The caves in the Cradle of Humankind (CoH), South Africa have yielded rich fossil assemblages of late Pliocene to early Pleistocene age, which include a range of hominin species (A. africanus, A. prometheus, A. sediba, P. robustus, H. ergaster, H. naledi and early Homo) and associated mammals, reptiles, and birds (e.g., Vrba, 1975, 1995; Brain, 1993; Tobias, 2000; Berger et al., 2010, 2015). For the past 3 million years, hominin-bearing deposits in caves formed in broadly similar settings, involving debris cone accumulations near cave openings (Partridge, 1973; Wilkinson, 1985; Brain, 1993; Pickering et al., 2007; de Ruiter et al., 2009; Dirks and Berger, 2013; Herries and Adams, 2013; Dirks et al., 2010, 2016b; Bruxelles et al., 2014; Stratford et al., 2014), with deposits cemented by carbonate-rich waters dripping from cave ceilings (e.g., Wilkinson, 1985; Pickering et al., 2011b). In contrast to all other hominin deposits in the CoH, the deposits that host H. naledi in Rising Star Cave are composed of largely unconsolidated, mud-clast breccia in a mud matrix with no evidence of coarse clastic sediment being carried in by water flow. This suggests a different depositional regime and timing for the sediments and the fossils (Dirks et al., 2015, Dirks et al., 2016a).

Rising Star Cave is situated in the Bloubank River valley, 2.2 km W of Sterkfontein Cave. The cave system comprises several kilometres of mapped passageways (Figure 1a) that are stratigraphically bound to a 20–30 m-thick, chert-poor dolomite horizon capped by a 1–1.3 m-thick chert unit that forms the roof to the cave system (Dirks et al., 2015). Geological mapping and laser-theodolite surveys indicate that this roof is intact and not penetrated by significant shafts that open to surface (Dirks et al., 2015; Kruger et al., 2016). The broader geological setting of the cave is discussed in Dirks et al. (2015), (Dirks et al., 2016a).

Figure 1

Download asset Open asset

The Dinaledi Chamber, which contains most of the fossils of H. naledi, is ~30 m below surface and ~80 m in a straight line from the nearest present-day opening to the surface (Figure 1a). The main cavity forming the Dinaledi Chamber is ~15 m long with variable widths not exceeding 2.5 meters (Figure 1b), and expands near the intersection with a crosscutting passage, which is the location of the main excavation site to date (Figure 1b). There is no evidence that the present entrance into the Dinaledi Chamber has significantly changed since the deposition of the fossil hominins, with sediment accumulating mostly near the current access point (Dirks et al., 2015, Dirks et al., 2016a; Figure 2). Samples for dating were collected from the various flowstone horizons and stratigraphic units exposed in the Dinaledi Chamber (Figures 1b, 2, 3, 4 and 5) as well as from fossil material itself (Figures 4, 6 and 7).

Figure 2

Download asset Open asset

The Dinaledi Chamber contains deposits of fine-grained, muddy sediments intercalated with flowstone drapes. The sediments include various types of orange, laminated mudstone and mud clast breccia distributed across three broad lithostratigraphic units (Units 1, 2 and 3; Dirks et al., 2015) that filled parts of the chamber over time. Based on variations in sediment composition, fossil content and/or stratigraphic position of each unit, we have divided Unit 1 into sub-units 1a, 1b and 1c, and Unit 3 into sub-units 3a and 3b, to more precisely define the stratigraphic packages targeted for dating (Figure 2). The units are separated by erosional unconformities or flowstone intercalations, but do not all necessarily occur in direct contact with one another due to the complex nature of caves as depositional systems (e.g., Brain, 1993; Martini et al., 2003). In addition, apart from sediments accumulating along the floor of the cave chamber, sediment in the form of orange mud deposits also accumulated inside fractures and along ledges higher up in the Dinaledi Chamber (Figure 2b), where it formed as a result of the combined effect of in situ weathering and deposition from water flowing down fractures and side walls. All units and sub-units are time-transgressive, meaning that they are lithostratigraphic units and not chronostratigraphic units that occur in strict temporal order. Periods of sedimentation alternated with periods of erosion, during which sediments were either redeposited or removed from the chamber via floor drains, resulting in erosional remnants of all units occurring in a variety of stratigraphic positions (Dirks et al., 2015).

Stalactites have formed at drip points along the roof and associated stalagmites formed below these points. In one area below the entrance to the chamber, these drip points repeatedly formed flowstone aprons over cave sediments that dip towards the deeper part of the chamber. Flowstone also formed as cascades and curtains that developed where water seeped down fractures and ran along the walls to locally spread out, horizontally, across the sediments comprising the cave floor (Dirks et al., 2015). The flowstones have preliminarily been sub-divided into three groups demarcating semi-contemporaneous generations of formation, which we named Flowstone Groups 1, 2 and 3 based on their appearance and relationships with each other, and with the floor sediments and other litho-stratigraphic units in the chamber. In making this subdivision it was realised that each group of flowstones will probably comprise a range of ages representing separate flowstone forming events (Dirks et al., 2015), a fact borne out by the ages presented below (Table 1).

Flowstone Group 1 (FS1 in Table 1; Figures 1b, 2 and 3) includes remnants of what are interpreted to be generally older flowstone units that were partly dissolved and resorbed to leave behind rims or aprons along the side walls of the cave chamber, some with sediment attached below them. Flowstone remnants interpreted as Flowstone Group 1 are mostly restricted to five staggered remnants (Flowstones 1a-e), one above the other in reverse stratigraphic order (oldest on top, youngest at the bottom), near the entry shaft into the Dinaledi Chamber (Figure 2b). Flowstone Group 2, the most extensive group of flowstones in the chamber (FS2 in Table 1, and Figures 1b, 2 and 3), comprises wall aprons and sheets that have spread out across the floor of the Dinaledi Chamber together with drip pools, cascades, curtains, stalactites and stalagmites that connect to these sheets, and, therefore, formed in conjunction with them. Flowstone Group 3 (FS3 in Table 1 and Figures 1b, 2 and 3) comprises the flowstone deposits that are actively forming below existing drip points, and include fresh growth of delicate crystals of aragonite and calcite in floor sediments and along cave walls.

Sedimentary deposits within the Dinaledi Chamber can be organized into three primary stratigraphic units (Dirks et al., 2015). Unit 1 consists of deposits of non-lithified, laminated, orange mud interpreted as suspension deposits in standing water (Facies 1a of Dirks et al., 2015), and laminated mud with fine sand containing small-scale ripple cross laminations and rodent remains (Facies 1b of Dirks et al., 2015), reflecting deposition by shallow, flowing water along the cave floor, with additional sandy material accumulating near local entry points, where fractures higher in the chamber act as sediment conduits (Dirks et al., 2015, Dirks et al., 2016a).

Within the Dinaledi Chamber Unit 1 deposits can be divided into three sub-units provisionally called sub-units 1a, 1b and 1c. It is assumed that Unit 1 is time-transgressive and future work may reveal additional sub-units. Sub-unit 1a is composed of laminated orange mudstone with isolated lenses of sandy material, occurs as erosion remnants along the cave floor, and is possibly more extensive beneath younger deposits in the chamber. Sub-unit 1b is dominated by sandy orange mud deposits that are rich in micro-faunal remain, stratigraphically overlies deposits of sub-unit 1a (Figure 2c and d), and may have formed through the partial erosion and re-deposition of sub-unit 1a. Deposits of sub-unit 1c are similar in appearance and composition to the laminated, muddy sediments of sub-unit 1a, but they occur along chert ledges, solution pockets and fractures in the chamber walls and along the entry shaft, higher up in the cave chamber (Figure 2). The orange mud is mostly the product of the cave formation process, representing the insoluble residue left over when cavities develop via dissolution of dolomite (Dirks et al., 2015). Some of the mud-bearing waters seeping out of the fractures would have flowed as water films along the cave walls to deposit mud on ledges and in fractures to form sub-unit 1c, whilst elsewhere this water would have dripped to the floor to contribute to the deposition of sub-unit 1a and 1b.

Unit 2 is composed of largely lithified mud clast breccia consisting of angular to sub-angular clasts of laminated orange mudstone (similar to that found in Unit 1), embedded in a brown mud matrix (Facies 2 of Dirks et al., 2015). The mud clasts are interpreted to be derived locally due to wetting and drying of orange mud deposits, which led to auto-brecciation, and subsequent erosion and re-deposition of angular mud clasts (Dirks et al., 2015). We hypothesize that the mud clasts forming Unit 2 are partly derived from erosion of deposits of sub-unit 1c, and partly from a yet unidentified unit that was likely deposited in fractures within and above the chamber entry zone. Two macro-fossils (partial shafts of long bones) that are non-specific, but not hominin, have been found in Unit 2.

Unit 2 sediments are only exposed as hanging remnants attached below the remains of a composite flowstone sheet (Flowstone 1a) near the entrance shaft into the chamber (Figure 2b; Dirks et al., 2015). Note that in Dirks et al. (2015) Unit 2 was originally defined to also include sediments below Flowstones 1b-e; however, based on our new dating results, the revised definition of Unit 2 has been narrowed to only include the more indurated and distinctly darker coloured erosional remnants of mud clast breccia under Flowstone 1a, which are notable for their absence of hominin fossils. Unit 2 sediments accumulated as a sloping debris cone of mud clast breccia below a vertical fracture system before being covered by flowstone (Flowstone 1a). The debris cone of mud clast breccia was subsequently eroded leaving behind hanging erosion remnants of Unit 2 below a flowstone apron (Figures 2b and 3l). The processes that caused erosion of the Unit 2 debris cone led to the deposition of Unit 3 sediment along the floor of the Dinaledi Chamber as shown in Figure 8.

