The reconstruction of ancestral dispersal patterns has major implications for the evolution of modern human behavior. Dispersal patterns determine which sex remains with kin and reaps the benefits of that inclusive fitness; this philopatric sex is more likely to form strong affiliative bonds with their same-sex siblings and cousins. In chimpanzees, these bonds form the foundation for male coalitions that influence everything from territory patrols to mating (Gilby et al., 2013; Mitani, 2009; Nishida and Hosaka, 1996). In most species of monkey, females are philopatric, and female coalitions defend resources and protect infants from potentially infanticidal males (Hrdy, 1999; van Schaik, 1989).

Strontium isotope ratios can be used to understand migration and landscape use of extant and extinct species (e.g., Hamilton et al., 2021; Makarewicz and Sealy, 2015; Vautour et al., 2015; Wallace et al., 2019). Strontium has an ionic radius similar to calcium and therefore can substitute for calcium in biotic tissues. Strontium isotope ratios (87Sr/86Sr) in soil, plants and local animals vary based on a variety of factors, but primarily based on bedrock age, composition and weathering (Bataille et al., 2020; Graustein, 1989; Hurst and Davis, 1981). Strontium isotope ratios are incorporated into bone and tooth enamel as they form and reflect the local 87Sr/86Sr of the foods eaten during formation. A comparison between tissues formed when animals are juveniles (such as tooth enamel) and local 87Sr/86Sr to where that animal lives as an adult (for example, from bones which are replaced throughout life or the plants growing in the local area) could identify those hominins which were philopatric and those which dispersed. This was the approach taken by Copeland et al. (2011), who examined strontium isotope ratios from tooth enamel of Australopithecus africanus and Australopithecus (Paranthropus) robustus and compared these to locally sampled vegetation to determine how many individuals from each species were ‘local’ vs. ‘non-local,’ and whether the probability varied by tooth size (a proxy for sex).

However, many areas surrounding fossil deposits are extremely geologically complex, and the simple ‘local’ versus ‘non-local’ distinctions for strontium isotope ratios may not be the most accurate way to capture the variation incorporated into tooth enamel in slow-growing, wide-ranging taxa such as the australopithecines. For example, studies on modern primates have shown that a species’ range size can impact the reliability of dispersal pattern reconstructions (Hamilton et al., 2021). The purpose of this study is to infer likely dispersal patterns for South African hominins and to compare these with previous dispersal reconstructions. To do this, we use a novel analytical method, recently validated in a modern primate community (Hamilton et al., 2021), to infer sex-biased dispersal based on previously published strontium isotope ratios from hominin tooth enamel and local plants (Balter et al., 2012; Copeland et al., 2011; Sillen et al., 1998).

In August of 1936, Robert Broom discovered the first adult A. africanus (TM 1511) at Sterkfontein Cave, a limestone quarry located about 50 km (km) northwest of Johannesburg, South Africa. In 1948, approximately 1 km away, the discovery of Australopithecus (Paranthropus) robustus specimens opened excavations at Swartkrans Cave. Together with over a dozen other caves in the vicinity, these fossil-rich deposits are a protected UNESCO World Heritage site referred to as the ‘Cradle of Humankind’ and are one of the most prolific paleontological fossil deposits in the world, yielding remains of several hundred hominins (Tobias, 2000). The deposits in Sterkfontein record the time period between 2.6 and 1.07 Ma ago across five geological members (Herries and Shaw, 2011; Partridge et al., 1999), while the four members of Swartkrans span from 2.0 to 0.6 million years ago (Herries et al., 2009).

Swartkrans and Sterkfontein caves are part of a system formed by the Blaauwbank River and today are located on either side of the small waterway, approximately 30 m above the modern river (Avery, 2001). The Cradle of Humankind is located on the western edge of an Archean basement gneiss formation referred to as the ‘Johannesburg Dome,’ a 3.1-billion-year-old basement complex (Anhaeusser, 2006; Robb et al., 2006). The Dome is surrounded by sedimentary and volcanic deposits including the fossil-bearing Malmani dolomites (Eriksson et al., 2006). The Malmani dolomites are part of the Malmani Subgroup of the Chuniespoort Group and Transvaal Supergroup. This supergroup is between 2.2 and 2.6 billion years old (Eriksson et al., 2001; Eriksson and Reczko, 1995). The Malmani Subgroup is characterized by stromatolitic dolomite with chert inbeds; the dolomite itself is rich in iron and manganese (Martini et al., 2003; Sutton, 2013). The dolomite of the Malmani Subgroup is composed of five formations: from base to top, the Oaktress, Monte Cristo, Lyttleton, Eccles and Frisco Formations (Kent, 1980). Swartkrans and Kromdraai caves are located entirely on the Monte Cristo Formation while Sterkfontein is at the contact between the Oaktree Formation and Monte Christo Formation. The Oaktree Formation is on top of Black Reef quartzite, which is primarily quartzite and conglomerate. The Oaktree Formation is characterized by chert-poor dolomite deposits and thick horizons of shale. By contrast, the Monte Christo Formation is rich in chert within the shale horizons (Dirks and Berger, 2013; Sutton, 2013).

