Two fossilized human crania (Apidima 1 and Apidima 2) from Apidima Cave, southern Greece, were discovered in the late 1970s but have remained enigmatic owing to their incomplete nature, taphonomic distortion and lack of archaeological context and chronology. Here we virtually reconstruct both crania, provide detailed comparative descriptions and analyses, and date them using U-series radiometric methods.
Apidima 2 dates to more than 170 thousand years ago and has a Neanderthal-like morphological pattern. By contrast, Apidima 1 dates to more than 210 thousand years ago and presents a mixture of modern human and primitive features. These results suggest that two late Middle Pleistocene human groups were present at this site-an early Homo sapiens population, followed by a Neanderthal population. Our findings support multiple dispersals of early modern humans out of Africa, and highlight the complex demographic processes that characterized Pleistocene human evolution and modern human presence in southeast Europe.
Southeast Europe is considered to be a major dispersal corridor as well as one of the principal European Mediterranean glacial refugia1-3. As such, the human fossil record of this region has previously been pro-posed to be more diverse than that of more isolated and less hospitable areas of Europe, reflecting the complexities of repeated dispersals, late survivals and admixture of human groups1,3. This hypothesis has been difficult to test, as palaeoanthropological finds from the Balkans are relatively scarce. The two fossilized human crania from Apidima, Mani (southern Greece)4, are among the most important finds from the region, yet remain little known. Here we applied the U-series dating method to elucidate their chronology and depositional history. We virtually reconstructed both specimens, correcting for taphonomic damage, and conducted detailed comparative description and geometric morphometric analyses.
Chronology
The Apidima specimens were discovered in a block of breccia wedged high between the cave walls of Apidima Cave A2,4-6 (Extended Data Fig.1), during research by the Museum of Anthropology, School of Medicine, National Kapodistrian University of Athens, which started in 1978. Owing to the lack of associated context, their geological age has been difficult to assess. Attempts to date the site radiometrically proved to be inconclusive7. However, geomorphology indicates a Middle-Late Pleistocene age, and a bracket between 190 and 100 thousand years ago (ka) has been proposed as the most-probable period for the deposition of the ‘skull breccia’6,8. Previous work calculated a minimum age of approximately 160 ka by U-series dating of an Apidima 2 bone fragment, which suggests a most-probable time of deposition of around 190 ka (transition between Marine Isotope Stage (MIS) 7 and MIS 6)5.We analysed three samples from the ‘skull breccia’, selected from fragments produced when cleaning the specimens from the matrix, using the U-series method. These included human bone fragments (subsamples 3720A and B of Apidima 2; and subsamples 3754 and 3755 of Apidima 1) and four unidentified bone subsamples (3757A-C and 3758, see Supplementary Information section 1). Our analyses show that both crania are older than the solidification of the matrix, which occurred around 150 ka. Despite their depositional proximity, Apidima 1 obtained its uranium in a considerably different environment than Apidima 2, during an accumulation event in MIS 7 (around 210 ka), whereas the uranium-uptake process of Apidima 2 took place in MIS 6 (around 170 ka) (see Methods ‘Depositional context’, Supplementary Information section 1). The crania and associated bones were therefore probably trapped on the surface of the talus cone, Apidima 1 around 210 ka and Apidima 2 around 170 ka, and were brought to their final position before the cementation and solidification of the sedimentary matrix around 150 ka (see Methods, ‘Depositional context’).
Description and comparative analyses
Apidima 2 (Fig.1a–c and Extended Data Fig.2) is the more complete and better known of the crania, and has previously been considered to be an early Neanderthal or Homo heidelbergensis4–6,9. It preserves an almost complete face and most of the vault (Supplementary Information section 2), but is taphonomically distorted. We produced four virtual manual reconstructions by two observers, following two different criteria, from a computed tomography scan of the original specimen (Extended Data Figs.3, 4 and Methods). Apidima 1 (Fig.1d–f) preserves the posterior cranium (Supplementary Information section 2). It shows no distortion; its virtual reconstruction therefore consisted of mirror-imaging the better-preserved side (Fig.1e, Methods and Extended Data Fig.5). Although no detailed study of this specimen has been conducted to date, it has been assumed to share the same taxonomic attribution as Apidima 2 (see, for example, a previously published study on the Apidima 2 chronology5).Apidima 2 shows Neanderthal-like features: a continuous, thick and rounded supraorbital torus with no break between the glabellar, orbital and lateral regions; a lack of break in plane between the glabellar and lateral regions in superior view; an anterior position of the nasal root; inflated infraorbital region; bilevel morphology of the inferior nasal margin; and rounded en bombe cranial profile in posterior view (Fig.1a-c and Extended Data Figs.2, 6, 7c, d). Most standard measurements (Supplementary Table 2) align it with Neanderthals. We conducted comparative geometric morphometric analyses of the face and neurocranium (analyses 1 and 2; Methods, Fig.2, Extended Data Table 1 and Supplementary Tables 4, 5), treating the Apidima 2
reconstructions and their mean configuration as separate individuals, projected into the principal component analysis (PCA). In both PCAs, the reconstructions plotted closest to Neanderthals or between Neanderthals and Middle Pleistocene Eurasians (MPEs). Linear discriminant analyses classified them as Neanderthal (except for reconstruction 2, which was classified as MPE only in analysis 1; Extended Data Table 1). The overall shape of the Apidima 2 reconstruction mean was closest to Gibraltar 1 in Procrustes distance in the face and to Spy 1 in the neurocranium, both of which are Neanderthals.By contrast, Apidima 1 does not have Neanderthal features; its linear measurements fall mainly in the region of overlap between taxa (Supplementary Information section 2 and Supplementary Table 3). It lacks a Neanderthal-like rounded en bombe profile in posterior view (Fig.1d and Extended Data Fig.7a, b). The widest part of the cranium is relatively low on the parietal; the parietal walls are nearly parallel and converge only slightly upwards, a plesiomorphic mor-phology that is common in Middle Pleistocene Homo10,11. It does not show the occipital plane convexity and lambdoid flattening associated with Neanderthal occipital ‘chignons’. Rather, its midsagittal outline is rounded in lateral view, a feature that is considered derived for modern humans12 (Fig.1e and Extended Data Fig.7b). The superior nuchal lines are weak with no external occipital protuberance. In contrast to some Middle Pleistocene specimens, the occipital bone is not steeply angled and lacks a thick occipital torus (Fig.1d, e and Extended Data Fig.7a). A small, very faint, depression is found above the inion (length, approximately 12 mm; height, approximately 4.55 mm; Extended Data Fig.7a). Although suprainiac fossae are considered derived for Neanderthals13, similar depressions occur among modern humans and in some African early H. sapiens14. The Apidima 1 depression does not present the typical Neanderthal combination of features. It is far smaller15 and less marked even than the ‘incipient’ suprainiac fossae of MPE specimens from Swanscombe and Sima de los Huesos, and is closest in size to the small supranuchal depression of the Eliye Springs cranium, a Middle Pleistocene African (MPA)16. Apidima 1 therefore lacks derived Neanderthal morphology, and instead shows a combination of ancestral and derived modern human features.We conducted a geometric morphometric analysis of the Apidima 1 neurocranium and its midsagittal profile (analyses 3 and 4; Fig.3, Extended Data Table 1 and Supplementary Tables 6, 7). In both analyses, Apidima 1 clearly clustered with H. sapiens in the PCAs and was classified as H. sapiens by the linear discriminant analyses (posterior probability 100% and 93.4% in analyses 3 and 4, respectively; Extended Data Table 1). Its overall shape was closest to Nazlet Khater 2 (analysis 3) and Dolní Věstonice 3 (analysis 4); both of which are modern humans. We calculated a neurocranial shape index based on the dataset from analysis 3 following a previous study17, using our Neanderthal and a modern African sample (n=15; Methods) and projecting Apidima 1 and all other specimens onto this axis (Fig.3c). Both fossil and recent H. sapiens are clearly separated from all archaic samples in this index. Apidima 1 fell within the range of fossil H. sapiens and just outside that of modern Africans, away from Neanderthals and Middle Pleistocene samples. Notably, the MPA crania from Jebel Irhoud, Morocco-which are considered to be early representatives of the H. sapiens lineage18-plotted with Neanderthals. The same analysis for the midsagittal profile dataset produced similar results (Extended Data Fig. 8).We compared the Apidima specimens for their common preserved anatomy. Although broadly similar in bi-auricular breadth, Apidima 2 is larger in its maximum cranial breadth, which reflects its en bombe outline in posterior view (Extended Data Figs.6, 7c). Apidima 1 is shorter antero-posteriorly and more rounded in lateral view (Extended Data Fig.9). The analysis of a restricted dataset of shared neurocra-nial landmarks and semilandmarks (analysis 5; Fig.4, Extended Data Table 1 and Supplementary Table 8) shows results similar to analyses 1-4. The Apidima 2 reconstructions fell with or close to Neanderthals along principal components 1 and 2 (PC1 and PC2) and were classified as Neanderthal (Extended Data Table 1). Their mean was closest in overall shape to Saccopastore 1, an early Neanderthal. Apidima 1 plotted closest to the H. sapiens convex hull, was classified as H. sapiens (posterior probability 92%, Extended Data Table 1) and was closest to Nazlet Khater 2 (a modern human) in Procrustes distance.
