The origins of the genusHomo remain elusive. Over 60 years ago, Leakey et al.¹ proposed the speciesHomo habilis with the discovery of fossils at Olduvai Gorge in Tanzania; they proposed that the species occupied a morphological gap betweenAustralopithecus andHomo erectus, and placed it at the base of the genusHomo. While the naming of this new species was initially controversial, subsequent discoveries of earlyHomo specimens at Olduvai, as well as other sites in Kenya, South Africa, and Ethiopia, gradually led to the general acceptance ofH. habilis as a valid species^2,3,4,5,6,7. Despite this acceptance, a number of key questions about earlyHomo remain unresolved, and in particular, we have a poor understanding of the phylogenetic relationship betweenH. habilis and other species. Additionally, recent work has emphasised hominin species diversity in the Pliocene and Pleistocene, including the co-existence of multiple species in close proximity^8,9,10,11,12, and thus a once influential account of hominin evolution, according to whichAustralopithecus evolved intoH. habilis, thenH. erectus, and ultimately modern humans^13,14, is now considered overly simplistic^15,16. In the case of the earlyHomo fossil record; a number of studies have suggested that the variation exceeds that expected of a single species and that a second species,Homo rudolfensis, is also present^{17,18,19,20,21}. Further support for the existence of this second species comes from a reconstruction of the mandible of the type specimen ofH. habilis, OH 7, which found the specimen shows a primitive long and narrow dental arcade that is incompatible with severalH. rudolfensis specimens²². Further, our understanding of the origins of the genus has been impacted by several discoveries ofHomo fossils older than 2 million years ago (Ma)^{7,23,24,25,26,27,28}, including a mandible from Ledi Geraru, which represents the earliest knownHomo at 2.8 Ma²⁹. It is also worth noting that some suggest that specimens currently assigned toH. habilis (particularly OH 13, OH 24, OH 62, and KNM-ER 1813) may in fact be better classified asAustralopithecus due to a lack of synapomorphies with laterHomo^30,31,32. Others have suggested that the entire hypodigms ofH. habilis and/orH. rudolfensis should be transferred out of the genusHomo^{10,33,34,35,36}, although several phylogenetic analyses support the monophyly of the genus as currently defined³⁷, with the inclusion ofAustralopithecus sediba^38,39, or with the exclusion of eitherHomo floresiensis³⁸ orHomo naledi³⁹.

TheH. habilis hypodigm includes numerous dental remains, with OH 7 preserving a nearly complete mandibular tooth row, OH 13 and OH 16 preserving most mandibular and maxillary teeth, and OH 24 preserving most postcanine maxillary tooth positions. Several features of the dentition have been suggested to be derived relative toAustralopithecus; the initial description ofH. habilis highlighted relatively large anterior teeth and buccolingually narrow postcanine teeth¹, while subsequent work has highlighted the small size of the postcanine teeth, the reduced premolar talonids, and a reduced M₃ that is similar in size to the M₂⁵. However, the usefulness of a number of these traits in distinguishingH. habilis fromAustralopithecus has been questioned^22,40,41, andH. habilis teeth are considered by some to be relatively generalised in their morphology¹⁹. Importantly, these studies have only considered the outer-enamel surface (OES) morphology of these specimens, the original morphology of which is frequently altered or removed by the effects of occlusal dental wear. For example, the lower first molars of OH 7, OH 13, and OH 16 are moderately to heavily worn, and consequently, analyses of crown morphology have focused on 2D measurements of the occlusal surface (e.g., linear dimensions, relative cusp areas, and crown outline shape). Critically, these measurements do not capture the vertical components of both crown height and cusp height, which have been shown to be useful in distinguishingAustralopithecus fromHomo^42,43.

In this study, we overcome the limitations of variable dental wear and 2D measurements through an analysis of the dentine crown that underlies the enamel cap. Hereafter referred to as the enamel-dentine junction (EDJ), it is a surface that is established early in tooth development and preserves the form of the basement membrane upon which enamel is deposited. In addition to retaining the vertical components of crown and cusp height, it is where the majority of features that are known to carry taxonomic relevance are formed during odontogenesis. Since the EDJ is not remodelled throughout an individual’s lifetime, it can provide insights into dental morphology that would otherwise not be possible due to the effects of tooth wear. A number of studies have demonstrated that the morphology of the EDJ is taxonomically informative in extant primates^42,44,45,46 and fossil hominoids^{43,47,48,49,50,51}, including a recent study in which Zanolli et al.⁵² suggested that theHomo status of a number of South African fossils from Sterkfontein, Drimolen and Swartkrans is not supported based on the morphology of the EDJ. Fossils from Olduvai Gorge are crucial to the issues surrounding earlyHomo systematics, but little is known about their internal dental morphology.

Here we assess the EDJ morphology of key specimens from Olduvai (Figs. 1 and2), including the type specimen ofH. habilis, OH 7, as well as earlyHomo fossils from Koobi Fora. Using 3D landmark-based geometric morphometrics (GM), we characterise the major EDJ shape differences betweenAustralopithecus and laterHomo (modern humans, Neanderthals, and Middle-PleistoceneHomo; Table 1). We find that keyH. habilis specimens including OH 7 retain the primitive condition and that several features previously suggested to characterise the dentition ofH. habilis are not evident at the EDJ. We also assess the homogeneity of theH. habilis hypodigm from a dental perspective, finding high levels of variation that in some cases exceed our expectations of a single species.

ThreeH. habilis tooth rows are shown, OH 7, OH 13, OH 16, OH 39. Each is shown in the occlusal view at the outer-enamel surface (left) and enamel-dentine junction (right). All teeth are shown as right-sided. Dating and stratigraphy information from refs.^69,72,73.

Full size image

ThreeH. habilis tooth rows are shown, OH 7, OH 13, and OH 16, as well as OH 22 which is attributed toH. erectus. Each is shown in the occlusal view at the outer-enamel surface (left) and enamel-dentine junction (right). All teeth are shown as right-sided. An example of the landmarking protocol for molars is shown in the bottom right (Red = EDJ fixed landmarks, yellow = EDJ ridge semilandmarks, blue = CEJ ridge semilandmarks). Dating and stratigraphy information from refs.^{69,72,73,87,88}.

Full size image

Overall shape patterns

Principal component analyses (PCA) summarise EDJ shape variation for three representative tooth positions; P₃ (Fig. 3a), P⁴ (Fig. 4a), and M¹ (Fig. 5a). PCAs for all postcanine tooth positions can be found in Supplementary Fig. 7 (interactive html versions of these plots are available in Supplementary Data Files 1–10). In each case, the first principal component (PC1) separatesAustralopithecus (Australopithecus afarensis andAustralopithecus africanus) from the laterHomo sample. For most tooth positions, and particularly the premolars, PC1 represents a high proportion of the total shape variation present in the sample. This corresponds to a number of shape differences present throughout the postcanine dentition of the laterHomo group that distinguish them fromAustralopithecus. These shape changes are summarised in Table 2 (see also Supplementary Tables 4 and5) and visualised using wireframes (Figs. 3b,4b,5b; Supplementary Figs. 8 and9). One notable shape difference betweenAustralopithecus and laterHomo is a relative increase in the height of the dentine body (see Supplementary Fig. 5 for an explanation of the terminology used to describe crown height at the EDJ) in the latter group; this is present across all postcanine tooth positions, although it is less pronounced in the upper molars. Further shape differences in laterHomo, when compared withAustralopithecus, include a reduction in the talon/talonid region in the premolars, a relatively taller and more distally placed protoconid in the lower molars, and a reduction in the distal marginal ridge in the upper molars (Table 2). The difference in EDJ shape between laterHomo andAustralopithecus is significant for all tooth positions in permutation tests (Supplementary Table 6).

a PCA plot of P₃ EDJ shape.Homo habilis specimens are represented by green points.Aus = Australopithecus. PC = principal component.b Wireframe images showing shape changes associated with the first principal component, showing landmark positions associated with extremes of PC1 (±1.5 standard deviations from the mean) in occlusal view (top) and lingual view (bottom).c Frequency plot of Procrustes distances between all possible pairs of individuals within (coloured lines) and between groups (grey fill). Vertical lines show the 95% limits of the within-group distributions. Below is shown the Procrustes distance between OH 16 and otherH. habilis or earlyHomo specimens. Red = Gorilla gorilla, Green = Pan troglodytes, Purple = Homo sapiens.H. sap = H. sapiens.d Same asc for dentine body height. Source data are provided as a Source Data file.

