Biochronology of South African hominin-bearing sites: A reassessment using cercopithecid primates

Significance

This study provides updated age estimates of major South African hominin sites based on faunal correlations of cercopithecid monkeys. Importantly, we demonstrate that molar size is highly correlated with geological age in the Theropithecus oswaldi lineage, a common fossil cercopithecid, providing a chronometric tool not available previously. Contrary to some recent analyses, we find no evidence for hominin sites in South Africa significantly older than 2.8 Ma. Our results also suggest that some hominin sites are older than recently estimated (e.g., Swartkrans Member 3, Cooper’s D). Where faunal estimates and still-developing chronometric methods conflict, further research is needed, but our current results have implications for the timing of human evolution in South Africa during the Plio-Pleistocene.

Abstract

Despite recent advances in chronometric techniques (e.g., Uranium-Lead [U-Pb], cosmogenic nuclides, electron spin resonance spectroscopy [ESR]), considerable uncertainty remains regarding the age of many Plio-Pleistocene hominin sites, including several in South Africa. Consequently, biochronology remains important in assessments of Plio-Pleistocene geochronology and provides direct age estimates of the fossils themselves. Historically, cercopithecid monkeys have been among the most useful taxa for biochronology of early hominins because they are widely present and abundant in the African Plio-Pleistocene record. The last major studies using cercopithecids were published over 30 y ago. Since then, new hominin sites have been discovered, radiometric age estimates have been refined, and many changes have occurred in cercopithecid taxonomy and systematics. Thus, a biochronological reassessment using cercopithecids is long overdue. Here, we provide just such a revision based on our recent study of every major cercopithecid collection from African Plio-Pleistocene sites. In addition to correlations based on shared faunal elements, we present an analysis based on the dentition of the abundant cercopithecid Theropithecus oswaldi, which increases in size in a manner that is strongly correlated with geological age (r² ∼0.83), thereby providing a highly accurate age-estimation tool not previously utilized. In combination with paleomagnetic and U-Pb data, our results provide revised age estimates and suggest that there are no hominin sites in South Africa significantly older than ∼2.8 Ma. Where conflicting age estimates exist, we suggest that additional data are needed and recall that faunal estimates have ultimately proved reliable in the past (e.g., the age of the KBS Tuff).

Accurate chronology is a key contextual component to understanding human evolution, yet the geochronology of many African Plio-Pleistocene hominin sites have proven elusive ever since the first Australopithecus fossil was described (1). The advent and implementation of absolute dating techniques over the past 60 y have greatly improved the chronological context of human evolution, particularly in eastern Africa where tectonic activity has provided a wealth of volcanic sediments amenable to radiometric dating. Nonetheless, a number of sites, especially in southern Africa, remain disputed in terms of their geological age (e.g., Taung, Sterkfontein Members [Mbs.] 2 and 4, Swartkrans [SK] Members 1 through 3, Kromdraai B, etc.). While recent geological studies and chronometric dating techniques appear promising in their potential to address some of these concerns (e.g., ref. 2 and citations therein), results still range widely in some cases. For example, samples collected from within the Silberberg Grotto, including the flowstones that are associated with and below portions of the Stw 573 skeleton in Sterkfontein Mb. 2, were dated to ∼2.3 to 2.1 Ma based on Uranium-Lead (U-Pb) analysis of speleothems (2), but 3.67 Ma based on Al/Be analysis of sediment (3). Estimates for Swartkrans Mb. 3 based on electron spin resonance spectroscopy (ESR) alone range between 4.38 and 0.036 Ma (4).

Thus, faunal correlation still serves as an important independent “control” and tool for age estimation in many cases, particularly when there are large differences in age estimates obtained through newer chronometric methods. One of the most famous examples involves the age of the KBS Tuff in eastern Africa, which was disputed on the basis of differing radiometric estimates (e.g., refs. 5–7). Ultimately, faunal correlation of suid taxa played a crucial role in resolving the debate, demonstrating that the older age estimates were almost certainly incorrect (8, 9). The faunal evidence was later corroborated by further, more sophisticated, radiometric analyses and geochemical correlation across the Turkana Basin confirming a younger age close to 1.9 Ma (e.g., refs. 10–12).

