The analysis of evolutionary relationships among hominin taxa is an ongoing process that requires revision and reassessment as new fossil information becomes available. This is particularly true regarding the documentation of new species, given that a number of paleontological discoveries in the last decade have greatly expanded the known diversity of extinct taxa. In particular, the discovery of Australopithecus sediba at the South African site of Malapa has reignited debate over the evolutionary history of the australopiths and the origin of the genus Homo. Evaluation of these evolutionary hypotheses is informed by the phylogenetic analysis of relevant morphological data. Here we incorporate A. sediba into the Mongle et al. (2019) character matrix by conducting independent morphological assessments of the MH1 and MH2 craniodental fossils. We further expand this character matrix by incorporating the Australopithecus anamensis cranium from Woranso Mille (MRD-VP-1/1, Haile-Selassie et al., 2019), as well as the newly described Paranthropus robustus cranial remains from Drimolen (DNH 7, DNH 155; Martin et al., 2021; Rak et al., 2021a).

The original description of A. sediba indicated that it was craniodentally derived toward Homo compared to earlier species such as Australopithecus afarensis and Australopithecus africanus (Berger et al., 2010). For example, in comparison to earlier australopiths, A. sediba was described as displaying a number of purported Homo-like characteristics, such as reduced postorbital constriction and a low degree of subnasal prognathism. However, as Berger et al. (2010) noted, A. sediba differed from later Homo in retaining a small cranial capacity and an australopith-like 'overall body plan'. This mosaic pattern of derived and primitive morphology led to the phyletic hypothesis that the Malapa hominin was appropriately placed within Australopithecus, but uniquely aligned with the genus Homo to the exclusion of other previously recognized australopith species.

Although Irish et al. (2013) proposed alternative phylogenetic relationships for A. sediba based on dental characteristics, the cladistic analyses conducted by Dembo et al. (2015) found results that were broadly consistent with the hypothesis articulated by Berger et al. (2010), in that A. sediba was recovered as the sister taxon to the genus Homo. However, the analysis conducted by Dembo et al. (2015) was not a truly independent test of the initial analysis by Berger et al. (2010), insomuch as the character data employed by Berger et al. (2010) were incorporated into the Dembo et al. (2015) data set. Moreover, the phylogenetic relationship posited by Berger et al. (2010) and corroborated by Dembo et al. (2015) has been questioned on the basis of character validity and redundancy (e.g., Kimbel, 2013; Carter et al., 2014; Mongle et al., 2019). Most notably, Kimbel and Rak (2017) have argued that the key evidence supporting a purported relationship between A. sediba and Homo is a consequence of the subadult status of the MH1 holotype. This contrasts with Carlson et al.'s (2016) geometric morphometric study which concluded that subsequent growth would not have affected the morphological affinities of MH1. However, Kimbel and Rak (2017) note that Carlson et al.'s (2016) results should be interpreted with caution in light of the incompleteness and deformation of several fossil specimens included in their study. Kimbel and Rak's examination of hominoid growth trajectories suggested that later ontogenetic changes in the juvenile MH1 cranium would have ultimately resulted in morphology which would have been "difficult to distinguish from the range of variation expressed by the existing sample of A. africanus adults" (Kimbel and Rak, 2017: 105). This led Kimbel and Rak (2017) to argue that many of the craniofacial dimensions that link A. sediba with Homo reflect the ontogenetic age rather than the phylogenetic affinity of MH1. Kimbel and Rak (2017) concluded that if A. sediba did indeed differ from A. africanus, it shared a special (sister) relationship with it.