Implications for human evolution
Our assessment of the overall features, linear measurements and shape analyses of the face and neurocranium of Apidima 2 support a Neanderthal or early Neanderthal attribution, consistent with its chronological age of approximately 170 thousand years under the ‘accretion hypothesis’19. By contrast, Apidima 1 lacks derived Neanderthal features despite postdating the establishment of the distinct Neanderthal morphology19. Instead it shows a rounded posterior cranium, which is considered derived for modern humans12. This morphology cannot be explained by ontogenetic age, sexual dimorphism or interindividual variability. Although these factors might produce attenuated Neanderthal characteristics, they should not result in a complete lack of Neanderthal occipital features20,21, nor in the presence of derived modern human traits. It might be hypothesized that Apidima 1 represents an early stage of the Neanderthal lineage, when facial morphology was established but derived features of the posterior cranium were not5,10. However, Apidima 1 differs not only from similarly dated early Neanderthals (for example, Saccopastore and Biache-St-Vaast), but also from earlier specimens from Sima de los Huesos, Swanscombe and Reilingen, which exhibit Neanderthal-like occipital features19. It also differs from MPE specimens such as Petralona (Northern Greece) or Ceprano, which show angulated occipitals and thickened tori; features that are absent in Apidima 1. Although the Steinheim MPE specimen appears relatively rounded in lateral view, it is heavily damaged (having suffered multidirectional distortions and erosion), which makes its morphology and taxonomic attribution uncertain14,22.Apidima 1, therefore, does not fit in the ‘accretional’ scheme of Neanderthal evolution19, which has been proposed as the main explan-atory model of human evolution in Europe. Rather, its combination of ancestral and derived modern human features and overall shape are consistent with a taxonomic attribution to early modern humans. If this interpretation is correct, it documents-to our knowledge-the earliest known presence of Homo sapiens in Eurasia, which indicates that early modern humans dispersed out of Africa starting much earlier, and reaching much further, than previously thought. It also suggests that contact with the Neanderthal lineage may also have occurred during the Middle Pleistocene, as postulated from ancient DNA evidence23. Together, the Apidima crania suggest a complex pattern of population dispersal and possible replacement for southern Greece that is not dissimilar to that proposed for the Levant24-26-a potential source area for the population represented by Apidima 1. In such a scenario, early modern humans who were present in the region in the late Middle Pleistocene were replaced by Neanderthals, whose subsequent pres-ence in southern Greece is well-documented27-29. The latter were them-selves replaced by Upper Palaeolithic modern humans, whose earliest appearance in the region-as documented by Upper Palaeolithic lithic industries30-32-dates to approximately 40 ka. Our results highlight both the scarcity of our knowledge of the human fossil record in southeast Europe and the importance of this region in understanding Pleistocene human evolution and modern human dispersals.As we completed this paper, we noted the publication of a new study33 of the partial crania of Apidima 1 and Apidima 2. The authors of that study conclude that the two crania represent a transitional population between European Homo erectus and Neanderthals, a conclusion that is not supported by our more comprehensive analyses.
Methods
Depositional context. The crania were encased in a small block of breccia (65cm×45cm×35cm)34, discovered in 1978 wedged between the walls and near the ceiling of Apidima Cave A (Extended Data Fig.1). In a previous study5, the minimum depositional date was calculated to be approximately 160 ka for a bone fragment from Apidima 2 by U-series dating, thus constraining the upper limit of this range, and a mostlikely time of deposition around 190 ka was proposed (during the transition between MIS 7 and MIS 6)5. The breccia block is interpreted as a remnant of an eroded steep talus cone that originally fanned out of the cliffs in front and above the cave6 (Extended Data Fig.1c). The talus had to be graded to a previously existing dryland surface, indicating that the sea level was much lower for most of the time of its formation, most likely during a glacial period.The U-series results (Supplementary Information section 1) show that both human samples are older than the solidification of the matrix at around 150 ka. This completely concurs with common sense. Apidima 1 accumulated its uranium in a considerably different environment than Apidima 2, during an accumulation event in MIS 7 (around 210 ka), whereas the uranium-uptake process of Apidima 2 took place in MIS 6 (around 170 ka). The crania and associated bones were probably trapped on the surface of the talus cone, first Apidima 1 around 210 ka and later Apidima 2 at around 170 ka. The two crania were then brought into their final position at a later time, before the cementation and solidification of the sedimentary matrix around 150 ka. Water that preferentially infiltrates along cave walls often produces sediment dissolution and down-washing, and the formation of open spaces between the cave walls and the sedimentary fill. These sedimentary traps are later filled with collapsed material from the overlying sedimentary sequence. The location of the finds-between the walls of Apidima Cave A, wedged near the ceiling-suggest a similar scenario, in which bone material from Apidima 2 could be dislocated in a sedimentary trap from the overlying sequence and could have mixed with Apidima 1 remains, which also entered the trap at a later stage. The bones seem to have been thoroughly mixed, perhaps by a mudflow creeping down the sedimentary trap before consolidating at around 150 ka. Computed tomography scanning and virtual manual reconstruction. The crania of Apidima 1 and Apidima 2 were scanned at the First Department of Radiology of the National and Kapodistrian University of Athens using a multidetector computed tomography scanner (Philips). The scanning parameters were as follows: tube voltage 120 kV, tube current–time product 599 mAs, 16× 0.75 collimation, 0.8-mm slice thickness, slice increment 0.4 mm, field of view 249 mm, matrix 768× 768, pitch 0.44, rotation time 0.75s, convolution kernel detailed (D) and ultra-high focal spot resolution. The computed tomography scans of both individuals show isotropic pixel sizes of 0.31 and 0.32 mm, respectively. Apidima 1 and Apidima 2 were virtually reconstructed by A.M.B. and C.R. In all cases, the reconstruction was manual and based on the preserved anatomical features. All reconstruction steps were carried out in the software environment of Avizo (Visualization Sciences Group). Before the multiple reconstructions of Apidima 2, each fragment was segmented separately to allow independent movement during the virtual reconstructions (Extended Data Figs.3, 4). Several thin and tiny fragments could not be segmented in a reproducible way, owing to minimal differences in the grey values of bone and sediment matrix, and were thus excluded from the reconstructions. In total, 66 fragments were segmented. It was possible to segment fragments of the posterior neurocranium with semi-automated processes, as there were sufficient density differences between bone and matrix in this area. Facial fragments were mostly segmented manually slice by slice, owing to small differences in density between bone and matrix, combined with a low thickness of the fragments.Four independent reconstructions of Apidima 2 were carried out by A.M.B. and C.R., each using two different protocols (for comparison, see a previous study35). Independent of the protocol used, matrix-filled cracks were not closed completely in the reconstructions, to account for possible alterations of the edges of the fragments. No reference cranium was used during the reconstructions of Apidima 2, to exclude the risk of driving the results in the direction of the chosen reference specimen.A shared feature of vertebrate crania is approximate bilateral symmetry. The first protocol was based on this principle and had the goal to restore this symmetry. The anterior right part of the neurocranium was chosen as a starting point, as it presented a low amount of taphonomic deformation. Fragments of the right neurocranium were reconstructed according to a biologically meaningful position relative to each other. All reconstructed fragments of the right side were duplicated and mirrored along the midsagittal plane onto the left side. This mirrored duplicate was used as reference for the reconstruction of the fragments from the distorted left side of the neurocranium. The reconstructed left side of the brain case was subsequently mirrored to the right side to reconstruct the missing right temporal bone. Following the same procedure, the area close to the midsagittal plane on the right and a part of the supraorbital region on the left were reconstructed (shown as grey areas in Extended Data Figs.3, 4). For restoring facial symmetry, the midsagittal plane of the neurocranium was used as a reference. The right facial side was reconstructed and mirrored to reconstruct the fragmented left side. The left nasal bone, right maxilla-zygomatic fragment, and the left side of the lower face were duplicated and mirrored to reconstruct missing areas (shown as grey areas in Extended Data Figs.3, 4).The second protocol exploited the assumption that the ectocranial surface should follow a smooth curvature, especially in the neurocranium. In this protocol, each fragment is spatially constrained by its neighbouring fragments. The anterior right part of the neurocranium was chosen as a starting point, as several frag-ments were located in positions relative to each other that almost preserved smooth curvature. After reconstructing the vault, the facial fragments were repositioned relative to each other to match the smoothness criterion. However, mirroring of the right side was necessary to check and correct the fragmented left side. When the position of fragments had to be corrected to deal with taphonomic distortion, smoothness was prioritized over bilateral symmetry. Finally, missing areas-such as the right temporal bone, the right nasal bone and the left maxilla-were reconstructed by duplicating and mirroring their preserved counterpart (shown as grey areas in Extended Data Figs.3, 4).As previously shown36,37, multiple reconstructions of the same specimen will typically show some shape differences and no single reconstruction can be considered to be ‘perfect’. As the different reconstructions might be considered equally plausible36, we treated them as separate individuals in all geometric morphometric analyses. Furthermore, we calculated the mean configuration of all four reconstructions and treated this as an additional individual in our analysis. The final Apidima 2 reconstructions retain some distortion with respect to the relationship between the face and the neurocranium. Therefore, these two anatomical regions were analysed separately (see ‘Comparative samples’).The reconstruction of Apidima 1 was carried out by first computing a plane through the preserved part of the sagittal suture. The slices of the computed tomography scan were resampled according to this computed plane. Subsequently, preserved parts of the right parietal bone and right side of the occipital bone were cropped out along the computed plane in the original scan volume. This allowed mirroring a duplication of the cropped scan volume along the midsagittal plane. As a result, the reconstruction of Apidima 1 is completely symmetrical (Extended Data Fig.5). Figures of the reconstructions were produced in Adobe Photoshop.