Full size image

a PCA plot of P⁴ EDJ shape.Homo habilis specimens are represented by green points.Aus = Australopithecus. PC = principal component.b Wireframe images showing shape changes associated with the first principal component, showing landmark positions associated with extremes of PC1 (±1.5 standard deviations from the mean) in occlusal view (top) and lingual view (bottom).c Frequency plot of Procrustes distances between all possible pairs of individuals within (coloured lines) and between groups (grey fill). Vertical lines show the 95% limits of the within-group distributions. Below is shown the Procrustes distances between OH 16 and otherH. habilis or earlyHomo specimens. Red = Gorilla gorilla, Green  = Pan troglodytes, Purple = Homo sapiens.H. sap = H. sapiens.d Same asc for dentine body height. Source data are provided as a Source Data file.

Full size image

a PCA plot of M¹ EDJ shape.Homo habilis specimens are represented by green points.Aus = Australopithecus. PC = principal component.b Wireframe images showing shape changes associated with the first principal component, showing landmark positions associated with extremes of PC1 (±1.5 standard deviations from the mean) in occlusal view (top) and lingual view (bottom).c Frequency plot of Procrustes distances between all possible pairs of individuals within (coloured lines) and between groups (grey fill). Vertical lines show the 95% limits of the within-group distributions. Below is shown the Procrustes distances between OH 16 and otherH. habilis or earlyHomo specimens. Green = Pan troglodytes, Purple =Homo sapiens.H. sap = H. sapiens.d Same asc for dentine body height. Source data are provided as a Source Data file.

Full size image

In these features that serve to distinguish betweenAustralopithecus and laterHomo at the EDJ, a number ofH. habilis specimens retain the primitive condition, including the holotype, OH 7. For example, the mandibular premolar talonids are not reduced and the relative dentine body height is either short (M₂) or intermediate (P₃, P₄, M₁). Table 2 summarises the condition observed in selected OlduvaiH. habilis specimens for these traits, and Supplementary Table 4 for specimens from Koobi Fora and A.L. 666-1. Other Bed I Olduvai specimens OH 4, OH 24, and OH 39 also approximate theAustralopithecus condition, as do Koobi ForaH. habilis specimens KNM-ER 1502, KNM-ER 1802 and KNM-ER 1813 from Koobi Fora. This is evident from the GM analysis, in which these specimens plot closely to specimens ofAustralopithecus. Furthermore, we found only a few significant differences in postcanine tooth shape or size betweenAustralopithecus andH. habilis in permutation tests (Supplementary Table 6). Several tooth positions of KNM-ER 5431 (Hominini gen. et sp. indet.) and A.L. 666-1 from Hadar (Homo aff.H. habilis) also show similar EDJ morphologies toAustralopithecus.

Among specimens commonly attributed toH. habilis, the most derived condition is seen in OH 16. This is particularly evident in the premolars, which show a tall relative dentine body height, reduced talons/talonids (P₃, P⁴), and an interrupted P₃ mesial marginal ridge, features that are frequent in laterHomo. The molars also show a tall dentine body, as well as taller molar dentine horns, particularly the mandibular molar protoconids and maxillary molar protocones (M² and M³; Table 2). These features contribute to the position of OH 16 in PCAs, which for some tooth positions is closer to laterHomo thanAustralopithecus. However, it is important to note that there are also a number of primitive aspects of the EDJ morphology in OH 16, such as the shape of the EDJ marginal ridges in the P₄, M₁, M¹, and P³, which are similar to that seen in otherH. habilis specimens andAustralopithecus. OH 13 shows a mixed pattern; the P₃ and M₂ are most similar to OH 7 and KNM-ER 1802, while the M² and M³ appear to be more derived, similar to OH 16.

The overall shape differences (measured using pairwise Procrustes distances; Figs. 3c,4c,5c; Supplementary Fig. 10) between OH 16 and otherH. habilis specimens are mostly within the 95% limit of extant groups, although P₄ shape is distinct from KNM-ER 1802, the shape of the M¹ is distinct from OH 21, and the P⁴ is distinct from KNM-ER 1813. However, the clearest differences between OH 16 and otherH. habilis specimens are in relative dentine body height. When this feature is considered in isolation, OH 16 is distinct from OH 7 (P₃-P₄), OH 21 (M¹), OH 39 (P³-P⁴), KNM-ER 1802 (P₃-P₄), KNM-ER 1813 (P⁴-M²) and KNM-ER 1502 (M₁). In these cases, the difference in relative dentine body height between OH 16 and other specimens attributed toH. habilis exceeds that which would be expected of a single species (Figs. 3d,4d,5d; Supplementary Fig. 11).

Some tooth positions provide better distinction between specimens ofAustralopithecus and earlyHomo than others in PCA plots. For example, while the M₂s ofH. habilis andH. erectus largely overlap with those ofAustralopithecus, there is a better distinction in other tooth positions, particularly the M³ (Supplementary Fig. 6) despite their generally more variable shape. This is well illustrated in the example of KNM-ER 1813. The P⁴, M¹, and M² of this specimen plot withAustralopithecus, while the M³ is clearly distinct. In this case,H. habilis specimens occupy the positive end of PC3, which corresponds to a mesiodistally shortened crown, with an asymmetrical EDJ ridge due to particularly reduced mesiolingual and distobuccal crown corners. The M² also provides a better distinction betweenH. habilis andAustralopithecus in most cases; only KNM-ER 1813 plots within theAustralopithecus range. This is also reflected in permutation tests, which find that the mean shapes of the M² and M³ differ significantly betweenAustralopithecus andH. habilis. There was also a significant mean difference between these two groups in P₃ shape; this is driven partly by the derived condition seen in OH 16, but otherH. habilis P₃s also have a slightly taller dentine body height than inAustralopithecus, as well as a more symmetrical crown base, and an internally placed protoconid dentine horn tip (this feature could not be assessed in OH 16 due to wear).

Canonical variate analysis (CVA) was carried out to further investigate shape differences betweenAustralopithecus,H. habilis, andH. erectus (laterHomo was excluded in this case in order to provide the best distinction between earlyHomo andAustralopithecus). CVA results are shown in Supplementary Figs. 12 and13, and cross-validated CVAs (cvCVA) in Supplementary Fig. 14. CVA plots show the distinction betweenAustralopithecus andH. habilis in several tooth positions, however, in the majority of cases, separation between these groups is severely reduced or removed in cvCVAs, suggesting they likely represent spurious group differences. The differences betweenH. habilis andAustralopithecus in the upper molars are maintained in cvCVAs however, indicating these represent real shape differences. In the M³, theH. habilis specimens are distinguished fromAustralopithecus along CV2, which represents several of the same features described above (a mesiodistally short crown, asymmetrical at the EDJ ridge). The M²s are similarly reduced mesiodistally, with an asymmetrical EDJ ridge due to a reduced mesiolingual and distobuccal crown corner, while the M¹s are distinguished along a combination of CV1 and CV2, corresponding to subtle differences in EDJ ridge shape around the metacone, crown height on the lingual side of the crown, and the shape of the cementum-enamel junction (CEJ; although it is worth noting that the shape difference betweenAustralopithecus andH. habilis M¹s were not significant in permutation tests).

