Following the University of the Witwatersrand’s fossil-numbering system (Zipfel and Berger, 2009), this second H. naledi locality has been designated U.W. 102. The chamber itself has been named the Lesedi Chamber, a word meaning ‘light’ in Setswana. By contrast, the Dinaledi Chamber was numbered site U.W. 101. Excavations in the Lesedi Chamber have been carried out in three areas, designated U.W. 102a, U.W. 102b, and U.W. 102c.

The Lesedi Chamber is in the central sector of the Rising Star system (Figure 2), at a depth of ~30 m from the surface directly above the chamber. All measurements reported here are approximate. The first fossil deposit to be recognized (U.W. 102a) is located just off the southwest corner of the North-South Fracture Passage, a northern arm of the Lesedi Chamber. This fossil deposit is approximately 60 m NNE in a straight line from the Dinaledi Chamber. There is no straight-line route between the Dinaledi and Lesedi Chambers, and the shortest traversable route between the two areas is almost 145 m. There are currently four access routes from the surface to the Lesedi Chamber. The most accessible of these currently follows an 86 m downward-sloping path with several narrow passages and short climbs, but only one squeeze and no significant crawls. This has been the main access route for excavators. The other three routes are each substantially more challenging.10.7554/eLife.24232.005Figure 2.

The Dinaledi Chamber (U.W. 101) is marked by a yellow circle, while three surface entrances into the system are marked by blue circles.

DOI: http://dx.doi.org/10.7554/eLife.24232.005

In addition to the first fossil deposit to be recognized in the chamber, two additional concentrations of skeletal material have been identified to date (Figure 3), and we have designated these as areas 102a, 102b, and 102c. We began investigating each of these areas because team members noticed hominin fossil material exposed on sediment surfaces. The discovery of 102a by Rick Hunter and Steven Tucker led to the initial scientific investigation of the chamber; discoveries of both 102b and 102c were made by Hannah Hilbert-Wolf during the course of geological sampling of the chamber. These three areas do not represent a systematic sampling of the chamber’s contents and we have excavated only a very small sediment volume, less than 200 L (<0.2 m³) in total from all three areas. The chamber contains a much greater volume of sediment and we do not know what density of fossil bone it may contain beyond our samples.10.7554/eLife.24232.006Figure 3.

DOI: http://dx.doi.org/10.7554/eLife.24232.006

Hominin material from the 102a area includes 118 identifiable specimens (Table 1; Figure 4). Fifty-seven of these are cranial and dental specimens that either refit directly or are morphologically compatible with a nearly complete fossil cranium, designated LES1 (Figure 5). Hominin postcranial remains from locality 102a include 61 identified specimens that represent a minimum of 31 postcranial elements, not counting ribs. These include a minimum of two partial femora, two partial humeri, one complete clavicle and two clavicular fragments, two partial ulnae, several fragments of scapula and radius, many rib fragments, a near-complete first rib, a partial sternum, four hand and wrist elements, an immature ilium and sacrum fragment, and a partial thoracic and lumbar vertebral column. Every anatomical region of the skeleton is represented with the notable exceptions of tibia, fibula and pedal remains.10.7554/eLife.24232.007Figure 4.

The adult cranial material from 102a all belongs to a single cranium; most of the adult postcranial material probably belongs to the same individual. The adult cranial and postcranial material is shown here, except for the U.W. 102a-001 femur. The possibility that the femora represent two adult individuals makes it unclear which femur may be attributable to the skeleton; for the purposes of illustration, the U.W. 102a-003/U.W. 102a-004 femur is included in this photograph.

DOI: http://dx.doi.org/10.7554/eLife.24232.007

The LES1 cranium is fragmented but is represented by most of the vault and part of the face (Figure 5). To date, we have successfully refitted the near-complete mandible, the near-complete right maxilla, a partial palate and a partial left maxillary dental row, and a partial vault including the near-complete frontal, left and right nasal and left lacrimal bones, near-complete left parietal and temporal, partial right parietal, and a portion of left occipital. LES1 has a complete adult dentition except for the crowns of the lower left central and lateral incisors. The face is reconstructed from the partial right maxillary bone, including the frontal process, which refits to the right nasal bone and frontal. The left mandibular ramus is well-enough preserved to allow a rough estimation of the condyle position, enabling an approximation of the midsagittal contour of the face (Figure 5).

