related to nutcracking behavior, including whether monkeys always chose the heaviest available stones to use as hammers in order to minimize the energetic cost of opening fruits. In addition, they assessed how anvil sites were chosen for hammering activity: the researchers found that monkeys chose anvils that had sufficient area for individuals to stand on the anvil for hammering (and place the hammers on the anvils); the presence of pores/holes so fruits could be placed in them once they were broken open and not roll away; adequate height to minimize energetic transport of hammers and fruits; and proximity to trees and vegetation, in case of need to escape the area (Visalberghi et al. 2007). Other characteristics also seem to influence anvil selection, including proximity to source trees for the fruits (to facilitate transport of fruit to the anvils), and anvils with enough surface area to leave fruits and hammers available for future hammering activity, again to minimize energetic cost (Visalberghi et al. 2007).

An experiment by Visalberghi et al. (2009) showed that Sapajus libidinosus chose the weight of stones as hammers based on the resistance of the fruit they were processing. Hammer stones are often a limited resource, and may be reused and transported from one anvil to another (Visalberghi et al. 2007). In another S. libidinosus population, Ferreira et al. (2010) showed that stones of different weights were used to process fruits of different sizes. Falótico et al. (2024) showed that S. libidinosus used hammers of different lengths to process different fruits at Ubajara National Park.

The timing of tool use throughout the year is also of interest in trying to understand how adaptive pressures may shape tool use. Researchers have found that wild Sapajus libidinosus use tools throughout the year (Falótico et al. 2017, 2018; Falótico and Ottoni 2023; Spagnoletti et al. 2011), rather than a more seasonal relationship of increased tool use in the season with fewer preferred fruits available. This evidence argues against the ‘necessity’ hypothesis for tool use in S. libidinosus as an adaptation to surviving in a resource-scarce environment (Fox et al. 1999; Spagnoletti et al. 2012).

For Sapajus libidinosus, nutcracking is widespread among many populations and routinely performed throughout the year. In 2009, Canale and colleagues published the first indirect evidence of wild tool use in the yellow-breasted capuchin monkey, S. xanthosternos, reporting six localities where they confirmed nutcracking sites for hammering open-encased fruit, as evidenced by anvil stones with hammer marks and remains of processed fruits. Additional S. xanthosternos tool use sites have been confirmed in Minas Gerais state (PAN PRINE Action Report 2019; Medeiros et al. 2025). Although populations of S. xanthosternos are found from the evergreen forests of the Northern Atlantic Forest to the seasonal forests in transitional zones within Cerrado and Caatinga (Lernould et al. 2012; Culot et al. 2019), tool use by this species has been found only within populations living in the drier environments (Canale et al. 2009; PAN PRINE Action Report 2019; Medeiros et al. 2025), despite intensive surveys and interviews within the Atlantic Forest (Canale et al. 2009) and long-term studies of wild S. xanthosternos there, such as at REBIO Una in Bahia (Suscke et al. 2021).

In the present study, we characterized the hammering sites used by Sapajus xanthosternos in two localities in northern Minas Gerais, Brazil, and we tested hypotheses about stone hammer use and anvil use for the first time in this species. We also compared tool use in S. xanthosternos to that reported in the literature for wild S. libidinosus and other Sapajus species. As was found for S. libidinosus in Visalberghi et al. (2007), we hypothesized that S. xanthosternos would use anvil stones (1) with a large enough surface area to accommodate the monkey's body and the hammers, (2) with the presence of depressions to facilitate positioning of fruit while hammering, (3) with low height, and (4) with proximity to source trees for the target fruit, as well as (5) proximity to trees that would provide visual cover and potential escape routes. We additionally hypothesized that monkeys would choose the heaviest available stones to use as hammers in order to minimize the energetic cost of opening fruits (Visalberghi et al. 2007), and in particular that the monkeys would select hammer stones on the basis of the size of the fruit they were breaking open (larger stones for larger or more difficult-to-open fruits) (Falótico et al. 2018; Ferreira et al. 2010). We also tested whether there was seasonality in tool use for S. xanthosternos, and if so, if tool use was more common in the season when fruits were scarce. Seasonal tool use related to fruit scarcity would provide evidence toward the ‘necessity’ hypothesis, that tool use has evolved in order for these capuchin populations to survive in an environment with seasonally scarce resources (Fox et al. 1999; Spagnoletti et al. 2012).