Unit 3 is composed of largely unlithified, clast-supported, mud clast breccia (Facies 2 of Dirks et al., 2015), dominated by reworked angular to sub-angular mud clasts, which are interpreted as being locally derived from the reworking of Units 1 and 2. Unit 3 accumulated below the hanging remnants of the Unit 2 debris cone near the entry shaft, and also extends along the current, sloping cave floor to the SW end of the chamber (Figures 2c and 8). Unit 3 sediments are dynamic in the sense that they are poorly lithified in most places and actively slump towards, and erode into, floor drains that occur in parts of the chamber where sediment is being washed down to deeper levels in the cave (likely as a result of fluctuations in the ground water level). Remains of Unit 3 sediment are attached to apron-like erosional remnants of Flowstones 1b-e near the entrance shaft (Figure 2a and 3l). Erosional remnants of Unit 3 under Flowstone 1c contain in situ long bones consistent with H. naledi, which are actively eroding out and accumulating along the present cave floor. Note that Dirks et al. (2015) originally included these erosional remnants as part of Unit 2. Everywhere else, Unit 3 deposits are spread across the cave floor as loosely packed, semi-moist, orange mud clasts of varying sizes in which bone material of H. naledi is distributed. Unit 3 is partly covered by sheets of Flowstone Groups 1, 2 and 3.

Unit 3 has been divided into a lower and an upper sub-unit, termed sub-unit 3a and 3b (Figure 2), based on the respective absence or presence of hominin fossils. Sediments belonging to sub-unit 3a are not directly exposed in the chamber, but their presence has been confirmed in the deepest part of the excavation area (Figure 2d). In contrast sub-unit 3b is exposed within the talus cone near the entry shaft and along the cave floor, and contains all of the known H. naledi fossils in the chamber (Figure 2c and d). The thickness of sub-unit 3b is thought to be no more than 20–30 cm (see below).

Most fossil deposits in the Cradle of Humankind that have been dated are between 0.5 and 3.7 Ma old and consist of bone material encased in well-cemented hard clastic rocks commonly referred to as cave breccia (e.g., Wilkinson, 1985; O'Regan and Reynolds, 2009; Herries et al., 2009; Pickering et al., 2011b; Granger et al., 2015). In the absence of volcanic deposits, it is generally difficult to obtain accurate ages for the fossils, not just because reliable techniques are few, but mostly because the stratigraphic sequences in the caves are complex, discontinuous and frequently reworked (e.g., Brain, 1993; Pickering et al., 2011a; Bruxelles et al., 2014; Stratford et al., 2014). Workers have relied on a combination of biochronology of faunal remains, palaeomagnetic work and a range of radiometric methods, including U-Pb, U-Th and ESR dating targeting flowstones and fossil teeth (e.g., Vrba, 1975; Partridge et al., 1999; Berger et al., 2002; Walker et al., 2006; Herries et al., 2006, 2013, 2014; Herries and Shaw, 2011; Dirks et al., 2010; Pickering and Kramers, 2010; Pickering et al., 2011a; Herries and Adams, 2013), as well as limited cosmogenic (¹⁰Be, ¹⁶Al) dating (e.g., Partridge et al., 2003; Granger et al., 2015; Dirks et al., 2016b). Whilst some of these techniques are well established, others such as the application of cosmogenic isochrons (e.g., Granger et al., 2015) are relatively new and not without significant analytical (and interpretative) challenges (Kramers and Dirks, 2017), and all efforts are strongly dependent on the stratigraphic interpretation of the fossils or units that are being dated.

Unlike other fossil deposits in the Cradle of Humankind, the remains in the Dinaledi Chamber are largely restricted to hominins. This makes it impossible to use biochronology as a preliminary technique to assess the age of the fossils. In addition, the fossils are contained in mostly unconsolidated muddy sediment with clear evidence of a mixed taphonomic signature indicative of repeated cycles of reworking and more than one episode of primary deposition (Dirks et al., 2015). This indicates that caution is required when interpreting the stratigraphy and the age of the fossils they contain.

In preparation for this study, trial dating of the deposits in the Dinaledi Chamber was undertaken to obtain an indication of the age of the deposit and the best techniques to apply. Preliminary work was focussed on assessing the viability of U-series techniques for flowstone dating, using ¹⁴C for dating bone fragments, and using OSL to test samples of quartz-bearing Unit 1 (Dirks et al., 2015). Initial tests were carried out at the University of Johannesburg (UJ) to assess suitability for U-Pb dating, which allows for the dating of older (>500 ka) flowstone material (e.g., Walker et al., 2006; Pickering et al., 2010; Pickering and Kramers, 2010), on the assumption that the H. naledi material could be older than 1 Ma based on its primitive morphology (Berger et al., 2015; Dembo et al., 2016; Hawks and Berger, 2016; Thackeray, 2016; Hawks et al., 2017). It was found that the older flowstones in the Dinaledi Chamber contained excessive common Pb caused by the inclusion of detrital material (mainly clays) making them unsuitable for U-Pb dating (Dirks et al., 2015). In contrast, preliminary tests with U-Th disequilibrium dating at James Cook University (JCU) returned promising results. U-Th dating is more precise in the <500 ka range than U-Pb dating, and is much less critically affected by detrital material. The initial tests with U-Th disequilibrium dating revealed that the fossils may be much younger than originally anticipated (e.g., Dembo et al., 2016; Thackeray, 2016), and mostly well within the range of the U-Th technique. Therefore, U-Pb dating was not pursued further.

Preliminary tests with OSL were conducted at the University of the Witwatersrand (Wits) on samples from Unit 1, which were assumed to be older than the fossils of H. naledi. These preliminary studies, and the results contained in this paper, are the first OSL results for cave sediments from the CoH, and again indicated that the H. naledi fossils were probably relatively young (i.e.,<500 ka).

Tests with radiocarbon (¹⁴C) dating were undertaken through a commercial facility (Beta Analytic Inc. in Florida), to ensure a fast turn-around time for results. At the time these dating tests were done, it was already known from U-Th and OSL tests that the H. naledi fossils would be too old to be dated by ¹⁴C. Nevertheless, analyses were carried out as part of the due diligence process, and the results of these tests are presented here. Following this initial work, no further radiocarbon studies were carried out.

The preliminary results have guided the subsequent dating strategy and sampling approach reported here. The dating strategy was designed to achieve three objectives: (i) establish a detailed stratigraphy for the cave sediments in the Dinaledi Chamber; (ii) date sedimentary units that potentially bracket the fossil-bearing deposits; and (iii) date the fossils directly.

To obtain an upper age limit for the fossil-bearing deposits of Unit 3 (i.e. sub-unit 3b), we conducted U-Th dating of flowstones that directly overlie Unit 3. A large number of such flowstones were sampled with the aim of finding the oldest flowstone directly overlying H. naledi fossils. To obtain a lower age limit for sub-unit 3b, erosional remnants of Unit 1 sediments that were at least partially covered by fossil-bearing sub-unit 3b sediments, were sampled for OSL dating on the assumption that sub-units 1a and 1b in these areas are older than sub-unit 3b (Dirks et al., 2015). This was done in the full knowledge that OSL dating of cave sediments is complex and difficult to interpret (e.g., Roberts et al., 2009), and probably imprecise. As an internal control, we also sampled flowstones that cover the outcrops of sub-units 1a and 1b from which OSL samples were taken. These flowstones were dated with U-Th with the expectation that they are younger than the underlying Unit 1 sediments. In addition to OSL, Flowstone 1a, which overlies Unit 2 sediments, was sampled for palaeomagnetic analyses. This flowstone was targeted, because it was expected to be the oldest flowstone in the chamber and possibly older than 780 ka, and hence could potentially record reverse magnetic polarity (e.g., Singer, 2014). In this case, this would constrain the minimum age of Unit 2.

The best age estimates for H. naledi can be obtained by directly dating fossil material. It was clear from preliminary tests that this could not be achieved with ¹⁴C, and instead combined ESR and U-Th disequilibrium dating techniques (US-ESR; Grün et al., 1988) were applied to three H. naledi teeth that were freshly collected from near the site of the original excavation (Figures 1, 2, 4 and 6), as well as a single baboon tooth (cf. Papio) that had been recovered from sub-unit 3a below the hominin-bearing horizon (Figures 2d and 7).

Once results were obtained for ESR and U-Th dating, it became apparent that OSL dating would only provide general age constraints that confirmed the ESR results, but in their own right did not return additional age constraints for the fossils. OSL results were also difficult to interpret in the complex cave environment that was strongly affected by Rn loss (see Discussion). It was, therefore, decided not to pursue more detailed OSL studies at this stage, even though we did carry out preliminary tests for single grain and feldspar analyses at the University of Wollongong, to assess the suitability of these techniques. Pilot results are encouraging, and suggest that future, detailed OSL studies are worth pursuing.

