Spatial signatures of carnivore-driven bone dispersal: A taphonomic analysis of modern proboscidean carcasses, and implications for Pleistocene elephant butchery sites

Spatial signatures of carnivore-driven bone dispersal: A taphonomic analysis of modern proboscidean carcasses, and implications for Pleistocene elephant butchery sites | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Spatial signatures of carnivore-driven bone dispersal: A taphonomic analysis of modern proboscidean carcasses, and implications for Pleistocene elephant butchery sites Manuel DOMÍNGUEZ-RODRIGO, Abel MOCLÁN, José Ángel CORREA CANO, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8799314/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This study integrates spatial point-process statistics and multivariate ordination to evaluate carcass modification and carnivore impact in modern elephant assemblages and to contextualize three archaeological sites (Áridos-2, Marathousa 1 and EAK). Using a suite of inhomogeneous spatial metrics and Minkowski functionals, modern elephant carcasses reveal a structured gradient from highly clustered, carnivore-impacted spatial patterns to more diffuse configurations associated with higher skeletal survival and various degrees of carnivore post-depositional disturbance. Principal component analysis demonstrates that spatial clustering, short inter-point distances, and fragmented carcass footprints covary with increased tooth-mark frequencies and reduced survival, whereas dispersed point patterns correlate with greater skeletal preservation. When archaeological assemblages are projected into this multivariate space, Áridos-2 and EAK (with caution, given its complex sedimentary nature) align closely with modern clustered carcass patterns indicative of either intense carnivore intervention with limited cluster disturbance, or early depositional stages (preserving a high degree of carcass articulation) or both, while Marathousa 1 occupies an intermediate position consistent with more limited carnivore access and greater structural integrity of carcass remains after disarticulation. Together, these results show that spatial point-pattern structure captures taphonomic processes at the carcass scale and provides a robust framework for inferring carnivore impacts in deep-time proboscidean assemblages. Biological sciences/Ecology Earth and environmental sciences/Ecology Biological sciences/Evolution Spatial statistics carnivore modification elephant carcass assemblages taphonomic processes Inhomogeneous spatial statistics Figures Figure 1 Figure 2 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction The study of elephant bone accumulations plays a pivotal role in reconstructing early hominin subsistence strategies, particularly in relation to the exploitation of megafauna [ 1 ]. Given their massive size and high nutritional return [ 2 , 3 ], proboscideans would have represented a significant resource for hominin populations if early access could be achieved, either through hunting or confrontational/active scavenging in highly competitive ecosystems, or even through opportunistic behaviors in low-competition habitats. Contrary to previous assumptions about prolonged meat availability from these carcasses [ 4 – 6 ], recent studies of proboscidean bulk flesh consumption by prides of lions in Zambia and Botswana show that carcasses can be mostly defleshed in five days before lions give way to other scavenging carnivorans [ 7 , 8 ]. So, hominin exploitation of proboscidean flesh may have occurred early in the consumption sequence. Alternatively, access to proboscidean resources may also have occurred later if targeting grease and flesh scraps instead. Determining the timing of access is a taphonomic challenge. The mere co-occurrence of elephant bones and stone tools does not inherently indicate a functional relationship between the two. In fact, most Pleistocene sites where these elements are found in spatial association are situated in alluvial contexts -environments that, as observed in modern African settings, foster highly productive taphocoenotic processes. These processes can result in spurious associations, where the proximity of bones and tools is the product of natural or marginal accumulation dynamics rather than intensive hominin proboscidean exploitation activities [ 9 – 12 ]. Fossil proboscidean sites occur in the vicinity of prehistoric lakes, river and ponds more abundantly than elsewhere [ 10 – 48 ], as they are also documented in modern African savanna ecosystems [ 49 ]. Understanding the spatial, anatomical, and taphonomic characteristics of elephant carcass accumulations and dispersions is thus essential to distinguish between natural processes and anthropogenic behaviors. This information can also be of relevance when reconstructing the timing of hominin access to such carcasses. One of the central challenges in this regard is disentangling the respective roles of carnivores and hominins in the access and modification of elephant remains. Elephant carcasses may persist on the landscape for extended periods, attracting a variety of taphonomic agents (including lions, hyenas, vultures, and environmental forces), which can all alter the bone assemblage [ 28 , 49 – 52 ]. Without careful assessment, these natural agents can mimic or obscure patterns that might otherwise suggest hominin involvement, such as tool-inflicted cut marks, percussion damage from marrow extraction, or the presence of lithic artifacts in spatial association with the bones. Recent neo-taphonomic studies of modern elephant death sites provide invaluable comparative data, offering baseline expectations for carcass modification in the absence of hominins [ 28 , 33 , 49 – 52 ]. These studies have shown that while carnivores such as hyenas can produce extensive damage on cancellous anatomical portions, their bone modification patterns may appear conspicuously limited to cancellous or trabecular tissue (Fig. 1 ). Thus, identifying felid and hyena impacts on proboscidean carcasses, and interpreting the subtle signatures of hominin exploitation (particularly early access to flesh) requires a detailed understanding of how elephant carcasses are naturally disarticulated, scattered, and consumed under undisturbed conditions. In modern African savanna ecosystems, the decomposition and transformation of elephant carcasses are shaped primarily by natural processes, especially the activity of durophagous carnivores such as hyenas. These powerful scavengers play a crucial role in the breaking, disarticulating, and scattering of skeletal elements, significantly altering the integrity and spatial configuration of carcasses over time. Other agents (i.e., smaller carnivorans) might have also played a relevant role in transporting bones away, but these agents have not received