Comparative samples. The samples used for our analyses included Neanderthals (MIS 8-3), earlier Middle Pleistocene specimens from Africa (MPA) and Eurasia (MPE), H. sapiens (including early anatomically modern human specimens and Upper Palaeolithic modern humans) and modern Africans (n=15) from the University of Witwatersrand Dart Collection. Severely taphonomically distorted and pathological specimens were excluded. The comparative summary statistics of the linear measurements reported in Supplementary Tables 2, 5 were based on data collected by C.S., supplemented by published values and by values collected from the Tübingen palaeoanthropology scan collection by K.H. and C.R. in Avizo (Visualization Sciences Group). The geometric morphometric comparative data were collected by K.H. Linear and three-dimensional measurements on the Apidima reconstructions were collected by K.H. and C.R. in Avizo (Visualization Sciences Group).
Analysis 1: the face of Apidima 2. This analysis comprised 25 facial landmarks: postorbital sulcus, glabella, nasion, infraspinale, prosthion, mid torus superior right and left, mid torus inferior right and left, dacryon right and left, zygoorbitale right and left, frontomalare right and left, infraorbital foramen right and left, zygomaxillare right and left, alare right and left, jugale right and left, frontomalare posterior right and left (landmark definitions have previously been published38). Comparative samples included 31 individuals: MPE, Arago 21 (as previously reconstructed36), Petralona, Sima de los Huesos 5; MPA, Bodo, Broken Hill, Irhoud 1; Neanderthals, La Chapelle-aux-Saints, Gibraltar 1, Guattari, La Ferrassie 1, Shanidar 1 and 5; H. sapiens, Abri Pataud, Chancelade, Cro-Magnon 1, 2, Dolní Věstonice 3, 13, 14, 15 and 16, Grimaldi, Hofmeyr, Mladeč 1, Muierii 1, Oase 2, Předmostí 3 and 4, Qafzeh 6 and 9, Wadi Kubbaniya.
Analysis 2: neurocranium of Apidima 2. This analysis included landmarks and curve semilandmarks outlining the supraorbital torus and midsagittal profile: gla-bella, bregma, lambda, frontomalare posterior (FMLP) right and left; 26 semiland-marks from glabella to bregma; 18 semilandmarks from FMLP right to FMLP left. Comparative samples included 41 specimens: MPE, Dali, Petralona, Sima de los Huesos 5; MPA, Broken Hill, Elandsfontein, Irhoud 1 and 2, Omo 2; Neanderthals, Amud 1, La Chapelle-aux-Saints, Feldhofer, La Ferrassie 1, Guattari, La Quina 5, Spy 1 and 2; H. sapiens, Abri Pataud, Brno, Chancelade, Cioclovina, Cro-Magnon 1, 2 and 3, Dolní Věstonice 3, 13, 15 and 16, Mladeč 1, 2 and 5, Muierii 1, Oase 2, Ohalo 2, Pavlov, Předmostí 3 and 4, Qafzeh 6 and 9, Skhul 5, Zhoukoudian Upper Cave 101 and 103. For Mladeč 2, the FMLP points were reconstructed using the entire sample as reference (see ‘Data processing’).
Analysis 3: neurocranium of Apidima 1. This analysis comprised 30 neurocra-nial landmarks and semilandmarks, including bregma, lambda and inion, as well as parietal notch, auriculare and porion bilaterally, and 21 semilandmarks from bregma to inion. Although the parietal of Apidima 1 is nearly complete in the midsagittal plane, the bregma is not preserved and was reconstructed on the basis of the entire fossil sample (see ‘Data processing’) in this and the next two datasets. The comparative sample comprised 38 fossil individuals: MPE, Dali, Petralona, Reilingen, Sima de los Huesos 5; MPA, Broken Hill, Eliye Springs, Irhoud 1 and 2, Omo 2; Neanderthals, Amud 1, La Chapelle-aux-Saints, La Ferrassie 1, Guattari, La Quina 5, Saccopastore 1; H. sapiens, Abri Pataud, Brno, Chancelade, Cioclovina, Cro-Magnon 1 and 2, Dolní Věstonice 3, 13, 15 and 16, Mladeč 1 and 5, Muierii 1, Nazlet Khater 2, Oase 2, Ohalo 2, Pavlov, Předmostí 3 and 4, Qafzeh 6 and 9, Skhul 5, Zhoukoudian Upper Cave 101.
Analysis 4: midsagittal profile of Apidima 1. This analysis comprised 24 land-marks and semilandmarks outlining the midsagittal profile from bregma to inion to analyse the parietal and occipital plane convexity of Apidima 1. The landmarks bregma, lambda and inion, and 21 semilandmarks from bregma to inion were included. The comparative sample consisted of 48 individuals: MPE, Dali, Petralona, Reilingen, Sima de los Huesos 5, Swanscombe; MPA, Broken Hill, Elandsfontein, Eliye Springs, Irhoud 1, 2, Omo 2; Neanderthals, Amud 1, Biache-st-Vaast, La Chapelle-aux-Saints, Feldhofer, La Ferrassie 1, Guattari, La Quina 5, Saccopastore 1, Spy 1 and 2; H. sapiens, Aduma, Abri Pataud, Brno, Chancelade, Cioclovina, Cro-Magnon 1, 2 and 3, Dolní Věstonice 3, 13, 15 and 16, Mladeč 1 and 5, Muierii 1, Nazlet Khater 2, Oase 2, Ohalo 2, Omo 1, Pavlov, Předmostí 3 and 4, Qafzeh 6 and 9, Skhul 5, Zhoukoudian Upper Cave 101 and 103.
Analysis 5: shared landmarks and semilandmarks of Apidima 1 and Apidima 2. This analysis included bregma and lambda, as well as parietal notch and auriculare (bilaterally), and 10 semilandmarks from bregma to lambda. The sample was the same as in analysis 3, but additionally comprised the Apidima 2 reconstructions.
Data processing. The fixed landmarks (type I, II and III) and curve semilandmarks (type IV) were collected from the reconstructions in Avizo 9.2.0 Lite (Visualization Sciences Group). The comparative data37,38 were collected by K.H. and processed with the dorsal-ventral-left-right fitting (DVLR) program (http://www.nycep.org/nmg/programs.html). The curve semilandmarks were calculated by resampling each curve as a predetermined number of equally spaced points, using Resample.exe (http://www.nycep.org/nmg/programs.html). As the bregma was not present in Apidima 1, but most of the bregma–lambda curve was preserved, this point was estimated using generalized Procrustes analysis (GPA) mean substitution in Morpheus39. This protocol first performs GPA to align the specimens. Then, grand-mean coordinate values are computed for the missing landmark using the non-missing points. The inverse scale, rotation and translation are subsequently applied to restore the original data. The same procedure was used to reconstruct the frontomalare temporale for Mladeč 2 in analysis 2. For the important, tapho-nomically deformed specimen Arago 21, the virtual reconstruction that had previously been produced36 was used in the comparative facial analysis of Apidima 2. Minimal reconstruction based on the surrounding anatomy was allowed during data collection, and landmarks that were missing on one side were reconstructed through reflected relabelling40, or by using a function in R41 based on a previously published study42. This function estimates a mirroring plane based on the unilateral landmarks. The missing landmarks are then reflected according to this plane. After the reconstruction of missing landmarks, the semilandmarks were slid along their respective closed curves using the Morpho package43 in R. Sliding was performed using the minimized bending energy algorithm44. After sliding, the data were exported in Morphologika format for further analysis45.
