Journal of Human Evolution 131 (2019) 76e95

Contents lists available at ScienceDirect

Journal of Human Evolution

journal homepage: www.elsevier.com/locate/jhevol

New bracketing luminescence ages constrain the Sima de los Huesos hominin fossils (Atapuerca, Spain) to MIS 12

* Martina Demuro a, , Lee J. Arnold a, Arantza Aranburu b, Nohemi Sala c, d, Juan-Luis Arsuaga d, e a School of Physical Sciences, Environment Institute, and Institute for Photonics and Advanced Sensing (IPAS), University of Adelaide, North Terrace Campus, 5005 Adelaide, SA, Australia b Departamento de Mineralogía y Petrología, Facultad de Ciencia y Tecnología, Universidad del País Vasco/EHU, Edificio F3, Barrio Sarriena S/n, 48940 Leioa, Bizkaia, Spain c Centro Nacional de Investigacion sobre Evolucion Humana, Avd. Sierra de Atapuerca, 3, 09002 Burgos, Spain d Centro Mixto Universidad Complutense-Instituto de Salud Carlos III de Evolucion y Comportamiento Humanos, Avd. Monforte de Lemos 5, (Pabellon 14), 28029 Madrid, Spain e Departamento de Paleontología, Facultad de Ciencias Geologicas, Universidad Complutense de Madrid, c/ Jose Antonio Novais, Ciudad Universitaria, 28040 Madrid, Spain article info abstract

Article history: Recent chronological studies of the Sima de los Huesos (SH) hominin fossil site, Atapuerca, Spain, have Received 7 September 2018 established a close minimum age of at least 430 ka for sedimentary material immediately overlying the Accepted 5 December 2018 human remains. However, a firm maximum age limit still needs to be established for the SH fossils in order to better constrain the timing for the onset of Neandertal speciation. In the present study, we address this important chronological gap at SH by providing direct ages for the sediment deposits that Keywords: host, and immediately underlie, the hominin fossils. Depositional ages were obtained using single-grain Sima de los Huesos thermally-transferred optically stimulated luminescence (TT-OSL), a technique that has yielded reliable Geochronology ‘ ’ Middle Pleistocene extended-range luminescence chronologies at several independently dated Atapuerca sites. Four ± ± ± ± Neandertal lineage single-grain TT-OSL depositional ages of 453 56 ka, 437 38 ka, 457 41 ka and 460 39 ka were Atapuerca obtained for the red clay lithostratigraphic units (LU-5 and LU-6) found underlying and encasing the SH Western Europe hominin bones. A Bayesian age-depth model was constructed using previously published chronologies, as well as the new single-grain TT-OSL ages for LU-5 and LU-6, in order to derive combined age esti- mates for individual lithostratigraphic units preserved at SH. The combined modeled ranges reveal that the hominin-bearing layer (LU-6) was deposited between 455 ± 17 ka and 440 ± 15 ka (mean lower and upper boundary 68.2% probability range ± 1s uncertainty, respectively), with a mean age of 448 ± 15 ka. These new bracketing ages suggest that the hominin fossils at SH were most likely deposited within Marine Isotope Stage (MIS) 12, enabling more precise temporal constraint on the early evolution of the Neandertal lineage. The SH fossils represent the oldest reliably dated hominin remains displaying Neandertal features across Eurasia. These Neandertal features are first observed in the facial skeleton, including the mandible and teeth, as well as the temporomandibular joint, and appear consistently across the SH collection. Our chronological findings suggest that the appearance of these Neandertal traits may have been associated with the climatic demise of MIS 12 and the ecological changes that occurred in Iberia during this period. Other Middle Pleistocene hominin fossils from Europe dated to MIS 12e11, or later, show different morphological trends, with some lacking Neandertal specializations. The latest SH dating results enable improved temporal correlations with these contrasting hominin records from Europe, and suggest a complex picture for hominin evolution during the Middle Pleistocene. © 2018 Elsevier Ltd. All rights reserved.

* Corresponding author. E-mail address: [email protected] (M. Demuro). https://doi.org/10.1016/j.jhevol.2018.12.003 0047-2484/© 2018 Elsevier Ltd. All rights reserved. M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 77

1. Introduction Detailed anatomical studies of the SH collection led Arsuaga et al. (2014) to conclude that there was a mosaic pattern of cra- 1.1. Site background and the need for an improved chronological nial evolution in Neandertals, with different anatomical and func- constraint tional modules evolving at different rates. This interpretation was based on the fact that the derived Neandertal traits are not uni- Located within the deep chambers of the Sierra de Atapuerca formly distributed in the cranial anatomy of the SH collection. karst system (northern Spain), the Sima de los Huesos (SH) site has Instead, Neandertal traits concentrate in a few regions of the cra- yielded thousands of hominin fossils (Arsuaga et al., 1997a) and nium and are consistently observed throughout the entire SH represents the largest individual Middle Pleistocene fossil assem- paleodeme, while other cranial regions lack Neandertal speciali- blage for the genus Homo globally. In total, more than 6900 fossils zations and are considered primitive. The most clear Neandertal relating to at least 28 individuals have been excavated from a single apomorphies in the SH collection are found in the facial skeleton, stratigraphic level (Bermúdez de Castro et al., 2004; Aranburu et al., including the mandible and teeth, as well as the temporomandib- 2017). Nuclear DNA sequencing and cranial morphological analyses ular joint (Arsuaga et al., 2014). The simultaneous occurrence of of these fossils have firmly placed the SH hominins at the beginning these apomorphies suggests that they are functionally related and of the Neandertal lineage (Arsuaga et al., 2014; Meyer et al., 2016) are part of the same adaptation, which probably reflects a masti- and have shown a mosaic pattern of evolution, with certain catory specialization in early Neandertals. A similar mosaic pattern Neandertal specializations (teeth, face, temporomandibular joint has been attributed to the evolution of modern humans based on and occipital bone) arising earlier than other Neandertal apomor- the Jebel Irhoud fossils, which have been radiometrically dated to phies. The SH fossils are also associated with an Acheulean han- 315 ± 34 ka (Hublin et al., 2017; Richter et al., 2017). Here again, daxe, making SH one of the few archaeological sites worldwide for modern human face, mandible and teeth specializations have which this iconic lithic industry can be directly related to a specific evolved earlier than neurocranium apomorphies, albeit in a hominin group (Carbonell and Mosquera, 2006; Daura et al., 2017). different direction to that of Neandertals. Thus, Neandertals and Establishing firm minimum and maximum ages for the SH fossils modern humans appear to have followed a similar evolutionary using numerical dating techniques is therefore important for un- mosaic pattern, one that could be described as ‘face first, brain derstanding the evolutionary and cultural history of these in- later’. Although the Jebel Irhoud fossils are younger than those at dividuals, and their temporal relationships with other Early and SH (Arsuaga et al., 2014; Richter et al., 2017), determining a precise Middle Pleistocene hominin records from Europe. bracketing age for the SH hominins is important not only to The completeness of the skeletal representation in the SH establish when the Neandertal clade branched out, but also to hominin assemblage, which preserves all anatomical components determine when the different morphofunctional units (modules) (Arsuaga et al., 1997a), is consistent with the interpretation that arose, and to appropriately compare the patterns and timing of entire corpses were deposited at the site, either as accidental falls evolution in Neandertals and modern humans. or as intentional anthropic accumulations. However, the origins Improved chronological constraints on the SH assemblage are and accumulation mechanisms of the SH hominins remain widely also critical for determining how past environmental change may debated, with proposed hypotheses ranging from catastrophic have influenced hominin biological and cultural evolution over events (Díez, 1990; Aguirre, 2000), the combined intervention of glacial-interglacial cycles, particularly those pertaining to the humans (collector agent), carnivores (transport agent) and mud Middle Pleistocene Neandertal lineage (e.g., Dennell et al., 2011; flows (reworking agent; Andrews and Fernandez-Jalvo,1997 ), other Herisson et al., 2016; Bermúdez de Castro et al., 2016; Roebroeks undetermined natural causes (Egeland et al., 2018), and the and Soressi, 2016). Orbitally-driven climate change has often intentional accumulation of corpses inside the chamber (Arsuaga been invoked as a major factor controlling Neandertal evolution in et al., 1993, 1997a; Arsuaga and Martínez, 2004). Several recent Europe via population expansions and crashes associated with geological, taphonomic and forensic studies have refuted a number glacial retractions and advances (Hublin and Roebroeks, 2009). of these proposed hypotheses. In particular, it has been shown that According to Hublin (2009:16025), “The fossil record suggests that the sedimentological features of the hominin-bearing unit (well- the cold episode(s) that triggered the process of divergence of the sorted, homogeneous muds devoid of extraclasts) do not support Neandertal clade most likely predates OIS 11”. During past glacial the hypothesis that the hominin bones were geologically trans- stages, icecap and permafrost advances in northern Europe are ported to SH from a distant locus of primary accumulation thought to have rendered large parts of the continent potentially (Aranburu et al., 2017). Recent taphonomic studies have also sug- uninhabitable and would have restricted hominin populations to gested that carnivores were not the primary accumulation agents the Mediterranean peninsulas (Dennell et al., 2011). This arrange- (Sala et al., 2014). Additionally, analysis of long bone fracture pat- ment would have likely led to the fragmentation of human groups terns has indicated that the vast majority of such breakages across a quasi-discontinuous habitable territory. The subsequent occurred after burial and were caused by overlying sediment retreat of glaciers during interglacial stages would have led to pressure, while analysis of cranial fractures does not support an population expansion and an increased chance for admixture accidental cause of hominin accumulation in the SH chamber (Sala among different human groups (e.g., Dennell et al., 2011; Steward et al., 2015a, b, 2016). Although it is difficult to determine the exact and Stringer, 2012). It is predicted that these repeated cycles of causes, duration (over human timescales) and nature (synchronous population expansion and contraction during the Middle Pleisto- versus repeated events) of hominin fossil accumulation at SH, it has cene lead to enhanced hominin diversity in Europe and a non- been possible to demonstrate that the hominin fossil deposit rep- anagenetic pattern of evolution (i.e., a pattern that was not resents a single accumulation episode on lithostratigraphic gradual and did not involve a single human population), since grounds (Aranburu et al., 2017). In this sense, the >28 individuals anagenetic evolution requires climatic stability over a continuous represented in the SH record have been considered to belong to the geographic region. Indeed, the emerging picture from the European same biological population, a paleodeme. The existence of such a Middle Pleistocene hominin fossil record suggests that a non- well-preserved and complete paleodeme at SH has provided anagenetic (likely cladogenetic) model is better suited to describe unique opportunities for studying Neandertal evolution based on a the pattern of hominin evolution on the continent. In particular, large collection of individuals from the same group, instead of hominin fossils that are potentially contemporaneous with, or even relying on a series of isolated specimens from disparate localities. younger than, those from SH, such as the Ceprano (~353 ka; 78 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95

Nomade et al., 2011; Manzi et al., 2011) or Arago (~450 ka, Falgueres supported by the relatively homogenous single-grain TT-OSL De et al., 2015) specimens, show less derived Neandertal conditions, distributions obtained for overlying SH deposits (Arnold et al., while the Aroeira cranium from Portugal, which has been dated to 2014). Similarly, there are no signs of re-transportation or post- 436e390 ka, shows a combination of Neandertal traits that is depositional alteration of the human fossils found at the base of the different to that observed at SH (Daura et al., 2017; Conde-Valverde SH chamber (Aranburu et al., 2017). These stratigraphical and et al., 2018). Clearly, a more refined timeline and further accurate taphonomic considerations indicate that luminescence dating of dating of Middle Pleistocene hominin fossils are necessary to reli- the red clay (LU-5 and LU-6) deposition event should provide un- ably unravel these complex phylogeographic patterns. ambiguous and tightly associated age constraint on the SH hominin fossils. 1.2. Improving the chronological framework of the SH hominin Previous luminescence dating attempts at SH have focused fossilsdApproach and aims exclusively on a debris flow deposit known as Cafe con Leche (LU- 7), which is found throughout the SH chamber and directly overlies Obtaining reliable and precise chronologies for Middle Pleisto- the hominin fossil-bearing red clays (Arnold et al., 2014; Arsuaga cene archaeological sites has traditionally proved difficult due to et al., 2014). This study aims to build on the existing minimum the limited number of radiometric dating techniques that are age estimates for the SH assemblage and to better constrain the applicable over this time-period (>125 ka) and because suitable timing for the onset of Neandertal speciation by establishing a more dating materials are not always readily available in stratigraphic precise bracketing chronology (i.e., a maximum age, as well as a association with target horizons. In this sense, the ubiquitous coeval age) for the hominin remains. To this end, we report on new presence of silicate minerals at most archaeological sites makes luminescence dating (single-grain TT-OSL) results obtained for the luminescence dating a viable option for indirectly dating biological sediment layers that encase and directly underlie the hominin ac- remains and cultural artifacts. Advances in luminescence dating cumulations (LU-6 and LU-5). We also derive combined age esti- over the last decade have also opened up new possibilities for mates for each depositional phase at SH by constructing a Bayesian obtaining finite and reliable age estimates on Middle Pleistocene model using our latest dating results and previously published sedimentary deposits (e.g., Duller and Wintle, 2012; Li et al., 2014; chronologies. In addressing important existing gaps in the SH Arnold et al., 2015, 2016), with at least two novel signals offering chronological record, we go on to discuss the climatic context of good potential for circumventing the limited upper age limit of early Neandertal evolution in Europe and examine the complex conventional optically stimulated luminescence (OSL); namely, evolutionary histories and phylogenetic relationships between SH thermally-transferred OSL (TT-OSL) signals from quartz grains and other European Middle Pleistocene hominin records. (Wang et al., 2006a, b) and post infrared (IR) stimulated lumines- cence (i.e., pIR-IRSL) signals from K-rich feldspar grains (Thomsen 2. Methods et al., 2008; Buylaert et al., 2012). Numerous paleoanthropolog- ical and archaeological sites in Europe have now been dated using 2.1. Study site, stratigraphy and existing chronology these methods (e.g., Rink et al., 2013; Frouin et al., 2014; Hernandez et al., 2014; Demuro et al., 2015; Olle et al., 2016; Arriolabengoa The Sima de los Huesos site is located in the Sierra de Atapuerca, et al., 2018). Furthermore, the reliability of these techniques have 14 km east from the city of Burgos, Spain, and comprises a small been successfully tested on several independently or semi- chamber situated at 985e990 masl in the lowermost sector of the independently dated stratigraphic sections from the Sierra de Cueva MayoreCueva del Silo multi-level karst system (Ortega et al., Atapuerca archaeological complex (Arnold et al., 2013, 2015; 2013; Fig. 1a). The site is divided into three interconnected sectors: Demuro et al., 2014, 2019), and confirm that the luminescence (i) Sima top section, a 1.9 m-long area located at the top of the properties of host deposits from this geological province are chamber and immediately below the 13 m vertical entry shaft; (ii) generally well-suited for single-grain TT-OSL and multi-grain pIR- Sima ramp, a linear section in the middle part of the chamber, with IRSL dating (e.g., Arnold et al., 2019). Systematic investigations of an inclination of 30; and (iii) Sima de los Huesos proper (SH), a single-grain TT-OSL properties have revealed that some sediment 32 m2 area located at the bottom of chamber. There are three layers at Atapuerca can contain subpopulations of quartz grains excavation areas along the ramp: one in the upper section called with unsuitable TT-OSL responses, including low thermal stabilities ‘Sima rampa alta’ (SRA), a second along the middle section called that can give rise to age underestimation if equivalent dose (De) ‘Sima rampa media’ (SRM), and a third excavation in the lower measurements are made on multi-grain aliquots (Arnold and ramp area called ‘Sima rampa baja’ (SRB). The fourth and lower- Demuro, 2015). To circumvent this problem, and to avoid other most excavation area corresponds to Sima de los Huesos proper types of averaging effects that may arise from simultaneously (SH) at the base of the ramp and is where the majority of the human measuring grains with different bleaching histories, signal com- fossils have been recovered (Fig. 2, inset map; Arsuaga et al., 1997a; positions or TT-OSL source trap properties, Arnold and Demuro Aranburu et al., 2017). (2015) recommended performing TT-OSL De measurements at the A revised stratigraphical framework for the site has recently individual grain scale where possible. The present study adopts been developed by Arsuaga et al. (2014) and Aranburu et al. (2017) these recommendations, and focuses on single-grain TT-OSL dating based on new sedimentological data and spatiotemporal correla- to establish new age constraint on the SH fossil deposits. tions of levels exposed at various locations within the chamber. The To reliably constrain the age of the SH hominins by lumines- new infill classification scheme, which is described in Aranburu cence dating requires firm knowledge of the stratigraphic re- et al. (2017:Table 1), is based on the concept of allostratigraphy, lationships of target sediments associated with, and surrounding, whereby lithostratigraphic units are grouped according to their the human fossils, as well as confidence in the overall stratigraphic bounding unconformities. Figure 1b summarizes the generalized integrity of the site (Rodríguez-Rey et al., 2015). Detailed prior stratigraphic framework, which consists, from the base upwards, of sedimentological analysis undertaken at SH suggests that deposi- 5 allostratigraphic units (AU) and 12 lithostratigraphic units (LU), as tion of the fossils and the encasing sediment (the red clays) follows: occurred coevally, and that any postdepositional movements within the chamber have been restricted to minor and localized AU1 A massive basal Miocene marl deposit (LU-1) capped by a events (Aranburu et al., 2017). The latter interpretations are further 10e15 cm-thick discontinuous flowstone speleothem (LU-2). M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 79