proper attention until now. Although one could think of elephant death sites as palimpsests spanning a large amount of taphonomic biotic processes through time, enabling the time-averaged overlying of the actions of successive carnivoran agents, controlled observations of such sites indicate a more nuanced reconstruction of this process. Most carnivore intervention at elephant carcasses occurs during the very initial phase of deposition and accumulation. After a few weeks (usually less than a month), carnivores impact the carcass minimally. This is the time in which no flesh is preserved, and some bone grease is accessible only to hyenas, but bones are already on a stage of drying up [ 7 ]. We have observed that it takes only a few weeks for some bones to start showing some sign of subaerial weathering, which would be expected much later according to observations made on smaller game [ 53 , 54 ]. Hyenas are among the few carnivores capable of crushing large bones, including those of elephants in specific anatomical locations (Fig. 1 ). Their exceptional bite force allows them to consume not only soft tissues but also to break into denser elements, such as long bone epiphyses, and even cranial structures. As a result, elephant skeletons rarely remain intact for long in the presence of active hyena populations. This is especially relevant in highly competitive ecosystems. The process of bone fragmentation, furrowing, and surface modification often begins within days or immediately after a carcass being abandoned by primary consumers like lions. Moreover, the dismemberment and transport of skeletal parts by hyenas contribute to the dispersion of bones across the landscape. Entire limbs or isolated elements may be dragged dozens of meters from the original site, resulting in spatially disaggregated assemblages. This scattering behavior not only affects the visibility of certain anatomical regions during taphonomic surveys, but also challenges efforts to reconstruct carcass use sequences or infer predator-prey dynamics from the fossil record. These patterns of carcass modification, particularly in large-bodied taxa like elephants, offer critical insights into the taphonomic signatures that aggregate under natural conditions. Understanding the role of durophagous carnivores, thus, becomes essential when interpreting fossil assemblages involving megafauna, especially in regions where early hominins may have interacted with similar carnivorans or exploited/competed for similar resources. Given the specific timeline of resource availability and carcass scattering/modification introduced during the post-defleshing scavenging phase, finding taphonomic signatures that could be linked to the timing of hominin intervention in elephant carcass exploitation results crucial. Here, we intend to specifically address this issue by using a taphonomic spatial approach. We will relate specific spatial patterns to degrees of post-depositional durophagous carnivore modification on bones and impact on bone assemblages (through bone survival estimates). We will then relate that to the timing of carcass access by hominins within the analysis of a selection of prehistoric butchery sites. The present work intends to address all these issues emphasizing an analytical approach, and it is innovative by doing so using spatial statistical taphonomy. Results Neo-taphonomic analysis For the purposes of this study, tooth-mark frequencies (TM) and bone survival percentages are used as complementary taphonomic variables to characterize carnivore impact on large-mammal carcasses (Table 1 ). TM represents the proportion of skeletal elements bearing carnivore tooth marks (or gnawing modifications) and provides a direct proxy for the intensity of carcass modification by carnivores during feeding. Bone survival reflects the proportion of the original skeletal assemblage that remains at the site after carnivore activity, integrating processes of bone removal, transport, consumption, and destruction. Together, these variables allow us to distinguish between assemblages dominated by intensive carnivore exploitation versus those experiencing more limited scavenging or disturbance. Table 1 Representation of skeletal parts (MNE: Minimum Number of elements) -except for ribs- in the Chobe and Okavango carcasses studied. Numbers in parentheses are for carnivore tooth-marked bones (numerator) and percentages per element (denominator). Tooth mark and bone survival frequencies are also included. Chobe National Park Okavango Delta Elephant_1 Elephant_2 Elephant_3 Giraffe Elephant_4 Elephant_5 Elephant_6 Cranium 1 1 1 1 1 1 1 Mandible 1 1 1 1 1 1 1 Ribs* 30 (6/20)* 35 (25/70)** 23 (10/43.4)*** 21 (9/39.1) 37 (24/64.8)**** 34 (8/23.5) 40 (6/15) Sternum 1 1 Vertebrae Atlas 1 1 1 1 Axis 1 1(1/100) 1 1 1 Cervical 3 6 (6/100) 1 6 2 6 6 Thoracic 14 (3/21.4) 15 (14/93.3) 7 (4/57.1) 9 (6/66.6) 4 (3/75) 20 18 Lumbar 2 (1/50) 1 (1/100) 6 5 Sacrum 1 1 Caudal 2 4 Innominates 2 (2/100) 2 (2/100) 2 2 (2/100) 2 2 Scapula 1 (1/100) 2 2 (1/50) 1 (1/100) 2 (1/50) 2 2 Humerus 2 (2/100) 1 1 (1/100) 2 (2/100) 2 2 Radius 2 1 (1/100) 1 (1/100) 1 1 (1/100) 2 2 Ulna 1 (1/100) 1 (1/100) 2 (1/50) 2 (2/100) 2 2 Femur 1 2 (1/50) 2 1 (1/100) 2 (2/100) 1 2 (1/50) Tibia 2 (1/50) 1 (1/100 1 1 2 2 2 Fibula 1 1 (1/100) 1 1 Calcaneum 1 1 2 2 Astragalus 2 1 Metacarpals/Metatarsals 11 13 Carpals/Tarsals 9 18 Phalanges 8 2 Tooth mark (TM) frequencies 26.5 84.3 36.9 40 61.6 6.6 5.3 Bone survival percentages 19.6 21.4 14.1 15 18.4 36.8 40 Estimates in MNE (Minimal Number of Elements). Parentheses: Numerator is number of toothmarked elements; denominator is the percentage of MNE bearing TM (pits/scores/furrowing and/or green breaks) * Out of 53 fragments **Out of 37 fragments ***Out of 34 fragments ****Out of 40 fragments The modern assemblages analyzed here show marked variability in both TM and survival, indicating substantial differences in carnivore behavior and access to carcasses. Elephant_2 exhibits extremely high TM values (84.3%) combined with low survival (21.4%), a pattern consistent with intense carnivore utilization leading to extensive bone modification and removal. Elephant_4 also shows high TM (61.6%) and low survival (18.4%), suggesting similarly strong carnivore pressure. In contrast, Elephant_5 and Elephant_6 present very low TM values (6.6% and 5.3%, respectively) coupled with higher survival (36.8% and 40%), indicating limited carnivore access, resulting in minimal bone modification and greater skeletal retention at the site. Intermediate cases, such as Elephant_1 (TM = 26.5%, survival = 19.6%) and Elephant_3 (TM = 36.9%, survival = 14.1%), reflect moderate carnivore involvement, with noticeable but not overwhelming impacts on carcass integrity. The giraffe assemblage occupies an intermediate position, with TM values of 40% and