While the differential diagnosis of A. sediba cranial morphology has been questioned due to the ontogenetic status of MH1, we continue to recognize the taxonomic validity of A. sediba on the basis of several dental and postcranial morphologies that serve to distinguish it as a species distinct from A. africanus (e.g., Kibii et al., 2011; Kivell et al., 2011; DeSilva et al., 2013) according to a conventional application of the phylogenetic species concept (Cracraft, 1983; see also Kimbel and Rak, 1993). For the purposes of our matrix, there are five dental characters that differ between A. africanus and A. sediba that could not have been impacted by ontogeny (SG 51, SG 52, SG 53, SG 60, and B 61; see Table 1 for character definitions, as well as http://morphobank.org/permalink/?P3730 for full character descriptions). Most notably, A. sediba appears to differ from A. africanus in its dental development. A. africanus has been described as displaying delayed dental maturation of the permanent incisors, premolars and canines relative to that of the M1 (Beynon and Dean, 1988), which is consistent with the delayed pattern exhibited by Pan and Gorilla (Dean, 1985, 1987, 1988; Smith, 1986, Smith, 1994). This contrasts with the accelerated pattern of maturation of the incisors, canines, and premolars in modern humans. By comparison, the extent of P4 development in the MH1 maxilla suggests that A. sediba had a higher degree of concordance in the developmental timing of the permanent dentition than A. africanus, though not so accelerated as Homo ergaster and modern humans (M.C. Dean, pers. comm.).

Of course, the question of the alpha taxonomy of the hominin fossils from Malapa is also of critical importance to address before undertaking a detailed investigation of the phylogenetic relationships of A. sediba. Hominin fossils have been recovered from sedimentary units referred to as facies D, E, and F at Malapa, with those that derive from Facies E and F being considerably rarer than those from Facies D (Val et al., 2018). The elements from Facies D represent the partial skeletons of two individuals, designated MH1 and MH2, with the former designated as the holotype and the latter as the paratype of A. sediba (Berger et al., 2010; Dirks et al., 2010). The specimens from Facies E and F (33 individual bones and fragments representing 7 individuals) were recovered from layered stratigraphic units (or ex situ calcified clastic sediment blocks that are lithological matches for those units) that were deposited atop facies D. Save for the right tibia (MH 4), most of these remains represent infants or duplicate elements from Facies D that comprise the MH1 and MH2 partial skeletons (Val et al., 2018). Although the fossils from Facies E and F have yet to be fully described and analyzed, there is no reason to suspect that they represent a species other than A. sediba.

The fossils that represent MH1 and MH2 have been described in detail in papers on the cranium (Berger et al., 2010; Kimbel and Rak, 2017; de Ruiter et al., 2018), the mandible (de Ruiter et al., 2013), the upper limb (Churchill et al., 2013, 2018a), the hand (Kivell et al., 2011), the pelvis (Kibii et al., 2011; Churchill et al., 2018b), the lower limb (DeSilva et al., 2018), the foot and ankle (Zipfel et al., 2011), ribs (Schmid et al., 2013), and the vertebrae (Williams et al., 2013, 2018, 2021; Meyer et al., 2017; Williams, 2018). These studies have all observed similarities between homologous elements comprising the MH1 and MH2 skeletons.

By contrast, Been and Rak (2014) and Rak and Been (2014) have proposed that the morphology of the lumbar spine and the morphology of the mandibular ramus indicate the presence of two taxa as represented by MH1 and MH2. Been and Rak (2014) identified two indices of vertebral body shape and spinal canal size and the relative size of the articular processes as distinguishing MH1 and Homo erectus from MH2 and Australopithecus. They suggested that MH1 represents Homo and MH2 is attributable to Australopithecus. Rak and Been (2014) and Rak et al. (2021b) have argued that the configuration of the sigmoid notch between the coronoid and condyloid processes and the shape of the anterior margin of the ramus (the presence or absence of a preangular notch) differ between MH1 and MH2. In this case, however, MH1 is seen as resembling australopiths, while MH2 displays the ‘generalized’ morphology of Homo.