Figure 1. A) Location of the Sima de los Huesos sites within the Atapuerca Cueva Mayor-Cueva del Silo karst system. B) Generalized stratgraphic sequence showing published chronologies (*) in Arsuaga et al. (2014) and new luminescence ages on quartz (single-grain TT-OSL) obtained in this study for LU-5 and LU-6. Abbreviations: AU ¼ allostratigraphic units; LU ¼ lithostratigraphic units; CCLC ¼ Cafe con Leche clay.

AU2 A layer of sands and silts (LU-3) of up to 1.2 m in thickness, flowstone in SRA and SH (LU-10), and is subsequently which have been classed as an allogenic fluviokarstic sedi- covered by a layer of dark clays composed of guano (LU-11). ment deposit of Early Pleistocene age (Pares et al., 2000). The entire sedimentary sequence is capped by collapsed These sands are overlain by a relict speleothem (LU-4). limestone blocks, coming from the roof and walls (LU-12). AU3 A layer of homogeneous light brown clays or mud divided into the 'lower red clay' (LU-5), which is devoid of clasts and A comprehensive review of the existing chronology at SH is fossils, and the 'upper red clay' (LU-6), which contains presented in Arnold et al. (2014) and Arsuaga et al. (2014). The intraclasts, human and carnivore remains, and in situ spe- oldest reliable age for the site comes from a speleothem corre- leothem rafts. Sedimentological, geochemical and fabric an- sponding to LU-2, which has been U-series dated to 856 þ 110/-65 alyses of these red clays indicate an external source for both ka (n ¼ 5). This speleothem underlies magnetically reversed sands LU-5 and LU-6. It has been suggested that the source was (LU-3), indicating that the start of the SH depositional sequence likely the overlying surface soils, which were deposited occurred >780 ka, possibly during the Matuyama chron (Pares within the cavity as suspended load (slackwater facies) or by et al., 2000; Arsuaga et al., 2014). The red clays underlying and drip water processes (backswamp facies) in low energy encasing the hominin remains (LU-5 and LU-6) were deposited ponding environments within the karst cavities (White, during the Brunhes chron (Arnold et al., 2014) and were sourced 2007; Aranburu et al., 2017). externally (Aranburu et al., 2017). Though LU-5 and LU-6 remain AU4 Dominated by a debris flow deposit composed of dark undated, a calcite concretion (cave raft speleothem) deposited on a yellowish orange mud and clast-supported clay matrix (LU- human cranium has been U-series dated to 434 þ 36/-24 ka (n ¼ 4; 7). The debris flow was sourced externally, but clasts have Arsuaga et al., 2014). The Cafe con Leche clay breccia (LU-7) that been derived from both the interior (limestone speleothem overlays the red clays and hominin remains has been dated to fragments and previously accumulated carnivore bones) and 427 ± 12 ka (average of 6 ages) using TT-OSL and pIR-IRSL dating of the exterior of the cavity (Miocene limestone pebbles and silt-sized quartz (n ¼ 3) and potassium feldspar grains (n ¼ 3), conglomerate fragments). This deposit, which is referred to respectively (Arnold et al., 2014). An age of 443 ± 90 ka has also as the Cafe con Leche clay, is overlain by a thin (<3 cm) and been obtained for one sediment sample from LU-7 using ESR dating laminated flowstone in the SRA sector (LU-8). of optically bleached quartz grains (Arsuaga et al., 2014). However, AU5 The base of this unit is a sandy clay deposit only present in two additional ESR quartz samples taken from LU-7 yielded SRA and SRM (LU-9), which is similar in composition to LU-7 significantly older ages of 945e1502 ka, and have been interpreted but does not contain any macrofossils. This unit, which is as maximum age estimates owing to potential methodological considered to be reworked Cafe con Leche, is overlain by a complications (Arsuaga et al., 2014, Fig. 1b). Towards the top of the 80 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95

Figure 2. Luminescence dating sampling positions in LU-5 (a, b) and LU-6 (c, d) at the Sima de los Huesos site. Magnification in (a) corresponds to human fossil fragment AT-6778. Sample holes shown in photos above sample SH13-1 in (b) and sample SH13-5 in (c) correspond to luminescence dating sample SH12-1A (b) and samples SH12-2A (left) and SH12- 3A (right) (c) presented in Arnold et al. (2014) and Arsuaga et al. (2014). sequence, U-series dating of the LU-8 and LU-10 speleothems has strategy on the red clays of LU-5 and LU-6 in order to determine produced ages of 68 þ 1.5/-1.4 ka and 68 ± 6 ka, respectively (see tighter maximum- and coeval depositional ages for the SH homi- Arsuaga et al., 2014:Fig. S31). There are also three additional U- nin fossils. Arnold et al. (2014) originally speculated that these series (TIMS) ages of 153 ± 5 ka, 257 þ 16/-14 and 281 þ 28/-23 ka deposits could have experienced potentially complex infilling and for LU-8 (Bischoff et al., 2003). Finally, Bischoff et al. (1997) reported bleaching histories, and therefore opted to focus their initial numerous 14C ages spanning 25e20 ka for bones embedded in the luminescence dating study on the overlying Cafe con Leche clay LU-10 speleothem, as well as U-series (alpha spectrometry) ages of deposits (LU-7). Continued sedimentological evaluations of the 40e25 ka for the LU-10 speleothem itself. LU-5 and LU-6 deposits over subsequent field campaigns, together The U-series ages for the calcite concretions attached to the with complementary TT-OSL dating of related infill deposits in the human cranium and the luminescence dating results for LU-7 adjacent Sala de los Cíclopes chamber (Demuro et al., 2016,in provide a firm minimum age of ~430 ka for the SH hominin fos- prep.) and analogous fine-grained, well-sorted slackwater facies sils. At present the only reliable maximum age for the human re- preserved elsewhere in the Atapuerca karst system (e.g., Demuro mains is provided by the palaeomagnetic data, which indicates et al., 2014, 2019), have since revealed a clearer understanding of that the red clays encasing and underlying the fossils were the SH infilling history and suggest greater potential for deriving deposited during the Brunhes chron (<780 ka). In the present reliable luminescence chronologies for LU-5 and LU-6. That said, study we have therefore focused our luminescence sampling the initial cautionary interpretations of Arnold et al. (2014) M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 81

Table 1 Environmental dose rate values for the quartz fractions measured in this study. Also shown are the U, Th and K radionuclide activities calculated from the in situ field gamma spectrometry measurements. The specific activities are ‘as measured’ in the field and incorporate the attenuating effects of sediment moisture content and spatial hetero- geneity in gamma-ray radioactivity.

Sample Sample Grain Water U Th K Gamma Beta dose Cosmic Internal dose rate Total dose depth (m) fraction (mm) contenta (Bq/kg)b (Bq/kg)b (Bq/kg)b dose rateb ratec dose rated for quartz (U þ Th)e rate (Gy/ka)f

SH13-1 0.60 90e125 45.4 24 ± 242± 2 555 ± 29 1.14 ± 0.04 1.97 ± 0.16 0.01 ± 0.01 0.03 ± 0.01 3.15 ± 0.30 SH13-2 0.36 90e125 30.6 24 ± 234± 2 386 ± 20 0.91 ± 0.03 1.49 ± 0.09 0.01 ± 0.01 0.03 ± 0.01 2.43 ± 0.18 SH13-3 0.90 90e125 29.0 27 ± 233± 2 424 ± 23 0.95 ± 0.04 1.69 ± 0.04 0.01 ± 0.01 0.03 ± 0.01 2.68 ± 0.20 SH13-5 0.65 90e125 26.4 26 ± 222± 1 267 ± 15 0.68 ± 0.03 1.11 ± 0.07 0.01 ± 0.01 0.03 ± 0.01 1.83 ± 0.13

a 'As-measured' water content values, expressed as % of dry mass of sample and assigned a relative uncertainty of ±20%. The 'as-measured' water content values are considered to be representative of the long-term water content for these samples. b Specific activities and gamma dose rates were calculated from in situ measurements made at each sample position with a NaI:Tl detector using the ‘energy windows’ method (Arnold et al., 2012a). c Beta dose rates were calculated using a Risø GM-25-5 low-level beta counter (Bøtter-Jensen and Mejdahl, 1988), after making allowance for beta dose attenuation due to grain-size effects and HF etching (Brennan, 2003). d Cosmic-ray dose rates were calculated following published procedures (Prescott and Hutton, 1994) and assigned a relative uncertainty of ±10%. e Assumed internal (alpha plus beta) dose rate for the quartz fractions are based on published238U and232Th measurements for etched quartz grains from a range of locations (Mejdahl, 1987; Bowler et al., 2003; Jacobs et al., 2006; Pawley et al., 2008) and an alpha efficiency factor (a-value) of 0.04 ± 0.01 (Rees-Jones, 1995; Rees-Jones and Tite, 1997). f Mean ± total uncertainty (68% confidence interval), calculated as the quadratic sum of the random and systematic uncertainties.

highlighted the need to focus on single-grain De determination for measurements were made directly in the luminescence sampling these potentially complex cave deposits. holes to account for any gamma radiation heterogeneity within the immediate (30 cm radius) vicinity of each sample. U, Th and K 2.2. Samples and luminescence dating procedure radionuclide concentrations were determined from the gamma-ray spectra using the ‘windows method’ (Arnold et al., 2012a) and A total of four new luminescence samples were dated in this converted to gamma dose rates using published conversion factors study, as shown in Figure 2. All of these samples were collected (Guerin et al., 2011). Estimates of the cosmic-ray dose rate were from the AU3 red clays that either encased (LU-6) or immediately made using theoretical calculations after taking into consideration underlay (LU-5) the hominin fossils. Sample SH13-1 was obtained site altitude and geomagnetic latitude, as well as the density, from within LU-5 at SRB, the lowermost section of the Sima ramp thickness and geometry of sediment and bedrock overburden that has been excavated (Fig. 2b). Sample SH13-2 was collected (Prescott and Hutton, 1994). from LU-5 within the main excavation area of the Sima chamber High-resolution gamma-ray spectrometry (HRGS) measure- (Fig. 2a), while sample SH13-3 was taken from LU-6 deposits ments were undertaken to assess the state of secular equilibrium in 238 232 exposed in an adjacent sector of the SH excavation (Fig. 2d). the U and Th decay chains. Measurements were made using a Sample SH13-5 was collected from LU-6 deposits exposed along high-purity germanium (HPGe) detector on the same homogenised the ramp in the SRM excavation area (Fig. 2c). The four dating sediment samples used for beta counting. Supplementary Online samples were collected from different parts of the SH chamber in Material (SOM) Table S1 shows the daughter-parent isotopic ra- 238 226 210 228 228 order to (i) make use of the most suitable stratigraphic exposures tios for the U, Ra, Pb, Ra and Th activities of each 238 232 of LU-5 and LU-6 available at the time, (ii) avoid damaging human sample. These HRGS results indicate that the U and Th chains fossils that were actively being excavated in certain areas, (iii) of the SH samples dated in this study are in present-day secular enable any potential spatial variations in LU-5 and LU-6 deposi- equilibrium (daughter-parent ratios are all consistent with unity at tional histories to be evaluated across the Sima chamber, (iv) target either 1s or 2s). different zones of the chamber in case some of the deposits had experienced localized dosimetric complexities (e.g., secular 2.4. Sample preparation and instrumentation disequilibrium or variable moisture histories), and (v) ensure that the samples collected in the present study can be related to the Luminescence dating samples were collected by inserting opa- localities examined by Arnold et al. (2014). Luminescence ages que PVC tubes into previously cleaned sediment profiles. For each were obtained by undertaking TT-OSL measurements on individ- dating sample, additional bulk sediment was collected from the ual quartz grains, following the successful application of this sampling hole and placed in a zip-lock bag for moisture content ‘extended-range’ luminescence dating technique on the overlying analysis and beta dose rate estimation. Purified quartz grains LU-7 level at SH and on several independently dated archaeolog- (90e125 mm) were prepared for burial dose estimation using ical layers at other Atapuerca sites (Demuro et al., 2014; Arnold standard sample preparation procedures (Aitken, 1998), including a et al., 2014, 2015). Post-IR-IRSL dating was not attempted on 48% hydrofluoric (HF) acid etch for 40 min to remove the alpha- these samples owing to the very low K-feldspar yields of the LU-5 irradiated outer (10 mm) rinds of the quartz grains. This etching and LU-6 clays (see Aranburu et al., 2017:Table 1). stage was followed by a 30% HCl treatment for 45 min to eliminate any acid-soluble fluoride precipitates, and re-sieving (using a 2.3. Environmental dose rate estimation 63 mm sieve) to remove any disaggregated or partially etched grains. Luminescence measurements were made using a Risø TL- The environmental dose rate of each sample was determined DA-20 reader equipped with blue LEDs (470 ± 20 nm, maximum using a combination of low-level beta counting and field gamma power of 84 mW cm 2) and an array of infrared (IR) LEDs (875 nm, spectrometry measurements (Table 1). Beta dose rate measure- maximum power of 151 mW cm 2). A focused green laser (532 nm, ments were made on dried and homogenized sediment using a Risø emitting 10 mW to a 20 mm diameter spot) was used to stimulate GM-25-5 beta counter (Bøtter-Jensen and Mejdahl, 1988). Gamma individual grains. Irradiation was carried out using a 90Sr/90Y beta dose rate estimates were based on in situ gamma-ray spectrometry source mounted on the reader. For single-grain measurements, measurements carried out with a NaI:Tl spectrometer. In situ spatial variations in beta dose rates across the disc plane were taken 82 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 into account by undertaking hole-specific calibrations using Table 3 gamma-irradiated quartz. Ultraviolet quartz emissions were SAR protocol used in this study to obtain single-grain quartz TT-OSL ages. Ln and Lx refer to the natural and regenerative-dose signal measurements, respectively. T and detected using an EMI 9235QA photomultiplier tube and a 7.5 mm- n Tx refer to the test dose signals measured after the Ln and Lx signals, respectively. thick Hoya U-340 filter pack. Each SAR measurement cycle was repeated for the natural-dose, three different TT-OSL measurements have been made by loading 90e125 um sized regenerative doses, a 0 Gy regenerative-dose (to measure OSL signal recu- quartz grains into standard single-grain aluminium discs drilled peration) and a replicate of the first regenerative-dose cycle (to assess the suitability with a 10 10 array of 300 mm-deep depressions (holes). Although of the test-dose sensitivity correction). The OSL IR depletion ratio (Duller, 2003)was measured separately and used to check for the presence of feldspar contaminants. this configuration results in ~18 grains being placed in each grain- hole position (Arnold et al., 2012b), we are reasonably confident Step Single-grain TT-OSL SAR protocol Observed a that the pseudo single-grain De measurements employed here 1 Give dose closely approximate true single-grain resolution (or otherwise are 2 Preheat to 260 C for 10 s minimally affected by any potential multi-grain averaging effects) 3 Stimulate with green laser at 125 C for 3 s (90% power) 4 Preheat to 260 C for 10 s because of the low frequency of grain-hole positions that produce 5 Stimulate with green laser at 125 C for 3 s (90% power) TT-OSL