SeveralH. erectus specimens included here also plot more closely toAustralopithecus than laterHomo in PCAs, although this varies by specimen and tooth position, and specimens from Sangiran are better distinguished fromAustralopithecus than those from Koobi Fora. KNM-ER 1507 is the least derived; the specimen plots within theAustralopithecus range along the first two PCs for P₃-M₂, and shows no talonid reduction in the premolars, a short or intermediate dentine body height, and a mesiodistally elongated crown in M₁ and M₂ (Supplementary Table 5). These features are more similar toAustralopithecus orH. habilis than to otherH. erectus, particularly specimens from Sangiran. KNM-ER 992 is more derived, and although several tooth positions overlap withAustralopithecus along PC1, there is a moderately tall dentine body in each tooth, there is some reduction of the premolar talonids, and the M₂ and M₃ show more rounded EDJ marginal ridges. Other Koobi ForaH. erectus specimens plot either on the periphery or just outside theAustralopithecus range in the first 2 PCs (KNM-ER 3733 P³, P⁴, KNM-ER 806 M₁) or are distinguished along PC2 (KNM-ER 806 M₂, M₃, KNM-ER 807 M₁, KNM-ER 1808 M₂) or PC3 (KNM-ER 807 M₃, KNM-ER 1808 M²; Supplementary Fig. 15). OH 22 is well-distinguished from other specimens within the sample in PCAs, including those attributed toH. erectus. In particular, the specimen occupies a positive position along PC2 for P₃-M₁, and a positive position along PC3 for M₂ (Supplementary Fig. 15). This reflects a number of features across the tooth row, including the presence of a low mesial marginal ridge in both the premolars and molars, a relatively elongated mesial fovea (present in all teeth, but particularly notable in the P₃) and a short metaconid.H. erectus specimens are clearly distinct from bothAustralopithecus andH. habilis in CVAs, and these group differences are maintained in cvCVAs for most tooth positions, except for the P³ where there is overlap between all three groups and the M² where KNM-ER 1808 and Sangiran 4 overlap withAustralopithecus. This distinction corresponds to a number of features across the dentition including premolars with a more symmetrical, oval-shaped CEJ (crown base), more rounded EDJ marginal ridges molars, and a reduced maxillary molar hypocone.

Typicality-based classifications for unclassified specimens are presented in Supplementary Table 7. KNM-ER 5431 is mostly classified asAustralopithecus. The premolars of KNM-ER 1590 and A.L. 666-1 are also classified asAustralopithecus, however, the molars are frequently not classified into any of the groups. KNM-ER 1507 is classified mostly asH. habilis orAustralopithecus. L26-1g, L398-573, and SK 847 are most often not classified into any group, while SK 27 is most often classified asHomo erectus, and Omo 75i-1255 is most often classified asH. habilis. OH 22 plots either close toH. erectus, or occupies a more extreme position along the CV that best separatesH. erectus from other taxa. In particular, the P₄ has an oval-shaped CEJ that is mesiodistally compressed and a clearly reduced talonid, while the M² has a rounded EDJ ridge, which is similar to otherH. erectus specimens but more extreme in their expression. As a result of these features, the teeth are either classified asH. erectus or none of the groups.

Variation in CEJ shape in each anterior tooth position is summarised in Supplementary Fig. 16. The mandibular incisors provide a good distinction betweenAustralopithecus and laterHomo along PC1; this refers to the extent of the apical extension of the enamel on the labial and lingual sides of the tooth, which is more marked inAustralopithecus, giving the CEJ a sinusoidal shape. In this respect, OH 7 and OH 16 retain theAustralopithecus condition, while twoH. erectus specimens (KNM-ER 820 and KNM-WT 15000) show a much flatter CEJ, even more so than in laterHomo specimens. This CEJ shape difference is less marked in the other anterior tooth positions, and as a result, there is more overlap between groups in these cases.

Buccolingually narrow postcanines

At the enamel surface, mostH. habilis postcanine teeth are buccolingually narrower than those ofAustralopithecus, particularly in the mandibular dentition (Supplementary Tables 1 and2). However, this pattern is less evident at the EDJ, instead, there is variation by specimen, tooth position, and by which part of the EDJ is being considered (Supplementary Figs. 1 and2). This trait is discussed in detail in Supplementary Note 1.

Size patterns across tooth row

Figure 6 shows the cervical size patterns across the tooth row inH. habilis when compared withAustralopithecus, H. erectus, and laterHomo (comparisons betweenH. erectus and other groups, includingHomo sp., are available in Supplementary Fig. 17, interactive html versions of each centroid size plot are available in Supplementary Data files 11-20 and all centroid size data is available in Supplementary Data 21). The size of the mandibular teeth ofH. habilis show a large degree of overlap with those ofAustralopithecus, and we found no significant size differences between the two taxa (Supplementary Table 6). In OH 7, all preserved tooth positions are within the range ofAustralopithecus. A possible exception to this pattern is OH 13, which is the smallestH. habilis specimen included here, and for which the canine and P₄ are outside theAustralopithecus range of variation. OH 13 also differs from otherH. habilis specimens in having a larger P₃ than canine; in OH 7 and OH 16 the P₃ is smaller. The P₃ > C pattern seen in OH 13 is similar to that seen in a number ofH. erectus specimens (KNM-ER 820, 992, 1507, and KNM-WT 15000). There is also an overlap in maxillary tooth size betweenH. habilis andAustralopithecus, although the postcanine teeth occupy the lower end of theAustralopithecus range, and some M²s and M³s are smaller. The teeth of OH 16 are large, similar to otherH. habilis andAustralopithecus. The anterior teeth inH. habilis are variable; OH 15 (C) and OH 16 (I¹) occupy the larger end or are slightly above, theAustralopithecus range of variation, while KNM-ER 1813 (I²-C) and OH 39 (I¹, C) occupy the lower end or are slightly below theAustralopithecus range.H. habilis specimens mostly have maxillary premolars that are similar in size (or in OH 39, a P⁴ that is slightly smaller), whereasAustralopithecus usually has a P⁴ > P³ pattern. However, there are also someA. afarensis specimens in which the P³ is larger (A.L. 199-1 and A.L. 200-1a).

Plots show the natural logarithm of the centroid size of the CEJ ridge landmark set, calculated separately fora the maxillary tooth row, andb the mandibular tooth row. In both plots,Australopithecus is shown in red, laterHomo in purple, andHomo habilis in green. Shaded areas represent the range of centroid sizes for each group, and lines connect teeth from the same individual.H. habilis specimens that preserve complete or nearly complete tooth rows are labelled and connected using black lines. Source data are available in Supplementary Data 21.