Today, Swartkrans and Sterkfontein are located on a strip of exposed Malmani dolomite approximately 10 km wide (Fig. 1). To the immediate northwest there are outcrops of the Pretoria Group, Timeball Hill shale and Hekpoort andesite/basalt dating to approximately 2.2 billion years ago. Beyond these outcrops are the Daspoort quartzite and alternating bands of diabase and undifferentiated Quaternary sedimentary deposits. To the southeast of the caves is the Witswatersrand Supergroup (3.0–2.7 billion years ago), underlain by 3-billion-year-old Archean granites which outcrop to the east of the caves. Much farther to the south of the cave systems are outcrops of the Karoo Supergroup (Late Carboniferous to Early Jurassic, ∼300–180 Ma; Johnson et al., 1996), which likely covered most of the Johannesburg Dome based on erosional remnants (Dirks and Berger, 2013).

It is important to explore how this current landscape geology may have been different 4–2 Ma. In earlier studies, Dirks et al. (2010) and Dirks and Berger (2013) report evidence for fairly rapid erosion of the dolomite surface. This implies that during the time of early hominin occupation, the Timeball Hill shale might have covered a more extensive area than today and that remnants of the Karoo Supergroup would have covered more of the land surface around 4–2 Ma. The Malmani dolomite exposure, on which Swartkrans and Sterkfontein caves sit, could have been “narrower by up to several kilometers” (Dirks and Berger, 2013:127) 2 Ma. Such differences would have dramatically altered the bioavailable strontium isotopes because the Timeball Hill shale has distinctly higher bioavailable isotopic values (∼0.75) than the dolomite (∼0.721–0.734) (Copeland et al., 2010a; b, 2011), while most Karoo Supergroup sediments have distinctly lower bioavailable isotopic values (∼0.709–0.718) (Stewart et al., 2020; Copeland et al., 2022).

Later geological work by Dirks et al. (2016) cast doubt on the rapid erosion rate of the dolomite and suggested instead that the land surface in the Cradle of Humankind has not changed by much in the last few million years. In agreement with this, the majority of fauna reported in Copeland et al. (2011), Balter et al. (2012) and Sillen et al. (1998) have strontium isotope values that fall below the range for the Timeball Hill shale and above that of the Karoo. Given this lack of isotopic matching, even in small rodents, it is unlikely that Timeball Hill and Karoo Supergroup outcrops formed a central part of the Swartkrans/Sterkfontein hominin home range.

Swartkrans and Sterkfontein caves are located along the banks of the Blaauwbank River, which today is approximately 4 m across at its widest point. However, 4–2 Ma this waterway would have been much more substantial, carving the valley that currently separates the two localities (Watson, 1993). Vrba’s broad hypothesis of increasing aridity throughout the last three million years in Africa is consistent with environmental reconstructions from these South African sites (Rowan and Reed, 2015; Vrba, 1975) and faunal isotopic studies (Lee-Thorp et al., 2007). Broadly speaking, paleoecological reconstructions of the Cradle of Humankind indicate mosaic environments, consisting of gallery forests along the edges of the larger rivers extending into more open grasslands or grassy woodlands with a mixture of C3 and C4 vegetation (Dirks and Berger, 2013). For example, the presence of Antilopini and Alcelephini bovids in all three members of Swartkrans cave suggests available grassland ecotones, while hippopotamus fossils indicate permanent sources of water, and larger fauna such as elephants and giraffes suggest some extensive open woodland habitat in the area as well (de Ruiter, 2003). These riverine and riparian forest environments generally decreased through time, leading to increasingly open habitats (Avery, 2001; Reed, 1997; Vrba, 1975; Watson, 1993).