relatively low survival (15%), suggesting substantial carnivore interaction comparable to heavily exploited elephant carcasses despite taxonomic differences. Overall, the inverse relationship observed between TM and survival across assemblages underscores the role of carnivores as primary agents structuring bone modification and assemblage composition. These modern patterns provide a quantitative baseline for interpreting fossil elephant assemblages, allowing inferences about the intensity and nature of carnivore involvement based on comparable combinations of tooth-mark frequencies and skeletal survival. Neo-taphonomic analysis Stage 1: Modern and archaeological samples The hyperframe that we used summarizes a rich and multiscale description of spatial organization across sites, combining local intensity structure (Linhom, Finhom), second-order interactions (PCF), neighborhood statistics (NND, Hopskel), global dispersion (quadrat test), and geometric–topological descriptors (Minkowski functionals, convex hull, distances) (Table 2 ; Supplementary Table 2; Fig. 2 ). The PCA distills this high-dimensional structure into a smaller number of orthogonal gradients, allowing us to understand which sites are truly similar in overall spatial behavior rather than in isolated metrics (Supplementary Tables 3–4). Since PCA was used to avoid the effects of collinearity, it was required that the resulting number of retained components should include a minimum of 90% of the sample variance. In order to do so, four components were retained, comprising 93.47% of the total variance. Across the dataset, PC1 clearly separates sites by overall spatial scale, density, and geometric footprint, while PC2 primarily captures differences in aggregation versus dispersion at local and intermediate scales, and PC3–PC4 reflect finer contrasts in topology (Euler, perimeter) and anisotropy/heterogeneity. PC1 summarizes scale and extent (Fig. 3 a). It primarily captures a gradient in overall spatial scale and footprint expansion versus compact clustering. This component loads positively on inhomogeneous intensity measures (Linhom_ml, Linhom_mn, Linhom_mlg) and negatively on geometric descriptors such as scatter area, maximum distance, mean pairwise distance, and Minkowski area and perimeter. Sites with strongly positive PC1 (40% of variance) scores (notably EAK, Áridos-2, Haynes_2, Giraffe, Marathousa 1, and White-Diedrich) are characterized in the hyperframe by high Linhom_ml values, elevated clustering indices, limited convex hull areas, and small maximum distances. These sites represent spatial systems that occupy limited windows and show substantial spatial structure beyond simple CSR. Áridos-2 and Haynes_2 cluster close to one another on PC1 because they share elevated Linhom values, strong clustering indices, and similar Minkowski Euler values, indicating fragmented but spatially restrictive point patterns. In contrast, negative PC1 scores (Elephant_2, Elephant_4, Elephant_5, Haynes_1) correspond to sites with larger spatial extent, lower intensity, and more compact footprints, despite sometimes having strong clustering at local scales. These sites are spatially constrained and dominated by short-range interactions rather than broad spatial organization. This axis thus separates archaeological assemblages formed within restricted depositional contexts from modern carcasses where bone displacement and carcass spread are more extensive. PC2 (30% of variance) introduces a second, orthogonal gradient (capturing local aggregation intensity) that is largely independent of overall size. Strong negative loadings on Linhom_clust, Finhom_nn, and both pair-correlation indices (pcf_index and pcf_index2) contrast with positive loadings on the Hopkins–Skellam statistic. Negative PC2 values correspond to pronounced clustering and spatial dependence, while positive values indicate weaker interaction and more regular or diffuse patterns. Sites with strongly negative PC2 scores (Elephant_1, Elephant_3, EAK) are characterized by very high PCF indices, strong local clustering, and low Hopskel values, indicating intense aggregation at short distances. Elephant_1 and Elephant_3 are particularly informative here. Despite moderate PC1 values, they cluster tightly on PC2 due to their very similar hyperframe profiles: extremely high PCF indices, high quadrat statistics, and strong Euler values. This suggests intense, fine-scale clustering within relatively constrained spatial domains. EAK shares this strong short-range clustering signature, explaining its negative PC2 position and highlighting its multiscale complexity. Conversely, sites with positive PC2 scores (Giraffe, Marathousa 1, Haynes_1) tend to show lower PCF values, higher Hopskel values, and weaker local clustering, consistent with more dispersed, less dense or mixed spatial patterns. PC3 (16.5% of variance) reflects a contrast between intensity-driven aggregation and spatial fragmentation linked to carcass modification and redistribution. Negative loadings on Linhom metrics and quadrat statistics oppose positive contributions from pair-correlation functions and NND_index. High positive PC3 scores indicate patterns with strong local aggregation but reduced large-scale coherence, whereas negative scores correspond to assemblages structured by intensity gradients and broader-scale density fields. In the scores, Áridos-2 and Marathousa 1 have positive PC3 values, suggesting aggregation driven by localized processes rather than simple density scaling. EAK, however, plots strongly negative on PC3, indicating a spatial pattern tightly structured by inhomogeneous intensity, consistent with a strongly constrained depositional environment. Modern elephant carcasses show mixed PC3 behavior, with Elephant_3 and Elephant_1 positive (localized aggregation) and Elephant_2 negative (more fragmented, intensity-driven distribution). PC4 (6.7% of variance) (which highlights secondary geometric contrasts) captures subtler variation, largely decoupled from size and clustering intensity. This component loads positively on nearest-neighbor structure (NND_index), Hopskel, and some interaction metrics, while loading negatively on the Euler characteristic, indicating sensitivity to the connectivity and perforation of occupied space. High positive PC4 scores indicate spatial configurations with many small clusters or disconnected small patches, whereas negative values reflect more topologically cohesive distributions. PCA4 also reflects differences in internal organization, such as whether clustering produces elongated structures versus compact ones, and how perimeter scales with area. Elephant_1 and Haynes_1, which load positively on PC4, show disproportionately large perimeters relative to area, suggesting elongated or irregular clustering. Negative PC4 values (e.g., Elephant_2, Elephant_6, Giraffe) correspond to more compact, isotropic spatial arrangements. Taken together, the PCA scores show that archaeological elephant assemblages (EAK, Áridos-2, Marathousa 1) occupy a distinct region of multivariate space compared to modern elephant carcasses. When the raw hyperframe data and the PCA are considered jointly, several robust site groupings emerge: Elephant_1 and Elephant_3 are among the most similar overall. They share: Very high PCF indices, strong quadrat deviation, and high Euler values.PCA confirms this similarity by placing them close together, especially along PC2 and PC3. They represent intensely clustered, fine-scale spatial systems, likely reflecting constrained accumulation processes. Elephant_4, Elephant_5, Elephant_6 form a loose cluster characterized by: Moderate clustering, longer scatter areas, and lower Euler fragmentation. PCA places them near one another on PC1 and PC2, indicating broadly similar spatial organization with moderate internal variation. Áridos-2 and Haynes_2 align on PC1 and PC3, driven by: High intensity, strong clustering indices, and large Euler values. Giraffe & Marathousa 1 cluster on PC2 and PC3, sharing: Lower PCF values, higher Hopskel values, and moderate Minkowski measures. Their spatial structure is less clustered and more dispersed, suggesting different formation dynamics from the elephant-dominated sites. EAK occupies an extreme position: high PC1, negative PC2, strongly negative PC3. This reflects a site that is simultaneously strongly clustered, and topologically cohesive, setting it apart from all others. Modern elephant carcasses span a broader range of PCA space, reflecting variability in carcass spread, scavenger-driven bone displacement, and landscape openness. Overall, the joint interpretation of loadings and scores indicates that spatial patterning in the archaeological sites cannot be explained simply as scaled-down versions of modern carcass dispersal, but instead reflects distinct formation dynamics involving spatial confinement, aggregation, and limited bone transport. A look at the distribution of the spatial point patterns shows that at least two (Elephants_1 and 3) of the three least conspicuously impacted by carnivores cluster together and separately (Fig. 3 b-c). We also noticed that time of carcass exposure is unrelated to either carnivore ravaging intensity, spatial point pattern or spread of the scattering distance. It is remarkable that the intensity of carnivore damage on carcasses shows a strong negative correlation with the skeletal element survival index (r = − 0.63, p-value = 0.1258). This is expected in the face of observational studies showing that the most intense damage to elephant carcasses occurs in the first few days of exposure, and that depending on the degree of food availability and carnivore competition in the ecosystem, proboscidean carcasses may remain unscattered and minimally impacted by durophagous carnivores, thus indicating that carcass damage and survival is ecologically-determined beyond the ecosystem predator carrying load [ 7 ]. To analyze this better, a second spatial analysis was carried out using only the experimentally documented modern carcasses, for which taphonomic information is available. Stage 2: Spatial and neo-taphonomic analysis of modern samples For this part of the analysis, only the modern carcasses that we studied were included in the analysis (Supplementary Tables 5–7). Taken together, the modern-elephant spatial results and the PCA provide a coherent and internally consistent picture of how different carcasses are distributed in space, how strongly they cluster, and how those patterns relate to movement scale, intensity structure, and inferred behavioral regimes such as exposure time and bone survival. What is especially powerful here is that the PCA does not contradict the raw spatial statistics; instead, it compresses them into a small number of interpretable spatial syndromes that map cleanly back onto the original measures. At the level of the raw spatial metrics (Supplementary Table 5), the modern elephants separate into two broad behavioral–spatial modes. Elephant_1 and Elephant_3 show extremely strong clustering signatures. Both exhibit very high Linhom_clust values (above 13), very large Finhom_nn values, and extreme deviations in pair-correlation indices (pcf_index and pcf_index2 both above 6). Their Hopskel values are very low, especially Elephant_3, indicating strong departure from CSR toward aggregation. Quadrat statistics are also very large, reinforcing the interpretation of highly clustered point patterns. Despite this strong clustering, their movement envelopes are not the largest in absolute terms: their scatter areas and convex-hull areas are moderate compared to some other elephants. This combination suggests limited transport by carnivores rather than wide-ranging dispersion. This seems to be unrelated to bone survival, which is reflected in low values for both carcasses. In contrast, Elephant_2, Elephant_4, and Elephant_5 form a second group characterized by weaker clustering and much larger spatial footprints. Their Linhom_clust values are substantially lower (around 4–8), Finhom_nn values are roughly half those of Elephant_1 and Elephant_3, and their pcf indices sit in the moderate range (roughly 2–3). At the same time, these individuals show very large M_area, scatter_area, and max_distance values, particularly Elephant_2 and Elephant_4, which have the largest convex hulls and pairwise distances in the dataset. This indicates extensive movement across space with less intense local reuse by carnivores, a pattern that aligns well with medium to short bone survival values and higher TM values, suggesting more transient or carnivore-impacted spatial behavior. Elephant_6 stands somewhat apart from both groups. It shows moderate clustering indices and relatively low pcf values compared to Elephant_1 and Elephant_3, but its spatial footprint is small, with the lowest scatter area, shortest max_distance, and lowest mean_pairwise distance among all elephants. This suggests a spatially constrained individual that nonetheless does not exhibit extreme point aggregation. Its very short TM and higher bone survival are consistent with a brief time exposure, spatially limited post-depositional disturbance, and more limited carnivore impact. The giraffe provides a useful external anchor. Its Linhom_clust is the lowest of all patterns, its pcf indices are well below 1, and its Finhom_nn is the smallest in the dataset. Quadrat statistics are also much lower than those of the elephants. At the same time, its spatial footprint is very small, with minimal scatter area and short pairwise distances. This combination reflects near-random or weakly regular spacing over a small area, fundamentally different