Full size image

The mandibular teeth ofH. habilis are generally larger than those of laterHomo, although the difference is only significant in the posterior tooth positions for which there are larger sample sizes (Supplementary Table 6). The maxillary teeth ofH. habilis show more overlap with those of laterHomo, but the premolars and molars were still found to be significantly different. There were no significant differences betweenH. habilis andH. erectus, although a number of SangiranH. erectus specimens and KNM-ER 820 (P₄-M₂) are below theH. habilis range. Most postcanine tooth positions inH. erectus were found to be significantly different to laterHomo (except M¹ and M³), and only a small number of tooth positions (M³, M₂, and M₃) were significantly different toAustralopithecus. The size of the teeth of OH 22 and OH 23 is similar to otherH. erectus specimens (Supplementary Fig. 17), except that both show a M₁ > M₂ pattern, whereas KNM-ER 806, KNM-ER 992, KNM-ER 1507, KNM-WT 15000 and Sangiran 1b show the opposite pattern. The teeth of KNM-ER 1590 are large compared to those ofH. habilis, and it has a canine that is similar in cervix size to the P³, which is rare inAustralopithecus, and not present in either of theH. habilis specimens that preserve these teeth (OH39 and KNM-ER 1813), or KNM-WT 15000, but is seen in some laterHomo. In A.L. 666-1, the premolars are slightly larger thanH. habilis, but the M¹ and M² are within theH. habilis range.

A suite of EDJ features distinguishes the postcanine teeth of laterHomo from those ofAustralopithecus (A. afarensis andA. africanus), including the relative height of the dentine body and reduction of the premolar talons and talonids. However, when considering these traits in earlyHomo, we find that keyH. habilis specimens such as OH 7, OH 24, and KNM-ER 1813 largely retain theAustralopithecus condition. The EDJ shape and size patterns across the tooth row in theseH. habilis specimens accord more closely withAustralopithecus than laterHomo, and we fail to find statistically significant differences between ourH. habilis andAustralopithecus samples in size in any tooth position, and only find significant shape differences in three tooth positions (P₃, M², M³) (Supplementary Table 6).

In fact, we find theH. habilis M² and M³ shapes to be distinct from those of other groups more broadly.Homo habilis M²s (n = 4), and especially M³s (n = 4), have a mesiodistally short, asymmetrical EDJ ridge. In the M²s, this is due to a reduced crown distobuccally, while the M³s are reduced mesiolingually and distobuccally. The M³s also frequently show a metacone dentine horn that is mesially and internally placed, but this is less pronounced in the M²s. This combination of features drives the separation ofH. habilis in CVAs for both tooth positions, and the separation of M³s along PC3 of the PCA (Supplementary Fig. 7), and distinguishes them fromAustralopithecus, laterHomo, andH. erectus (H. erectus M³s also have an EDJ ridge that is shortened mesiodistally, but they are less asymmetrical and more rounded in occlusal view). This morphology is evident in some of the generally more derivedH. habilis specimens (OH 13 and OH 16) but is also present in the M³ of KNM-ER 1813, which is otherwise veryAustralopithecus-like. Distal molars are known to be highly variable, while the M¹ in particular has a more stable morphology^53,54. This lower level of variation in the M¹ can be considered useful for hominin systematics as it reduces intraspecific variation^55,56,57,58, however in this case it seems that differences in EDJ morphology between our samples ofH. habilis andAustralopithecus are clearer in the distal molars (M² and M³). This could suggest that relatively minor morphological differences in the M¹s may be exaggerated along the molar row such that they are clearer in the M² and M³, although this requires further investigation. There was also a significant difference in P₃ shape between our samples ofH. habilis andAustralopithecus; driven by a slightly increased dentine body height (most extreme in OH 16), symmetrical crown base, and internally placed protoconid dentine horn tip.

There are derived OES dental traits inH. habilis, including a buccolingual narrowing of the postcanine dentition^1,41,59 (Supplementary Tables 1 and2). However, at the EDJ this feature is less clear and is variable by specimen, tooth position, and by which part of the tooth crown contributes to the shape difference. It is possible that differences in the distribution and relative thickness of enamel across the crown contribute to this feature, although this requires further investigation. Similarly, we do not find consistent evidence for large anterior teeth inH. habilis at the crown base, when compared withAustralopithecus or laterHomo (Supplementary Table 6). For a more detailed discussion of both of these traits, see Supplementary Note 1. Quam et al. also found that the M¹ of KNM-ER 1813 and OH 21 show derived relative cusp areas compared withAustralopithecus and otherH. habilis specimens⁵⁵; this is not clear at the EDJ, where both specimens show a primitive overall M¹ shape in PCAs, although the relationship between EDJ morphology and relative cusp areas is likely to be complex, with contributions from multiple aspects of EDJ shape, as well as the thickness and regional distribution of enamel. It is possible that there are additional dental traits not identified here that are autapomorphic forH. habilis, or that effectively distinguish earlyHomo fromAustralopithecus. In particular, this may be the case for aspects of morphology that are not captured in our GM analysis (such as the occlusal basin or lateral faces of the EDJ), or discrete dental traits. However, as shown recently for lower molar accessory cusps, such discrete dental traits frequently show complex expression patterns that can be problematic in studies of hominin systematics⁶⁰.

The lack of clearly derivedH. habilis dental traits at the EDJ in most tooth positions, combined with the overall primitive postcanine EDJ morphology in a number of keyH. habilis specimens, suggest an overallAustralopithecus-like endostructural dental morphology in the earliest members of the genusHomo. Some of the specimens displaying a more primitive morphology, such as KNM-ER 1813 and OH 24, have been suggested by some authors to belong toAustralopithecus rather thanHomo^30,31,32. OH 7 is important in this respect, however, as it combines a largelyAustralopithecus-like dental and gnathic pattern with an enlarged endocranial volume that is consistent with other early members of the genusHomo^22,61. A recent study by Zanolli et al.⁵² analysed the EDJ morphology of 23 specimens from South African cave sites that have been suggested to belong to the genusHomo, finding that only four of the specimens analysed could be unambiguously attributed toHomo, while others show a dental pattern clearly more similar toAustralopithecus orParanthropus. Unlike eastern African specimens such as OH 7, OH 24, and KNM-ER 1813, many of the southern African specimens analysed are isolated teeth, and therefore assessments of their taxonomic status necessarily rely heavily on dental morphology. Since the eastern African Early Pleistocene hominin fossils generally preserve more anatomical structures across the skeleton, future studies will have to integrate all the available morphology: dental, cranial, and postcranial, when assessing the taxonomic affinities of individual hominin specimens.

Alternatively, some have advocated for the entire hypodigms ofH. habilis and/orH. rudolfensis to be transferred out of the genusHomo^{10,33,34,35,36}. They suggest that the inclusion of these species inHomo could render the genus paraphyletic, which is supported by some older phylogenetic analyses^33,62. However, it is important to mention that several more recent phylogenetic studies have supported the monophyly ofHomo even with the inclusion ofH. habilis/rudolfensis^37,63, although in certain instances, maintaining this monophyly might necessitate reconsideration of the generic status ofA. sediba,H. naledi and/orH. floresiensis^38,39. It has also been argued that species included inHomo should occupy a coherent adaptive grade³³, and that plesiomorphic features across theH. habilis skeleton (e.g., refs.^22,64,65,66) suggest that the adaptive strategy ofH. habilis is more similar toAustralopithecus thanH. sapiens^33,36. Others have argued against the requirement for genera to be adaptively unified¹⁶, emphasising that the adaptations seen inH. sapiens evolved in a mosaic, stepwise pattern and that the distinction betweenAustralopithecus and the earliest members of the genusHomo is necessarily slight^15,16. The results presented here highlight a number of plesiomorphic features across theH. habilis dentition that emphasise the species’ morphological and adaptive distinction from later members of the genusHomo, but whether this justifies removing the species fromHomo depends on the definition of the genus.

