The late Early Pleistocene site of Fuente Nueva-3 (Guadix-Baza Depression, SE Spain): A hyena latrine developed on a quicksand trap for megaherbivores?

The late Early Pleistocene archaeological site of Fuente Nueva-3 (FN3), which lies in the Guadix-Baza Depression (SE Spain) and is dated to ~ 1.4 Ma, contributes some of the oldest evidence of hominin presence in Western Europe, including a huge tool assemblage of Oldowan tradition, manuports (i.e., unmodi�ed stones used as percussion tools) and abundant fossils of large mammals, some of which preserve anthropogenic marks related to de�eshing, butchering and marrow processing. In addition, there are bones with tooth marks produced by scavenging carnivores. The fertile layers of the FN3 section have been grouped in a lower archaeological level (LAL) and an upper archaeological level (UAL). Both levels preserve abundant skeletal remains and lithic tools. However, the LAL shows a high density of manuports, which suggests that hominin activity was more intense at this level, while the UAL preserves many remains of megaherbivores, particularly proboscideans (Mammuthus meridionalis), and almost all coprolites unearthed from the site, which points to a greater involvement of the giant hyenas (Pachycrocuta brevirostris). In this paper, we (i) test for statistical differences in the composition of the faunal assemblages preserved in the UAL and LAL; and (ii) analyze particle size in the fertile layers of both archaeological levels. Our results show that megaherbivores are comparatively overrepresented in the UAL, specially by young elephants, while other medium-to-large and large-sized ungulates, particularly equids, are more abundant in the LAL, showing a predominance of adult individuals. Concerning the sedimentology, layers 2–3 of the LAL show a predominance of silts and clays, with sands representing a minor fraction. In contrast, layer 5 of the UAL is composed of two-thirds of �ne and very �ne sands, with the remaining third consisting of silts and clays. These data and the paleoenvironmental reconstruction of layer 5 suggest that it may be interpreted as a paleo-quicksand in which megaherbivores were


Introduction and background
The late Early Pleistocene archaeological site of Fuente Nueva-3 (FN3) lies in the north-eastern sector of the Guadix-Baza Depression (SE Spain, Figs.1A-B), which is the largest intramontane basin of the Betic Cordillera (García-Tortosa et al., 2011; García-Aguilar and Palmqvist, 2011; Palmqvist et al., 2022a).This inland basin extends over an area of ~ 4000 km 2 and is surrounded by the highest reliefs of the Alpine-Betic orogeny, which rise up to an altitude of ~ 3482 m above mean sea level (AMSL).During the Miocene, the sedimentary depression was connected to the Mediterranean Sea by the Almanzora Corridor and to the Atlantic Ocean by the Guadalquivir Basin.Both corridors were closed between the end of the Tortonian and the Lower Messinian, which led to a change from exorheic conditions to continental sedimentation in an endorheic basin (Soria et al., 1999;Hüsing et al., 2010).Isostatic uplifting, with an average uplift rate of ~ 200 m/Ma as estimated from Late Neogene coastal marine conglomerates and coral reefs (Braga et al., 2003), resulted in an elevation of the uppermost sedimentary level of the basin (i.e., the glacis geomorphological surface) to ~ 1000 m AMSL.
The Guadix-Baza Depression developed a network of endorheic drainage during most of the Pliocene and Pleistocene, being subject to intense tectonic subsidence in relation to the surrounding mountain ranges.This led to the accumulation of a ~ 550 m thick sedimentary record composed of lacustrine, alluvial and uvial deposits (limestones, gypsum, marls, shales, sands, and conglomerates) Tectonically induced hydrothermal activity in the basin provided a mild and productive environment for the terrestrial fauna (García-Aguilar et al., 2014, 2015;Palmqvist et al., 2022a).This favored the development of rich and diverse paleocommunities of mammals through the Plio-Pleistocene, whose fossil remains were abundantly preserved in many paleontological and archaeological localities (Maldonado-Garrido et al., 2017; Martínez- Navarro et al., 2018;Palmqvist et al., 2022a).The Orce area (Figs.1B-C), a satellite basin in the north-eastern sector of the depression with an extent of ~ 170 km 2 , preserves a unique fossil record of the late Early Pleistocene (Late Villafranchian) fauna, including two archaeological sites dated to ~ 1. 4 Ma that mark the earliest presence of hominins in Western Europe, Barranco León-D (BLD) and FN3 (Martínez- Navarro et al., 1997Navarro et al., , 2014;;Espigares et al., 2013Espigares et al., , 2019Espigares et al., , 2023;;Toro-Moyano et al., 2013;Palmqvist et al., 2016Palmqvist et al., , 2022a, a), a).
There are several age estimates for BLD and FN3.For example, a combination of biostratigraphy, magnetostratigraphy and the U-series/ESR dating method applied to optically bleached quartz grains and fossil mammalian teeth provided age estimates of 1.43 ± 0.38 Ma for BLD and 1.19 ± 0.21 for FN3 (Duval et al., 2012;Toro-Moyano et al., 2013).In addition, an age of 1.50 ± 0.31 Ma was derived for FN3 based on cosmogenic nuclides (Álvarez et al., 2015).Moreover, the absence of suids from the rich faunal assemblages of BLD and FN3 contributes signi cant biochronological information (Martínez- Navarro et al., 2015): Suids are not present in Europe during the 1.8-1.2Ma interval until the arrival of a derived form of Sus strozzi during the Epivillafranchian (Cherin et al., 2018(Cherin et al., , 2020)).This species is rst recorded in association with hominin remains at unit TE9 of Sima del Elefante in the Atapuerca karstic complex, which is dated to 1.22 ± 0.16 Ma on the basis of cosmogenic nuclides (Carbonell et al., 2008), and later in other sites of Jaramillo age like Untermassfeld in Germany, Vallonnet in France, or Vallparadís in Spain (Moullé et al., 2006;Madurell-Malapeira et al., 2010, 2014;Cherin et al., 2018Cherin et al., , 2020)).This proposal, however, has recently been challenged by Iannucci (2024) based on a distal fragment of a left 4th suid metatarsal from a fossil collection of the XIX century.The specimen, which is housed in the Natural History Museum, London (NHMUK PV OR 27621), comes from Peyrolles, a locality of uncertain age of the French Central Massif.Although a fossiliferous level of Peyrolles has been dated by the 40 Ar/ 39 Ar method to 1.47 Ma (Nomade et al., 2014), it has been suggested that the short faunal list from this site is biochronologically coeval with the Upper Valdarno (Valli et al., 2006), which points to an age close to 1.8 Ma.Therefore, according to the evidence discussed above, the most parsimonious age for the two Orce sites, BLD and FN3, would be ~ 1. 4 Ma (see review in Palmqvist et al., 2016).
The activity of P. brevirostris is particularly evident in the case of the upper archaeological level of FN3, in which hyena coprolites are abundantly preserved (Espigares, 2010;Espigares et al., 2019Espigares et al., , 2023))