TT-OSL signals for these samples. This is evident from Table 2, Ln or Lx which shows that 75e80% of grain-hole positions did not produce 6 Stimulate with blue LEDs at 280 C for 400 s any statistically distinguishable TT-OSL T signal when measuring 7 Give test dose (200 Gy) n ~18 grains per hole for the four SH dating samples, and only 2% of 8 Preheat to 260 C for 10 s 9 Stimulate with green laser at 125 C for 3 s (90% power) grain holes resulted in signals from which meaningful De estimates 10 Preheat to 260 C for 10 s could be derived. Similarly low proportions of TT-OSL-producing 11 Stimulate with green laser at 125 C for 3 s (90% power) TT-OSL quartz grains have been reported for other Atapuerca infill de- Tn or Tx posits by Arnold et al. (2014) and Demuro et al. (2014). Demuro 12 Stimulate with blue LEDs at 290 C for 400 s 13 Return to 1 et al. (2013) have also shown that pseudo single-grain De mea- a surements are not likely to induce any significant grain-hole aver- Step omitted when measuring the natural signal (Ln). aging effects for samples characterised by such low yields (<20%) of luminescent grains. (e.g., Fig. 3c, d). The latter are generally well-represented by a single saturating exponential function for the SH samples.

2.5. De estimation Single-grain TT-OSL De estimates were not considered suitable for final age calculation if they exhibited one or more of the

TT-OSL equivalent dose (De) values were determined using the following properties: (i) the net intensity of the natural test dose single-aliquot regenerative-dose (SAR) protocol described in signal (Tn)was<3s above the late-light background signal (i.e., Arnold et al (2014; Table 3), which has been modified to enable grains with low luminescence sensitivity); (ii) the recycling ratio measurement of individual quartz grains. In this protocol, sensi- (i.e., sensitivity-corrected luminescence responses [Lx/Tx] for two s tivity changes affecting the natural (Ln) and regenerated (labora- identical regenerative doses) was not consistent with unity at 2 ; tory-irradiated) TT-OSL signals (Lx) are monitored and corrected for (iii) the recuperation ratio, calculated as the ratio of the sensitivity- using a TT-OSL test dose (Tn or Tx) signal. Net TT-OSL signals were corrected 0 Gy dose point (L0/Tx) to the sensitivity-corrected nat- calculated from the first 0.25 s of laser stimulation after subtracting ural (Ln/Tn), was >5% at 2s; (iv) the OSL IR depletion ratio (Duller, a late-light background derived from the last 0.25 s of stimulation. 2003), which was measured separately using two conventional s Individual De estimates were calculated by interpolating the single-grain OSL SAR cycles, was less than unity at 2 ; (v) the net Tn > sensitivity corrected natural (Ln/Tn) signal of each grain onto its signal had a relative error of 30%; (vi) the sensitivity-corrected corresponding sensitivity-corrected Lx/Tx dose-response curve natural signal (Ln/Tn) did not intercept the sensitivity-corrected

Table 2

The number and proportion of single-grain measurements that were rejected from the final De estimation after applying the various SAR TT-OSL quality assurance criteria.

Natural De values Dose recovery test Bleachability test Bleached grains Bleached þ dose grains Modern analogue

SH13-1 SH13-2 SH13-3 SH13-5 SH13-5 SH13-5 SH12-5A

No. of %of No. of %of No. of %of No. of %of No. of %of No. of %of No. of %of grains grains grains grains grains grains grains grains grains grains grains grains grains grains

Total measured grains 4300 100 4000 100 3600 100 1500 100 900 100 1000 100 300 100 SAR rejection criteria:

Tn < 3s background 3365 78 3010 75 2740 76 3101 78 738 82 804 80 201 67 Recycling ratio s 1at±2s 236 6 260 7 212 6 221 6 48 5 44 4 14 5

0GyLx/Tx >5% Ln/Tn 10 <117<112<19<11<12<100 OSL-IR depletion ratios <1at±2s 00 00 00 00 1 <10 000 Additional rejection criteria:

Net Tn signal had a relative 596 14 608 15 519 14 572 14 99 11 123 12 11 15 error of >30%

Non-intersecting grains (Ln/Tn > 3 <112<19<17<10 0 2 <100 dose response curve saturation)

Saturated grains (Ln/Tn z dose 2 <1002<14<10 0 0 0 0 0 response curve saturation) Anomalous dose response/ 8 <19<13<1000 0 0 0 0 0 unable to perform Monte Carlo fit Sum of rejected grains 4220 98 3917 98 3497 97 3914 98 887 99 975 97 259 86 Sum of accepted grains 80 2 83 2 103 3 86 2 13 1 25 3 41 14 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 83

Figure 3. A) Modified log transformed radial plot (Galbraith and Roberts, 2012) showing the single-grain TT-OSL residual De distribution for modern analogue sample SH12-5A. The De distribution has been plotted using a modified log transformation of z ¼ log(De þ a) (Galbraith and Roberts, 2012), to more easily accommodate both the large and small (negative and near zero Gy) De values observed in this dataset. The standard errors of this modified log transformed dataset are given relative to De þ a, where a ¼ 40 Gy. The shaded band on the radial plot is centred on the TT-OSL CAMUL De values. B) Single-grain TT-OSL signal brightness plot for the samples collected from LU-5 and LU-6 in this study. The results of sample SH12-3A collected from LU-7 and published previously in Arnold et al. (2014) is shown for comparison. C, D) Examples of sensitivity-corrected dose-response curves and TT-OSL decay curves for two typical quartz grains that passed the SAR rejection criteria in this study.

dose-response curve; (vii) the Ln/Tn value intercepted the saturated 2.6. Bayesian modeling part of the dose-response curve (Ln/Tn values were equal to the Imax saturation limit of the dose-response curve at 2s); and (viii) the In order to integrate all of the stratigraphically reliable chro- dose-response curve displayed anomalous properties and/or poor nological information within a unified statistical framework and Monte Carlo fits (i.e., zero or negative response with increasing derive combined age estimates for individual stratigraphic units, dose) or very scattered Lx/Tx values that could not be successfully we have constructed a Bayesian age-depth model for SH using fitted with the Monte Carlo procedure and, hence, did not yield OxCal v4.2.4 (Bronk Ramsey, 2009). This Bayesian modeling finite De values and uncertainty ranges. approach incorporates the latest numerical dating results (likeli- Individual De estimates are presented with their 1s error ranges, hoods) presented in the current study and in Arsuaga et al. (2014), which are derived from three sources of uncertainty: (i) a random as well as relative stratigraphic information (priors) previously uncertainty term arising from photon counting statistics for each recorded for the sedimentary sequence (Arnold et al., 2014; TT-OSL measurement, calculated using Eq. 3 of Galbraith (2002); Arsuaga et al., 2014; Aranburu et al., 2017), to derive combined (ii) an empirically determined instrument reproducibility uncer- (posterior) chronological datasets for individual lithostratigraphic tainty of 1.6% for each single-grain measurement (calculated for the units. The combined likelihood dataset for SH comprises 14 age specific Risø reader used in this study following the approach estimates spanning lithostratigraphic units LU-2 to LU-11, and it has outlined in Jacobs et al., 2006); and (iii) a dose-response curve been derived using multiple dating techniques (extended-range fitting uncertainty determined using 1000 iterations of the Monte luminescence dating, ESR, U-series and palaeomagnetism), modern Carlo method described by Duller (2007) and implemented in analytical approaches (e.g., single-grain TT-OSL, pIR-IRSL) and Analyst v. 3.24 (Aberystwyth). different preserved materials (quartz and K-feldspar grains, calcite 84 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95

rafts, flowstones and bulk clays). Full details of the Bayesian ), g , f 56 41 modeling approach are provided in SOM S1. 39 31 ± ± ± ± FMM 453 457 3. Results age (ka) SOM Table S3 f 104109 214 7460 199

3.1. SAR suitability tests ; see 1 ± ± ± ± (Gy) e i FMM , e D To test the appropriateness of the De determination methods used in this study (i.e., grain size choice, SAR procedure, data analysis and rejection criteria), we undertook a single-grain dose 1212 673 1430 77 532 1224

recovery test on SH13-5 using the same procedures outlined in the ± ± ± ± FMM results e m previous section. For this purpose, a subset of prepared 90 125 m 28 72 19 81 Proportion quartz grains from sample SH13-5 was set aside and exposed to of grains (%) sunlight for 7 days to optically reset a significant proportion of the

natural TT-OSL signal. A known laboratory dose (900 Gy) of similar ) n nal ages. fi magnitude to the expected D was added on top of the remaining (k e FMM residual natural signal for a portion of these bleached grains 1 2 1 2 k k components (n ¼ 500), and the TT-OSL signals were subsequently measured h , g , using the SAR procedure shown in Table 3. A second batch of f 38 39 35 k 41 k ± ± ± ¼ ± bleached grains (n 600) was measured without any prior dosing components. The FMM parameter values shown here were obtained CAM n

to determine the residual dose remaining after sunlight exposure. 437 460 The central age model (CAM) was used to calculate the weighted age (ka) e e 50 386 42 38 mean De values for each group of grains, and the recovered dose 37 364 f ± ± ± was then calculated by subtracting the residual De of the undosed ± (Gy) ed number of k grains from the measured De of the bleached and dosed grains. CAM D fi

The results of the single-grain dose-recovery test are shown in cantly skewed if the weighted skewness value is greater than the corresponding critical fi CAM results SOM Figure S1 and Table S2. A weighted mean De of 1262 ± 103 Gy ± 4 1033 4 1063 5 842 5 1148

and an overdispersion of 27 8% were obtained for the subset of . ± ± ± ±

grains that had been bleached and given a dose of 900 Gy, while the (%) value (see main text for details). undosed grains produced a weighted mean residual De of e distribution of SH13-5 and SH13-2 were shown to contain a single dose component (k 297 ± 40 Gy and an overdispersion of 27 ± 15%. Approximately 35% Overdis-persion e of the weighted mean natural dose of SH13-5 remained after 1 d week of exposure (compare natural De value of 842 Gy in Table 4), ). 0.4827 37 0.5377 20 0.5252 27 which is consistent with TT-OSL daylight bleaching rates reported 0.5477 41 Critical ± ± ± ± (95% CI) elsewhere for quartz samples from the Iberian Peninsula (Demuro skewness c et al., 2015; Duval et al., 2017). A dose recovery ratio of 1.07 ± 0.09 Arnold and Roberts (2009) was calculated by subtracting the weighed mean residual D from

e 8of distributions are considered to be signi e 0.4875 0.4427 0.3884 the De obtained for the bleached and dosed grains. These results e Weighted skewness provide assurance that, under controlled laboratory conditions at .D ). Using this approach, the D ). b least, the SAR procedure used in this study is suitable for single- 2 Galbraith and Green, 1990 grain TT-OSL De estimation. and k values 1 e ) between 5 and 25% and incrementally increasing the speci measured Accepted/ D k values using Eq. 7

3.2. TT-OSL signal bleachability s e a

TT-OSL signals are known to be optically reset at a slower rate nal ages for the Sima de los Huesos samples collected from LU-5 and LU-6. Values in bold represent fi

than conventional OSL signals (e.g., Jacobs et al., 2011; Demuro Bailey and Arnold (2006) type and Arnold and Roberts, 2009 nite mixture model ( e et al., 2015), meaning that it is important to consider site-specific fi resolution D ¼ 2% associated with laboratory beta-source calibration.

bleaching assessments as part of TT-OSL dating studies. Encourag- ± ingly, single-grain TT-OSL bleaching assessments undertaken by ); FMM 0.20 SG (18 grains) 103/3600 0.18 SG (18 grains) 83/40000.13 SG (18 0.0777 grains) 86/4000 Arnold et al. (2019) on 13 modern analogue samples from different 0.30 SG (18 grains) 80/4300 ± ± ± sedimentary environments, including surface samples collected ± Total dose adjacent to several Atapuerca cave site entrances, have shown that rate (Gy/ka) residual TT-OSL signals can be naturally reduced down to insig- dence interval), calculated as the quadratic sum of the random and systematic uncertainties. values, overdispersion and fi m) nificant levels (0e24 Gy) when compared to the natural dose range e m 125 2.68 125 2.43 125 1.83 125 3.15 t with the lowest BIC score; e e e e of interest for most TT-OSL dating applications. As part of the SH fi Grain size (