All additional cranial fragments in the present 102a collection are non-duplicative with this refitted vault and face, and where they represent the opposite side of the vault, they match in morphological detail. However, many of the fragments lack clear refits with the existing vault or maxillary portions. Further physical reconstruction of the cranium will await fragments that may emerge from excavation in the future. The refitted vault, with the application of virtual mirror reconstruction, is sufficient to allow an estimate of endocranial volume of approximately 610 ml (Figure 6).10.7554/eLife.24232.009Figure 6.

The refitted calvaria was mirrored and filled, resulting in a volume estimate of 610 ml. Scale sphere = 10 mm.

DOI: http://dx.doi.org/10.7554/eLife.24232.009

U.W. 102a-021 is a nearly complete right clavicle, missing only the articular surface of the sternal end, where trabecular bone is exposed over the entire articular area, including a small bit of the anterior surface (Figure 14). The shaft is broken into two pieces near the midshaft but the two pieces conjoin cleanly. There is also a small bit of damage to the acromial end. On the posterior surface, the medial part of the crest for the conoid tubercle is broken off. The specimen exhibits a dark surface coating on the anterior aspect of the sternal half and patchy areas of black staining on its acromial half. There are fine hairline longitudinal cracks on much of the acromial half of the bone.10.7554/eLife.24232.018Figure 14.

Left, from top: superior, anterior, inferior, posterior views. Right, from top: medial and lateral views. Scale bar = 2 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.018

U.W. 102a-002 is a proximal shaft fragment of a right humerus (Figure 15). The head and greater tubercle are missing, as is all but the very distal base of the lesser tubercle. From this metaphyseal region, the fragment preserves approximately 50–60% of the shaft, with a total fragment length of 85 mm. U.W. 102a-013 includes two fragments identified as humeral head, each with some articular subchondral bone, which may derive from the same element as U.W. 102a-002. They appear to be consistent in curvatures of the articular surface with U.W. 102a-257. The specimen is mostly coated with a brown to dark-brown mineral patina, the surface is unweathered with only slight surface removal on the distal end of the anterior surface. The breaks, both proximal and distal, are sharp.10.7554/eLife.24232.019Figure 15.

From left: posterior, medial, anterior and lateral views. Right, from top: Scale bar = 5 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.019

U.W. 102a-015 is a right proximal ulna, on which much of the trochlear notch and olecranon process are preserved in addition to the proximal half of the diaphysis (Figure 17). There is erosion to the anterior tips and margins of the coronoid and olecranon process. The surface of the shaft is well-preserved and exhibits very slight hairline longitudinal cracks. The break to the distal end is sharp and cleanly transverse. The olecranon process is mediolaterally narrow and the trochlear notch appears to have opened anterosuperiorly, as in modern humans but not Neandertals. While there is only one fragmentary (and probably immature) proximal ulna from the Dinaledi Chamber, U.W. 102a-015 generally compares well with U.W. 101–560 in terms of morphology and gracility.10.7554/eLife.24232.021Figure 17.

From left: anterior, medial, posterior and lateral views. Right from top: proximal and distal views. Scale bar = 2 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.021

Four adult hand and wrist elements have been recovered from the 102a locality (Figure 18). U.W. 102a-028 is a right fourth metacarpal (RMc4), with a small base and a relatively radioulnarly broad head (Figure 15). The metacarpal shaft shows substantial curvature and is relatively robust for its length, although it still falls within the upper range of variation found in modern humans (Figure 16). U.W. 102a-117 is a complete right scaphoid; U.W. 102a-476 is a complete right capitate; and U.W. 102a-477 is a partial right lunate. The scaphoid, lunate, and capitate are consistent in size and appear to match each other when placed in anatomical articulation; the RMc4 is likewise a good match in size, with a lateral base matching in dorsopalmar contour the base of the capitate. Thus, these four bones may represent the right hand of one individual.10.7554/eLife.24232.022Figure 18.

From left: dorsal, ulnar, palmar and radial views. Right from top: distal and proximal views. Scale bar = 1 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.022

Seven vertebrae have been recovered in the 102a assemblage, all from the lower thoracic and lumbar region of the spine. These vertebrae are roughly equivalent in preservation. The thin cortical bone of the vertebral bodies is eroded in large patches on these elements with exposure of underlying trabeculae. They have minimal surface staining or patination, and where the vertebral arches are present, the cortical surface is well-preserved.

