Introduction

Burials have been recognized in the archaeological record as pits within the earth that were intentionally dug to inter the remains of the dead (Pearson, 1999; Pettitt and Anderson, 2020). Burials are one of many kinds of mortuary practices, which are diverse across human societies and have meanings and functions that are in many cases related to their forms (Metcalf and Huntington, 1991; Robben, 1991). Late Pleistocene modern humans and Neandertals had varied mortuary practices including burials. They interred bodies within anthropogenic pits, natural depressions, and rock niches, in varied postures, sometimes alone and sometimes with multiple bodies together (Martinón-Torres et al., 2021; Tillier, 2022; Maureille, 2022).

It can be challenging to test whether ancient remains were part of a burial due to the varied forms of burials and subsequent diagenesis of sediments. The oldest known human burial in Africa was at Panga ya Saidi cave in Kenya, dating back to 78.3 ± 4.1 ka, where the partial skeleton of a child was intentionally buried in a pit dug within the earth, with precise articulation and excellent anatomical association of skeletal elements indicating in-situ decomposition of the body (Martinón-Torres et al., 2021). In that case the burial pit was delimited, and the burial fill mixture of ferruginous silt and sand was found to be different from the sediment layer where the pit was excavated, but similar to the two overlying layers (Martinón-Torres et al., 2021).

Here we describe burials from the Rising Star cave system (Berger et al., 2015; Dirks et al., 2015; Fig. 1) containing bodies of Homo naledi individuals. We uncovered and partially excavated two delimited burial features on the floor of the Dinaledi Chamber (Dirks et al., 2015; Fig. 1) and extracted en bloc a delimited burial feature from the floor of the Hill Antechamber (Elliott et al., 2021; Fig. 1). Analysis of the stratigraphy, textures, geochemical composition and granulometry of the sediments around and within the burial features on the Dinaledi Chamber floor suggests that the two burials occur in pits that were intentionally dug.

Figure 1.

Results

Dinaledi Feature 1

The Dinaledi Feature 1 (Figs. 2 & 3) was uncovered in 2018 in an excavation unit within the Dinaledi Chamber immediately to the north of our 2013–2014 excavation area that produced abundant skeletal remains of H. naledi (Fig. 4). Within this new excavation, at a depth of 8 cm, a concentration of bones and accompanying soil disturbance forms a roughly oval-shaped distribution. On three sides away from the earlier excavation, this oval is clearly delimited from surrounding sediment in which few small bone fragments occur. On the fourth side, the space between this feature and the earlier excavation area included three long bone fragments. The feature measures approximately 50 cm in total length and 25 cm in breadth (Fig. 1). This area of the chamber floor has a slope of ∼11 degrees from magnetic North to South, and the feature itself is on a horizontal level beneath this sloping surface. We excavated 108 skeletal specimens from directly above or within the feature as we exposed its horizontal extent; all are anatomically consistent with H. naledi (SI 2.1). During this excavation, we recognized that the developing evidence was suggestive of a burial, due to the spatial configuration of the feature and the evidence that the excavated material seemed to come from a single body. We halted further excavation and left the feature with identifiable skeletal remains in place across its surface. At the southernmost extent of the feature, we excavated a profile that reveals a remaining depth of skeletal material of 3cm in this location; we do not know whether this depth is consistent across the remainder of the feature.

Figure 2.

Figure 3.

Figure 4.

A combination of stratigraphic, anatomical, and taphonomic evidence supports this feature as a burial. The key observations are (1) the difference in sediment composition within the feature compared to surrounding sediment; (2) the disruption of stratigraphy; (3) the anatomical coherence of the skeletal remains; (4) the matrix-supported position of some skeletal elements; and (5) the compatibility of non-articulated material with decomposition and subsequent collapse.

The feature intrudes into the unlithified mud clast breccia (Fig. 3a-c, Supplementary Table 1) of sub-unit 3b deposits, the fossil-bearing sub-unit in the Dinaledi Chamber (Dirks et al., 2017; Wiersma et al., 2020; Robbins et al., 2021). The angular mud clasts found in this subunit are fragments of the laminated orange-red mud facies from Unit 1 deposits (Fig. 3d) (Wiersma et al., 2020). These angular mud clasts were incorporated into the unlithified mud clast breccia in a mostly chaotic and unstructured way, without any observable microstratigraphy (Fig. 3c). However, the abundance of laminated orange-red mud clasts with grain sizes >2 cm around Feature 1 seems to be related to a laterally continuous layer where these clasts emanate (Fig. 3b). This layer was not noted during our previous excavations where the clasts were reported (Dirks et al., 2015, 2017), but it is continuous in the profile immediately to the east of the feature; it is disrupted in the sediment profile at the southern extent of the feature (Fig. 3b). Some orange-red clasts are visible within the exposed south face of the feature and isolated clasts occur within the feature itself. We carried out petrographic analysis of sediment from within the feature as well as surrounding it and compared with sediments from other parts of the Rising Star cave (SI 1). The bulk chemistry results of the sediments were evaluated using principal components analysis (PCA). The PCA yields three principal components (Fig. 5). Principal component 1 (PC1) explains 51% of the total variation in the major element chemistry with a dominance in the positive axis of Ti, Al, K and Ba vs Ca and Mg in the negative direction (Figs. 5 and 6). The dominance of Al and K is consistent with the clay-rich mudstone sediments across the Rising Star cave (Dirks et al., 2015, 2017; Wiersma et al., 2020). Sediments in the Lesedi chamber and other areas in the Dinaledi Subsystem have higher negative loadings in PC1 indicating more calcite cementation whereas sediments from the Dinaledi Chamber do not exhibit pronounced negative loadings (Ca and Mg) in PC1 which supports their uncalcified nature (Fig. 6). PC2 describes only 8% variability with a dominance of Si, P and S in the positive direction vs all other elements, but mainly Mn and Fe in the negative axis (Figs. 5 and 7). PC2 distinctly delineates fossil-bearing sediments from sterile sediments based on the positive loadings of P and S, which we assume to be proxies of the fossil bones (Fig. 7). For example, around the Dinaledi Feature 1, PC2 shows that the floor sediment collected during the excavation that uncovered the feature has sterile areas with no positive loadings of P and S, and fossil-bearing areas have positive loadings of P and S (Figs. 3 and 7). While different from surrounding sediment on PC2, the infill of the feature resembles the composition of other parts of the Dinaledi Subsystem and Lesedi Chamber where remains of H. naledi have been unearthed (Fig. 7). Sediments around the feature known to be sterile (group SA samples, Figure S1a) do not show convincing positive loadings of P and S, but sediments inside and below the feature (group SB and SE samples, Figure S1) as well as between Features 1 and 2 (group SC samples) show convincing positive loadings of P and S. PC3 also explains 8% variability with dominance of Mn, Na, Ba, Fe and Si in the positive domain vs mainly Ti, Al, Ca, Mg and K in the negative direction (Figs. 5 and 8). The PC3 positive loadings seem to be indicators of secondary processes that promote the breakdown of LORM to form UMCB. This is based on the pervasive deposition of Mn and Fe as Mn-Fe-oxihydroxide by microbial activity during the autochthonous transformation of LORM to UMCB (Wiersma et al., 2020).

Figure 5.

Figure 6.

Figure 7.

Figure 8.

The skeletal representation and spatial relationship of elements indicate that Feature 1 contains predominantly the remains of a single body including the eighty-three identifiable bone fragments and teeth that we recovered above and within the exposed circumference of the feature. Some of these were in direct contact with underlying fossil material, while others were above the feature separated by up to 5 cm of sediment. The excavated hominin remains (SI 2.1) include a fragmentary adult left hemi-mandible with associated dentition, partial adult right femur, fragments of both left and right humeri, left and right ulnae, left and right radii, vertebral and rib fragments, ischium fragment, and some cranial fragments. Three of the fragments that we excavated directly above and in contact with Feature 1 represent two elements of at least one additional individual: an immature proximal right femur and shaft fragment compatible with an immature femur, and a fragment of an immature proximal left humerus. Although some skeletal material is obscured by sediment, many elements remaining within the unexcavated portion of the feature are identifiable (Fig. 1d, Fig. 9). None of the visible material duplicates elements already recovered from the feature. Fragments of a right hemi-mandible are present that are consistent with the excavated left hemi-mandible. Much of a cranial vault is also present. Some elements visible on the surface of Feature 1 are in articulation and most are near related anatomical elements (Fig. 9; see Materials and Methods).

Figure 9.

The spatial arrangement of the skeletal remains is consistent with primary burial of a fleshed body, covered in sediment, followed by decomposition and post-depositional collapse. Some elements remain in a vertical or sub-vertical orientation reflective of decomposition with matrix support (Bolter et al., 2018). Such matrix-supported elements predominantly occur near the edge of the feature. A portion of ribcage is visible with some ribs oriented near vertical, others broken and collapsing along a craniocaudal axis, all apparently constrained by matrix support along the edge of the feature (Fig. 1d). Near the south end of the feature, cranial fragments are supported by matrix in sub-vertical orientations (Fig. 1d; Fig. 9). The decompositional collapse of the structures making up the thorax and pelvis, and void-space resulting from this and the decomposition of the soft tissues, would have brought overlying sediment into spaces. This process may have allowed the extraneous bone fragments from at least one other individual into the feature from overlying sediment. We also cannot exclude that some physical disturbance may have resulted from H. naledi activity within the chamber subsequent to the burial, such as digging nearby burials, or trampling. Such disturbance may also explain the breakage and spatial displacement of the two halves of the mandible.

The decomposition and dissociation of this body are consistent with an original burial environment substantially like the current sedimentary conditions in the chamber (Dirks et al., 2015). The feature includes angular laminated orange-red mud clasts that are inconsistent with saturation of the sediment by water after interment (Dirks et al., 2015, 2017; Wiersma et al., 2020). Without a water-saturated environment, the decomposition of soft tissue resulted in voids, into which the bones comprising anatomical units (head, torso, limbs) could collapse, resulting in displacement of fragments and joint articulations. By contrast, the decomposition of a body within a natural depression on the cave floor results in a different pattern of displacement. The matrix-supported elements at the edges of this burial, the evidence of disruption of the laminated orange-red mud clast layer, and the inclusion of mud clasts within the skeletal remains are also consistent with deposition of the body within a pit dug by Homo naledi and covered rather than exposed upon the cave floor within a natural depression.

At the western extreme of the excavation unit containing Dinaledi Feature 1, a second concentration of bone is visible, which we have designated Dinaledi Feature 2 (Fig. 1), separated from Dinaledi Feature 1 by approximately 20 cm of sterile soil. This feature enters the west wall of the excavation unit and we do not know its full extent. The small number of elements recovered from Feature 2 are compatible with a single individual being present and its visible form is consistent with that described for Feature 1. We have left Feature 2 otherwise undisturbed for future investigation.

Hill Antechamber Feature

The Hill Antechamber feature was uncovered during excavations conducted in 2017. The excavations were of a debris cone situated in the northeast part of the chamber, near the base of the vertical entrance (termed “the Chute”) used by our team to access the Dinaledi Subsystem (Fig. 10). At a depth 20 cm below datum, and 5 cm below the sloping surface, the team encountered a localized accumulation of bone fragments and skeletal elements across a roughly triangular space some 38 cm by 48 cm. Due to the fragile state of the material, the team excavated around the edges of this bone concentration, pillaring it to a depth of approximately 20 cm. To remove it safely through the Chute, the team separated the feature into one larger and two smaller subsections, each of which was encased within a plaster jacket for removal. The two smaller subsections were scanned on a micro-computed tomography (micro-CT) scanner (see Materials and Methods). The largest subsection was too large for this micro-CT unit and so we obtained a clinical CT scan of this specimen with 0.5 mm voxel size (see Materials and Methods).