LU-7 single-grain TT-OSL dating study, Arnold et al. (2014) collected Galbraith et al., 1999 a modern sediment sample from the hillslope located directly

t (i.e., the outside the entrance to Cueva Mayor (El Portalon). This sample fi

(SH12-5A) was considered to be a useful analogue for the single-grain measurements containing ~18 grains per grain-hole position. tted by varying the common overdispersion parameter ( ¼ depth (m) measurements that passed the SAR rejection criteria/total number of grains analyzed. fi

geomorphic and transportation history of the clay breccia sedi- e ments accumulated in SH, as the allochthonous cave infill deposits total uncertainty (68% con central age model (

originate from surface soils immediately surrounding the cave ± ¼ entrance. It was considered unlikely that there would have been sufficient daylight exposure to reset the previously accumulated SG (18 grains) Number of D Weighted skewness scores have beenCritical calculated skewness on scores log-transformed have D been calculated using Eq. 16 of CAM Mean Total uncertainty includes a systematicFinal component single-grain of TT-OSL ages ofThe samples FMM was SH13-2 and SH13-5 (shown in bold) have been calculated using the CAM D i f c SH13-3 6 0.90 90 Sima proper (SH) SH13-2 5'Sima rampa media' (SRM) 0.35SH13-5 6 90 0.65 90 Sample LU Sample 'Sima rampa baja' (SRB) SH13-1 5 0.60 90 a e g b d fl h skewness value. whereas samples SH13-1 and SH13-3 contain two discrete dose populations (k Table 4 Summary of the single-grain TT-OSL D TT-OSL signals of all sediment grains during the mud ow from the optimum FMM M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 85 transportation events themselves. However, the eolian source procedures being employed in the present study. The resulting sediments from which these mudflow deposits were derived are single-grain TT-OSL residual De values for sample SH12-5A are thought to have received prolonged and direct sunlight exposure shown in Figure 3a. In total 300 quartz grains were measured, and prior to entrainment and transportation into the cave, as shown by 41 (14%) of these passed the SAR rejection criteria (Table 2). A Berger et al. (2008). Providing that the source sediments had weighted mean (unlogged central age model, CAMUL; Arnold et al., accumulated relatively recently before being washed into the cave 2009)De value of 2.3 ± 1.3 Gy was calculated for the accepted grain (i.e., compared with the expected Middle Pleistocene ages of the SH population. The individual De values vary widely (from negative fossil deposits), it was argued that the SH sediments could have values that overlap with 0 Gy at 2s,uptoDe values of 117 Gy), but been fully bleached at deposition, or else retained only very minor the majority of grains in this modern analogue sample show residual luminescence signals. This interpretation was borne out by adequate TT-OSL signal bleaching; 93% of accepted grains (n ¼ 38) the multi-grain (1100-grain aliquot) TT-OSL residual De value of produced residual doses that were within 2s of zero, albeit with 7.3 ± 1.2 Gy obtained for modern surface sediment sample SH12-5A large uncertainties (typically >40%), while only 2% (n ¼ 1) produced collected from outside the cave mouth (Arnold et al., 2014:Table 5). aDe value > 100 Gy. Assuming that the original sediment sources This residual dose would have translated to an age offset of 3e4ka (i.e., surface eolian and slopewash deposits) and past depositional for the Sima dating samples (assuming that the sediment being processes (i.e., mudflow and alluvial entry into the cavity) were dated had entered the cave via similar processes in the past), and similar to those observed at present, these results support the was considered of minor significance because it was well within the notion that partial bleaching of the SH source sediments prior to 1s uncertainties of the final TT-OSL ages for LU-7. their transportation into the cave system is unlikely to have been a Following on from the original multi-grain TT-OSL assessment of significant problem; particularly as the expected De values of these SH12-5A, we have additionally assessed the single-grain TT-OSL Middle Pleistocene deposits is on the order of 1000 Gy. resetting properties of the SH modern analogue sample using the It is worth noting that analysis of modern analogue sediments 90e125 mm quartz fraction and the single-grain De determination from the cave exterior surface does not necessarily provide direct

Table 5 Summary of Bayesian modeling results for Sima de los Huesos. The likelihood (unmodeled) and posterior (modeled) age ranges are presented for each of the numerical dating samples. Posterior (modeled) age ranges are also shown for the boundaries of each stratigraphic unit. Posterior ages are presented as the 68.2% and 95.4% highest probability density ranges. The mean and 1s uncertainty ranges of the modeled posterior distributions are shown for comparison (assuming a normally distributed probability density function). The unmodeled and modeled age estimates have been rounded to the nearest 50 years.

Boundary Dating Unmodeled age (years)a,b Modeled age (years) Agreement Posterior sample index (A ) outlier 68.2% range 95.4% range Mean ± 1s 68.2% range 95.4% range Mean ± 1s i (%) probability (%)

LU-11 top 0e12450 0e29600 10300 ± 9200 LU11 bottom 5650e30450 1850e50600 23200 ± 13200 LU-10 top 15300e57500 11300e61200 33550 ± 14100 CPV 61900e74100 56000e80000 68000 ± 6000 21850e61200 18100e63350 39050 ± 13650 15.6 74 isochron LU-10 bottom 27000e63150 21700e65300 43100 ± 12950 LU-9 top 42400e65400 29000e67700 49850 ± 11200 LU-9 bottom 53600e68100 39600e69700 57450 ± 8500 LU-8 top 62850e69600 52350e71250 64200 ± 5250 LU-8 1a-3 66500e69500 65100e70900 68000 ± 1450 66900e69800 65450e71250 68350 ± 1500 102.9 0 LU-8 bottom 65750e92400 64300e378400 108700 ± 70250 LU-7 top 386400e425100 347050e439800 398400 ± 27350 SH12-3A 396000e446400 371750e470650 421200 ± 24700 393350e425250 372850e439100 407250 ± 16400 108.4 6 SH12-3B 351200e534800 263000e623000 443000 ± 90000 395750e430850 381350e442550 412450 ± 14700 131.7 5 SH12-2A 398500e451650 372950e477150 425050 ± 26050 404750e431050 390150e444100 417450 ± 13250 123.9 4 SH12-1A 412200e461750 388400e485550 436950 ± 24300 408800e435750 394950e449100 422100 ± 13350 108.2 5 LU-7 bottom 440300e411700 398400e454800 426200 ± 14300 LU-6 top 424700e454550 411100e470300 440450 ± 14950 SH94-4 403400e464600 374000e494000 434000 ± 30000 431450e461450 417650e477650 447250 ± 14850 120.6 2 SH13-5 419950e499350 381800e537450 459650 ± 38900 431700e462450 417300e479600 448000 ± 15200 127.2 4 SH13-3 414850e499050 374450e539450 456950 ± 41250 431650e462200 419650e479100 447900 ± 15250 130.6 5 LU-6 bottom 437600e470050 423100e488900 455200 ± 16700 LU-5 top 449950e488950 433900e511950 471750 ± 19800 SH13-2 398200e475650 360950e512850 436900 ± 38000 459200e501350 442100e527150 483350 ± 21200 74.1 2 SH13-1 396850e510800 342100e565550 453811 ± 55850 459300e505100 440900e534750 485750 ± 23650 116.2 4 SH(T16) <777150e778850 <776300e779700 <778000 ± 850 <777150e778850 <776300e779700 <778000 ± 850 100 LU-5 bottom 464650e525350 444350e585450 506300 ± 38450 LU-4 top 495200e639050 467200e751900 594400 ± 78150 LU-4 bottom 589300e776500 516000e865350 690500 ± 92000 LU-3 top 696700e864050 599900e950350 778950 ± 86700 SH Area A >777150e778850 >776300e779700 >778000 ± 850 >777150e778850 >776300e779700 >778000 ± 850 100 LU-3 bottom 778300e887150 777150e986850 864950 ± 63750 LU-2 top 834900e972850 791750e1048300 918700 ± 70050 SRA-3 766750e945250 681000e1031000 856000 ± 87500 884900e1024900 828900e1101000 962350 ± 68900 73.2 3 C1-K10 LU-2 bottom 922600e1500000 907950e1500000 1164900 ± 184650

a The original dating results are all expressed in years before AD2012 (average sample collection date) for modeling purposes to ensure the likelihoods are directly comparable. b The combined (unmodeled) ages for SH12-1A, SH12-2A and SH12-3A have each been calculated from their replicate TT-OSL and pIR-IRSL ages using the OxCal ‘Combine’ function. See SOM S1 text for further details. 86 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 constraint on any additional sediment transportation and deposi- 3.3. Luminescence signal characteristics tion complexities that may have taken place within the endokarst chamber. For example, it is possible that transportation of the In total, 4000 'pseudo' single-grain measurements (~18 grains predominantly well-bleached, externally derived sediments per hole) were made on each sample. Approximately 2e3% of these through a closed cave system may have simultaneously resulted in measured grains passed the SAR rejection criteria and were the entrainment of some grains from pre-existing cave sediments, accepted for final age estimate (Table 2). The majority of grains and their subsequent translocation to the SH cavity along with the were rejected due to poor TT-OSL signal sensitivity (~90% of grains; more recently bleached grain populations. That said, we expect that criteria (i) and (v)) and failure of the recycling ratio test (~5% of the single-grain De analysis employed in this study would enable grains). Representative decay and dose-response curves for two sample-by-sample detection of any such minor populations of re- accepted grains are shown in Figure 3c, d. The TT-OSL signals entrained, older grains in otherwise well-bleached cave deposits, typically show fast decay rates (decaying down to background as demonstrated by Arnold et al. (2019). In rare cases where sedi- levels within 0.5 s of stimulation) consistent with charge transfer ment retransportation occurs within a closed cave system without into the most readily bleached (so-called ‘fast’) OSL dating trap, and the entrainment of pre-existing (older) grain populations (e.g., the dose-response curves exhibit high saturation dose limits previously accumulated chimney plugs that are subsequently dis- (characteristic saturation dose or D0 values are typically >1000 Gy). lodged by gravity and fall into new depositional positions as intact However, the TT-OSL signal intensities of the LU-5 and LU-6 red clay lumps), it may not be possible to detect autochthonous cave samples are uniformly low; especially when compared to those reworking complexities in empirical De datasets. However, such obtained for the LU-7 (overlying unit) samples by Arnold et al. methodological complications should be readily discernible by (2014), for which the signals were an order of magnitude brighter comparing the resultant luminescence ages with those of strati- (Fig. 3b). The proportion of TT-OSL producing grains is also lower graphically related (bracketing and replicate) luminescence sam- for the red clay samples. On average, 23% of measured grain-hole ples or with independent age control (e.g., sample SH12-4A of positions produced a statistically detectable TT-OSL signals (i.e., Arnold et al., 2014). Both of these strategies (i.e., analyses of net Tn TT-OSL intensities >3s of background), while just 15e20% of single-grain De distribution and comparisons with stratigraphically grain-hole positions accounted for 95% of the summed TT-OSL related dating samples) are employed in the present study to detect signal. These distinctive luminescence properties likely reflect dif- any potential internal reworking complexities with the LU-5 and ferences in the original source geology of the LU-7 versus LU-5 and LU-6 deposits. LU-6 deposits, which is consistent with the sedimentological and Geochemical analyses performed on the SH red clays indicate geochemical interpretations of Aranburu et al. (2017). an external (allochthonous) source for the sediments being dated in this study (Arsuaga et al., 2014). Aranburu et al. (2017) originally 3.4. De distributions proposed that the red clay accumulation within the SH chamber was potentially the result of topsoil being transported through The single-grain TT-OSL De distributions of SH13-5, from LU-6 at limestone cracks and fissures by dripping water, though deposi- SRM, and sample SH13-2, from LU-5 at SH proper, display limited tion as low-energy suspended load could not be discounted. scatter and contain a single dose population, with the vast majority However, recent investigations into the source material for the red of individual De values being well represented by the weighted De clays have focused on a debris flow deposit currently blocking a value (falling within the shaded band in the radial plots; Fig. 4b, d). paleoentrance in Sala de los Cíclopes, the large chamber found Neither of these samples are considered to be significantly posi- immediately adjacent to SH (Fig. 1; see Arsuaga et al., 1997a:Figs. 3 tively skewed according to the criterion outlined by Arnold and and 5). This debris flow is composed of red/orange clay supporting Roberts (2009; Table 4). A low overdispersion values of 20 ± 4% small lacustrine limestone gravels (mud supported breccia), which was also obtained for sample SH13-2, which is in agreement at 2s originate from the cave exterior and extend towards the interior of with the value obtained in the single-grain TT-OSL dose-recovery the cavity as a colluvial deposit through a low gradient opening test for SH13-5, as well as with average overdispersion values (Arsuaga et al.,1997a). Although large portions of this deposit were that have been reported for well-bleached single-grain TT-OSL and originally removed during an erosive episode associated with the OSL samples elsewhere (e.g., 21 ± 2% and 20 ± 1%, respectively; emptying of a large section of the Sala de los Cíclopes chamber, a Arnold and Roberts, 2009; Arnold et al., 2019). Sample SH13-5 sequence of this relict material is preserved 20e30 m from the yielded a low-to-moderate overdispersion of 27 ± 5%, which is opening of the vertical shaft that leads to SH. Consequently, it is consistent with that obtained for sample SH13-2 at 1s. These De now thought possible that a large portion of the red clays depos- distribution characteristics suggest that samples SH13-2 and SH13- ited in SH were sourced during this debris flow accumulation 5 were not significantly affected by partial bleaching or any major event and were transported further into the chamber as sus- postdepositional complications (e.g., sediment mixing or beta dose pended load slackwater deposits. The currently absent talus cone rate heterogeneity). Consequently, representative single-grain TT- of this debris flow deposit in Sala de los Cíclopes, which was OSL burial dose estimates have been calculated using the central removed by the erosive episode, could also have provided past age model (CAM) of Galbraith et al. (1999). human populations with direct access to the SH chamber and the The remaining two samples, which were collected from the LU- main vertical opening chimney C2 (see details in Aranburu et al., 6 red clays in the Sima proper (SH13-3) and the LU-5 red clays at 2017) from the cave exterior. Under this scenario, in which the SRB (SH13-1), show more scattered De distributions and a more red clays were primarily deposited as an extension of the debris noticeable proportion of the individual De values fall outside the 2s flow deposit located in Sala de los Cíclopes (although not excluding standardized estimate of the CAM burial dose (Fig. 4a, c). Both De some input from water dripping through fissures), and the distributions are significantly negatively skewed according to the modern-analogue surface source deposits yield a low residual dose criteria outlined by Arnold and Roberts (2009), and both samples (2.3 ± 1.3 Gy), it seems feasible that there could have been suffi- exhibit moderate to high overdispersion values (37 ± 4% and cient potential for TT-OSL signal resetting of the LU-5 and LU-6 41 ± 5%) that do not overlap at 2s with that of the seemingly well- sediments shortly before their final deposition within SH (at bleached and unmixed sample from LU-5 (SH13-2; Table 4). least considering the 2s uncertainties of our TT-OSL dating Application of the finite mixture model (FMM; Galbraith and Green, results). 1990) reveals the presence of two discrete dose populations in both M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 87

