The dental metrics used here are part of a database assembled by the Research and Dissemination Center in Human Evolution of the Institute of Advanced Studies of the University of São Paulo (NPDEH-IEA-USP). The construction of this database was achieved through citation tracking and a systematic review of the literature published in the last three decades, as detailed in a previous publication [40]. The hominin species included in the data span from the Miocene (7 Ma) to the European Upper Paleolithic (30 thousand years ago; ka).

The data reviewed here are based on 22 sources listed in S1 Table, and include mesiodistal (MD) and buccolingual (BL) dimensions of postcanine teeth from the maxilla and mandible of 1,080 specimens (1,572 teeth). These dimensions were taken by different authors using the conventional procedure in biological anthropology of positioning the caliper at the maximum crown length (BL) and width (MD). Although calculating interobserver error is generally recommended in metric-focused dental studies that use published literature like ours [41], this was not possible here due to the amount of sources and the small number of specimens overlapping within these publications. However, it is expected that this error will represent a small portion of the variance, since differences among hominin species are relatively large, especially between genera, and therefore interobserver error should have a small impact on the interpretation of results.

Specimens without a species designation, those preserving only anterior dentition, and those missing more than 50% of the posterior dentition in either the maxilla or the mandible were excluded from the original dental database. From the five Dmanisi hominin fossils included in the data, only the D4500-D2600, D2282-D211, and D2700-D2735 specimens were kept in the analyses. The final dataset of the comparative specimens consists of 241 maxillary teeth from 51 specimens (Table 1) and 342 mandibular teeth from 71 specimens (Table 2). Missing data (5.4% in the maxillae and 3.6% in the mandibles) were estimated using multiple linear regression that considered the existing teeth as predictor variables, following the methodology recently described by us [42].

The MD and BL dimensions of premolars and molars were used to calculate dental areas by multiplying both dimensions, following the procedure in previous studies [24,26]. The mandibular and maxillary data were kept separated in the analyses to maximize the representation of dental area for fossils in which only the maxillae or the mandible were present. Table 3 presents the sample sizes of teeth used in the analyses by species.

Multivariate Linear Discriminant Analysis (LDA) was used to explore the morphological affinities of the D4500-D2600, D2282-D211 and D2700-D2735 specimens with 15 well-accepted hominin species (Tables 1–3). Discriminant functions were calculated for each of the datasets without the inclusion of the Dmanisi specimens, and were used to classify them a posteriori. Classification of the Dmanisi hominin fossils was based on the posterior probability of belonging to each of the hominin species in the reference data. To visualize the morphological affinities of the Dmanisi specimens, the scores of the first two discriminant functions were used to create a scatterplot illustrating the morphospace occupied by all the individuals in the analyses. All analyses were conducted in R [43]. LDA was implemented with the MASS package [44]. Data visualization was achieved using the ggplot2 and ggrepel packages [45,46].

As morphological differences in Dmanisi hominin fossils were suggested to be the result of sexual dimorphism, we also tested if the ratios between individual dental size of these fossils falls within the range of sexual dimorphism in extant apes. The ratios of the mandibular postcanine dentition of the three Dmanisi specimens were compared with the ratios between 15 male and 14 female gorillas, and 11 male and 11 female chimpanzees. Only mandibular data were considered because the comparative data used is limited to a sample with only mandibular data. However, it is expected that the size ratios of postcanine teeth between sexes of the great apes will not differ significantly between maxillary and mandibular dentition. The ratios of mandibular postcanine dentition of the Dmanisi specimens compared to gorillas and chimpanzees were represented visually using ggplot2 in R [45].

The comparative analysis of the dental crown area of the postcanine teeth from the three Dmanisi specimens included in our analyses support their classification in more than one species, as recently proposed [11,15–19]. The classification results (Tables 4–6 and Figs 1 and 2) demonstrate that the posterior dentition of these fossils is extremely diverse when compared to well-characterized hominin species. The D4500-D2600 specimen showed strong similarity to australopiths, while D2282-D211 and D2700-2735 specimens demonstrated stronger affinity with early Homo. This pattern is observed in both dental arcades, but the differentiation is more evident in the maxillary dentition, in which the larger specimen shows a very distinct classification pattern, based on posterior probabilities, and a clearly different position in the morphospace (Fig 1).