The late Early Pleistocene site of FN3
The stratigraphy of FN3 consists of a ~ 5 m thick section formed by the alternation of sub-horizontal limestones with lutites and ne sands.This stratigraphic section was rst described by Turq et al. (2006) and later by Oms et al. (2011), who described twelve layers (see Figs. 2A-C), numbered from bottom to top and grouped into three sedimentary cycles.Each cycle begins in a lower part constituted by greenish marly lutites and clays, as well as sands with iron oxides and carbonate nodules, and ends in an upper part characterized by a limestone/marly limestone stratum with evidence of microkarsti cation, together with intraclasts and diagenetic markers (Fig. 2D).These lutitic-carbonate sediments were deposited in a shallow lacustrine-toswampy environment of variable depth, with occasional uvial inputs (Turq et al., 2006;Oms et al., 2011;Espigares et al., 2023).Speci cally, the clays and marly lutites were deposited under a semi-permanent water table while the sands correspond to lake shore deposits.
The rst sedimentary cycle is only represented by its upper part (layer 1), a whitish limestone stratum with a soft-sediment deformation structure that shows a rounded, pillow-shaped surface (Fig. 3).The second cycle (layers 2-4) and the third (layers 5-12) are both preserved complete.Layer 2 is an irregular bed of green clays that conforms to the surface of layer 1, showing a similar soft-sediment deformation structure.This layer is of variable thickness (10-20 cm) and includes small intercalations of sand.Layer 3 was deposited in concordance with layer 2 and consists of brown-whitish clays with carbonate nodules.Layers 1-3 show plastic deformation and small synsedimentary normal faults with a vertical extent of a few centimeters (Fig. 2C).In all probability, these structures were formed by earthquakes (Oms et al., 2011), which are frequently recorded in the Baza Basin (Alfaro et al., 1997(Alfaro et al., , 2010;;Gibert et al., 2005).Layer 4 is a whitish limestone that seems to have been less affected by deformation.The bottom of this layer conforms to the paleosurface, while its top is eroded by the deposition of the overlying sediments.The third sedimentary cycle consists of layers 5-12 (Fig. 2B).Layer 5 is of variable thickness (20-30 cm, although it can occasionally be thicker) and consists of greenish sands with lenticular shapes, which can be interpreted as the result of lowenergy uvial ows with su cient energy to erode the unconsolidated limestones of layer 4. A high percentage (68.2%,5466 out of 8014) of the fossil remains of large mammals unearthed from FN3 are from layer 5 (Espigares et al., 2023: Table 1).Layers 6-10 are composed of marly lutites and clays, which show differences in color and thickness.The boundary between layers 7 and 8 shows an important irregularity, which may be indicative of an erosional process resulting from subaerial emersion.Finally, layers 11-12 are horizontal whitish limestones that close the third sedimentary cycle and the stratigraphic section of FN3.  1).Up to the 2015 excavation season, 1367 Mode 1 lithic artifacts knapped in situ were unearthed from FN3.In addition, there are 375 unmodi ed stones or manuports (i.e., non-aked, allochthonous blocks and cobbles from the Jurassic carbonate rocks of the site surroundings).Nearly half of these limestone blocks display percussion marks (Barsky et al., 2015(Barsky et al., , 2016(Barsky et al., , 2022)), which suggests they were probably intentionally brought to the FN3 site by the hominins, who used them as percussion tools for fracturing the bones of large mammals.
In a recent study, Espigares et al. (2023: Table 1) showed that fossil bones and lithic tools are abundantly preserved in both archaeological levels of FN3 (see Fig. 4): 2234 bones and 507 lithic tools come from the LAL, while 5780 bones and 482 tools are from the UAL.However, there are differences between both archaeological levels.For example, the density of manuports is higher in the LAL than in the UAL: 329 out of 375 manuports (87.7%) come from layers 2-3 of the LAL and only 46 (12.3%) from layer 5 of the UAL.This suggests that hominin activity concentrated in the LAL.In contrast, hyena coprolites are more abundantly represented in the UAL than in the LAL (Fig. 4): 98.2% (216 out of 220) coprolites preserved in FN3 come from layer 5 of the UAL and only three were recovered from layers 2-3 of the LAL (Espigares et al., 2023: Table 1).This suggests that hyaenid activity was particularly intense at the UAL.
According to Espigares et al. (2023), the lithological differences between the LAL and UAL could tentatively explain the differences in the composition of their bone assemblages of large mammals.Speci cally, the fossiliferous layers of the UAL (layers 5-7; see Fig. 2) are composed of greenish ne sands and marly mudstones (layer 5), dark brown clays (layer 6), and greenish-brown dark marly mudstones (layer 7).In contrast, the fertile layers of the LAL (layers 2-3) are composed of green clays (layer 2) and brown whitish clays with nodules (layer 3).The composition of layer 5, which preserves most of the skeletal remains from the UAL, led Espigares et al. (2023) to propose that this layer could have behaved as quicksand that became unstable when it was forced to move by the pressure exerted by very large sized animals like megaherbivores, which remains are particularly abundant in this layer.For this reason, Espigares et al. (2023) interpreted layer 5 of the UAL, which preserves a high number of hyena coprolites, as a latrine of the giant hyena P. brevirostris developed in an open-air setting.

Quicksand as a potential deadly trap for megafauna
Quicksand can potentially be a deadly trap for wildlife.Given its composition ( ne sand, silt and salt water), quicksand is extremely sensitive to even small changes in stress, which can result in signi cant variations in its steady-state viscosity, thus causing the sediment to liquefy.For this reason, the higher the stress, the more uid the quicksand becomes, so movement by a trapped body causes it to sink (Khaldoun et al., 2005(Khaldoun et al., , 2006)).
After an animal sinks into quicksand, the viscosity of the sediment increases due to the separation of the quicksand into two phases: a water-rich phase and a sand-rich phase (Khaldoun et al., 2005).This causes an animal trapped in quicksand to sink deeper when it moves trying to escape from the quicksand trap (Khaldoun et al. 2005).In such an environment, viscosity can reach such high levels that an animal may require a force up to three times its weight to free itself from the sediment (Kadau et al., 2009).
If artesian conditions exist, quicksand is formed by spring water issuing from an underlying permeable stratum fed by a head of water at higher elevation (Matthes, 1953).Surface quicksand is likely to occur in hilly country, especially in regions of calcareous rocks like the Baza Basin, which often contain solution openings, thus favoring the occurrence of springs (Patchick, 1966).In these environments, the sun bakes a thin crust on the sand (Matthes, 1953), which hides the risk of this substrate for animals moving in the vicinity.
The coe cient of friction and the penetration hardness of a quicksand depend on the water contents it incorporates (Fall et al., 2014;Liefferink et al., 2020).At low water contents the sand shows a pendular regime and can withstand relatively high loads before reaching the critical penetration hardness.Deformation of the granular material causes a reorientation of the grains, including the breaking and rebuilding of capillary liquid bonds.An increase in water content causes coalescence of the liquid bonds and a shift to a capillary regime, which further decreases the plastic response of the granular material.As a result, the coe cient of friction increases, resulting in an increased risk of entrapment for heavy animals with large limb sections.In addition, the high body mass of a trapped megaherbivore like an elephant makes it more di cult for it to counteract the negative suction of the quicksand to lift its pillar-like legs out of the sand (see Ren et al., 2010).In this scenario, an increase in salinity increases the cohesion and viscosity between the sand grains, which in turn increases the hardness of penetration into the quicksand (Fall et al., 2014).In a warm region like the Baza Basin, the evaporation of water from a sand with salt water results in the formation of salt crystals that ll the spaces between the sand grains.This process binds the sand grains together more tightly, creating a stiffer material that is more resistant to penetration stresses.The increase in strength is particularly noticeable in the pendular regime, where the amount of water (and thus dissolved salt) is relatively low but su cient to form liquid bridges which, when evaporating, leave salt crystals that function as solid bridges in the surface crust.This is a double-edged trap because the surface crust is fragile.For this reason, if the fracture limit at the surface is exceeded (Nguyen et al., 2016), a death trap occurs when the limbs of a large sized animal move into the sand layer with an elevated water content, where the penetration hardness is lower and the friction is higher, thus trapping the animal.This would explain the difference of behavior between the LAL and UAL of FN3, and also why hyenas and hominins were able to roam the surface of quicksand without facing excessive risk of sinking into it, as explained below.
Experimental data on quicksand from salt-lake environments composed of ne sand, clay and salt water, similar in composition to those found in the sediments deposited in the oligosaline lacustrine systems recorded at the Orce sites (García-Aguilar et al., 2014, 2015; Palmqvist et al. 2022a), show that the quicksand acts as a trap for the animals caught.Espigares et al. (2023) proposed that this scenario would explain the high abundance of skeletal remains of megaherbivores, particularly elephants, in the UAL of FN3, and also the high density of hyena coprolites, because the half-sunken carcasses of the animals that regularly got stuck in the muddy sands would have attracted the scavenging hyenas.
Another way of uidization for a quicksand can be vibrations from an earthquake (Kadau et al., 2009).Of interest here, paleo-seismites are frequently recorded in the Orce sediments and throughout the whole Baza Basin, which shows that this region was tectonically and seismically active during the Quaternary (Alfaro et al., 1997(Alfaro et al., , 2010;;Gibert et al., 2005; García-Aguilar et al., 2014).In fact, the dome-shaped surface of the limestone layers of FN3 can be interpreted as resulting from earthquakes (see Fig. 3).
Most research on quicksand focuses on the danger it poses to man-made structures (Craig, 1997