Figure 4. Radial plots showing the single-grain TT-OSL De distributions obtained for the LU-5 and LU-6 samples. The shaded bar in plots (B) and (D) are centred on the central age model (CAM) De used to derive the final age estimate. The shaded bar in plots (A) and (C) are centred on the finite mixture model (FMM) De population containing the largest proportion of grains, which has been used to derive a final age. datasets (Table 4; SOM Table S3). The dominant FMM components could have occurred in parts of the SH chamber. However, field (i.e., those containing the highest proportion of individual De observations do not show evidence of mixing between sediment values; 81% of grains for SH13-3 and 72% of grains for SH13-1) yield layers (e.g., human bones are confined to LU-6) and the lithos- ages in stratigraphic agreement with the published luminescence tratigraphic unit boundary remains clearly preserved above the age of 427 ± 12 ka for the overlying LU-7 deposits and the U-series SH13-1 sample position at SRB (Fig. 2b). The low dose FMM com- age of 434 þ 36/-24 ka on the hominin skull concretions (Arsuaga ponents of these two samples could relate to grains that were et al., 2014; Fig. 1b). However, the minor dose FMM components sitting adjacent to marls, carbonates or bones, and hence received (containing 19e28% of grains) yield significantly lower De values of lower-than-average dose rates in comparison to grains that were 530e670 Gy and underestimate the published minimum ages for surrounded by purer clays. Beta dose heterogeneity may have LU-6 and LU-5 by >50%. played a minor role in sample SH13-3, which was collected from a The younger-than-expected ages obtained using the minor dose section of SH containing a high-density of bones (Fig. 2d), though it FMM components of samples SH13-1 and SH13-3 indicate that seems unlikely that this type of beta dose heterogeneity could have partial bleaching is unlikely to explain the higher overdispersion given rise to the extreme and discrete low dose components seen and inter-grain scattered observed for these De datasets. Given the for sample SH13-3. This source of extrinsic scatter is also unlikely to generally well-stratified nature of the LU-5 and LU-6 deposits, it be relevant for sample SH13-1 from LU-5, which was taken from a also seems unlikely that postdepositional mixing could explain the very pure clay horizon that was devoid of clasts and bones (Fig. 2b). multimodal De distributions of samples SH13-1 and SH13-3. Minor Comparisons with previous Atapuerca TT-OSL studies suggest and localized mixing of younger grains within a given layer by that intrinsic sources of De scatter may partly or wholly explain the plastic deformation of clays (shrink/swelling) or trampling by bears complex single-grain TT-OSL datasets of samples SH13-1 and SH13- 88 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95

3. Similar low dose FMM components have been observed for a small number of Atapuerca samples collected from Gran Dolina (unit TD6; Arnold et al., 2015), and also potentially from Galería (samples AT10-2 and ATG10-7). For these published datasets, it was possible to discount beta microdosimetry and sediment mixing as explaining the low De scatter since the samples were collected from homogeneously fine-grained and well-bedded layers (Demuro et al., 2014). Subsequent TT-OSL characterization experiments un- dertaken on the Gran Dolina TD6 samples (Arnold and Demuro, 2015) revealed that the low dose FMM components were attrib- utable to inter-grain variations in TT-OSL signal characteristics; specifically, the presence of grain subpopulations that displayed unsuitably low TT-OSL thermal stabilities. Such unfavorable TT-OSL properties may relate to the type of source traps associated with TT- OSL signal in different grain populations, which may in turn reflect variability in geological provenance and formation histories be- tween quartz grains in a given sample (e.g., differences in thermal history, type and density of point defects/impurities, extent of sedimentary recycling, and nature of weathering processes). The single-grain TT-OSL characterization studies undertaken so far also suggest that the presence of a 280 C TL trap and/or slowly bleaching OSL signals may adversely affect TT-OSL dating suitability for some grains/samples (e.g., Arnold and Demuro, 2015; Bartz et al., 2019). In the absence of any clear extrinsic explanations for the low De scatter observed in the two SH samples, we consider that the low dose FMM components may be similarly related to grains that are poorly suited to the SAR protocol or grains that exhibit thermally unstable TT-OSL signals; though we acknowledge that further characterization studies would be needed to defini- tively ascertain the exact cause of this intrinsic De scatter. To derive representative burial doses (and ages) for samples SH13-1 and SH13-3 we have opted to use the dominant FMM component, which contains >70% of the measured De values. The suitability of our age model choice is supported by the consistency of the resultant ages for the four samples from LU-5 and LU-6. In contrast, use of the CAM De for samples SH13-1 and SH13-3 would yield stratigraphically inconsistent ages when compared with the ages obtained for samples SH13-2 and SH13-5 (Table 4).

3.5. Ages

The resulting single-grain TT-OSL luminescence ages for sam- ples SH13-1, SH13-2, SH13-3 and SH13-5 are 453 ± 56 ka, 437 ± 38 ka, 457 ± 41 ka and 460 ± 39 ka, respectively (Table 4). The cor- responding ages for LU-6 and LU-5 (upper and lower red clays) are statistically indistinguishable at 2s and are in close agreement with the published luminescence and U-series age of ~430 ka age for sedimentary material overlying the fossils (Arsuaga et al., 2014). The integration of the existing minimum age and the new coeval/ maximum ages most likely places the accumulation of the SH hominins within marine isotope stage (MIS) 12 (478e424 ka; Lisiecki and Raymo, 2005).

3.6. Refining the chronology framework of Sima de los Huesos using Bayesian modeling

The Bayesian modeling results are summarized in Table 5 and Figure 5. Bayesian modeling results for the Sima de los Huesos lithostratigraphic Figure 5. All modeled age ranges have been rounded to the nearest sequence. The prior age distributions for the dating samples (likelihoods) are shown as light blue probability density functions (PDFs). The modeled posterior distributions for 50 years and are reported as the 68.2% and 95.4% highest proba- the dating samples and stratigraphic unit boundaries are shown as dark blue and gray bility density function (PDF) ranges, and the mean and 1s uncer- PDFs, respectively. Unmodeled and modeled ages are shown on a calendar year tainty ranges of the modeled posterior distributions. The Bayesian timescale and both are expressed in years before AD2012 (average sample collection date). The white circles and associated error bars represent the mean ages and 1s uncertainty ranges of the PDFs. The 68.2% and 95.4% ranges of the highest posterior interpretation of the references to color in this figure legend, the reader is referred to probabilities are indicated by the horizontal bars underneath the PDFs. (For the Web version of this article.) M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 89 modeled posterior age ranges were used to constrain the beginning hominin fossil accumulation and represent a high-energy debris and end periods of each depositional event at Sima de los Huesos, flow event, were deposited between 398.4 ± 27.4 ka and as well as the likely age and duration of each lithostratigraphic unit, 426.2 ± 14.3 ka, and have a mean age of 412.3 ± 20.0 ka (Tables 5 and potential occurrences of statistically significant hiatuses be- and 6; Fig. 6). The modeled duration range of the CafeconLeche tween sedimentary deposits. These chronological relationships are deposits overlaps with that of the hominin-bearing clays at the summarized in Figure 6 and Table 6, and have been calculated from 95.4% credible interval, suggesting that the preserved erosional the posterior probabilities of the upper and lower (top and bottom) boundary between LU-6 and LU-7 spanned a relatively short time boundaries of each lithostratigraphic unit using the ‘difference’ and period (14.3 ± 12.0 ka) that cannot be fully resolved beyond the ‘date’ query functions in OxCal (see Table 6 footnote for further modeled likelihood uncertainty ranges. The only statistically sig- details). nificant temporal hiatus identified in the depositional sequence The results of the Bayesian analyses reveal that the dated ho- took place between the accumulation of the CafeconLechede- rizons from LU-6, which contain the hominin fossil remains, were posits (LU-7) and the overlying flowstone layer (LU-8). The ‘dif- deposited between 455.2 ± 16.7 ka and 440.5 ± 15.0 ka (mean ference’ query indicates a mean temporal gap of 289.8 ± 74.3 ka lower and upper boundary 68.2% probability range ± 1s uncer- between these two units, which may reflect a prolonged period of tainty), and have a mean age of 447.9 ± 15.5 ka (Tables 5 and 6; non-deposition and/or a statistically significant amount of un- Fig. 6). The underlying sterile red clays of LU-5, which are identical dated (either non-sampled or missing) material in this part of the in mineralogy and texture to LU-6, have upper and lower profile. boundary ages of 471.7 ± 19.8 ka and 506.3 ± 38.4 ka, and a mean ± ‘ ’ depositional age of 489.0 28.6 ka. Application of the difference 4. Discussion query to the posterior probabilities of the LU-5 and LU-6 bound- fi fi aries con rms the absence of any statistically signi cant temporal 4.1. Comparisons with ancient nuclear DNA divergence estimates gap between these successive units. This observation is consistent with the lack of a preserved erosive surface between LU-5 and LU- The new Bayesian modeled age range of 455 ± 17 to 440 ± 15 ka 6, and suggests that their combined accumulation took place over (1s uncertainties) for the SH fossils has implications for under- a time frame comparable to our likelihood uncertainty ranges. The standing the nature of human evolution in Middle Pleistocene Cafe con Leche deposits (LU-7), which immediately overlie the Europe, not least because the SH specimens represent the earliest

Figure 6. Bayesian-modeled durations of the lithostratigraphic units at Sima de los Huesos. The PDFs have been calculated from the modeled posterior probabilities of the upper and lower boundaries of each stratigraphic unit (shown as gray PDFs in Fig. 5) using the ‘date’ query function in OxCal v4.2.4. 90 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95

Table 6 Bayesian modeled posterior age ranges and depositional durations for the Sima de los Huesos stratigraphic units. Posterior ages/durations are presented as the 68.2% and 95.4% highest probability density ranges. The mean and 1s uncertainty ranges of the modeled posterior distributions are shown for comparison (assuming a normally distributed probability density function).

Unit/boundary Time variable Modeled age (years)a Modeled duration (years)a,b

68.2% range 95.4% range Mean ± 1s 68.2% range 95.4% range Mean ± 1s

LU-11 age range 2400e21350 300e39650 16750 ± 11050 LU-11 duration 10e15450 10e37400 12900 ± 11300 LU-11 e LU-10 boundary duration 10e12500 10e29700 10300 ± 9150 LU-10 age range 21350e60050 16900e63050 38350 ± 13350 LU-10 duration 10e11600 10e27950 9600 ± 8750 LU-10 e LU-9 boundary duration 10e7900 10e21250 6750 ± 6700 LU-9 age range 48050e66650 34500e68600 53650 ± 9700 LU-9 duration 10e8950 10e23650 7400 ± 7400 LU-9 e LU-8 boundary duration 10e7800 10e21150 6750 ± 6650 LU-8 age range 62700e79450 183750e51100 86450 ± 42350 LU-8 duration 10e29500 10e219250 44500 ± 71600 LU-8 e LU-7 boundary duration 286400e357750 372200e110450 289750 ± 74300 LU-7 age range 399600e431700 374500e447950 412300 ± 20000 LU-7 duration 10e31300 10e82450 27800 ± 29050 LU-7 e LU-6 boundary duration 25e17400 10e38700 14250 ± 12000 LU-6 age range 431400e461900 417750e479050 447850 ± 15450 LU-6 duration 10e17500 10e42900 14750 ± 13550 LU-6 e LU-5 boundary duration 10e20050 10e46500 16550 ± 14550 LU-5 age range 458500e506800 438550e546000 489050 ± 28550 LU-5 duration 6e39000 -6e108650 34550 ± 36100 LU-5 e LU-4 boundary duration 21e109550 21e234650 88100 ± 72400 LU-4 age range 539700e707500 492750e804500 642450 ± 83800 LU-4 duration 21e119800 21e254400 96100 ± 78750 LU-4 e LU-3 boundary duration 21e109700 21e235050 88450 ± 72650 LU-3 age range 751100e886400 675750e981450 821950 ± 74350 LU-3 duration 3e106550 30e231850 86000 ± 72700 LU-3 e LU-2 boundary duration 9e65000 9e151300 53750 ± 47250 LU-2 age range 875050e1106500 827400e1325850 1041900 ± 134800 LU-2 duration 12e282200 7e598350 246150 ± 195850