U.W. 102a-036 is a largely complete antepenultimate thoracic vertebra, inferred as T10, with limited erosion to the anterior surface of the vertebral body and some damage distally on the transverse processes, particularly on the right side, and to the posterior end of the spinous process (Figure 20). The vertebral body is ovoid to kidney-shaped and the ring apophyses are relatively thick, covering approximately three-quarters of the vertebral body surface. The spinal canal is ovoid in shape and about one-third the size of the vertebral body. Facets for the tenth rib are posteriorly positioned, almost entirely on the pedicles. The pedicles themselves are transversely thick, as are the transverse processes, which are strongly posteriorly oriented. Together, the transverse processes and pedicles form nearly continuous, robust lateral structures for anchoring epaxial muscles and for transmitting forces to and from the vertebral body, respectively.10.7554/eLife.24232.024Figure 20.

Clockwise from top left: posterior, superior, inferior, left, right, and anterior views. Scale bar = 2 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.024

U.W. 102a-250 is a nearly complete right first rib, with erosion and breakage to the head, tubercle, lateral border and distal end. The neck is flattened in its superior-inferior dimension and descends in the vertebro-inferior direction. The tubercle and the posterior angle coincide. The facet of the articular tubercle was damaged post-mortem.

Two partial first ribs (U.W. 101–083 and U.W. 101–621) of H. naledi are preserved in the Dinaledi Chamber hominin sample, but neither rib has its head nor enough of the shaft preserved to allow accurate estimation of curvature (Williams et al., 2017). The angulation and shape of these fragments appears comparable to those of MH2 Au. sediba and A.L. 288–1 Au. afarensis. U.W. 102a-250 is more complete than the Dinaledi first rib fragments, and is similar in morphology in the overlapping regions. This rib is slightly more curved than the Sterkfontein first rib, StW 670 (Tawan et al., 2016). The anatomy of U.W. 102a-250 is entirely compatible with attribution to H. naledi, although the bone is also similar in morphology and size to known australopith first ribs.

Thirteen additional specimens from 102a are partial ribs or rib fragments, none are identifiable to element and none present anatomical information that is useful for testing the taxonomic affiliation of the sample.

U.W. 102a-138 (Figure 23) is a fragmentary right ilium of an immature individual (as evident by the presence of triradiate cartilage, by an unfused apophysis at the anterior inferior iliac spine, and by very small overall size). The fragment is very light, with thin cortical bone, and is eroded around margins of the acetabular portion. The iliac blade is mostly missing, but the auricular surface, greater sciatic notch, acetabulosacral buttress, and anterior margin of the iliac blade are present. Despite the thin and fragile nature of this element, the surface is well-preserved.10.7554/eLife.24232.027Figure 23.

The medial view is at the center. Clockwise from top: superior, lateral, inferior and anterior views. The unfused triradiate suture is notable.

DOI: http://dx.doi.org/10.7554/eLife.24232.027

U.W. 102a-001 is a proximal right femur, in which much of the head and neck, and the proximal subtrochanteric shaft are preserved (Figure 24). The head is badly eroded, especially anteriorly, and only a few small patches of subchondral articular bone are preserved on the posterior aspect. The posterior side of the neck is fairly well preserved from the head all the way to the lesser trochanter, which is planed off, with only the base remaining. The anterior side of the neck is missing. Trabecular bone is exposed from the anterior head all the way to the lateral surface at the base of the greater trochanter. The greater trochanter is missing entirely, save for a small bit of its distal lateral surface. The surface overall is marred by areas of post-depositional damage, including a number of transverse scratches on the shaft.10.7554/eLife.24232.028Figure 24.

From left: posterior, medial, anterior and lateral views. Right from top: proximal and distal views. Scale bar = 5 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.028

The hominin material from 102a appears to represent a minimum of two adult individuals and one immature individual. The inference of two adults is based upon the morphological incongruence of the left (U.W. 102a-003) and right (U.W. 102a-001) femoral elements (discussed above). Still, no adult element is clearly duplicated in the collection. The U.W. 102a-138 ilium, along with an immature sacrum fragment and two immature long bone fragments not described here, demonstrates the presence of at least one immature individual.

The lack of duplication of elements suggests that much (but not all) of the adult material may represent a single individual skeleton, which parsimoniously would also include the LES1 cranium. We accept this hypothesis provisionally. All elements in the current 102a collection were recovered from within an excavation area less than 50 cm x 70 cm, and 40 cm deep. Two vertebrae (U.W. 102a-154a and U.W. 102a-154b) were in articulation in situ. The articular morphology and sizes of seven vertebrae suggest strongly that they represent a single individual; the remainder of these elements were recovered in close physical proximity but not in articulation. All fragments attributed to the LES1 cranium were likewise recovered from a small area. The hand and wrist material is consistent with a single right hand on the basis of articular morphology. None of the other elements lend themselves to an evaluation of articular compatibility, but they are consistent in size. We consider it unlikely that the number of elements in the current collection would be recovered from a commingled assemblage consisting of substantial parts of multiple skeletons without also introducing duplicate elements.