2 Material and Methods

2.2 Study Locations

This study took place in two localities in northern Minas Gerais state: the Santa Rosa de Lima (SRL) district and Lapa Grande State Park (LGSP) (Figure 1). These locations were selected because of reports of stone tool use by monkeys near there from Canale et al. (2009), and personal communication with the local workers and the community of the region, leading to our own field surveys for primates at these sites (Medeiros et al. 2025).

The forest fragment within the SRL district, located 59 km south of the county seat of Montes Claros municipality (16°19′03.6″ S, 43°58′58.8″ W), has approximately 1000 ha of native vegetation, interspersed with areas of human disturbance. LGSP is an integral protection conservation unit created in January 2006, administered by the State Forest Institute (IEF—North regional office), in the Montes Claros municipality, northern Minas Gerais (16°42′23.7″ S, 43°56′34.4″ W), near an urban center, and about 4 km from the county seat. It historically included about 7900 ha and has recently been increased to 15,320 ha. LGSP has extensive green conserved areas (IEF 2018).

Both study locations are seasonal deciduous or dry forests (Mata Seca) with rocky outcrops. The landscape is predominantly hilly, characterized by limestone massifs, with an abundance of stones and pebbles of all sizes scattered across the ground. The dominant climate is sub-humid hot tropical, with two well-defined seasons: a dry season (May to September) and a wet season (October to April). The mean annual temperature is 24.2°C, with a mean annual maximum of 29.4°C and a mean annual minimum of 16.3°C (IEF 2018). The mean annual rainfall is 1074 mm (IEF 2018).

2.3 Data Collection

We carried out 21 days of surveys (for 2–3 days per month, with researchers divided into two teams to survey both study locations at the same time) between January and September 2018. We systematically walked through both research locations in order to find sites with evidence of stone tools used for hammering open-encased fruits. We used pre-existing trails, or if necessary opened new trails, that in total traversed the entire length of the forest (about 5 km of trails). Our surveys lasted from about 9 a.m. to 5 p.m., or when it started to get dark. In SRL we walked along trails looking for trees that had encased fruits that might require tool use to open. When a tree with encased fruits was found, we looked for rocky outcrops surrounding the tree in a 100 m radius to search for potential anvil sites. In LGSP, we focused on all the areas that park guards suggested for Sapajus xanthosternos stone tool use to break open fruits (about 500 m of trails). These areas are located near the park administration buildings and on trails used by visitors. Some parts of both parks were inaccessible, but we did walk in all the areas with potential tool use sites (i.e., with rocky outcrops, trees with encased fruits, and/or reports of monkey tool use from locals).

We characterized sites as stone tool use sites if they had a flat stone embedded in the substrate (anvil) associated with at least two of the following items: (1) white strike marks in the stone showing that it had been hit repeatedly with another stone; (2) smaller loose stone(s) with white marks (hammer) indicating that they had been used to strike multiple times, found on top of the anvil or in a 30 cm radius from it (Visalberghi et al. 2007; Falótico et al. 2017); and (3) remains of the shells of the fruit processed on top of the anvil or in a 30 cm radius from it (Canale et al. 2009; Mendes et al. 2015) (Figure 2). When a stone tool use site was identified, we registered the following data: composition (type of rock) and size of anvil (area, perimeter, and height); material and weight of hammers; presence or absence of processed fruits; species of fruit processed; distance from the anvil to the closest tree; distance from the anvil to the nearest source tree; and estimated degree of vegetation cover (see below). In this study, we qualitatively defined a “Nutcracking Area (NCA)” as a concentration of stone tool use sites associated with one or more source trees and in close proximity to one other.

We measured the dimensions of the anvil and distance to trees with 50 m fiberglass tape. The anvil area was calculated considering the shape of an ellipse, measuring the major and minor diagonal (see Moraes 2013). The height of the anvil was calculated by selecting and measuring the tallest of the four sides (Mendes et al. 2015). The hammers were weighed using a digital balance with a precision of 5 g and a maximum capacity of 50 kg. On rare occasions, we found two stone fragments next to one another, almost together, that were clearly the broken pieces of a larger hammer; in these cases, we weighed the pieces together and registered the total weight as one single hammer stone. For all the other stones on anvils, if it was not possible to be certain that smaller stones had been part of a larger hammer stone, all the stones were weighed individually. The distances from the anvil to the trees were measured by positioning the tape measure on the tree of interest and extending the tape to the closest edge of the anvil.

