Contributed by Juan Luis Arsuaga, January 27, 2017 (sent for review November 21, 2016; reviewed by William Henry Gilbert and Giorgio Manzi)
We describe a recently discovered cranium from the Aroeira cave in Portugal dated to around 400 ka. This specimen is the westernmost Middle Pleistocene cranium of Europe and is one of the earliest fossils from this region associated with Acheulean tools. Unlike most other Middle Pleistocene finds, which are of uncertain chronology, the Aroeira 3 cranium is firmly dated to around 400 ka and was in direct association with abundant faunal remains and stone tools. In addition, the presence of burnt bones suggests a controlled use of fire. The Aroeira cranium represents a substantial contribution to the debate on the origin of the Neandertals and the pattern of human evolution in the Middle Pleistocene of Europe.
The Middle Pleistocene is a crucial time period for studying human evolution in Europe, because it marks the appearance of both fossil hominins ancestral to the later Neandertals and the Acheulean technology. Nevertheless, European sites containing well-dated human remains associated with an Acheulean toolkit remain scarce. The earliest European hominin crania associated with Acheulean handaxes are at the sites of Arago, Atapuerca Sima de los Huesos (SH), and Swanscombe, dating to 400–500 ka (Marine Isotope Stage 11–12). The Atapuerca (SH) fossils and the Swanscombe cranium belong to the Neandertal clade, whereas the Arago hominins have been attributed to an incipient stage of Neandertal evolution, to Homo heidelbergensis, or to a subspecies of Homo erectus. A recently discovered cranium (Aroeira 3) from the Gruta da Aroeira (Almonda karst system, Portugal) dating to 390–436 ka provides important evidence on the earliest European Acheulean-bearing hominins. This cranium is represented by most of the right half of a calvarium (with the exception of the missing occipital bone) and a fragmentary right maxilla preserving part of the nasal floor and two fragmentary molars. The combination of traits in the Aroeira 3 cranium augments the previously documented diversity in the European Middle Pleistocene fossil record.
Ongoing research and excavations since 1987 at the Almonda cluster of paleoanthropological localities in central Portugal (Fig. 1, Fig. S1, and SI The Gruta da Aroeira Site) have yielded human remains and rich archaeological levels of the Lower, Middle, and Upper Paleolithic as well as Early Neolithic and later prehistoric periods (1⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓–12). Within the Almonda karst system, the Gruta da Aroeira was first investigated from 1998–2002 (13), revealing a rich lithic assemblage with Acheulean bifaces (Fig. S2) associated with faunal remains and two human teeth (Fig. S3): Aroeira 1 (a left mandibular canine) and Aroeira 2 (a left maxillary third molar). Aroeira 1 is moderately large, especially compared with the Atapuerca (SH) sample, and Aroeira 2 is among the larger of the Middle Pleistocene upper right third molars (6, 14). They fit morphologically within the known variation of European Middle Pleistocene dentitions, although Aroeira 2 has a relatively large hypocone (6, 15).
(A) Geographical location of Gruta da Aroeira and main sites mentioned in the text. (B) Detail of the excavation area and provenance of the ARO2 U-series sample. (C) Stratigraphic profile and cranium provenance (denoted by its field inventory no. 606).
(A and E) Almonda escarpment with the position of Gruta da Aroeira and the Almonda River in the foreground. (B) General view of Gruta da Aroeira and the location of the main dated speleothems. (C) Estremadura Limestone Massif with the position of the Almonda spring, the Tagus River, and the Serra d’Aire. (D) Schematic cross-section of the Almonda karst system. (F) Acheulean biface (flint) from level Xb/c. (G) Gruta da Aroeira site plan.
Acheulean handaxes (bifaces) from level Xb/c. (A) Flint. (B–E) Quartzite.
(A and B) The Aroeira 1 left mandibular canine (A) and the Aroeira 2 left maxillary third molar (B) in mesial (Mes), distal (Dist), buccal (Bucc), and lingual view (Ling). (C) Aroeira 3 maxilla in medial view. Arrows indicate the lateral nasal crest (1), turbinal crest (2); nasal floor (3), and root of the inferior nasal concha (4). (Scale bar, 2 cm.)
Renewed fieldwork in 2013, focused on establishing the chronology of the sequence via U-series dating of interstratified flowstone deposits (Figs. 1 and 2), led to the discovery of a human cranium (Aroeira 3) encased in hard breccia toward the base of the sequence (Fig. 3 and Fig. S4).
Stratigraphic longitudinal profile of the Gruta da Aroeira, with the location of dating samples (nos.1–8) and human remains indicated. Dating samples are referred to in Table S1.
