The site of Kromdraai (26°00′41″S, 27°44′60″E; altitude: 1470 m) is located in one of the paleokarsts that flank the Blaauwbank River and its tributaries in Gauteng Province, South Africa (Braga et al., 2016a, Braga et al., 2016b; Braga et al., 2017, 2022b, Fig. 1). The highest portion of the Kromdraai paleokarst has suffered considerable erosion through the formation of sinkholes, dissolution and the collapse of cave walls and roof. This erosional process has resulted in the exposure of at least one system of extensive unroofed dolomitic cave chambers, as illustrated in Fig. 1. This vast system was filled with deposits of both consolidated (breccias) and unconsolidated sediments, mainly rudites, containing a remarkable array of micro and macro vertebrate specimens, including early hominins (Broom, 1938a, 1938b, 1941; Brain, 1981; Grine, 1982, 1988; Braga et al., 2013, 2022a, 2023). These deposits, which were previously separated into two distinct localities known as ‘Kromdraai A’ (KA) and ‘Kromdraai B’ (KB), have been integrated into a unified stratigraphic succession (Braga et al., 2022b; Bruxelles et al., 2016). This stratigraphic succession extends from Unit A, which lacks fossils, at its base, to Unit Z at its uppermost layer. Hominin fossils have been found in situ in this stratigraphic sequence, within Units O, P, Q–R and X–Y.

Some fossiliferous deposits appear to have been discovered in 1895 by David Draper, who collected some calcified cave sediment (breccia) containing micromammal teeth from the ‘Kromdraai farm’ and sent it to the British Museum of Natural History (now, the Natural History Museum) in London (de Graaff, 1961; Malan, 1959). However, the site of Kromdraai became known only some 43 years later, when it was brought to the attention of Robert Broom. Broom (1938a) designated the hominin fossils that had been found there by a schoolboy, Gert Terblanche, as the holotype (TM 1517) of P. robustus. Within two years, Broom initiated paleontological investigation at Kromdraai, with an excavation of calcified sediments close to the block from which the holotype had been extracted. His excavations uncovered a juvenile hominin mandible, TM 1536 “within 6 feet of the spot where” the TM 1517 skull had been found (Broom, 1941:607).

Broom thus initiated the first of several phases of excavation and research activity to be conducted at Kromdraai. Subsequently, work was carried out under the direction of C.K. Brain (1955–1956; e.g., Brain, 1958, 1981). Brain was the first to undertake a controlled excavation and produce a map and geological interpretation of the site where he distinguished the KA and KB localities. After a 20-year hiatus in activity, excavations of KB were renewed under the direction of E.S. Vrba (1977–1980; e.g., Vrba, 1981; Partridge, 1982). She established an excavation grid over the site and divided it into separate KB East and KB West deposits separated by a septum of dolomite (Vrba, 1981; Vrba and Panagos, 1982). Vrba's excavation work concentrated on the solid breccia of the KB East using a feather and wedge technique, and from a series of five boreholes drilled through the KB East and KB West breccias recognized five sedimentary members (Partridge, 1982; Vrba, 1981). Two of the five hominin fossils discovered by Vrba (1981)—KB 5223 and KB 5226—derive from a single block of calcified breccia in ‘Member 3’ (Partridge, 1982; Vrba, 1981), subsequently subdivided into two distinct deposits by Bruxelles et al. (2016), and now designated as Units Q–R in the new system of nomenclature for the Kromdraai site (Braga et al., 2022b). ‘Member 3’ also contained a previously unrecognized hominin specimen—KB 6067—discovered in the collections of the Ditsong Museum of Natural History by one of us (J.B.) (Braga et al., 2013). The few other hominin specimens discovered by Vrba (1981) were taken from decalcified ‘Member 3’ deposits. ‘Member 3’ was also the source of most of the faunal elements recovered by Vrba (1981).

The fourth phase of activity at Kromdraai was initiated by J.F. Thackeray (1993–2002) in collaboration with L.R. Berger and N.J. van der Merwe (e.g., Berger et al., 1994; Thackeray et al., 2002, 2003, 2005). A considerable portion of this effort was directed at the first systematic excavation of breccias from KA; their work at KB produced a single isolated and unprovenanced hominin tooth (KB 5503). Thackeray's work at KB yielded the second attempt to establish the geochronological age of the Paranthropus-bearing deposits using paleomagnetic readings from a flowstone and from a series of drill cores (Thackeray et al., 2002). Jones et al. (1986) had produced the first attempt at a paleomagnetic age determination; this was based solely on solid breccia samples and was largely unsuccessful. Although they obtained some reversed polarity readings, most of them were ‘intermediate.’ Thackeray et al. (2002) recorded a normal to reversed sequence within a speleothem located at the eastern end of the site where Units X–Y are exposed (Braga et al., 2022b; Bruxelles et al., 2016). Thackeray et al. (2002) also analyzed seven breccia samples from Units X–Y; four of these displayed normal polarity, while the other three displayed reversed polarity.