The main objective of this analysis is to provide a revised assessment of African Plio-Pleistocene biochronology across hominin sites using cercopithecid monkeys, which are well suited to this task for several reasons. First, they are widespread and abundant members of African Plio-Pleistocene faunas and are found at nearly all hominin sites of this age (e.g., ref. 13). Second, they are a diverse group, and undergo a significant amount of both anagenetic evolution and species turnover during the Plio-Pleistocene (14). Third, a significant portion of their fossil record derives from sites with precise, independent geochronological control (e.g., refs. 15–27), which allows for well-documented genus and species age-ranges in certain cases. Finally, cercopithecids have previously been used and cited as key indicator taxa in faunal correlation analyses of hominin sites in eastern and southern Africa (28, 29).

Over the 30-plus years since the last major biochronological study using cercopithecid taxa was published, a number of new fossil African cercopithecids have been discovered and named (e.g., refs. 13, 30). Many new specimens have been collected or identified that represent additional locality and age range data for previously recognized taxa (e.g., refs. 30–33). Significant modifications have also been made to previously known species (30, 33, 34). Furthermore, geochronological control has improved for many assemblages (see above) and many entirely new assemblages have been described (e.g., refs. 31, 35–48). Therefore, an updated biochronological analysis of cercopithecid fossils is timely and provides an independent chronological control on recent estimates for South African hominin sites.

Results

Based on our updated assessment of the African Plio-Pleistocene cercopithecid fossil record, we provide a zonation and refined correlations between well-constrained eastern African localities and South African hominin-bearing fossil sites (Figs. 1 and 2 and Table 1). These African cercopithecid faunal zones build on and modify those originally proposed by Delson (28, 29) (see Materials and Methods for details). We further improve the accuracy and precision of age estimates in a number of cases based on a regression analysis of the well-documented evolutionary trend for increasing molar size in the Theropithecus oswaldi lineage (Fig. 3 and Table 2) combined with other independent constraints from the literature, including U-Pb dates and updated paleomagnetic correlations.

Fig. 1.

Ranges of select African Plio-Pleistocene cercopithecids from radiometrically dated, stratified sites in eastern Africa. For a complete list of all cercopithecid species and their ranges see SI Appendix, Table S3. Vertical axis indicates geological age in millions of years (Ma); a geomagnetic polarity timescale (81) is also shown for reference. Boxes represent chronological ranges of eastern African occurrences based on well-constrained and readily taxonomically diagnosable material, while whiskers indicate possible extensions due to less taxonomically diagnostic and/or less well-constrained occurrences. Note C. coronatus includes Cercopithecoides kimeui (30). Left column indicates African Cercopithecid zones identified in this analysis.

Fig. 2.

Distribution of select cercopithecids across South African sites. Left lists localities analyzed here in approximate chronometric order from oldest to youngest based on results from this study. Placement of different sites in African Cercopithecid zones from Fig. 1 are indicated on the Right.

Table 1.