Figure 10.

From the CT scans of the feature, we identified a minimum of 90 skeletal elements and 51 dental elements. The dental elements include at least 30 teeth from a single individual, which we have designated as Individual 1. All maxillary teeth of this individual are spatially contiguous, in correct occlusal order, and are consistent with being positioned within what was once a complete maxilla that either has degraded or does not present a contrast in density with surrounding sediment (Fig. 10). The mandibular teeth of Individual 1 are in two disjunct parts; the left mandibular dentition and right I₁, I₂, P₃, and P₄ are in correct order and in situ within a mandibular fragment that is approximately 1 cm from occlusion with the left maxillary dentition; the right mandibular molars are in near occlusion with the right maxillary molars, making clear that the mandible of this individual was in articulation with the skull when these teeth arrived at their current positions. The third molars of Individual 1 are partial crowns without roots consistent with a late juvenile developmental stage (Bolter et al., 2018). At least fourteen teeth represent a second late juvenile individual designated Individual 2.

These teeth are localized within the feature, all within a 10-cm area, but are not in any anatomical order. The identifiable elements are all maxillary teeth. Based on the incomplete roots of the canines, premolars, and second molars, Individual 2 was a slightly younger juvenile than Individual 1, at or shortly after second molar eruption. Four additional teeth represent a third individual designated Individual 3. Individual 3’s remains preserve left and right permanent maxillary canines, a right permanent mandibular canine, and a probable mandibular incisor, all localized within an 8-cm area. Nearly translucent objects that may be developing crowns of a much younger individual are identified as probable Individual 4.

The identifiable postcranial elements in the feature are nonduplicative and consistent with a single individual. Based on the spatial distribution of the remains we deem it most likely associated with the Individual 1 skull. Many elements are in articulation or sequential anatomical position, including a substantially articulated right foot, ankle, and adjacent lower limb bones, two series of partial ribs, a partial right hand, and ulna and radius. Other elements attributable to this skeleton are in anatomical proximity; for example, most left hand elements are within a 7cm radius of each other, fragments of right tibia and fibula are adjacent to each other and the right ankle, and elements of the left foot are localized within a small radius. Elements with thin cortical bone such as vertebral bodies, tarsals, and pelvic elements are less evident in the CT data; this may either reflect a lack of density contrast with surrounding sediment or the postdepositional degradation of these elements similar to the excavated material from the Dinaledi Chamber (Dirks et al., 2015). In addition to these identifiable elements, a number of small unidentifiable diaphyses can be seen in the northwest corner of the feature; these are consistent with a young immature individual and may be attributable to Individual 4. Synchrotron scanning of these specimens is planned as part of future work.

The configuration of the skeletal remains in this feature is consistent with the body of Individual 1 being in a flexed position at the time of interment, with the right foot and right hand near or at their current spatial positions. The proximity and anatomical attitude of the left humerus, radius, and ulna, and left hand and wrist material is consistent with flexion of the left upper limb. Based on their current position, we conclude that the decomposition of muscle and integument created a void space into which the elements collapsed, with further spatial displacement following loss of connective tissues within the joints, suggesting the limb was resting at a somewhat higher level during the early post-mortem period.

Collectively, the remains of the torso and upper limbs of Individual 1 now lie in a relatively flat plane that represents their collapsed state. These are separated from the lower limb and right foot by sediment that must have been in place prior to decompositional slumping of the upper body. The absence of such a matrix layer (i.e., if the burial was configured as an open pit) would have ensured that the torso and limb groups would have settled in a single layer at the base of the cut (Mickleburgh and Wescott, 2018). Instead, the vertical separation between lower limb and upper body material is indicative of sedimentary matrix support.

Decompositional slumping of the body and/or subsequent matrix compression has resulted in post-depositional fracturing visible on all identifiable rib bones. The upper limb elements exhibit multiple transverse and step fractures consistent with the effects of post-depositional matrix compression and other effects from the burial environment or our movement of the specimen. The maxillary dentition, with the mandible in or near articulation, is presently in a semi-inverted position, located above several rib shafts and near the right hand. This position is consistent with the displacement of the skull from a higher position, following decomposition of the nuchal musculature and cervical connective tissue, allowing the head to disarticulate from the axial skeleton, accompanied by rolling, inversion, and some horizontal displacement.

This feature presents clear sedimentary evidence that a pit was dug through existing stratigraphy (Fig. 10, SI 3.1). The present floor of the Hill Antechamber is a steep slope of approximately 30 degrees from the passages in the north wall toward the two narrow passageways leading to the Dinaledi Chamber. The Hill Antechamber feature was excavated at the top of this slope, near the northeast wall. As in the Dinaledi Chamber, the floor of the Hill Antechamber is composed of subunit 3b unlithified mud clast breccia, including laminated orange-red mud clasts. The excavation profiles at the east and south edges of the feature show distinct layering of orange-red mud clasts parallel to the current chamber floor. The CT data show that these layers of clasts are interrupted within the feature; the feature exhibits two clast-rich zones. A bowl-shaped concave layer of clasts and sediment-free voids makes up the bottom of the feature. At the south end of the feature this clast layer slopes in the opposite direction as the chamber floor. The right foot and lower limb material is supported by this concave layer, sloping southwest-to-northeast approximately 15 degrees. Above the foot is additional sediment with a high density of clasts and small air pockets or voids, separating it from the upper limb and torso material. This skeletal material slopes at 10 degrees northeast-southwest, opposite the slope of the foot and leg, and different from the 30 degree slope of the chamber floor. The texture and internal stratigraphy of this sediment infill is not compatible with slow incremental deposition or the percolation of small particles around and into skeletal remains that were once exposed on the antechamber floor (Fig. 10). The combination of these lines of evidence indicates that a pit was dug into existing strata, and then a body was placed into it and buried prior to the decomposition of soft tissue.

In their current situation it is not possible to inspect the surfaces of the Hill Antechamber bones for other signs of taphonomic modification, and so we cannot rule out further involvement of hominins in the removal of the skull or disarticulation of the remains. It is worth noting that the amplitude of movement away from the original anatomical position will depend upon the position of the remains when deposited, and the available space in which any movement can take place; such spaces can be created by the decomposition of surrounding soft tissues in cases of interment in sediment, and the amplitude of movement represents the extent to which a bone can move in any of the possible directions. Primary dispositions (especially burials) tend to place bones in a relatively stable position, and thus there will be little if any movement of individual bones or elements following the loss of soft tissue. When disposition occurs where bones are placed in an unstable position they will move in accordance with gravity and the shape of the surrounding space once soft tissue is lost (Roksandic, 2002). As such, the spatial distribution and alignment of the Hill Antechamber remains are consistent with decompositional processes taking place after their deposition into a sedimentary-filled pit feature (Mickleburgh and Wescott, 2018).

In addition to the skeletal remains, the feature contains a single stone artifact in close contact with the articulated hand and wrist material (SI 3; Figs. 11 and 12). It is emplaced within the feature at an angle of 25 degrees from horizontal in a southeast-northwest slope, in a different orientation from with any of the skeletal material or the slope of the chamber floor, and it does not rest upon the bowl-shaped bottom of the feature. We obtained a high-resolution scan of this artifact from the European Synchotron Research Facility; it is described in Supplementary Information 3.

Figure 11.

Figure 12.

Discussion

These burial features meet evidentiary standards used for recognizing burials of H. sapiens (Martinón-Torres et al., 2021). The recognition of burials in these chambers within the cave system prompts us to evaluate the broader array of H. naledi remains for evidence of mortuary activities. Our 2013–2014 excavation was localized between 1 and 2 meters to the southwest of Dinaledi Feature 1 (Fig. 1c). Within an excavation area of 80 cm by 80 cm we recovered remains attributable to a minimum of 5 individuals, with many elements highly fragmented and others in articulation with matrix support preserving anatomical regions (Berger et al., 2015). Our study of the spatial distribution of this material identifies concentrations that are spatially identifiable, although with substantial commingling and loss of anatomical ordering of much of the material (Kruger, 2017) (Figs. 4 and 13). A parsimonious explanation for this configuration of skeletal remains is that these remains may be a palimpsest of burials that have sequentially disrupted each other. In this hypothesis, early burials were disturbed when pits were dug for subsequent burials. Other occurrences of remains outside of the Dinaledi Chamber and Hill Antechamber (Hawks et al., 2017; Brophy et al., 2021) are discussed as possible evidence of mortuary practices in SI 4.2. Instances where parts of individuals occur in remote narrow passages cannot be explained as a result of carnivore or water transport (Elliott et al., 2021; Brophy et al., 2021), making it necessary to consider that H. naledi may have placed these partial remains in these locations, possibly representing a form of funerary caching. It is possible that H. naledi used certain parts of the cave system for burials and other mortuary practices in contrast to other kinds of behaviors, and further exploration of the cave system may assess that hypothesis.

Figure 13.

The existence of diverse practices in the placement or interment of individuals within the Rising Star cave system is similar to the diverse practices noted within other sites of multiple burials of H. sapiens including Qafzeh Cave and Skhūl Cave (Tillier, 2022). Based on current dates associated with the Dinaledi and Hill Antechamber remains of H. naledi, this complex treatment of the dead of H. naledi may pre-date the earliest evidence of burials by H. sapiens in Africa by as much 160,000 thousand years or more (Martinón-Torres et al., 2021). The Rising Star burials are, however, younger than the mortuary behavior by early Neandertals found at Sima de los Huesos, Spain at ∼430,000 years (Sala et al., 2022; Arsuaga et al., 2014).

Some authors have argued that mortuary behavior is unlikely for H. naledi, due to its small brain size (Val, 2016). The evidence demonstrates that this complex cultural behavior was not a simple function of brain size. While we cannot at this time exclude H. naledi as part of the ancestral makeup of humans, its overall morphology suggests that its common ancestors with today’s humans and Neandertals go back a million years or more (Dembo et al., 2016; Argue et al., 2017; Caparros and Prat, 2021; Thackeray, 2015). This raises the possibility that burial or other mortuary behavior may have arisen much earlier than present evidence for them, or that such behaviors evolved convergently in minds different from our own. Understanding such behaviors will require comparative study of all hominin lineages in which they occur.

Materials and Methods

Fossil material and its location

The features and fossils described here have been found within naturally formed chambers of the Rising Star Cave System (Fig. 1), two of which are sub-chambers of the Dinaledi Subsystem known as the Hill Antechamber U.W. 107 and Dinaledi Chamber U.W. 101 (Figure S2-5) (Elliott et al., 2021). We also discuss material recovered from the Lesedi Chamber U.W. 102 (Figure S6) (Hawks et al., 2017) and a fourth locality in a remote passage of the Dinaledi Subsystem, U.W. 110 (Figure S5) (Brophy et al., 2021) in the University of the Witwatersrand’s fossil locality catalogue system (Zipfel and Berger, 2009). An augmented virtual reality flythrough of the cave is provided as Movie S1 that allows a detailed examination of the Hill antechamber and Dinaledi Subsystem.

The features described here have been excavated to varying extents ranging from partial exposure to total excavation and extraction from the cave. It is important to note that once it was recognized during excavations that some of the fossil bones were contained in discrete features that appeared to be pits dug into the floor of the cave, the decision was made to preserve a significant portion of these features for future studies. This was done in order to preserve evidence for developing additional hypotheses as well as to await the development of anticipated advances in analytical technologies. The Hill Antechamber feature is fragile and where we exposed its surface it had already experienced significant deterioration. We therefore extracted the entire feature in three parts by exposing the sediments surrounding its edges and then encasing all contents in three plaster jackets, one large and two smaller encasements. This was done so that the resulting sections would be small enough to be brought through the narrow confines of the Chute during ascent from the Hill Antechamber. This material was CT scanned and the larger block designated for synchrotron scanning. The material remains in these plaster jackets in order to protect them from the environment and allow the application of technology to examine them in situ.