The degree of vegetation cover was estimated visually, along a scale from 0 to 4, in which 0 represented no vegetation cover and 4, a fully enclosed canopy (Visalberghi et al. 2007). All the fruit species processed on the anvils were already known to the researchers, and they were collected for confirmation of botanical identification. All the anvils and hammer stones were marked with permanent ink so as not to be re-registered in new searches for tool use sites. Marking the hammers also allowed us to verify if the same hammers were moved over time onto different anvils. In terms of phenology at the field sites, we qualitatively characterized nut availability on the ground for each month of study.

Additionally, we installed camera traps (Brand Bushnell, Model Trophy cam. 20 MP, 720p HD video with audio, 32 LED night vision flash, full color, high-resolution picture settings [3, 8 or 20 MP], 0.7 s frame speed, temperature range [−5° to 140° F]) in the study areas, placing them at hammer sites with evidence for heavy use by the monkeys, based on the quantity of processed fruit remains found around the hammer sites. Cameras were installed monthly and always removed after a week of data collection. The objective of camera trap use was to provide direct evidence of stone tool use by unhabituated capuchin monkeys (see Torralvo et al. 2017; Barrett et al. 2018). Prior to this study, Sapajus xanthosternos had only been inferred to use tools through indirect methods, such as observations of assemblages of anvils, hammer stones, and processed fruit in areas S. xanthosternos was known to inhabit (Canale et al. 2009).

3 Results

In LGSP, we found only one Nutcracking Area (NCA 1) (Figure 3), that included 50 stone tool use locations, all with processed remains of Macaúba (Acrocomia aculeata) (Figure 4). Sixteen Macaúba palm trees were present in this area and associated with one or more tool use locations, at a mean distance of 7 m between the palm tree and the tool use site.

In SRL, we encountered six distinct Nutcracking Areas (NCA 2–7, Figure 5). NCA 2 included 61 tool-use locations, and NCA 3 included 14 tool-use locations; these all included remains from processing Cansanção (Cnidoscolus pubescens) (Figure 4). In NCA 2 and 3, there were only two Cansanção trees associated with all 61 and 14 tool-use sites, respectively, at an average distance of 12 m from the palm tree to the anvil.

NCA 4, 5, 6, and 7 contained respectively 2, 5, 25, and 12 tool use site locations, all with processed remains of Guariroba (Syagrus oleracea) (Figure 4). The Guariroba palms also were associated with one or more NCA, with a mean distance of 13.5 m from the palm tree to the anvil. NCA 4 and 6 had only one Guariroba palm associated with 2 and 25 respective tool use sites. NCA 5 had 5 Guariroba palms associated with 5 tool-use locations, and NCA 7 included 12 stone tool-use localities associated with 4 source trees.

Across the seven Nutcracking Areas in the study, we found a total of 169 anvils and 244 hammer stones used by Sapajus xanthosternos. Table 1 provides the mean dimensions of hammers and anvils. In all areas, the anvils were made of limestone and were embedded in the ground (Figure 6). Anvil dimensions included a mean area of 1.45 m² (±4.15), a mean perimeter of 3.84 m (±3.91), and a mean height of 33.93 cm (±68.57). The anvil with the largest area that we registered in the study was from NCA 3, at 23.62 m². This anvil provided an extensive and continuous rocky substrate, on which eight hammers were found, along with a large amount of processed fruit remains and strike marks on various points of its surface. None of the anvils had depressions like those described for S. libidinosus in Fazenda Boa Vista. In general, anvils had irregular surfaces, with a lot of pores, as is characteristic of limestone. The pores often contained fruit shell remnants from stone tool processing.

Out of all of the anvils, 85.7% were within 2 m proximity to a tree (Table 2). The mean distance between an anvil and the closest tree was 1.17 m (±1.03) (Table 2). Only 3 anvils were more than 4 m from a tree. In contrast, the mean distance of anvils from source trees was 11.81 m (±10.08), with nearly half of anvils (46.2%) more than 10 m from a source tree (Table 3). In total, only 18.93% of anvils did not have any remains of processed fruit on or near them.