This raises the problem of what seems to be a mismatch of the U.W. 102a-001 and U.W. 102a-003 femora. These two are similar in size, and the difference in their shaft dimensions is not greater than that found in 95% of a large sample of paired left and right human femora. However, they are different in muscle attachments and diaphyseal morphology, to the extent that would represent unusual asymmetry in a single individual. The data do not allow us to discard the hypothesis that one of the femora represents a second adult individual in 102a, albeit an individual of similar body size. The descriptive and measurement data do not indicate which (if either) of these femora may belong to the individual represented by LES1, nor which (if any) of the other postcranial elements are associated with either femur. On the basis of the non-duplication of elements, it seems likely that if there is a second individual, this second individual is represented by only a small number of elements, possibly just the femur. The two humerus specimens, U.W. 102a-002 and U.W. 102a-257, differ slightly in shaft diameter where it can be compared, but no other morphological differences are apparent on the preserved fragments, and a slight degree of upper limb asymmetry is not unusual in humans or in fossil Homo. We conclude that most of the 102a adult material probably represents a single skeleton, but we cannot assume that the ratios of either femur with other elements in the sample would reflect the proportions of a single individual. To be conservative, any consideration of proportions should allow for the possibility that multiple individuals are present.

The material from area 102b includes 12 specimens identified as hominin. Most of this collection consists of small fragments of cranium, many of which are identifiable to element, but which preserve no diagnostic morphology that would assist in taxonomic assessment. One partial mandible and five possibly associated teeth preserve morphological characters that are useful for taxonomic attribution.

U.W. 102b-438 is a fragment of right mandibular corpus from an immature individual (Figure 29). The alveoli for the RM₁ are present; the alveoli for the rdm₂ are also present and reveal the crypt with RP₄ crown intact within. The fragment is broken anteriorly at the crypt for RP₃ and posteriorly at the crypt for RM₂, neither permanent tooth is present. The preserved corpus height at the midpoint of rdm₂ is 14.9 mm; this is perhaps 1–2 mm less than the true value because of the erosion of the alveolar bone surface. Corpus breadth at the anterior edge of M₁ is 15.2 mm; total length of the fragment as preserved is 43.5 mm.10.7554/eLife.24232.033Figure 29.

From left: basal, lingual, occlusal, buccal and anterior views. The RP₄ is within its crypt. Scale bar = 2 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.033

The 102c deposit is a very small (~2 L) volume of sediment enclosed within a dissolution cavity in the cave wall. Only one morphologically identifiable hominin specimen has been recovered from this area. U.W. 102c-589 is a fragment of left mandibular corpus with worn LM₁ and LM₂ in situ (Figure 31). These teeth have less occlusal wear than those of the LES1 mandible, but are morphologically very similar. The mandibular corpus is broken irregularly at the alveoli and the base of the mandible is not present. Anteriorly, it is broken at the mesial M₁ root; posteriorly, it is broken at the mesial alveolus for the M₃. The root of the ascending ramus is preserved and becomes independent at approximately the midpoint of M₂. Corpus breadth at M₂ is approximately the anatomical value at 19.8 mm; the broken corpus does not permit this measurement elsewhere.10.7554/eLife.24232.035Figure 31.

Top row, from left: lingual view, buccal view and occlusal view. Top right: two H. naledi teeth from the Dinaledi Chamber that present comparable occlusal morphology and wear to the U.W. 102 c-589 teeth: U.W. 101–297 RM₁ (top, reversed to represent left side) and U.W. 101–284 LM₂ (bottom). Bottom row, from left: U.W. 102 c-589 posterior view, anterior view and basal view. Scale bar = 2 cm.

DOI: http://dx.doi.org/10.7554/eLife.24232.035

The general preservation of the Lesedi Chamber hominin material resembles that of the skeletal assemblage from the Dinaledi Chamber (Dirks et al., 2015). The Lesedi skeletal material has a surface coloration that ranges from light grey to red-brown, and internal structures or cortex at fresh break points are colored pale buff to off-white, contrasting with unbroken adjacent surfaces. The remains present no evidence of calcite crystal formation. Specific comments on preservation have been included with the specimen descriptions above, and taphonomic observations on each specimen are summarized in Supplementary file 5.