Various stages during the in situ extraction and restoration process of the fossil, and reconstruction of the Aroeira 3 cranium after cleaning. (A) Outline of the cranium embedded in the breccia. (B) Location of the cranium after its protection with gauze coating and polyurethane resin. (C) Cutting of the breccia with a circular saw to remove the cranium. (D) Protection of the fossil with wooden boards during the fracturing of the speleothem adhering to the wall with a pneumatic hammer. (E) Final stage of the extraction of the cranium. Wooden boards were placed along both sides of the wall while the breccia block containing the cranium was cut along the bottom. (F) Main portion of the cranium embedded within the breccia block. (G) Isolated cranial fragments before restoration. (H) Main portion of the cranium during the removal of the hardened sediments from the endocranial surface. (I) Main portion of the cranium during the cleaning process (Left) and reconstruction of some of the isolated cranial fragments (Right). (J) Detail of the cranial base and temporal bone during the cleaning process. The arrow indicates a thin layer of speleothem coating which remains on the superior portion of the petrous pyramid. (K) Fragments comprising the cranium before reconstruction. (L) Manually joining the fragments together with adhesive. (M) Endocranial view of the reconstructed cranium. (N) Lateral view of the reconstructed cranium. (Scale bars, 5 cm.) Images in K–N are from J. Trueba (photographer).
The Aroeira stratigraphy spans a thickness of ∼4 m and comprises three major stratigraphic units (Fig. 2 and SI Stratigraphic Outline). The cranium was recovered from unit 2, a 2.2-m-thick mud-supported breccia rich in angular and subangular clasts. This unit corresponds to the Acheulean layer X–Xb/c [the upper and lower parts of a single layer excavated in 1998–2002 (X) and in 2013–2015 (Xb/c)], and the overlying ARO2 flowstone has yielded a minimum age of 417.7 + 37.3/−27.5 ka (SI U-Series Results and Table S1) (9). A further uranium-thorium (U-Th) age of 406 ± 30 ka for the outer layer of a stalagmitic column (SI Chronostratigraphy) covered by unit 2 provides a maximum age for the sequence and allows correlation of it to Marine Isotope Stage (MIS) 11. Two additional U-Th ages of 390 ± 14 ka and 408 ± 18 ka for calcitic crusts that formed on the cranium provide additional, consistent minimum age constraints for the cranium itself (SI Chronostratigraphy). Thus, the Aroeira 3 cranium most likely dates to 390–436 ka.
U-Th results from Aroeira speleothems, calcite samples adhering to the cranium, and the stalagmite column
Our excavation of layer X–Xb/c covered an area of 6 m2 and a depth of 1 m. The lithic assemblage recovered (n = 387) includes handaxes and other bifacial tools (n = 17), other retouched tools (n = 27), cores (n = 43), flakes or flake fragments and debris (n = 180), and tested or untested cobbles (possibly manuports) (n = 114). Quartzite is the raw material of choice, whereas flint is scarce, but both are represented among the handaxes (Fig. S2). The Levallois method is absent.
The faunal remains are highly fragmented, mainly consisting of isolated teeth, phalanges, carpal/tarsal bones, and antler fragments. Among the 209 piece-plotted faunal remains from layer X–Xb/c, cervids [number of identified specimens (NISP) = 58], including both Dama and Cervus, and equids (NISP = 46) predominate. Rarer species include Rhinocerotidae (NISP = 2) (likely Stephanorhinus cf hundsheimensis), and bear (NISP = 4) (Ursus sp.), as well as a large bovid (Bos/Bison), a caprid (Caprinae), and a tortoise (Testudo sp.) (NISP = 1 each). Several burnt bone fragments were recovered at the base of layer Xb/c in association with the stone tools and the human cranium.
Gruta da Aroeira (39° 30′ 20′′ N; 08° 36′ 57′′ W) is located in Torres Novas, central Portugal (Fig.1). The cave is part of the Almonda karst system, a labyrinthine network of passages containing a number of former entrances with Pleistocene infillings completely sealed by roof collapses. This system is formed of passages excavated at different elevations whose intersections with the 70-m-high escarpment rising above the extant spring of the Almonda River, a tributary of the Tagus River, correspond to fossil outlets of its subterranean course (Fig. S1).
Published reports of previous excavation work, carried out between 1998 and 2002, have used the designation of “Galerias Pesadas” for this site (6). However, this designation corresponds to inner conduits of the karst system situated at this elevation that, at present, remain unconnected to the exterior. In 1991, speleoarcheological surveys identified a cone of sediments in these conduits with Pleistocene fauna and stone tools (3) and eventually led to the location of a collapsed, sediment-sealed entrance—the Gruta da Aroeira. The 1998–2002 excavations, as well as the new phase of the project begun in 2013, were carried out here (Fig. 1 and Fig. S1).