U-Th dating of 17 flowstone samples (Figure 3) has yielded minimum depositional age estimates for the sedimentary units they overlie, and has provided insights into the timing of flowstone formation events (Tables 1, 2 and 3). Three separate checks were built into the U-Th dating strategy to ensure robust results would be obtained. Independent dates for the same samples were obtained by laboratories at JCU and at the University of Melbourne (UoM), with results displaying a high degree of concordance. In instances where samples were obtained from the same flowstone layer, but at different stratigraphic levels (e.g., sample pairs RS13 and RS18, RS22 and RS23, RS16 and RS17, and RS11 and RS21) all ages are consistent with stratigraphic order, and blind duplicates of the same sample (RS1 and RS15) returned identical results within error, indicating that results are both accurate and precise. Double blind U-Th results from JCU and UoM are shown in Tables 2 and 3, respectively. The distribution of flowstone ages across the Dinaledi Chamber is shown in Figures 1b and 2.

Figure 3

Download asset Open asset

The oldest dated flowstone in the assemblage is Flowstone 1a overlying Unit 2, which yields age estimates of 502 +181/–53 ka (RS 23) and 478 +107/–41 ka (RS 22) (Table 1). The next oldest age comes from a flowstone interpreted as Flowstone Group 1 overlying sediment of sub-unit 1b to the W of the excavation pit (Figure 1b) with an age of 290 ± 6 ka (RS5; Table 1). This age is younger than the OSL age derived from sub-unit 1a (OSL5), but is slightly older than OSL ages derived from sub-unit 1b in this location (OSL4; Figure 2c). This suggests that sub-unit 1a was deposited prior to precipitation of this flowstone and that sub-unit 1b formed out of stratigraphic order due to erosion and redeposition of the top of sub-unit 1a beneath this flowstone. However, the U-Th date for RS5 should be interpreted with caution as the flowstone has a porous texture (Figure 3a), which probably indicates some degree of dissolution and/or recrystallization of the primary calcite, and may have affected the U-Th systematics (see Discussion).

Flowstone samples that overlie sub-unit 3b, which contains the H. naledi fossils, yield age estimates that fall within four distinct time periods: ~242 ka (RS18 = 242 ± 5 ka [JCU] and 242.9 ± 6.6 ka [UoM]) for Flowstone 1c; 88–106 ka (RS1, RS8, RS13-17); 50 ka (RS6); and 24–32 ka (RS10, RS19-21) for Flowstone Group 2 deposits. The results for RS18 provide a minimum age for the H. naledi fossils in this part of the cave.

An actively forming (i.e., dripping) stalagmite of Flowstone Group 3 (RS11) on top of an older base of Flowstone Group 2 returned a younger age of 9–10 ka (RS11 = 9.05 ± 0.06 [JCU] and 9.95 ± 0.06 ka [UoM]). The final flowstone sample that was dated in this study was collected on the land surface above the cave system. It was sampled from a ~14 cm wide vertical flowstone-filled fracture exposed on the surface above the southern end of the Dinaledi Chamber itself. This is the only possible alternative entry-way into the Dinaledi Chamber that we have observed. The replicate samples (RS9) analysed at JCU and UoM both indicated secular equilibrium, which confirms that the flowstone sealed this fracture sometime before ~600 ka, eliminating this thin flowstone-filled fracture as a possible alternative entrance for H. naledi into the Dinaledi Chamber. It should also be noted that no evidence of a talus cone or any other evidence of sediment entry into the chamber below this point has been observed.

The spatial distribution of the flowstones belonging to the different age groups (Figure 1b) indicates that the oldest flowstones (Flowstone Group 1) occur near the entry zone into the chamber and as an erosional remnant (RS 5) near the back of a WNW-trending fracture W of the excavation pit. The 88–106 ka flowstones formed in three separate parts of the chamber (Figure 1b): (a) on top of older flowstones near the entry; (b) as wall drapes above a drain 6 m SW of the entry; and (c) deep within a N-trending fracture, 8 m N of the excavation pit. The ~ 50 ka flowstones originate from a W-trending fracture, 6 m W of the entry shaft where it overlies sub-unit 1a and sub-unit 3b sediments. The 24–32 ka flowstones all originate from around the area where the excavation pit is located at the intersection point of three fracture sets (Figure 1b). The youngest flowstone sample comes from below an active drip point, 1.5 m E of the excavation pit, and similar actively forming flowstones can be seen in other parts of the chamber. The flowstone age groupings indicate that episodic wet periods in the Dinaledi Chamber alternated with periods during which no flowstone was deposited.

U-Th disequilibrium analyses of four tooth samples were conducted to constrain U uptake models into dental tissues used in ESR dating. The analyses were also used to provide apparent U-Th age estimates (Tables 4 and 5). Analyses of all four teeth (samples 1767, 1788 and1810 from H. naledi, and sample 1841 from Papio sp.) were performed at the University of Wollongong (UoW), in collaboration with Southern Cross University (SCU). Duplicate analyses of two of the tooth samples (samples 1788 and 1810) were conducted at Griffith University (GU) in collaboration with the Australian National University (ANU).

Summaries of the U-Th analytical data and ages are reported in Table 4 (SCU-UoW) and Table 5 (GU-ANU). In Table 4 only closed system dates are reported, while Table 5 also lists dates based on the continuous diffusion model of Sambridge et al. (2012). In both datasets, the U content in enamel is much lower than in dentine. Note that apparent U-Th ages for the teeth are likely to provide apparent age estimates, which will approach the age for U uptake events that affected the teeth during wet periods in the chamber, typically after deposition. These ages should, therefore, be regarded as minimum age estimates for the teeth, and do not represent depositional ages for the fossils.

Figure 4

Download asset Open asset

Figure 5

Download asset Open asset

Figure 6

Download asset Open asset

Sample 1767: This extremely worn upper premolar crown (Figure 6a) is heavily weathered, and only a small fragment of enamel was left attached to the dentine. It could, therefore, only be dated once (at SCU-UoW). Both dentine and enamel yield consistent results with apparent U-Th ages of 44.5 ± 0.6 ka for dentine and 46.1 ± 0.7 ka for enamel, and initial ²³⁴U/²³⁸U activity ratios at 6.99 ± 0.01 and 6.99 ± 0.04 respectively. These results suggest that a single uptake event is dated. The tooth is characterised by an extremely high U content in the enamel when compared to the other teeth (Table 4). Uranium concentration gradients show the effects of diffusion into the enamel from all external surfaces, with enrichment at the Enamel Dentine Junction (EDJ).

Sample 1788: This lower right second molar was covered by a thin layer of sediment and is well-preserved (Figure 6b). Uranium concentrations in enamel and dentine vary across the surface, with minor hotspots and leaching zones near enamel cracks and along the dentine canal. The EDJ is enriched in U, showing a diffusion gradient into the enamel, and resulting in elevated U concentrations in spots located close to the EDJ (Figure 9). The U uptake history appears complex and heterogeneous, and probably involved several episodes. Most of the U concentrations in the enamel are too low to provide a meaningful age. However, parts of the enamel and the dentine yielded consistent measurements for which SCU-UoW provide ages with mean values of 71.5 ± 0.6 ka for dentine and 67.0 ± 5.4 ka for enamel with initial ²³⁴U/²³⁸U activity ratios of 7.62 ± 0.03 and 6.94 ± 0.17 respectively. GU-ANU obtained a combined age of 95.1 ± 3.2 ka for two enamel spots with higher U concentrations, with parts of the enamel with lower U-enrichment yielding a combined age of 56.2 ± 6.5 ka. GU-ANU also provide a consistent mean apparent age of 87.9 ± 2.6 ka, associated with initial ²³⁴U/²³⁸U activity ratios of 7.98 ± 0.06 (individual spots agreeing within error) for dentine which is interpreted as the age of an U uptake event.

Sample 1810: This lower left molar or premolar from the excavation pit (Figures 4 and 6c), is near complete and only moderately weathered. Uranium diffusion patterns show U accumulating at the EDJ with slow diffusion into the enamel tissue. The U concentrations in most of the enamel are too low to calculate a meaningful age. Areas of enamel with higher U concentrations return older ages (Tables 4 and 5). SCU-UoW provide a mean age for high-U spots in enamel of 128.8 ± 7.2 ka and an associated initial ²³⁴U/²³⁸U activity ratio of 7.60 ± 0.16. GU-ANU report a mean apparent U-Th age of 195.7 ± 13.8 ka, which is much higher than the adjacent dentine spots (at 81.1 ± 2.7 ka), but is coupled with realistic initial ²³⁴U/²³⁸U activity ratios of 7.24 and 7.84 with overlapping error limits. This indicates that a secondary overprint of the dentine took place for which the U source had a similar ²³⁴U/²³⁸U composition as the source of the initial U uptake event. Dentine measurements are consistent along the measured sections with small regions affected by leaching and enrichment near cracks and the pulp cavity. The dentine analyses done by SCU-UoW (Table 4) yield similar ages with consistent initial ²³⁴U/²³⁸U activity ratios of around 7.04 ± 0.03, and a mean apparent age of 72.3 ± 1.0 ka. The combined analytical data for dentine from GU-ANU in samples 1810A and 1810B yield apparent U-Th ages of 80.1 ± 2.5 ka and 82.1 ± 2.7 ka respectively, coupled with consistent initial ²³⁴U/²³⁸U activity ratios (Table 5).