from the elephant patterns and validating that the analytical framework is sensitive to genuine ecological or behavioral contrasts rather than producing uniform outputs. When these same variables are projected into PCA space, the underlying structure becomes clearer (Supplementary Tables 6–7) (Fig. 4 a). PC1 primarily captures a scale-versus-dispersion gradient. Variables associated with large spatial extent (scatter_area, max_distance, mean_pairwise distance, M_area, and M_perimeter) load strongly and negatively on PC1, whereas intensity-related measures such as Linhom_ml, Linhom_mn, Linhom_mlg, NND_index, and Survival load positively. As a result, Elephant_2, Elephant_4, and Elephant_5, which have very large spatial footprints, plot strongly negative on PC1, while Elephant_6 and especially the giraffe plot strongly positive. Elephant_1 and Elephant_3 occupy intermediate PC1 positions, reflecting their moderate spatial extent but high internal clustering. PC2 separates patterns according to clustering structure rather than scale. Linhom_clust, Finhom_nn, and both pcf indices load strongly and positively on PC2, while Hopskel loads negatively. High PC2 scores therefore correspond to intense clustering across multiple spatial statistics. Elephant_1 and Elephant_3 sit very high on PC2, matching their extreme pcf and Linhom_clust values. Elephant_4 and Elephant_5 are negative on PC2, consistent with weaker clustering, while Elephant_2 occupies an intermediate position. The giraffe again lies apart, with low clustering metrics translating into a negative PC2 score. PC3 introduces a contrast between quadrat-scale aggregation and topological structure. Quadrat_stat and Hopskel load strongly and positively, while M_euler loads strongly and negatively. This component differentiates patterns where aggregation manifests as large-scale count heterogeneity versus patterns where the occupied space becomes highly perforated or fragmented. Elephant_3, which shows extreme quadrat values and low Hopskel, is strongly negative on PC3, whereas Elephant_1 is positive, reflecting subtle but meaningful differences in how clustering is organized spatially despite similar pcf behavior. PC4 and higher components refine these interpretations by incorporating carnivore damage (TM), survival, and finer-scale geometric properties. PC4 shows strong positive loadings for TM, mean_pairwise distance, and Linhom measures, separating Elephant_1 (positive PC4, long TM and long survival) from Elephant_2 (negative PC4, short TM and short survival). PC5 and PC6 further isolate Elephant_6, whose unusual combination of limited carnivore damage, constrained space use, and moderate clustering gives it extreme loadings on these axes. The giraffe consistently remains isolated across components, reinforcing its role as a structurally distinct spatial pattern rather than simply an endpoint along an elephant gradient. Viewed holistically, the fusion of raw spatial statistics and PCA reveals that modern elephant space use is not defined by a single continuum but by at least two interacting dimensions: the scale of movement across the landscape and the intensity of local clustering within that space. Elephant_1 and Elephant_3 converge in PCA space because they share extreme clustering signatures, even though they differ in bone survival and carnivore damage. Elephant_2, Elephant_4, and Elephant_5 cluster together because their dominant trait is large-scale dispersion with relatively weaker aggregation. Elephant_6 emerges as a distinct spatial strategy characterized by confinement without strong clustering. The giraffe stands apart as a fundamentally different spatial process. This integrated perspective strengthens confidence in the results (Fig. 4 b-c). The PCA clarifies how multiple spatial statistics co-vary and how individual animal carcasses impacted by different degrees of carnivore impact and preservation lead to consistent spatial patterns. In this sense, the modern-elephant dataset provides a well-resolved reference framework that can be directly compared to fossil or archaeological assemblages, allowing differences in clustering, dispersion, and space-use intensity to be interpreted, not as isolated metrics, but as coherent spatial models with taphonomic implications. Stage 3: Spatial analysis with homogeneous windows After controlling window size by restricting all point patterns to the same 10 × 10 m area, the PCA isolates true differences in spatial organization rather than differences driven by observation scale (Supplementary Tables 8–10). This normalization is critical: previously, large scatter areas and long inter-point distances strongly structured the PCA, but here those effects are intentionally minimized. As a result, the first PCA primarily reflects local clustering intensity, spatial heterogeneity, and topological structure, making the comparison between modern and archaeological assemblages more meaningful at a different level; that is, excluding the overall effect of dispersion captured by the heterogeneous windows and concentrating at agglomerative and segregation processes at the dense cluster local level (Fig. 5 ). PC1, which separates sites most strongly along the horizontal axis, is dominated by intensity-related variables (Linhom_ml, Linhom_mn, Linhom_mlg) loading positively and by Minkowski and distance-related measures (M_area, scatter_area, max_distance, mean_pairwise) loading negatively. High positive PC1 scores, therefore, characterize assemblages with very dense local clustering within the standardized window, whereas negative PC1 values indicate more locally spatially dispersed or fragmented clusters even within the same area. Áridos-2 and EAK occupy the extreme positive end of PC1, indicating that their densest 100 m² zones are characterized by exceptionally high local intensity relative to background, closely resembling tightly packed accumulations. This is due to carcasses at these localities being in articulated state (Áridos-2) or semi-articulated and additionally tectonically compressed, further increasing the clustering intensity (EAK). In contrast, Elephant_4, Elephant_5, Haynes_1, and Elephant_6 fall on the negative side of PC1, reflecting more diffuse or broken spatial structure even at peak density, as would be expected from disarticulated carcasses and assemblages modified by dispersing carnivore behavior. PC2 is shaped almost entirely by second-order interaction statistics, with strong negative loadings from Linhom_clust and Finhom_nn and similarly strong contributions from both pair correlation indices. This axis captures the degree of spatial interaction beyond simple intensity, distinguishing patterns dominated by strong clustering and aggregation at short distances (negative PC2) from those closer to randomness or weak interaction (positive PC2). EAK