Objectives
The two main goals of this study are: (i) to test for statistical differences in the composition of the fossil assemblages of large mammals preserved in the LAL and UAL of FN3; and (ii  1) for those performed during the years 2017-2020.Species abundances were recorded as numbers of identi able specimens (NISP) and estimates of minimum numbers of individuals (MNI).The latter were distributed among two major age categories, juveniles with milk teeth and adults with permanent dentition.Statistical comparison between the frequencies of taxa and size/age categories in the LAL and UAL were based on the z-test (Lawley, 1938).
Three samples of sediment were taken from the FN3 site for analyzing particle size.One sample is from layer 5 of the UAL, which is homogeneous in grain size, color and texture (Figs.2B-C).As indicated above, this layer preserves a huge accumulation of skeletal remains of megaherbivores and hyena coprolites (see Espigares et al., 2023).The other two samples are from layers 2-3 of the LAL, in which other herbivores are comparatively more abundant and manuports are better represented.Of the latter samples, the rst corresponds to the bottom of layer 2, which sediment shows a light tone (Fig. 2C), and is hereafter referred to as sample 2. The second sample was taken at the top of layer 2, which has a darker tone.Given that the upper part of layer 2 shows an irregular bedding surface due to soft-sediment deformation (Fig. 2C), the sample also includes sediments from the bottom of layer 3.For this reason, this sample is hereafter referred to as sample 3.
Part of the sediment contents of the three samples studied were washed using 0.710 mm, 0.500 mm and 0.125-mm mesh sieves.After drying the fractions recovered, an Olympus SZH-10 binocular magni er was used in order to recognize ostracod and mollusk remains.Scanning Electron Microscopy (SEM) images of microfauna were taken using a JEOL JSM-6490LV device from 'Servicios Centrales de Apoyo a la Investigación' (SCAI, University of Málaga, Spain), equipped with W lament.
Particle size was also measured in the samples by sieving and laser diffraction.In this case, sieve aperture ranged from − 1φ to 4φ.Larger sieve sizes were omitted due to the huge amount of sample that would have been required to ensure a representative analysis.Particle size analysis of the silt and clay fractions was performed using a Beckman Coulter LS13 320 Laser Diffraction Particle Size Analyzer.Particles were classi ed among size classes following Blott and Pye (2012) and data were analyzed using a custom GRADISTAT sheet (Blott and Pye, 2001), modi ed to calculate the porosity following the formula of Wu and Wang (2006).These analyses were performed at the Geology Laboratory of 'Centro Nacional de Investigación de la Evolución Humana' (CENIEH, Burgos, Spain).

Differences between the large mammal assemblages of the LAL and UAL
Table 1 shows data on the composition of the bone assemblages unearthed from FN3, distributed among archaeological levels and excavation seasons.According to this information, the abundance of skeletal remains of megaherbivores (i.e., those species with an adult body mass of > 1000 kg, including elephant M. meridionalis, hippo H. antiquus and rhino S. hundsheimensis) relative to other large mammals (i.e., those species weighing > 5 kg) is higher in the UAL than in the LAL.Speci cally, megaherbivores represent 63.9% (1060/1658) of the NISP of large mammals in the UAL and only 17.0% (105/616) in the LAL.A z-test (Lawley, 1938) shows that this difference in frequencies is statistically highly signi cant (z = 19.879,p < 0.00001).
Similarly, megaherbivores account for 32.6% (28/86) of the MNI calculated for the large mammals in the UAL and for only 11.8% (6/51) in the LAL.This difference is also statistically signi cant, albeit to a lesser degree (z = 2.724, p = 0.0065) given that the MNI estimates are lower than the NISP counts.These data unequivocally con rm that the frequency of megaherbivores is about three times higher in the UAL than in the LAL, as  1) showed that most hyena coprolites preserved in FN3 come from the UAL (98.6%, 217 out of 220), while most manuports are from the LAL (87.7%, 329 out of 375) (see Fig. 4).Both differences are statistically highly signi cant (z = 20.404,p < 0.00001 for coprolites; z = 20.667,p < 0.00001 for manuports), which suggests that different site formation processes were involved in the two main archaeological levels of FN3.

Differences of grain size composition between the fertile layers of the LAL and UAL
There is a clear difference in grain size between layer 5 of the UAL, which shows a predominance of sands, and layers 2-3 of the LAL, in which ne-grained particles predominate (Fig. 5, Table 2).Speci cally, layer 5 shows an elevated proportion of ne and very ne sands (64.4%), with a mode at 94 µm, a very low frequency of medium, coarse, and very coarse sands (only 2.5%), and a moderate proportion of silt and clay (33.1%).In the case of samples 2-3, ne and very ne sands represent lower fractions (12.9% and 18.5%, respectively) than in layer 5, while silts and clays show much higher frequencies (84.9% and 79.4%, respectively).According to the grain size distribution of Table 2, ne silts are represented by a higher proportion in layer 2 than in layer 3, with a mode at 6 µm, and clay contents (29.5%) are also more abundant.In contrast, layer 3 has more ne sands and coarse silts, with modes at 187.5 and 24 µm, respectively, and somewhat lower clay contents (21.2%).These results show that the overall particle size distribution of sample 3 shows some similarities with that of layer 5, although it has much lower sand contents and a remarkably higher frequency of very ne-grained sediments.The stratigraphic section of FN3 not only shows signi cant differences in lithology and granulometry, but also in the invertebrate contents of the fertile layers of the LAL and UAL.Layers 2-3 of the LAL show differences in the composition of their ostracod and mollusk assemblages.Speci cally, layer 2 preserves shells of the ostracods Ilyocypris bradyi and I. gibba (Fig. 6A), as described by Anadón et al. (2003) in a study based on a more extensive sampling of this layer.Moreover, Anadón et al. ( 2003) also identi ed in this layer the presence of the ostracod Candona angulata.Shells of I. bradyi and I. gibba are also found in layer 3, as well as scarce remains of the ostracod Heterocypris incongruens (Fig. 6B).In contrast, the UAL shows a predominance of the ostracod Cyprideis torosa, which is represented by both smooth and node-reticulated valves (Fig. 6C).The latter valves are less abundant, at least as far as layer 5 is concerned.Anadón et al. (2003) also highlighted the presence in several stratigraphic layers of FN3 of the ostracods Heterocypris salina and Ilyocypris sp., as well as operculum remains of the freshwater snail Bithynia sp., which are recorded in low frequencies.Our study shows that these remains are associated in layer 5 to shell accumulations of the land snail Limax sp. (Fig. 6D).