The Lesedi skeletal assemblage exhibits varying degrees of post-mortem damage. Most specimens are fragmented or broken to some degree, with areas of cortical bone removal or abrasion (Supplementary file 5). All fractures observed in the assemblage are consistent with post-mortem (dry bone) failure (Supplementary file 5), and there are no spiral or incomplete fractures indicative of green or wet bone (L’Abbé et al., 2015; Symes et al., 2014). Fractures in the assemblage include transverse or right-angled breaks on the shafts of long bones, showing block-comminution between major breaks and step-fractures following the longitudinal grain of the long bone, which are distinct from cracking or crazing resulting from weathering (Symes et al., 2014).

The Lesedi remains display no evidence of sub-aerial weathering processes (sensu Behrensmeyer, 1978; see also Lyman and Fox, 1989 and Junod and Pokines, 2014, indicating that the dry bones were not subject to surface processes before deposition in the cave environment. Sub-aerial weathering is primarily characterized by both cracking and delamination. Within the Lesedi assemblage, there is some evidence of surface cracking, but it generally does not penetrate deep into the cortex. No indications of the secondary features of weathering, such as delamination, deep patination, bleaching or cortical exfoliation, were observed in any of the bone fragments from Lesedi, suggesting that the bones were not affected by surface exposure (Supplementary file 5). A comparison with human bone derived from forensic or sealed archaeological contexts (Figure 32) suggests that the cracking and fracture patterns observed in the Lesedi assemblage can be explained most parsimoniously by the effects of the burial environment (Pokines, 2014), and were exacerbated in specimens where fluctuations in moisture content caused swelling and shrinkage of the cortical structure (Conard et al., 2008; Dirks et al., 2016; Dirks et al., 2015).10.7554/eLife.24232.036Figure 32.

Top: right corpus and ramus of LES1 mandible prior to reconstruction, showing patterns of cracking consistent with the effects of sediment loading and wetting during the burial process and during skeletal decomposition. This pattern of taphonomic alteration is superficially similar to sub-aerial weathering processes, but is independent of surface exposure and occurs in both deep- and shallow-buried deposits. Middle: forensic known-history case for comparison. This specimen was recovered from a deep known-history inhumation. The fleshed body was deposited at a depth of 1.5 metres and recovered after a 30-year burial period. Note the similarities in surface texture, and superficial cracking, which follows the biomechanical stress lines (grain) of the bone. Bottom: U.W. 102a-010 acromial fragment showing surface modification with cortical pitting and punctate marks. Note that many of these marks (circled in yellow) penetrate pre-existing layers of manganese oxy-hydroxide deposited on the bone surface; the damage was therefore produced inside the Lesedi Chamber on bones that were already covered in coatings of manganese mineral.

DOI: http://dx.doi.org/10.7554/eLife.24232.036

In addition to the hominin skeletal material, some faunal remains have been recovered from the Lesedi Chamber. Over 80 faunal elements or fragments were collected from U.W. 102a. U.W. 102b yielded 23 specimens of fauna, while a single rodent tooth was recovered from U.W. 102c. In-depth analyses of the faunal material are ongoing, and comprehensive descriptions are in preparation. We provide here a preliminary list of the taxa identified for reference (Table 3).10.7554/eLife.24232.037Table 3.

Mammal species recorded in the Lesedi Chamber. Several specimens from the classes Aves, Amphibia and Reptilia were also recovered, but individual counts and taxonomic identifications are pending further examination.

DOI: http://dx.doi.org/10.7554/eLife.24232.037

The Lesedi faunal assemblage includes micromammal, small to mid-size mammal, and non-mammalian remains. With respect to the micromammals (prey body mass <500 g), 5 genera of rodents and 1 genus of shrew were identified out of 28 craniodental specimens (21 MNI). There are also potentially two additional murine genera and one soricid genus present in the assemblage, although lower-level identification is not possible at this time. Interestingly, all of the non-hominin fauna are of relatively small species. The largest mammalian specimens come from dental material attributed to Canis aff. C. familiaris. The size range of this material is outside the range of modern C. mesomelas but the morphology is definitively not Lycaon (Hartstone-Rose et al., 2010; Wayne, 1986). The felid material is also small, falling in the size range of the African wildcat. None of these individuals is likely to have exceeded 10–15 kg (Kingdon, 2015) and the rest of the assemblage consists almost entirely of animals smaller than 3 kg, including four specimens from the family Herpestidae. Aside from a single lagomorph specimen, the macro-mammalian material comes exclusively from the order Carnivora, a situation that is unusual in the fossil record (Werdelin and Sanders, 2010).