Sample 1841 The baboon tooth consists of an enamel crown that is structurally intact, but the enamel is friable and weathered (Figure 7). The U distribution within the enamel appears homogenous, however, Th concentrations are low and the resolution of the elemental distribution is poor, which impairs the quality of the U-Th age estimates. A recent U uptake event may have occurred affecting enamel in contact with sediment, resulting in a mean apparent U-Th age estimate of 115.6 ± 1.9 ka with a mean initial ²³⁴U/²³⁸U activity ratio of 7.76 ± 0.05 (Table 4).

Figure 7

Download asset Open asset

GU-ANU also provide age estimates in which the continuous diffusion assumptions of Sambridge et al. (2012) have been applied. The results obtained for samples 1788 and 1810 are about 20% older than the closed system ages (Table 5), but show much less consistency and are not further considered.

Collectively, the results show that the teeth are older than 70 ka, and considering sample 1810 with a minimum age of around 200 ka, that the H. naledi fossils are probably older than 200 ka (see Discussion).

Combined U-series and ESR dating (US-ESR; Grün et al., 1988) of three hominin teeth (samples 1767, 1788 and 1810) and a baboon tooth (sample 1841) was performed at SCU. Blind duplicate analyses of two of the hominin samples (samples 1788 and 1810) were performed at the ‘Centro Nacional de Investigación sobre la Evolución Humana’ (CENIEH), Spain in collaboration with GU (CENIEH-GU). In obtaining the ages, each laboratory carried out independent sample preparation, and ESR and U-series analyses of the fossil teeth. Estimates for the environmental dose rates used in the age calculations were standardized for both laboratories (Table 6) in order to produce comparable results (see discussion and methodology sections for details).

Analyses and results from both labs are presented in Table 7 and Figures 10, 11 and 12. Results are presented for two scenarios: scenario 1 in which the teeth are fully covered in sediment that contains 25 ± 10% water and experienced 80% Rn loss; and scenario 2 in which the teeth are fully covered in sediment that contains 25 ± 10% water and experienced no Rn loss. Scenario 1 reflects the measured present-day situation and is interpreted as a maximum age estimate. Scenario 2 provides a minimum age estimate (Table 7). Together these scenarios provide our best estimate for the age range of the fossil teeth.

Combined US-ESR ages determined by SCU for samples 1810, 1788 and 1767 under scenario 1 conditions (i.e., the maximum age scenario) are 284 ± 51 ka, 247 ± 41 ka and 104 ± 29 ka (2σ uncertainty), respectively (Table 7). Combined US-ESR ages determined by CENIEH-GU for samples 1810 and 1788 under scenario 1 conditions are 267 ± 68 ka and 211 ± 28 ka (2σ uncertainty), respectively (Table 7).

Combined US-ESR ages determined by SCU for samples 1810, 1788 and 1767 under scenario 2 conditions (i.e., the minimum age scenario) are 230 ± 40 ka, 194 ± 34 ka and 87 ± 22 ka (2σ uncertainty), respectively (Table 7). Combined US-ESR ages determined by CENIEH-GU for samples 1810 and 1788 under this scenario are 210 ± 50 ka and 163 ± 24 ka (2σ uncertainty), respectively (Table 7).

Results for sample 1767 are based on anomalously high (~20 times) U concentrations in enamel, and probably yield anomalously low age estimates (Duval et al., 2012). This result is, therefore, considered to be unreliable and has been excluded from final age estimates (see Discussion).

The observed difference in age estimates obtained by SCU and CENIEH-GU for samples 1788 and 1810, are most likely explained by natural dose variations within the tested enamel layers (see Discussion and methodology sections), and we have no reason to prefer one age result over another. The optimal age estimate for the H. naledi fossils, therefore, combines the results from both laboratories with average maximum (i.e., scenario 1) age estimates for samples 1788 and 1810 of 229 + 60/–46 ka and 276 + 59/–77 ka (2 σ uncertainty) respectively, and average minimum (i.e., scenario 2) age estimates of 179 + 49/–40 ka and 220 + 50/–60 ka (2 σ uncertainty) respectively. Together these results provide an age range of 139–335 ka for the H. naledi remains, although dating of flowstone encasing H. naledi bones helps to better constrain this range (see Discussion).

Combined US-ESR ages for the baboon tooth (sample 1841) determined by SCU using scenario 1 and 2 conditions are 723 ± 181 ka and 635 ± 148 ka respectively. The tooth contained no inner dentine (Figure 7), and was filled with sediment. In calculating the age it was, therefore, assumed that sediment occurred on both sides of the enamel layer. Sample 1841 was recovered from sub-unit 3a directly below the occurrence of articulated H. naledi remains in the excavation pit (Dirks et al., 2015). The age results provide an upper age limit for the deposition of the H. naledi bearing layer, and mark an earlier stage of deposition of mud clast breccia in the cave assigned to sub-unit 3a, which predates the entry of H. naledi fossils into the cave.

Optically stimulated luminescence (OSL) dating of three samples of sediment from Unit 1 in the Dinaledi Chamber (samples OSL3 and OSL4 from sub-unit 1b, and sample OSL5 from sub-unit 1a; Figure 5) was performed at the University of the Witwatersrand (Wits). The measurements were carried out on small aliquots containing ~30 grains. Summaries of the OSL analytical data and ages are reported in Table 8.

The reported dose rates for the samples range from 0.7 to 0.9 Gy ka⁻¹ (Table 8), with significant within-sample scatter, resulting in uncertainties on age estimates of 15–18%. Overdispersion in D_e ranges from 50–70%, which is much higher than would be expected for a well-bleached sample, and indicates that it is most appropriate to apply a Minimum Age Model (MAM) to the dataset, in which the MAM age is likely to represent a maximum estimate for the age of the sediments (see Discussion). As with ESR, significant Rn loss was detected in the samples of Unit 1, and corrections to the measured U concentrations were applied (Table 8). The MAM calculations for the three samples yield maximum age estimates for the sediments of 231 ± 41 ka (OSL3), 241 ± 37 ka (OSL4), and 353 ± 61 ka (OSL5). The MAM apparent ages for OSL3 and OSL4 were obtained from the sandy facies of sub-unit 1b sediments and yield ages that are younger than the age of a Flowstone 1 sheet (sample RS5 at 290 ± 4 ka; Table 1) that covers the outcrop of sub-unit 1a from which sample OSL4 was taken (Figure 2c). This discrepancy can be attributed to inverted stratigraphy associated with erosion of the top of the older sub-unit 1a after the deposition of Flowstone 1 (RS5) by running water and subsequent deposition of sub-unit 1b between sub-unit 1a and the flowstone (Figure 2c). Sample OSL5 was obtained from muddy sediment of sub-unit 1a, and yields an older age than the covering flowstones (RS5 and RS20). Note that if a Central Age Model (CAM) is applied to the OSL data, results are significantly older at 560 ± 102 ka (OSL3), 546 ± 79 ka (OSL4), and 849 ± 132 ka (OSL5), however, this model is considered unrealistic within a cave environment (Galbraith et al., 1999).

Palaeomagnetic analysis of one composite sample of Flowstone 1a (Figure 13), covering erosional remnants of Unit 2 near the entry shaft into the Dinaledi Chamber, was performed at La trobe University, Melbourne (LTU). The palaeomagnetic results for Flowstone 1a are presented in Figure 13 and Table 9. The palaeomagnetic sample from Flowstone 1a comprises three distinct phases of flowstone formation (from base to top: A, B and C; Figure 13a,b) that have been analysed for their palaeomagnetic orientation (Figure 13c). Note that the palaeomagnetic sample of Flowstone 1a was taken up-dip from the position where the U-Th sample of Flowstone 1a was taken (i.e., RS22 and RS23). The basal, phase A flowstone observed in the palaeomagnetic sample tapers out down-dip (Figure 13a) and is not present in the U-Th sample; thus, RS23 at the base of the U-Th sample corresponds to phase B carbonate in the palaeomagnetic sample, and RS22 to phase C carbonate (Figure 2b).

Phase A flowstone at the base of the sample records a consistent reversed magnetic polarity, with a clear overprint that is anti-parallel to the characteristic, reversed polarity remanence (ChRM) and is removed by ~10 mT. The reversed polarity ChRM is then stable between 10 and 40 mT, although within this range there are two reversed polarity components that can be identified in some samples (between 10 and 19 mT and then between 20 and 40 mT; Figure 13c). Both components have almost identical declination values, but the lower field component has a shallower inclination. This is not uncommon within speleothems (Herries and Shaw, 2011) because a single subsample of 2.5 cm depth is measuring the remanence recorded in multiple layers of speleothem as well as multiple layers of detrital contamination. Phase B flowstone records a weak, normal polarity ChRM that is consistently isolated between 7 and 36mT (Figure 13c). The results from this phase have the highest mean maximum angular deviation (MAD) values (Table 9) due to the small and oddly shaped nature of the samples that provide less consistent measurements between each spin in the magnetometer. Phase C flowstone also records a normal ChRM that is similar to that seen within phase B (Figure 13c), although with slightly steeper inclinations (Figure 13c, Stereo Plot).

The samples do not have strong secondary overprints as seen in many Plio-Pleistocene palaeo-cave deposits from the region (Dirks et al., 2010), which may indicate that they are younger or have a different sediment source and, thus, mineralogy holding the remanence. The coercivity of phase B and C flowstones is similar, and distinct from phase A flowstone, although all the demagnetisation spectra suggest the dominant mineral holding the remanence is ferrimagnetic (magnetite/maghaemite) as at many other CoH sites (Herries et al., 2006; Herries and Shaw, 2011; Herries et al., 2014).