Cercopithecids identified from Plio-Pleistocene sites in South Africa

Site	Cercopithecid monkey taxa	Hominin taxa	African Cercopithecid Zone
Sterkfontein Member 6/Post Member 6	Papio hamadryas ssp.	Homo sp.	Zone 8
Elandsfontein/Hopefield	Theropithecus oswaldi leakeyi	Homo sp.	Zone 6/7
Gladysvale*	cf ?Papio izodi Papio hamadryas angusticeps cf Cercopithecoides coronatus Theropithecus oswaldi leakeyi Papionini sp. indet. LARGE	Australopithecus cf africanus	Zones 4 to 7
Bolt's Farm Pit 10	Theropithecus oswaldi leakeyi Papionini sp. indet. LARGE	None	Zone 6
Swartkrans Member 3	Papionini sp. indet. LARGE Theropithecus oswaldi oswaldi Papionini sp. indet. VERY LARGE	Paranthropus robustus early Homo	Zone 5
Cooper’s D	Theropithecus oswaldi oswaldi Papionini sp. indet. LARGE Papionini sp. indet. VERY LARGE Colobinae sp. indet.	Paranthropus sp.	Zone 5
Swartkrans Member 2	Papio cf robinsoni Theropithecus oswaldi oswaldi Papionini sp. indet. VERY LARGE	Paranthropus robustus early Homo	Zone 5
Kromdraai A	Papio hamadryas angusticeps Gorgopithecus major Papionini sp. indet. SMALL cf Cercopithecoides sp. indet.	None	Zone 5
Swartkrans II	Papio sp. indet. Cercopithecoides sp. indet.	None	Zone 5
Cooper’s A-B	Theropithecus sp. indet. Papio hamadryas angusticeps cf Gorgopithecus major Cercopithecoides sp. indet.	None	Zone 5
Bolt's Farm Pit 6	Papio hamadryas angusticeps Cercopithecoides coronatus	None	Zone 5
Haasgat	Papio hamadryas angusticeps Cercopithecoides haasgati Cercopithecoides coronatus	None	Zone 5
Kromdraai B	Papio cf robinsoni Cercopithecoides coronatus	Paranthropus robustus early Homo	Zone 5
Skurweberg/Hoogland	Papio robinsoni Dinopithecus ingens T. o. oswaldi †	None	Zone 5
Swartkrans Member 1	Papio robinsoni Theropithecus oswaldi oswaldi Dinopithecus ingens Gorgopithecus major Papionini sp. indet. SMALL Cercopithecoides williamsi Cercopithecoides coronatus	Paranthropus robustus early Homo	Zone 5
Sterkfontein Member 5 StW53 Infill	Theropitheucs oswaldi oswaldi	early Homo	Zone 5
Drimolen Main	Papio robinsoni Cercopithecoides coronatus	Paranthropus robustus early Homo	Zone 5
Malapa Facies B and D	Papio hamadryas angusticeps	Australopithecus sediba	Zone 5
Bolt's Farm Pit 23	Parapapio broomi Papio robinsoni Cercopithecoides williamsi	None	Zone 4
Sterkfontein Member 4	Parapapio jonesi Parapapio broomi Parapapio whitei ?Papio izodi Papio sp. indet. Dinopithecus ingens ‡ Cercopithecoides williamsi	Australopithecus africanus and/or Australoithecus prometheus	Zone 4
Taung	Procercocebus antiquus ?Papio izodi Papionini sp. indet. SMALL Cercopithecoides cf williamsi	Australopithecus africanus	Zone 4
Sterkfontein Member 2/Silberberg Grotto	Parapapio jonesi Parapapio broomi ?Papio izodi Cercopithecoides williamsi	Australopithecus prometheus and/or Australopithecus africanus	Zone 4
Makapansgat	Parapapio jonesi Parapapio broomi Parapapio whitei Theropithecus oswaldi darti Cercopithecoides williamsi Colobinae sp. indet. LARGE	Australopithecus africanus and/or Australoithecus prometheus	Zone 4

Notes: Bold species names indicate the type locality for that taxon.

^*The fauna at Gladyvale is mixed and could span from ∼3.0 to 0.5 Ma over various external and internal deposits (e.g., see also SI Appendix, SI Text for recent summaries). We list the fossil monkeys found across all Gladyvale deposits here but note that they are not all found contemporaneously.

^†T. o. oswaldi is described from Hoogland, which may or may not be the same deposit as Skurweberg.

^‡D. ingens is identified at Sterkfontein Member 4 from specimens with STS numbers (STS 265, STS 365). Specimens with STS numbers are assumed to be from Member 4, but could possibly include mixed specimens from Member 5. See SI Appendix, SI Text and Tables S3 and S4 for further details.

Fig. 3.

Model for evolution of molar size in T. oswaldi lineage used to estimate geological age based on analysis of covariance using age in millions of years (Ma) as the independent variable (horizontal axis) and log₁₀ of molar area (mm²) as the dependent variable (vertical axis). Area calculated as indicated in the Lower Right Inset (length × the average of the mesial and distal breadths). Upper molars are shown by solid circles with solid trend lines and lowers by hollow circles and dashed trend lines. Color denotes tooth position: gold, 1; orange, 2; and red, 3. Approximate geological age ranges for recognized subspecies of the T. oswaldi lineage are shown by the alternating shaded boxes and labeled at Top. See SI Appendix, SI Text and Table S1 for further details.

Table 2.