Data analysis. The compiled datasets were imported in Morphologika45 and super-imposed using GPA, which translates the specimen configurations to common origin, scales them for size and rotates them to best fit. Procrustes distances among specimens are a measure of overall shape difference. The superimposed coordinates of the comparative samples, excluding the Apidima specimens, were used as var-iables in a PCA, performed in the Past 3.04 software46. The resulting eigenvectors (principal component loadings) were used to compute the principal component scores for the Apidima specimens to plot them into the PCA graphs after the latter had been calculated on the basis of the comparative samples only. PCA plots were processed using Adobe Illustrator and extracted as Adobe PDF files. Furthermore, linear discriminant analyses (LDAs) and classification analyses were performed in Past 3.04 using the principal components as variables, in each case treating the reconstructions of Apidima 1 and Apidima 2 as unknown. The number of principal components included in the LDA for each of the 5 analyses included the first 7, 8, 8, 4 and 4 principal components, accounting for 70.72%, 91%, 88.6%, 85.4% and 78.2% of the total variance, respectively. Posterior probabilities were calculated with the SPSS software package (IBM, version 24 for Windows). We investigated whether the datasets used met the LDA assumptions47. We verified that all variables (principal component scores) showed an approximately normal distribution on the basis of both histograms and normal probability plots47. We removed potential outliers from the analysis by excluding pathological or taphonomically distorted specimens. On the basis of z-score analyses47, we found that outliers were absent in all variables, except for one case in PC3 of analysis 2: the MPA individual Omo 2, for which the z-score was 0.08 points over the maximum acceptable limit47 of 3.29. Given the limited number of well-preserved MPA crania in the fossil record, we decided to maintain this specimen in the analysis to maximize the representation of this group. Finally, the covariance matrices were similar among groups in all analyses, and Box’s M-tests showed that they were homogeneous for the samples used in analyses 4 and 5 (resulting P values were 0.19 and 0.07, respectively)47. However, this assumption could not be tested using Box’s M-test for most analyses owing to the small sample sizes of certain fossil groups, a common problem in palaeontology48. Because of these limitations, the results of the LDAs must be approached with caution, and not be interpreted in isolation, but in the context of all analyses presented here.
Visualization. Shape changes along principal component axes were visualized in Morphologika45. To further aid in visualization of shape differences between Apidima 1 and Apidima 2 (Extended Data Fig.9), we conducted manual super-impositions of their three-dimensional models in the software environment of Avizo 9.2.0 Lite (Visualization Sciences Group). Apidima 2 stayed in its original configuration and manipulations were carried out on Apidima 1. In the first step of superimposition, Apidima 1 was scaled to the biauricular breadth of Apidima 2. The transmeatal axes of both specimens were matched by translating and rotating Apidima 1. In the last step, Apidima 1 was rotated around the transmeatal axis to match the orientations of the external auditory meatus and the supramastoid crest of Apidima 2.
Shape index. The globular shape of the modern human neurocranium is consid-ered derived for modern humans and differentiates them from Neanderthals and other archaic Homo. It has recently been shown17 that a less-globular cranial shape in modern Europeans is related to the presence of specific Neanderthal alleles in their genome. We calculated the shape index for the posterior neurocranium of Apidima 1, to approximate the globularization index of this previous study17. We calculated an axis between the mean shapes of our Neanderthal sample and a Neanderthal-unadmixed, modern African sample (Zulu, Dart Collection, University of the Witwatersrand, n=15) and projected all other specimens (Apidima 1, MPE, MPA and fossil H. sapiens) onto this axis, to further evalu-ate the degree of globularity of the Apidima 1 neurocranium (Fig.3c, Extended Data Fig.8).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this paper. Data availabilityThe data that support the findings of this study are available from the corresponding authors upon reasonable request. 34.
Apidima Cave fossils provide earliest evidence of Homo sapiens in Eurasia
Katerina Harvati1,2,3*, Carolin Röding1, Abel M. Bosman1,2, Fotios A. Karakostis1, Rainer Grün4, Chris Stringer5, Panagiotis Karkanas6, Nicholas C. Thompson1,3, Vassilis Koutoulidis7, Lia A. Moulopoulos7, Vassilis G. Gorgoulis8,9,10* & Mirsini Kouloukoussa
Source: Apidima Cave fossils provide earliest evidence of Homo sapiens in Eurasia
1 Paleoanthropology, Senckenberg Centre for Human Evolution and Palaeoenvironment, Eberhard Karls University of Tübingen, Tübingen, Germany.
2DFG Centre of Advanced Studies ‘Words, Bones, Genes, Tools’, Eberhard Karls University of Tübingen, Tübingen, Germany.
3Museum of Anthropology, Medical School, National and Kapodistrian University of Athens, Athens, Greece.
4 Australian Research Centre for Human Evolution, Griffith University, Nathan, Queensland, Australia.
5Centre for Human Evolution Research, Department of Earth Sciences, The Natural History Museum, London, UK. 6Malcolm H. Wiener Laboratory for Archaeological Science, American School of Classical Studies at Athens, Athens, Greece.
7First Department of Radiology, National and Kapodistrian University of Athens, Athens, Greece.
8Department of Histology and Embryology, Medical School, National and Kapodistrian University of Athens, Athens, Greece.
9Biomedical Research Foundation of the Academy of Athens, Athens, Greece.
10Division of Cancer Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, Manchester Cancer Research Centre, NIHR Manchester Biomedical Research Centre,, University of Manchester, Manchester, UK. *e-mail: katerina.harvati@ifu.uni-tuebingen.de; vgorg@med.uoa.gr
1. Dennell, R. W., Martinón-Torres, M. & Bermúdez de Castro, J. M. Hominin variability, climatic instability and population demography in Middle Pleistocene Europe. Quat. Sci. Rev. 30, 1511–1524 (2011).
2. Tourloukis, V. & Harvati, K. The Palaeolithic record of Greece: a synthesis of the evidence and a research agenda for the future. Quat. Int. 466, 48–65 (2018).
3. Roksandic, M., Radović, P. & Lindal, J. Revising the hypodigm of Homo heidelbergensis: a view from the Eastern Mediterranean. Quat. Int. 466, 66–81 (2018).
4. Pitsios, T. K. Paleoanthropological research at the cave site of Apidima and the surrounding region (South Peloponnese, Greece). Anthropol. Anz. 57, 1–11 (1999).
5. Bartsiokas, A., Arsuaga, J. L., Aubert, M. & Grün, R. U-series dating and classification of the Apidima 2 hominin from Mani Peninsula, Southern Greece. J. Hum. Evol. 109, 22–29 (2017).
6. Harvati, K., Stringer, C. & Karkanas, P. Multivariate analysis and classification of the Apidima 2 cranium from Mani, Southern Greece. J. Hum. Evol. 60, 246–250 (2011)
7. Liritzis, Y. & Maniatis, Y. ESR experiments on quaternary calcites and bones for dating purposes. J. Radioanal. Nucl. Chem. 129, 3–21 (1989).
8. Rondoyanni, T., Mettos, A. & Georgiou, C. Geological–morphological observations in the greater Oitilo-Diros area, Mani. Acta Anthropol. 1, 93–102 (1995).
9. Coutselinis, A., Dritsas, C. & Pitsios, T. K. Expertise médico-légale du crâne pléistocène LAO1/S2 (Apidima II), Apidima, Laconie, Grèce. L’Anthropologie 95, 401–408 (1991).
10. Arsuaga, J. L. etal. Neandertal roots: cranial and chronological evidence from Sima de los Huesos. Science 344, 1358–1363 (2014).
11. Stringer, C. The origin and evolution of Homo sapiens. Phil. Trans. R. Soc. Lond. B 371, 20150237 (2016).
12. Galway-Witham, J. & Stringer, C. How did Homo sapiens evolve? Science 360, 1296–1298 (2018).
13. Harvati, K. in Handbook of Paleoanthropology (eds Henke, W. & Tattersall, I.) 2243–2279 (Springer, 2015).
14. Balzeau, A. & Rougier, H. Is the suprainiac fossa a Neandertal autapomorphy? A complementary external and internal investigation. J. Hum. Evol. 58, 1–22 (2010).
15. Verna, C., Hublin, J.-J., Debenath, A., Jelinek, A. & Vandermeersch, B. Two new hominin cranial fragments from the Mousterian levels at La Quina (Charente, France). J. Hum. Evol. 58, 273–278 (2010).
16. Bräuer, G. & Leakey, R. L. The ES-11693 cranium from Eliye Springs, West Turkana, Kenya. J. Hum. Evol. 15, 289–312 (1986). 17. Gunz, P. etal. Neandertal introgression sheds light on modern human endocranial globularity. Curr. Biol. 29, 120–127 (2019). 18. Hublin, J.-J. etal. New fossils from Jebel Irhoud, Morocco and the pan-African origin of Homo sapiens. Nature 546, 289–292 (2017). 19. Hublin, J.-J. The origin of Neandertals. Proc. Natl Acad. Sci. USA 106, 16022–16027 (2009). 20. Caspari, R. The Krapina occipital bones. Period. Biol. 108, 299–307 (2006). 21. Arsuaga, J. L., Martínez, I., Gracia, A. & Lorenzo, C. The Sima de los Huesos crania (Sierra de Atapuerca, Spain). A comparative study. J. Hum. Evol. 33, 219–281 (1997).