During the time of the A. africanus and A. robustus specimens included in previous strontium isotope studies in the area, Sterkfontein was most likely an open- to medium-density woodland with nearby areas of brush and thicket with a riparian forest along the banks of the Blaauwbank (Avery, 2001; Reed, 1997; Vrba, 1975). Swartkrans was likewise in or near a riparian forest and/or edaphic grassland (Reed, 1997; Watson, 1993), but with C4 food sources nearby (Leichliter et al., 2017). The presence of these riprian forests is important in our interpretation of isotopic results. Animals with a preference for food found in riparian areas can differ in their strontium isotope ratios (Hamilton et al., 2019; Sillen et al., 1998) because of geochemical dynamics unique to those riparian zones. The isotopic composition of surface water is primarily due to chemical weathering (Capo et al., 1998); because some minerals dissolve more readily than others, flowing water often does not have an isotope signature that matches the underlying bedrock. Isotopically, riparian zones can look quite different from the surrounding non-riparian area (Aubert et al., 2002; Banner, 2004; Bentley, 2006; Montgomery et al., 2009). It is therefore important to consider the potential influence of riparian food sources when interpreting data from early hominins in the area.

Strontium isotope ratios vary according to the age and type of bedrock in a given area. The underlying isotopic ratio of the bedrock is incorporated into the soil and the biologically available strontium is taken up by plants growing in the area (Bataille et al., 2020; Beard and Johnson, 2000; Bentley, 2006; Graustein, 1989; Holt et al., 2021; Hurst and Davis, 1981). Animals acquire most of the strontium in their tissues from their diet (Nielsen, 2004); therefore, local plants at the base of a food chain provide a reliable baseline for the biologically available strontium isotope ratios across a given area. Calcium-bearing faunal tissues reflect the local 87Sr/86Sr consumed over the time of that tissue’s formation. 87Sr is formed as the daughter isotope of 87Rb; given that 87Rb has a half-life of 48.8 billion years, the ratio of 87Sr/86Sr in a given lithology is stable over timescales of interest to human evolution (Bentley, 2006; Faure and Mensing, 2005; Holt et al., 2021). Thus, the 87Sr/86Sr ‘fingerprint’ in faunal tissues (bones, hair, tooth enamel, etc.) reflects the geologic area in which the animal lived (or at least ate) during that tissue’s formation (Makarewicz and Sealy, 2015; Price et al., 2002). Because different tissues form at different developmental periods and turn over at different rates throughout life, it is possible to track changes in location through the life cycle by comparing the 87Sr/86Sr ratios of different tissues to each other or to the surrounding environment.

Sex biases in dispersal at sexual maturity is a defining socioecological characteristic of primates, influencing other behavioral traits ranging from rates of aggression and affiliation (Goodall et al., 1979), to patterns of alloparenting (Matisoo-Smith et al., 1997), to coalition formation (Mitani, 2009). The ancestral condition for the human clade is currently unresolved, with data pointing to numerous disparate conclusions (Hill et al., 2011; Murdock, 1981; Wilkins and Marlowe, 2006). An accurate reconstruction of dispersal patterns in early hominin species would contribute substantially to this persistent question.

A few previous studies have attempted to reconstruct such dispersal patterns for earlier hominin species (e.g., Balter et al., 2012; Copeland et al., 2011; Lockwood et al., 2007; Sillen and Balter, 2018). Copeland et al. (2011) used laser ablation to measure the strontium isotope ratios from the enamel of 11 A. robustus teeth from Swartkrans cave dating to about 2 Ma and eight A. africanus teeth from Sterkfontein cave dating to 2.2–0.6 Ma (Herries et al., 2009; Herries and Shaw, 2011). They concluded that more of the small teeth (50%, n = 10) had ‘non-local’ strontium isotope signals compared to the large teeth (11%, n = 9) or any other included fauna (15%, n = 37). Due to high levels of sexual dimorphism in australopithecines, researchers assumed that the small teeth represented females of each species and the large teeth represented males. Therefore, Copeland et al. (2011) concluded that, similar to chimpanzees today, males remained in their natal communities throughout life while females dispersed to other communities to reproduce.