In the rainy season, all anvils had vegetational cover, with more than 60% of the canopy closed. However, in the dry season, all the anvils were found with no or very little vegetational cover. Across all the NCAs, it was most common for an anvil to be associated with just one hammer stone (47% of anvils), but anvils could have up to 8 hammer stones (see Supporting Information). Out of all of the anvils, only 4% did not have a hammer on or near them, although these anvils without hammers had strike marks and remains of processed fruits, indicating that they were also likely used as hammering sites. All the hammer stones marked in the study remained on the same anvil for the duration of the study. Hammer stones were all limestone cobbles (Figure 6). The mean hammer stone weight was 1.198 kg (±0.883). Table 4 provides the mean hammer stone weight for each Nutcracking Area (NCA) in the study. The heaviest hammer recorded was from NCA 3, at 7.675 kg (Table 4), found on an anvil with remains of Cansanção fruit shells. Hammers used to process Cansanção fruits had the highest mean weight (1.438 kg ± 1.031) and those used to process Macaúba (at LGSP) had the lowest mean weight (0.972 kg ± 0.728) (Table 5). The GLM testing the differences in stone tool weight used across the three encased fruits showed a significant effect of resource species on hammerstone weight. The stones used to process Cansanção were heavier than the ones used to process Guariroba and Macaúba (GLM, Intercept 6.952e-04, df = 239, t-value Guariroba 2.775, p < 0.01; t-value Macaúba 3.42, p < 0.001) (Figure 7).

During the entire study from January to September in the two locations, we found evidence of stone tools used for nutcracking only during the dry season, from May to September. From January through April, no new processed fruit shells were observed, and no new positioning of hammers on anvils, nor new hammers were observed. In terms of phenology, the nuts for the three tree species of interest were only present during the dry season; no wet or dry fruits from these species were found on the ground at either of the two field sites during the wet season. In other words, the monkeys were limited to the dry season for nutcracking, because that is the only time these nuts were present.

Throughout the project, our camera traps recorded several short videos (60 s long) showing Sapajus xanthosternos breaking encapsulated fruits using hammer stones (Embedded Media: Videos 1-6). All of these videos were recorded between May and September, during the regional dry season. These videos provide the first direct records of stone tool use in this species.

We also compared the hammer size used by Sapajus xanthosternos to break open Macaúba fruits in our study with the hammer size for S. libidinosus using tools to access the same fruit species. Comparing hammer size for Acrocomia aculeata at LGSP in our study and Ubajara National Park (Falótico et al. 2024), we found no significant difference in hammer weight for S. xanthosternos and S. libidinosus when opening the same fruit species (GLM: df = 323, t-value 1.679, p = 0.094).

4 Discussion

In this study, we characterize for the first time the stones used as hammers and anvils to break open resistant fruits by wild populations of Sapajus xanthosternos in two locations in northern Minas Gerais state, Brazil. When comparing our findings for S. xanthosternos to that in the literature for S. libidinosus, the mean area of anvils used by S. xanthosternos was a little smaller than those used by S. libidinosus at Fazenda Boa Vista, Piauí state, Brazil (Visalberghi et al. 2007). Most of our anvils were relatively small and low to the ground (often around 1 cm high), so a monkey could not fit on top of them but could stand next to the anvil to break fruit upon it. Even the small anvils usually had hammer stones on top of them. S. xanthosternos left hammers on anvils so they were available for future use, reducing energetic costs for searching and moving hammers.

Another hypothesis of Visalberghi et al. (2007) relevant to our study is the use of anvil properties to secure the fruits during nutcracking as a strategy to economize energy spent hammering. The various shallow pores, characteristic of the limestones used as anvils by S. xanthosternos in our study, provided the same function as the depressions in anvil stones found for S. libidinosus at Fazenda Boa Vista (Visalberghi et al. 2007). S. xanthosternos appeared to take advantage of these pores in the rock to fix the fruits in place to hammer them more easily, so nuts would not roll when hit. In our videos from the camera traps, we were able to see clearly that monkeys placed fruits in the rock pores before hitting them with the hammer stone.

The mean anvil height in this study was even shorter than the mean 83 cm found at Fazenda Boa Vista for Sapajus libidinosus (Visalberghi et al. 2007). S. xanthosternos used rocks lower to the ground as anvils, minimizing the energetic costs of carrying the hammers on top of them, as described for S. libidinosus. The anvils from our study that differed greatly from the mean height had fewer hammers on them as well as lighter hammers; this suggests a relationship between the height of the anvil and the weight or quantity of hammers found on the anvil.

The distance between anvils and the nearest source tree at SRL was often quite far, with distances above 30 m found in the study (Table 3). This is different than the findings for Sapajus libidinosus at Fazenda Boa Vista (Visalberghi et al. 2007), Serra das Confusões, Piauí state (Falótico et al. 2018) and in Goias state (Mendes et al. 2015). The distance from the anvil to the nearest source tree at our other site LGSP was more similar to that found for S. libidinosus sites, with most source trees found within 3 m of the anvils, fitting better with the hypothesis that anvil sites are preferentially chosen for their proximity to source trees, to economize the energy spent transporting fruits to the anvils. It is worth noting that most of our SRL Nutcracking Areas were in very steep regions, where a source tree was often found in the highest part. When fruits fell from those trees, they could roll closer to the Nutcracking Areas, which helps to explain the fact that we found some anvils quite distant from a source tree.