Williams et al. (2018) have effectively dismissed the vertebral distinctions recognized by Been and Rak (2014). Likewise, Ritzman et al. (2016) concluded that the magnitude of the distinctions between MH1 and MH2 in ramal morphology did not fall outside the 95% confidence intervals for large samples of extant hominid species. As such, the MH1 and MH2 jaws were acknowledged as representing a single, variable species in rameal morphology. Indeed, Hawks and Berger (2022) have extended the comparisons to fossil samples, showing that the differences between the MH1 and MH2 rami are mirrored by other fossil specimens comprising the H. erectus sample from Tighenif, Algeria (Tighenif 2 and 3), and the pre-Neanderthal (Homo heidelbergensis) sample from Sima de los Huesos, Spain (AT 605 and AT950). The evidence is overwhelmingly supportive of the conspecificity of MH1 and MH2 and their attribution to the species A. sediba.

Similarly, the validity of the commonly accepted hypodigm of fossils from the sites of Sterkfontein and Makapansgat as representing the species A. africanus must also be considered when evaluating any evolutionary relationship between it and A. sediba. Claims that this hypodigm is not monospecific (e.g., Clarke, 1988, 1998, 2002) have been challenged (Grine, 2013, 2019), as there is no consistent evidence related to the skull, dentition, or postcranial skeleton for the presence of two taxa. As such, for the purpose of the current analysis, we continue to treat A. africanus as a single, variable hypodigm as it has been previously defined (Strait et al., 1997; Strait and Grine, 2004; Mongle et al., 2019). Although the recently described Stw 573 (‘Little Foot’) skeleton from the older Member 2 deposits at Sterkfontein has been used to bolster arguments for the existence of two australopiths in this deposit (Beaudet et al., 2019; Clarke and Kuman, 2019), the presence of a species other than A. africanus in the Silberberg Grotto does not necessarily impact the alpha taxonomy of the hominin fossils from Member 4. Regardless, the taxonomic status of this specimen warrants a full analysis by an independent research group, which was beyond the scope of the current study. Future phylogenetic analysis may be warranted following such an assessment.

The objective of the present study is to re-assess hominin phylogenetic relationships, with particular emphasis on the placement Australopithecus sediba. In doing so, we conduct an independent character assessment of the craniodental morphology of A. sediba and evaluate whether the ontogenetic status of MH1 may have affected its purported Homo-like characteristics. We also expand previously available fossil hypodigms to incorporate new discoveries, including the A. anamensis cranium from Woranso-Mille (MRD-VP-1/1, Haile-Selassie et al., 2019), as well as the newly described P. robustus cranial remains from Drimolen (DNH 7, DNH 155, Martin et al., 2021; Rak et al., 2021a) and other fragmentary specimens from the site. Using this updated character matrix, we test the hypothesis that A. sediba is the sister taxon to A. africanus, as well as the alternative hypothesis that A. sediba is uniquely related to the genus Homo to the exclusion of other species of Australopithecus.

Throughout the present analyses, we maintain the same operational taxonomic units employed by Mongle et al. (2019). However, several notable paleontological discoveries have added specimens to the hypodigms of a few constituent species, and these have necessitated a revision to that character matrix. In particular, the addition of the Woranso-Mille cranium (MRD-VP-1/1) has introduced previously missing anatomical information for A. anamensis.

Because Haile-Selassie et al. (2019) characterized the MRD-VP-1/1 cranium using the Strait and Grine (2004) character matrix, states for this specimen were incorporated directly from their Supplementary table 1. However, we did not adopt the proposed modifications to character state definitions therein. This resulted in 42 newly coded characters (see Table 1 for the definition of characters), which expanded the data coverage for A. anamensis from 30% to 65% of the total matrix. In addition to these new character states, the Woranso-Mille cranium introduced variability into the A. anamensis hypodigm for one character. Anterior palate depth (SG 12) was previously coded as shallow (state 0) in A. anamensis on the basis of the morphology present in KNM-KP 29283. However, Haile-Selassie et al. (2019) have characterized MRD-VP-1/1 as deep and "slightly flexed anteriorly … though not extreme, as seen in Paranthropus and Homo" (Haile-Selassie et al., 2019; Supplementary Table 1). As such, we now code A. anamensis as state 1 (variable) for this character.