Fossil material excavated prior to the 2017 discovery of Feature 1 in the Dinaledi Chamber were typically extracted entirely from their original context using methods described in Excavation Methods below or in papers describing these fossils (Berger et al., 2015; Dirks et al., 2015; Hawks et al., 2017; Brophy et al., 2021; Elliott et al., 2021). Images and notes for this work have either been published or are held by the University Curator of Collections at the University of the Witwatersrand, Johannesburg, South Africa. All sediments recovered from these excavations have been catalogued and preserved and are held on-site at the Rising Star site and are available for study.

The fossil remains described here, and from all previous excavations are available for study by application to the Fossil Access Committee of the University of the Witwatersrand through the Curator of Collections. 3D shape files of more identifiable material may be downloaded at https://Morphosource.org under a Creative Commons License by attribution.

Dinaledi Chamber excavation

The Dinaledi Chamber is the location of our excavations from 2013 to 2016. During the course of these excavations, our team collected skeletal material from across the present-day surface of the chamber and opened an excavation unit that finally reached 80 cm by 80 cm to a depth of 20 cm, and a small sondage to a total depth of 70 cm at its center (Fig. 4). This excavation yielded abundant skeletal remains of Homo naledi. The remains of at least five individuals are commingled in this excavation unit. Many of the skeletal elements were articulated at the time of excavation or fragmented and disarticulated near each other. The overall distribution of the remains within this excavation unit exhibited clear horizontal and vertical spatial clustering, calculated using a 3D density-based clustering algorithm (Kruger et al., 2016) (Fig. 13). The presence of more steeply angled fossils, such as a near vertical femur fragment, and orientations at every direction suggests skeletal materials were also buried and have not been substantially moved through post-depositional soil movement. In comparison, the surface material tends to be at shallower angles corresponding to the shallow slopes of the chamber floor (Fig. 4b). Our work to associate skeletal remains with each other to reconstruct individuals has continued (Huber, 2011). The sondage was largely sterile, but in its lowest level produced a single baboon molar crown. The excavation and sondage were central to the dating of the H. naledi-bearing deposit (Girardeau-Montaut, 2011); uranium-series-ESR combined dating on H. naledi teeth from this excavation unit provided a maximum age of 335,000 years, while the baboon tooth from a depth of 45–50 cm, around 25–30 cm below the H. naledi material, yielded US-ESR age estimates of 723 ± 181 ka and 635 ± 148 ka under two different uranium uptake scenarios (Dirks et al., 2017; Wiersma et al., 2020).

The stratigraphy noted within the 2013–2016 excavation unit and sondage was described by Dirks et al. (2015). They interpreted the brown sediment comprising the floor of the cave as one stratigraphic unit divided into two sub-units, sub-unit 3a and 3b. Sub-unit 3b is the fossil-bearing stratigraphic sub-unit present across much of the present-day floor surface, while sub-unit 3a is stratigraphically below 3b, devoid of H. naledi material. Both sub-unit 3a and 3b contain angular clay clasts of a reddish-orange color. These clasts are interpreted as deriving from sub-units 1a and 1b, which occur in some places near the cave wall and in remnants adhering to the cave walls or in dolomite pockets. The erosion of these sub-units resulted in redeposition of clasts into the Unit 3 sediments.

In November, 2018, our team initiated renewed.,/ investigation of the Dinaledi Chamber. At that time it was apparent that fossil discoveries in the Lesedi Chamber, Hill Antechamber, and distal fissure network of the Dinaledi Subsystem had each presented situations in which skeletal remains of single individuals were localized. These situations appeared to contrast with the commingled assemblage of skeletal remains in the initial Dinaledi Chamber excavation unit. We opened two new 50 cm by 50 cm excavation units within the Dinaledi Chamber, flanking the previous excavation area on the northwest corner (S950W550) and southeast corner (S1050W500). Our goals were (1) to test whether other portions of the floor of the Dinaledi Chamber have similar subsurface concentrations of H. naledi skeletal material as found in the previous excavation unit; (2) to assess whether subsurface skeletal material bears any relation to remains that were recovered from the present-day surface; and (3) to further understand the stratigraphic situation in the chamber.

The excavation of the southeast unit continued in 5 cm levels until a depth of 20 cm was reached, bringing its base into line with the original excavation unit. This yielded 58 fossil specimens, mostly fragmentary including cranial fragments and long bone fragments, although a few complete pedal elements and four teeth or teeth germs consistent with a single juvenile H. naledi individual were also found. The skeletal material in this unit was localized near the boundary with the larger excavation unit from the earlier excavations, across the complete range of depths. Elements within an anatomical region were in some cases clustered spatially, such as the three teeth from an immature individual. However, the evidence does not make clear whether all this material may represent a single individual, and also whether it may comprise one or more parts of the more complex array of remains from the original excavation unit.

The northwest unit (S950W550) presented a more definitive picture. This unit has a large rock jutting into its southwestern corner, which may be a projection of the chamber’s bedrock floor. At 10 cm below surface, long bone elements including a partial right femur, partial tibia shaft, and partial humerus shaft were exposed across the unit just east of the rock within a single horizontal level. At the same level on the northernmost part of the unit, a semi-circular concentration of skeletal material appeared, extending into the north wall of the unit. This concentration of bone was not in direct contact with the three long bone fragments.

At this stage we elected to open an additional 50 cm unit to the north of this one, to investigate the extent of the bone concentration. In level 2 of this unit (S900W550), between 5 and 10 cm in depth, poorly preserved bone material was exposed including a left demi-mandible with premolars and molars, a partial tibia shaft, and other long bone fragments.

Below this, we encountered a northward continuation of the bone concentration from the adjoining unit. We exposed the horizontal extent of this feature (Dinaledi Feature 1) at the same level as in the adjacent S950W550 unit. During this phase of excavation, we collected bone fragments and elements that excavation exposed; these specimens are listed in SI 2.1.

After the pause in excavation, we returned to the Dinaledi Chamber to sample sediment within and around Feature 1. Our aim was to better characterize the local sediment profile and provide material for analysis of the sediment composition. We cleaned a profile at the south end of the feature, without disrupting any of the remaining skeletal material.

At this time, we also worked to identify skeletal material exposed on the surface of the feature. Many of the remaining elements are partially obscured by sediment and we have not removed sediment for the purpose of identification. At the southern end of the exposed feature are cranial vault fragments consistent with a single partially crushed cranium. Vault fragments at the edge of the feature are in angled positions supported by underlying matrix. Near this and extending into the center of the feature are vertebral remains, including two vertebral bodies in articulation. A partial ribcage is present at the west center of the feature, with rib elements remaining in situ in a vertical or near-vertical position supported by matrix. Toward the east within the feature are visible long bone shafts that correspond in size to femur or tibia. A partial fibula shaft is also exposed in this part of the feature. Toward the north of these hindlimb fragments are identifiable portions of the right mandibular body, including at least four premolar and molar teeth in place (Fig. 9).

We compared the remaining in situ identifiable material to the skeletal material excavated from the feature to evaluate their compatibility, finding that all material is compatible with a single adult individual with the exception of three fragments of immature elements. Most of the identifiable elements within Feature 1 were positioned in ways that follow anatomical logic for a whole body with arms and legs flexed into a compact orientation. Large fragments of cranial vault are localized near the south end of the feature and are matrix-supported. Above this fragmented vault were upper limb elements. Lower limb elements remain in the central portion of the feature, below and to the east of a partial ribcage with matrix-supported rib fragments. The fragments of the mandible are separated from the position of the vault, with the fragments of right mandibular corpus and associated teeth in roughly the center of the feature and a fragment of left mandibular corpus above and separated from the concentration of bone, near its north end. One zygomatic fragment and small (<20mm) fragments of vault are also in this area. Without fully excavating the feature, it is difficult to test hypotheses about the original position and subsequent decompositional collapse of the skeleton. We can tentatively suggest that the body was in a tightly flexed position with the head higher toward the south end of the feature and arms flexed near the face. From this position, the mandible disarticulated and collapsed into the center of the thorax. It is possible that the further fragmentation of the mandible may reflect disturbance of the body after decomposition. It is also possible that small fragments may have been part of the sediment load above the burial and pertain to other individuals.

Hill Antechamber excavation

The Hill Antechamber U.W. 107 is the portion of the Dinaledi Subsystem nearest to the present-day entrance access into the subsystem (Elliott et al., 2021). This space is accessible today via the ca. 12 m vertical passage known as the Chute, which descends into its northern edge. There are two exit avenues, both narrow (< 0.5 m) passageways that lead in parallel to the Dinaledi Chamber; there are also small (< 0.2 m) fissure passages exiting this chamber to the west and north. The floor of the Hill Antechamber forms a ca. 30 degree slope with the highest point near the Chute and lowest point at the southwest edge of the chamber (Figure S3). Our team recovered one tooth and a small number of bone fragments from the floor surface of the Hill Antechamber in 2013; these are included in the sample published in Berger et al. (2015).

In 2017 the team began work in the Hill Antechamber with the goal of understanding the interaction of Homo naledi with the Chute. At that point we had collected more than 1500 fossil specimens of Homo naledi from the Dinaledi Subsystem, and all but a handful of these came from the Dinaledi Chamber. The largest concentration of bone at that point had been excavated in a single 80 cm square unit within the Dinaledi Chamber, approximately 12 meters from the Chute (Figure S14). The closest point of the Dinaledi Chamber to the Chute is ca.8 meters, and at least 5 meters of that distance is through one of the narrow (< 0.5 m) passages. We considered the hypothesis that Homo naledi skeletal material may have been deposited into the Chute, either fleshed or unfleshed, accumulating at the base of the Chute in sufficient quantity for gravity to bring portions into the Dinaledi Chamber. Under this hypothesis, we predicted that the sloping floor of the Hill Antechamber should correspond to a talus deposit including Homo naledi body parts and possibly other material including artifacts. The distance and physical constraints of this part of the subsystem made it hard to understand how a substantial concentration of Homo naledi skeletal material, including articulated limbs, could occur in the subsurface of the Dinaledi Chamber at such a distance from the Chute. We also considered the hypothesis that living individuals of Homo naledi had entered the subsystem via the Chute. This hypothesis would help to explain the concentration of material within the Dinaledi Chamber but did not make any clear prediction about the subsurface content of the Hill Antechamber.

In September 2017, we established a datum for the subsystem near the northeast extreme of the Hill Antechamber and established a grid based on a local north extended throughout the subsystem (Figure S2). This local north diverges from magnetic north by 35 degrees NW; maps here are presented relative to magnetic north. The grid system was projected into the spaces with laser level and laser rangefinding equipment. The spatial location of each fossil fragment was recorded from measurements prior to recovery, with scaled photographs taken in situ. Photography for photogrammetry was carried out approximately daily and sometimes at shorter intervals. Metashape 1.8.1 Standard (Agisoft) was used to generate 3D models of the excavation from these series of photographs. The resulting 3D models were correlated with 3D point cloud data generated by a Faro laser instrument.

We initiated fieldwork in the Hill Antechamber with a surface search, which identified additional small fossil fragments. Within the Antechamber, the team opened three 50 cm excavation units. One of these at grid unit N100W50 was a half unit against the north wall of the Antechamber. This unit produced only a small number of unidentifiable fossil fragments. A second unit, at N50W100 was the nearest full 50 cm square to the base of the Chute. It is this second unit that produced the Hill Antechamber feature. The third unit, S150W150, was lower on the slope near the opening to the parallel narrow passages that lead to the Dinaledi Chamber. This area was covered with flat rocks (Figure S15). After documenting and lifting of the rocks, excavation in this unit revealed some skeletal material, including a partial maxillary dentition of an adult individual of H. naledi and fragments of cranial and postcranial skeletal material, including some attributable to tibiae or femora. It is unclear if the rocks are associated with the presence of the underlying skeletal material. In addition to these three excavation units, the team investigated a small pocket of sediment on the north wall of the chamber, and excavated the thin layer of sediment upon the flowstone shelf at the base of the Chute. No skeletal material came from these two areas.