When consideringH. erectus, there is also some overlap withAustralopithecus in PCAs for several tooth positions, however,H. erectus is better distinguished fromAustralopithecus in CVA analyses, and there are significant differences in EDJ shape between the two taxa in all mandibular tooth positions, as well as P⁴ and M².H. erectus specimens are distinguished fromAustralopithecus by a number of features, including a reduction of the talon/talonid and a more symmetrical CEJ shape in the premolars (except P³), and more rounded EDJ ridges in the molars. There are also significant differences in size for some posterior molars (M³, M₂, M₃). Overall, these differences are more pronounced than the differences betweenAustralopithecus andH. habilis, and are indicative of a more derived dental morphology inH. erectus, even in the Early Pleistocene specimens included here. The small size of the posterior molars is consistent with the finding thatH. erectus is derived in having relatively reduced M₃s (relative to the M₁), while earlyHomo specimens are more similar toAustralopithecus⁶⁷ (although theH. erectus sample in ref.⁶⁷ is mostly younger than the Early Pleistocene specimens included here). OneH. erectus specimen, KNM-ER 1507, was found to have a more primitive EDJ morphology, with P₃-M₂ resemblingH. habilis orAustralopithecus more than other specimens ofH. erectus, which is supported by CVA-based classifications (Supplementary Table 7). It is possible that this specimen would be better attributed toH. habilis, however, this should be further assessed with reference to the morphology of the mandible, as well as the deciduous molars, particularly as previous attributions of this specimen toH. erectus have relied on these aspects of morphology^5,68. Previous analyses of this specimen have also outlined similarities with another juvenile mandible, KNM-ER 820, which could not be included in EDJ analyses here due to poor tissue distinction in scans. OH 22 shares some similarities withH. erectus specimens from Koobi Fora and Sangiran, but is also distinct in features such as the elongation of the mesial fovea (particularly in the P₃), and shows more extreme expression of some features seen in earlyH. erectus specimens, such as talonid reduction in the P₄ and rounded M₂ EDJ marginal ridges. This combination is consistent with the more recent age of the specimen compared to the otherH. erectus specimens included here, although comparison with youngerH. erectus specimens would be necessary to fully evaluate the taxonomic affinities of OH 22.

We find a considerable level of variation within theH. habilis dental hypodigm that in some cases exceeds that expected of a single species (Figs. 3,4, and5; Supplementary Figs. 10 and11). Bed I Olduvai specimens generally show a moreAustralopithecus-like morphology; OH 7 (~1.84 Ma⁶⁹) and OH 24 (~1.86 Ma⁶⁹) are close to or within theAustralopithecus range of variation for all preserved tooth positions (Supplementary Fig. 6), while OH 39 ( ~ 1.82 Ma⁶⁹) is within theAustralopithecus range for both premolars but has a more derived M², and has generally small teeth, some of which are smaller than those ofAustralopithecus (I¹, M²). Of the Koobi Fora specimens, KNM-ER 1802 is the oldest at 1.98-2.09 Ma⁷⁰, and the postcanine teeth are consistentlyAustralopithecus-like in overall morphology, as well as being larger than any otherH. habilis specimen. Similarly, the teeth of KNM-ER 1813 (1.78−1.95 Ma⁷¹) are largely primitive in morphology, albeit with a distinct M³ and generally small teeth.

The best-preservedH. habilis specimens from Olduvai Bed II are OH 13 and OH 16, both of which preserve entire mandibular and maxillary postcanine tooth rows. OH 13 is from the middle of Bed II and is the youngest specimen assigned toH. habilis, with a suggested age of 1.65 Ma^72,73. OH 16 is older, deriving from the lower section of Bed II. The specimen comes from the above marker Tuff I^F, which makes it no older than 1.8 Ma^69,72. Some early criticisms of the naming of the speciesH. habilis suggested that these Bed II specimens represented a different taxa to the older Bed I material^74,75, while others suggested the Bed II specimens could be distinguished from each other, with either OH 13^76,77 or OH 16⁷⁸ more similar toH. erectus specimens known at the time. We find that OH 16 is clearly more derived than otherH. habilis specimens in several aspects of postcanine EDJ morphology, and the specimen displays a suite of derived traits including an increased relative dentine body height in the postcanine dentition, which is typical of laterHomo. In this regard OH 16 differs significantly from OH 7 and KNM-ER 1802 (lower premolars), OH 39 (upper premolars), and KNM-ER 1813 (P³-M²); these differences in dentine body height exceed the level of variation expected within a species, based on extant hominines (Supplementary Fig. 11). However, it is important to note that there are key differences between OH 16 and laterHomo, in particular, the specimen retains several features that are clearly primitive (as seen in otherH. habilis specimens), particularly in the EDJ marginal ridge shape of the P₄, M₁, P³, and M¹.

The derived features present in the OH 16 dentition could suggest evolution within theH. habilis lineage over time. However, despite its younger age, OH 13 is not as derived; The P₄, M₁, and M² are similar to OH 16, but for a number of tooth positions such as the P₃, the morphology is more similar to Bed I specimens such as OH 7. The specimen therefore shows an intermediate morphology overall. It is possible that the differences between it and OH 16 represent normal variation within the Bed II sample; the differences between the two specimens are within the expected limits for a species based on most extant samples for all tooth positions (Supplementary Figs. 10 and11). It should be noted that another specimen that is suggested to derive from lower Bed II, OH 21, does significantly differ from the M¹ of OH 16 (in Procrustes distance and dentine body height) and is most similar in morphology to KNM-ER 1813. However, there is some uncertainty over the provenance of this specimen, which was found on the surface of a disturbed deposit⁴¹. Overall, while it is clear that there is substantial variation within specimens attributed toH. habilis, arguably too much to be subsumed within a single species, it is not clear to what degree time depth influences this variation, if at all.

The high level of variability in the earlyHomo fossil record has led many authors to suggest that multiple species are represented^{17,19,20,21,22}, although there is little agreement over the arrangement of specimens into groups. Unfortunately, KNM-ER 1470, which is central to this debate, does not preserve any tooth crowns, while the OH 65 maxilla, suggested to strongly resemble KNM-ER 1470³¹, could not be included in this sample. Equally, several specimens suggested to belong to this second species of earlyHomo such as KNM-ER 60000, KNM-ER 62000, and KNM-ER 62003²¹, could not yet be included in this sample as they have not yet been microCT scanned. KNM-ER 1590, which was included here, has been likened to KNM-ER 1470, although this is mostly based on the large size of the vault^5,79. We find that the teeth of this specimen are more derived along PC1 than KNM-ER 1813, and the M² is particularly derived. All teeth are larger in size than those ofH. habilis, which is consistent with a recent study of canine size that found that including KNM-ER 1590 in a sample ofH. habilis s.l. led to greater variability than seen in modern humans⁸⁰. The EDJ shape is not as derived as OH 16 in dentine body height, talonid reduction, or a reduction in cusp height, but shows an intermediate morphology in these features. In CVAs, the premolars of KNM-ER 1590 are found to be similar toAustralopithecus, and classify with the group. The molars show a more mixed signal, with some classifications withHomo, but more often not classifying with any group. This would be consistent with the specimen belonging to a second species of non-erectus earlyHomo. However, a further study including some of the specimens mentioned above would be necessary for a detailed dental assessment of this issue. Another important specimen is A.L. 666-1, a 2.3 Ma maxilla from Hadar. The specimen has been attributed toHomo aff.H. habilis²⁵, but more recently the arcade shape was found to be a poor match for OH 7 (and thereforeH. habilis), and instead shows shortened postcanine tooth rows as seen in earlyH. erectus andH. sapiens²². We find that the premolar morphology of this specimen is quite primitive, with both premolars within the range ofAustralopithecus, but also close to those of OH 39. The premolars also classify asAustralopithecus in CVA analyses. As in other earlyHomo specimens, the M² is more derived, showing among other features, a mesiodistally shortened EDJ ridge, and plotting between theAustralopithecus and laterHomo groups. The M² classifies as eitherAustralopithecus,H. habilis, or none of the groups in CVA analyses, depending on how many PCs are included. The tooth morphology is not found to be similar to the earlyH. erectus specimens included here. This suggests that this specimen combines an overall primitive dental morphology, similar toAustralopithecus and some other earlyHomo specimens, with a significantly more derived arcade shape. Of the remaining unattributed orHomo sp. specimens, there were no clear attributions for L26-1g (M₁), L398-573 (M³), or SK 847 (M³); both M³s occupy an intermediate position in CVA plots, however, theH. erectus comparative sample here is only 2 specimens, so this analysis is unlikely to capture the full range of variation in this taxa. However, Omo 75i-1255 (P₃) was found to be similar toH. habilis, while SK 27 (M¹) was found to be most similar toH. erectus.