We were also interested in whether, as seen in Sapajus libidinosus in Fazenda Boa Vista (Visalberghi et al. 2007), S. xanthosternos would use tool use sites with dense vegetational cover, that might afford protection from aerial predators, and that were close to at least one tree that could serve as an escape route from the ground from other predators. Our results were not similar to those from S. libidinosus in terms of vegetation cover. Instead, the Seasonal Deciduous Forest (Mata Seca) characterizing our sites lost about 85% of the canopy during the dry season (Pezzini et al. 2008), exactly the time of year that the monkeys used tools with the highest frequency. For this reason, tool use by S. xanthosternos most often occurred when there was little to no vegetation cover at the Nutcracking Areas. However, we did find that anvils in our study almost always had a nearby tree (less than 2 m away) that could serve as an escape route, even closer than the average distance from the anvil to the nearest tree (about 3 m) found by Visalberghi et al. (2007) for S. libidinosus at Fazenda Boa Vista. In other words, S. xanthosternos used anvil stone locations that included a nearby tree as a potential escape route. As S. xanthosternos tool use behavior occurred in areas without vegetation cover, this escape route strategy might be all the more necessary, since vulnerability to predators would likely be increased during noisy hammering behavior near or on the ground.

Even though there were ample stones of various sizes and shapes available throughout our study areas, the hammers we found in the study had a higher mean weight than those described in most Sapajus libidinosus studies (Visalberghi et al. 2007; Ferreira et al. 2010; Falótico and Ottoni 2016; Falótico et al. 2018, 2022), with the exception of Parque Nacional da Chapada dos Veadeiros (Falótico et al. 2022) (see Table 6), and the mean hammer weight in this study was heavier than that previously reported for S. xanthosternos (Canale et al. 2009). However, when we compared hammer weight between S. libidinosus and S. xanthosternos to access the same resource (Acrocomia aculeata nuts), there was no difference in hammer weight.

We recorded the heaviest hammer stone (7.675 kg) reported for tool use in the genus Sapajus to date, exceeding the heaviest weight of a hammer stone previously reported for S. xanthosternos (3.025 kg) (Canale et al. 2009) or for S. libidinosus (5.7 kg at Chapada dos Veadeiros National Park) (Falótico et al. 2022). We found this stone on an anvil along with remains of processed fruit, with strike marks on both the anvil and the stone itself. Our inclusion of this stone as a hammer was based on indirect evidence, using the same criteria for indirect evidence as for the other hammers in this study and as used in other studies (Canale et al. 2009). The fruit processed in this Nutcracking Area (NCA 3) is Cansaçao, a thorn-covered fruit that is not consumed by people in the region. An adult male S. xanthosternos can reach 4.8 kg (ICMBio 2018), so it is likely that such an individual could lift a stone of 7.6 kg. S. xanthosternos tends to be larger and heavier than S. libidinosus, which may allow this species to lift heavier stones. However, it is important to note that we did not measure force or body weight for the animals in this study. In any case, our results suggest that S. xanthosternos may be able to exploit food resources bigger and more difficult to open than those explored by S. libidinosus when these resources are available.

At SRL the hammers used to break open Cansanção were significantly heavier on average than those used to break open the smaller fruit Guariroba, providing evidence in favor of the hypothesis proposed by Visalberghi et al. (2007) that capuchin monkeys select hammers on the basis of the characteristics of the target fruits. Sapajus xanthosternos, like S. libidinosus as described in Ferreira et al. (2010) and Falótico et al. (2019), seemed to be able to judge and choose which stone tool was ideal for breaking open a given fruit, as larger, more resistant fruits could be more easily opened with heavier stones, but avoiding excess force, on the other hand, as smaller fruits may not require heavy stones for efficient processing.

In contrast to Sapajus libidinosus at Boa Vista (Visalberghi et al. 2007), S. xanthosternos was not observed to transport the same hammer from one anvil to another. Throughout the length of our study, all the hammer stones stayed at their original anvils, only changing position on that same anvil. Our camera traps did not register any transport of hammers either. The behavioral difference may be explained by differences in resource availability in the two studies. In the work of Visalberghi et al. (2007), hammer stones were a limited resource and transported from anvil to anvil, unlike our two study areas where stones of different weights were left on the same anvils over time, possibly to facilitate future hammering sessions.