22. Prossinger, H. etal. Electronic removal of encrustations inside the Steinheim cranium reveals paranasal sinus features and deformations, and provides a revised endocranial volume estimate. Anat. Rec. 273B, 132–142 (2003).
23. Posth, C. etal. Deeply divergent archaic mitochondrial genome provides lower time boundary for African gene flow into Neanderthals. Nat. Commun. 8, 16046 (2017).
24. Mercier, N. etal. Thermoluminescence date for the Mousterian burial site of Es-Skhul, Mt. Carmel. J. Archaeol. Sci. 20, 169–174 (1993).
25. Hershkovitz, I. etal. The earliest modern humans outside Africa. Science 359, 456–459 (2018). 26. Stringer, C. & Galway-Witham, J. When did modern humans leave Africa? Science 359, 389–390 (2018).
27. Harvati, K., Panagopoulou, E. & Karkanas, P. First Neanderthal remains from Greece: the evidence from Lakonis. J. Hum. Evol. 45, 465–473 (2003).
28. Harvati, K. etal. New Neanderthal remains from Mani peninsula, Southern Greece: the Kalamakia Middle Paleolithic cave site. J. Hum. Evol. 64, 486–499 (2013).
29. Tourloukis, V. etal. New Middle Paleolithic sites from the Mani peninsula, Southern Greece. J. Field Archaeol. 41, 68–83 (2016).
30. Elefanti, P., Panagopoulou, E. & Karkanas, P. The transition from the Middle to the Upper Paleolithic in the Southern Balkans: the evidence from Lakonis 1 Cave, Greece. Eurasian Prehistory 5, 85–95 (2008).
31. Douka, K., Perlès, C., Valladas, H., Vanhaeren, M. & Hedges, R. E. M. Francthi Cave revisited: the age of the Aurignacian in south-eastern Europe. Antiquity 85, 1131–1150 (2011).
32. Lowe, J. etal. Volcanic ash layers illuminate the resilience of Neanderthals and early modern humans to natural hazards. Proc. Natl Acad. Sci. USA 109, 13532–13537 (2012).
33. De Lumley, M. A. Les Restes Humains Anténéanderthaliens Apidima 1 et Apidima 2 (CNRS, 2019).
34. Kormasopoulou-Kagalou, L., Protonotariou-Deilaki, E. & Pitsios, T. K. Paleolithic skull burials at the cave of Apidima. Acta Anthropol. 1, 119–124 (1995).
35. Zollikofer, C. P. etal. Virtual cranial reconstruction of Sahelanthropus tchadensis. Nature 434, 755–759 (2005).
36. Gunz, P., Mitteroecker, P., Neubauer, S., Weber, G. W. & Bookstein, F. L. Principles for the virtual reconstruction of hominin crania. J. Hum. Evol. 57, 48–62 (2009).
37. Harvati, K., Hublin, J.-J. & Gunz, P. Evolution of middle–late Pleistocene human cranio-facial form: a 3-D approach. J. Hum. Evol. 59, 445–464 (2010).
38. Harvati, K., Gunz, P. & Grigorescu, D. Cioclovina (Romania): affinities of an early modern European. J. Hum. Evol. 53, 732–746 (2007).
39. Slice, D. E. Morpheus et al., Java Edition. http://morphlab.sc.fsu.edu/ (The Florida State University, 2013).
40. Mardia, K. V., Bookstein, F. L. & Moreton, I. J. Statistical assessment of bilateral symmetry of shapes. Biometrika 87, 285–300 (2000).
41. R Development Core Team. R: A Language and Environment for Statistical Computing. http://www.R-project.org/ (R Foundation for Statistical Computing, 2008).
42. Claude, J. Morphometrics with R (Springer Science & Business Media, 2008).
43. Schlager, S. in Statistical Shape and Deformation Analysis (eds Zheng, G. etal.) 217−256 (Academic, 2017).
44. Bookstein, F. L. Landmark methods for forms without landmarks: morphometrics of group differences in outline shape. Med. Image Anal. 1, 225–243 (1997).
45. O’Higgins, P. & Jones, N. Morphologika: Tools for Shape Analysis. Version 2.2 https://sites.google.com/site/hymsfme/resources (Hull York Medical School, 2006).
46. Hammer, Ø., Harper, D. A. T. & Ryan, P. D. PAST: paleontological statistics software package for education and data analysis. Palaeontol. Electronica 4, 1–9 (2001).
47. Field, A. Discovering Statistics using SPSS (Sage, 2013).
48. Brown, P. Nacurrie 1: mark of ancient Java, or a caring mother’s hands, in terminal Pleistocene Australia? J. Hum. Evol. 59, 168–187 (2010).
Acknowledgements
This research was supported by the European Research Council (ERC CoG no. 724703) and the German Research Foundation (DFG FOR 2237). We thank all curators and their institutions for access to original specimens or casts used in this study; T. White, B. Asfaw, M. López-Soza, V. Tourloukis, D. Giusti, G. Konidaris, C. Fardelas and O. Stolis for their input and assistance; A. Balzeau (Muséum National d’Histoire Naturelle; MNHN), E. Delson (New York Consortium in Evolutionary Primatology; NYCEP), L. Leakey (africanfossils.org) for providing access to three-dimensional models of specimens used in our figures. C.S.’s research is supported by the Calleva Foundation and the Human Origins Research Fund. We are grateful to S. Benazzi, E. Delson and I. Hershkovitz for their comments and suggestions. Author contributions K.H., M.K. and V.G.G. designed the research; V.K. and L.A.M. carried out the computed tomography scans; C.R. and A.M.B. generated the virtual reconstructions; K.H., C.S. and C.R. collected comparative data; K.H., C.R., A.M.B., F.A.K. and N.C.T. processed and analysed the data; R.G. dated the specimens; P.K.and R.G. provided stratigraphic and geological interpretations; all authors contributed to compiling the manuscript.
Depositional context. The crania were encased in a small block of breccia (65cm×45cm×35cm)34, discovered in 1978 wedged between the walls and near the ceiling of Apidima Cave A (Extended Data Fig.1). In a previous study5, the minimum depositional date was calculated to be approximately 160 ka for a bone fragment from Apidima 2 by U-series dating, thus constraining the upper limit of this range, and a mostlikely time of deposition around 190 ka was proposed (during the transition between MIS 7 and MIS 6)5. The breccia block is interpreted as a remnant of an eroded steep talus cone that originally fanned out of the cliffs in front and above the cave6 (Extended Data Fig.1c). The talus had to be graded to a previously existing dryland surface, indicating that the sea level was much lower for most of the time of its formation, most likely during a glacial period.The U-series results (Supplementary Information section 1) show that both human samples are older than the solidification of the matrix at around 150 ka. This completely concurs with common sense. Apidima 1 accumulated its uranium in a considerably different environment than Apidima 2, during an accumulation event in MIS 7 (around 210 ka), whereas the uranium-uptake process of Apidima 2 took place in MIS 6 (around 170 ka). The crania and associated bones were probably trapped on the surface of the talus cone, first Apidima 1 around 210 ka and later Apidima 2 at around 170 ka. The two crania were then brought into their final position at a later time, before the cementation and solidification of the sedimentary matrix around 150 ka. Water that preferentially infiltrates along cave walls often produces sediment dissolution and down-washing, and the formation of open spaces between the cave walls and the sedimentary fill. These sedimentary traps are later filled with collapsed material from the overlying sedimentary sequence. The location of the finds-between the walls of Apidima Cave A, wedged near the ceiling-suggest a similar scenario, in which bone material from Apidima 2 could be dislocated in a sedimentary trap from the overlying sequence and could have mixed with Apidima 1 remains, which also entered the trap at a later stage. The bones seem to have been thoroughly mixed, perhaps by a mudflow creeping down the sedimentary trap before consolidating at around 150 ka. Computed tomography scanning and virtual manual reconstruction. The crania of Apidima 1 and Apidima 2 were scanned at the First Department of Radiology of the National and Kapodistrian University of Athens using a multidetector computed tomography scanner (Philips). The scanning parameters were as follows: tube voltage 120 kV, tube current–time product 599 mAs, 16× 0.75 collimation, 0.8-mm slice thickness, slice increment 0.4 mm, field of view 249 mm, matrix 768× 768, pitch 0.44, rotation time 0.75s, convolution kernel detailed (D) and ultra-high focal spot resolution. The computed tomography scans of both individuals show isotropic pixel sizes of 0.31 and 0.32 mm, respectively. Apidima 1 and Apidima 2 were virtually reconstructed by A.M.B. and C.R. In all cases, the reconstruction was manual and based on the preserved anatomical features. All reconstruction steps were carried out in the software environment of Avizo (Visualization Sciences Group). Before the multiple reconstructions of Apidima 2, each fragment was segmented separately to allow independent movement during the virtual reconstructions (Extended Data Figs.3, 