At the back of the cave, where bedrock probably has been reached, the Aroeira stratigraphy spans a thickness of ∼4 m and comprises three major stratigraphic units. Uppermost unit 1 is an infilling breccia capped by the ARO1 flowstone. Unit 2 is a 2.2-m-thick mud-supported breccia rich in angular and subangular clasts, corresponding to the Acheulean layer X–Xb/c (upper and lower parts of a single layer, excavated in 1998–2002 and 2013–2015, respectively), and is capped by the ARO2 flowstone. The human cranium (inventory no. ARO-606, found in square H6, at a depth of 608 cm below datum) was recovered at the base of this unit (Fig. 1 and Fig. S4). Basal unit 3 is an endokarst fluvial deposit and comprises two layers: XI is a 0.4-m-thick layer of silty sand with scattered gravel and faunal remains but no artifacts, and XII is a 0.5-m-thick, archeologically sterile, layer of slightly gravelly sand.
U concentrations of flowstones ARO1 and ARO2 are between 40 and 90 ng/g (Table S1). The base of stalagmite BL1 has a higher U concentration of 171 ng/g. The calcite crystals that formed in sediment voids (ARO14-03 and ARO14-04) have U concentrations of 191 and 129 ng/g, respectively; the outer layer of the stalagmite column (ARO14-H6-727) has a concentration of 105 ng/g; and a postunit 2 sedimentation crust on top of the column has much higher U concentration of 613 ng/g. The calcite crusts that formed on the cranium (ARO-SK4 and ARO-SK6) have U concentrations of 305 and 273 ng/g, respectively.
232Th is a proxy for detrital components in the calcium carbonate, and the degree of contamination is assessed by today’s 230Th/232Th activity ratio. A correction for the detrital contribution is done using a detrital 238U/232Th activity ratio of 0.8 with 50% uncertainty and also assuming secular equilibrium of the 238U decay chain in the detritus. ARO1 has a high degree of detrital components, and two of three dating results (ARO1-1 and ARO1-3) are dominated by detritus, resulting in high uncertainties. ARO1-2 also has a high detrital contribution, but the corrected age is robust. The results for the column and the precipitates on the cranium also show a high detrital contribution, with 230Th/232Th around 30 for ARO-SK-4 and ARO-SK-6 and >10 for the column sample. However, the detrital correction is not significant for samples with ages around 400 ka, so the corrected and uncorrected results agree within uncertainties, and the results are robust. All the other samples have low detrital components, and the correction is not significant. The uncorrected and corrected ages are reported in Table S1 and are discussed in the next section.
A minimum of five site formation phases can be differentiated (Fig. 2 and Table S1). Episode 5 corresponds to the period predating the opening of the cave and includes layer XII (unit 3) and the formation of a massive stalagmitic column that, based on available cross-sectional views, would seem to have grown from its top. The column’s last layer of calcite, formed before the accumulation of the deposit that eventually buried it, dates to 406 ± 30 ka (sample no. 8) and provides a minimum age for the sequence of events subsumed under episode 5 and a maximum age for the deposition of the Aroeira 3 cranium inside the cave.
Episode 4 is represented by the accumulation of layer X (the Acheulean deposit) and layer XI. The age of the cranium, found at the base of layer X, is bracketed by the age of the outermost flowstone layer of the stalagmitic column formed during the previous phase, against which it leaned and the basal age of the ARO2 flowstone (sample no. 2.1; 418/+37/−27 ka); therefore, the age of the cranium must lie in the 390–436 ka interval. The dating of calcite crystals sampled from the cranium’s breccia fill (samples no. 3 and 4) has provided minimum ages (390 ± 14 ka and 408 ± 18 ka, respectively) consistent with this stratigraphic bracketing (Table S1).
Episode 3 corresponds to a sedimentation hiatus, probably because of the infilling of the entrance to the cave. During the hiatus, the ARO2 flowstone formed over the then-extant cave floor. In areas where the fill was not fully sealed, and as a result of subsidence/suffosion, water percolating from above precipitated calcite in the voids and along cracks and fissures formed between sediment and cave wall or between sediment and the massive stalagmitic column of episode 1. This water percolation would explain the results of 300 ± 3 ka for crystals formed in voids toward the top of layer X (samples no. 5 and 6) and of 236 ± 3 ka for calcite formed between the sediment column and the outer, sediment-dirty rim of the stalagmitic column (sample no. 7). Together with the dates obtained for the upper part of the ARO2 flowstone (sample no. 2.2.; 326.4 ± 13.4 ka) and for the BL1 stalagmite (sample BL1.1; 278.5 ± 12.7 ka) found exteriorly at the Brecha das Lascas locus but in the same stratigraphic position, these results place the end of this episode after 239 ka.
Episode 2 corresponds to a collapse of the roof in the area of rows 10–14 of the grid. A second round of infill deposits eventually filled the newly formed shaft.
Episode 1 corresponds to the formation of the ARO1 flowstone (sample no. 1). This flowstone is found at the top of the fill in the interior space behind the new infill deposit, for which it provides a minimum age of 44.8 ± 2.0 ka.