The expanded Drimolen sample adds information for three characters that were previously unavailable for P. robustus (SG 21, SG 67, SG 36), while introducing variation in several others (SG 13, SG 18, SG 32, SG 68, KRJ 11, KRJ 13). Additionally, we note four character state changes from Mongle et al. (2019) that were necessitated by the observations of DNH 7 (SG 2, SG 27, SG 29, SG 34), as well as the new observations of deciduous teeth from Drimolen (SG 49). These changes are discussed in detail in the (Supplementary Online Material (SOM) S1).

The character matrix has been expanded to include two new characters described by Berger et al. (2010) that are not redundant with those comprising the Mongle et al. (2019) matrix. Characters are referenced by the initials of the authors of key papers in which they were used (e.g., SG, CW, KRJ). These two new characters are indicated in the matrix with a ‘B’ and numbered according to their order of appearance in Berger et al. (2010: Table 1). They relate to the prominence of the zygomatic and the development of grooves on the buccal surfaces of the upper premolars.

Berger et al. (2010) described the development of the zygomatic prominence as either slight or prominent across hominin taxa. Within their matrix, this differentiates australopiths from Homo, with the former displaying an anterior inflation of the zygomatic process (i.e., a zygomatic tubercle) and the latter exhibiting a relatively flat zygomatic body with minimal anterior projection. We have adopted this character (B 35) with modified states (state 0: absent to slight; state 1: present). We also incorporate a new character (B 61) that describes the buccal grooves of the maxillary premolars. However, the states for this character differ slightly from those described by Berger et al. (2010), and incorporate a variable state (state 0: marked; state 1: variable; state 2: weak to moderate).

In addition to these two new characters from Berger et al. (2010), we have revised the definitions of two existing characters in the matrix (SG 26, SG 38) in order to be more consistent with widely adopted cranial metrics. Strait et al. (1997) previously measured postorbital constriction (SG 26) as an index of minimum frontal breadth to superior frontal breadth using available fossil casts. However, these measurements differ slightly from the index of minimum frontal breadth to superior facial breadth (FT–FT) measurements recorded from a larger sample of original fossil specimens by Chamberlain (1987) and Wood (1991). Maintaining consistency with these more widely used, and arguably standardized metrics requires a slight revision to the character states for SG 26, which are now as follows: Index values <67.5% are considered marked (state 0), index values ranging between 67.5 and 77% are considered moderate (state 2), and index values >77% are considered as having slight postorbital constriction (state 3). We include a variable, intermediate state (1), for taxa whose values span the marked and moderate states.

Similarly, we have also modified the definition of SG 38 to be more consistent with the standard measurement of petrous orientation. There is considerable variation in the shape of the anterior aspect of the petrous that affects the position of the petrous apex (the most anterior point on the inferior surface of the petrous). This, in turn, affects the measurement of petrous orientation as defined by Dean and Wood (1981). Strait et al. (1997) measured petrous orientation as the angle between the bi-carotid line and the central axis of the petrous as it tapers anteriorly, although they did not explicitly note this in their character description. This has the effect of making measurements of petrous orientation more coronal in specimens in which the petrous apex is positioned relatively lateral (as when the anterior petrous is ‘squared’ rather than narrowing to a tip). Although we consider the definition used by Strait et al. (1997) to be reasonable, the original definition of Dean and Wood (1981) is the standard measurement used by researchers describing new fossil specimens. We adopt the Dean and Wood (1981) definition of this trait and have re-coded the character accordingly. This requires a slight modification to the states for SG 38, which are now defined as sagittally oriented when petrous orientation is greater than 55° (state 0), and coronally oriented when petrous orientation is less than 55° (state 2). An intermediate variable state (1) is included for taxa which display petrous orientations across the two categories. The resultant changes to the character states of individual taxa are described in the supplementary online material for each of these two characters (SOM S2).

We conducted an independent character assessment of the craniodental morphology of A. sediba using personal observations and measurements of the original MH1 and MH2 fossils housed at the University of the Witwatersrand. The morphological assessment of this taxon was conducted in two stages. First, we coded A. sediba for any characters in which the relevant morphology could be observed on MH1 and MH2 (regardless of ontogenetic status).