We do not presently know whether some or all of these faunal remains may be contemporaneous with any of the hominin fossil material. Faunal remains have been recovered both on the surface and also from within sediments near hominin remains. However, the Lesedi Chamber is not a completely isolated environment, and sediment deposits are currently eroding from their original depositional contexts, with evidence for slumping and reworking in the chamber. Therefore, we cannot yet comment on the relative timing of deposition of the hominin and faunal material. Additional tests, including attempts to date both hominin and faunal elements directly, will help us answer these questions and to relate the faunal material to the chronological and environmental context of H. naledi in the Lesedi Chamber.

Formal excavations began in the Lesedi Chamber in May 2014, after initial surface and ex situ material had been recovered. Like the Dinaledi Chamber, the access route to the Lesedi Chamber is very restricted. Additionally, the excavation area available in the Lesedi Chamber is considerably more confined than that in the Dinaledi Chamber. As a result, protocols followed those used in the Dinaledi Chamber (Dirks et al., 2015), with a few exceptions. Specifically, attempts to use some of the surface-scanning methods applied in the Dinaledi Chamber (Kruger et al., 2016) failed due to the limited working distances of the scanner, the convoluted surfaces, and the poor contrast of the surrounding dolomite. However, high-resolution laser scan data have been acquired for the Lesedi Chamber and will form the basis of future spatial work. Consequently, documentation of the excavations and provenience of the material was conducted using traditional archaeological methods, including written descriptions, photography and drawings. Excavations were conducted in tandem with geological mapping and with sedimentologic and taphonomic analyses of the site.

For excavations in U.W. 102a, the area was divided into 20–40 cm sections that were based on the contours of the sloped surface and the width of the tunnel at that point. Sections were excavated in 2–5 cm levels. In some areas where few specimens were being recovered, levels were expanded to 10 cm.

In the U.W. 102b area, excavations began on the surface of the sediment deposit in which the first material was recovered, and proceeded in 2–5 cm levels until the chert layer on which the sediment had accumulated was reached. When this area was exhausted, excavations moved along the chert shelf, removing sediments from the surface, to the chert, in 20 cm horizontal sections. A sediment pocket above the fossil deposit (on Chert 3) was also excavated to determine whether material had trickled down from above. Preliminary excavations were also begun on the antechamber floor below the primary fossil deposit.

Owing to the confines of the dolomite recess in which the hominin material was found, U.W.102c was not divided into sections. However, it was also excavated in 2–5 cm levels, from the top of the sediment accumulation to the chert shelf at its base.

For all three areas, all sediments were collected for every section and level separately. Each bag of back dirt was labeled, removed from the cave and dry screened to recover small elements, fragments and microfaunal remains. All sediments were retained and are currently stored at the University of the Witwatersrand’s Evolutionary Studies Institute.

Taxonomic identification of the Lesedi Chamber hominin material depended on a series of systematic comparisons to other hominin species. Many of these observations were carried out on the original fossil specimens by the authors; some have relied upon observations taken from research-grade casts of fossil specimens available at the University of the Witwatersrand and elsewhere, while in some cases, observations were only available from the literature. The fossil comparative samples employed in this study are essentially the same as those described in Berger et al. (2015), with the addition of the Dinaledi Chamber sample of H. naledi (Berger et al., 2015; Kivell et al., 2015; Marchi et al., 2017; Feuerriegel et al., 2017; Laird et al., 2017; Williams et al., 2017; Zipfel and Berger, 2009). The gross morphological appearance and dental dimensions of the remains immediately made it clear that they were inconsistent with Paranthropus, and with Australopithecus afarensis, Australopithecus africanus, or Australopithecus garhi, and so we focused our comparisons upon Homo and Australopithecus sediba. The composition of samples of other hominin species is as follows, largely repeated from Berger et al. (2015).

Australopithecus sediba. The partial skeletons MH1 and MH2 from Malapa, South Africa were included in this study on the basis of examination of the original specimens by the authors.