Based on these results, we propose the following correlation of the sequence with the global record of MIS: episode 5, MIS 11c or earlier; episode 4, MIS 11c, ∼390–420 ka; episode 3, MIS 7a–11b, ∼190–390 ka; episode 2, MIS 4–7a, ∼60–190 ka; and episode 1, MIS 3c ∼40–60 ka.
The Aroeira 3 cranium was painstakingly extracted from the hard calcareous breccia and restored (SI In Situ Extraction of the Fossil and Fig. S4). The cranium is taphonomically broken obliquely to the sagittal plane, with the preserved bone margin running diagonally from the left supraorbital arch anteriorly, crossing the midline just anterior to bregma, and continuing posteriorly toward the right asterion. Approximately half of the right parietal bone and the right half of the frontal bone are preserved. A circular portion of the right frontoparietal region was originally present but was destroyed in situ in the act of discovery (Fig. 3).
In addition, portions of the sphenoid and the nearly complete temporal bone are preserved, as well as the medial portion of the left supraorbital arch, the interorbital pillar (including the superior portions of both nasal bones), and most of the right supraorbital arch. The outer surface of the supraorbital torus is preserved only in the glabellar region and in the medial half of the right supraorbital arch (we use the term “supraorbital arch” to refer to the part of the supraorbital torus lateral to the glabellar region; thus, it comprises the supercilliary arch, the supraorbital margin, and the lateral trigone). In addition, a fragment of the right maxilla includes the lower border of the nasal aperture and a part of the anterior nasal floor. A small portion of the alveolar process of the right maxilla is also present, with two fragmentary molars partially preserved.
The cranial landmarks nasion, glabella, right asterion, right auriculare, and right porion are preserved. Although the bregma is not preserved, its position can be estimated accurately, because the coronal suture is preserved up to a point very close to the bregma. A remnant of the metopic suture is preserved near the glabella, as well as the right parietomastoid and occipitomastoid sutures and the segment of the lambdoid suture on the right parietal bone closer to the asterion. Internally, the frontal crest, foramen cecum, and crista galli are preserved in the anterior cranial fossa.
The coronal suture is fully fused, and there are no traces of the suture on the endocranial surface. In addition, the preserved teeth show fully formed roots, with closed apices, and the broken tooth crowns are worn flat. Although the enamel is present over nearly the entire preserved crown surface in both teeth, it is not possible to study the cusp pattern or details of occlusal anatomy, precluding a phylogenetic analysis of the dental morphology. Nevertheless, these observations collectively indicate an adult age for this individual.
Two main supraorbital torus morphologies can be found in European Middle Pleistocene fossils. In many of them, the supraorbital arches are curved mediolaterally (in frontal view) and rounded on their anterior surface. The two arches can fuse completely in a swollen glabellar region (Fig. 3 and Fig. S5), or they can remain more or less separated in the midplane by a glabellar depression. This supraorbital morphology, with different degrees of glabellar fusion, is found in the large Atapuerca Sima de los Huesos (SH) sample, in the Bilzingsleben B1, Steinheim, and Petralona crania, and in the Late Pleistocene Neandertals (16). It is also seen in the Early Pleistocene Atapuerca Gran Dolina specimen ATD6-15. However, the Arago 21 and Ceprano specimens depart from this condition, resembling the Middle Pleistocene African specimens from Kabwe and Bodo in which the two supraorbital arches are well separated at the glabella and are flatter and less curved (17, 18). Despite the loss of the outer surface over much of the supraorbital torus, it is clear to us that the supraorbital arches in Aroeira 3 are fused in a swollen glabella (i.e., unlike the Ceprano and Arago 21 crania, the supraorbital torus in Aroeira 3 is not medially concave). Although the precise morphology of the supraorbital arches is more difficult to assess, the better-preserved right side seems to show a rounded condition, and the Bilzingsleben B1 specimen represents the closest Middle Pleistocene match to the Aroeira 3 supraorbital torus (Fig. S5).
Virtual reconstruction of the Aroeira 3 cranium in frontal (A), posterior (B), superior (C), and endocranial (D) views. The frontal sinus in D is exposed in a parasagittal section located 4 mm to the right of the sagittal plane. (E) Virtual reconstruction of the Aroeira 3 cranium in a three-quarters view compared with Bilzingsleben B1 (cast).
Numerous Middle Pleistocene fossils, including the Kabwe, Bodo, Arago 21, and Petralona specimens (and perhaps the Steinheim specimen, although it has some deformation in this region) exhibit a nasion that is depressed with respect to glabella. In contrast, Neandertals and the Atapuerca (SH) sample show a nasion that projects to the same degree as glabella. The Aroeira 3 cranium, like the Bilzingsleben B1 specimen, is intermediate in this trait (Figs. S5–S7).