Discussion
Espigares et al. (2023) suggested that the lithological differences between the LAL and UAL, the two main archaeological levels of FN3, could explain the differences in composition between their bone assemblages of large mammals.Speci cally, layers 2-3, the two fossiliferous layers of the LAL, are composed of green clays (layer 2) and brown-whitish clays with nodules (layer 3).In contrast, layer 5, which preserves most skeletal remains of the UAL, is composed of ne greenish sands (the greenish tone probably results from the clay contents of this layer) and marly mudstones (see also Oms et al., 2011).For this reason, Espigares et al.
(2023) proposed that the sediments of layer 5, which are mostly represented by ne sands and clays, (i) may have behaved like quicksand that became unstable when forced to move by the pressure exerted by very large ungulates; and (ii) the entrapped carcasses of these large-sized animals lured scavenging carnivores like the hyenas and hominins.In the case of hyenas, they regularly visited the site during this time interval and defecated in the surroundings of the carcasses of megafauna on which they fed, thus giving rise to a latrine of P. brevirostris in an open-air setting (Espigares et al., 2023).
Our statistical comparison of the bone assemblages preserved in the LAL and UAL of FN3 shows that the abundance of megaherbivores relative to other large mammals is between 2.8 and 3.7 times higher in layer 5 than in layers 2-3, depending on whether the comparison is based on MNI estimates or NISP counts (Table 1), respectively.In addition, megaherbivores are represented in layer 5 by juveniles in a proportion 2.5 times higher than that calculated for other ungulates.
Our grain size analysis (Fig. 5, Table 2) con rms that layer 5 of the UAL shows a predominance of very ne and ne sands, which account for nearly two-thirds of the sediments that make up this layer, and that silt and clay represent most of the remaining fraction.In contrast, layers 2-3 of the LAL are mainly composed of silt and clay, showing < 20% of sand contents.All these data tentatively support the hypothesis of Espigares et al.
(2023) that layer 5 of FN3 could have represented a death trap for megafauna in quicksand.
The micropaleontological contents of the fossiliferous layers indicate that the faunal assemblage of layers 2-3 of the LAL consists of ostracods that lived in a permanent lacustrine environment with ephemeral fresh to oligosaline waters (Fuhrmann, (Fuhrmann, 2012), while the presence of the land snail Limax sp.suggests the subaerial emersion of these sediments.Therefore, compared to the situation depicted in layers 2-3 of the LAL, the mixture of faunas and the shell concentrations of Limax sp. in the UAL suggest a periodic drop of the lake waters, which would led to occasional emersions.This would imply that the sediments of the UAL remained saturated in water with moderate to high salinity, as particularly evidenced in the case of layer 5, which is relevant for our interpretation of this layer as a potential quicksand trap for megaherbivores, as discussed below.

The upper archaeological level of FN3 as a natural trap for megaherbivores
As noted before, experimental data on quicksand composed of ne sand, clay and salt water from salt-lake environments similar to the oligosaline lacustrine systems of the Baza Basin (García-Aguilar et al., 2014, 2015; Palmqvist et al. 2022a) show that the quicksand can trap animals wandering on its surface.However, the objects with a density of 1 g/ml (i.e., the average density of mammalian tissues) are not completely sucked beneath the surface but sink halfway into the quicksand (Khaldoun et al., 2005(Khaldoun et al., , 2006)).Therefore, a quicksand scenario would explain the entrapment of megaherbivores in the UAL of FN3 and the high numbers of coprolites unearthed from this archaeological level (Fig. 4), because the partly sunk carcasses of very large animals that regularly became entrapped in the muddy sands of layer 5 would have attracted the hyenas, which fed on these carcasses and defecated in their surroundings (Espigares et al. 2023).
Whether or not a given sandy sediment behaves like a quicksand depends on its lithology (e.g., granulometry and sand composition) and also on a number of factors (e.g., seismic movements, porosity, water saturation, water salinity, and pressure) of which some do not fossilize.Given these constraints, the interpretation of a paleo-quicksand relies on sedimentological data but must also be corroborated by other sources of evidence, The clay-size fractions of samples from layers 2-3 of the LAL (Fig. 5, Table 2) are well above the limit (10%) for classifying them as quicksand.In contrast, the sample from layer 5 of the UAL shows a low clay content of < 10%, being primarily composed of ne and very ne sands.Therefore, this layer satis es the rst and second constraints, being a good candidate to be considered as a potential paleo-quicksand.-C).After ~ 1.4 Ma of diagenetic compaction, layer 5 ranges in thickness between 10 and 30 cm.Considering the porosity calculated for this layer (48.6%; see Table 2), it would be expected that the original thickness was at least 25% greater (Lundegard, 1992), an estimate that would satisfy the requirement of a sediment thickness of > 25 cm over most of the lateral extent of layer 5.
The coe cient of uniformity (determined by the ratio D10:D60) calculated for the sample from layer 5 (37.2) is higher than the maximum limit for a quicksand, which is set at 5. Layer 5 primarily consists of ne sands showing a mode of 94 µm (Table 2).These sands were deposited by a low-energy ow, as evidenced by the erosive bottom of layer 5 and the lenticular shape observed in the stratigraphic section (Fig. 2).Layer 5 also has a silt and clay fraction in excess of that expected for a quicksand.In contrast, our fraction of ne sands in the sample from layer 5 could correspond to the mostly sandy sublayers located immediately below a partial skeleton of the proboscidean M. meridionalis (microfacies 2 and 3 of Yravedra et al., 2024).In this way, the coe cient of uniformity of these sandy sublayers with low silt and clay contents would be compatible with a quicksand.
Another possible explanation for the presence of ne-grained sediments in layer 5 is that they could have entered in this layer after the original accumulation of the deposit, which was followed by soft-sediment deformation and diagenesis.In fact, layer 6 consists of mudstones (Oms et al., 2011) and has similar characteristics to layer 3, with analogous proportions of silts and clays.This opens the possibility that the nest sediments of layer 5 come from layer 6, the overlying stratum, by edaphic processes that lixiviated the ne-grained sediments of layer 6 and percolated them into layer 5.In addition, the silts and clays of layer 5 (Fig. 5, Table 2) are consistent with the grain size distribution of the global aeolian dust (Lawrence and Neff, 2009), which means that aeolian transport cannot be ruled out as a source of sediment.Of interest here, the Iberian Peninsula continuously receives African dust, which is deposited in the Mediterranean soils (Bout-Roumazeilles et al., 2007; Escudero et al., 2011).Loess deposits consisting of coarse and ne silts and clays have been identi ed in the Granada Basin from the Pliocene to the present (Günster et al., 2001;Wagner et al., 2015).Therefore, we can assume that aeolian sediments were also deposited on the surface of layer 5, which would also potentially account for the most ne-grained fraction of this layer.
Post-depositional processes such as pedogenesis and diagenesis are another potential source of ne-grained sediments.Primary soil-forming mechanisms for Mediterranean soils include clay illuviation, carbonate redistribution and aeolian inputs (Yaalon, 1997).These processes, coupled with the weathering of siliciclastic sediments, generate ne-grained sediments.In fact, the pedogenetic processes observed in the Granada Basin result in an increase in the abundance of ne-grained sediments at the expense of the coarser fraction (Günster et al., 2001;Wagner et al., 2015).Sand diagenesis can lead to mineral compaction, dissolution and reprecipitation, which results in a decrease in particle size distribution as smaller minerals are precipitated replacing the original ones (Mar l and De la Peña, 1989).These processes would increase the proportion of ne-grained sediments in layer 5 after the burial of the stratum.If we tentatively assume that there were no ne-grained sediments during the subaerial exposure of layer 5, the coe cient of uniformity of the sand would then be 1.59, a value in agreement with the limit of < 5 required for a quicksand.
Porosity was calculated assuming a unimodal distribution.Layer 5 is near to unimodal distribution, with a mode in 94 µm and a porosity of 48.6% (Fig. 5, Table 2).If the silt and clay fractions are considered, porosity may vary to some degree.However, the nest sediments may have been generated later due to pedogenetic and diagenetic processes, as discussed before.
The ostracod assemblage of layer 5 suggests oligosaline conditions in the lake waters, with periodic variations of salinity over time.The shells of C. torosa are indicative of a shallow lacustrine environment near a coastal zone (Anadón et al., 1994;Granados et al., 2021), which was in uenced by the arrival of ne-grained detrital sediments.These inputs of freshwaters are evidenced by the presence of Ilyocypris.The shell accumulations of terrestrial gastropod Limax sp.indicate periods of subaerial exposure in this area, which would have allowed the transit of megafauna and their predators from the surrounding environments.Therefore, the sediments of layer 5 in the UAL were likely in uenced by brackish waters, which salt contents favored the development of a quicksand trap, as evidenced by the experiments of Khaldoun et al. (2005).
The constraints related to water contents and permeability [i.e., those addressed above in points (vi), (vii) and (viii)] are not recorded in the stratigraphic sequence, but they can be inferred from the paleoenvironment of FN3, which is interpreted as a lake margin subject to small, low-energy ows.This lake may have experienced periods of ooding and recession, which would cause the lake margin to become a water-saturated soil, in agreement with the environmental reconstruction of Espigares et al. (2013Espigares et al. ( , 2019)), Yravedra et al. (2021Yravedra et al. ( , 2024) ) and Palmqvist et al. (2023a).Layer 4 is a limestone bed (Oms et al., 2011, Espigares et al., 2023) and these limestones may have acted as a low permeability layer, which may have eventually led to a poor drainage of layer 5. Thus, it is likely that layer 5 became saturated with oligosaline waters due to insu cient drainage, which satis es conditions (vi), (vii) and (viii).
In summary, both the sedimentological data and the paleoenvironmental reconstruction of layer 5 indicate that it may be interpreted as a paleo-quicksand deposit that may have functioned under certain conditions as a natural trap for very large animals, which would explain the high frequency of megaherbivores recorded at the UAL of FN3.