is again extreme here, with a very strong negative PC2 score, indicating pronounced short-range clustering within its densest cluster. Áridos-2 also shows positive clustering structure but less extreme. Among modern assemblages, Elephant_1, Elephant_2, Elephant_3, and Giraffe occupy positive PC2 values, suggesting more moderate or heterogeneous short-range interactions once window size is fixed. PC3 is primarily structured by quadratic and topological contrasts, with quadrat_stat and Linhom_mn loading negatively and hopskel loading weakly. This axis separates assemblages with high internal heterogeneity and patchiness from those with smoother internal organization. Áridos-2 stands out again with a strong negative PC3 score, suggesting that even within its densest cluster, bone distribution is internally irregular and segmented, consistent with repeated disturbance or selective removal. In contrast, Giraffe, Elephant_2, and Haynes_2 load positively on PC3, indicating more internally coherent spatial clusters. PC4 highlights nearest-neighbor structure and Euler topology, with strong positive loadings from NND_index and Hopskel and strong negative loadings from M_euler. This component differentiates patterns dominated by many small disconnected voids and holes (highly negative Euler values) from those with more continuous occupied space. Elephant_2 and Giraffe score high on PC4, reflecting compact clustering, whereas Elephant_1 and White-Diedrich score negatively, indicating fragmented spatial structure even at high density. Higher components (PC5–PC8) capture more subtle contrasts, including trade-offs between nearest-neighbor spacing and clustering strength (PC5), smoothing versus fragmentation effects (PC6), and quadrat-based regularity versus Euler complexity (PC7). These axes refine differences among modern elephant assemblages but do not drive the main archaeological–modern separation. When viewed together, the PCA reveals a clear and consistent pattern: after standardizing window size, Áridos-2and EAK remain distinct from modern elephant carcasses, not because of scale but because of intrinsic spatial organization. Their densest clusters are simultaneously more intense, more tightly aggregated, and more topologically complex than any modern carcass pattern, because they are not completely disarticulated as the modern samples are. Marathousa 1, by contrast, falls closer to the modern elephant cluster, especially Elephant_1 and Elephant_3, suggesting that its spatial organization is more consistent with limited disturbance and localized accumulation, after initial disarticulation. The main result of the Stage 3 analysis is that window size was not driving earlier interpretations. Even when controlling spatial extent and focusing only on the densest 100 m² areas, archaeological assemblages (especially Áridos-2 and EAK) retain spatial signatures indicative of non-random accumulation, limited modification, and controlled carnivore-mediated disturbance, whereas modern elephant carcasses exhibit spatial structures consistent with a more disarticulation (probably reflecting time-averaging) and probably more carnivore impact. The first argument can be defended in sites like Áridos-2 and EAK exhibiting barely any bone weathering caused by subaerial exposure (in contrast with several of the carcasses studied). The second argument is more speculative, since it can only be effectively assessed over broader areas than those unearthed at the three archaeological sites. This certainly limits behavioral interpretations about the depositional history of those elephant carcasses, their impact by carnivores and additional taphonomic information potentially crucial to unravel the timing of hominin access to them. Discussion In our philosophy of trying to implement an analytical taphonomy of proboscideans (to counter the more traditional descriptive taphonomy), we adopted a taphonomic spatial approach, which was previously successful at differentiating degrees of post-depositional disturbance of experimental assemblages subjected to hydraulic flows [ 55 ]. Taken together, the PCA structure and the original spatial statistics reveal a coherent and taphonomically meaningful relationship between point-process structure and both carnivore tooth‐mark frequencies and bone survival, with scattering processes also possibly reflecting time averaging intervention of dispersing agents. Rather than acting independently, these taphonomic variables align consistently with gradients of spatial clustering, scale, and geometric complexity captured by the point‐pattern descriptors. At one extreme of the ordination lie point processes characterized by large spatial extent, long interpoint distances, and expansive convex hulls; patterns with high scatter area, large maximum distance, and high mean pairwise distances. In the modern dataset, Elephant_2 and Elephant_4 fall clearly into this domain, loading strongly in the negative direction of PC1 and PC2. These processes are spatially diffuse: points are spread over wide areas, nearest-neighbor distances are large, and Minkowski area and perimeter values are high. Importantly, these same cases are associated with relatively high TM values and reduced bone survival. This association suggests that when carcass remains are spatially dispersed, carnivores have had more opportunity to transport, disarticulate, and selectively remove skeletal elements. Spatial diffusion therefore appears to be a strong spatial signature of intensive carnivore intervention. By contrast, point processes clustering on the opposite side of PC1 (most clearly Elephant_6 and Giraffe) exhibit compact spatial configurations (Fig. 4 a). These assemblages show small scatter areas, short maximum distances, and low mean pairwise distances, coupled with lower Minkowski area and perimeter. Their point patterns are dense but spatially constrained, indicating limited lateral transport. Correspondingly, these cases show low TM values and higher bone survival. In practical terms, this implies that when bones remain spatially concentrated, carnivores either had limited access or engaged primarily in localized consumption rather than prolonged transport. The PCA, thus, captures a consistent spatial–taphonomic signal: compact point processes are associated with better skeletal survival and fewer carnivore modifications. A second important gradient emerges along PC2 and PC3, which are strongly influenced by measures of clustering intensity and spatial heterogeneity, including Linhom_clust, Finhom_nn, pcf_index, and pcf_index2. Assemblages such as Elephant_1 and Elephant_3 score high on PC2, reflecting strong small-scale clustering and pronounced departures from spatial randomness (Fig. 4 a). These sites also exhibit elevated TM values but intermediate bone survival. This combination