The Aroeira 3 cranium compared with Atapuerca SH Cranium 4 and Cranium 5. Atapuerca SH Cranium 5 is oriented in the Frankfurt horizontal orientation; this orientation is estimated for the Aroeira 3 cranium and Atapuerca SH Cranium 4. Aroeira 3 shows a slightly depressed nasion (n) with respect to the glabella (g), whereas in the Atapuerca SH Cranium 5 the nasion and glabella are located in the same vertical plane. The mastoid projection measured from the parietal incisure is shorter in the Aroeira 3 cranium than in the Atapuerca (SH) crania, but the projection is similar in all three crania when measured from the level of the porion. It is possible that the relatively high position of the transverse sulcus in the Aroeira 3 cranium (see Text) is related to the low position of the parietomastoid suture and asterion (a). A clear torus angularis (t.a.) is present is both the Aroeira 3 cranium and Atapuerca SH Cranium 4 but can only be insinuated in Atapuerca SH Cranium 5.
The Aroeira 3 cranium compared with Atapuerca SH Cranium 5 and the Steinheim specimen. The Frankfurt horizontal orientation is estimated in both the Aroeira 3 cranuim and the Steinheim specimen. The nasion seems depressed relative to the glabella in the Steinheim cranium, but the strong deformation of this specimen precludes a conclusive assessment. The Aroeira 3 cranium is reminiscent of the Steinheim specimen in both the low position of the parietomastoid suture (compared with Atapuerca SH Cranium 4 and 5) and the corresponding short mastoid process when measured from the parietal incisure.
Despite some abrasion of the outer surface, the right and left supraorbital arches are thick as compared with the majority of the European or African Middle or Late Pleistocene fossils (16, 19). The maximum midorbit thickness of the torus (19.0 mm) can be taken on the right side and is similar to that of the Bodo and Ceprano crania (each 17.5 mm), although the torus of the Bilzingsleben B1 cranium (21–22 mm, right side, cast measurement) is even thicker. The interorbital pillar is very broad. Although the dacryon and maxillofrontale landmarks cannot be located precisely, the distance between the two inner orbital borders is large (34–35 mm) and similar to the Atapuerca SH Cranium 4 (38.0 mm) and the Kabwe (32.0 mm), Bodo (37.5 mm), and Bilzingsleben B1 (35.5 mm, on cast) specimens. The frontal sinuses in the Aroeira 3 cranium are well developed (Fig. S7) but are not as large laterally (in the torus) or superiorly (in the frontal squama) as in the Petralona specimen.
The lateral wall of the nasal cavity preserves the root of the inferior nasal concha as well as a sharp lateral nasal crest and a smooth turbinal crest (Fig. S3). An “internal nasal margin” (a “frame” made by the fusion of the turbinal and spinal crests) and a “medial projection” (bone swelling above the inferior nasal concha) described in Neandertals (20) are lacking in the Aroeira 3 cranium. This apparently derived internal nasal morphology in Neandertals is also absent in the Atapuerca (SH) sample but has been identified in the Steinheim cranium (20).
The nasal floor in the Aroeira 3 cranium is sufficiently preserved to determine that it is not bilevel. The Aroeira 3 cranium resembles the Gran Dolina and Sima de los Huesos hominins in showing a level or sloped configuration (the ancestral condition for Homo) but differs from the bilevel form present in most Late Pleistocene Neandertals and in the Steinheim, Petralona, east Asian archaic Homo, and some African Middle Pleistocene crania (21, 22).
Virtual reconstruction of the Aroeira 3 cranium by mirror-imaging the right side (Fig. S5) shows that the parietal walls are nearly vertical. However, the maximum cranial breadth is located at the supramastoid crest, as in other earlier Middle Pleistocene European fossils. This morphology departs from the ancestral condition seen in Homo erectus of strongly convergent parietal walls superiorly and also from the more circular contour in posterior view of late Middle and Late Pleistocene Neandertals (16).
The auriculo-bregmatic height and frontal sagittal chord (with the bregma reconstructed) can be measured directly on the original fossil, and a number of bilateral measurements can be estimated in the virtual reconstruction (Table S2). Compared with the Atapuerca (SH) sample, the Aroeira 3 cranium is closer to Cranium 5 (1,090 cm3) in the transverse diameters taken at both the auricular point and the supramastoid crest (i.e., on the temporal bone) and is intermediate or closer to Cranium 4 (1,390 cm3) in the rest of the measurements. Thus, a cranial capacity above 1,100 cm3 can safely be established.