Estimated geological age ranges for Plio-Pleistocene sites in South Africa

Site	Theropithecus mean (95% CI)	Combined final estimated range	Cradle sedimentation zones (2)
Elandsfontein/Hopefield	0.9 Ma (∼1.2–0.6 Ma)	1.2–0.6 Ma	-
Bolt's Farm Pit 10	1.3 Ma (1.9–0.7 Ma)	1.4–0.9 Ma	Sedimentation Zone 6 (<1.32 Ma)
Swartkrans Member 3	1.5 Ma (1.8–1.2 Ma)	1.8–1.4 Ma	Sedimentation Zone 5 (1.63–1.41 Ma)
Cooper's D	1.8 Ma (1.9–1.6 Ma)	2.0–1.6 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Swartkrans Member 2	1.9 Ma (2.7–1.2 Ma)	2.1–1.6 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Kromdraai A	—	2.1–1.6 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Swartkrans II	—	2.1–1.6 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Cooper's A-B	—	2.1–1.6 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Bolt's Farm Pit 6	—	2.1–1.6 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Haasgat	—	2.1–1.7 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Kromdraai B	—	2.1–1.7 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Skurweberg/Hoogland	—	2.1–1.7 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Swartkrans Member 1	1.9 Ma (2.0–1.8 Ma)	2.1–1.7 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Sterkfontein Member 5 StW53 infill	2.2 Ma (2.5–1.8 Ma)	2.1–1.7 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Drimolen Main	—	2.1–1.9 Ma	Sedimentation Zone 4 (2.0–1.82 Ma)
Malapa Facies B	—	2.3–2.0 Ma	Sedimentation Zone 3 (2.17–2.12 Ma)
Bolt's Farm Pit 23	-	2.6–2.0 Ma	Sedimentation Zone 3 (2.17–2.12 Ma) Sedimentation Zone 2 (2.62–2.28 Ma)
Sterkfontein Member 4	—	2.6–2.0 Ma	Sedimentation Zone 3 (2.17–2.12 Ma) Sedimentation Zone 2 (2.62–2.28 Ma)
Taung	—	2.8–2.0 Ma	—
Sterkfontein Member 2/Silberberg Grotto	—	<3.0 Ma-2.0 Ma	Sedimentation Zone 2 (2.62–2.28 Ma) Sedimentation Zone 1 (3.08–2.83 Ma)
Makapansgat	2.4 Ma (2.5–2.3 Ma)	2.7–2.6 Ma	-

Notes: “-” indicates Theropithecus data are unavailable or the site in question is outside of the Cradle of Humankind. Cradle Sedimentation Zones are listed as hypothesized in Pickering et al. (2).

The relationship between T. oswaldi molar dimensions and absolute geological age is tight and highly significant (P < 0.0001) across radiometrically dated sites, with geological age explaining more than 83% of the overall variance in molar area (Fig. 3 and SI Appendix, Table S1). Thus, given a reasonable sample size of T. oswaldi molar specimens at any African Plio-Pleistocene hominin site lacking absolute dates, our analyses demonstrate that basic dental dimensions provide an excellent chronometric tool, producing statistically constrained age estimates that were absent from previous analyses (Table 2). In addition, our African cercopithecid zonation (Figs. 1 and 2 and SI Appendix, Table S2) is reinforced by the regression results from the T. oswaldi data. Recent paleomagnetic correlations and U-Pb dates largely corroborate our cercopithecid-based assessments as well (Tables 1 and 2).

Estimated Ages for South African Hominin Sites.

Our estimated age ranges for South African hominin sites are broadly consistent with previous biochronological analyses (9, 28, 29, 49–51), but with a number of refinements and some notable differences (Fig. 4 and Table 2). Makapangat is likely the oldest site (∼2.7 to 2.6 Ma), but rather younger than previously suggested (see also ref. 52). Sterkfontein Members 2 and 4 are somewhat younger, but faunally similar, despite recent arguments that they are far older (see below). These are followed by a cluster of sites close in age to Swartkrans Member 1, ranging between ∼2.1 and 1.6 Ma (Fig. 2 and Table 2). The classic deposits surrounding the inferred hominin locality at Taung are still difficult to place but appear to be closest in age to Sterkfontein Members 2 and 4. A few sites, e.g., Swartkrans Member 3 and Hopefield, are younger. For a full assessment of all South African sites analyzed, see SI Appendix, SI Text. Below, we summarize our results for some of the best-known and best-constrained hominin sites.

Fig. 4.

Consensus estimates for the ages of select South African fossil sites indicated by vertical bars, black for sites in the Cradle, blue for others. Vertical axis indicates geological age (Ma), with the geomagnetic polarity timescale (after ref. 82) and shading indicating flowstone forming (FGI) and sedimentary (SED) intervals (after ref. 2) for reference. Red stars mark mean age estimates based on T. oswaldi lineage molar sizes. Yellow dots mark U/Pb estimates. Black and white dots mark paleomagnetic samples. For further details on data behind estimates see SI Appendix, SI Text.

Older sites (∼2.8 to 2.0 Ma).