To conclude, we find that the postcanine EDJ morphology of a number of key specimens ofH. habilis, including the type specimen OH 7, is very similar to that ofAustralopithecus. These results suggest that dental changes associated with laterHomo were not present in the earliest members of the genus. The EDJ morphology ofH. habilis is for the most part generalised, which is consistent with the primitive nature of the dental arcade and mandibular corpus in the species²², as well as important aspects of the postcrania such as the hand morphology^64,65 and relative limb proportions⁶⁶. As such, features linking specimens such as OH 7 to the genusHomo are largely limited to the neurocranium. Considering the skeleton as a whole, this indicates thatH. habilis possessed a number of plesiomorphic features, which is relevant for ongoing discussions over the genus-level designation ofH. habilis^16,36,81,82. Finally, our results highlight excessive levels of dental variation within the hypodigm as currently defined; Bed II specimen OH 16 shows a suite of derived dental traits that clearly distinguish it from otherH. habilis specimens, but do not appear to align the specimen with early AfricanH. erectus.

Permissions, outreach, and engagement

Scan data for specimens from Olduvai Gorge was collected for this study after acceptance of a research permit (#2017-182-NA-2016-304) to the Tanzanian Commission for Science and Technology (COSTECH) and under a memorandum of understanding signed between MMS and the National Museum of Tanzania. All of the original CT scan data, a high-end computer workstation (with an installation of the Avizo Software package), and 3D prints derived from these scans were deposited with the museum. A training internship at the University of Kent under the supervision of MMS was provided to AG to become proficient in the processing and analysis of CT scans. Additionally, outreach activities were organised by AG at the museum for local primary school student visits, using custom teaching materials available for download on the Human Fossil Record Archive (https://human-fossil-record.org/index.php?/category/6687). This included custom-designed interactive activity sheets created and funded by MMS focused on human evolution andHomo habilis in Tanzania, specifically.

Study sample

The hominin study sample is summarised in Table 1, and a detailed list of hominin specimens can be found in Supplementary Data 21. The sample consists of 712 hominin teeth (91 incisors, 54 canines, 212 premolars, and 355 molars. Additionally, a sample of extant ape postcanines (Gorilla gorilla andPan troglodytes,n = 199) was collected in order to compare the levels of variation inH. habilis with extant taxa (alongside theHomo sapiens specimens included in the main study sample); full details of this sample can be found in Supplementary Data 22.

The hominin sample includes 48 teeth assigned toH. habilis from Olduvai Gorge (OH 4, 7, 13, 15, 16, 21, 24, 39, 45; Figs. 1 and2), and 13 from Koobi Fora (KNM-ER 1502, 1802, 1813; Supplementary Figs. 3 and4). A full list of specimens preserving whole tooth crowns that are considered here to belong toH. habilis is available in Supplementary Table 3; for the small number of specimens that could not be included in the present study, the reasons are outlined in this table.

We also include several specimens that have been suggested to belong toH. habilis by some authors but do not have secure species-level attributions (SK 27, SK 847, L26-1g, L398-573, and Omo 75i-1255). These specimens are separated from theH. habilis sample for comparative purposes. Additionally, KNM-ER 1590 is included, which has been suggested to belong toH. rudolfensis through its cranial similarities with KNM-ER 1470^5,19. A.L. 666-1 is a 2.3 Ma maxilla from Hadar attributed toHomo aff.H. habilis²⁵, but is here grouped separately asHomo sp. KNM-ER 5431 has not been given a genus-level designation but has been suggested to show features resemblingAustralopithecus afarensis or earlyHomo^5,23,83.

TheH. erectus sample is limited to Early PleistoceneH. erectus specimens from Koobi Fora (KNM-ER 806, 807, 992, 1507, 1808, 1812, 3733) and Sangiran (Sangiran 1a, 4, 7-3a-c, 7-10, 7-20, 7-26, 7-27, 7-30, 7-32, 7-53, 7-65, 7-89) as earlyH. erectus provides the most appropriate comparison for the earlyHomo material. KNM-ER 807 has alternately been attributed toHomo sp⁵. orH. erectus⁸⁴, but is included here asH. erectus. The EDJ morphology of KNM-ER 1507 appears to differ from otherH. erectus specimens, and so was excluded from permutation tests and projected into CVA analyses to test its group affinities. OH 22, OH 23, and OH 29 are also assigned toH. erectus, but are separated from the mainH. erectus sample in figures as they are younger than the specimens from Koobi Fora and Sangiran (all are from Bed III or above, <1.14 Ma; refs.^{72,85,86,87,88}). We assess this earlyHomo dental hypodigm against a pooledAustralopithecus sample, consisting ofA. afarensis andAustralopithecus africanus, and a pooled laterHomo sample, consisting of modern humans, Neanderthals, and Middle-PleistoceneHomo. These groups are pooled to allow us to identify the overall EDJ shape changes that have taken place during the evolution of the genusHomo, and subsequently identify which of these traits are present inH. habilis.

Terminology

Throughout the manuscript, we refer to the morphology of the EDJ rather than the OES unless otherwise specified. At the EDJ it is clear that the height of the crown can be divided into two components: dentine body height and dentine horn height. Dentine body height refers to the distance between the cervix and the marginal ridges that encircle the occlusal basin, while dentine horn height refers to the distance between the marginal ridges and the tip of the tallest dentine horn (Supplementary Fig. 5). Total crown height is the combination of the two. Unless otherwise specified, these aspects of crown height are described relative to overall crown shape, rather than as an absolute measure.

Microtomography

Microtomographic scans of the sample were obtained using either a SkyScan 1172 or SkyScan 1173 at 100–130 kV and 90–130 microA, a BIR ACTIS 225/300 scanner at 130 kV and 100–120 microA, or a Diondo d3 at 100–140 kv and 100–140 microA and reconstructed as 16-bit tif stacks with an isometric voxel resolution ranging from 13-45 microns. SevenH. erectus teeth (Sangiran 7-3b, 7-3c, 7-20, 7-26, 7-53, 7-65, 7-89) were imaged using neutron microtomography at the ANTARES imaging facility located at the e FRM II reactor of the Technical University of Munich, Germany (for technical details see ref.⁸⁹).