Principal measurements (in millimeters) in Aroeira 3 compared with the Atapuerca (SH) sample
An angular torus is present in the mastoid angle of the parietal bone in the Aroeira 3 cranium (Fig. S6). Among European Middle Pleistocene specimens, this feature is found only in a few large, robust individuals, including the Ceprano and Arago 47 crania and Atapuerca SH Cranium 4 (18, 23). In the Aroeira 3 cranium the parietal bone is ∼9 mm thick near the bregma and is 10.2 mm thick at the thickest point along the break (in the area of the parietal boss). These values are within the range of the Atapuerca (SH) sample. The thickness in the Aroeira 3 cranium near the asterion, at the angular torus, is large (14.6 mm) as it is in the other fossils that show this feature, including Atapuerca SH Cranium 4 (17.0 mm) and the Arago specimen (13.5 mm, on cast).
The temporal bone is nearly complete and preserves several phylogenetically relevant features. Although the squamosal portion is largely preserved, abrasion along the superior margin makes the height and curvature difficult to discern. The styloid process is fused to the basicranium, and there is a large and triangular postglenoid process. In both these features, the Aroeira cranium is different from Asian H. erectus but resembles the Atapuerca (SH) specimens (24). On the other hand, the articular eminence in the Aroeira 3 cranium is raised, unlike the derived flattened articular eminence in the Atapuerca (SH) crania, Steinheim and Petralona specimens and the Neandertals (25). The sphenoid bone contributes slightly to the medial wall of the glenoid fossa, but there is variation in this trait in the Atapuerca (SH) sample (24) and Neandertals. Other Neandertal derived features are also absent in the Aroeira 3 temporal bone, including an anterior bridge in the digastric groove or an external auditory meatus located at the same level as the zygomatic process root (26, 27).
In inferior view, the right mastoid process in the Aroeira 3 cranium projects well beyond the level of the occipitomastoid suture, whereas in most Neandertals the mastoid process characteristically does not project beyond the basicranium (24). Although different measurement techniques for mastoid projection have been proposed, we have measured the projection of the mastoid tip from the parietal incisure, because this measurement is not dependent on orienting the specimen in the Frankfurt horizontal orientation and has been found to differentiate Neandertals from other groups (25). When measured from the parietal incisure (Figs. S6 and S7), the mastoid projection in the Aroeira 3 cranium is low (33 mm), close to the Neandertal mean (36.4 ± 4.3) and to that of the Middle Pleistocene Steinheim specimen (31 mm, left side, on cast) and well below the range of values (40.0–50.0 mm) in the Atapuerca (SH) sample (25). The digastric groove is deep and lacks a paramastoid crest (a bony ridge located between the mastoid process and the occipitomastoid suture, sometimes called the “juxtamastoid eminence”), but there is variation of this feature in modern and fossil populations, including the Atapuerca (SH) sample.
Internally, the transverse sulcus crosses the parietal bone above the asterion before entering the temporal bone (Fig. S5). This condition is observed in modern humans and some Neandertal specimens, but the transverse sulcus generally does not cross the parietal bone in earlier hominins, including the Atapuerca (SH) sample (16, 28).
The cranium, heavily fossilized and well preserved, was contained within a cemented breccia (Fig. S4). Because of the extreme hardness of the sediments and the difficulty of excavation, several fragments of the cranium were separated from the main portion at the moment of discovery by the impact of the heavy-duty demolition hammer being used at the time (hence, the circular hole apparent in Fig. 3 and Figs. S4–S7). The detached fragments and the contour and thickness of the sectioned cranial vault made it immediately apparent that a human fossil had been hit by the tool. Work was interrupted on the spot, all the detached fragments were collected, and preparations for extraction of the remainder from the breccia were initiated. The visible sections of the main portion of the fossil were protected with gauze coating impregnated with Paraloid B-72 (Rohm & Haas) at 5–15% acetone (CH3COCH3) concentration (47⇓–49). Subsequently, the fossil was covered with a polyurethane resin to protect it further during the rock-cutting, with appropriate machinery, of a large block of the hard calcareous breccia that contained it (Fig. S4).
Subsequent to the acquisition of the Portuguese Heritage authority's pertinent temporary export permit, the block containing the main portion of the cranium and the detached fragments were transported to the Conservation and Restoration Laboratory at the Centro de Investigación Universidad Complutense de Madrid-Instituto de Salud Carlos III sobre la Evolución y Comportamiento Humanos in Madrid for further preparation. The cranium was painstakingly extracted from the breccia and restored over a period of two years (Figs. S5 and S6). The most recent criteria for conservation and restoration were followed, including prioritizing the conservation of the fossil over its restoration and using products that are homogeneous and compatible with the bone as well as reversible to facilitate future interventions and avoid additional damage. A detailed diagnosis of the specimen was carried out before each intervention, and all steps during the conservation and restoration process were documented (50, 51).