Excavation of the Hill Antechamber units followed methodology previously described for work in the Rising Star cave system (Berger et al., 2015; Hawks et al., 2017). No metal instruments were used for excavation. On the highly sloped units in the Hill Antechamber we proceeded with a terraced approach to bring the excavation surface down in horizontal steps of 5 cm. During the excavation of the unit at N50W100, we encountered a substantial area of fragmented and powdery bone material beginning at a depth 5 cm below surface, near the north end of the unit (Figure S16). Leaving this material in place, we continued to excavate within the designated unit to reveal its extent. As we exposed this material further toward the south, we found that it included some more complete fossil elements and extended across more than half the horizontal area of the 50 cm × 50 cm unit. At this stage we decided to delineate the edge of this feature and excavate more deeply around it to form a pedestal that could be removed en bloc from the cave system for further work in the laboratory (Figure S17). The material extended to the east and south edges of the N50W100 square, and so we opened half of the neighboring grid squares at N50W50 and S50W100, and the northwest quadrant of S50W50 to fully expose the boundaries of the feature. By the end of excavation in March 2018, an area of 75 cm × 75 cm was excavated down to a horizontal level 40.5 cm below datum, leaving the bone-containing feature as a pedestal. Only a small number of fossil elements and fragments occurred within the surrounding excavated area.

All material and people entering and exiting the Dinaledi Subsystem must pass through the constraints of the Chute, which places a severe size limit on any excavated blocks of material. The overall size of the Hill Antechamber feature was incompatible with removal from the subsystem in a single block. We therefore planned for a segmentation of the feature into a minimal number of pieces. The contour of the surface of the feature presented an upper surface dense with bone elements and bone material through most of the western two-thirds, while the eastern third of the feature lacked any bone at this level and presented deeper, less uniform bone material (Figure S18). This suggested a division of the feature to separate the eastern portion would be most appropriate. We applied a plaster jacket to the east wall of the feature and allowed it to solidify. We then excavated into the feature gradually to separate this portion from the overall mass, using toothpicks to confine its lower and western boundaries. Once separated, this block (U.W. 101-2074) measured 9 by 23 by 9 cm. We followed a similar approach to remove a second somewhat larger block that partially underlay the first.

The resulting block, U.W. 101-2075, measured 21 by 26 by 11.5 cm. Finally, we jacketed the top and sides of the remaining mass of the feature (Figure S19 jacketed block). We used two thin metal sheets to separate the feature from the underlying sediment, then inverted it, applied a rigid plastic sheet to the bottom surface, and covered this bottom with plaster. The resulting mass, designated as U.W. 101-2076*, measured 40 cm by 49 cm by 19 cm at its maximum dimensions. We wrapped all plaster jacketed masses in bubble wrap and additional padding, placing them in waterproof cave haul bags. The team brought them out of the cave system without incident (Figure S20).

In the lab we prioritized the plaster-jacketed masses for imaging. The smaller jacketed masses, U.W. 101-2074 and U.W. 101-2075, were suitable for imaging with the Nikon Metrology XTH 225/320 microtomography (microCT) scanner housed at the Evolutionary Studies Institute of the University of the Witwatersrand. These were scanned at a voxel size of 72 microns. The larger jacketed mass, U.W. 101-2076, was not suitable for scanning with this instrument, and so we obtained a clinical CT from a Siemens Somatom Definition AS 40 (Erlangen, Germany) with larger gantry size in the Charlotte Maxeke Johannesburg Academic Hospital (CMJAH), with a 0.5mm voxel size. We segmented these models with a combination of manual and density threshold segmentation strategies.

A possible stone artefact was recognized on the clinical CT (Figure S21; SI 3.2). It was then scanned as a focused target at the European Synchrotron Radiation Facility (ESRF) in Grenoble France. This scan was performed using propagation phase-contrast on the beamline BM18, using an isotropic voxel size of 16.22um and a propagation distance of 6m. The average detected energy was 156 keV, obtained by filtering the white beam of the BM18 tripole wiggler with 2.62mm of molybdenum. We used an indirect X-ray detector based on a LuAG:Ce scintillator of 2mm thickness, coupled to a PCO edge 4.2 CLHS camera through a Hasselblad 120mm macro objective.

Analytical and computational methods used for studying sediments

We used x-ray diffraction (XRD) and x-ray fluorescence (XRF), respectively, to study the mineralogy and chemistry of the samples. To study the textures, we used scanning electron microscopy (SEM) on the sediments mounted in epoxy resin to create polished blocks. On the SEM, backscattered electron (BSE) imaging was used to observe the textures of the sediments and energy dispersive spectroscopy (EDS) was used for semi-quantitative chemical mapping and spot analysis of the observed mineral grains. The methodologies, software, and measurement parameters/conditions for XRD, XRF and SEM were as used for Dirks et al. (2015) at the Spectrum Analytical Facility of the University of Johannesburg (UJ). The XRF results were treated with Principal Component Analysis (PCA) computed using the MATLAB Statistics and Machine Learning Toolbox PCA function pca on 12 major elements (Al, Ba, Ca, Fe, K, Mg, Mn, Na, P, S, Si, Ti). PCA is the standard method for unmixing mixed variables. The output of the function pca includes the principal component loads, the scores, and the variances. The data correlations were implemented using the function corrplot which uses Pearson’s correlation coefficient. The particle-size distribution (PSD) was carried out at the Department of Metallurgy at UJ using the Microtrac S3500 laser-diffraction particle size analyzer. The obtained results were reduced using GRADISTAT, the grain size and statistics package (Blott and Pye, 2001), and selecting the Folk and Ward Method to obtain the mean grain size, sorting, skewness and kurtosis in addition to the percentages of clay, silt and sand in each sample.

Acknowledgements

Permits to conduct research in the Rising Star Cave system are provided by the South African National Research Foundation (LRB). Permission to work in the Rising Star cave is given by the LRB Foundation for Research and Exploration. We would like to thank the University of the Witwatersrand’s Primate Fossil Access Committee for allowing access to the original material for study and the University Curator and Assistant Curator of Collections for assisting with logistics in studying the material. We would like to thank the Department of Diagnostic Radiology at the Charlotte Maxeke Academic Hospital for allowing access to their clinical CT scanner. We acknowledge the European Synchotron Research Facility for providing access to facilities including the BM18 beamline. We would like to thank Matthew Caruana for discussions related to the lithic associated with the Hill Antechamber burial. The Authors would like to acknowledge the funders of the various expeditions that recovered the fossil material and information, including 3d scanning and the production of AVR material described in this paper including the National Geographic Society (LRB, CJ, KB), the Lyda Hill Foundation (LRB) and the National Research Foundation of South Africa (LRB). Laboratory work and travel was funded by the National Geographic Society (LRB, ME, AK, CJ, KB), the Lyda Hill Foundation (LRB, ME, AK) the Centre for Excellence in PalaeoSciences at the University of the Witwatersrand (now GENUS) (AK), the Fulbright Scholar Program (JH), and National Research Foundation of South Africa (TVM), University of Johannesburg Research Council (TVM), Spectrum Analytical Facility at University of Johannesburg (TVM).

Author and article information

Author contributions

Conceptualization: LRB, JH

Methodology: LRB, JH, TM, ZJ, ME, PRQ, BC, GC, KJ, CJ, KB, AK, CW

Investigation: LRB, TM, KM, AK, PRQ, ME, BP, AF, PT, VB, KD, ZJ, ABG, JB, GC, PHMD, EF, AG, NH, LH, RH, KJ, HM, EP, MR, ER, JSS, MT, ST, DVR, KW, MK, PS, JH

Visualization: LRB, TM, KM, AK, PRQ, ME, BP, PT, VB, KD, ZJ, ABG, KB, GC, EF, AG, NH, LH, RH, KJ, CJ, MR, HM, JSS, MT, ST, DVR, KW, CW, JH

Funding acquisition: LRB, ME, AK, CJ, KB, TVM

Project administration: LRB

Writing – original draft: JH, LRB

Writing – review & editing: LRB, TM, KM, AK, PRQ, ME, BP, AF, PT, VB, KD, ABG, KB, JB, GC, PHMD, EF, AG, NH, LH, RH, KJ, CJ, HM, EP, MR, ER, JSS, MT, ST, DVR, KW, CW, MK, PS, JH

Competing interests

The authors declare that they have no competing interests with the production or publication of this research.

Provenience

All fossil and archaeological material studied in this work was recovered by the authors. Details of their context are presented within this paper. The materials are curated at the University of the Witwatersrand. All materials and scans are available for study through application to the Primate Fossil Access Committee of the University of the Witwatersrand. 3D shape files and images of HAA1 are available via Creative Commons License by Attribution on https://Morphosource.org. Movies and Augmented Virtual Reality data are available for download via Figshare.

Additional files

Supplementary Movies

SM1. Augmented Virtual Reality of the Hill Antechamber and the Dinaledi Chamber

SM2. HAA1 visualization movie

SM3. Hill Antechamber Block movie

Supplementary Information (SI) file contents

Supplementary Information 1: Petrography and geochemistry of sediments

Supplementary Information 1.1: Unlithified Mud Clast Breccia (UMCB) in passages of the Dinaledi Subsystem compared to UMCB on the Dinaledi floor

Supplementary Information 1.2. Grain size analysis of sediments around Feature 1 and the Lesedi Chamber

Supplementary Information 2: Hominin skeletal material and element representation

Supplementary Information 2.1: Identification and assessment of skeletal remains from Dinaledi Feature 1

Specimen list

Supplementary Information 2.2. Identification and assessment of skeletal remains from Dinaledi Feature 2

Supplementary Information 3: Hill Antechamber Feature and Artefact

Supplementary Information 3.1. Stratigraphic situation of the Hill Antechamber feature

Supplementary Information 3.2: Description of the Hill Antechamber Artefact 1 associated with the Hill Antechamber

Feature

Supplementary Information 4: Materials and Methods

Supplementary Information 4.1: Previous Excavations in the Dinaledi Chamber Supplementary Information

4.2. Previous work in other parts of the cave system

Supplementary Table 1.

Supplementary Table 2.

Supplementary Figures

Supplementary Information 1: Petrography and geochemistry of sediments

Supplementary Information 1.1: Unlithified Mud Clast Breccia (UMCB) in passages of the Dinaledi Subsystem compared to UMCB on the Dinaledi floor

The petrography, specifically the mineralogy, chemistry and textures, of the Homo naledi-bearing unlithified mud clast breccia (UMCB) on the Dinaledi Chamber (UW101) floor was reported by Dirks et al. (2015), Makhubela et al. (2017) and Wiersma et al. (2020). Here we supplemented those results by carrying out particle-size distribution (PSD) and x-ray fluorescence (XRF) of 57 samples of the UMCB sediments collected during excavations (DF group in Supplementary Tables 1 and 1), and 22 in situ samples of the UMCB that is still spatially around Dinaledi Feature 1 on the Dinaledi floor (see Figure S1 and SI 4.2). The DF group samples were subsampled from bags of sediments above the burial Features 1 and 2, after sieving for recovery of fossils not recognized at the time of excavation. Each bag of sediment corresponds to a particular area and depth level of the gridded excavation pit (Figure S2). Further, we report the mineralogy, textures and geochemistry of 16 sediment samples from four fossil-bearing narrow passages of the Dinaledi Subsystem formed from W-NW, N and NW trending fracture systems (Elliott et al., 2021). These passages, namely: UW108, UW109, UW110 and UW111, are isolated locations and extremely difficult to access, and here we compare their sediments with the UMCB on the Dinaledi floor (Figure S3). The mineralogy and chemistry were studied using x-ray diffraction (XRD) and XRF, respectively. To study the textures, we used scanning electron microscopy (SEM). The petrography of the sediments in the other fossil-bearing chambers of the Dinaledi Subsystem, namely Hill Antechamber and Chaos Chamber, have not yet been studied. But for the Lesedi Chamber we have done XRF and PSD on 3 samples of UMCB sediments on a ledge and 5 samples of UMCB sediments from the floor.