Using biochronological estimates which suggested that the KB fauna dated to between ca. 2.0–1.5 Ma (e.g., Delson, 1988; McKee, 1995; Vrba and Panagos, 1982), Thackeray et al. (2002) concluded that their paleomagnetic data recorded an episode of normal polarity within the Olduvai SubChron (C2n) at 1.95–1.78 Ma (Ogg, 2020) and reversed polarities immediately prior to it (C2r.1r at 2.13–1.95 Ma) and after it (C1r.3r at 1.78–1.19 Ma) in the Matuyama chron. Thus, their paleomagnetic results might suggest that Units X–Y were deposited during the Olduvai event. However, given the uncertainties associated with the faunal age estimates, there is little reason to exclude an interpretation that would see these data reflecting instead the Reunion SubChron (C2r.1n; 2.13–2.15 Ma) and the reversed polarities immediately prior to it (C2r.2r at 2.58–2.15) and after it (C2r.1r at 2.13–1.95) within the Matuyama Chron. Indeed, Pickford's (2013) assessment of the Kromdraai suid fauna indicates the presence of Phacochoerus modestus and Metridiochoerus andrewsi (Stage III) from KA, suggesting an age of 1.8–1.6 Ma, whereas Potamochoeroides hypsodon from KB indicates an earlier age for these deposits, equivalent to some of the deposit at Makapansgat. Most faunal estimates place Makapansgat Members 3 and 4 between about 3.0 and 2.6 Ma (Delson, 1988; Vrba, 1982, 1985; White et al., 2006), although Reed's (1996) assessment of the overall large mammal fauna suggests a somewhat older range (ca. 3.3–3.1 Ma). The faunal ages for the fossiliferous units at Makapansgat would thus accord with paleomagnetic observations at the site that might correlate with C2An.2n at 3.207–3.116 Ma or perhaps more reasonably with C2An.1n at between 3.032 and 2.581 Ma (Herries et al., 2013). Thus, there are two possible interpretations of the paleomagnetic results obtained by Thackeray et al. (2002) from KB—and especially those recorded from the speleothem associated with Units X–Y— that would be concordant with the age estimates based on faunal correlations. It is perhaps more conservative to regard the flowstone as recording the Olduvai subchron (1.78–1.95 Ma), as initially interpreted by Thackeray et al. (2002) and Herries et al. (2009:22).

Thackeray's stewardship of research at KB also witnessed the first (and only) foray into dating the site using electron spin resonance (ESR) estimates from mammalian tooth enamel. Curnoe et al. (2002) reported ESR dates for a single bovid tooth from a block of solid breccia that was stratigraphically just above the speleothem examined by Thackeray et al. (2002). Curnoe et al. (2002) obtained an average date of 568 ± 27 ka using an early uptake model, a date of 814 ± 32 ka using a linear uptake model, and an isochron estimate of 668 ka owing to the large difference between two estimates. Although there may be Middle Pleistocene deposits at KA or KB, it is unlikely that these are represented by Units X–Y.

The latest and ongoing phase of systematic excavation at Kromdraai was initiated in 2002 under the auspices of the Kromdraai Research Project (KRP) (Braga et al., 2016b; Braga et al., 2017) (Fig. 1). The first aims of the KRP were to clean the in situ solid breccias that had been exposed by Vrba with acetic acid (10%) and high pressure washing to better reveal their textures in order to identify their stratigraphic context (Braga et al., 2016b; Bruxelles et al., 2016; Braga et al., 2017, 2022a, 2022b). Subsequently, the KA and KB infillings were integrated into a single stratigraphic sequence (Braga et al., 2017, 2022b; Bruxelles et al., 2016).