Makapansgat (here including fossils from Members 2 to 4) contains a large sample (n = 30 specimens with molars) of T. o. darti specimens, which, combined with other faunal evidence, places Makapansgat in Cercopithecid zone 4 and provides a well-constrained estimate (∼2.7 to 2.6 Ma) close to the Gauss–Matuyama boundary. Several lines of evidence support this assessment. First, Members 2 to 4 record normal polarity throughout the sequence (52), and the T. o. darti mean estimate (∼2.4 Ma) is closest to the young end of the 3.03 to 2.60 Ma normal polarity chron. Furthermore, previous analyses of fossil Suidae provide a best-fit correlation to unit C-4 of the Shungura Formation, ∼2.74 Ma, which again suggests an age close to the Gauss–Matuyama boundary at 2.60 Ma (27, 51, 53). Finally, the presence of Cercopithecoides williamsi, which is known from eastern African sites ranging from ∼2.3 to 1.56 Ma (but not older), also suggests an age toward the younger end of the 3.03 to 2.60 Ma normal Upper Gauss Subchron and close to the transition from Cercopithecid Lower Zone 4 to Upper Zone 4, is most likely.

Based on correlations with Makapansgat combined with reversed paleomagnetic samples and multiple U-Pb dates, Sterkfontein Members 4 and 2 (including the Silberberg Grotto and Stw 573) are best estimated between ∼2.6 and 2.0 Ma. The combination of Parapapio jonesi, Parapapio broomi, Parapapio whitei, and C. williamsi found in Member 4 is otherwise only known at Makapansgat. All of the taxa known from Member 2 (Pp. jonesi, Pp. broomi, C. williamsi, and ?Papio izodi) are also found in Member 4 (see also ref. 54). Other fauna within Member 2 support a similar correlation with Sterkfontein Member 4 as well; for example, all extinct felid and hyaenid taxa recognized at Member 2 are found in Member 4 (55, 56). In fact, the Member 2 hominin Stw 573 itself also provides a correlation to Sterkfontein Member 4 and Makapansgat regardless of preferred classification (i.e., Australopithecus prometheus and/or Australopithecus africanus are both recognized at Sterkfontein Member 4 and Makapansgat) (57). Given the faunal correlation with Makapansgat ∼2.7 to 2.6 Ma, the closest reversed paleomagnetic interval is ∼2.60 to 2.0 Ma. This age range is consistent with U-Pb dates ranging from ∼2.8 to 2.7 Ma at the base of Members 2 and 4 to ∼2.3 to 2.1 Ma in the Silberberg Grotto to ∼2.0 Ma at the top of Member 4 (2). Thus, we find no support for the proposed ∼3.67 Ma date for the Stw 573 skeleton in the Silberberg Grotto found in association with Pp. broomi (58), a monkey otherwise only known from Member 4, Makapansgat, Bolt’s Farm Pit 23, and possibly Uraha, Malawi (43). Our results are also inconsistent with recent Al/Be estimates near 3.5 Ma for Member 4 (59). The presence of true Papio (P. sp. indet.) documented by specimens with secure provenience in Member 4 (e.g., SWP 31, SWP 35) demonstrates at least some of the fauna is ∼2.3 Ma or younger (34). This also suggests that other fauna indicative of an age <3.0 Ma (e.g., Equus, Metridiochoerus, and Antidorcus) cannot be dismissed as intrusive from Member 5 in the absence of more specific evidence. Thus, any date over ∼3.0 Ma for either Member 2 or 4 is biochronologically unlikely (see SI Appendix for further details).

Taung is faunally most similar to Sterkfontein Members 2 and 4, sharing two out of three cercopithecid taxa (Cercopithecoides cf williamsi and ?P. izodi) and a hominin (A. africanus) as well. The geological situation at Taung is complicated (SI Appendix), but regardless of the debate over the major fossiliferous units (60–62), the paleomagnetic data at Taung record a transition from normal to reversed polarity. Given the correlation with Sterkfontein Members 2 and 4, this normal-reversed sequence could represent the Gauss–Matuyama boundary (60, 62), suggesting that the Taung deposits cover a span of time before and after this reversal (e.g., ∼2.8 to 2.4 Ma). Alternatively, the reversal could represent the Reunion (∼2.2 Ma), Feni (∼2.14 to 2.12 Ma), or Huckleberry Ridge events (∼2.08 Ma), resulting in an overall broad estimate between ∼2.8 and 2.0 Ma. There are no monkey taxa present suggestive of an age equal to or greater than ∼3.0 Ma or younger than 2 Ma (SI Appendix).