The UMCB of the Dinaledi floor is dark and uncalcified, but it is heavily cemented and does not disintegrate easily because of the very high MnO and Fe₂O₃ content, which can be more than 10 wt.% for each of these two oxides (Supplementary Table 1) (Dirks et al., 2015). The bulk geochemistry of the 57 Dinaledi floor UMCB analyzed here is consistent with the three samples analyzed by (Dirks et al., 2015). The average MnO and Fe₂O₃ contents are 4.36 wt.% and 9.88 wt.%, respectively. A sample (2280 on Supplementary Table 1) of UMCB shows the pervasive nature and abundance of MnO and Fe₂O₃ in different forms of Mn- and Fe-oxides and -hydroxides (Mn-Fe-oxihydroxide) concretionary infilling or impregnation and coatings (Figure S4). The CaO content of the Dinaledi UMCB ranges from 0.33 to 6.90 wt.%, and the element map of CaO shows that there is little CaO compared to MnO and Fe₂O₃. The average CaO content is 1.41 wt.% and it is lower than the averages of the samples from the Lesedi Chamber and the four passages of the Dinaledi Subsystem, which have 4.49 wt.% and 5.18 wt.%, respectively (see below and Supplementary Table 1).

The UW108 passage is 20-35 m wide and 5-6 m long (Figure S3), but has three varied UMCB sediments on the floor and unaltered LORM sediments on a ledge about 1.5 m above the floor. The first type of UMCB (L01) is identical to the Dinaledi floor UMCB, but shows less MnO and Fe₂O₃ content under the SEM (Figure S5). The second type of UMCB (L02) in this passage has abundant Mn-oxihydroxide occurring as discrete grains and as infilling and coating on mud clasts (Figure S6). In addition, this UMCB contains abundant fragmentary grains of dolomite (Figure S6). In some aspects, L01 and L02 are visually different under the SEM, but chemically similar on XRF bulk chemistry (Supplementary Table 1). For example, L01 shows a high Fe₂O₃ content in the mud clasts while L02 does not, but they both have similar Fe₂O₃ content of 9.96 wt.% and 9.14 wt.%, respectively (Supplementary Table 1). The differences in the dolomite fragments can also be seen chemically where L01 has CaO and MgO of 0.62 wt.% and 2.36 wt.%, respectively, but L02 CaO and MgO of 6.10 wt.% and 7.03 wt.%, respectively (Supplementary Table 1). The third type of UMCB (L03) is a highly calcified version of L02 due to a thin (<5 mm) flowstone that partially covers the floor (Figure S7) (Elliott et al., 2021). L03 contains less mud clasts than dolomite fragments cemented by calcite, and this can also be seen chemically where MnO and Fe₂O₃ content are low (2.15 wt.% and 4.14 wt.%, respectively) but CaO and MgO are much higher (25.24 wt.% and 9.70 wt.%, respectively). The unaltered LORM mudstone from the ledge has a very fine-grained texture of clay particles that contain abundant K, Mg and Fe content but does not contain any observable Mn and Ca content under the SEM (Figure S8) (Dirks et al., 2015). The bulk chemistry of L04 is consistent with the textures (Supplementary Table 1).

The UW110 locality is parallel to UW109 in an E-W trending fracture that connects to the Chaos Chamber and is the locality where 28 cranial fragments of a juvenile Homo naledi were found (Figure S3) (Elliott et al., 2021, Brophy et al., 2021). At UW110 there are five observable chert horizons that are broken up and form ledges along the walls. On the floor and the ledge of the first chert unit about 0.7 m from the floor there is UMCB sediments that contain abundant metallic luster. PF01 is from the floor and contains less metallic luster than PF02 from the ledges, and this is also reflected in the bulk chemistry of these samples (Supplementary Table 1). PF02 has the highest combination of MnO and Fe₂O₃ content of all samples analyzed, but texturally does not look different from the other UMCB sediments from the other passages and chambers (Figure S9; Supplementary Table 1). A lot of the Fe and Mn in this sample seems to be within the clay fraction of the mud clasts and not as Mn-Fe-oxihydroxide (Figure S9). On the second ledge about 1.5 m from the floor the sediments are a muddy sandstone (PFL01) unique to this area alone (Figure S10). This unit composed of mostly quartz sand with very little mud clasts, Fe and Mn (Figure S10). Interestingly, it contains abundant gypsum grains (CaSO₄) not seen in any other samples, making PFL01 to have a very high SO₃ content of 6.86 wt.% while in other samples it is less than 1 wt.% or undetectable (Figure S10).

The UW111 locality is almost 16 m SW of the Dinaledi Chamber where a NE-SW fracture terminates at intersection with a N-S fracture (Figure S3) (Elliott et al., 2021). In the N-S fracture passage we use to access UW111, the sediments are unaltered LORM mudstones (sample HAP1) similar to those of L04 from UW108 and they contain considerable very-fine quartz sand (Figure S11). There is no observable Mn under the SEM and in the bulk chemistry it is low at 0.95 wt.%, but Fe is noticeable in the mud clasts and is represented by 8.67 wt.% Fe₂O₃ content (Supplementary Table 1). On the floor of the UW111 locality is UMCB (HD01-03) that is partially covered by a thin flowstone like in UW108. Only HD03 has considerable CaO content of 16.79 wt.% but under the SEM it does not look calcified as the Ca seems to be in discrete grains of dolomite, and this is consistent with MgO content of 12.66 wt.% (Figure S12). The mud clasts have abundant Fe and Mn incorporated into their very fine-grained nature and not as Mn-Fe-oxihydroxide infilling (Figure S12).

There are no characteristic differences with all the major elements and loss on ignition (LOI) in terms of concentration range and averages in the geochemistry of the sediments sampled from inside Feature 1, the sterile area around Feature 1, and the sediments removed above Feature 1 (Supplementary Table 1). For example, even an element such as P₂O₅ closely linked with the fossil bone shows similar averages of 0.36 wt.% and 0.38 wt.% for the sterile sediments (SA group) around Feature 1 and the sediments removed above the Feature 1 (DF group), respectively. These two groups of samples have higher average P₂O₅ content than the sediments inside Feature 1 (SB group), between features 1 and 2 (SC group) and those from the vertical profile (SE group), which have 0.23 wt.%, 0.17 wt.% and 0.16 wt.%, respectively (Supplementary Table 1).

Supplementary Information 1.2. Grain size analysis of sediments around Feature 1 and the Lesedi Chamber

The sediments in the Dinaledi and Lesedi Chambers have similar mean grain-size centered around 350 µm and variable between 200 µm and 600 µm (Figure S13, Supplementary Table 2). The deviations from 350 µm seem to be related to the percentage sand or silt proportion of the samples, which may be linked to the proportion of the LORM mud clasts vs silt matrix of the UMCB (Supplementary Table 2). The SA and SE groups were sampled in situ at various depths and so their mean grain sizes show patterns of upward fining sequences (Figure S13). The DF group of samples are also from various depths of 0 to 15 cm, but the sediments of each sample were mixed during sieving. They show a mixture of upward and downward fining sequences.

Supplementary Information 2: Hominin skeletal material and element representation

In this section we present a preliminary identification of skeletal elements within the burial features, with an assessment of skeletal part representation. We note the presence of articulated or spatially contiguous elements where this has been observed.

In both the Dinaledi and Hill Antechamber, we have left much of the evidence described here in situ or en bloc without unnecessary destructive excavation or preparation. The skeletal material that has been excavated from Dinaledi Features 1 and 2 has not been subjected to solvents or consolidants at this time. The Hill Antechamber feature rests within three plaster-jacketed blocks at the present time. These decisions limit the number of observations that we can gather on the skeletal remains from these features. Here it is worth a brief comment to clarify these protocols and discuss their relevance to the study of the burials.

Archaeological controlled excavation is a fundamentally destructive process (17), one in which the spatial relationships between buried objects (including hominin fossils) is lost once the matrix surrounding such remains is disturbed and the buried objects lifted. Spatial relationships are necessary to determine temporal relationships in archaeological settings; they are also fundamental to understanding the processes of alteration that assemblages may have undergone after burial. Article 5 of the International Charter for Archaeological Heritage Management states that archaeological investigations can be carried out using a wide array of methods from non-destructive remote sensing, through sampling, to total excavation. It is widely recognized that total excavation is a recourse that should be adopted after consideration of other less destructive means of study (18).

3D modeling or non-invasive imaging is the preferred method for recording and understanding fragile, friable buried contexts where the precise spatial relationships between clastic components is considered more important rather than the necessity to view the surface morphology of such remainsi. Such a spatially-based approach has been used and advocated in recording of mainstream archaeology (19), cremains (20), bone taphonomy (21, 22), 4D documentation of pit burials (23), virtual autopsy (24), recording fragile funerary goods and human remains (25, 26), through to whole burial chambers (27), where information derived from the spatial relationships within the burial environment outweighs that from direct analysis of the bones or artefacts themselves (28).

In the Rising Star cave system we have employed study protocols that minimize the extent of excavation areas and maximize the collection of spatial data by multiple modalities (1, 5, 9, 11). The total excavation surface in the Dinaledi Chamber to date is 1.55 m² and in the Hill Antechamber is 0.75 m². Homo naledi skeletal and dental material has come from these excavated areas and from collection of skeletal remains from the floor surfaces of these chambers. That material now numbers more than 2500 pieces, many of them small fragments but including a large number of complete and well-preserved elements. Nonetheless much of this skeletal material is highly fragile with loss of elements with thin cortical bone or substantially trabecular bone content. Our excavations in 2013 and 2014 in the Dinaledi Chamber identified buried skeletal material with articulated limb, cranial, and vertebral elements (5, 6, 11) including portions that could be attributed based on spatial and developmental data to a partial juvenile skeleton (8). This extensive evidence of skeletal morphology and spatial positioning of material has informed our more recent work in the system in a variety of ways (1).

In many paleoanthropological contexts, the recovery of morphological or developmental observations on hominin skeletal material is a very high research priority. Additionally, some settings pose taphonomic questions that can be addressed only through collection of data from bone surfaces. To recover such data, hominin skeletal material has often been subjected to intensive preparation, cleaning, and consolidation. Much Homo naledi skeletal material from our earlier work has been treated with conventional methods of skeletal conservation and analysis, which has included microscopic examination of bone surfaces (5). Having this published evidence in hand gives us the possibility of taking a less destructive approach to studying the burial features that we identified in subsequent work. For these reasons, once we identified a high probability that the Dinaledi Feature 1 and Hill Antechamber Feature were burial contexts, we planned for non-destructive methods to record and preserve the spatial relationships between any remains entombed within these features. In the Hill Antechamber case, this involved lifting the feature en bloc for study, scanning, and possible preparation in the laboratory. In the case of Dinaledi Feature 1, the prior recovery of a similar burial feature from the Hill Antechamber argued for leaving this feature in situ within the cave system and recording data on its spatial and sedimentary context noninvasively.

Supplementary Information 2.1: Identification and assessment of skeletal remains from Dinaledi Feature 1

Here we list brief anatomical identification of all catalogued skeletal elements and fragments excavated within and above the area of Feature 1. These specimens all come from grid units S950W550 and S900W550. Some other material excavated from unit S900W550 is associated with Feature 2 and these are listed separately below.