Segmentation

For the majority of specimens, TIFF stacks were filtered using only a mean of least variance filter (kernel size one), or a 3D median filter (kernel size of three) followed by a mean of least variance filter (kernel size of three), implemented using MIA open source software⁹⁰. Filtered image stacks were used to segment enamel from dentine using a seed-growing watershed algorithm employed via a custom plugin employed in Avizo 6.3 (Visualisation Sciences Group), before being checked manually (the custom Avizo 6.3 plugin is not publicly available, but modern versions of Avizo, e.g., Avizo 3D 2022.1, have similarly built-in watershed segmentation modules). In specimens with particularly poor contrast between tissue types, enamel and dentine were segmented using the magic wand tool in Avizo v.8.0 (FEI Visualisation Sciences Group) using interpolation and manual corrections based on information from the filtered and unfiltered image stacks. In regions where enamel and dentine could not be precisely demarcated, interpolation tools were used to extend from the surrounding regions in which better tissue contrast was present. In all segmentations, particular attention is afforded to the marginal ridges, where landmarks are placed, with checks against unfiltered and filtered image stacks if necessary. Once enamel and dentine were segmented, a triangle-based surface model of the EDJ was produced using the unconstrained smoothing parameter in Avizo and then saved in polygon file format (.ply). In some cases, dental wear removes dentine horn tips. In the case of specimens with minimal wear, the missing portion of the dentine horn was reconstructed following a previously published procedure⁹¹. This methodology is similar to correcting for interstitial wear and involves inferring the shape of the dentine horn tip from the preserved anatomy of the dentine horn. This was restricted to cases in which the wear was less than wear level 3 as defined by Molnar⁹², and the observer was confident of the original position of the dentine horn using their experience, anatomical knowledge and preserved EDJ morphology. The reconstruction procedure was completed using the surface modification tools in Geomagic Studio 2014 (https://www.geomagic.com), and through manually adjusting the position of landmarks placed on dentine horn tips in Avizo 6.3. Dentine horn reconstructions inH. habilis specimens are available in Supplementary Fig. 6.

Landmark collection and derivation of homologous landmark sets

3D landmarks were collected in Avizo 6.3 in three distinct sets: ‘EDJ main’, ‘EDJ ridge’, and ‘CEJ ridge’ (Fig. 2). EDJ main consists of fixed landmarks placed on the tips of the main cusps (Mandibular premolars—1) protoconid 2) metaconid; Mandibular molars—1) protoconid, 2) metaconid, 3) entoconid, 4) hypoconid; Maxillary premolars—1) protocone, 2) metacone; Maxillary molars—1) protocone, 2) paracone, 3) metacone, 4) hypocone). EDJ ridge consists of sliding landmarks placed along the marginal ridges encircling the tooth, following the same order as the EDJ main landmarks. Both EDJ sets were collected directly on the EDJ surface models, but in specimens with poor contrast between tissue types, EDJ landmarks were checked against, or in some cases placed directly onto, the unfiltered stack to ensure that they are situated correctly along the marginal ridges. In the CEJ ridge set, landmarks were placed along the cementum-enamel junction (CEJ), beginning at the middle of the buccal face of the tooth (premolars and maxillary molars) or at the mesiobuccal corner of the crown, below the protoconid (mandibular molars), and continuing mesially. These landmarks were placed on an isosurface rendering of the unfiltered TIFF stack, or in cases where the CEJ is not visible on the isosurface due to calculus build-up or the presence of an adjacent tooth, landmarks were placed directly on the unfiltered stack.

A cubic spline function was used to fit a smooth curve through both ridge landmark sets, and the EDJ main landmarks were projected onto the EDJ ridge curve to divide it into sections (two sections in premolars, four in molars). A fixed number of equidistantly spaced landmarks were placed on each section of the curves (in premolars, 20 landmarks were placed in the first EDJ ridge section, and 25 in the second, while in molars, each EDJ ridge section has 20 landmarks. The CEJ ridge set has 40 landmarks all tooth positions). Homologous landmarks were then derived in R using the packages Morpho⁹³ and princurve⁹⁴ using a freely available R-based software routine⁹⁵, and following previously published protocols for premolars^42,43 and molars^49,96. EDJ main landmarks were fixed, while those in EDJ ridge and CEJ ridge sets were treated as semilandmarks and allowed to slide along their curves to reduce the bending energy of the thin-plate spline interpolation function calculated between each specimen and the Procrustes average for the sample⁹⁷. Sliding was performed twice, with landmarks projected back onto the curve after each step, before landmarks were considered geometrically homologous and converted into shape coordinates using generalised least-squares Procrustes superimposition, which removes scale, location, and orientation information from the coordinates^{98,99,100,101}.

For incisors and canines, the same process was followed, but with landmarks placed around the CEJ only. Landmarks were placed starting on the midpoint of the buccal face and continuing mesially, and as in other tooth positions, 40 equidistant CEJ ridge landmarks were derived, prior to sliding.

Analysis of EDJ shape and size

Principal component analysis (PCA) was carried out using the Procrustes coordinates of each specimen in shape space. This was completed separately for each postcanine mandibular and maxillary postcanine tooth position using both EDJ and CEJ landmark sets. For the anterior teeth, the same was done using only the CEJ landmark sets. In all cases, PCAs were used as an initial step to explore variation in EDJ shape and to identify EDJ features that distinguish betweenAustralopithecus and laterHomo. These shape changes were visualised using wireframe images of EDJ shape change across principal components. TheH. habilis specimens were then assessed according to their position in PCA plots, as well as by manually assessing the expression of EDJ features that were identified as important (e.g., those that distinguishAustralopithecus from laterHomo).

Canonical variate analysis (CVA) was performed on theAustralopithecus,H. habilis, andH. erectus samples, using a small subset of principal components (between 5 and 10) in order to ensure that the number of specimens exceeded the number of variables. In each case, the number of PCs with the highest cross-validated classification accuracy was selected for use in figures, and wireframe images were produced to visualise EDJ shape change across CVs. Cross-validated CVAs (cvCVAs) were also produced in order to check for spurious group separation. The group affinities of specimens without a specific attribution were assessed by projecting them into CVAs before calculating typicality probabilities; this was repeated for each subset of PCs (5-10, resulting in 6 CVAs) in order to ensure that the choice of PC subsets did not influence the results. Group affinities were assessed using a conservative threshold of 0.1; if typicality probabilities for all three groups were below this threshold, the specimen was considered to be unclassified.

The size of specimens was analysed using the natural logarithm of centroid size in two separate analyses, one for mandibular teeth and one for maxillary teeth. Only the CEJ landmark sets were used for these analyses to allow comparisons of size patterns across the entire tooth row.