4). Several thin and tiny fragments could not be segmented in a reproducible way, owing to minimal differences in the grey values of bone and sediment matrix, and were thus excluded from the reconstructions. In total, 66 fragments were segmented. It was possible to segment fragments of the posterior neurocranium with semi-automated processes, as there were sufficient density differences between bone and matrix in this area. Facial fragments were mostly segmented manually slice by slice, owing to small differences in density between bone and matrix, combined with a low thickness of the fragments.Four independent reconstructions of Apidima 2 were carried out by A.M.B. and C.R., each using two different protocols (for comparison, see a previous study35). Independent of the protocol used, matrix-filled cracks were not closed completely in the reconstructions, to account for possible alterations of the edges of the fragments. No reference cranium was used during the reconstructions of Apidima 2, to exclude the risk of driving the results in the direction of the chosen reference specimen.A shared feature of vertebrate crania is approximate bilateral symmetry. The first protocol was based on this principle and had the goal to restore this symmetry. The anterior right part of the neurocranium was chosen as a starting point, as it presented a low amount of taphonomic deformation. Fragments of the right neurocranium were reconstructed according to a biologically meaningful position relative to each other. All reconstructed fragments of the right side were duplicated and mirrored along the midsagittal plane onto the left side. This mirrored duplicate was used as reference for the reconstruction of the fragments from the distorted left side of the neurocranium. The reconstructed left side of the brain case was subsequently mirrored to the right side to reconstruct the missing right temporal bone. Following the same procedure, the area close to the midsagittal plane on the right and a part of the supraorbital region on the left were reconstructed (shown as grey areas in Extended Data Figs.3, 4). For restoring facial symmetry, the midsagittal plane of the neurocranium was used as a reference. The right facial side was reconstructed and mirrored to reconstruct the fragmented left side. The left nasal bone, right maxilla-zygomatic fragment, and the left side of the lower face were duplicated and mirrored to reconstruct missing areas (shown as grey areas in Extended Data Figs.3, 4).The second protocol exploited the assumption that the ectocranial surface should follow a smooth curvature, especially in the neurocranium. In this protocol, each fragment is spatially constrained by its neighbouring fragments. The anterior right part of the neurocranium was chosen as a starting point, as several frag-ments were located in positions relative to each other that almost preserved smooth curvature. After reconstructing the vault, the facial fragments were repositioned relative to each other to match the smoothness criterion. However, mirroring of the right side was necessary to check and correct the fragmented left side. When the position of fragments had to be corrected to deal with taphonomic distortion, smoothness was prioritized over bilateral symmetry. Finally, missing areas-such as the right temporal bone, the right nasal bone and the left maxilla-were reconstructed by duplicating and mirroring their preserved counterpart (shown as grey areas in Extended Data Figs.3, 4).As previously shown36,37, multiple reconstructions of the same specimen will typically show some shape differences and no single reconstruction can be considered to be ‘perfect’. As the different reconstructions might be considered equally plausible36, we treated them as separate individuals in all geometric morphometric analyses. Furthermore, we calculated the mean configuration of all four reconstructions and treated this as an additional individual in our analysis. The final Apidima 2 reconstructions retain some distortion with respect to the relationship between the face and the neurocranium. Therefore, these two anatomical regions were analysed separately (see ‘Comparative samples’).The reconstruction of Apidima 1 was carried out by first computing a plane through the preserved part of the sagittal suture. The slices of the computed tomography scan were resampled according to this computed plane. Subsequently, preserved parts of the right parietal bone and right side of the occipital bone were cropped out along the computed plane in the original scan volume. This allowed mirroring a duplication of the cropped scan volume along the midsagittal plane. As a result, the reconstruction of Apidima 1 is completely symmetrical (Extended Data Fig.5). Figures of the reconstructions were produced in Adobe Photoshop.
Comparative samples. The samples used for our analyses included Neanderthals (MIS 8-3), earlier Middle Pleistocene specimens from Africa (MPA) and Eurasia (MPE), H. sapiens (including early anatomically modern human specimens and Upper Palaeolithic modern humans) and modern Africans (n=15) from the University of Witwatersrand Dart Collection. Severely taphonomically distorted and pathological specimens were excluded. The comparative summary statistics of the linear measurements reported in Supplementary Tables 2, 5 were based on data collected by C.S., supplemented by published values and by values collected from the Tübingen palaeoanthropology scan collection by K.H. and C.R. in Avizo (Visualization Sciences Group). The geometric morphometric comparative data were collected by K.H. Linear and three-dimensional measurements on the Apidima reconstructions were collected by K.H. and C.R. in Avizo (Visualization Sciences Group).
Analysis 1: the face of Apidima 2. This analysis comprised 25 facial landmarks: postorbital sulcus, glabella, nasion, infraspinale, prosthion, mid torus superior right and left, mid torus inferior right and left, dacryon right and left, zygoorbitale right and left, frontomalare right and left, infraorbital foramen right and left, zygomaxillare right and left, alare right and left, jugale right and left, frontomalare posterior right and left (landmark definitions have previously been published38). Comparative samples included 31 individuals: MPE, Arago 21 (as previously reconstructed36), Petralona, Sima de los Huesos 5; MPA, Bodo, Broken Hill, Irhoud 1; Neanderthals, La Chapelle-aux-Saints, Gibraltar 1, Guattari, La Ferrassie 1, Shanidar 1 and 5; H. sapiens, Abri Pataud, Chancelade, Cro-Magnon 1, 2, Dolní Věstonice 3, 13, 14, 15 and 16, Grimaldi, Hofmeyr, Mladeč 1, Muierii 1, Oase 2, Předmostí 3 and 4, Qafzeh 6 and 9, Wadi Kubbaniya.
Analysis 2: neurocranium of Apidima 2. This analysis included landmarks and curve semilandmarks outlining the supraorbital torus and midsagittal profile: gla-bella, bregma, lambda, frontomalare posterior (FMLP) right and left; 26 semiland-marks from glabella to bregma; 18 semilandmarks from FMLP right to FMLP left. Comparative samples included 41 specimens: MPE, Dali, Petralona, Sima de los Huesos 5; MPA, Broken Hill, Elandsfontein, Irhoud 1 and 2, Omo 2; Neanderthals, Amud 1, La Chapelle-aux-Saints, Feldhofer, La Ferrassie 1, Guattari, La Quina 5, Spy 1 and 2; H. sapiens, Abri Pataud, Brno, Chancelade, Cioclovina, Cro-Magnon 1, 2 and 3, Dolní Věstonice 3, 13, 15 and 16, Mladeč 1, 2 and 5, Muierii 1, Oase 2, Ohalo 2, Pavlov, Předmostí 3 and 4, Qafzeh 6 and 9, Skhul 5, Zhoukoudian Upper Cave 101 and 103. For Mladeč 2, the FMLP points were reconstructed using the entire sample as reference (see ‘Data processing’).
Analysis 3: neurocranium of Apidima 1. This analysis comprised 30 neurocra-nial landmarks and semilandmarks, including bregma, lambda and inion, as well as parietal notch, auriculare and porion bilaterally, and 21 semilandmarks from bregma to inion. Although the parietal of Apidima 1 is nearly complete in the midsagittal plane, the bregma is not preserved and was reconstructed on the basis of the entire fossil sample (see ‘Data processing’) in this and the next two datasets. The comparative sample comprised 38 fossil individuals: MPE, Dali, Petralona, Reilingen, Sima de los Huesos 5; MPA, Broken Hill, Eliye Springs, Irhoud 1 and 2, Omo 2; Neanderthals, Amud 1, La Chapelle-aux-Saints, La Ferrassie 1, Guattari, La Quina 5, Saccopastore 1; H. sapiens, Abri Pataud, Brno, Chancelade, Cioclovina, Cro-Magnon 1 and 2, Dolní Věstonice 3, 13, 15 and 16, Mladeč 1 and 5, Muierii 1, Nazlet Khater 2, Oase 2, Ohalo 2, Pavlov, Předmostí 3 and 4, Qafzeh 6 and 9, Skhul 5, Zhoukoudian Upper Cave 101.