In conclusion, in our study, Sapajus xanthosternos showed a degree of planning in hammering behavior, as described for S. libidinosus. We characterized tool use for the first time in this species, and we confirmed the presence of a group of this Critically Endangered primate using tools in a protected area, namely, the LGSP. Conservation prioritization for this park should be extremely high, as it offers an enormous opportunity to study the origins of a non-human primate culture including tool use. Protecting these populations of S. xanthosternos which use tools in order to access fallback or seasonally preferred foods in dry environments is important, especially considering that climate change is predicted to cause an expansion and intensification of the dry environment within their distribution (Graham et al. 2016). Also urgent would be increasing the connectivity of habitats across the already highly fragmented range of S. xanthosternos, to encourage cultural diffusion of tool-use behaviors as forested areas become more arid. Further studies could clarify if these nuts provide essential energy intake during the dry season, and if they are treated as fallback or preferred foods. As Izar et al. (2022, 4090) suggest, stone tool use in capuchins “improves diet quality by providing a rich source of fat and carbohydrate energy which stabilizes dietary macronutrient balance.”

Hammering with stone tools has now been observed directly in the wild for five Sapajus species (S. libidinosus, S. flavius, S. nigritus cucullatus, S. apella, S. xanthosternos) (Steinberg et al. 2022; Lima et al. 2024; dos Santos Gutierres et al. 2025; Medeiros et al. 2025). The most commonly observed tool use type across Sapajus is the use of hammer stones to break open hard-encased fruits or shellfish on anvils, and there is archeological evidence that S. libidinosus has been using stone tools for nutcracking at Serra de Capivara for at least 3000 years (Falótico et al. 2019). Stone tool use has also been observed in two species of gracile capuchins, Cebus imitator (Barrett et al. 2018; Monteza-Moreno et al. 2020) and C. yuracus (Araujo et al. 2022), but is absent in many well-studied populations of wild Sapajus species (Steinberg et al. 2022), and apparently absent in most Cebus species, although very few of them have been studied in the wild. In this paper, we characterize stone tool use in S. xanthosternos, the most distantly related robust capuchin to S. libidinosus, and we find similarities between the two species' tool use behavior including that both use heavier tools for nuts that are more difficult to break open, and both tend to have anvil sites near trees that may serve as escape routes. We also found some distinct differences, such as seasonal tool use in S. xanthosternos rather than the year-round tool use seen in S. libidinosus.

When we consider these similarities and differences in an evolutionary context, trying to understand how tool use changed and spread during hominin evolution, our study shows that many stone tool use characteristics are conserved and shared across the radiation of extant Sapajus, suggesting that their shared ancestor had the cognitive capability (and perhaps behavioral predisposition toward banging objects against hard surfaces, ubiquitous among capuchins) that permitted or facilitated developing this tradition independently in several different dry environments. The traditions themselves seem to have evolved in parallel (versus being a synapomorphy of the genus), given that they are present in some populations of some species and are completely absent in other species, at least in today's wild populations. There may have been an evolutionary environment in which the ancestral population acquired this capacity, that has been retained across all the diverse species within the genus. This means there may have been the cognitive capability for stone tool use over the last 1.5–3.3 million years in this genus, based on estimated divergence times for S. libidinosus and S. xanthosternos (Lima et al. 2018; Martins et al. 2023). What is still unclear is how continuous stone tool use has been as an adaptation for robust capuchin populations that exist in dry habitats, and whether cultural diffusion may have occurred between different capuchin species at borders of their ranges over evolutionary history. In relation to hominin evolution, this study helps to clarify that cognitive capability for stone tool use for nutcracking may be present in populations where tool use is absent; it suggests that tool use does not necessarily persist in populations once acquired, but instead, the ability is maintained and the behavior may be able to resurface when environmental conditions change to require it.

Author Contributions

Rayssa Durães Mainette: formal analysis (lead), investigation (lead), methodology (equal), writing – original draft (equal). Jessica W. Lynch: conceptualization (equal), investigation (equal), writing – original draft (equal), writing – review and editing (lead). Patrícia Izar: conceptualization (equal), supervision (equal), writing – review and editing (equal). Gustavo Canale: conceptualization (equal), writing – review and editing (equal). Waldney Pereira Martins: conceptualization (equal), funding acquisition (lead), investigation (equal), methodology (equal), project administration (lead), supervision (lead), writing – original draft (equal), writing – review and editing (equal).