suggests a scenario in which carcasses remained spatially aggregated, yet were repeatedly revisited by carnivores. Rather than extensive transport, carnivore activity here appears to have been focused on intensive modification within a restricted area, producing high tooth‐mark frequencies without complete removal of remains. The role of geometric complexity, captured by Minkowski functionals (especially Euler characteristic) adds another layer of interpretation. High absolute Euler values, indicating fragmented or topologically complex spatial structures, tend to align with higher TM and lower survival. This is particularly evident for Elephant_2 and Elephant_3, where complex spatial topology coincides with strong evidence of carnivore processing. In contrast, assemblages with simpler spatial topology (lower Euler magnitudes) tend to preserve skeletal material more effectively. This suggests that spatial fragmentation itself may be a downstream consequence of carnivore behavior, reflecting repeated disturbance, displacement, and reorganization of remains across the landscape. Importantly, TM and Survival do not map onto a single spatial metric but instead emerge from the combined effect of scale, clustering, and geometry. TM aligns most strongly with dimensions capturing local clustering intensity and interaction (PC2 and PC3), whereas survival aligns more closely with global spatial extent and dispersion (PC1). This distinction is biologically intuitive: tooth marks accumulate through repeated localized interactions between carnivores and bones, while bone loss is driven primarily by transport beyond the observation window. All this spatial information shows that the PCA reveals that spatial point-process structure is not merely descriptive but directly informative about underlying taphonomic processes. Assemblages with compact, tightly clustered point patterns tend to reflect limited carnivore impact and high survival, whereas spatially extensive, fragmented patterns are strongly associated with intensive carnivore modification and bone removal. Intermediate configurations capture mixed behavioral regimes, where carnivores repeatedly modify remains without fully dispersing them. Together, these results demonstrate that spatial statistics provide a powerful framework for disentangling carnivore behavior from the spatial organization of archaeological and paleontological assemblages. When the spatial signatures derived from the modern elephant carcasses are taken as a comparative baseline, they provide a powerful interpretive framework for evaluating carnivore impact and additional taphonomic processes in archaeological elephant assemblages. In modern cases, the point patterns represent well-understood ecological processes: carcasses are initially deposited by natural mortality or predation, and then progressively modified by carnivore feeding, transport, and bone removal. The PCA shows that these processes leave consistent spatial fingerprints. Assemblages characterized by high tooth-mark frequencies (TM) and low survival percentages tend to occupy regions of PCA space associated with dispersed point patterns, large scatter areas, long maximum distances, and elevated mean pairwise distances. Conversely, assemblages with lower carnivore impact cluster in PCA regions dominated by higher local intensity contrasts, stronger clustering indices, and tighter spatial footprints. In this sense, the modern elephant carcasses define a continuum from spatially compact, minimally resedimented bone concentrations to highly dispersed, carnivore-modified assemblages. Unfortunately, all of this cannot be captured through the small windows of the archaeological sites. To interpret archaeological proboscidean sites adequately, we need therefore to exclude the larger picture obtained through the analyses of the extended windows and focus on the small windows represented by the sites themselves. Since no proboscidean site has been excavated beyond 100 m2 (approximately EAK´s size) [ 1 ]. This requires us to limit our referents to their densest smallest windows (Stage 3 analysis). When the modern spatial–taphonomic gradient derived from the small-window approach of modern elephant carcasses is used to contextualize the archaeological assemblages, Áridos-2 emerges as an extreme but internally dense spatial configuration, rather than a highly dispersed one. Its strongly positive scores along PC1 (driven by high Linhom intensity measures and reduced spatial extent) indicate a concentration of remains within a relatively compact area characterized by elevated local point density. At the same time, the combination of a high quadrat statistic and strong deviation from complete spatial randomness suggests that this compactness is not simply the product of homogeneous accumulation, but instead reflects structured aggregation within the observation window. This fits well with the almost fully articulated nature of the preserved remains, despite the carnivore damage documented on some of the bones [ 48 ]. In PCA space, Áridos-2 occupies a position that partially overlaps with modern carcass configurations associated with substantial carnivore modification, though through a different spatial pathway than simple dispersal. Rather than exhibiting extensive spatial spread, Áridos-2 shows evidence for intensive impact within a confined area, consistent with repeated carnivore access, trampling, and localized bone rearrangement. Its proximity to modern carcasses with elevated tooth-mark frequencies and reduced skeletal survival supports an interpretation of significant carnivore involvement, likely involving bone consumption and short-distance redistribution rather than wholesale removal across a broad landscape. This pattern is compatible with a scenario in which carnivores acted repeatedly on an initially concentrated accumulation, generating spatial complexity and fragmentation without producing a highly dispersed carcass field. Marathousa 1, by contrast, occupies a very different position relative to the modern baseline. Its point pattern is comparatively compact, with low pcf indices, modest scatter area, short maximum distances, and low Finhom-based neighborhood estimates. In PCA space, Marathousa 1 aligns more closely with assemblages showing limited dispersion and relatively stable spatial structure. This configuration resembles modern scenarios with lower carnivore transport intensity, where bones remain near the original deposition area and spatial patterning is dominated by primary accumulation processes rather than extensive secondary modification. Under this light, Marathousa 1 appears consistent with a scenario in which carnivores were present but did not extensively dismantle or disperse the assemblage, allowing spatial coherence to persist. EAK represents a third and more complex