Entrapment of megaherbivores in the upper archaeological level of FN3
Mud entrapment is a relatively common phenomenon today in Africa, because elephants and other large mammals venture during drought episodes into drying dams in their quest for drinking water and taking mud baths.Giant naked-skinned animals such as elephants, rhinoceroses and hippos usually wallow in mud to thermoregulate and protect themselves from insect bites and parasites.However, their huge size poses a risk of being mired in the mud. Figure 7 depicts one of these entrapment events, in which a mother elephant and her calf became stuck in a muddy waterhole during the height of the 2019 dry season in Mana Pools National Park, Zimbabwe.Eventually, the calf was eaten alive by a pack of spotted hyenas (Crocuta crocuta) and the mother succumbed days later, probably due to dehydration.
Increased exposure of megaherbivores to the risk of entrapment in quicksand results from the high weight per unit area supported by their limbs.Given that the three species of megaherbivores recorded at FN3 are extinct, there are no direct measurements on the ground support surface of their feet.However, reliable estimates can be retrieved from their modern analogues and also from fossil footprints (although it should be noted that footprint dimensions can overestimate or underestimate the actual foot dimensions; for review and references, see Masao et al., 2016).
In the case of elephant Mammuthus meridionalis, Pasenko (2017) provided useful data on body mass and footprint diameters of the forefeet and hindfeet for a number of extant African (Loxodonta africana) and Asian (Elephas maximus) elephants, which are useful for estimating the weight per unit area supported by their feet.For example, the average footprint length and width of the manus (forefoot), 48.5 x 38 cm, and pes (hindfoot), 47 x 30 cm, of an adult male of L. africana of 4899 kg allow estimating the area of the footprint ellipses of its manus (1432.6 cm 2 ) and pes (1083.9cm 2 ), which results in a ground support surface for all four feet of 5033 cm 2 .Considering the body mass of this individual, the weight per unit area supported by its feet would be 0.97 kg/cm 2 .Following this procedure, an adult female of E. maximus of 3962 kg shows average length and width diameters of the manus (42 x 38 cm) and pes (39 x 28 cm) that result in a ground support surface for all four feet of 4180.4 cm 2 , which provides a similar estimate of weight per unit area, 0.95 kg/m 2 .In a ve-years old male of L. africana of 1515 kg, the manus and pes diameters (34.5 x 24.5 cm and 35 x 21 cm, respectively) allow estimating a ground support surface for all four feet of 2350 cm 2 , which results in a lower weight per unit area, 0.64 kg/m 2 .Finally, the manus (31 x 22.5 cm) and pes (29.5 x 19.5 cm) diameters of the footprints of a 1.5-years juvenile male of L. africana of 726 kg translate in a ground support surface for all four feet of 1929.7 cm 2 , which corresponds to a weight per unit area of only 0.38 kg/m 2 .These measurements of footprints for extant elephants agree with those reported by Altamura et al. (2023) for fossil footprints produced by young and adult individuals of the straight-tusked elephant Palaeoloxodon antiquus in the Middle Pleistocene site of Schöningen (Germany), dated to ~ 321 ka, and also with the footprint tracks of this species described by Neto de Carvalho et al. (2021) in the Late Pleistocene (MIS 5) site known as the Matalascañas Trampled Surface in Huelva (Spain).In the latter site, 34 footprints attributed to newborns, calves, juveniles, and adults were measured (Neto de Carvalho et al., 2021: Table 1).The length and width diameters of these footprints range between 9.6 x 7.6 cm in calves of < 1 year and 54.5 x 31.6 cm in adults of > 60 years, values that are similar to those measured in extant elephants.If we consider the body masses calculated for these specimens by Neto de Carvalho et al. ( 2021), the estimates of ground support surface for all four feet would be comprised between 0.24 kg/m 2 in calves and 1.03 kg/m 2 in old adults, respectively.
In the case of the giant, short-limbed Hippopotamus antiquus (Palmqvist et al., 2022b), Behrensmeyer and Laporte (1981) reported on the nding of single footprints and trails of large vertebrates in a 12 cm thick unit of sandy muds (sand:silt:clay = 12:59:29), dated to 1.6-1.5 Ma and deposited in a marginal lacustrine environment of the Koobi Fora Formation (Kenia).Among these footprints, there are several deep, four-toed impressions of hippopotamus feet and a shallow depression resembling a human footprint.According to Behrensmeyer and Laporte (1981), the larger footprints (25-32 cm wide on average) can be attributed to the giant Hippopotamus gorgops, a huge extinct sister species with a size and anatomy similar to H. antiquus (~ 3200 kg; Palmqvist et al., 2022b), while the smaller ones (18-20 cm wide) may be from juveniles of this species or even belong to the fossil pigmy species, aff.Hippopotamus aethiopicus (Boisserie, 2005).The larger footprints, each describing an ellipse with an area of 628.3 cm 2 , result in a minimum ground support surface for all four feet of 2513.2 cm 2 .Given the mass of H. gorgops, this would provide a weight per unit area of 1.27 kg/cm 2 .Similarly, Bennet et al. (2014: Fig. 9D) gured one footprint attributed to H. gorgops from an ichnological surface close to Koobi Fora (GaJi10), which is dated to 1.4 Ma.The digitization of this footprint with ImageJ software provided an area of 814 cm 2 , which allows estimating a ground support surface for all four feet of 3256 cm 2 .This would in turn result in a weight per unit area of 0.98 kg/cm 2 .In contrast to H. gorgops, the weight per unit area is lower in the living H. amphibius, almost half of the estimate obtained for the giant, short-limbed extinct hippo.Speci cally, Altamura et al. ( 2017) published a tracksite presumably made by H. amphibius in the Middle Pleistocene site of Gombore II-2 (Melka Kunture, Upper Awash, Ethiopia), which is dated to ~ 0.7 Ma.Digitization with ImageJ of the best-preserved footprint (Altamura et al., 2017: Fig. 8B) provided an area of 693 cm 2 , which results in a ground support surface for all four feet of 2772 cm 2 .Given that the body mass of the modern common hippo (~ 1420 kg; Palmqvist et al., 2022b) is less than half that estimated for H. antiquus, this would translate in a weight per unit area of only 0.51 kg/cm 2 , which would represent a lower risk of entrapment for the living hippo compared to the giant extinct hippo.
Concerning the extinct rhino Stephanorhinus hundsheimensis, Leakey and Hay (1979) provided average measurements for the forefoot and hindfoot footprints of two fossil rhinos, the black rhino Diceros bicornis (~ 1100 kg; Mendoza et al., 2006), and the white rhino Ceratotherium simum (~ 1600 kg; Mendoza et al., 2006).These fossil footprints are preserved in the Late Pliocene Upper Laetolil Beds (Tanzania), which are dated to 3.85 − 3.63 Ma (Deino, 2010).In the case of the footprints attributed to D. bicornis, the average dimensions of the forefoot (24.6 x 24.8 cm) and hindfoot (25.6 x 22.8 cm) translate in a ground support surface for all four feet of 1734.2 cm 2 , which results in a weight per unit area of 0.63 kg/cm 2 .The larger trails, attributed by Leakey and Hay (1979) to C. simum, have average dimensions of the forefoot (28.5 x 31 cm) and hindfoot ( 40x 27 cm) that allow estimating a ground support surface for all four feet of 3084.2 cm 2 .These estimates would in turn result in a weight per unit area of 0.52 kg/cm 2 .Footprint area values for modern black rhino in Mwabvi Game Reserve can be derived from average diameters of forefeet (23 cm) and hindfeet (19.7 cm) provided by Jachmann (2001: Table 10.1; the last individual track was excluded because it was probably made by a juvenile).These measurements allow estimating a ground support surface for all four feet of 1440.6 cm 2 , which results in a weight per unit area of 0.76 kg/cm 2 .
In summary, the elevated weight per unit area supported by the feet of the largest megaherbivores of FN3, which would uctuate between 0.38 kg/cm 2 in the juvenile elephants and 1.27 kg/cm 2 in the adult giant hippos, would represent an increased risk of entrapment in quicksand compared to other ungulates of medium-to-large size.
In contrast to the megaherbivores, the weight per unit area supported by the feet of the two main scavenging carnivores at FN3, the giant hyena Pachycrocuta brevirostris and Homo sp.(Espigares et al., 2013(Espigares et al., , 2019;;Palmqvist et al., 2011Palmqvist et al., , 2023a)), would be lower, which means a lower risk of being stuck in the quicksand.In the case of the hyena, the average size of the footprints measured by Leakey and Hay (1979) in four relatively long trails of the Laetolil Beds attributed to Hyaena bellax (a sister taxon of P. brevirostris; Palmqvist et al., 2011), 11.5 x 9.6 cm, translates in a support surface for all four feet of 442 cm 2 .If we consider the body mass estimated for P. brevirostris, ~ 110 kg (Palmqvist et al., 2011), this results in a weight per unit area of 0.25 kg/cm 2 .It is worth noting, however, that the European specimens of P. brevirostris show a larger size than their African counterparts, which means that their footprints were probably greater (and their weight per unit area lower) than these estimates.Concerning Homo sp., the dimensions (26 x 10 cm) of the human footprints of Koobi Fora reported by Behrensmeyer and Laporte (1981)  The above data refer to the weight per unit area supported by the feet of the animals under static conditions.However, the pressure exerted by the limbs increases when the animals are moving, depending on their speed.Speci cally, the kinetic energy depends on the body mass (BM) of the animal and its velocity (v): Considering a moderate velocity of 3.07 m/s (Ren and Hutchinson, 2008) for the adult male of L. africana of 4899 kg cited before, this provides an E k estimate of 23,086.3J.This in turn results in an increase of pressure of 45,870 Pa.Given the ground support surface measured for all four feet in this specimen (5033 cm 2 ), this translates in an additional pressure of 0.47 kg/cm 2 .If we sum this value to the static pressure, the weight per unit area supported by the elephant's feet under kinetic motion would be 1.44 kg/cm 2 (i.e.50% higher than our previous estimate under static conditions).A speed of 5.56 m/s (20 km/h) would result in an increase of 1.53 kg/cm 2 , which provides a weight per unit area of 2.5 kg/cm 2 (i.e., 1.5 times higher than the estimate for static conditions).Similarly, the increase of pressure for a speed of 3.07 m/s would be 0.612 kg/cm 2 in H. antiquus, which would result in a weight per unit area of ground support of 1.88 kg/cm 2 .In contrast, the pressure under kinetic motion at this speed would rise in P. brevirostris and Homo sp. to 0.37 and 0.19 kg/cm 2 , respectively, values that are both well below the penetration hardness of a quicksand (1-3 kg/cm 2 according to Nguyen et al., 2016: Fig. 6).These data suggest that while adult megaherbivores, particularly elephants and hippos, could occasionally become trapped in the quicksand of layer 5 (Fig. 8), with the juveniles remaining in the vicinity of their mothers, the hyenas and hominins would have been much less exposed to this risk.For this reason, they could have scavenged the half-sunken carcasses of the megaherbivores that became stuck in quicksand.Interestingly, this is documented by a partial skeleton of an old female elephant in the UAL of FN3 (Fig. 9A).The skeleton preserves the mandible, vertebral column, ribs, and pelvis in anatomical connection, and is partly surrounded by akes and hyena coprolites (Espigares et al., 2013(Espigares et al., , 2023)).A similar scenario has been described for the remains of a Late Pleistocene American proboscidean (Mammut americanum) preserved in peaty silt and clay underlying till in a gypsum karst near Hillsborough, New Brunswick (Harington el al., 1993).In this case, nearly all of the postcranial elements on the right side of the body were preserved, whereas much the bones from the left side were missing or broken, except for the teeth and tusks.This suggests that this megaherbivore eventually sank deeply into the mud on its right side, leaving its exposed left side to be destroyed by the scavengers (Harington el al., 1993).