Most researchers support a European origin and a local evolution for the Neandertals (23, 29⇓–31). However, despite the relatively good fossil record of Neandertals and their recently sequenced nuclear genome (32, 33), considerable debate still exists regarding the time of divergence of the lineages leading to modern humans and Neandertals and regarding the name, hypodigm, and geographic distribution of the stem species. One of the main reasons for this ongoing debate is the scarcity, generally fragmentary preservation, and often unclear chronology of most European Middle Pleistocene fossils, and the taxonomic classification of many of these Middle Pleistocene European fossils (as well as those from Africa and Asia) remains controversial (23, 31, 34⇓⇓–37).
At present three different cranial morphologies can be recognized in the European Middle Pleistocene hominin record. One is almost fully Neandertal, with nearly the entire suite of derived traits, and occurs during the last part of this period, mainly <200 ka. A second cranial configuration shows many Neandertal traits in the face, supraorbital torus, temporal bone, and mandible, but the general shape of the neurocranium (in both lateral and posterior views) is not Neandertal-like, indicating a mosaic nature for Neandertal cranial evolution.
The Atapuerca (SH) sample conforms to this cranial morphology (23), and ancient DNA analysis has shown that the Atapuerca (SH) hominins are members of the Neandertal clade (38). The incomplete braincase from Swanscombe should probably be grouped with the Atapuerca (SH) hominins as well, even though it is more Neandertal-derived in its excavated suprainiac area and its bilaterally projecting occipital torus. The Swanscombe specimen is dated to MIS 11 (29), and the Atapuerca (SH) sample is dated to MIS 11 or MIS 12 (39). The supraorbital morphology, especially in the glabellar region, where it is better preserved, and the mastoid process projection may indicate that the Aroeira 3 cranium (dated to MIS 11) also belongs in this category, even though the combination of traits is not the same as in the Atapuerca (SH) crania or any other fossil in this group.
Finally, there are other European partial crania, such as the Arago 21 and Ceprano specimens, that do not show Neandertal-derived traits in the preserved regions or in which the features are more ambiguous (36). The Aroeira 3 cranium resembles these specimens in its well-developed angular torus (also present in Atapuerca SH Cranium 4) and its lack of a flattened articular eminence.
Although the taxonomic identity of the Arago and Ceprano hominins is debated, some authors prefer to group them with other Middle Pleistocene fossils from Africa and Asia in a separate species (Homo heidelbergensis) (34, 40). This view sees the Neandertals as evolving out of H. heidelbergensis in Europe and posits a largely anagenetic (linear) evolutionary scenario. Other researchers (23) have argued for a high degree of morphological diversity in the Middle Pleistocene European hominin record, a scenario that is incompatible with an anagenetic evolutionary pattern. Evolutionary scenarios that posit a series of temporally successive grade shifts are likely to be largely a product of the general paucity and poor chronological control of the European Middle Pleistocene fossil record. Elucidating nonlinear evolutionary patterns in the hominin fossil record relies on fossil morphology, as well as geography and chronology, and the addition of relatively complete, well-dated fossils, such as the Aroeira 3 specimen, will help establish a more robust evolutionary scenario.
The Aroeira 3 cranium shows several features characteristic of European earlier Middle Pleistocene crania. However, the combination of traits in the Aroeira 3 cranium is not seen in any other Middle Pleistocene individual. The Aroeira 3 cranium shows a continuous and thick supraorbital torus similar to that of the Bilzingsleben cranium, a short mastoid process as in the Steinheim specimen, and a large, triangular postglenoid process as in the Atapuerca (SH) sample. These features are combined with a raised articular eminence, which contrasts with the flatter articular eminence generally seen in the Atapuerca (SH) sample and in the Steinheim cranium. It has been argued that a flattened articular eminence is a feature that appears very early in Neandertal evolution (23, 41).
The Aroeira, Atapuerca (SH), and Arago sites are relatively close to one another in time (400–450 ka) and space (southwestern Europe), but the fossils from these sites are clearly different. These differences suggest that intra- or interdeme hominin diversity and complex population dynamics characterized this period, including variable population replacement with varying levels of isolation and admixture (23). In fact, it has been argued that archaic paleodemes (e.g., Ceprano) could have persisted in eco-geographic refugia (36) along with more evolved paleodemes (e.g., Atapuerca Sima de los Huesos) showing Neandertal apomorphies in other regions. This same time period also documents two major technological innovations: the expansion of the Acheulean tradition (42) and the first evidence for widespread, systematic controlled use of fire (43). Both are present at the Aroeira site, whose geographic situation in extreme southwestern Europe suggests that these innovations spread quickly throughout the European continent and were largely independent of hominin morphological diversity [although with the arrival of the Acheulean industry to Western Europe, the possibility of gene flow from outside Europe should also be taken into account (38)]. Well-dated fossils, such as the Aroeira 3 cranium, with a clear technological and ecological context are crucial to building a robust evolutionary scenario during the European Middle Pleistocene.