Three weathered bone fragments of H. naledi were analysed via radiocarbon dating at Beta Analytic Inc. (Florida, USA). Analyses indicated that no collagen was present in any of the samples and that the bone appeared possibly cremated. This was investigated with a bone carbonate extraction technique. Tests did not support cremation and indicated that extensive secondary CaCO₃ replacement had occurred, providing ages of 33.0 ± 0.2 ka and 35.50 ± 0.16 ka for two of the fragments. We interpret these ages to relate to late calcite precipitation in the bones that may reflect a wet period in the cave.

The dating techniques applied in this study do not all work in the same way, and hence the results must be viewed differently. U-series analysis on carbonates, ¹⁴C analysis of bone and palaeomagnetic analyses of flowstones (e.g., Herries and Shaw, 2011; Taylor and Bar-Yosef, 2014; Hellstrom and Pickering, 2015) are well established dating techniques requiring few, if any, à priori assumptions. In contrast, ESR and OSL results are strongly dependent on model assumptions for the environmental conditions that affected the locations from which the samples were taken.

The importance of a deep understanding of the stratigraphic position of the fossils and the geological processes that led to their deposition cannot be overstated considering the extremely complex nature of sedimentary cave fill in many cave systems (e.g., Wilkinson, 1985; Brain, 1993; Sasowsky, 1998; Stock et al., 2005; Stratford et al., 2014; Sutikna et al., 2016) involving repeated cycles of deposition, erosion and reworking, leading to complex and sometimes contradictory age results (e.g., Granger et al., 2015; Kramers and Dirks, 2017). This problem is well illustrated with the on-going debate on the age of Stw 573 (‘little foot’) in the nearby Sterkfontein Cave, where after 20 years of dating efforts no definitive age is yet established (see Partridge et al., 1999, 2003; Berger et al., 2002; Walker et al., 2006; Herries and Shaw, 2011; Pickering and Kramers, 2010; Granger et al., 2015; Kramers and Dirks, 2017). Another good example illustrating the difficulties of linking cave stratigraphy to a definitive age for the hominin fossils they contain is presented by the H. floresiensis remains in the Liang Bua cave, Indonesia (Morwood et al., 2004; Roberts et al., 2009; Sutikna et al., 2016).

The stratigraphy within the Dinaledi Chamber has been previously described by Dirks et al. (2015). The ages presented here help to resolve outstanding questions about the stratigraphy in the Dinaledi Chamber, and allow us to more closely define the distribution of correlative stratigraphic units (Figure 14), and thus constrain the age of the H. naledi fossils.

The oldest ages returned from the Dinaledi Chamber are from the baboon tooth that is embedded in sediment attributed to sub-unit 3a, followed by U-Th ages for Flowstone 1a, which directly covers erosional remnants of Unit 2 (Figure 14). Flowstone 1a consists of at least three generations of flowstone growth, named, from oldest to youngest, phase A, B and C (Figures 2b, 13a,b and 14). We dated phases B and C via U-Th, at 502 ka (RS23; Table 1) and 478 ka (RS22; Table 1) respectively, but the uncertainties are large because the ages are close to the upper dating limit of the U-Th technique. The oldest, phase A layer at the base of Flowstone 1a records reversed magnetic polarity indicating that it formed before 780 ka (Singer, 2014). The three age estimates are consistent with the stratigraphic position of the three phases in Flowstone 1a, and they indicate that the erosion remnant of Unit 2 encrusted by Flowstone 1a is also older than 780 ka. The fact that the erosion remnants of Flowstone 1a dip into the chamber suggests that at the time of formation of all three phases of Flowstone 1a, the debris cone of Unit 2 sediment was still in place; that is, erosion of Unit 2 sediment from below Flowstone 1a would have occurred sometime after 585 ka (i.e., the older age limit for phase C in Flowstone 1a) and possibly as late as 437 ka (i.e., the younger age limit of phase C in Flowstone 1a). Unit 2 contains rare fossils of macrofauna, including a long bone in the erosion remnant, which must be older than 780 ka as well. We interpret the floor sediments of sub-unit 3a that contain the baboon tooth (sample 1841) to represent, at last in part, the reworked remains of the debris cone that once existed below Flowstone 1a (Figure 8) as material was removed from the chamber via floor drains, thereby undercutting the debris cone, which responded by slowly slumping into the chamber – a process ongoing today. The baboon tooth could be part of the original, Unit 2 debris cone and, therefore, older than 780 ka, although the US-ESR age indicates that it could also be younger (with a mid point age of ~ 679 ka between a maximum age of 723 ± 181 ka and the minimum age of 635 ± 148 ka, with a possible age range of 487 ka to 904 ka). If the tooth is younger than 780 ka, it would suggest that it was derived from different sedimentary deposits not tested in this study, or that it may have entered the chamber separately during erosion of the Unit 2 debris cone, and its presence may reflect more direct entry points from surface into the chamber now sealed by flowstone. This possibility was tested with sample RS9, a flowstone sample filling a thin (<14 cm wide) fracture in the dolomite on the surface that occurs above the Dinaledi Chamber. This flowstone yielded equilibrium U-Th results meaning that it formed before ~ 600 ka (Table 1), which is consistent with the interpretation that the Dinaledi Chamber was closed to direct entry of coarser-grained sediment from the surface prior to the entry of H. naledi into the cave system and remained closed until the present (Dirks et al., 2016a).

Below Flowstone 1a are five other flowstones (Flowstones 1b-e and Flowstone Group 2; Figure 2b), which each cover erosional remnants of sediments that we originally grouped as Unit 2 (Dirks et al., 2015). The geochronology results presented here (Table 1) now permit a better evaluation of the flowstone stratigraphy in the chamber, and it is evident that Unit 2 represents a significantly older stratigraphic unit that is restricted to deposits directly below Flowstone 1a, but not to the sediment deposits below Flowstones 1b-e. Flowstone 1c returns an age of ~243 ka (Table 1), suggesting that the sediments below Flowstones 1b-e are significantly younger than Unit 2 sediments below Flowstone 1a (Figure 14). It was noted before that the sediments below Flowstones 1b-e are less indurated and less weathered than the sediments below Flowstone 1a, and that they contain H. naledi material (Dirks et al., 2015). Their distinct appearance and fossil content is now confirmed with the dating. Therefore, we reinterpret the H. naledi-bearing sediments below Flowstones 1b-e as part of Unit 3 (sub-unit 3b), which means that all H. naledi-bearing sediments in the chamber are now part of sub-unit 3b (Figure 8).

Figure 8

Download asset Open asset

Apart from the poorly consolidated erosional remnants below Flowstones 1b-e, hominin-bearing Unit 3 sediments also cover most of the floor of the Dinaledi Chamber. The layer of sub-unit 3b sediment has been interpreted as a relatively thin sheet (~20 cm) of rubbly mud clast breccia material mixed with H. naledi fossils based on outcrops in the excavation pit and preliminary ground penetrating radar results (Naidoo, 2016; Figures 2 and 8). Age brackets for sub-unit 3b were obtained by dating underlying outcrops of sub-units 1a and 1b via OSL, and by dating overlying flowstone units. In this context sample OSL5 is the most relevant for obtaining a maximum age estimate for sub-unit 3b, because it was taken from an outcrop of Unit 1 that is overlain by Unit 3. Here, a maximum age limit of 414 ka (the upper error limit of OSL5) can be assigned to sub-unit 3b if the OSL ages are taken at face value. Note however, that the maximum age limits for sub-unit 1b as determined from samples OSL3 and OSL4 are significantly less at 272 ka and 278 ka (upper error limits), respectively (Table 8), which supports the interpretation that sub-unit 1b formed due to erosion and redeposition on top of sub-unit 1a. The minimum age limit of sub-unit 3b can be more confidently constrained as it is overlain by flowstones with age ranges between 97 ka and 24 ka (Table 1). More importantly, a hanging remnant of sub-unit 3b with hominin material is covered by Flowstone 1c with a lower age limit of 236 ka obtained from the core of a stalactite overlying the rim of the flowstone remnant. This age provides the best minimum age estimate for sub-unit 3b, and by extension a minimum age for the H. naledi fossils.

U-Th dating of an erosional remnant of Flowstone 1 that occurs directly above an outcrop of sub-unit 1b from which sample OSL4 was taken, and appears to cover it, provides an age of 290 ± 6 ka, suggesting that Unit 1 in this location must be older than 284 ka. The OSL results (using the MAM model), however, suggest that the Unit 1 sediments in this location must be younger than 278 ka (Table 8; using MAM age models). This apparent paradox may indicate that the OSL ages are unreliable, but could mean that sub-unit 1b was deposited in an inverse stratigraphic order in relation to the flowstone. This is supported by physical evidence in the chamber in which a gap between the flowstone drape and underlying sediment widens towards the back of the outcrop (Figure 2c), implying that sub-unit 1b was deposited below an erosional remnant of Flowstone 1, and is, therefore, younger than this flowstone. Recrystallization-dissolution textures, and anomalously high ²³⁴U/²³⁸U ratios (>3) suggest that the age reported for this flowstone sample (RS5) should be treated as a minimum age.