Second, following the arguments presented by Kimbel and Rak (2017), as well as an examination of character state transformations between juvenile and adult Pan troglodytes (SOM Table S1), we compiled a list of characters that might be expected to change to a different character state based on hypothesized growth trajectories of MH1 (Table 1). We limited this assessment to characters that were preserved and observable on MH1. Of the 70 characters that could be coded for A. sediba, only seven were recognized as being potentially affected by continued ontogenetic growth to the attainment of adulthood. Of these seven, there are two (SG 20 and SG 21, sagittal crest present in presumptive males, and compound temporo-nuchal crest present in presumptive males) that arguably should not be considered missing data for A. sediba, inasmuch as neither are present in presumptive males of A. africanus, the species to which A. sediba has been likened (Kimbel and Rak, 2017). Nevertheless, to err on the side of caution since we do not actually know the morphology of an adult male cranium of A. sediba in regard to these entheseal crests, we have also treated these as missing data for A. sediba in the final iteration of analysis. This list of characters was used to evaluate the impact of the ontogenetic status of MH1 on the phylogenetic placement of A. sediba.

All character states recorded for A. sediba are listed in Table 1. In total, we were able to evaluate character data for 65 ontogenetically stable characters, representing 61% of the total craniodental character matrix. Our morphological assessments of these characters largely conform to the published character state descriptions of A. sediba (Berger et al., 2010: Table 1), with the following seven exceptions.

Berger et al. (2010) coded the presence of an anterior pillar and canine jugum as a single character for A. sediba. We consider the presence of a canine jugum to be independent from the presence or absence of an anterior pillar (SG 4). Therefore, while we agree that there is a prominent canine jugum present laterally in MH1, we regard A. sediba as lacking an anterior pillar (state 0).

The zygomatic morphology of the MH1 cranium has been described as displaying an affinity with Homo in both its orientation and prominence (Berger et al., 2010). We agree that the zygomatics of MH1 are posterior to the piriform aperture (SG 11, state 0) but differ from Berger et al. (2010), and follow Kimbel and Rak (2017) in recognizing a slight zygomatic prominence formed by its anterior inflation relative to the anterior terminus of the frontozygomatic suture.

Berger et al. (2010) characterized the mandibular symphysis as vertical in A. sediba. However, the degree of symphyseal slope exhibited by both MH1 and MH2 falls within the range of variation encompassed by our determination of the intermediate state for character SG 45. This is especially apparent with regard to the cross-section of the MH2 symphysis.

Berger et al. (2010) described the post-incisive planum as weak in A. sediba. However, our post-incisive planum character (CW 2) describes both the angle and length of the post-incisive planum. Under our definition of this character, the post-incisive planum of A. sediba falls within the intermediate category (state 1). This is consistent with the character state exhibited by most hominins, including both A. afarensis and H. ergaster. Indeed, within our matrix, only Homo sapiens is considered to display a short and steeply inclined post-incisive planum.

Berger et al. (2010) characterized A. sediba as displaying an arched supraorbital contour. However, we consider the supraorbital contour of MH1 to be comparable to that of specimens such as Sts 5 and Sts 71. This differs from the arched supraorbital contour exhibited by specimens such as Stw 505. Accordingly, we have coded A. sediba as state 2 (‘less arched’) for character CW 26.

Berger et al. (2010) described the patency of the premaxillary suture as ‘trace’ in A. sediba. However, we do not distinguish between trace and fully obliterated (only between patent and obliterated). Within the states we have defined, A. sediba displays an obliterated premaxillary suture (state 2).

Lastly, while the pneumatization of the squamous temporal of A. sediba was considered reduced by Berger et al. (2010) based on its external anatomy, Balzeau (2019) has demonstrated the presence of extensive temporal pneumatization in MH1 using synchrotron microtomography. We follow Balzeau (2019) in coding temporal pneumatization as extensive in A. sediba (SG 27, state 0).