Two analyses were conducted in order to assess the levels of variation expected within a species. In the first, Procrustes distances were calculated between all possible pairs of individuals within each of the extant species (P. troglodytes, G. gorilla, H. sapiens), as well as between each possible between-group pair forPan—Gorilla andPan—H. sapiens. These were plotted as frequency distributions in order to visualise inter- and intra-group variation and were compared to the differences observed betweenH. habilis specimens (specifically, OH 16 vs otherH. habilis). This was done separately for each postcanine tooth position (except for the third molars). For the second analysis, a specific aspect of EDJ shape is considered; dentine body height. Dentine body height refers to the distance between the cervix and the occlusal basin of the tooth (equivalent to the total height of the crown minus the height of the dentine horns, also referred to as cervical height¹⁰²). In premolars, a relatively taller dentine body is known to distinguish laterHomo fromAustralopithecus^42,43, and in modern humans is found to correlate strongly with tooth size. Dentine body height was quantified here by taking the centroid of a subset of landmarks on the mesial and distal marginal ridges of each tooth, as well as the centroid of a corresponding subset of landmarks on the mesial and distal sides of the cervix, then calculating the distance between the two. The two subsets of landmarks were chosen separately for each tooth position in order to capture the lowest portion of the marginal ridges, between dentine horns, in these specimens. On lower molars, the hypoconulid often occupies a distal position on the crown, so instead two distal subsets were used, one on either side of the hypoconulid. In each case, the same number of landmarks were used in the mesial and distal subsets. This was completed after Procrustes superimposition, and as such describes dentine body height relative to overall crown shape, rather than as an absolute measure. Within- and between-group differences in dentine body height were then calculated and visualised in the same manner as described above for Procrustes distances.

We also tested for shape and size differences between groups using permutation tests, carried out in R using 10,000 permutations, separately for each postcanine tooth position. For shape, this was done using Procrustes coordinates from the full EDJ analysis, and for size, this was done using centroid size from the CEJ analysis to allow the inclusion of anterior teeth. The Benjamini-Hochberg procedure was used to control the false-discovery rate¹⁰³.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

The geometric morphometric landmark data generated in this study have been deposited in the Publications section of The Human Fossil Record (https://human-fossil-record.org/index.php?/category/14230). The raw scan data used in this study are curated by the museums and institutes that curate the original fossil material. These data were used under a MOU for the current study, and so are not publicly available, but can be accessed by research application to the relevant curatorial institution (see Supplementary Data 21). The source data for Figs. 3,4,5, and Supplementary Figs. 7,10,11,12,13,14,15, and16 are provided as a Source Data file. The source data for Fig. 6 and Supplementary Fig. 17 are available in Supplementary Data 21. Supplementary Tables 1 and2 are based on previously published data, which is available in references listed in Supplementary Note 1. Source data are provided in this paper.

R code used for geometric morphometric analyses is available from:https://zenodo.org/doi/10.5281/zenodo.10255288⁹⁴.

For access to specimens and microCT scans, we would like to thank Emma Mbua and Samuel Muteti (National Museums of Kenya), Audax Mabulla (National Museum of Tanzania), the Tanzania Commission for Science and Technology, Metasebia Endalemaw, Yared Assefa (Ethiopian Authority for Research and Conservation of Cultural Heritage), Bernhard Zipfel, Sifelani Jira (Evolutionary Studies Institute, University of the Witwatersrand), Miriam Tawane (Ditsong Museum), Davorka Radovčić (Croatian Natural History Museum), Michel Toussaint (ASBL Archéologie Andennaise, Royal Belgian Institute of Natural Sciences), Jean-Jacques Cleyet-Merle (Musée National de Préhistoire des Eyzies-de-Tayac), Véronique Merlin-Langlade (Musée d’Art et d’Archéologie du Périgord), Christine Feja (Institut für Anatomie der Universität Leipzig), Antonio Rosas (Museo Nacional de Ciencias Naturales), Flora Gröning (Institute of Medical Sciences, University of Aberdeen), Dejana Brajković (Croatian Academy of Sciences and Arts), Jean-François Tournepiche (Musée d’Angoulême), Patrick Périn (Musée d’Archéologie nationale de Saint-Germain-en-Laye), Paul Tafforeau (European Synchrotron Radiation Facility), David Hunt (National Museum of Natural History), Ullrich Glasmacher (Institut für Geowissenschaften, Universität Heidelberg), Mirjana Roksandic (University of Winnipeg), Dušan Mihailović, Bojana Mihailović (Department of Archaeology, Belgrade University and the National Museum), Eleni Panagopoulou (Ephorate of Palaeoanthropology & Speleology of Southern Greece), Katarina Harvati (University of Tübingen), Martin Fischer (Phyletisches Museum Jena), Irina Ruf, Friedemann Schrenk and Christine Hemm (Senckenberg Forschungsinstitut und Naturmuseum, Frankfurt), Frieder Mayer and Christiane Funk (Museum für Naturkunde—Leibniz Institute for Evolution and Biodiversity Science), Emmanuel Gilissen (Royal Museum for Central Africa, Tervuren), Christophe Boesch and Roman Wittig (Taï Chimpanzee Project), and the Senckenberg, Forschungsstation für Quartärpaläontologie. We express our gratitude to the Werner Reimers Foundation in Bad Homburg (Germany), which provides the Gustav Heinrich Ralph von Koenigswald collection as a permanent loan for scientific research to the Senckenberg Research Institute and Natural History Museum Frankfurt. For µCT-scanning and technical assistance, we thank Kudakwashe Jakata, David Plotzki, and Heiko Temming. We thank the ANTARES imaging facility located at the FRM II reactor of the Technical University of Munich, Germany and Burkhard Schillinger for the possibility of neutron scanning of the specimens from Sangiran. For segmentation assistance, we thank Lia Schurtenberger, Rhianna Drummond-Clarke and Mykolas Imbrasas. This project has received funding from the Max Planck Society (T.W.D., P.G., J.-J.H.), the European Research Council (grant agreement No. 819960; M.M.S.) and the Calleva Foundation (U.K.; F.S.).

Open Access funding enabled and organized by Projekt DEAL.

1. Deceased: William H. Kimbel.

Authors and Affiliations

1. Department of Human Origins, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

   Thomas W. Davies, Philipp Gunz & Fred Spoor

2. School of Anthropology and Conservation, University of Kent, Canterbury, UK

   Thomas W. Davies, William P. Plummer & Matthew M. Skinner

3. Centre for Human Evolution Research, Natural History Museum, London, UK

   Fred Spoor

4. Department of Organismal Biology and Anatomy, University of Chicago, Chicago, IL, USA

   Zeresenay Alemseged

5. Department of Cultural Heritage, Ngorongoro Conservation Area Authority, P. O. Box 1, Ngorongoro Crater, Arusha, Tanzania

   Agness Gidna

6. Collège de France, Paris, France

   Jean-Jacques Hublin

7. Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

   Jean-Jacques Hublin & Matthew M. Skinner

8. Institute of Human Origins, and School of Human Evolution and Social Change, Arizona State University, Tempe, AZ, USA

   William H. Kimbel

9. Palaeobiology and Environment workgroup, Institute of Ecology, Evolution, and Diversity, Goethe University, Frankfurt, Germany

   Ottmar Kullmer

10. Division of Palaeoanthropology, Senckenberg Research Institute and Natural History Museum Frankfurt, Frankfurt am Main, Germany

    Ottmar Kullmer

11. Univ. Bordeaux, CNRS, MCC, PACEA, UMR 5199, 33600, Pessac, France

    Clément Zanolli

12. Centre for the Exploration of the Deep Human Journey, University of the Witwatersrand, Johannesburg, South Africa

    Matthew M. Skinner

Contributions

T.W.D., P.G., F.S., Z.A., A.G., J.-J.H., W.K., O.K., W.P., C.Z., and M.M.S. designed research and wrote the paper; T.W.D., P.G., F.S., W.P., C.Z., M.M.S. performed research and analysed data; F.S., Z.A., A.G., J.-J.H., W.K., M.M.S contributed to microCT scanning of fossils.

Corresponding author

Correspondence toThomas W. Davies.

Competing interests

The authors declare no competing interests.

Peer review information

Nature Communications thanks Sandrine Prat and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visithttp://creativecommons.org/licenses/by/4.0/.

Reprints and permissions