Although the age constraints for sub-units 1a and 1b are imprecise they do suggest that Unit 1 in this part of the chamber is younger than Unit 2, and that the red mud clasts forming Unit 2 sediment were derived from source material matching our description of sub-unit 1a, but positioned higher up in the cave. This source material was possibly part of sub-unit 1c, or it could represent part of an older and as yet undefined sub-unit of Unit 1. No age assessment for sub-unit 1c deposits were done, because accumulations are too small to be tested with OSL or too difficult to access.

The U-Th ages from flowstones and teeth place constraints on the changing physical environment experienced in the cave chamber over time. Flowstone deposition in the Dinaledi Chamber occurred during discrete periods including 24–32 ka, 50 ka, 88–105 ka, and during older events around 242 ka, 290 ka, between ~437 ka and 683 ka, and >780 ka. The flowstones are associated with relatively low initial ²³⁴U/²³⁸U ratios of 1.8–2.4 (Tables 2 and 3). Flowstones formed in different parts of the chamber as drip points shifted, but no clear pattern in age distribution is apparent (Figure 1b) other than that the flowstones deeper in the cave generally appear to be younger. The U-Th ages for the teeth also define a number of apparent U uptake events at around 43–48 ka, 70–75 ka, 80–90 ka, 100–120 ka and ~200 ka (Tables 4 and 5). The initial ²³⁴U/²³⁸U ratios associated with each of these events is similar and anomalously high (6.9–8.1).

The periods of time during which flowstones formed in the cave, by and large, do not overlap with the periods of time during which U-uptake appears to have occurred in the teeth, although both types of events were probably associated with wet periods in the chamber. The systematic difference in the initial ²³⁴U/²³⁸U ratios obtained from flowstones, as compared to teeth, indicates that an isotopically distinct water source led to U-uptake in the teeth, such as groundwater in the dolomitic aquifer of the Malmani Group with reported anomalously high initial ²³⁴U/²³⁸U ratios (e.g., Kronfeld et al., 1994). Working on the assumption that the high initial ²³⁴U/²³⁸U ratios are derived from the groundwater reservoir, the observed age groupings suggest that U uptake events in teeth represent (partial) inundation events of the Dinaledi Chamber, whilst the flowstone formation events reflect periods during which the groundwater table had dropped below floor level, but extensive drip still occurred within the Dinaledi Chamber caused by water derived from the surface with generally lower ²³⁴U excess. The age groupings also indicate that episodic wet periods in the Dinaledi Chamber alternated with periods during which no flowstone was deposited.

Figures 1b and 14 summarize the results of all dating methods applied to the Dinaledi Chamber during the course of this study. It is clear from these results that the H. naledi assemblage in Unit 3 is of mid- to late-Middle Pleistocene age. The best age estimates for the H. naledi fossils come from the averaged US-ESR ages for samples 1788 and 1810: 229 +60/–46 ka (maximum) and 179 +49/–40 ka (minimum) for sample 1788, and 276 +59/–77 ka (maximum) and 220 +50/–60 ka (minimum) for sample 1810, with an age range of 139 ka to 335 ka. The maximum age scenario provides an average age for both teeth of 253 +82/–70 ka, and the minimum age scenario provides an average age of 200 +70/–61 ka. Considering the observed Rn loss in the cave sediments, the maximum age estimate is considered to be closer to the true age.

The lower age limit of 139 ka must be disregarded and shifted to ~200 ka considering a U-Th minimum age estimate for parts of enamel on tooth sample 1810 (Table 5). This minimum age limit can be constrained further to 236 ka based on U-Th age estimates for Flowstone 1c (242.0 ± 5.0 ka; 242.9 ± 6.6 ka; Table 1), which directly covers fossil material of H. naledi, noting that these ages may represent minimum age estimates for the flowstone as a result of possible U uptake during a period of raised groundwater levels (Figure 14).

The maximum age limit of 335 ka for the H. naledi fossils relies on 80% Rn loss throughout the burial history of the fossils and is probably an over-estimate. A separate maximum age estimate for sub-unit 3b can be obtained from the baboon tooth (sample 1841) in sub-unit 3a sediments devoid of hominin fossils, underlying the partly articulated remains of H. naledi, with US-ESR age estimates varying from 635 ± 148 ka (minimum) to 723 ± 181 ka (maximum) with an age range of 487 ka to 904 ka and a mid-point age of 679 ka. The older age estimate is considered to be more likely (considering the measured Rn loss). These maximum age estimates are consistent with the direct age estimates for the H. naledi fossils, but do not further constrain the upper age limit (Figure 14).

A further maximum age estimate for the H. naledi fossils can be obtained from age estimates for sub-unit 1a using sample OSL5. This sample provides a maximum age limit for sub-unit 3b, of ~414 ka (Table 8; using MAM). Although OSL ages are poorly constrained, and do not provide very precise age limits, this estimate is broadly consistent with the estimated age for reworking of Unit 2 sediments from below Flowstone 1a (437–585 ka) and the US-ESR age estimates for the teeth.

Considering all age results presented here the most parsimonious age estimate for the H. naledi fossils is sometime between 236 ka and 335 ka. More work will be needed in future to constrain these ages further (Figure 14).

Until now, it has been generally assumed that morphologically primitive hominins like H. naledi (Berger et al., 2015) did not survive into the later parts of the Pleistocene in Africa. This general assumption has commonly guided the interpretation of fossil discoveries with poor geological or stratigraphic context, including initial estimates for the age of the H. naledi fossils (Thackeray, 2016; Dembo et al., 2016). The new age estimates for H. naledi show that an approximate age for the hominin fossil fragments cannot be simply deduced from their morphology (Thackeray, 2016; Dembo et al., 2016). Detailed geological investigations are critical before any attempt to ascribe an age to the fossils is made, and even then great care must be taken in interpreting results, which may not always be conclusive (e.g. see famous examples such as Sterkfontein or Liang Bua; Partridge et al., 2003; Wilkinson, 1985; Walker et al., 2006; Roberts et al., 2009; Pickering and Kramers, 2010; Granger et al. 2015; Sutikna et al., 2016; Kramers and Dirks, 2017).

It is generally assumed that all African fossil hominins producing Middle Stone Age archaeological industries in the past 300 ka were part of a single variable species of early H. sapiens or an immediate precursor (e.g., Mcbrearty and Brooks, 2000; Lahr and Foley, 2001; Stringer, 2002). The new ages now show that H. naledi existed at the same time as the first Middle Stone Age tools were produced in southern and eastern Africa, whilst skeletal evidence shows that H. naledi was probably capable of tool use (Berger et al., 2015; Hawks et al., 2017). This raises the possibility of H. naledi being responsible for some of the MSA traditions. The implications of the new ages for H. naledi are discussed in detail in Berger et al., 2017.

A total of seventeen flowstone samples (RS1, RS5-6, RS8, RS10-11, and RS 13–23) from the Dinaledi Chamber were dated via U-Th geochronology, including one set of blind duplicates (RS1 and RS15; Table 1, Figures 1b and 3). In addition, one flowstone sample (RS9) was taken on the surface (WGS84 571240–7121866) from a shallow pit about 11 m SW of the projected surface position of the excavation pit in the Dinaledi Chamber (Figure 3d). For each sample a powder was prepared and then split, with one half being dated at JCU and the other half being dated at UoM; that is, for each sample both JCU and UoM dated the same material.

In the Dinaledi Chamber samples of Flowstone Groups 1, 2 and 3 were collected from a variety of stratigraphic positions (Figures 1b, 2 and 3). All the sampled flowstones formed as sheets, crusts, or drapes overlying older sediment units, with the exception of samples RS13 and RS18, which were taken from a small stalactite that formed along the lip of an erosional remnant of Flowstone 1c. Additionally, samples RS14, RS16 and RS17 were taken from a flowstone drape along a dolostone side-wall of the chamber. In all instances the flowstones have a free upper surface, that is, they are not overlain or covered by sediment, and all flowstones are interpreted to be younger than the sediment units they cover. The one exception to this rule could be RS5, which comes from a partly resorbed erosion remnant of Flowstone 1 that appears to overlie Unit 1, but is separated from the top surface of Unit 1 by a small opening that widens with depth (Figure 2c). This leaves the possibility that this flowstone is an erosional remnant, and that Unit 1 (sub-unit 1a) sediments built up below it after it had been deposited. The location of each sample within the Dinaledi Chamber is shown in Figure 1b, and outcrop and close-up photos of the sampled flowstones are shown in Figure 3.

RS1 and RS15 are blind duplicate samples taken from a hanging remnant of Flowstone Group 2 that occurs in a N-trending fracture, ~10 m N of the excavation pit and ~3 m N of a major outcrop of sub-unit 1a sediment from which sample OSL3 was collected. The flowstone overlies largely unconsolidated floor sediments of Unit 3, which in this locality have eroded from underneath the flowstone to leave a hanging remnant ~8 cm above the current floor level (Figure 3i). The flowstone consists of a 15–18 mm-thick layer of calcite overlying an irregular surface of mud clast breccia, locally incorporating and growing around large mud clasts that were lying on the palaeo-surface. The flowstone is grey-white in colour and preserves 3–6 mm scale laminations visible due to subtle colour variations. The flowstone is recrystallized with elongated, acicular crystals of calcite growing from the base to the top of the layer across all internal laminations. The upper surface of the flowstone has a rough, pitted appearance as a result of partial resorption or dissolution of calcite along the grain boundaries of the needle-like crystals. Samples RS1 and RS15 were taken from a 3 mm-thick zone, 3 mm above the basal contact of the flowstone layer (Figure 3i).