The differences among the Dmanisi specimens have been traditionally explained as a product of sexual dimorphism [5,8,9,12–14]. We show that these differences fall within the range of chimpanzees and gorillas, suggesting that they could be the product of sexual dimorphism if the Dmanisi hominin fossils came from a species with a similar level of sexual dimorphism to these great apes. These results indicate that we are unable to reject the hypothesis that Dmanisi specimens represent males and females of a single species based on crown area ratios. However, despite these results, we argue that the differences among the Dmanisi specimens are more parsimoniously explained by the existence of more than one species in the Dmanisi site. The strong association of the D4500-D2600 specimen with Australopithecus species is not only a function of the larger size of the teeth, but also that this specimen has a relative large M³ (Table 1), which goes in opposition to the trend in later Homo of showing smaller third molars [37]. This degree of separation can hardly be explained by sexual dimorphism alone, as it is not just a function of the size of the teeth. Moreover, while the area ratio between the largest Dmanisi specimen and the two smaller ones falls within the range of great apes, it still would represent more morphological diversity than the one observed among other Homo species, as well illustrated in Fig 1. This variety contrasts with the Homo erectus dental classification proposed initially for the Dmanisi hominin fossils [9], even though a delayed formation in the posterior dentition has recently been shown in D2700-D2735 specimen [47]. Therefore, similar to the results of cranio-morphological classification [11,16], the postcanine dental crown area of the three Dmanini specimens analysed here supports the taxonomic classification of the D4500-D2600 specimen as Homo georgicus, and the classification of the D2282-D211 and D2700-D2735 specimens as Homo caucasi.

Although our analyses did not formally test the phylogenetic history of the Pleistocene Georgian hominins, the proposal of more than one species in the Dmanisi fossil assemblage has implications for the dispersal of the genus Homo out of Africa in the beginning of the Pleistocene [15,16,48]. It is traditionally accepted that the Homo erectus migration started in Kenya (Turkana) around 1.89 Ma, reached Georgia (Dmanisi) around 1.77 Ma, continued into eastern Eurasia (Yuanmou) around 1.7 Ma, and finally arrived to Indonesia (Sangiran) by ~1.57 Ma [49,50]. However, the speciation events that led to the evolution of more than one species in Dmanisi requires that lineages were separated for long periods after leaving Africa, and were likely also evolving in response to different selective environments.

In recent years, several hypotheses have been proposed to explain the motivations behind the Homo erectus dispersal out of Africa [51,52]. Brain expansion has been suggested as the primary driver of Homo expansion, as increased cognitive capacity associated with efficient bipedal locomotion would allow Homo erectus to expand into new ecological niches. Cultural exclusion, which suggests that the emergence of Acheulean technology may have displaced Oldowan tool-using populations, has also been suggested as a main driver for expansion. Other hypotheses have been ecological in nature. For instance, it has been proposed that shifts in African fauna led associated consumers, including hominins, to move toward Eurasia [51,52].

If the Dmanisi specimens cannot be taxonomically grouped with Homo erectus [11,16], it raises the possibility that early Homo evolution had multiple episodes of cladogenesis, where some of them may have started in Africa, and others outside Africa. Of particular interest to this discussion is the high similarity between the D4500-D2600 specimen and australopiths, which suggests either a retention of the ancestral dental proportions of australopiths in Dmanisi, or an evolutionary convergence after the initial differentiation of early Homo. With the evidence available, it is not possible to properly evaluate if Homo georgicus and Homo caucasi evolved from Homo erectus ancestors, or if they evolved from australopith-like ancestors, but alternative scenarios are worth exploring and considering as new early Homo fossils are discovered in Asia.

Recent discoveries have been published and support alternative scenarios of Homo migration out of Africa. The new ²⁶Al/¹⁰Be ages from Yuanmou and Sangiran suggest that Homo erectus may have reached the farthest regions of Asia as early as 1.8 Ma [53,54], which contradicts the traditional route of dispersal [49,50]. Moreover, regions such as the Middle East and the southern fringes of Eurasia may have been more ecologically and biogeographically integrated with the African landscape than traditionally assumed, potentially creating favorable climatic conditions for the development of new hominin species [55]. Recent discoveries of Oldowan tools and associated cut marks in Jordan and Romania, respectively, predate the arrival of Homo erectus to these regions, offering further support for the presence of earlier hominin species in the north of or even outside of Africa [16,56,57]. The diversity of the Dmanisi hominin fossils, and the possibility that they represent more than one species, adds to this discussion demonstrating that a revision of our current models for the expansion of Homo out of Africa is required.