These brief identifications do not focus on comparative morphology beyond the identification to element where possible. Many of the remains are small fragments that are not identifiable to element and these are included here to present a complete record. In the few cases where diagnostic morphology of H. naledi is present, this is indicated. Where evidence for developmental stage is present, this is indicated.

We have not fully cleaned sediment from this skeletal material or applied fixatives or other chemicals, in anticipation of possible analysis of biochemicals or trace evidence. This precludes a full examination of possible surface modifications or other surface taphonomy.

Nearly all elements identified in this list are compatible with belonging to a single adult individual. All identifiable adult elements are size-compatible; they do not appear to represent a mixture of individuals of different sizes. None of the adult elements duplicate each other; nor is there any duplication with identifiable elements that remain in situ within Feature 1 in the Dinaledi Chamber.

There are three notable exceptions that are immature elements and inconsistent with the single adult represented by the rest of this material. Two of these are fragments of juvenile femur, including the right proximal femur fragment included in U.W. 101-2250 and a fragment of femur shaft U.W. 101-2260. These two elements were excavated overlying and in physical contact with the bone concentration of Feature 1, both within 5 cm of each other near the center of the feature. The proximal right immature femur duplicates the proximal right adult femur fragment, also included within U.W. 101-2250. These two elements were excavated in immediate contact with each other with the immature femur fragment overlying the adult fragment. The other inconsistent element is an immature proximal left humerus fragment, U.W. 101-2243. This element duplicates another proximal left humerus within Feature 1, U.W. 101-2231. These two elements contrast in developmental status, with U.W. 101-2243 having evidence of an unfused proximal epiphysis and U.W. 101-2231 compatible with adult status. The two elements are also in different situations relative to the feature: U.W. 101-2231 is near other compatible humerus and upper limb material at the south end of the feature, while U.W. 101-2243 was out of anatomical placement overlying the uppermost part of the north side of the feature. We interpret these three fragments of immature elements as bone fragments that were on the surface or within pre-existing deposit that comprised the sediment fill of Feature 1.

Specimen list

U.W. 101-2114

Thin plate-like bone fragment, broken on all edges, less than 15 mm.

U.W. 101-2115

Two cranial bone fragments. One is possibly zygomatic or root of zygomatic arch on temporal, length 23.8. The other is possibly zygomatic at orbit, or frontal above orbit, bone surface is concave and slightly wavy. Preserved size 16.5 × 16.5.

U.W. 101-2116

Phalanx fragment with adhering clump of sediment. Fragment size is 18.9 × 9.0

U.W. 101-2117

Right proximal radius fragment. The head is present but broken around the visible edge, and one side is still obscured by sediment. The bone is very comparable in size with U.W. 101-2246 and they may be antimeres. 38.8 mm length

U.W. 101-2118

Two bone fragments less than 15 mm

U.W. 101-2119

Bone fragment less than 15 mm

U.W. 101-2120

Two cranial bone fragments less than 20 mm

U.W. 101-2121

Right proximal radius fragment, representing nearly the entire circumference of the bone. The head is not present. The distal break on this fragment is consistent with the break at one end of U.W. 101-2240 and may be the same bone. Preserved length 49.6. Shaft diameter 10.0 × 9.4.

U.W. 101-2136

Cranial or mandibular fragment, possibly zygomatic arch. Length 20.4

U.W. 101-2137

Three large clumps of sediment with spongy bone fragments embedded within them. Possibly distal femur or pelvic.

U.W. 101-2138

Sediment clump with embedded bone fragments. ca. 15 mm

U.W. 101-2139

Sediment clump with embedded fragments of cortical bone. ca. 15 mm

U.W. 101-2140

Tooth root. 11.5 mm

U.W. 101-2141

Tooth root. 13.2 mm.

U.W. 101-2142

Cortical bone fragment, 16.0 mm.

U.W. 101-2143

Three bone fragments less than 15 mm.

U.W. 101-2144

Mandibular incisor. Probably left I1, possibly I2. Occlusal wear is extensive and remaining crown height above cervix on distal face is only 2.2 mm. The wear is at an angle of approximately 60 degrees compared to labial face. The wear has exposed dentin with MD 2.0 mm, LL 1.3 mm. LL breadth 5.9, MD 4.6. Root is nearly complete with small section of distal tip missing due to fracture.

U.W. 101-2145

Bone or dentin fragments, less than 5 mm, with some sediment.

U.W. 101-2146

Bone fragments in sediment clump, less than 10 mm.

U.W. 101-2147

Bone fragment with large cancellous structure in sediment clump. Less than 20 mm.

U.W 101-2148

Two bone fragments with sediment adhering. Smaller fragment 13.0, larger 21.3.

U.W. 101-2168

Sediment with bone fragments less than 5 mm.

U.W. 101-2170

Tibia shaft portion, 121 mm in length, represents midshaft. Consistent with adult. Diameter at point where posterior shaft flattens is 21.4 × 17.2. Other small shaft fragments, all less than 20 mm.

U.W. 101-2171

Long bone shaft fragment, 34.3 × 10.4

U.W. 101-2172

Fourteen long bone shaft fragments. Most represent a part of the bone circumference and are between 30 and 40 mm in length, 10-15 in width. Additional fragments in bag.

U.W. 101-2173

Two bone fragments less than 15 mm.

U.W. 101-2174

Mandibular fragments. Two tooth crowns are present with their roots, and three additional root fragments. The two crowns are both left mandibular molars. One is markedly more worn, with large dentin pools at protoconid and endoconid, and metaconid is entirely fragmented away. The rest of the enamel rim has large fragments missing. BL 11.1 MD 12.1. This tooth has a distal interproximal facet, I interpret it as an M2. The other tooth is worn approximately flat with dentin exposed on mesial four cusps. BL 10.8 MD 12.1. This tooth is more triangular in shape distally and does not have an interproximal facet distally; I interpret it as an M3. One of the broken root fragments is a molar root, the others appear likely to be premolar roots. The morphology and dimensions of these teeth are compatible with H. naledi.

U.W. 101-2175

Lower left fourth premolar. Root is largely intact, broken at tip. Single root. Enamel is worn flat with large dentin pools for both buccal and lingual cusps. BL 10.7, MD 9.2. The morphology and size of this tooth are compatible with H. naledi.

U.W. 101-2176

Lower left third premolar. Two roots, both intact. BL 10.6, MD 9.4. Crown is worn flat, dentin exposed at protoconid and metaconid. This tooth possesses diagnostic anatomy for H. naledi.

U.W. 101-2177

Ulna shaft portion, 48 mm.

U.W. 101-2178

Two fragments. One is fibula shaft, 37.5 mm. The other is a flat fragment less than 20 mm.

U.W. 101-2179.

Bone fragment, possibly tarsal, maybe medial cuneiform fragment. Two additional pieces are bone fragments less than 10 mm.

U.W. 101-2180

Distal ulna shaft fragment, 44.6 mm.

U.W. 101-2181

Cranial fragment consistent with right zygomatic bone including orbital border. 34.8 × 24.7 mm.

U.W. 101-2182

Fibula or metacarpal shaft portion, 27 mm

U.W. 101-2183

Radius or humerus shaft portion 43.6 mm. Additional fragments of shaft.

U.W. 101-2184

Base of right mandibular corpus. 52.5 mm.

U.W. 101-2185

Enamel fragment, 5 mm.

U.W. 101-2186

Enamel fragment, corresponding to rim of molar, 8.9 mm

U.W. 101-2187

Tooth root fragment, 10.1 mm

U.W. 101-2188

Cranial vault fragment. One side is obscured by sediment. 21.4 × 15.4 mm, 5.4 thick.

U.W. 101-2189

Mesial part of mandibular molar, with mesial root complete. Very little enamel on this crown. This looks like a mirror image of the U.W. 101-2174 M3 mesial root. Several additional cranial or mandibular fragments, mostly less than 20 mm.

U.W. 101-2190

Fragment of humerus head, with small slice of surface 22.0 × 8.3 mm.

U.W. 101-2191

Partial phalanx, probably manual intermediate phalanx. Proximal articular surface is eroded but present, shaft 16.5. Embedded in sediment chunk.

U.W. 101-2199

Phalanx shaft portion, or possibly fibula shaft portion. 23.5 mm.

U.W. 101-2200

Fibula shaft or metatarsal shaft portion, triangular cross section. 22.6 mm

U.W. 101-2201

Three bone fragments, with some articular surfaces.

U.W. 101-2202

Bone fragment, shaped like possibly zygomatic or mandibular fragment, seems too thick to be vertebral or rib. 22.7 mm length.

U.W. 101-2203

Shaft fragment consistent with metacarpal. Length 28.5, diameter 5.9 × 5.2.

U.W. 101-2204

Bone fragment. 25.8 long, 14.8 broad, 6.8 thick.

U.W. 101-2205

Long bone shaft fragment, based on size and curvature consistent with humerus, representing approximately 40% of the circumference. Length 43.5, width 13.3.

U.W. 101-2206

Bone fragment, one side exposed trabeculae, the other rough cortical surface. Possibly distal femur or ilium. 24.5 × 17.9

U.W. 101-2207

Metatarsal or manual phalanx shaft fragment. 23.6 long, 7.9 × 5.3 near base.

U.W. 101-2208

Long bone shaft fragment, based on size and curvature probably humerus, representing 30% of the circumference. 25.4 × 13.9.

U.W. 101-2209

Long bone shaft fragment, based on size and curvature probably humerus, representing 30% of the circumference. 29.1 × 11.4.

U.W. 101-2210

Bone fragments, probably cranial, less than 30 mm.

U.W. 101-2211

Rib fragment. 34.3 long, diameter 8.6 × 6.1

U.W. 101-2221

Bone fragments.

U.W. 101-2222

Cranial vault fragment. 22.4 × 14.8.

U.W. 101-2223

Long bone fragment. 38.7 × 13.0

U.W. 101-2224

Left distal humerus fragment. The lateral supracondylar crest is well preserved and evident. Length 54.4.

U.W. 101-2225

Long bone shaft fragment with rounded circumference compatible with humerus or radius, including 50% of shaft circumference. 29.9 long, 12.0 wide.

U.W. 101-2226

Flat long bone shaft fragment. Consistent with distal humerus or proximal tibia. 38.5 × 15.8.

U.W. 101-2227

Long bone shaft fragment consistent with humerus. 29.1 × 15.1

U.W. 101-2228

Shaft fragment consistent with small long bone, metacarpal or metatarsal. 21.3 × 9.5

U.W. 101-2229

Long bone shaft fragment consistent with radius, ulna, or fibula. 27.4 × 10.2

U.W. 101-2230

Bone fragments in sediment.

U.W. 101-2231

Proximal left humerus shaft. Length 53. Diameter at surgical neck 18.2 × 15.8. A chip of enamel is adhering within the sediment that fills the proximal end of this fragment. The enamel chip is approximately 3 mm × 5 mm and could be molar or incisor. A second fragment refits the distal end of the first. Length 24.8 × 13.7.

U.W. 101-2232

Bone fragment 19 × 9 mm.

U.W. 101-2233

Bone fragment 18.7 × 12.8.

U.W. 101-2234

Bone fragment 20.6 × 8.4.

U.W. 101-2235

Tooth root, slightly bilobate toward crown. 15.1 mm.

U.W. 101-2236

Bone fragment with concave surface exposed on one side, other side trabecular fragments with sediment obscuring detail. 21 × 17.

U.W. 101-2237

Maxillary left molar. BL 12.3, MD 11.6. All roots are present and intact, all three curve distally, and the buccal two roots curve into each other strongly distally. Based on roots and crown morphology this resembles an M3 more than either other molar. The crown is bigger than any of the M1s and those teeth do not have the posteriorly directed lingual root that this one has. Occlusal surface has slight wear, no dentin exposure and all cusps are salient.