Analysis 4: midsagittal profile of Apidima 1. This analysis comprised 24 land-marks and semilandmarks outlining the midsagittal profile from bregma to inion to analyse the parietal and occipital plane convexity of Apidima 1. The landmarks bregma, lambda and inion, and 21 semilandmarks from bregma to inion were included. The comparative sample consisted of 48 individuals: MPE, Dali, Petralona, Reilingen, Sima de los Huesos 5, Swanscombe; MPA, Broken Hill, Elandsfontein, Eliye Springs, Irhoud 1, 2, Omo 2; Neanderthals, Amud 1, Biache-st-Vaast, La Chapelle-aux-Saints, Feldhofer, La Ferrassie 1, Guattari, La Quina 5, Saccopastore 1, Spy 1 and 2; H. sapiens, Aduma, Abri Pataud, Brno, Chancelade, Cioclovina, Cro-Magnon 1, 2 and 3, Dolní Věstonice 3, 13, 15 and 16, Mladeč 1 and 5, Muierii 1, Nazlet Khater 2, Oase 2, Ohalo 2, Omo 1, Pavlov, Předmostí 3 and 4, Qafzeh 6 and 9, Skhul 5, Zhoukoudian Upper Cave 101 and 103.
Analysis 5: shared landmarks and semilandmarks of Apidima 1 and Apidima 2. This analysis included bregma and lambda, as well as parietal notch and auriculare (bilaterally), and 10 semilandmarks from bregma to lambda. The sample was the same as in analysis 3, but additionally comprised the Apidima 2 reconstructions.
Data processing. The fixed landmarks (type I, II and III) and curve semilandmarks (type IV) were collected from the reconstructions in Avizo 9.2.0 Lite (Visualization Sciences Group). The comparative data37,38 were collected by K.H. and processed with the dorsal-ventral-left-right fitting (DVLR) program (http://www.nycep.org/nmg/programs.html). The curve semilandmarks were calculated by resampling each curve as a predetermined number of equally spaced points, using Resample.exe (http://www.nycep.org/nmg/programs.html). As the bregma was not present in Apidima 1, but most of the bregma–lambda curve was preserved, this point was estimated using generalized Procrustes analysis (GPA) mean substitution in Morpheus39. This protocol first performs GPA to align the specimens. Then, grand-mean coordinate values are computed for the missing landmark using the non-missing points. The inverse scale, rotation and translation are subsequently applied to restore the original data. The same procedure was used to reconstruct the frontomalare temporale for Mladeč 2 in analysis 2. For the important, tapho-nomically deformed specimen Arago 21, the virtual reconstruction that had previously been produced36 was used in the comparative facial analysis of Apidima 2. Minimal reconstruction based on the surrounding anatomy was allowed during data collection, and landmarks that were missing on one side were reconstructed through reflected relabelling40, or by using a function in R41 based on a previously published study42. This function estimates a mirroring plane based on the unilateral landmarks. The missing landmarks are then reflected according to this plane. After the reconstruction of missing landmarks, the semilandmarks were slid along their respective closed curves using the Morpho package43 in R. Sliding was performed using the minimized bending energy algorithm44. After sliding, the data were exported in Morphologika format for further analysis45.
Data analysis. The compiled datasets were imported in Morphologika45 and super-imposed using GPA, which translates the specimen configurations to common origin, scales them for size and rotates them to best fit. Procrustes distances among specimens are a measure of overall shape difference. The superimposed coordinates of the comparative samples, excluding the Apidima specimens, were used as var-iables in a PCA, performed in the Past 3.04 software46. The resulting eigenvectors (principal component loadings) were used to compute the principal component scores for the Apidima specimens to plot them into the PCA graphs after the latter had been calculated on the basis of the comparative samples only. PCA plots were processed using Adobe Illustrator and extracted as Adobe PDF files. Furthermore, linear discriminant analyses (LDAs) and classification analyses were performed in Past 3.04 using the principal components as variables, in each case treating the reconstructions of Apidima 1 and Apidima 2 as unknown. The number of principal components included in the LDA for each of the 5 analyses included the first 7, 8, 8, 4 and 4 principal components, accounting for 70.72%, 91%, 88.6%, 85.4% and 78.2% of the total variance, respectively. Posterior probabilities were calculated with the SPSS software package (IBM, version 24 for Windows). We investigated whether the datasets used met the LDA assumptions47. We verified that all variables (principal component scores) showed an approximately normal distribution on the basis of both histograms and normal probability plots47. We removed potential outliers from the analysis by excluding pathological or taphonomically distorted specimens. On the basis of z-score analyses47, we found that outliers were absent in all variables, except for one case in PC3 of analysis 2: the MPA individual Omo 2, for which the z-score was 0.08 points over the maximum acceptable limit47 of 3.29. Given the limited number of well-preserved MPA crania in the fossil record, we decided to maintain this specimen in the analysis to maximize the representation of this group. Finally, the covariance matrices were similar among groups in all analyses, and Box’s M-tests showed that they were homogeneous for the samples used in analyses 4 and 5 (resulting P values were 0.19 and 0.07, respectively)47. However, this assumption could not be tested using Box’s M-test for most analyses owing to the small sample sizes of certain fossil groups, a common problem in palaeontology48. Because of these limitations, the results of the LDAs must be approached with caution, and not be interpreted in isolation, but in the context of all analyses presented here.
Visualization. Shape changes along principal component axes were visualized in Morphologika45. To further aid in visualization of shape differences between Apidima 1 and Apidima 2 (Extended Data Fig.9), we conducted manual super-impositions of their three-dimensional models in the software environment of Avizo 9.2.0 Lite (Visualization Sciences Group). Apidima 2 stayed in its original configuration and manipulations were carried out on Apidima 1. In the first step of superimposition, Apidima 1 was scaled to the biauricular breadth of Apidima 2. The transmeatal axes of both specimens were matched by translating and rotating Apidima 1. In the last step, Apidima 1 was rotated around the transmeatal axis to match the orientations of the external auditory meatus and the supramastoid crest of Apidima 2.
Shape index. The globular shape of the modern human neurocranium is consid-ered derived for modern humans and differentiates them from Neanderthals and other archaic Homo. It has recently been shown17 that a less-globular cranial shape in modern Europeans is related to the presence of specific Neanderthal alleles in their genome. We calculated the shape index for the posterior neurocranium of Apidima 1, to approximate the globularization index of this previous study17. We calculated an axis between the mean shapes of our Neanderthal sample and a Neanderthal-unadmixed, modern African sample (Zulu, Dart Collection, University of the Witwatersrand, n=15) and projected all other specimens (Apidima 1, MPE, MPA and fossil H. sapiens) onto this axis, to further evalu-ate the degree of globularity of the Apidima 1 neurocranium (Fig.3c, Extended Data Fig.8).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this paper. Data availabilityThe data that support the findings of this study are available from the corresponding authors upon reasonable request. 34.
Katerina Harvati1,2,3*, Carolin Röding1, Abel M. Bosman1,2, Fotios A. Karakostis1, Rainer Grün4, Chris Stringer5, Panagiotis Karkanas6, Nicholas C. Thompson1,3, Vassilis Koutoulidis7, Lia A. Moulopoulos7, Vassilis G. Gorgoulis8,9,10* & Mirsini Kouloukoussa
Source: Apidima Cave fossils provide earliest evidence of Homo sapiens in Eurasia
1 Paleoanthropology, Senckenberg Centre for Human Evolution and Palaeoenvironment, Eberhard Karls University of Tübingen, Tübingen, Germany.
2DFG Centre of Advanced Studies ‘Words, Bones, Genes, Tools’, Eberhard Karls University of Tübingen, Tübingen, Germany.
3Museum of Anthropology, Medical School, National and Kapodistrian University of Athens, Athens, Greece.
4 Australian Research Centre for Human Evolution, Griffith University, Nathan, Queensland, Australia.
5Centre for Human Evolution Research, Department of Earth Sciences, The Natural History Museum, London, UK. 6Malcolm H. Wiener Laboratory for Archaeological Science, American School of Classical Studies at Athens, Athens, Greece.
7First Department of Radiology, National and Kapodistrian University of Athens, Athens, Greece.
8Department of Histology and Embryology, Medical School, National and Kapodistrian University of Athens, Athens, Greece.
9Biomedical Research Foundation of the Academy of Athens, Athens, Greece.
10Division of Cancer Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, Manchester Cancer Research Centre, NIHR Manchester Biomedical Research Centre,, University of Manchester, Manchester, UK. *e-mail: katerina.harvati@ifu.uni-tuebingen.de; vgorg@med.uoa.gr
1. Dennell, R. W., Martinón-Torres, M. & Bermúdez de Castro, J. M. Hominin variability, climatic instability and population demography in Middle Pleistocene Europe. Quat. Sci. Rev. 30, 1511–1524 (2011).
2. Tourloukis, V. & Harvati, K. The Palaeolithic record of Greece: a synthesis of the evidence and a research agenda for the future. Quat. Int. 466, 48–65 (2018).
3. Roksandic, M., Radović, P. & Lindal, J. Revising the hypodigm of Homo heidelbergensis: a view from the Eastern Mediterranean. Quat. Int. 466, 66–81 (2018).
4. Pitsios, T. K. Paleoanthropological research at the cave site of Apidima and the surrounding region (South Peloponnese, Greece). Anthropol. Anz. 57, 1–11 (1999).