RS5 was sampled from a thin sheet of Flowstone Group 1 in a WNW-trending fracture, ~4 m W of the excavation pit (Figure 1b). The flowstone overlies orange sandy mudstone belonging to sub-unit 1b from which sample OSL4 was taken (Figure 3a), but appears to be separated from this unit by a narrow opening that widens to the back of the outcrop (Figure 2c). The flowstone consists of an 8–22 mm thick, cream white layer of carbonate with a sponge-like, porous, sugary texture that largely masks (due to recrystallization of the primary calcite) underlying mm-scale laminations. The flowstone layer appears to be partly dissolved along its stratigraphic top with the external surface of the layer truncating internal laminations. Sample RS5 was taken from a 5 mm-thick horizon, 3 mm below the top of the flowstone layer (Figure 3a).

RS6 was sampled from a thin sheet of Flowstone Group 2 in a WNW-trending fracture, ~6 m W of the entry zone (Figure 1b). The flowstone overlies erosion remnants of Unit 1 and Unit 3. At the sample site, sediments of Unit 3 have been partly eroded from underneath the flowstone leaving a hanging remnant, 5–10 cm above the current floor level (Figure 3b). The flowstone consists of an 8–12 mm thick crust overlying an irregular sediment surface, incorporating fine (<3 mm) mud clasts in the base of the layer. It is white-grey in colour and semi-transparent, with limited evidence of internal layering except for a slightly lighter coloured basal layer that is several mm thick. The entire layer is recrystallized with fine, radiating acicular crystals of calcite growing upward from the basal contact. The upper surface has a rough pitted appearance as a result of partial resorption/dissolution along grain boundaries of the acicular crystals. Sample RS6 was taken from a 3 mm-thick horizon at the bottom of the flowstone (Figure 3b).

RS8 was sampled from a thin sheet of Flowstone 2 on the floor of the entry zone below the stack of hanging remnants of Flowstone 1a-e. The flowstone overlies sediments of Unit 3 that are partly eroded from underneath the flowstone leaving a 5–10 cm gap between the flowstone sheet and the current floor (Figure 3c). The flowstone consists of a 5–22 mm thick layer overlying an irregular surface of mud clast breccia. The grey-white flowstone preserves 3–5 mm thick layering visible as subtle colour variations, despite partial recrystallization and replacement by elongated acicular crystals of calcite. The basal 10 mm is composed of a mesh of fine aragonite needles. Above this zone acicular calcite replaces the aragonite. The upper surface of the flowstone has a rough, pitted appearance as a result of partial resorption/dissolution along grain boundaries of the acicular crystals. Sample RS8 was taken from the bottom 3–5 mm of the flowstone layer (Figure 3c).

RS9 was sampled from a surface outcrop in a shallow mine pit that occurs about 11 m SW of the projected surface position of the excavation pit in the Dinaledi Chamber (Figure 3d). The sample site occurs along strike of the same fracture along which the Dinaledi Chamber was formed at depth, and it is therefore possible that this surface outcrop in the past could have linked to the cave system below. The flowstone consists of a 12–25 mm thick layer of carbonate overlying an irregular surface of consolidated mud clast breccia similar to sediments of Unit 2 in the Dinaledi Chamber. The grey-white flowstone preserves 1–5 mm scale laminations visible due to colour variations and the presence of several thin brown marker surfaces. The laminations are locally recrystallized and overgrown by radiating, elongated crystals of calcite growing from the base to the top of the flowstone layer. Additionally, fibrous aragonite needles are widely distributed in sheaf-like patterns. The upper surface of the flowstone has a rough, pitted appearance as a result of partial resorption/dissolution along the grain boundaries of the acicular crystals. Sample RS9 was taken from a 4 mm-thick white horizon, 2 mm above the basal contact of the flowstone layer (Figure 3d).

RS10 was sampled from a thin crust of Flowstone Group 2, about 2 m W of the excavation pit and 0.2 m W of the location where tooth sample 1810 was found (Figure 3e). The flowstone directly overlies mud clast breccia of Unit 3 and hominin bone fragments. It consists of a 4–8 mm-thick laminated flowstone crust overlying an irregular sediment surface and incorporates fine (<3 mm) mudstone clasts within its basal laminae. This flowstone is white to grey in colour and is finely laminated and partly recrystallized, with recrystallization visible as white, fine, radiating needles of aragonite growing upward from the basal contact, along an irregular alteration front into grey-white, laminated calcite near the top of the layer. RS10 was taken from the laminated bottom 3 mm of the flowstone layer (Figure 3e).

RS11 and RS21 represent two samples taken from the top and bottom, respectively, of the same flowstone unit developed along the floor below an active drip point, 1.5 m E of the excavation pit. This flowstone consists of a basal layer of grey flowstone interpreted as Flowstone Group 2 overlying Unit 3 sediments, covered by a white flowstone layer containing small stalagmites interpreted as Flowstone Group 3 (Figure 3f). The sample collected for dating consists of a 20–28 mm thick, finely laminated flowstone including the two distinct layers described above (Figure 3f). The basal layer (RS21; Figure 3f) is 3–12 mm thick, and consists of brown-grey calcite resting on Unit 3 sediment. This layer is strongly recrystallized with radial, acicular crystals of calcite growing upward from the lower contact, which locally appears to have replaced an earlier generation of acicular aragonite. These crystals overgrow mm-scale laminations defined by white to brown colour variations. RS21 was sampled from the basal-to-central segment of this layer. The basal layer is overlain by an 8–16 mm thick layer of finely laminated (sub-mm scale) white speleothem. The basal 2–3 mm of this layer is recrystallized with fine aragonite needles radiating out from the base. Above that the flowstone consists of fine lamellae defined by subtle grey to white colour variations. Each lamina is composed of botryoidal aggregates of acicular aragonite. RS11 was taken from the top 3 mm of this white flowstone (Figure 3f).

RS13 and RS18 are two samples taken from the same stalactite developed along the lip of an erosional remnant of Flowstone 1c near the entrance into the Dinaledi Chamber (Figure 1b). In this location, Flowstone 1c covers the erosional remains of a mud clast breccia containing a long bone consistent with the H. naledi assemblage (Figure 3g), and interpreted as Unit 3. The stalactite connects to part of the speleothem layers that cover the bone.

The stalactite preserves well-developed internal layering, with layers asymmetrically developed around a core, in which layers thicken towards the outward facing side of the stalactite. From core to rim the stalactite reaches a maximum thickness of 53 mm and includes three separate zones that can be distinguished based on internal texture, layering, and colour (Figure 3g). The innermost zone forms an 8 mm-thick, finely laminated (sub-mm scale) core centred on a small mud clast and is terminated by a thin brown, mud-rich rim. This core is surrounded by a ~23 mm-thick central zone that consists of more coarsely layered (3–10 mm-thick), cream to grey-white coloured calcite with layering preserved as subtle colour variations, which at its base shows replacement by radiating sheaves of aragonite needles. This zone is mantled by a ~22 mm-thick outer zone of white calcite preserving only remnants of internal layering. This outermost zone is characterized by extensive replacement of abundant, older aragonite needles by coarse-grained calcite, creating a patchy texture. The sample has been affected by recrystallization resulting in the formation of radiating acicular crystals of calcite that grow from the core outward. The white outer layer also shows evidence of further recrystallization, with the formation of coarse (2–4 mm), equant calcite grains, many with highly irregular grain boundaries that overgrow the acicular grains. The outer surface of the stalactite has a rough, pitted appearance as a result of partial resorption/dissolution along the surface with dissolution along the grain boundaries of the acicular crystals. RS18 was sampled from a 3 mm-thick horizon near the base of the central zone. RS13 was sampled from the outermost part of the outer zone (Figure 3g).

RS14 was sampled from an irregularly shaped, cascade-like crust of Flowstone Group 2, along the side-wall of the main floor drain, between the entry shaft and excavation pit (Figures 1b and 3h). The sample of flowstone crust consists of an 8–11 mm thick layer of cream coloured carbonate displaying mm-scale laminations and a sugary, recrystallized texture with numerous, fine pore spaces along laminar surfaces. Sample RS14 was taken from a 3 mm-thick zone encompassing several fine laminations, ~5 mm above the base of the flowstone layer (Figure 3h).

RS16 and RS17 are two separate samples from a cascade-like crust of Flowstone Group 2 that formed within the main floor drain between the entry shaft and excavation pit (Figure 1b). This flowstone developed on top of the dolomite back-wall of the drain (Figure 3h). The sample consists of a mostly cream coloured flowstone that is up to 73 mm thick, with a massive sugary, recrystallized texture, preserving mm-scale laminations visible due to subtle colour variations. Many of the equant calcite grains contain remnant aragonite needles, reflecting an earlier phase of aragonite growth. Laminations along the basal 3–4 mm of this flowstone are brown in colour, whereas the top 10–12 mm of this flowstone layer consists of what is possibly a separate, younger unit of laminated grey-brown carbonate with small pore spaces developed along some of the laminar surfaces. Sample RS16 was taken from a 3–4 mm-thick layer directly above the basal zone with brown laminations. Sample RS17 was taken from a 4 mm-thick zone at the top of the cream-coloured laminated flowstone, immediately below the darker coloured top unit (Figure 3h).