Work by the KRP soon revealed that fossiliferous sediments extended for a considerable distance to the north and east than had been previously recognized by either Brain or Vrba. Thus far, some 600 m2 of deposits have been exposed in this ‘Extension Site,’ and these correspond to consolidated and unconsolidated sediments forming an extensive talus cone during their accumulation, which was subsequently truncated by erosion. The oldest part of the Kromdraai stratigraphic sequence —Units O and P— is represented in the central part of this talus cone and ‘Extension Site’. Unit P is considered to be equivalent to the decalcified deposits that were designated as ‘Member 2’ by Vrba (1981) (see also Bruxelles et al., 2016; Partridge, 1982). There is also a large stalagmite, the base of which formed on an unconformity separating Unit O from Unit P above (Braga et al., 2022b). The younger sediments that have been designated Units Q–R (previously referred to as ‘Member 3’) are separated from the underlying Unit P by another unconformity and another flowstone that was recognized and illustrated by Vrba (1981). The youngest part of the Kromdraai stratigraphic sequence —Units X–Z—is represented only at the periphery of the site, on its eastern, northern and western ends. This latter end was previously designated ‘KA’.

The Kromdraai hominin assemblage that was available prior to the commencement of excavations by the KRP consists of some 17 individually numbered craniodental specimens and six postcranial bones, five of which may derive from a single individual (TM 1517; Thackeray et al., 2001). The 2014–2017 KRP excavations of the Unit P deposits have produced an additional 30 individually catalogued hominin craniodental fossils (Table 1) that are attributable to P. robustus (see below), 31 bone and stone pseudo-tools (sensu Brain, 1975, Brain, 1981), a number of which were found in situ (Braga et al., 2017). These excavations also uncovered an abundant assemblage of faunal remains, most of which derive from Unit P (e.g., Braga et al., 2022b). The spatial patterning of this fossil hominin and non-hominin assemblage from Unit P was described in Ngoloyi et al. (2020).

The new discoveries of P. robustus from Kromdraai are significant for two reasons. Firstly, these findings entail the first fully adult specimens from this site (e.g., KW 9900). They also constitute the primary sample of this species from Unit P, positioned in the earliest segment of the stratigraphic sequence. Unit P (previously referred to as ‘Member 2’) was considered as ‘sterile’ (Vrba, 1981) prior to the commencement of new excavations at Kromdraai in 2014. Unit P is separated from the stratigraphically younger Units Q–R, where KB 6067 was discovered (Braga et al., 2013), by an unidentified time span and a layer of flowstone, as noted by Vrba (1981) and Braga et al. (2017, 2022b). Therefore, the new P. robustus assemblage from Kromdraai Unit P predates all prior conspecific specimens found at this site, including the holotype of this species, TM 1517, and other specimens recovered from this site before 2014.

Among the 30 fossil hominins from Kromdraai Unit P described in this study (Table 1), seven specimens have featured in one or more research articles in which they have been attributed to P. robustus (Braga et al., 2016b, 2021, 2022a, b, 2023; Zimmer et al., 2023), but none has been afforded complete morphological description. In particular, the bony labyrinths contained within the KW 9600 partial cranium, the KW 9700 petrosal and the KW 9900 and KW 10840 cranial fragments were included in Braga et al. (2021, 2022a) in their analyses of cochlear shape and semicircular canal allometry. Similarly, the KW 6420 mandible, KW 9000 maxillae, KW 9600 partial cranium and the KW 10840 cranial base with associated teeth have been employed in analyses of early ontogenetic development in P. robustus (Braga et al., 2023) and symphyseal morphology (Zimmer et al., 2023). The specimens, which were not previously classified as P. robustus before this study, were taxonomically attributed based on the dental and cranial morphological features outlined in this study.

The aim of this paper is to provide detailed description, illustration, and relevant mensural data of all the new P. robustus craniodental fossils recovered between 2014 and 2017 from the Unit P deposits at Kromdraai. This study is also intended to provide researchers with a comprehensive list of the P. robustus specimens available, their preservation and association in order to assess the minimum number of individuals (MNI). Together with the descriptions, basic measurements and summary statistics are provided. The enamel-dentine junction of the dental remains examined here will be presented in a separate paper. The basic mesiodistal (MD) and buccolingual (BL) diameters of the P. robustus dental sample from Kromdraai Unit P are compared with the other relevant South African samples. This is done to place the new P. robustus sample from Kromdraai in a broader perspective and to begin to explore its significance relative to conspecifics from Swartkrans, Drimolen, Cooper's, Gondolin and Sterkfontein Member 5. We discuss our results in the context of previous interpretations of differences among P. robustus samples due to sexual dimorphism (Lockwood et al., 2007; Rak et al., 2021), the presence of a South African ‘robust’ lineage (Braga et al., 2013, 2022a; Martin et al., 2021) and differences in age distribution, or ‘paleodemography’ (Mann, 1975; Riga et al., 2019).