U-Th dating was carried out on eight speleothem specimens found in stratigraphic relationship with the excavated units or the cranium itself (Table S1). We dated flowstones ARO1 and ARO2, which were capping units 1 and 2, respectively, and a basal section of a stalagmite (BL1), which formed over the flowstone that caps the Pleistocene fill exteriorly, at the Brecha das Lascas locus (Figs. 1 and 2). Results for these samples were presented and discussed in ref. 9. For the present study, we dated two additional calcite crystals which precipitated inside sediment voids of unit 2 (ARO14-03 and ARO14-04) (Table S1), providing a minimum age for the sediment accumulation. We furthermore analyzed the outer layer and a postsedimentation overgrowth of a stalagmitic column (ARO14-H6-727) covered by unit 2. We finally analyzed two calcite crusts (ARO-SK4 and ARO-SK6) that precipitated on the Aroeira 3 cranium, providing minimum ages for the specimen (Table S1). Subsamples were cut from the collected specimen using a microdrill fitted with a diamond cutting disk. The CaCO3 pieces were cleaned in an ultrasound bath and dried. Chemical separation and purification was done following previously described protocols (44). Purified U and Th fractions were analyzed in 0.5 M HCl solution by multicollector-inductively coupled plasma mass spectroscopy (MC-ICPMS). Analytical protocols for MC-ICPMS and data reduction are presented in detail in ref. 45.
The conservation process in the laboratory consisted of extracting the cranium from the breccia and adhering speleothem coating and cleaning and removing sediment and limestone pebbles between the fossil and the speleothem. This cleaning was done using two different types of drills with different steel bits and ultrasound techniques. After cleaning, the fragments were joined together using an acrylic resin adhesive, Paraloid B-72 (Rohm & Haas), at 15–30% acetone concentration (Fig. S4) (46). Once the fossil was reconstructed, a final thin layer of consolidant Paraloid B-72 (Rohm & Haas) at 3% acetone concentration was applied to the entire surface. Endocranially, a thin layer of possible speleothem remains coating the superior portion of the petrous pyramid (Fig. S4), because the benefits of removal were outweighed by possible damage to the fossil.
The Aroeira 3 cranium was subjected to high-resolution CT scanning using a YXLON MU 2000-CT scanner housed at the University of Burgos, Burgos, Spain, with the following scanning parameters: 160 kV, 4 mA, 0.5-mm slice thickness, 0.3-mm interslice distance, and a field of view of 221.88 mm. Six hundred twelve slices were obtained as a 1,024 × 1,024 matrix of 32-bit Float format with a final pixel size of 0.162 mm. Virtual reconstruction of the cranium (Fig. S5), relying on mirror-imaging across the sagittal plane, was carried out using the Mimics v.18 (Materialise, N.V.) software program. The virtual reconstruction initially aligned the two halves of the cranium relying on the recognition of homologous landmarks. Subsequently, a “best fit” of the overlapping mesh surfaces, consisting of 500,000 triangles, was carried out relying on ∼50 automated iterations. Because of the lack of direct contact, no attempt was made to situate the maxillary fragments with respect to the cranial vault.
We thank J. Trueba for photographic documentation of the restored cranium. CT scanning of the Aroeira 3 cranium was carried out at the University of Burgos Scientific Park by R. Porres and in collaboration with J. M. Carretero and L. Rodríguez from the Laboratory of Human Evolution at University of Burgos. This research was supported by the Ministerio de Economía y Competitividad of the Government of Spain, Projects HAR2014-55131, CGL2012-38434-C03-01, and CGL2015-65387-C3-2-P (MINECO/FEDER). Fieldwork was funded by Câmara Municipal de Torres Novas and Fundação para a Ciência e Tecnologia with logistical support by Fábrica de Papel A Renova. J.D. was supported by Postdoctoral Grant SFRH/BPD/100507/2014 from the Portuguese Fundação para a Ciência e a Tecnologia using funding from the European Social Fund/Operational Programme for Human Potential, and M.S was supported by Juan de la Cierva Postdoctoral Grant FJCI-2014-21386.
Author contributions: J.D., M.S., J.L.A., D.L.H., R.M.Q., E.T., and J.Z. designed research; J.D., M.S., J.L.A., D.L.H., R.M.Q., S.G., A.R., L.V., P.S., J.M., F.R., A.F., P.G., E.T., and J.Z. performed research; J.D., M.S., J.L.A., D.L.H., R.M.Q., M.C.O., E.S., E.T., and J.Z. analyzed data; and J.D., M.S., J.L.A., D.L.H., R.M.Q., M.C.O., E.S., E.T., and J.Z. wrote the paper.
Reviewers: W.H.G., California State University, East Bay; and G.M., Sapienza University of Rome.
The authors declare no conflict of interest.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1619040114/-/DCSupplemental.
Freely available online through the PNAS open access option.