U.W. 101-2238

Mandibular ramus fragment of right coronoid process. 35.9 × 14.8.

U.W. 101-2239

Shaft fragment of femur or tibia, relatively flat, consistent in size and thickness with U.W. 101-2259 femur. Length 48.3, width 19.4.

U.W. 101-2240

Shaft portion of radius or ulna. 43.1 long, diameter obscured by sediment. This fragment is consistent with the break at the distal end of the U.W. 101-2121 radius fragment and may be the same bone. A small thin and slightly rounded bone fragment was adhering to one end, not in anatomical position. This is thinner than the shaft fragment and probably belongs to some other bone.

U.W. 101-2241

Bone fragment, thin with adhering sediment on one side. 19.9 × 13.6

U.W. 101-2242

Proximal right ulna, lacking olecranon process. Length 50.2, diameter below coronoid process 12.9 × 12.4. Two additional shaft portions refit the larger fragment.

U.W. 101-2243

Proximal left humerus shaft. The proximal end of this includes a small portion of metaphyseal surface across the lateral 25% of this end of the bone. There is additionally a small (10 mm) piece of epiphysis that was adhering in position to part of this lateral edge of the proximal end. This is now detached and its surface does correspond to the opposing surface of the diaphysis. Length 46.7. Shaft diameters at surgical neck 18.5 × 15.5.

U.W. 101-2244

Proximal end of a rib with the articular part of the head missing but the tubercle present. The neck is round in cross section and the break just lateral to the tubercle has a rib-like cross section. I tend to think this is left mid-to upper thorax. Length 25.9, diameter 6.3.

U.W. 101-2245

Bone fragment consistent with vertebral lamina. 13.3 × 9.3 × 5.6 thick at thickest point.

U.W. 101-2246

Proximal left radius, lacking head. Radial tuberosity and approximately 25% of shaft are present. The neck is broken and does not preserve an metaphyseal surface. It is size-consistent with adult material. That makes it comparable to the humerus elements nearby including U.W. 101-2243. Length 55.7, diameter of neck 10.1 × 9.2.

U.W. 101-2247

Phalanx or rib fragment. 20.6 × 8.2 × 4.9.

U.W. 101-2248

Rib fragments less than 20 mm.

U.W. 101-2249

Right distal humerus, lacking trochlea and capitulum. Length 88. Diameter above epicondylar ridges 14.1 × 15.7.

U.W. 101-2250

Long bone remains that represent at least two different elements. One of these is a right proximal femur, including base of neck, lesser trochanter, broken below greater trochanter. Subtrochanteric diameters AP 20.9, ML 28.5. Length of fragment 68.4. A second large fragment is a portion of long bone shaft, 62.5 long, 16.0 in diameter. The cortical bone is thin relative to the shaft diameter, suggesting that it is more consistent with juvenile femur or tibia rather than adult humerus. In size, this shaft fragment may be compatible with the immature proximal femur fragment U.W. 101-2260 but there is no refit between these pieces. This bone is highly dark stained with iridescent sheen and edge, it has the appearance of burned bone. Smaller additional fragment of same 27.3 × 14.0. Additional bag contains bone fragments in sediment.

U.W. 101-2258

Rib fragment 23.1 long.

U.W. 101-2259

Femur shaft. Refits right proximal femur in U.W. 101-2250. Fracture is stained, not fresh. Length 152.9. Diameter same as 101-2250.

U.W. 101-2260

Immature proximal right femur. Includes neck, head is missing. It is possible that the metaphyseal surface for head is present here but sediment and possible erosion mask whether this is the case. The metaphyseal surface for the greater trochanter is partially present. None of the head is present, and the broken portion of the neck does not retain any of its metaphyseal surface. Lesser trochanter is projecting, shaft is broken away irregularly and lateral border of shaft missing. Fragment 41 mm, neck 19.6 × 15.1.

U.W. 101-2261

Fragment with concave surface on one side, convex obscured by sediment on the other. Fragment 19.7 × 15.2 × 6.6 thick.

U.W. 101-2262

Long bone shaft fragment, ulna. 39 × 13.4.

U.W. 101-2263

First metacarpal head and 40% of shaft. Includes diagnostic morphology of H. naledi. 28.5 mm.

U.W. 101-2264

Shaft fragment of phalanx or metacarpal. 25 mm × 8.1 mm.

U.W. 101-2265

Bone fragment, possibly carpal fragment less than 15 mm.

U.W. 101-2266

Bone fragment, less than 20 mm.

U.W. 101-2267

Bone fragments within clump of sediment, possibly rib fragments, 32.7 mm.

U.W. 101-2268

Bone fragment, possibly rib fragment, 25.4 × 9.4.

U.W. 101-2269

Bone fragment with morphology and possible metaphyseal surfaces, possibly immature long bone fragment? 18.4 mm.

U.W. 101-2270

Bone fragments

U.W. 101-2271

Bone fragment less than 20 mm.

U.W. 101-2272

Long bone fragment 25.2 × 16.6 × 6.9 thick.

U.W. 101-2273

Shaft fragment of long bone, 28.5 × 17.4. Other associated fragments.

U.W. 101-2274

Manual phalanx. Head is eroded, base is missing. Probably proximal phalanx based on length. 28.0 mm long, diameters at midshaft 8.8 × 4.9.

U.W. 101-2275

Ischium fragment including lunate surface of acetabulum, broken superior to ischial tuberosity. 25.0 × 13.0 mm of lunate surface present. Fragment length 21.1 from acetabular border to inferior edge. Subacetabular sulcus 6.9 mm.

U.W. 101-2276

Mandibular corpus fragment, including the angulation between the base of the corpus and either the external or internal surfaces. Based on curvature this seems likely to be external. This fragment is possibly consistent with U.W. 101-2184, which is clearly right mandibular corpus including the base and the swelling at the base of the ramus. Fragment dimensions

36.5 × 13.5.

U.W. 101-2277

Shaft portion of ulna or fibula. Length 37.3, diameter 9.9 × 9.4.

U.W. 101-2278

Left proximal ulna with shaft fragments. Olecranon process is present but eroded, coronoid process appears to be broken or missing. The coronoid process is certainly missing because of a break. This is consistent with adult ulna due to the extent of the olecranon process that remains without evidence of metaphysis, but its most proximal extent is abraded.

Measurements below coronoid 11.9 × 11.5. Length 47.6.

Supplementary Information 2.2. Identification and assessment of skeletal remains from Dinaledi Feature 2

Very little skeletal material has been excavated from or above Dinaledi Feature 2. The feature remains largely undisturbed, with only a small semi-circular concentration of bone visible and an unknown portion remaining within the unexcavated S950W600 grid square. The identifiable elements are consistent with a single adult individual but very little evidence has been recovered to date.

U.W. 101-2134

Bone fragment, thin cortical flake, 13.9 mm

U.W. 101-2135

Bone fragment consistent with cranium or mandible, 12.7 mm

U.W. 101-2166

Bone fragment consistent with cranium or mandible, 34.9 × 20.9 mm.

U.W. 101-2167

Bone fragment in sediment, possibly phalanx shaft. Less than 20 mm.

U.W. 101-2198

Flat thin fragment consistent with zygomatic arch or vertebral lamina. 15.2 × 7.4.

U.W. 101-2220

Femur or tibia shaft portion. Surface morphology and cross-section obscured by sediment. Length 80.7, diameter 24.9 x

Supplementary Information 3: Hill Antechamber Feature and Artefact

Supplementary Information 3.1. Stratigraphic situation of the Hill Antechamber feature

During the course of excavation, it was possible to make a number of observations about the sediments and stratigraphy of the units surrounding the feature. The sediment is a brown color with a fine texture similar to that encountered in the Dinaledi Chamber (5) (Figure S24), with laminated orange-red mud (LORM) clasts present within the sediment. The sediments along the north wall of the N50W50 and N50W100 units were loose in texture and with few large clay clasts. Around 5 cm from this north wall, LORM clasts became more frequent, and sediment is more indurated. Toward the south within these units, this LORM forms layers, from a thin veneer to 2 cm in thickness. This layering is visible in profile on the east wall of the present excavation (Figure S25). The layers are roughly parallel with the north-south slope of the floor surface of the chamber. The west wall of the excavation is dominated by a large dolomite boulder that remains in place. South of this boulder the layering is present and similar in orientation to the east wall, although not exhibiting as prominent contrasts in coloration. To the north of the boulder, clay clasts are less frequent, and the profile is similar to the north wall of the excavation area. Finally, the sediments to the south of the feature extend downslope from the feature itself. Our excavation of the S50W100 and S50W50 grid squares identified the southern boundary of the feature and excavated downward approximately 15 cm to leave a vertical wall representing the sediment profile immediately against the feature itself. This wall revealed a layered profile with alternating reddish clay layers, similar to the east wall of the excavated area (Figure S25).

The upper 10 cm of this wall exhibited horizontal layering, and in the lowest 5 cm this layering transitioned into a subhorizontal slope becoming a 20 to 30 degree slope running downward from east to west. These lower layers were continuous further to the south where they are visible in the excavated horizontal surface at the base of the S50W100 and S50W50 units.

The stratigraphic situation of the feature itself can be examined directly from CT data as well as from the excavation record (Figure S26). LORM clasts are highly visible in the CT images due to their higher density than surrounding sediment; their near-uniform density and irregular shapes make them easily distinguished from fossil material. Clay clasts occur sporadically throughout the feature. More of these clasts occur in the lower half of the feature on the eastern edge. Here they are arranged in layers that have a slope similar in orientation that seen in the surrounding sediments. These layers do not characterize the rest of the feature. Instead, most of the volume of the feature is poor in clasts and rich in fossil material and sediment. This fill has a shallow bowl-shaped bottom that is marked by clay clasts and frequent voids. For example, a jumble of clasts and voids is evident in east-west transverse sections at 50% and 60% of the feature’s length (Figure S27), which contrast with the homogeneous, higher density sediment above this bowl-shaped bottom. Excavation records show that voids were also noted during the excavation at the bottom of the feature, as we worked to define the edge of the feature. These voids were air-filled, not filled with lower-density sediment. With this fill, we face the key question of whether the fill is stratigraphically above an existing layer or whether it cuts through layers, interrupting them . In the transverse CT sections from 10% to 20%, the layering is indistinct due to the relative lack of LORM clasts. In the section at 30%, a well-defined layer slopes from the east edge of the block continuously down to the bottom without interruption. Further to the south, from 40% up through 60% of the feature’s length it is evident that the bowl-shaped bottom of the feature fill interrupts the more layering that is seen at the easternmost part of the feature. The layering near the lower east side of the feature is less easily visible in the north-south sagittal sections, but also appears to be interrupted by a jumble of clasts and voids in the sections from 30% to 60% where the layering no longer appears.

This bowl-shaped interface of the feature lies immediately below several fossil specimens including an articulated right foot and ankle. It is notable that these articulated elements are surrounded by a halo of sediment with lower density than the surrounding fill; additionally, some small voids occur immediately above this articulated foot (Figure S29). The lower boundary of this fill is less clearly defined near the north edge of the feature, due to the lower number of clay clasts, however this fill still contains fossil material including a partial articulated hand and upper limb material, as well as the stone artifact. In north-south sagittal sections, the bowl-shaped lower border of the fill is also evident. Near the center of the U.W. 101-2076 block, the bowl reaches the bottom of the block itself and the orientation of the articulated foot suggests that the volume of fill likely extended slightly below the excavation boundary of this block. The excavation records provide additional context for this observation. After the removal of the feature, a short remnant of the sediment pillar remained, which the team excavated to bring to the same level as the surrounding excavation units. Near the center of the feature was an oval of bright orange LORM (Figure S30). This bright orange LORM patch was also evident within the bottom surface of the feature after inverting it, prior to jacketing (Supplementary Figure S30), and the portion remaining in the site was less than 1 cm in thickness. Further excavation revealed a very dark layer of sediment, looser and with a less compact texture, of 0.5–1 cm thickness, giving way to more typical brown sediment below. A small number of fossil fragments were recovered from this area after the removal of the jacketed blocks. These areas suggesting a textural and color contrast were localized beneath the lowest portion of the bowl-shaped contrast within the feature itself.