Younger sites (∼2.1 Ma and younger).

Among the younger South African hominin sites, Swartkrans (SK) Member 1 is well-constrained to ∼2.1 to 1.7 Ma by a sizeable T. o. oswaldi sample (n = 18 with molars) combined with other data. The mean T. o. oswaldi estimate is ∼1.9 Ma with a 95% CI between ∼2.0 and 1.8 Ma, and this age range is consistent with the overlapping age ranges of Gorgopithecus major (∼1.88 to 1.84 Ma), C. williamsi (∼2.3 to 1.56 Ma), and Cercopithecoides coronatus (∼2.04 to 1.56) in Cercopithecid Zone 5 at eastern African sites. U-Pb dates obtained from basal and capping flowstones at SK Member 1 suggest a time range between ∼2.2 and 1.7 Ma (2), independently corroborating our biochronological estimate.

Drimolen Main Quarry (DMQ) contains two cercopithecid taxa (Papio robinsoni and C. coronatus). Both species are also found together in Swartkrans Member 1, suggesting a similar age between ∼2.1 and 1.7 Ma. An independent age estimate for DMQ between ∼2.04 and 1.95 Ma was recently obtained through a combination of U-Pb and paleomagnetic data (63), supporting our biochronological age range. Thus, the consistency of the cercopithecid fauna and the other dating methods point to a likely age range between ∼2.1 and 1.9 Ma.

Cooper’s D contains a relatively large sample of T. o. oswaldi (n = 13 with usable molars) and is thus constrained between approximately 2.0 and 1.6 Ma (95% CI) with a mean estimate of 1.8 Ma. The other monkeys present at Cooper’s D are not readily identifiable, but most consistent with G. major and Papio sp., suggesting a combination of taxa otherwise known only from Swartkrans Member 1 and consistent with the estimate from T. oswaldi molars. In addition, postcrania allocated to T. oswaldi from Cooper’s D (64) are most similar in size to T. o. oswaldi postcrania from Olduvai Bed I (∼1.9 to 1.8 Ma) and notably smaller than those from Middle and Upper Bed II or Konso Intervals 4 to 5 (∼1.6 to 1.3) (65–67). Other fauna at Cooper’s D better fit the T. o. oswaldi estimate as well. In particular, the distinctive felid Dinofelis cf aronoki is most similar to other specimens belonging to this species at Drimolen Main Quarry and the Upper Burgi Member at Koobi Fora (68), two well-constrained sites between ∼2.1 and 1.9 and ∼2.0 and 1.87 Ma, respectively. Thus, in contrast to previous age estimates younger than 1.4 Ma based on U-Pb dates from a flowstone at Cooper’s D, the fauna strongly suggests a range between ∼2.0 and 1.6 Ma.

Numerous other South African sites share the same cercopithecid taxa as in SK Member 1, DMQ, and Cooper’s D in some combination, and are therefore also likely between ∼2.1 and 1.6 Ma in age (Tables 1–2 and SI Appendix, SI Text). These localities include the hominin sites of SK Member 2, Sterkfontein Member 5 Stw 53 Infill, and Kromdraai B.

SK Member 3 contains a relatively small sample of T. o. oswaldi (n = 4), along with two other large papionin species consistent in size with 1) P. robinsoni and 2) Gorgopithecus or Dinopithecus. All of these taxa are also found in Swartkrans Member 1 and/or SK Member 2 (∼2.1 to 1.7). The mean age estimate of SK Member 3 based on the T. o. oswaldi sample is ∼1.5 Ma, with a 95% CI between ∼1.8 and 1.2 Ma. Thus, the T. o. oswaldi molars from Swartkrans Member 3 are larger than those from Swartkrans Member 1, but clearly smaller than T. o. leakeyi specimens younger than 1.4 Ma in eastern Africa and the Theropithecus oswaldi leakeyi sample from Hopefield (Elandsfontein), South Africa (see below). Therefore, a likely age for Swartkrans Member 3 is between ∼1.8 Ma and 1.4 Ma. This estimate contrasts with U-Pb dates of less than 1.0 Ma based on bovid tooth enamel (4), but is consistent with previous faunal estimates (28, 29, 69).