Homo habilis. Samples from Olduvai Gorge, East Lake Turkana, the Omo Shungura sequence, Hadar, and Sterkfontein were considered within the hypodigm of H. habilis for this study. Original Olduvai Gorge and East Lake Turkana fossils were examined first-hand, whereas for the Omo and Hadar materials, we relied on our original observations on casts and originals and published reports (Tobias, 1991; Boaz et al., 1977; Kimbel et al., 1997). As in the initial announcement of H. naledi (Berger et al., 2015), in this paper we adopt a conservative approach that follows a more conventional hypodigm, thereby encompassing a maximum amount of variation in this taxon; for a more detailed discussion of the probable hypodigms of early Homo species, see de Ruiter et al., 2017. We therefore include the following fossils in the hypodigm of H. habilis: A.L. 666–1, KNM-ER 1478, KNM-ER 1501, KNM-ER 1502, KNM-ER 1805, KNM-ER 1813, KNM-ER 3735, O.H. 4, O.H. 6, O.H. 7, O.H. 8, O.H. 13, O.H. 15, O.H. 16, O.H. 21, O.H. 24, O.H. 27, O.H. 31, O.H. 35, O.H. 37, O.H. 39, O.H. 42, O.H. 44, O.H. 45, O.H. 62, OMO-L894-1, and Stw 53.

Homo rudolfensis. Samples from Olduvai Gorge, East Lake Turkana, and Lake Malawi were considered as part of the hypodigm of H. rudolfensis for this study. The East Lake Turkana fossils available prior to 2010 were examined first-hand, while for the Olduvai and Lake Malawi fossils and for KNM-ER 60000, 62000, and 62003, we relied on original observations of fossils and casts as well as published reports (Blumenschine et al., 2003; Schrenk et al., 1993; Leakey et al., 2012). As above, and in the initial announcement of H. naledi (Berger et al., 2015), in this paper we adopt a conservative approach that follows a more conventional hypodigm, thereby encompassing a maximum amount of variation in this taxon; for a more detailed discussion of the probable hypodigms of early Homo species, see de Ruiter et al. (2017). We include the following fossils in the hypodigm of H. rudolfensis: KNM-ER 819, KNM-ER 1470, KNM-ER 1482, KNM-ER 1483, KNM-ER 1590, KNM-ER 1801, KNM-ER 1802, KNM-ER 3732, KNM-ER 3891, KNM-ER 60000, KNM-ER 62000, KNM-ER 62003, O.H. 65, and UR 501.

Homo erectus. Samples from Buia, Chemeron, Daka, Dmanisi, East and West Lake Turkana, Gona, Hexian, Konso, Mojokerto, Olduvai Gorge, Sangiran, Swartkrans, Trinil, and Zhoukoudian were included in the hypodigm of H. erectus for the purposes of this study. South African material is of special interest in this comparison because of the geographic proximity, and because of the difficulty of clearly identifying Homo specimens within the large fossil sample from Swartkrans. In particular, the following specimens from Swartkrans are considered to represent H. erectus: SK 15, SK 18a, SK 27, SK 43, SK 45, SK 68, SK 847, SK 878, SK 2635, SKW 3114, SKX 257/258, SKX 267/2671, SKX 268, SKX 269, SKX 334, SKX 339, SKX 610, SKX 1756, SKX 2354, SKX 2355, SKX 2356, and SKX 21204. We considered ‘Homo ergaster’ (and also ‘Homo aff. erectus’ from Wood, 1991) to be synonyms of Homo erectus for this study; Turkana Basin specimens that are attributed to H. erectus thus include KNM-ER 730, KNM-ER 820, KNM-ER 992, KNM-ER 1808, KNM-ER 3733, KNM-ER 3883, KNM-ER 42700, KNM-WT 15000. Olduvai specimens include O.H. 9, O.H. 12 and O.H. 28. Original fossil materials from Chemeron, Lake Turkana, Swartkrans, Trinil, and Dmanisi were examined first-hand by the authors, whereas the remainder were based on casts and published reports (Abbate et al., 1998; Gilbert and Asfaw, 2008; Wood, 1991; Weidenreich, 1943; Suwa et al., 2007; Antón, 2003; Rightmire et al., 2006, 2017; Martinón-Torres et al., 2008 ; Spoor et al., 2007). 

A large number of postcranial specimens have been collected from the Turkana Basin and appear consistent with the anatomical range otherwise found in Homo, and inconsistent with known samples of Australopithecus and Paranthropus from elsewhere. These include KNM-ER 1472, KNM-ER 1481, KNM-ER 3228, KNM-ER 737, KNM-ER 5881 (Ward et al., 2015), and others.