While this fossil-containing sediment fill makes up the majority of the feature’s volume, the uppermost aspect of the feature is a sub-horizontal layer that follows a northeast-southwest slope of approximately 10 degrees. This slope is at variance with the slope of the layering present around the feature, which is principally east-west and a steeper northeast-southwest slope. It does not correspond to the bowl-shaped bottom of the fill within the feature, and it does not correspond to the orientation of the articulated foot, stone artifact, or other material present within this fill. This sub-horizontal layer is composed entirely of fossil material. This material includes complete elements in articulation, in near-articulation, or in anatomical positioning relative to each other. It is this layer that defined our excavation of the feature, and it therefore extends to the edges of the blocks we recovered. The skeletal material itself is described below. From the perspective of stratigraphy, this uppermost portion of the feature was clearly defined relative to overlying sediment. The undistorted and complete elements within this layer are small compact elements including the upper and lower dentition of a single individual, carpals, metacarpals, and phalanges. Ribs and upper limb long bones are present within this layer, in relative anatomical order, and all exhibit crushing and flattening. The demi-mandible containing part of the lower dentition also is crushed with small bone fragments separated from each other by matrix-filled gaps. A large mass of undifferentiable bone material occurs near the north end of the feature; the proximity of this large mass near the upper and lower dentition and partial mandible suggests that this is the crushed remains of a calvaria. It is evident from the crushing of cortical bone of long bone shafts and mandible, and the complete flattening of a calvaria, that this upper body was subject to gravity and possibly additional mechanical forces that flattened it, distorting its original depositional situation into a sub-horizontal layer.

Supplementary Information 3.2: Description of the Hill Antechamber Artefact 1 associated with the Hill Antechamber Feature

The Hill Antechamber Artifact 1 (HAA1) is a stone located near articulated hand bones in the feature. This is the only stone greater than centimeter size evident within the feature. HAA1 still remains within the larger plaster jacket containing most of the Hill Antechamber feature. Examination of the CT data revealed this stone in close contact with the articulated bones of a Homo naledi hand, and we prioritized obtaining higher-resolution scan data of the stone to evaluate its morphology, as described in SI 1.3. This section presents a description of the artifact. Orientation is described based on an arbitrarily defined position of features illustrated in Figure 11. All measurements and descriptions are taken from both 3D images produced from the synchrotron scans or from examination of high-resolution 3D prints of HAA1. Both scans and 3D shape files of HAA1 are available for download on https://Morphosource.org. A high resolution movie of HAA1 is available as Supplementary Movie 2.

The shape of HAA1 is distinctive in comparison to other rocks on the surface or encountered during excavation of the Hill Antechamber or Dinaledi Chamber. HAA1 is 138.5 mm in total length. Its greatest supero-inferior height is 49mm just left of the middle of center. Its greatest width is 26.3mm. HAA1 is roughly crescent-shaped coming to a sharp point laterally left and a more rounded but antero posteriorly sharp edge right laterally. It presents a large flake scar of ca. 80mm in length on the anterior-inferior surface from approximately 10mm left laterally of its midline that travels to within 12mm of the right lateral end. The lunate area of this flake scar occupies approximately three quarters of the body of the artifact for most of its extent. A 6-7mm wide ledge is observed in the superior one third of the artifact that is created from the removal of rock flake ca. 84.5mm in length that leaves a prominent hump occupying the middle 65mm of the artifact. We cannot determine the type of stone unambiguously from scan data but we hypothesize based on its surface characteristics that it is likely composed of dolomite.

Within the area of the lunate flake scar on the anterior surface, and opposite this area on the posterior surface, striations are visible that appear to be use wear or erosional marks (Figure 12). Some of these lines are as much as 1.5mm wide but most are sub-millimeter in width and are 20 to 25mm in length. They travel predominantly in an lower-left to upper-right direction across this surface. These lines do not pass outside of the lunate flake scar area and are not visible in synchrotron segments of the rock below the surface, indicating that they are surface features confined to the area of the flake scar (Figure 12).

The posterior surface of HAA1 is dominated by a prominent ridge that runs from the left lateral point supero-laterally at an approximate 15 degree angle, reaching the superior edge about 50mm from the right lateral edge. The peak of the ridge forms the greatest width of the artifact and this occurs approximately 55mm from the left lateral end. Supero-inferiorly oriented erosional or wear lines are found on the right lateral half of the artifact roughly mirroring those on the anterior surface. There is a depressed area that may represent a worn flake removal in this same region that is not as large nor prominent as that found on the anterior surface.

The inferior edge of the artifact along the area of the anterior lunate flake scar is very sharp and under high resolution imaging by synchrotron (6μm), irregular serration is visible across the whole of the surface (Figure 12). No other serration is obvious along the edges of other sharp areas.

Supplementary Information 4: Materials and Methods

Supplementary Information 4.1: Previous Excavations in the Dinaledi Chamber

During the first two excavations undertaken in the Dinaledi Chamber (November 2013 and March 2014), a combination of structured white-light surface scans and high-resolution laser point cloud scans were used to collect 3D data on the spatial position of fossils, from which relative location and orientation of specimens could be derived (Figure S22). Most specimens, both on the surface of the chamber and in the excavation unit, were scanned in situ once uncovered and prior to collection. These spatial data were compiled and form the basis of this study. Work by Kruger and colleagues (11) details the workflow for the scanning process and resulting geo-referencing.

By combining scans from the surface of the chamber, as well as the main excavation pit, a 3D record of all collected and excavated fossil material was created. Individual white-light source photogrammetry scans, taken during excavations, were converted to the E57 point cloud File Format for 3D Imaging Data Exchange (12). These E57 point cloud data structures were aligned to the broader context of the Dinaledi Chamber (11) using CloudCompare (13). Each scan was aligned in CloudCompare by manually picking three or more reference points which were common to both the individual scan and the georeferenced Terrestrial Laser Scan (TLS) data collected from the Dinaledi Chamber. Reference points were a combination of fixed survey pins (15 in total) in the Dinaledi Chamber, together with permanent geological features of the chamber walls and floor that can be repeatedly identified in multiple scans. It is often convenient to refer to the locations of objects in terms of these reference points. As excavation had only been conducted in one area of the chamber, the remaining pin locations are useful in the present study only in reference to the material recorded and collected from the chamber surface.

Reference points were a combination of fixed survey pins (15 in total) in the Dinaledi Chamber, together with permanent geological features of the chamber walls and floor that can be repeatedly identified in multiple scans. It is often convenient to refer to the locations of objects in terms of these reference points. As excavation has only been conducted in one area of the chamber, the remaining pin locations are useful in the present study only in reference to the material recorded and collected from the chamber surface.

Scans were aligned from the most recent (from the excavation during March 2014) to the oldest (November 2013). Once the scans were rotated, aligned and scaled, fine registration was applied, to provide an optimum alignment. Once scans were accurately aligned, the 3D transformation was then applied and the individual scan’s coordinate system was then updated to the global coordinates of the chamber, and thus any other scan’s coordinates within the broader framework of the Dinaledi Chamber. This method of data collection provides a virtual snapshot of the excavations at any one point in time.

Individual bones and fossil fragments were then plotted, and their coordinates recorded by visualizing the transformed E57 point clouds in Autodesk’s Recap 360 software. This was accomplished by using high-resolution white-light source photogrammetry scans, visualized in Artec Studio 10 Professional (and later Artec Studio 11 Professional), in conjunction with the point cloud data, as a guide for placement of coordinate data. A surface resolution of approximately 0.5 mm and a three-dimensional point accuracy of approximately 0.1 mm ensured accurate placement of bone material into the global coordinate system. A centroid datum of these data, in the form of 3D points (x, y and z), were recorded for individual teeth, fragments and other small elements. In the case of long bones, a coordinate for the distal, centroid and proximal end of each bone or portion was recorded. Thus, when possible, three coordinates were recorded for each long bone, allowing for the directionality (strike and dip) of these bones to be attributed in spatial analysis in relation to the rest of the fossil assemblage. For long bone material for which it was not possible to record 3 coordinates (“distal”, “proximal” and “center”), a centroid was calculated based on the coordinates of endpoints (“distal” and “proximal” points).

Supplementary Information 4.2. Previous work in other parts of the cave system

The evidence from the Lesedi Chamber (2) we now recognize may represent a situation of possible mortuary practice by H. naledi. The LES1 partial skeleton was excavated from the 102a locality, which is within a blind tunnel approximately 1.8 m in length, with a tapering width of c50 cm and a floor located above the current floor of the Lesedi Chamber. Portions of this skeleton were exposed on the sediment surface at the time of discovery, and a few elements were discovered up to 6 m away on the surface, but most were contained within a 10cm layer of weakly stratified, unlithified mud clast breccia beneath a 2 cm-thick lighter brown colored mudstone, like that found in the Dinaledi Chamber. Some volume of the sediment containing skeletal remains had slumped downward from this alcove onto the cave floor (2). The recovered material includes anatomically contiguous elements, with a small number found in articulation. Sediment from the Lesedi Chamber excavation is similar in chemical content to that associated with Dinaledi Feature 1, including the significant positive loadings of P and S on PC2 (Figure 7). The combined evidence suggests that the LES1 individual was also buried with soft tissue present, with later surface exposure of some elements due to sediment slumping from the alcove. Although we cannot rule out that other features may occur within the floor of the Lesedi Chamber, the location of the LES1 individual in this blind tunnel or alcove clearly placed it in a different situation from the chamber floor.

We have described elsewhere (1, 3) the occurrence of isolated remains of a H. naledi juvenile individual within the U.W. 110 locality, which lies within the Dinaledi Subsystem approximately 10m beyond the Dinaledi Chamber within a narrow (< 20cm) fissure passage. The bones comprise parts of a skull with some teeth, absent the mandible or any postcranial remains. The material presents no signs of bone surface modifications indicative of mechanical damage caused by postdepositional transport, nor are the sedimentary and stratigraphic situation in the subsystem compatible with flooding or other conditions that would enable water transport (5). It is plausible that H. naledi placed the skull in this location and this may reflect a mortuary practice different from the graves seen in the other chambers. We have located three other occurrences of H. naledi fossils in similarly remote situations. Though analyses of these remains and their situations have not been concluded, each is challenging to explain without the involvement of H. naledi in their placement.

Supplementary Information: Tables

Supplementary Table 1.

Supplementary Table 2.

Supplementary Information: Figures

Supplementary Figure S1.

Supplementary Figure S2.

Supplementary Figure S3.

Supplementary Figure S4.

Supplementary Figure S5.

Supplementary Figure S6.

Supplementary Figure S7.

Supplementary Figure S8.

Supplementary Figure S9.

Supplementary Figure S10.

Supplementary Figure S11.

Supplementary Figure S12.

Supplementary Figure S13.

Supplementary Figure S14.

Supplementary Figure S15.

Supplementary Figure S16.

Supplementary Figure S17.

Supplementary Figure S18.

Supplementary Figure S19.

Supplementary Figure S20.

Supplementary Figure S21.

Supplementary Figure S22.

Supplementary Figure S23.

Supplementary Figure S24.

Supplementary Figure S25.

Supplementary Figure S26.

Supplementary Figure S27.

Supplementary Figure S28.

Supplementary Figure S29.

Supplementary Figure S30.