Hyenas and hominins in the upper archaeological level of FN3
As noted before, fossil remains of large mammals and lithic tools are abundantly preserved in layers 2-3 of the LAL and layer 5 of the UAL (Fig. 4).Moreover, although cut marks are relatively scarce, they are equally represented in the LAL and UAL (Espigares et al., 2019: Table S12).However, manuports are more abundantly recorded in the LAL, while hyena coprolites and skeletal remains of megaherbivores, particularly elephants, are concentrated in the UAL (Espigares et al., 2023: Table 1).Carnivore tooth marks also predominate in this archaeological level (Yravedra et al., 2021: Table 5).These data suggest that although there is evidence in support of the involvement of hominins and hyenas in both archaeological levels, hominins exploited more intensively the resources available at the LAL, while hyenas concentrated their activity on the carcasses of megafauna at the UAL, for which sedimentological evidence suggests a quicksand trap as the most plausible sedimentary environment.In fact, the high frequency of coprolites in the UAL evidences that this area was frequently occupied by the hyenas during the period of time in which the UAL was formed.This suggests a partial ecological segregation between hyenas, which had more nocturnal and twilight habits, and hominins, who were predominantly diurnal.
Three scenarios have been discussed for carcass acquisition by the FN3 hominins (see review in Palmqvist et al., 2023a): (i) that the hominins had primary access to ungulate carcasses by hunting prey on their own; (ii) that they had early access to the prey of sabertooth cats and other primary predators after engaging them in confrontational scavenging (i.e., kleptoparasitism); or (iii) that they were able to locate the carcasses of animals dead by various causes (including predation) before other scavengers did.
No direct method allows to test if early Homo had primary access to ungulate carcasses by hunting or by early scavenging (Potts, 1991).However, the limited technological skills of the FN3 hominins make it di cult to argue that they could have hunted medium-to-large, large and very large prey on a regular basis: the small size of the int akes unearthed from FN3, few centimeters in most cases (Turq et al., 1996;Martínez-Navarro et al., 1997;Toro-Moyano et al., 2011;Espigares et al., 2013Espigares et al., , 2019)), suggests the lack of an effective weaponry for killing large ungulate prey (Palmqvist et al., 2023a), which makes it di cult to conceive how these hominins could have had a direct impact on megafauna hunting large-sized animals on their own (for a review on the ecology of Oldowan hominin carnivorous activity, see Blumenshine and Pobiner, 2007).
The second scenario, confrontational scavenging, is also unlikely because it would be dangerous for the hominins (Treves and Palmqvist, 2007).However, it must be noted that modern African agro-pastoralists successfully use rudimentary weapons like simple stones or long sticks to drive away lions, leopards and cheetahs from their kills and steal them (Treves and Naughton-Treeves, 1999).Moreover, it has been suggested that stone-throwing was an early adaptive strategy of Homo, which would explain the huge number of allochthonous cobbles found in the Early Pleistocene site of Dmanisi (Lordkipanidze, 2015;Coil et al., 2020).
In the case of FN3, bone remains are also intermingled with abundant limestone cobbles (manuports), particularly in the LAL, in which hyena coprolites are very scarce (Fig. 4).This opens the possibility that the hominins transported the cobbles to the site and used them as weapons (i.e., as " ying stones") for forcing predators like the sabertooth cats and wild dogs (or other scavengers like the hyenas, jackals, and vultures) to surrender the prey carcass (Espigares et al., 2013(Espigares et al., , 2023)).These cobbles could also be used later as percussion tools for fracturing bones and accessing their medullary cavities, as suggested by the nding of percussion marks in half of these manuports (Barsky et al., 2015(Barsky et al., , 2016(Barsky et al., , 2022).
The third possibility, i.e. that the hominins had primary access to intact carcasses of animals that died through natural causes (e.g., disease and starvation) or secondary access to the prey leftovers abandoned by primary predators, seems the most reasonable given the anatomy of early Homo as well as their apparently limited technological skills (Espigares et al., 2019;Palmqvist et al., 2023a).Apart from vultures, the giant hyena P. brevirostris and the jackal-sized dog C. mosbachensis would indeed be worthy competitors of the FN3 hominins in the access to carrion (Palmqvist et al., 2011(Palmqvist et al., , 2023b;;Iannucci et al., 2021).Modern hyenas and jackals often rely upon visual clues such as circling vultures to identify scavengeable resources, and then run long distances to secure the carcass (Bramble and Liberman, 2004).Compared with these terrestrial carnivores, which patterns of activity are more twilight and nocturnal, the hominins were likely predominantly diurnal.This means that the information provided by the aerial scavengers was probably more helpful for the early members of Homo than for other terrestrial scavengers.Given that hominins and vultures occupied the same habitats and scavenged the same medium-to-large, large and very large mammals, they likely competed for access to carcasses alongside other terrestrial scavengers like hyenas and jackals (Morelli et al., 2015).
Vultures are such e cient scavengers that in some areas of Africa they consume more meat than all other carnivores combined (Ogada et al., 2012), which explains their high diversity.For this reason, it has been suggested that vultures used as beacons for meat were particularly important to the hominins that dispersed out of Africa, facilitating their occupation of new landscapes when they arrived in Europe (see review in Morelli et al., 2015).
According to Liesenberg (2006), endurance or persistence running (i.e., the ability to run many kilometers at relatively low speed over extended time periods using aerobic metabolism) could have been an adaptation of hominins for tracking down and chasing to exhaustion small-to-medium sized ungulates.The reason is that although humans are comparatively poor sprinters, they perform well at running long distances due to several anatomical and physiological adaptations.These include (i) a long tibia, which results in a long stride length combined with the presence of long spring-like tendons connected to short muscle fascicles, which economically generate substantial force; (ii) a long plantar arch and enlarged articular surface areas in most of the joints of the lower body, which dissipate the high impact loads generated in running; (iii) derived features that enhance trunk stabilization, such as an expanded area on the sacrum for attachment of enlarged gluteus muscles; and (iv) thermoregulatory adaptations for heat dissipation that help to maintain stable body temperature, including a multiplication of eccrine sweat glands for evapotranspiration, reduced body hair, mouth breathing, and an elaborated cranial venous circulation for brain cooling (see reviews in Bramble and Liberman, 2004;Liesenberg, 2006;Ruxton and Wilkinson, 2012).Endurance running evolved early in the genus Homo, as shown by the modern human-like body proportions and lower limb morphology of the Dmanisi hominins, which are indicative of the ability for long-distance traveling (Lordkipanidze et al., 2007).
The human adaptations discussed above for persistence hunting of small-to-medium sized prey are equally useful for scavenging carcasses of large and very large animals in open habitats during the day, particularly in the dry season, when other terrestrial scavengers like hyenas and jackals avoid running long distances due to thermoregulatory constraints (Bramble and Liberman, 2004;Liesenberg, 2006;Lieberman et al., 2007;Ruxton and Wilkinson, 2012).For this reason, endurance running may have been the key to hominin success in the access to carcasses of large-sized ungulates at FN3 before eventually surrendering them on the arrival of potentially dangerous hyenas or other terrestrial scavengers (Palmqvist et al., 2023a).This scenario is suggested by the nding of a partial skeleton of elephant M. meridionalis in the UAL (Fig. 9A), which is partially surrounded by akes and hyena coprolites (Espigares et al., 2013).Although Yravedra et al. (2021) argued that no tooth marks or anthropogenic marks were observed on the elephantid bones, thus not conclusively proving the interaction of humans or carnivores with the proboscidean carcass (however, for a new interpretation see Yravedra et al., 2024), the spatial association of lithic tools and coprolites with the elephant carcass did not result from independent episodes that coalesced into a palimpsest (Palmqvist et al., 2023a): >95% of the bones unearthed from FN3 preserve intact their cortical surface, showing weathering stage 0 (Yravedra et al., 2021), which evidences a very short time of exposure before burial, less than one year according to the studies of weathering rates by Behrensmeyer (1978) in Amboseli National Park, Kenya.According to Espigares et al. (2013), this tentatively suggests that the hominids were the rst to scavenge the elephant carcass, dismembering and transporting its limbs for later consumption in a safer place, and that the hyenas arrived later and consumed the rest of the carcass, defecating in its surroundings.
All these arguments apply to the bone assemblage accumulated in the UAL of FN3.However, the situation depicted at the LAL was quite different.For this reason, it is worth considering the insights on the role played by hominins and hyenas provided by Palmqvist et al. (2023a), who studied the patterns of bone survival in FN3 and compared them with those detected in the hyena den of VM (Arribas and Palmqvist, 1998).Speci cally, Palmqvist et al. (2023a) showed that in both sites there is: (i) a direct relationship between the abundance of skeletal remains of large mammals and their estimated mineral densities; and (ii) an inverse relationship between the survival of major limb bone epiphyses and their estimated marrow yields.However, FN3 shows considerable scatter around both regression lines, which renders these relationships statistically less signi cant than in VM.According to Palmqvist et al. (2023a), these results suggest that exploitation by the hominins of bone marrow after hammerstone breakage was a usual activity during carcass processing at FN3.As a result, few bones retained their marrow contents after the exploitation of marrow by hominins, which made them less attractive for hyenas.This explains the lower frequency of tooth-marked bones in FN3 compared to VM, where the bone-cracking agency was the hyena P. brevirostris (the frequency of tooth-marked bones in VM3, the main excavation quarry of the VM site, is 29.3%;Palmqvist et al., 2022b).The lower selectivity in bone fracturing evidenced at FN3 probably results from the fact that hominins fractured the bones assisted by their limestone manuports.This suggests that the mineral density of the skeletal elements was not as much of a problem for breaking them in the case of hominins as it was in the case of hyenas, which performed this task using their robust jaws and enlarged premolar teeth (Arribas and Palmqvist, 1998;Palmqvist et al., 2011Palmqvist et al., , 2022b)).This forced the hyenas to be more selective than the hominins when choosing which skeletal elements to fracture, in order to avoid breaking their teeth or dislocating their jaws when fracturing the densest bones (Palmqvist et al., 2023a).Palmqvist et al. (2023a) also calculated an index of skeletal completeness that allows estimating ravaging intensity in a skeletal assemblage, the ratio of axial bones to appendicular bones.The value of this ratio in FN3 (0.44) is higher than in VM (0.13), which indicates that the bone assemblage of VM was comparatively more ravaged due to selective transport of the skeletal remains by the hyenas.According to Palmqvist et al. (2023a), the living hyenas do not cooperate in the transportation of large portions of a carcass, which limits the individuals in what they can move and forces them to be more selective.The analysis of skeletal representation for ungulate taxa in VM showed that hyenas selectively transported herbivore carcasses and body parts to their maternity den as a function of the mass of the ungulates scavenged (Palmqvist and Arribas, 2001;Palmqvist et al. 2011): small-to-medium sized ungulates were transported as whole carcasses to the denning area while large and very large-sized species were dismembered and the hyenas preferentially transported the limbs that provided higher marrow yields.This resulted in a low ratio of axial elements to limb bones at VM.In contrast, the higher value of this ratio at FN3 suggests that the carcasses of the ungulates scavenged were less biased by selective transport, which points to cooperative transport of these ungulate carcasses by the hominins (Palmqvist et al., 2023a).
However, evidence of hominin activity concentrates in the LAL, as evidenced by the abundance of both Oldowan tools and manuports, while most hyena coprolites are found in the UAL, which suggests that the FN3 area was frequently occupied by hyenas during the period of time in which the sediments of layer 5 accumulated.This layer, whose lithology agrees with the expectations from a quicksand, preserves abundant remains of megafauna, particularly young elephants, which suggests the frequent entrapment of large-sized animals in the sediments of layer 5. Given that the resources of these carcasses would remain accessible to the scavengers for longer, the giant hyenas could have made greater use of them, which explains the concentration of coprolites in the UAL of the FN3 site.