Copeland et al. (2011) used geologically defined boundaries to delineate areas with potentially different bioavailable strontium isotope ratios. Many of these are extremely constrained areas, particularly when considering the large home range potential for a large, bipedal hominin. The Malmani dolomite on which the Swartkrans and Sterkfontien caves are located, for example, is a long but narrow ribbon of geological bedrock often no more than 10 km wide. Both A. africanus and A. robustus were bipedal (Stern and Susman, 1983; Wood and Constantino, 2007; Haile-Selassie et al., 2010; Senut et al., 2018), a highly efficient form of locomotion for covering large distances (Alemseged, 2023). Even quadrupedal modern chimpanzees have home ranges that range from 15 km2 in Kibale National Forest (Chapman and Wrangham, 1993) to 27.4 km2 in Mahale (Nakamura et al., 2013). Both Kibale and Mahale are forested habitats; in open habitats such as Fongoli, Senegal, chimpanzees can range over 85 km2 (Pruetz, 2006). The mean estimated local group range for African non-equestrian hunter-gatherers such as the Hadza is 471 km2 (with hunting behavior driving a large portion of this range size; Marlowe, 2005). Bipedal australopithecines in an open habitat are likely to have had home ranges intermediary between knuckle-walking chimpanzees and hunting modern humans. This larger range, likely one that would have encompassed numerous geologic zones, would have produced an ‘averaging’ effect on the isotopic composition of their tissues, potentially distorting a local/non-local signal defined purely by geology.

The isotopic ranges of each of the geologically defined units surrounding the South African fossil sites also overlap extensively (Fig. 1). If the geologic substrate immediately adjacent to the Malmani dolomite, the Timeball Hill shale, had not been included in the ‘non-local’ category in Copeland et al.’s (2011) analysis—which is reasonable considering the likely daily ranging patterns of bipedal hominins—then none of the A. africanus teeth and only two of the A. robustus teeth become ‘non-local.’ Given the small sample size and most likely ecological conditions at the time, it is imperative to discuss this variability when drawing any conclusions. Methods that assign residency using only geologic boundaries are constrained by the assumption that those boundaries correlate with geospatially, geochemically meaningful differences. We propose that there is more potential to accurately reconstruct mobility patterns when relying on models that instead use independently derived isotopic clusters to form local boundary definitions.

Hamilton et al. (2021) developed and tested such a method using modern primates and environmental data from Kibale National Park in Uganda and showed that strontium isotopes can be used to reliably predict the primates' known philopatry and dispersal patterns. The researchers used hierarchical cluster analysis to statistically group together strontium isotope ratios of plants from across the park. This created isotopically distinct, geographically continuous clusters which were used to define local areas of unique, non-overlapping bioavailable strontium isotope ratios. They also devised an alternative method to the binary categorization of ‘local’ versus ‘non-local’ as a way to identify dispersal patterns. They calculated the difference (or ‘offset’) between a primate’s molar tooth enamel 87Sr/86Sr (which forms before any sex-biased dispersal events; Smith et al., 1994) and the average 87Sr/86Sr ratio of plants within the local isotopic cluster on which that primate was found. These offsets are then used to assign philopatric or dispersing individual status to each sample by comparing them to the offsets of other individuals of that same species, thus controlling for home range size, diet, habitat selection and other confounding variables. Those with offsets greater than the species' mean are categorized as dispersers while those falling below the species' mean are categorized as philopatric individuals.

Hamilton et al. (2021) found that using isotopic clusters to define local bioavailable strontium combined with the offset method led to accurate identification of the philopatric versus dispersing sex for all primate species, including chimpanzees. They were able to correctly assign philopatric/dispersing sex status to 71% of individuals in the study, which included six primate species of varying home range sizes and dispersal patterns. The model had a lower success rate for accurately identifying non-local or dispersing individuals compared to local/philopatric individuals (for example, 82% correct assignment of local individuals vs. 50% correct assignment of dispersing individuals for chimpanzees, n = 17). This pattern of errors makes the model inherently conservative; we are more likely to misidentify dispersing individuals as philopatric and thus be unable to draw any conclusions about sex bias in dispersal patterns, rather than the opposite. It is notable that at the species level, even with small samples sizes, the method correctly identified which sex dispersed and which remained philopatric for all primates included in the study. For all six species, a higher proportion of the dispersing sex had offsets above the species’ mean than below. Thus, while the method is only partially reliable at the individual level for assigning dispersal status, it functions quite reliably at the species level.

Additionally, Hamilton et al.’s (2021) method also showed predictable variability based on the home range size of the species: species with larger home ranges, such as chimpanzees, showed less offset separation between the sexes while those with smaller home ranges, such as olive baboons, showed greater separation of offsets between the sexes. This is currently the only published method for assigning philopatric or disperser status using strontium isotope ratios that has been tested for accuracy on a known primate population. The objective of the current study is to apply this method to previously published strontium isotope data from South African hominins and landscapes.