5. Bartsiokas, A., Arsuaga, J. L., Aubert, M. & Grün, R. U-series dating and classification of the Apidima 2 hominin from Mani Peninsula, Southern Greece. J. Hum. Evol. 109, 22–29 (2017).
6. Harvati, K., Stringer, C. & Karkanas, P. Multivariate analysis and classification of the Apidima 2 cranium from Mani, Southern Greece. J. Hum. Evol. 60, 246–250 (2011)
7. Liritzis, Y. & Maniatis, Y. ESR experiments on quaternary calcites and bones for dating purposes. J. Radioanal. Nucl. Chem. 129, 3–21 (1989).
8. Rondoyanni, T., Mettos, A. & Georgiou, C. Geological–morphological observations in the greater Oitilo-Diros area, Mani. Acta Anthropol. 1, 93–102 (1995).
9. Coutselinis, A., Dritsas, C. & Pitsios, T. K. Expertise médico-légale du crâne pléistocène LAO1/S2 (Apidima II), Apidima, Laconie, Grèce. L’Anthropologie 95, 401–408 (1991).
10. Arsuaga, J. L. etal. Neandertal roots: cranial and chronological evidence from Sima de los Huesos. Science 344, 1358–1363 (2014).
11. Stringer, C. The origin and evolution of Homo sapiens. Phil. Trans. R. Soc. Lond. B 371, 20150237 (2016).
12. Galway-Witham, J. & Stringer, C. How did Homo sapiens evolve? Science 360, 1296–1298 (2018).
13. Harvati, K. in Handbook of Paleoanthropology (eds Henke, W. & Tattersall, I.) 2243–2279 (Springer, 2015).
14. Balzeau, A. & Rougier, H. Is the suprainiac fossa a Neandertal autapomorphy? A complementary external and internal investigation. J. Hum. Evol. 58, 1–22 (2010).
15. Verna, C., Hublin, J.-J., Debenath, A., Jelinek, A. & Vandermeersch, B. Two new hominin cranial fragments from the Mousterian levels at La Quina (Charente, France). J. Hum. Evol. 58, 273–278 (2010).
16. Bräuer, G. & Leakey, R. L. The ES-11693 cranium from Eliye Springs, West Turkana, Kenya. J. Hum. Evol. 15, 289–312 (1986). 17. Gunz, P. etal. Neandertal introgression sheds light on modern human endocranial globularity. Curr. Biol. 29, 120–127 (2019). 18. Hublin, J.-J. etal. New fossils from Jebel Irhoud, Morocco and the pan-African origin of Homo sapiens. Nature 546, 289–292 (2017). 19. Hublin, J.-J. The origin of Neandertals. Proc. Natl Acad. Sci. USA 106, 16022–16027 (2009). 20. Caspari, R. The Krapina occipital bones. Period. Biol. 108, 299–307 (2006). 21. Arsuaga, J. L., Martínez, I., Gracia, A. & Lorenzo, C. The Sima de los Huesos crania (Sierra de Atapuerca, Spain). A comparative study. J. Hum. Evol. 33, 219–281 (1997).
22. Prossinger, H. etal. Electronic removal of encrustations inside the Steinheim cranium reveals paranasal sinus features and deformations, and provides a revised endocranial volume estimate. Anat. Rec. 273B, 132–142 (2003).
23. Posth, C. etal. Deeply divergent archaic mitochondrial genome provides lower time boundary for African gene flow into Neanderthals. Nat. Commun. 8, 16046 (2017).
24. Mercier, N. etal. Thermoluminescence date for the Mousterian burial site of Es-Skhul, Mt. Carmel. J. Archaeol. Sci. 20, 169–174 (1993).
25. Hershkovitz, I. etal. The earliest modern humans outside Africa. Science 359, 456–459 (2018). 26. Stringer, C. & Galway-Witham, J. When did modern humans leave Africa? Science 359, 389–390 (2018).
27. Harvati, K., Panagopoulou, E. & Karkanas, P. First Neanderthal remains from Greece: the evidence from Lakonis. J. Hum. Evol. 45, 465–473 (2003).
28. Harvati, K. etal. New Neanderthal remains from Mani peninsula, Southern Greece: the Kalamakia Middle Paleolithic cave site. J. Hum. Evol. 64, 486–499 (2013).
29. Tourloukis, V. etal. New Middle Paleolithic sites from the Mani peninsula, Southern Greece. J. Field Archaeol. 41, 68–83 (2016).
30. Elefanti, P., Panagopoulou, E. & Karkanas, P. The transition from the Middle to the Upper Paleolithic in the Southern Balkans: the evidence from Lakonis 1 Cave, Greece. Eurasian Prehistory 5, 85–95 (2008).
31. Douka, K., Perlès, C., Valladas, H., Vanhaeren, M. & Hedges, R. E. M. Francthi Cave revisited: the age of the Aurignacian in south-eastern Europe. Antiquity 85, 1131–1150 (2011).
32. Lowe, J. etal. Volcanic ash layers illuminate the resilience of Neanderthals and early modern humans to natural hazards. Proc. Natl Acad. Sci. USA 109, 13532–13537 (2012).
33. De Lumley, M. A. Les Restes Humains Anténéanderthaliens Apidima 1 et Apidima 2 (CNRS, 2019).
34. Kormasopoulou-Kagalou, L., Protonotariou-Deilaki, E. & Pitsios, T. K. Paleolithic skull burials at the cave of Apidima. Acta Anthropol. 1, 119–124 (1995).
35. Zollikofer, C. P. etal. Virtual cranial reconstruction of Sahelanthropus tchadensis. Nature 434, 755–759 (2005).
36. Gunz, P., Mitteroecker, P., Neubauer, S., Weber, G. W. & Bookstein, F. L. Principles for the virtual reconstruction of hominin crania. J. Hum. Evol. 57, 48–62 (2009).
37. Harvati, K., Hublin, J.-J. & Gunz, P. Evolution of middle–late Pleistocene human cranio-facial form: a 3-D approach. J. Hum. Evol. 59, 445–464 (2010).
38. Harvati, K., Gunz, P. & Grigorescu, D. Cioclovina (Romania): affinities of an early modern European. J. Hum. Evol. 53, 732–746 (2007).
39. Slice, D. E. Morpheus et al., Java Edition. http://morphlab.sc.fsu.edu/ (The Florida State University, 2013).
40. Mardia, K. V., Bookstein, F. L. & Moreton, I. J. Statistical assessment of bilateral symmetry of shapes. Biometrika 87, 285–300 (2000).
41. R Development Core Team. R: A Language and Environment for Statistical Computing. http://www.R-project.org/ (R Foundation for Statistical Computing, 2008).
42. Claude, J. Morphometrics with R (Springer Science & Business Media, 2008).
43. Schlager, S. in Statistical Shape and Deformation Analysis (eds Zheng, G. etal.) 217−256 (Academic, 2017).
44. Bookstein, F. L. Landmark methods for forms without landmarks: morphometrics of group differences in outline shape. Med. Image Anal. 1, 225–243 (1997).
45. O’Higgins, P. & Jones, N. Morphologika: Tools for Shape Analysis. Version 2.2 https://sites.google.com/site/hymsfme/resources (Hull York Medical School, 2006).
46. Hammer, Ø., Harper, D. A. T. & Ryan, P. D. PAST: paleontological statistics software package for education and data analysis. Palaeontol. Electronica 4, 1–9 (2001).
47. Field, A. Discovering Statistics using SPSS (Sage, 2013).
48. Brown, P. Nacurrie 1: mark of ancient Java, or a caring mother’s hands, in terminal Pleistocene Australia? J. Hum. Evol. 59, 168–187 (2010).
Acknowledgements
This research was supported by the European Research Council (ERC CoG no. 724703) and the German Research Foundation (DFG FOR 2237). We thank all curators and their institutions for access to original specimens or casts used in this study; T. White, B. Asfaw, M. López-Soza, V. Tourloukis, D. Giusti, G. Konidaris, C. Fardelas and O. Stolis for their input and assistance; A. Balzeau (Muséum National d’Histoire Naturelle; MNHN), E. Delson (New York Consortium in Evolutionary Primatology; NYCEP), L. Leakey (africanfossils.org) for providing access to three-dimensional models of specimens used in our figures. C.S.’s research is supported by the Calleva Foundation and the Human Origins Research Fund. We are grateful to S. Benazzi, E. Delson and I. Hershkovitz for their comments and suggestions. Author contributions K.H., M.K. and V.G.G. designed the research; V.K. and L.A.M. carried out the computed tomography scans; C.R. and A.M.B. generated the virtual reconstructions; K.H., C.S. and C.R. collected comparative data; K.H., C.R., A.M.B., F.A.K. and N.C.T. processed and analysed the data; R.G. dated the specimens; P.K.and R.G. provided stratigraphic and geological interpretations; all authors contributed to compiling the manuscript.