Declarations
Compliance with Ethical Standards Disclosure of potential con icts of interest: On behalf of all authors, the corresponding author states that there is no con ict of interest.Abundance (Z-axis) in the two archaeological levels of the late Early Pleistocene site of Fuente Nueva-3 (FN3) of fossil remains (mostly large mammals; see Table 1), lithic tools ( akes, debris, cobbles, and debitage), manuports (unmodi ed limestone blocks intentionally brought to the site by the hominins), hyena coprolites, and skeletal remains of elephant Mammuthus meridionalis (data from Espigares et al., 2023: Fig. 12 and Table 1).The limit of the layers included in the upper archaeological level (UAL, layers 4-7) and lower archaeological level (LAL, layers 2-3) is not depicted in this gure, because these layers show irregular bedding surfaces (see Fig. 2).analyzed in this study (data from Table 2).Sample 5 is from layer 5 of the upper archaeological level (UAL).
Samples 2-3 are from the lower archaeological level (LAL).Speci cally, sample 2 was taken at the bottom of layer 2, which sediments show a light tone, while sample 3 is from the top of layer 2, which sediments show a dark tone, and from the bottom of layer 3.  Two elephants (Loxodonta africana), mother and calf, got stuck in the mud while searching for water during the 2019 dry season in Mana Pools Natural Park, Zimbabwe.Soon after, spotted hyenas (Crocuta crocuta) came and roamed around them (A).The calf, which could barely breathe by sticking its trunk out of the mud (B), was eaten alive that night to the helplessness of its mother (C).The hyenas fed on the carcass until the next morning (D), dismembered it (E) and carried the limbs to be consumed at a distance (F).The vultures (hooded vulture, Necrosyrtes monachus, in G; white-backed vulture, Gyps africanus, in H) fed on the remains of the calf carcass while the mother elephant tried to chase them away by throwing mud (G).Finally, the vultures landed on the mother's body, waiting for her death (H).The adult elephant was rescued from the mud with a crane but did not survive stress and dehydration.Montage made by the authors with photographs taken by Jens Cullmann (https://jenscullmann.de),reproduced with permission of the author.
Entrapment of an elephant in the quicksand of layer 5 of the upper archaeological level of the late Early Pleistocene site of Fuente Nueva-3.Abbreviations: E k , kinetic energy; W k , impact transfer work, in J; W lift , mechanical work required for the upward thrust of the limbs, in J; Q fr , friction coe cient; P (-) , negative pressure.Elephant silhouettes were obtained with an AI image generator (https://chat.openai.com/g/g-tvADFu1PW-generador-de-imagenes).).The mandible, vertebral column, ribs, and pelvis were all preserved in anatomical connection, although some ribs and vertebrae were slightly displaced by post-depositional processes.The elephant skeleton is partially surrounded by 34 hyena coprolites and 17 akes.The tone of these coprolites, which are darker than other unearthed from layer 5, evidences that the hyenas consumed large amounts of the elephant's meat and fat.The coprolites and akes partially overlap in their distribution, suggesting competition between the hominins and hyenas for the elephant carcass (see discussion in Espigares et al., 2013Espigares et al., , 2023;;Palmqvist et al., 2023a; for an opposite interpretation, see Yravedra et al., 2024).B: Adult female elephant and her two young caught in a mud trap in Kenya (photograph taken by Kieran Avery, reproduced with permission of the Sheldrick Wildlife Trust).
License:   This work is licensed under a Creative Commons Attribution 4.0 International License.Read Full License Version of Record: A version of this preprint was published at Journal of Iberian Geology on May 26th, 2024.
to analyze particle size in the main fertile layers of the LAL (layers 2-3) and UAL (layer 5), which preserve most skeletal remains of large mammals unearthed from the FN3 site, in order to test the hypothesis of entrapment of megaherbivores in quicksand proposed by Espigares et al. (2023) for layer 5 of the UAL. 2 Material and Methods Published data on the composition of the vertebrate assemblages unearthed from the two main archaeological levels of the FN3 site were compiled from Espigares et al. (2019: Tables S7-S8) for the excavation seasons of the years 1999-2015, and from Yravedra et al. (2021: Table taphonomic and tectonic factors.According to current studies on quicksand (SUCTC, 2012; Gao et al., 2018), a substrate should show the following properties for considering it as a potential quicksand: (i) a predominance of ne sands and silts; (ii) a clay-size fraction of < 10%; (iii) a thickness of > 0.25 m; (iv) a coe cient of uniformity of < 5; (v) a porosity of > 43%; (vi) water contents of > 30%; (vii) a soil in a saturated state; and (viii) a low permeability and poor drainage.