Hopefield (Elandsfontein) contains a sample of T. oswaldi, in this case the very large chronosubspecies T. o. leakeyi (n = 6 with molars). The molar dimensions of T. o. leakeyi at Hopefield result in a mean age estimate of ∼900 Ka, with a 95% confidence interval between ∼1.2 Ma and 590 Ka. This is consistent with previous estimates based on the other fauna, between ∼1.4 and 0.6 Ma or ∼1.0 and 0.6 Ma (70, 71).

Discussion

In summary, data from multiple sources, including our present analyses, help to refine and provide confidence in an emerging consensus around the likely ages of numerous South African hominin sites such as Makapansgat (∼2.7 to 2.6 Ma), Sterkfontein Member 4 (∼2.6 to 2.0 Ma), Drimolen Main Quarry (∼2.1 to 1.9 Ma), and Swartkrans Member 1 (∼2.1 to 1.7 Ma) (Fig. 4 and Table 2). We find no cercopithecid faunal evidence to support an age >3.0 Ma at any South African hominin site, in contrast to some previous suggestions (3, 57–59). The cercopithecid taxa known from Cercopithecid Zone 3 in eastern Africa (∼3.8 to 3.0 Ma) include Parapapio ado, Cercopithecoides meaveae, Paracolobus chemeroni, and a much smaller/more primitive form of T. oswaldi (T. o. cf darti), none of which is found at any South African hominin locality (Figs. 1 and 2, Table 1, and SI Appendix, Table S2). Instead, taxa such as C. williamsi and a larger/more derived form of T. oswaldi (T. o. darti), otherwise only known from Cercopithecid Zone 4, are found at the oldest South African hominin localities (Fig. 1 and Table 1 and SI Appendix, Table S2). Given that there are numerous cercopithecid taxa shared across eastern and southern Africa in Cercopithecid Zone 4 and younger (e.g., T. oswaldi ssp., Papio sp., Pp. jonesi, G. major, Soromandrillus quadratirostris, C. williamsi, C. coronatus), it is most likely that South African hominin sites do not sample earlier cercopithecid zones. This inference is consistent with most other geochronological data, including U-Pb on speleothems, paleomagnetic correlation, and other fauna (2, 52), but contrasts with recent Al/Be analyses (3, 59). Interestingly, our age estimates for nearly all of the South African sites are also broadly consistent with the pattern of sedimentation recently hypothesized at a number of South African hominin sites (2); in general, our estimates overlap the proposed sedimentation zones, and a number of the T. oswaldi mean estimates fall within hypothesized sedimentation zones as well (Fig. 4 and Table 2).

T. oswaldi is widely recognized by cercopithecid experts as an evolving lineage through geological time across eastern and southern Africa regardless of whether populations such as T. o. darti are ranked as species or subspecies (31, 37, 41, 65, 72–80). Here, we adopt the chronosubspecies approach (e.g., refs. 37, 74–76 and see Fig. 1) and document that among the clear anatomical trends through time present in T. oswaldi is a steady and unchanging increase in molar size beginning at least 3.6 Ma (Fig. 3). The documented utility of T. oswaldi molar dimensions in estimating geological age provides an important and highly accurate tool in African Plio-Pleistocene biochronology moving forward. Where conflicting signals exist, we point to the historical accuracy of previous faunal estimates relative to still developing chronometric methods, as famously illustrated by the suid example and the KBS Tuff (see Introduction) and suggest that our current cercopithecid-based estimates should be given considerable weight in subsequent studies. Future geochronological work should be aimed at resolving conflicting age estimates from key localities such as Sterkfontein Members 2 and 4, Cooper’s D, and Swartkrans Member 3.

Materials and Methods

We began with a comprehensive taxonomic assessment of the Plio-Pleistocene African cercopithecid fauna, including species distributions across fossiliferous deposits (SI Appendix, Tables S3 and S4). To this end, we studied all available original fossil material housed at museums around the world (see Acknowledgments for a comprehensive list). Qualitative and quantitative taxonomic comparisons were made using standard qualitative/quantitative characters and metrics across a large extant and fossil comparative sample (13, 28, 30, 31, 33, 34, 36, 37, 44, 72, 73). Figs. 1 and 2 show the occurrences of biochonologically informative taxa from eastern and southern Africa, respectively, based on our assessment (see also SI Appendix, Tables S3 and S4).