a Modeled age ranges were calculated from the posterior probabilities of the upper and lower boundaries (top and bottom) of each stratigraphic unit (see Table 5) using the ‘date’ query function in OxCal v4.2. Modeled durations were calculated using the ‘difference’ query function, and provide the temporal range between the posterior probability density distributions of successive stratigraphic boundaries. b The OxCal ‘difference’ function can be used to test whether or not the posterior probability distributions of successive stratigraphic boundaries are significantly different from each other at a given confidence interval. When the ‘difference’ function is applied to adjacent boundaries of two different stratigraphic units (e.g., LU-6eLU-5 boundary), it provides a statistical indication of the presence or absence of potential depositional hiatuses (given the available dating evidence). Calculated duration ranges that overlap with 0 at the 95.4% confidence interval suggest that the boundaries of successive units are not separated by a statistically significant temporal hiatus. Calculated duration ranges that do not overlap with 0 at the 95.4% confidence interval indicate potentially missing material and/or a temporal gap between the boundaries of successive units. When the ‘difference’ function is applied to the upper and lower boundaries of the same stratigraphic unit (as opposed to the adjacent boundaries of successive units), a calculated duration of >0 years indicate a statistically significant temporal difference between the onset and termination of the depositional event. In contrast, calculated duration ranges that overlap with 0 indicate statistically indistinguishable onset and termination ages for the depositional event at a given confidence interval. stages of speciation in the Neandertal lineage. As detailed in the 4.2. Contextualizing the Sima de los Huesos fossils within the Introduction section, morphological studies of the SH hominin climatic record of the region fossils have revealed the presence of derived Neandertal traits in the dentition, mandible and crania of the entire SH assemblage The new bracketing ages established in this study suggest that (Rosas, 1987; Bermúdez de Castro, 1993; Arsuaga et al., 1993, the SH hominin fossils most likely accumulated during MIS 12. This 1997b, 2014; Martinon-Torres et al., 2012). Nuclear DNA analysis glacial period represents one the most extreme cold phases globally has also confirmed the Neandertal ancestry of the SH fossils and within the past 1 Myr (Voelker et al., 2010), and its termination indicated that they postdate the split from a common ancestral coincided with the end of a long transition in the global climatic lineage with the Denisovans (Meyer et al., 2016). At present, the structure (the EarlyeMiddle Pleistocene transition or EMPT from age estimates for the split time between the archaic and modern 1.4 to 0.4 Ma), which saw an increment in average global ice vol- human lineages, as well as the Denisovan and Neandertal line- ume, as well as an increase in the amplitude (more extreme) and ages, are based on nuclear DNA data obtained from late Nean- length of glacial/interglacial cycles (from ~41 kyr to ~100 kyr in dertal, Denisovan and modern human fossil samples (Prüfer et al., duration), and the establishment of asymmetric global ice volumes 2014). These nuclear DNA results place the split of the modern (Head and Gibbard, 2015). This long transition culminated in the human lineage from the Neandertal/Denisovan lineage as occur- Mid-Brunhes Event (MBE) between MIS 11 and MIS 12, when the ring between 765 ka and 550 ka (Prüfer et al., 2014). Our modeled present-day features of the glacial/interglacial cycles became age bracket for the SH hominins of 447.9 ± 15.5 ka (1s uncer- established (Augustin et al., 2004). Sediment cores recovered from tainty) is compatible with (i.e., postdates) this genetic divergence the North Atlantic, off the western coast of Iberia, record at least 8 estimate and the inferred timing for the beginning of the Nean- cold Heinrich-type events during the 580e300 ka period, with dertal lineage. Similarly, our combined age for the SH fossils is southward intrusions of polar waters being more frequent during compatible with the genetic estimate for the split time between MIS 12 than in the previous glacial stage (Rodrigues et al., 2011). Neandertal and Denisovan populations at 473e381 ka (Prüfer The first cold Heinrich-type event of MIS 12 occurred at ~475 ka et al., 2014). after a long (530e479 ka) and stable interglacial (MIS 13) stage, and M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 91 was followed by three more closely spaced cold events between least in the preserved regions). The latest SH dating results there- ~450 and ~430 ka. The extreme conditions of MIS 12 are also fore point to a complex phylogeographic pattern, with several lin- recorded in other marine cores from the North Atlantic and eages coexisting in the same continent for perhaps a long period. terrestrial records from Europe, which provide evidence that the An increasing number of Middle Pleistocene hominin fossil sites Fennoscandian and British ice sheets coalesced for the first time across Europe now have reliable age control based on radiometric during this glacial to form a single ice sheet that effectively changed dating techniques. One of the oldest dated hominin remains for this fluvial discharge patterns across the northern portion of the time period in Europe is the Mauer mandible, the holotype of Homo continent (Pawley et al., 2008; Toucanne et al., 2009; Roskosch heidelbergensis, which has been constrained to 609 ± 40 ka (likely et al., 2014). Although Iberia remained largely ice-free throughout MIS 15) by infrared radiofluorescence dating of bracketing sedi- the Middle Pleistocene, pollen analysis of marine cores collected ments and combined ESR-U-series dating of herbivore teeth from the Gulf of Cadiz suggest that the region experienced a sig- (Wagner et al., 2010, 2011). Chronologically, the next set of well- nificant increment in semi-desert vegetation during MIS 12 when dated hominin fossils available in Europe belong to the long and compared to pre-MBE glacial stages (i.e., MIS 30), with vegetation cold glacial stage MIS 12 and include SH and Arago in France composition being similar to that described for MIS 2 (Sanchez (Falgueres et al., 2015). The Ceprano specimen from Italy (Manzi Goni~ et al., 2016). Within this paleoenvironmental and climatic et al., 2010), which lacks Neandertal traits, has also been attrib- context, the emergence of Neandertal specializations in the SH uted to MIS 12, although revised Ar/Ar dating of volcanic feldspar hominins may be associated with the aridification of Iberia and grains indicate that this hominin fossil may have accumulated emergence of distinct cold conditions during the Middle Pleisto- within MIS 10 (352 ± 4 ka; Nomade et al., 2011). The partial skulls cene. These distinctive ecological changes likely occurred in the from Swanscombe (England) and Aroeira (Portugal), as well as the region for the first time during the MIS 12 glacial climatic regime, >397 ka Balanica (Serbia) mandible, have been attributed to MIS 11 which was more pronounced and protracted than the preceding (Rink et al., 2013; Daura et al., 2017). Several Central European cycle, and which interrupted a long period of ~240 kyr (between hominin crania likely belong to the MIS 9 interglacial, including the beginning of MIS 15 and the end of MIS 13) of less severe cli- Steinheim, Reilingen and Bilzingsleben (Germany), and matic conditions (Voelker et al., 2010; Rodrigues et al., 2011). Vertossz oll€ os€ (Hungary), although firm numerical age control is The local climate characteristics of MIS 12 may have also played lacking at some of these localities and the latter site may be younger a key role in triggering the geomorphic processes (sediment input, than MIS 9 (Vandermeersch and Garralda, 2011). The hominin accumulation and flushing) that lead to the accumulation of clay- fossils from Ehringsdorf-Weimar (Germany), Biache (France), Api- rich sediments within the SH chamber. The sediments that dima (Greece) and Cova Negra (Spain) likely belong to MIS 7 encompass the SH hominins (LU-6) are dominated by characteristic (Guipert et al., 2011; Bartsiokas et al., 2017; Richard et al., 2018). The red clays seen preserved as the allochthonous Cíclopes red/orange Petralona cranium has been dated to 150e250 ka by U-series and breccia and exterior topsoil deposits, which likely reached the SH ESR dating (Grün, 1996). However, a comparative study of an Ursus chamber via low-energy (flowing or dripping) water trans- deningeri cranium recovered in the vicinity of the Petralona homi- portation. An increase in exposed surface sediments (including soil nin cranium, and within the same cave chamber, points to a clays) in the exterior environment and the predominance of debris- primitive morphology that is similar, in evolutionary terms, to that flow processes near the cave openings could have occurred during displayed by the large SH bear collection. These biostratigraphic past climatic regimes characterized by low vegetation cover, as well comparisons are not compatible with a MIS 7e6 age attribution for as intense rainstorm or rapid snowmelt (Decaulne et al., 2005). the Petralona bear specimens (Santos et al., 2014). As such, the These environmental conditions could have been prevalent at hominin fossil from Petralona could potentially be older than the Atapuerca during the various closely spaced, cold Heinrich-type existing dating suggests (perhaps MIS 12 or 11). events of MIS 12, particularly given the high elevation of the lo- Within the increasingly robust chronological framework being cality (985e990 masl; Sanchez Goni~ and d'Errico, 2005; see established for the hominin fossil record of Europe, it is now Gonzalez-Samp eriz et al., 2010 for review on last glacial pollen possible to clarify patterns of human evolution across the continent records of central Iberia). Rapid snowmelt induced by rainfall or and to reconstruct more detailed scenarios for the origin of Nean- sudden temperature increases could have acted to intensify water dertals. The combined continental fossil record indicates that infiltration into the surrounding soils. Mass movements triggered Neandertal cranial specializations became firmly established by MIS by solifluction and other periglacial processes in this region could 7e6. The European hominin fossils pertaining to MIS 12, which have also resulted in the formation of clay-rich deposits (Lundberg include SH, Arago and possibly Ceprano, are morphologically and McFarlane, 2007). Under this climatic scenario, snowmelt may different, with the SH paleodeme exhibiting clear Neandertal traits have acted to recharge the shallow fractured limestone and raise in the face (including supraorbital torus), mandible, teeth and the pore-pressure beneath the soils, thus triggering debris flow temporomandibular joint, as well as in the suprainiac area of the events. It seems feasible, therefore, that the sedimentation taking occipital bone. These Neandertal features are in general incipient, place in SH before (LU-5), during (LU-6) and after the accumulation and less marked than those observed in the MIS 7 specimens, but of the hominin bones (LU-7) was significantly influenced by the they clearly indicate that the SH paleodeme belongs to the Nean- prevailing MIS 12 climate, which would have been locally charac- dertal clade, as confirmed by nuclear DNA sequencing (Meyer et al., terized by extensive cold periods, pronounced cycles of snowmelt 2016). In the case of the Arago specimens, the general morphology and intense rain, and reduced evapotranspiration losses of soil appears to be non-Neandertal, with some researchers considering water in the Atapuerca region. these fossils to be a separate taxon (i.e., Homo erectus tautavelensis; de Lumley, 2015). The Ceprano calvaria, which is either contem- 4.3. Neandertal evolution in Europe poraneous with or younger than the SH fossils, also exhibits primitive neurocranium morphology, and is similar to the Arago 21 Our new ages for LU-5 and LU-6 confirm that the SH hominins specimen. However, the Ceprano fossil is incomplete and does not are the earliest known palaeodeme exhibiting clear Neandertal preserve the main skeletal regions that display the earliest Nean- features. Significantly, several other European Middle Pleistocene dertal traits at SH (face and mandible); hence the phylogenetic fossils of supposedly similar or younger age do not show the same position of the Ceprano fossil is uncertain. The Aroeira 3 partial clear Neandertal features that are apparent in the SH individuals (at cranium, dated to MIS 11, also lacks the facial skeleton, but the 92 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 supraorbital torus is continuous in the glabellar region, as observed presented luminescence ages and previously published chronolo- in Neandertals, as well as the SH fossils and the Bilzingsleben gies obtained on underlying and overlying lithostratigraphic units. specimen. The temporomandibular joint of the Aroeira 3 cranium The modeled depositional age range for LU-6 and the hominin fossil also lies at the primitive end of the SH range of morphologies accumulation is between 455 ± 17 ka and 440 ± 15 ka (at the 1s (Daura et al., 2017). The Swanscombe partial cranium, which has uncertainty), with a mean age of 448 ± 15 ka (MIS 12). likewise been dated to MIS 11 based on geomorphic correlations, This latest geochronology study confirms that the SH hominins shows a very similar morphology (in the preserved region) to that of represent the oldest known specimens displaying Neandertal fea- the SH specimens. In contrast, the Balanica mandible (Serbia), also tures. Our bracketing age for the SH hominins is in agreement with likely MIS 11 in age, lacks any Neandertal traits (Rink et al., 2013). the proposed split time for archaic and modern human lineages The purportedly MIS 9 crania from Steinheim and Reilingen, in (765e550 ka), as well as for the Neandertal and Denisovan lineages central Europe, are generally reminiscent of the SH morphology. (473e381 ka), which have been independently estimated using However, other fossils that are suggested as belonging to this time nuclear DNA sequencing. Available climatic records indicate that period (Bilzingsleben and Vertossz oll€ os),€ although incomplete, do MIS 12 was a severe glacial period proceeded by ~240 kyr not show the incipiently Neandertal occipital morphology seen in (encompassing MIS 15, MIS 14 and MIS 13) of milder and more the SH, Swanscombe, Steinheim and Reilingen fossils. uniform climate conditions across the region. We propose that the In sum, there are a number of European fossils that are appar- emergence of Neandertal traits may be associated with the drastic ently younger than SH but nevertheless look less derived towards environmental changes that likely occurred in Iberia during MIS 12, the Neandertal condition. As the known hominin fossil sites are though additional paleoenvironmental and fossil records from this isolated and fragmentary, the co-occurrence of remains with time period would help to confirm potential evolutionary links. Neandertal-derived features and others preserving more primitive Comparison of the latest SH dating results with other European features may be explained by intra-deme variation. However, there Middle Pleistocene hominin fossil records reveals a complex phy- is scarce variation in the phylogenetic traits of interest within the logeographic pattern, with several lineages coexisting across the SH assemblage, which would argue against the interpretation that continent over potentially long periods of time. The development of different individuals of the same population exhibited different more widespread, reliable age control is now critical for further combinations of primitive features and Neandertal derived traits. refining evolutionary histories and phylogenetic relationships of As a consequence, we favor an alternative explanation that seems the European Middle Pleistocene hominin record. The new chro- more consistent with the SH data: that the evolution of the nologies presented in this study demonstrate the potentially useful Neandertal lineage during the European Middle Pleistocene was role that extended-range luminescence dating methods could play non-anagenetic. In other words, not all of the European populations in this area of research. evolved at the same pace towards classic Neandertals and there was not one single European population during the Middle Pleis- Acknowledgements tocene. Rather, there were several paleodemes existing until to- wards the end of the Middle Pleistocene, some of which were more M.D. is funded by Australian Research Council (ARC) Discovery primitive and some of which show clear Neandertal apomorphies; Early Career Researcher Award DE160100743 and L.J.A. is funded by it is likely that some paleodemes exhibiting Neandertal apomor- ARC Future Fellowship project FT130100195. NS is funded by a Juan phies also show their own autapomorphies, which would exclude de la Cierva-Incorporacion Fellowship (IJCI-2017-32804). This them from the direct ancestry of the Pleistocene Neandertals, and research has also received support from the Spanish Ministerio de this is an important issue to be investigated in the future. There- Economía y Competitividad (project CGL2015-65387-C3-2-P- after, the classic Neandertal constellation of specializations in the MINECO/FEDER). The authors wish to thank the Atapuerca research masticatory apparatus and neurocranium became dominant. As and excavation team, especially those involved in the excavations at postulated by Hublin (2009), climatic instability prevailing the Sima de los Huesos site. Fieldwork at the Sierra de Atapuerca throughout the Middle Pleistocene probably had a major role in this sites was financed by the Junta de Castilla y Leon and the Fundacion evolutionary process. Our refined ages for SH suggest that the Atapuerca. mosaic pattern of Neandertal evolution likely originated during a long and very harsh glacial period (MIS 12), which inevitably would Appendix A. Supplementary Online Material have resulted in a drastic reduction of the hominin geographic range across Europe. This population contraction would likely have Supplementary online material related to this article can be produced a (series of) genetic bottleneck(s) that may have led to the found at https://doi.org/10.1016/j.jhevol.2018.12.003. origin of the Neandertal lineage (perhaps in the Iberian Peninsula), or at least accelerated its evolution. References

5. Conclusion Aguirre, E., 2000. Sima de los Huesos. Escenarios de la formacion del yacimiento, crítica y sesgo demografico. In: Dobon, L.C., Otero, H.R., Compadre, B.S., Martínez, B.L., Blanco, M.J. (Eds.), Tendencias Actuales de Investigacion en la This study has addressed important gaps in existing chrono- Antropología Física Espanola.~ Universidad de Leon, Leon, pp. 31e42. logical control at SH by establishing a firm maximum age and Aitken, M.J., 1998. An Introduction to Optical Dating: The Dating of Quaternary coeval age for the hominin remains. Depositional ages of 453 ± 56 Sediments by the Use of Photon-Stimulated Luminescence. Oxford University ka, 437 ± 38 ka, 457 ± 41 ka and 460 ± 39 ka were obtained on the Press, Oxford, p. 267. Andrews, P., Fernandez-Jalvo, Y., 1997. Surface modifications of the Sima de los red clays underlying and encasing the hominin remains (LU-5 and Huesos fossil humans. Journal of Human Evolution 33, 191e217. LU-6, respectively) using single-grain TT-OSL dating of sedimentary Aranburu, A., Arsuaga, J.L., Sala, N., 2017. The stratigraphy of the Sima de los Huesos (Atapuerca, Spain) and implications for the origin of the fossil hominin accu- quartz grains. The new ages for LU-5 and LU-6 are consistent with e s mulation. Quaternary International 433, 5 21. each other at 1 and are in close stratigraphic agreement with the Arnold, L.J., Roberts, R.G., 2009. Stochastic modelling of multi-grain equivalent dose published luminescence and U-series age of ~430 ka age for sedi- (De) distributions: Implications for OSL dating of sediment mixtures. Quater- mentary material overlying the hominin fossils (Arsuaga et al., nary Geochronology 4, 204e230. Arnold, L.J., Demuro, M., 2015. Insights into TT-OSL signal stability from single-grain 2014). To derive combined age estimates for each depositional analyses of known-age deposits at Atapuerca, Spain. Quaternary Geochronology phase at SH we have constructed a Bayesian model using the newly 30B, 472e478. M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 93