Specimens from the latest Lower Pleistocene and Middle Pleistocene of Europe and Africa that cannot be attributed to H. erectus were also included in the comparisons in this study. These include fossils that have been attributed to H. heidelbergensis, H. rhodesiensis, ‘archaic H. sapiens’ or ‘evolved H. erectus’ by a variety of other authors. Specimens include KNM-ES 11693, Arago 2, Arago 13, Arago 21, Atapuerca 1, Atapuerca 2, Atapuerca 4, Atapuerca 5, Atapuerca 6, Cave of Hearths, Ceprano, SAM-PQ-EH1, Kabwe, Mauer, Ndutu, Salé, Petralona, Reilingen-Schwetzingen, and Steinheim. We also included Neandertal samples from Krapina, Vindija, La Chapelle-aux-Saints, La Ferrassie, Monte Circeo, Saccopastore, and Feldhofer.

Homo floresiensis. The hypodigm of H. floresiensis used in this study includes specimens from Liang Bua, Flores, as described by Brown et al. (2004), Brown and Maeda (2009), Tocheri et al. (2007), Orr et al. (2013), Jungers et al. (2009a, 2009b), Larson et al. (2009), Morwood et al. (2005), Falk et al., 2005, Kaifu et al. (2011), and Kubo et al. (2013).

The LES1 ECV was estimated virtually using 3D surface scans of the reconstructed partial cranium. These methods were similar to those performed on the previously published Dinaledi H. naledi crania (Berger et al., 2015). The LES1 partial cranium was scanned using a NextEngine 3D Scanner and associated ScanStudio HD Pro software. Two 360-degree scans were collected using 16 divisions and 1000 dpi with the cranium in different orientations in order to capture the maximum ectocranial and endocranial surfaces. The two 360-degree scans were then merged and exported as a .ply file. The .ply file was opened in GeoMagic Studio, converted to a point cloud, and re-wrapped to minimize the number of edges in the model (created from the individual surface scans). The LES1 virtual model was then duplicated and mirrored, and the mirror-image (right portion) was aligned with the original model (left portion) using both the Manual and Global Registration functions. In doing this, the program uses an iterative process to find the greatest congruency between the region of overlap in the two models, minimizing deviations. Convergence was detected after 20 iterations, with an average deviation of 0.47 mm and a standard deviation of 0.43 mm. Some deviation is expected given the natural anatomical asymmetry and refitting of the fragments. A deviation map illustrates that areas with the highest deviation were fairly localized (Figure 37). We inspected the mirror image in comparison to the preserved but not refitted fragments that represent portions of the right parietal and right temporal bones of LES1, finding that they are similar in size and curvature. The LES1 original images and mirror-image were then merged.10.7554/eLife.24232.043Figure 37.

Congruency between the original LES1 3D scan and an aligned mirror-image. Scale bar = 2 cm. (A) Frontal view of a 3D scan of the original LES1 specimen (brown) aligned with the mirror-image (grey), illustrating congruency between overlapping regions. (B) Deviation map of the internal view of the frontal region. Deviation scale bar in mm.

DOI: http://dx.doi.org/10.7554/eLife.24232.043

The hominin assemblage from the Lesedi Chamber was analysed using the taphonomic protocols and methods applied to the U.W. 101 Dinaledi fossils, and detailed in Dirks et al. (2015): 22–24 and 32–33. Specimens were viewed macroscopically, and fragments >50 mm diameter were imaged on all surfaces using a Canon 70D DSLR fitted with a Canon EF-S 60mm f2.8 macro lens and ring-flash. Criteria for scoring taphonomic observations are detailed in Supplementary file 5. Additional taphonomic analyses are underway and will be described in future publications.

The taxonomic classification of the micromammal remains follows that of Wilson and Reeder, 2005, using published descriptions and images (Avery, 2007; Coetzee, 1972; Davis, 1965; De Graaff, 1981; Meester and Setzer, 1971; Reppening, 1967; Skinner and Chimimba, 2005), as well as comparisons with a photographic database of southern African rodents from various museum collections. As genus is the lowest taxonomic level at which most material can be identified accurately (Reed, 2005, (2007); Reed and Geraads, 2012), specimens were identified to this level or higher. Macrovertebrate remains were also identified using published reference guides (Walker, 1985) and visual comparisons with mammalian collections at the University of the Witwatersrand. Wherever possible, specimens were attributed to species. In cases where this was not possible, the next highest taxonomic level was used. Identification of the non-mammalian remains was limited to the level of class or order, pending more detailed analyses.

All fossil material from the Lesedi Chamber is available for study by researchers upon application to the Evolutionary Studies Institute at the University of the Witwatersrand where the material is curated. Three-dimensional surface renderings and other digital data are available from the MorphoSource digital repository (http://morphosource.org).