RS19 was taken from a wedge-shaped sample of Flowstone Group 2 that formed along the lip of a drip pool directly overlying Unit 3 sediments, 1.2 m S of the excavation pit (Figure 1b). The flowstone is cream-coloured and up to 30 mm thick, and overlies a flat sediment surface. It preserves complex internal layering and a variety of textures (Figure 3j). A basal 2–4 mm-thick layer characterised by fine calcite needles growing at right angles to layering is overlain by an irregular mass of darker cream coloured, unstructured calcite with a sugary texture and numerous fine voids. Towards its top this unstructured mass is interlayered with and covered by several 1–4 mm-thick lamellae characterised by fine calcite needles growing at right angles to layering similar to the basal layer, but containing pore spaces. These in turn are overlain by an 8 mm-thick zone of finely laminated flowstone with a fine-grained (<0.5 mm) sugary texture and abundant fine pore spaces. RS19 was taken from the basal flowstone layer with radiating calcite crystals (Figure 3j).

RS20 was sampled from a thin sheet of Flowstone Group 2 in a WNW-trending fracture, ~2 m W of the excavation pit (Figure 1b). The flowstone forms a cascade-like drape that directly overlies orange mudstone of sub-unit 1a from which sample OSL5 was taken (Figure 3k). RS20 occurs directly above sample RS10, which was taken at the foot of the cascade (Figure 1b). The sampled flowstone layer is irregular in shape as it covers sediments with topography. It consists of a finely laminated 15–22 mm thick crust overlying sub-unit 1a sediment and includes a ~6 mm thick, laminated basal unit of white flowstone that incorporates small (3–10 mm) mud clasts, overlain by more coarsely laminated (mm-scale) grey brown carbonate free of inclusions. The flowstone is internally locally recrystallized with acicular aragonite crystals growing within ~3 mm thick zones at right angles to layering. RS20 was sampled from the basal layer of the sample (Figure 3k).

RS22 and RS23 are two samples from an eroded rim of Flowstone 1a near the entry shaft into the Dinaledi Chamber (Figure 1b). This flowstone overlies erosional remnants of well-indurated mud clast breccia assigned to Unit 2 (Figure 3l), and mostly consists of coarsely recrystallized white carbonate. The flowstone sample from which RS22 and RS23 were taken occurs down dip from the palaeo-magnetic sample taken from Flowstone 1a as described below (Figures 2b and 13). The flowstone layer is ~25 mm thick and comprises a 6 mm thick, basal unit of grey to brown calcite that is recrystallized into fine (sub-mm scale), equigranular grains of calcite overgrowing an older acicular texture. The basal layer corresponds to phase B carbonate described for the palaeo-magnetic sample, and is overlain by a clean white calcite unit (i.e., phase C in the palaeo-magnetic sample) that is coarsely recrystallized with 2–4 mm equidimensional calcite grains overgrowing (and partly destroying) an older texture formed by acicular grains. Small voids occur in the interstitial spaces between the coarse calcite grains. RS23 was taken from the grey basal unit (i.e., phase B) of the flowstone layer and RS22 was sampled from the central part of the upper white recrystallized unit (i.e., phase C, Figures 3l and 13b). In this sample, flowstone belonging to the older phase A carbonate is no longer present, as this layer pinches out along dip between the point where the palaeomagnetic sample was taken and the point where the U-Pb sample was taken as can be seen in Figures 2b and 13.

Three H. naledi teeth (samples 1810, 1767 and 1788) and one baboon tooth (sample 1841) were collected from Unit 3, in the Dinaledi Chamber (Figures 1b, 4, 6 and 7) for ESR dating and U-Th analysis. Sample 1767 (full catalogue number U.W.101–1767) is an extremely worn upper premolar crown (Figure 6a) obtained from approximately 1 m SW of the excavation pit (Figure 4), and occurred on surface surrounded by mud clast fragments of Unit 3. This tooth is deeply weathered, and preserved only a small rim of enamel on the buccal margin with a maximum height of 4.5 mm. Otherwise the crown is a concave dentine surface worn to the cervix, with bright white, highly bleached dentine. The tooth is brittle and appears strongly affected by water action. Sample 1788 (full catalogue number U.W.101–1788) is a lower right second molar (Figure 6b) obtained from approximately 2 m WSW of the excavation pit (Figure 4), embedded within loosely packed, mud clast breccia of Unit 3, ~ 2 cm below the ground surface level. This tooth is partly broken at the root, but has an otherwise well preserved crown with thick, light-grey enamel. The distal root of the tooth is present and complete, but the lingual root is broken off just below the cervix. The tooth is morphologically consistent as an antimere of U.W.101–284. The dentine is highly bleached and brittle and appears affected by water action. Sample 1810 (full catalogue number U.W.101–1810) is a lower left third premolar or possibly lower left fourth premolar (Figure 6c) obtained from the SE corner of the excavation pit (Figure 4), embedded within sediments of Unit 3, ~5 cm below the original ground surface level. This tooth is well-preserved with thick, light-blue-grey enamel, and shows little evidence of bleaching or weathering, with only a very slight polishing wear on the distal crest of the protoconid. The morphology of the crown is similar to other lower third premolars in the collection, however, it is slightly shorter than many of those. Sample 1841 (full catalogue number U.W.101–1841) is a well-preserved tooth crown, morphologically consistent with a lower left second molar of Papio sp. This is a complete enamel crown of an unerupted tooth with no wear facets or interproximal facets in evidence, and the roots had not formed (Figure 7). The specimen was recovered from a sediment sample taken at a depth of 55–60 cm below the original ground surface of the cave floor near the base of the sondage dug in the centre of the excavation pit (Figures 1b, 2d and 4). This tooth occurs in sub-unit 3a, ~40 cm below the stratigraphically lowest occurrence of partly articulated remains of H. naledi, and represents the only non-hominin macrofossil recovered from Unit 3.

Three OSL samples (OSL3, OSL4 and OSL5) were collected in the Dinaledi Chamber from erosion remnants of sub-unit 1a, in which sandy, laminated mudstones are exposed (Figures 1b and 5). These sandy intercalations were targeted because they contain fine-grained quartz and feldspar grains that can be extracted for analysis. For each sample a 30 cm length of aluminium piping with a diameter of 5 cm was hammered into the sediments in a horizontal direction or parallel to sedimentary laminations visible within the units. A core sample within the pipe was extracted for OSL analyses together with a sediment sample from the same unit to determine background radiation from measured values of U, Th and K.

Sample OSL3 comes from an erosion remnant of sediments of sub-unit 1b collected near the intersection point of two fractures trending N and E respectively, ~6 m N of the excavation pit (Figure 1b). Samples OSL4 and OSL5 come from an erosion remnant of Unit 1 sediment along an ENE-trending fracture, ~3 m W of the excavation pit. Sample OSL4 is obtained from sub-unit 1b directly below a thin, partly resorbed flowstone sheet attributed to Flowstone Group 1 (Figure 3a). Sample OSL5 occurs as an erosional remnant of sub-unit 1a, 1 m E, and stratigraphically 10–20 cm below sample OSL4. The Unit 1 sediments in this location are partly covered by Unit 3 sediments and a cascade of Flowstone Group 2 (Figure 3k).

The speleothem sampled for palaeomagnetic analysis comes from Flowstone 1a near the entry zone into the Dinaledi Chamber (Figure 2b). The sample is layered and comprises three distinct phases (from base to top: A-C) separated by thin clastic horizons that mark disconformities (Figure 13a,b).

The lower phase (phase A) is interstratified with visible clastic laminations. The internal laminations are truncated along the lower surface of the sample indicating that phase A flowstone was partly dissolved during a phreatic event after the flowstone had been deposited. The middle phase (phase B) consists of intercalated flowstone speleothem and detrital sediment layers, with the detrital layering concentrated towards the downslope part of the sample. Detrital material is generally less than in phase A, indicating decreasing amounts of clastic contamination. Like phase A, the internal layering of phase B is truncated along the lower surface of the sample, indicating that the phreatic dissolution event occurred after deposition of phase B. There is no apparent truncation surface between phase A and B flowstone, indicating that phase B formed on top of phase A after a period of non-deposition during which detrital material accumulated on top of phase A. The upper surface of phase B truncates internal layering reflecting dissolution during a phreatic event. This was followed by a period of non-deposition of flowstone during which a thin layer of detrital clastic material accumulated, before deposition of the upper layer (phase C) of flowstone occurred. Phase C comprises younger, cleaner flowstone which displays extensive recrystallization along the base of the unit with the formation of elongated calcite crystals. This zone of recrystallization could potentially also indicate a younger infill of a cavity that had formed between phase B and C flowstone, and has been avoided during sampling. Phase C flowstone contains little to no detrital inclusions, and suggests that sediment influx into the chamber was not occurring during its deposition.

Phase B and phase C flowstone correlate with samples RS23 and RS22 respectively, that were collected for U-Th dating. The top of the sample was oriented in the cave to magnetic N (−18.2^o degrees from true N at this location with a −62.9^o inclination). The inclination was accounted for by marking the sample on a completely flat surface of the block.

Three weathered bone fragments of H. naledi were collected for radiocarbon dating including: (i) a tibia shaft fragment, 53 mm in length (U.W. 101–567); (ii) a femur shaft fragment comprising the whole circumference of the shaft, 79 mm in length (U.W. 101–857), and; (iii) a metatarsal or metacarpal shaft with no articular morphology, 48 mm in length (U.W. 101–065). All samples were collected from the surface of Unit 3 as isolated fragments near the excavation pit.