Arnold, L.J., Roberts, R.G., Galbraith, R.F., DeLong, S.B., 2009. A revised burial dose Bermúdez de Castro, J.M., Martinon-Torres, M., Rosell, J., Blasco, R., Arsuaga, J.L., estimation procedure for optical dating of young and modern-age sediments. Carbonell, E., 2016. Continuity versus discontinuity of the human settlement of Quaternary Geochronology 4, 306e325. Europe between the late Early Pleistocene and the early Middle Pleistocene. The Arnold, L.J., Duval, M., Falgueres, C., Bahain, J.-J., Demuro, M., 2012a. Portable mandibular evidence. Quaternary Science Reviews 153, 51e62. gamma spectrometry with cerium-doped lanthanum bromide scintillators: Bischoff, J.L., Fitzpatrick, J.A., Leon,L.,Arsuaga,J.L.,Falgueres,C.,Bahain,J.J., suitability assessments for luminescence and electron spin dating applications. Bullen, T., 1997. Geology and preliminary dating of the hominid-bearing Radiation Measurements 47, 6e18. sedimentary fill of the Sima de los Huesos Chamber, Cueva Mayor of the Arnold, L.J., Demuro, M., Navazo Ruiz, M., 2012b. Empirical insights into multi-grain Sierra de Atapuerca, Burgos, Spain. Journal of Human Evolution 33, averaging effects from ‘pseudo’ single-grain OSL measurements. Radiation 129e154. Measurements 47, 652e658. Bischoff, J.L., Shamp, D.D., Aramburu, A., Arsuaga, J.L., Carbonell, E., Bermúdez de Arnold, L.J., Demuro, M., Navazo, M., Benito-Calvo, A., Perez-Gonz alez, A., 2013. OSL Castro, J.M., 2003. The Sima de los Huesos hominids date to beyond U/Th dating of the Middle Palaeolithic Hotel California site, Sierra de Atapuerca, equilibrium (>350 kyr) and perhaps to 400e500 kyr: new radiometric dates. north-central Spain. Boreas 42, 285e305. Journal of Archaeological Science 30, 275e280. Arnold, L.J., Demuro, M., Pares, J.M., Arsuaga, J.L., Aranburu, A., Bermúdez de Brennan, B.J., 2003. Beta doses to spherical grains. Radiation Measurements 37, Castro, J.M., Carbonell, E., 2014. Luminescence dating and palaeomagnetic age 299e303. constraint on hominins from Sima de los Huesos, Atapuerca, Spain. Journal of Bronk Ramsey, C., 2009. Bayesian analysis of radiocarbon dates. Radiocarbon 51, Human Evolution 67, 85e107. 337e360. Arnold, L.J., Demuro, M., Pares, J.M., Perez-Gonz alez, A., Arsuaga, J.L., Bermúdez de Bøtter-Jensen, L., Mejdahl, M., 1988. Assessment of beta dose-rate using a GM Castro, J.M., Carbonell, E., 2015. Evaluating the suitability of extended-range multicounter system. Nuclear Tracks Radiation Measurements 14, 187e191. luminescence dating techniques over Early and Middle Pleistocene time- Bowler, J.M., Johnston, H., Olley, J.M., Prescott, J.R., Roberts, R.G., et al., 2003. New scales: Published datasets and case studies from Atapuerca, Spain. Quaternary ages for human occupation and climate change at Lake Mungo, Australia. Na- International 389, 167e190. ture 421, 837e840. Arnold, L.J., Duval, M., Demuro, M., Spooner, N.A., Santonja, M., Perez-Gonz alez, A., Buylaert, J.-P., Jain, M., Murray, A.S., Thomsen, K.J., Thiel, C., Sohbati, R., 2012. 2016. OSL dating of individual quartz 'supergrains' from Ancient Middle A robust feldspar luminescence dating method for Middle and Late Pleistocene Palaeolithic site of Cuesta de la Bajada, Spain. Quaternary Geochronology 36, sediments. Boreas 41, 435e451. 78e101. Carbonell, E., Mosquera, M., 2006. The emergence of a symbolic behaviour: the Arnold, L.J., Demuro, M., Spooner, N.A., Prideaux, G.J., McDowell, M.C., Camens, A.B., sepulchral pit of Sima de los Huesos, Sierra de Atapuerca, Burgos, Spain. Reed, E.H., Pares, J.M., Arsuaga, J.L., Bermúdez de Castro, J.M., Carbonell, E., 2019. Comptes Rendus Palevol 5, 155e160. Single-grain TT-OSL bleaching characteristics: insights from modern analogues Conde-Valverde, M., Quam, R., Martínez, I., Arsuaga, J.-L., Daura, J., Sanz, M., and OSL dating comparisons. Quaternary Geochronology 49, 45e51. Zilhao,~ J., 2018. The bony labyrinth in the Aroeira 3 Pleistocene cranium. Journal Arriolabengoa, M., Iriarte, E., Aranburu, A., Yusta, I., Arnold, L.J., Demuro, M., of Human Evolution 125, 105e116. Arrizabalaga, A., 2018. Reconstructing the sedimentary history of Lezetxiki II Daura, J., Sanz, M., Arsuaga, J.L., Hoffmann, D.L., Quam, R.M., Ortega, M.C., Santos, E., cave (Basque Country, northern Iberian Peninsula) using micromorphological Gomez, S., Rubio, A., Villaescusa, L., Souto, P., Mauricio, J., Rodrigues, F., analysis. Sedimentary Geology 372, 96e111. Ferreira, A., Godinho, P., Trinkaus, E., Zilhao,~ J., 2017. New Middle Pleistocene Arsuaga, J.L., Martínez, I., 2004. Atapuerca y la Evolucion Humana. Fundacion Caixa hominin cranium from Gruta da Aroeira (Portugal). Proceedings of the National Catalunya, Barcelona. Academy of Sciences USA 114, 3397e3402. Arsuaga, J.L., Martínez, I., Gracia, A., Carretero, J.-M., Carbonell, E., 1993. Three new Decaulne, A., Sæmundsson, þ., Petursson, O., 2005. Debris flow triggered by rapid human skulls from the Sima de los Huesos Middle Pleistocene site in Sierra de snowmelt: a case study in the Glei .arhjalli area, northwestern Iceland. Geo- Atapuerca, Spain. Nature 362, 534e537. grafiska Annaler Series A Physical Geography 87, 487e500. Arsuaga, J.L., Martínez, I., Gracia, A., Carretero, J.M., Lorenzo, C., García, N., de Lumley, M.-A., 2015. L’homme de Tautavel. Un Homo erectus europeen evolu e. Ortega, A.I., 1997a. Sima de los Huesos (Sierra de Atapuerca, Spain). The site. Homo erectus tautavelensis. L'Anthropologie 119, 303e348. Journal of Human Evolution 33, 109e127. Demuro, M., Arnold, L.J., Froese, D.G., Roberts, R.G., 2013. OSL dating of loess de- Arsuaga, J.L., Martínez, I., Gracia, A., Lorenzo, C., 1997b. The Sima de los Huesos posits bracketing Sheep Creek tephra beds, northwest Canada: Dim and prob- crania (Sierra de Atapuerca, Spain). A comparative study. Journal of Human lematic single-grain OSL characteristics and their effect on multi-grain age Evolution 33, 219e281. estimates. Quaternary Geochronology 15, 67e87. Arsuaga, J.L., Martínez, I., Arnold, L.J., Aranburu, A., Gracia-Tellez, A., Sharp, W.D., Demuro, M., Arnold, L.J., Pares, J.M., Perez-Gonz alez, A., Ortega, A.I., Arsuaga, J.L., Quam, R.M., Falgueres, C., Pantoja-Perez, A., Bischoff, J., Poza-Rey, E., Pares, J.M., Bermúdez de Castro, J.M., Carbonell, E., 2014. New luminescence ages for the Carretero, J.M., Demuro, M., Lorenzo, C., Sala, N., Martinon-Torres, M., García, N., Galería Complex archaeological site: Resolving chronological uncertainties on Alcazar de Velasco, A., Cuenca-Bescos, G., Gomez-Olivencia, A., Moreno, D., the Acheulean record of the Sierra de Atapuerca, northern Spain. PLoS One 9, Pablos, A., Shen, C.-C., Rodríguez, L., Ortega, A.I., García, R., Bonmatí, A., Ber- e110169. múdez de Castro, J.M., Carbonell, E., 2014. Neandertal roots: cranial and chro- Demuro, M., Arnold, L.J., Pares, J.M., Sala, R., 2015. Extended-range luminescence nological evidence from Sima de los Huesos. Science 344, 1358e1363. chronologies suggest potentially complex bone accumulation histories at the Augustin, L., Barbante, C., Barnes, P.R.F., Barnola, J.M., Bigler, M., Castellano, E., Early-to-Middle Pleistocene palaeontological site of Huescar-1 (Guadix-Baza Cattani, O., Chappellaz, J., Dahl-Jensen, D., Delmonte, B., Dreyfus, G., Durand, G., basin, Spain). Quaternary International 389, 191e212. Falourd, S., Fischer, H., Flückiger, J., Hansson, M.E., Huybrechts, P., Jugie, G., Demuro, M., Arnold, L.J., Aranburu, A., Arsuaga, J.L., 2016. Extended-range lumi- Johnsen, S.J., Jouzel, J., Kaufmann, P., Kipfstuhl, J., Lambert, F., Lipenkov, V.Y., nescence dating of sedimentary infill sequences of the Atapuerca karst com- Littot, G.C., Longinelli, A., Lorrain, R., Maggi, V., Masson-Delmotte, V., Miller, H., plex: determining access history to the Sima de los Huesos site. Proceedings of Mulvaney, R., Oerlemans, J., Oeter, H., Orombelli, G., Parrenin, F., Peel, D.A., the European Society for the Study of Human Evolution 5, 76. Petit, J.-R., Raynaud, D., Ritz, C., Rith, U., Schwander, J., Siegenthaler, U., Demuro, M., Arnold, L.J., Aranburu, A., Gomez-Olivencia, A., Arsuaga, J.L., 2019. Souchez, R., Stauffer, B., Steffensen, J.P., Stenni, B., Stocker, T.F., Tabacco, I.E., Single-grain OSL dating of the Middle Palaeolithic site of Galería de las Estatuas, Udisti, R., van de Wal, R.S.W., van den Broeke, M., Weiss, J., Wilhelms, F., Atapuerca (Burgos, Spain). Quaternary Geochronology 49, 254e261. Winther, J.-G., Wolff, E.W., Zucchelli, M., 2004. Eight glacial cycles from an Dennell, R.W., Martinon-Torres, M., Bermúdez de Castro, J.M., 2011. Hominin vari- Antarctic ice core. Nature 429, 623e678. ability, climatic instability and population demography in Middle Pleistocene Bailey, R.M., Arnold, L.J., 2006. Statistical modelling of single grain quartz De dis- Europe. Quaternary Science Reviews 30, 1511e1524. tributions and an assessment of procedures for estimating burial dose. Qua- Díez, J.C., 1990. Estudio tafo-zooarqueologico del Pleistoceno Medio. Aplicacion a la ternary Science Reviews 25, 2475e2502. Sima de los Huesos. Sierra de Atapuerca. Burgos. In: I Jornadas Burgalesas de Bartsiokas, A., Arsuaga, J.L., Aubert, M., Grün, R., 2017. U-series dating and classifi- Historia Introduccion a la Historia de Burgos en la Edad Media. Monografias de cation of the Apidima 2 hominin from Mani Pensinsula, Suthern Greece. Journal Historia Castellano-Leonesa, Burgos, pp. 517e530. of Human Evolution 109, 22e29. Duller, G.A.T., 2003. Distinguishing quartz and feldspar in single grain luminescence Bartz, M., Arnold, L.J., Demuro, M., Duval, M., King, G.E., Rixhon, G., Alvarez measurements. Radiation Measurements 37, 161e165. Posada, C., Pares, J.M., Brückner, H., 2019. Single-grain TT-OSL dating results Duller, G.A.T., 2007. Assessing the error on equivalent estimates derived from single confirm an Early Pleistocene age for the lower Moulouya River deposits (NE aliquot regenerative dose measurements. Ancient TL 25, 15e24. Morocco). Quaternary Geochronology 49, 138e145. Duller, G.A.T., Wintle, A.G., 2012. A review of the thermally transferred optically Berger, G.W., Perez-Gonz alez, A., Carbonell, E., Arsuaga, J.L., Bermúdez de stimulated luminescence signal from quart for dating sediments. Quaternary Castro, J.M., Ku, T.L., 2008. Luminescence chronology of cave sediments at the Geochronology 7, 6e20. Atapuerca palaeoanthropological site, Spain. Journal of Human Evolution 55, Egeland, C.P., Domínguez-Rodrigo, M., Pickering, T.R., Menter, C.G., Heaton, J.L., 300e311. 2018. Hominin skeletal part abundances and claims of deliberate disposal of Bermúdez de Castro, J.M., 1993. The Atapuerca dental remains. New evidence corpses in the Middle Pleistocene. Proceedings of the National Academy of (1987-1991 excavations) and interpretations. Journal of Human Evolution 24, Sciences USA 115, 4601e4606. 339e371. Falgueres, C., Shao, Q., Han, F., Bahain, J.J., Richard, M., Perrenoud, C., Moigne, A.M., Bermúdez de Castro, J.M., Martinon-Torres, M., Lozano, M., Sarmiento, S., Muela, A., Lumley, H. de, 2015. New ESR and U-series dating at Caune de l'Arago, France: a 2004. Palaeodemography of the Atapuerca-SH Middle Pleistocene hominid key-site for European Middle Pleistocene. Quaternary Geochronology 30, sample. A revision and new approaches to the paleodemography of the Euro- 547e553. pean Middle Pleistocene population. Journal of Anthropological Research 60, Frouin, M., Lahaye, C., Hernandez, M., Mercier, N., Guibert, P., Brenet, M., Folgado- 5e26. Lopez, M., Nertran, P., 2014. Chronology of the Middle Palaeolithic open-air site 94 M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95