To construct an overall South African Plio-Pleistocene biochronology, we identified age ranges for as many fossil cercopithecid taxa as possible based on their occurrence at well-constrained sites in eastern and southern Africa. We considered stratified eastern African sites with ⁴⁰Ar/³⁹Ar dates based on volcanic deposits and/or with unambiguous paleomagnetic correlations to be well constrained in terms of geological age (Fig. 1). We likewise considered South African sites to be well constrained if they either had well-supported radiometric dates based on U-Pb analyses of speleothems and/or paleomagnetic correlations with agreed upon stratigraphic relationships to fossiliferous deposits or include widely accepted biomarkers such as Equus. Based on these known occurrences, we constructed a revised list of African cercopithecid (AC) faunal zones, built on and modified from those proposed by Delson (28, 29). While Delson’s original AC Zone 1 included Parapapio ado, the earliest Theropithecus, and the early small colobine from Laetoli (cf Kuseracolobus), we have divided this into three AC zones: AC Zone 1 (>4.2 Ma) is defined here as pre-Theropithecus; AC Zone 2 (∼4.2 to 3.8 Ma), which includes the earliest Theropithecus and small non-Theropithecus papionins typically referred to Pp. ado or cf Parapapio ado; and AC Zone 3 (∼3.8 to 2.9 Ma), defined by fauna from sites such as Laetoli and Hadar, including the earliest members of the T. oswaldi lineage (T. o. cf darti), the earliest Soromandrillus, definitive Parapapio (Pp. ado and Pp. jonesi), the earliest Cercopithecoides (C. meaveae), Paracolobus, and possible Rhinocolobus. AC Zone 4 (∼2.9 to 2.0 Ma) is defined by the appearance of definitive T. o. darti and the transition into T. o. oswaldi ∼2.5 Ma, with this transition dividing the zone into lower and upper halves. The whole zone could be considered the Parapapio “acme” zone in South Africa (following Delson), and Soromandrillus is also found throughout this time period. The lower half is defined by T. o. darti and the upper half is characterized by T. o. oswaldi and the colobines Paracolobus mutiwa, Rhinocolobus turkanensis, and C. williamsi. AC Zone 5 (∼2.0 to 1.4 Ma) is defined by the continued presence of T. o. oswaldi, the definitive first appearance of the large colobine C. coronatus, the presence of Papio in the South African record, and the appearance of the large papionin G. major. AC Zone 6 is defined by the appearance of T. o. leakeyi at ∼1.4 Ma and the last appearance of Cercopithecoides as defined by Cercopithecoides alemayehui ∼1.0 Ma. AC Zone 7 is defined by the last occurrences of T. o. leakeyi combined with the appearance of modern Papio hamadryas ssp. in East Africa. AC Zone 8 marks the beginning of exclusively modern fauna, ∼0.5 Ma (Figs. 1 and 2). Correlation of well-known eastern and southern African Plio-Pleistocene sites to these faunal zones is provided in SI Appendix, Table S2.

We estimated geological ages for sites with samples of T. oswaldi, which is known to show a clear trend of increasing molar size with geological age (41, 65, 66, 78–80). Standard measurements for cercopithecid molars were collected for a sample of 1,263 molars of T. oswaldi (Dataset S1). The measurements included mesial breadth, distal breadth, and length (Fig. 3). From these three we also calculated molar areas by multiplying the length by the average of the two breadths to produce a fourth measure of molar size (66). For each of these four molar variables, we used separate regression models based on the T. oswaldi material from well-constrained sites using geological age as the independent variable to measure the effect of age on the change in molar size through time (Fig. 3 and SI Appendix, Table S1). Regression models for tooth positions were compared using ANCOVA in SAS 9.1.4 (SAS Institute) to test for homogeneity of slopes. Tooth positions not significantly different to parallel were pooled to their common slope (SI Appendix, Table S1). The intercepts of tooth positions with common slopes could then be compared and used in the predictive model. For each of the four molar variables, we calculated the variance at each age within each site and weighted them by sample size to produce a pooled variance (81). For sites not included in the model (i.e., those without well-corroborated independent age estimates) we used this model to predict geological ages for each tooth. We then calculated mean ages based on all teeth available from each site and 95% confidence ranges using the pooled variances for each measurement and the number of individual specimens at that site (81). The final estimates reported in Table 2 were generated using the model based on molar area as this number included data from the other measurements and this model generally had the highest percentage of variance explained by the model and narrowest 95% confidence ranges (Fig. 3). Results from the individual molar measurements are reported in SI Appendix, Table S1.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.