Layer 5
shows an uneven thickness due to the erosion and levelling of the irregular top of layer 4 (Oms et al., 2011; Espigares et al., 2023), and is affected by soft-sediment deformation (Figs.2B In a recent study,Yravedra et al. (2024) indicate the presence in layer 5 of ve sublayers with different microfacies, of which some show a predominance of sands while others are rich in silts and clays.Speci cally, the silt and clay fraction identi ed in our study of a sample of sediment from layer 5 (Table2) could correspond to the basal sublayer and upper sublayers described by Yravedra et al. (2024) in this stratigraphic level, which they characterize as sandy clays and mudstones (microfacies 1, 4 and 5 of Yravedra et al., 2024).
allow estimating a ground support surface for the two feet of 408.4 cm 2 .Considering a body mass of ~ 53 kg for Homo sp. in Orce (see Rodríguez-Gómez et al., 2016), this would result in a weight per unit area of only 0.13 kg/cm 2 .
Research involving Human Participants and/or Animals: This research involved no human participants and/or animals.Informed consent: All authors participated in this research, contributed to elaborating the manuscript contents and agreed in submitting it for publication in Journal of Iberian Geology.

Figures
Figures

Figure 1 A
Figure 1

Figure 7
Figure 7 Figure composed and edited by A. Perez-Ramos using Adobe illustrator.

Figure 9 A
Figure 9

Table 1
Vertebrate assemblage unearthed from the Fuente Nueva-3 (FN3) site during the systematic excavations of the years 1999-2015 (data from Espigares et al., 2019: Tables S7-S8) and 2017-2020 (data from Yravedra et al., 2021: Table 1) distributed according to archaeological levels.Abundances of large mammals are reported as numbers of identi able specimens (NISP) and estimates of minimum numbers of individuals [MNI, juveniles/adults (j/a)].
Espigares et al. (2023)et al., 2016;Sun, 2016).However, with the only exception ofHungerbühler (1998), who proposed a quicksand-like trap as an explanation for the nding of 17 partial skeletons of the Late Triassic prosauropod dinosaur Sellosaurus gracilis, as far as we know no paleo-quicksand has been reliably identi ed in the sedimentological record.For this reason, it is of great interest to test if layer 5 of the UAL of FN3 could have behaved as a quicksand in which very large animals became entrapped, as proposed byEspigares et al. (2023).
Anadón et al., 2003)ón et al., 2015) 2021)2021).In contrast, layer 5 of the UAL shows a predominance of species indicative of a brackish lacustrine environment, as evidenced by the presence of C. torosa with smooth valves (Fig.6C) (seeAnadón et al., 1994;Granados et al., 2021).The nding of a low frequency of shells of this species showing nodose valves would indicate occasional reductions of salinity in the environment(Fuhrmann, 2012;Anadón et al., 2015).In addition, the presence of Ilyocypris sp.(reported inAnadón et al., 2003)is probably indicative of uvial inputs from a nearby spring