of Combe Brune 2 (Dordogne, France): a multi luminescence dating approach. Ortega, A.I., Benito-Calvo, A., Perez-Gonz alez, A., Martín-Merino, M.A., Perez- Journal of Archaeological Science 52, 524e534. Martínez, R., Pares, J.M., Aramburu, A., Arsuaga, J.L., Bermúdez de Castro, J.M., Galbraith, R., 2002. A note on the variance of a background-corrected OSL count. Carbonell, E., 2013. Evolution of multilevel caves in the Sierra de Atapuerca Ancient TL 20, 49e51. (Burgos, Spain) and its relation to human occupation. Geomorphology 196, Galbraith, R.F., Green, P.F., 1990. Estimating the component ages in a finite mixture. 122e137. Nuclear Tracks and Radiation Measurements 17, 197e206. Pares, J.M., Arsuaga, J.L., Perez-Gonz alez, A., Weil, A., 2000. On the age of the Galbraith, R.F., Roberts, R.G., 2012. Statistical aspects of equivalent dose error hominid fossils at Sima de los Huesos, Sierra de Atapuerca: paleomagnetic calculation a nd display in OSL dating: An overview and some recommenda- evidence. American Journal of Physical Anthropology 111, 451e461. tions. Quaternary Geochronology 11, 1e17. Pawley, S.M., Bailey, R.M., Rose, J., Moorlock, B.S.P., Hamblin, R.J.O., et al., 2008. Age Galbraith, R.F., Roberts, R.G., Laslett, G.M., Yoshida, H., Olley, J.M., 1999. Optical limits on Middle Pleistocene glacial sediments from OSL dating, north Norfolk, dating of single and multiple grains of quartz from Jinmium rock shelter, UK. Quaternary Science Reviews 27, 1363e1377. northern Australia: Part I, experimental design and statistical models. Prescott, J.R., Hutton, J.T., 1994. Cosmic ray contributions to dose rates for lumi- Archaeometry 41, 339e364. nescence and ESR dating: large depths and long-term time variations. Radiation Gonzalez-Samp eriz, P., Leroy, S.A.G., Carrion, J.S., Fernandez, S., García-Anton, M., Measurements 23, 497e500. Gil-García, M.J., Uzquiano, P., Valero-Garces, B., Figueiral, I., 2010. Steppes, sa- Prüfer, K., Racimo, F., Patterson, N., Jay, F., Sankararaman, S., Sawyer, S., Heinze, A., vannahs, forests and phytodiversity reservoirs during the Pleistocene in the Renaud, G., Sudmant, P.H., de Filippo, C., Li, H., Mallick, S., Dannermann, M., Iberian Peninsula. Review of Palaeobotany and Palynology 162, 427e457. Fu, Q., Kircher, M., Kuhlwilm, M., Lachmann, M., Meyer, M., Ongyerth, M., Grün, R., 1996. A re-analysis of electron spin resonance dating results associated Siebauer, M., Theunert, C., Tandon, A., Moorjani, P., Pickrell, J., Mullikin, J.C., with the Petralona hominid. Journal of Human Evolution 30, 227e241. Vohr, S.H., Green, R.E., Hellmann, I., Johnson, P.L.F., Blanche, H., Cann, H., Guerin, G., Mercier, M., Adamiec, G., 2011. Dose-rate conversion factors: update. Kitzman, J.O., Shendure, J., Eichler, E.E., Lein, E.S., Bakken, T.E., Golovanova, L.V., Ancient TL 29, 5e8. Doronichev, V.B., Shunkov, M.V., Derevianko, A.P., Viola, B., Slatkin, M., Reich, D., Guipert, G., de Lumley, M.-A., Tuffreau, A., Mafart, B., 2011. A late Middle Pleistocene Kelso, J., Pa€abo,€ S., 2014. Nature 505, 43e49. hominid: Bianche-Saint-Vaast 2, North France. Comptes Rendus Palevol 10, Rees-Jones, J., 1995. Optical dating of young sediments using fine-grain quartz. 21e33. Ancient TL 13, 9e14. Head, M.J., Gibbard, P.L., 2015. EarlyeMiddle Pleistocene transitions: Linking Rees-Jones, J., Tite, M.S., 1997. Optical dating results for British archaeological sed- terrestrial and marine realms. Quaternary International 389, 7e46. iments. Archaeometry 39, 177e187. Herisson, D., Brenet, M., Cliquet, D., Moncel, M.-H., Richter, J., Scott, B., Van Richard, M., Falgueres, C., Pons-Branchu, E., Foliot, L., Guillem, P.M., Martínez- Baelen, A., Di Modica, K., De Loecker, D., Ashton, N., Bourguignon, L., Valle, R., Eixea, A., Villaverde, V., 2018. ESR/U-series chronology of early Delanges, A., Faivre, J.-P., Folgado-Lopez, M., Locht, J.-L., Pope, M., Raynal, J.-P., Neanderthal occupations at Cova Negra (Valencia, Spain). Quaternary Roebroeks, W., Santagata, C., Turq, A., Van Peer, P., 2016. The emergence of the Geochronology 49, 283e290. Middle Palaeolithic in north-western Europe and its southern fringes. Quater- Richter, D., Grün, R., Joannes-Boyau, R., Steele, T.E., Amani, F., Rue, M., Fernandes, P., nary International 41, 233e283. Raynal, J.-P., Geraads, D., Ben-Ncer, A., Hublin, J.-J., McPherron, S.P., 2017. The Hernandez, M., Mercier, N., Rigaud, J.-P., Texier, J.-P., Delpech, F., 2014. A revised age of the hominin fossils from Jebel Irhoud, Morocco, and the origins of the chronology for the Grotte Vaufrey (Dordogne, France) based on TT-OSL dating of Middle Stone Age. Nature 546, 293e296. sedimentary quartz. Journal of Human Evolution 75, 53e63. Rink, W.J., Mercier, N., Mihailovic, D., Morley, M.W., Thomspon, J.W., Roksandic, M., Hublin, J.J., 2009. The origins of Neandertals. Proceedings of the National Academy 2013. New radiometric ages for the BH-1 hominin from Balanica (Serbia): Im- of Sciences USA 106, 16022e16027. plications for understanding the role of the Balkans in Middle Pleistocene hu- Hublin, J.J., Roebroeks, W., 2009. Ebb and flow or regional extinctions? On the man evolution. PLoS One 8, e54608. character of Neandertal occupation of northern environments. Comptes Rendus Rodrigues, T., Voelker, A.H.L., Grimalt, J.O., Abrantes, F., Naughton, F., 2011. Iberian Palevol 8, 503e509. Margin sea surface temperature during MIS 15 to 9 (580e300 ka): glacial Hublin, J.-J., Ben-Ncer, A., Bailey, S.E., Freidline, S.E., Neubauer, S., Skinner, M.M., suborbital variability versus interglacial stability. Paleoceanography 26, Bergmann, I., Le Cabec, A., Benazzi, S., Harvati, K., Gunz, P., 2017. New fossils PA1204. from Jebel Irhoud, Morocco and the pan-African origin of Homo sapiens. Nature Rodríguez-Rey, M., Herrando-Perez, S., Gillespie, R., Jacobs, Z., Saltre, F., Brook, B.W., 546, 289e292. Prideaux, G.J., Roberts, R.G., Cooper, A., Alroy, J., Miller, G.H., Bird, M.I., Jacobs, Z., Duller, G.A.T., Wintle, A.G., Henshilwood, C.S., 2006. Extending the Johnson, C.N., Beeton, N., Turney, C.S.M., Bradshaw, C.J.A., 2015. Criteria for chronology of deposits at Blombos Cave, South Africa, back to 140 ka using assessing the quality of Middle Pleistocene to Holocene vertebrate fossil ages. optical dating of single and multiple grains of quartz. Journal of Human Evo- Quaternary Geochronology 30, 69e79. lution 51, 255e273. Roebroeks, W., Soressi, M., 2016. Neanderthals revised. Proceedings of the National Jacobs, Z., Roberts, R.G., Lachlan, T.J., Karkanas, P., Marean, C.W., Roberts, D.L., 2011. Academy of Sciences USA 113, 6372e6379. Development of the SAR TT-OSL procedure for dating Middle Pleistocene dune Rosas, A., 1987. Two new mandibular fragments from Atapuerca/Ibeas (SH site). a and shallow marine deposits along the Cape coast of South Africa. Quaternary reassessment of the affinities of the Ibeas mandibles sample. Journal of Human Geochronology 6, 491e513. Evolution 16, 417e427. Li, B., Jacobs, Z., Roberts, R.G., Li, S.-H., 2014. Review and assessment of the potential Roskosch, J., Winsemann, J., Polom, U., Brandes, C., Tsukamoto, S., Weitkamp, A., of post-IR IRSL dating methods to circumvent the problem of anomalous fading Bartholomaus,€ A., Henningsen, D., Frechen, M., 2014. Luminescence dating of in feldspar luminescence. Geochronometria 41, 178e201. ice-marginal deposits in northern Germany: evidence for repeated glaciations Lisiecki, L.E., Raymo, M.E., 2005. A Pliocene-Pleistocene stack of 57 globally during the Middle Pleistocene (MIS 12 to MIS 6). Bores 44, 103e126. distributed benthic d18O records. Paleoceanography 20, PA1003. Sala, N., Arsuaga, J.L., Martínez, I., Gracia-Tellez, A., 2014. Carnivore activity in the Lundberg, J., McFarlane, D.A., 2007. Pleistocene depositional history in a periglacial Sima de los Huesos (Atapuerca, Spain) hominin sample. Quaternary Science terrane: A 500 ka record from Kent's Cavern, Devon, United Kingdom. Geo- Reviews 97, 71e83. sphere 3, 199e219. Sala, N., Arsuaga, J.L., Martínez, I., Gracia-Tellez, A., 2015a. Breakage patterns in Sima Manzi, G., Magri, D., Milli, S., Palombo, M.R., Margari, V., Celiberti, V., Barbieri, M., de los Huesos (Atapuerca, Spain) hominin sample. Journal of Archaeological Barbieri, M., Melis, R.T., Rubini, M., Ruffo, M., Saracino, B., Tsedakis, P.C., Science 55, 113e121. Zarattini, A., Biddittu, I., 2010. The new chronology of the Ceprano calvarium Sala, N., Arsuaga, J.L., Pantoja-Perez, A., Pablos, A., Martínez, I., Quam, R.M., Gomez- (Italy). Journal of Human Evolution 59, 580e585. Olivencia, A., Bermúdez de Castro, J.M., Carbonell, E., 2015b. Lethal interper- Manzi, G., Magri, D., Palombo, M.R., 2011. EarlyeMiddle Pleistocene environmental sonal violence in the Middle Pleistocene. PLoS One 10, e0126589. changes and human evolution in the Italian peninsula. Quaternary Science Sala, N., Pantoja-Perez, A., Arsuaga, J.L., Pablos, A., Martínez, I., 2016. The Sima de los Review 30, 1420e1438. Huesos crania: Analysis of the cranial breakage patterns. Journal of Archaeo- Martinon-Torres, M., Bermúdez de Castro, J.M., Gomez-Robles, A., Prado-Simon, L., logical Science 72, 25e43. Arsuaga, J.L., 2012. Morphological description and comparison of the dental Sanchez Goni,~ M.F., d'Errico, F., 2005. La historia de la vegetacion y el clima del último remains from Atapuerca-Sima de los Huesos (Spain). Journal of Human Evo- ciclo climatico (OIS5-OIS1, 140.000-10.000 anos~ BP) en la Península Iberica y su lution 62, 7e58. posible impacto sobre los grupos paleolíticos. In: Montes Barquín, R., Lasheras Mejdahl, V., 1987. Internal radioactivity in quartz and feldspar grains. Ancient TL 5, Corruchaga, J.A. (Eds.), Actas de la Reunion Científica: Neandertales Cantabricos, 10e17. Estado de la Cuestion. Museo Nacional y Centro de Investigacion de Altamira. Meyer, M., Arsuaga, J.L., de Filippo, C., Nagel, S., Aximu-Petri, A., Nickel, B., Ministerio de Cultura, Santander, pp. 115e129. Martínez, I., Gracia, A., Bermúdez de Castro, J.M., Carbonell, E., Viola, B., Kelso, J., Sanchez Goni,~ M.F., Llave, E., Oliveira, D., Naughton, F., Desprat, S., Ducassou, E., Prüfer, K., Pa€abo,€ S., 2016. Nuclear DNA sequences from the Middle Pleistocene Hodell, D.A., Hernandez-Molina, F.J., 2016. Climate changes in south western Sima de los Huesos hominins. Nature 531, 504e507. Iberia and Mediterranean Outflow variations during two contrasting cycles of Nomade, S., Muttoni, G., Guillou, H., Robin, E., Scardia, G., 2011. First 40Ar/39Ar age of the last 1 Myrs: MIS 31eMIS 30 and MIS 12-MIS 11. Global and Planetary the Ceprano man (central Italy). Quaternary Geochronology 6, 453e457. Change 136, 18e29. Olle, A., Verges, J.M., Rodríguez, X.P., Caceres, I., Angelucci, D.E., Vallverdú, J., Santos, E., Garcia, N., Carretero, J.M., Arsuaga, J.L., Tsoukala, E., 2014. Endocranial Demuro, M., Arnold, L.J., Falgueres, C., Bennasar, M., Lopez-García, J.M., traits of the Sima de los Huesos (Atapuerca, Spain) and Petralona (Chalkidiki, Blain, H.-A., Banuls-Cardona,~ S., Burjachs, F., Exposito, I., Lopez-Polín, L., Lopez- Greece) Middle Pleistocene ursids. Phylogenetic and biochronological implica- Ortega, E., 2016. The Middle Pleistocene site of La Cansaladeta (Tarragona, tions. Annales de Paleontologie 100, 297e309. Spain): Stratigraphic and archaeological succession. Quaternary International Stewart, J.R., Stringer, C.B., 2012. Human evolution out of Africa: the role of refugia 393, 137e157. and climate change. Science 335, 1317e1321. M. Demuro et al. / Journal of Human Evolution 131 (2019) 76e95 95

Thomsen, K.J., Murray, A.S., Jain, M., Bøtter-Jensen, L., 2008. Laboratory fading rates Wang, X.L., Wintle, A.G., Lu, Y.C., 2006a. Thermally transferred luminescence in fine- of various luminescence signals from feldspar-rich sediment extracts. Radiation grained quartz from Chinese loess: Basic observations. Radiation Measurements Measurements 43, 1474e1486. 41, 649e658. Toucanne, S., Zaragosi, S., Bourillet, J.F., Gibbard, P.L., Eynaud, F., Turon, J.L., Wang, X.L., Lu, Y.C., Wintle, A.G., 2006b. Recuperated OSL dating of fine-grained Cremer, M., Cortijo, E., Martinez, P., Ressignol, L., 2009. A 1.2 Ma record of quartz in Chinese loess. Quaternary Geochronology 1, 89e100. glaciation and fluvial discharge from the West European Atlantic margin. Wagner, G.A., Krbetschek, M., Degering, D., Bahain, J.-J., Shao, Q., Falgueres, C., Quaternary Science Reviews 28, 2974e2981. Voinchet, P., Dolo, J.-M., Garcia, T., Rightmire, G.P., 2010. Vandermeersch, B., Garralda, M.D., 2011. Neanderthal geographical and chrono- Radiometric dating of the type-site for Homo heidelbergensis at Mauer, logical variation. In: Condemi, S., Weniger, G.-C. (Eds.), Continuity and Germany. Proceedings of the National Academy of Sciences USA 107, Discontinuity in the Peopling of Europe: One Hundred Fifty Years of Neander- 19726e19730. thal Study. Springer, Heidelberg, pp. 113e125. Wagner, G.A., Maul, L.C., Loscher,€ M., Schreiber, H.D., 2011. Mauer e the type site of Voelker, A.H.L., Rodrigues, T., Billups, K., Oppo, D., McManus, J., Stein, R., Hefter, J., Homo heidelbergensis: palaeoenvironment and age. Quaternary Science Reviews Grimalt, J.O., 2010. Variations in mid-latitude North Atlantic surface water 30, 1464e1473. properties during the mid-Brunhes (MIS 9-14) and their implications for the White, W.B., 2007. Cave sediments and palaeoclimate. Journal of Cave and Karst thermohaline circulation. Climate of the Past 6, 531e552. Studies 69, 76e93.