case. Its exceptionally high Linhom values, strong clustering index, and large quadrat statistics point to intense local aggregation of remains, while at the same time its Finhom and pair-correlation indices suggest non-trivial spatial structuring across multiple scales. In PCA space, EAK falls near modern assemblages that combine high carnivore interaction with strong spatial anchoring (patterns seen when carcasses are heavily exploited but bones are repeatedly revisited rather than widely transported). This dual signal is consistent with scenarios involving prolonged carnivore access at a relatively fixed location, such as denning or repeated scavenging locales. Compared to Áridos-2, EAK shows less evidence for extreme dispersal, but more evidence for sustained carnivore-driven modification than Marathousa 1. The tectonic deformation of the assemblage might also have impacted the original clustering, making it extraordinarily and artificially compact, as detected in the spatial statistical analysis. The interpretation of these archaeological assemblages is just preliminary, and must be carried out with extreme caution, since they are dimensionally different from the analogical modern carcasses studied in this work, despite our interpretation being limited to the small-window analysis (Stage 3). Crucially, however, the interpretive patterns that were derived in this study -especially the contrasts between modern elephant carcasses and archaeological assemblages- remain robust in a relative sense. While window size affects absolute values, it is unlikely to reverse the observed gradients: modern carcasses consistently show various degrees of clustering, spatial spread, and carnivore-driven removal signatures, whereas Áridos-2, Marathousa 1, and EAK exhibit within-window limited dispersion, some spatial fragmentation, and signatures consistent with prolonged carnivore modification. As long as windows are taphonomically meaningful, and comparisons are made cautiously across similarly defined spatial frames, window size modulates magnitude but does not invalidate the taphonomic interpretations. Therefore, window size matters greatly for scale, magnitude, and comparability, but when handled consistently, it does not undermine the core conclusions about spatial organization, site formation processes, or carnivore impact. Conclusion The main hypothesis of carnivore impact and its related scattering point processes was tested successfully. This is reflected in all carcasses having undergone high carnivore impact (with tooth marked bones in frequencies > 20%), showing low bone survival values (< 25%), and displaying a higher degree of scattering as well as lower spatial clustering (Elephant_2, 4 and 5). However, the opposite is not true; carcasses with high degree of clustering can also display intense carnivore impact (reflected both in tooth marking frequencies and low bone survival (e.g., Elephant_1 and 3). Even carcasses with low carnivore impact (as reflected in high bone survival rates and low tooth mark frequencies) can be locally scattered to the point of not creating significant clustering assemblages (e.g., Elephant_6 and Giraffe). An alternative way of interpreting the documented point patterns is that (regardless of how much in locus gnawing tooth place) all of them have undergone variable but intensive carnivore impact, since > 50% (in some cases up to > 85%) of the original carcass has been removed from its depositional spot. Under the modern observational/experimental referential point, patterns generated by the study of the modern elephant (and giraffe) carcasses, and relating them to the taphonomic evidence of carnivore impact, under two different (but complementary) processes (bone transport signaled by the carcass bone survival rates) and in locus bone consumption indicated by gnawing intensity and the resulting bone modification and tooth-marking), it can be argued that the documented point patterns find matching with higher or lower degrees of impact of carnivores on elephant carcasses. When used as a reference for the archaeological elephant butchery assemblages of EAK, Áridos-2, and Marathousa 1, as a preliminary testing ground, the archaeological assemblages link to specific referents showing different degrees of taphonomic modification by carnivores. The intense clustering of Áridos-2 and EAK indicates a limited time of exposure, and probably, an early access by hominins in the early stages of carcass resource exploitation targeting bulk flesh extraction. This is convincingly shown in the presence of cut marks on the pelvic and axial portion of the elephant skeleton at Áridos-2 [ 48 ]. This first testing spatial taphonomic experiment for archaeological interpretation is promising. It warrants further and more detailed analysis of modern and archaeological elephant carcasses, so that we can understand how non-primate carnivores and hominins interacted in the consumption of elephants. Given that most of the historically-documented elephant butchery leaves very few traces on bones, the lack of direct evidence underscores the need of developing alternative and complementary methods to detect in which moment hominins had access to elephant carcasses and what resources they extracted from them. Ultimately, interpreting elephant bone accumulations through a taphonomic lens enables researchers to evaluate key hypotheses about the economic capabilities and ecological roles of early humans, including their capacity for timely accessing key proboscidean flesh and fat anatomical sections, extent of their cooperative behavior, and the use of technology to access high-yield resources. Given the energetic potential of a single elephant carcass, even rare instances of access could have had profound effects on hominin diet, social organization, and territoriality. The integration of controlled taphonomic observations with fossil assemblages is therefore indispensable for advancing our understanding of hominin-megafauna interactions during the Pleistocene. Methods Neo-taphonomic samples The present study was conducted in the Chobe National Park and in the Okavango delta (Botswana) during the month of June 2024. We sampled seven carcasses (six elephants and one giraffe for control as to what the taphonomic differences could be when compared to a smaller megafaunal taxon) (Table 1 ). All the carcasses that were examined and whose information was collected had been either discovered by local rangers or game drivers and approximate dates of death were provided. The exposure time of the carcasses ranges from three and a half months to more than two years (Supplementary Table 3). Given that all the carcasses spotted were inside both National parks and reserves, we limited our study to the documentation of the conspicuous elements that were visually observed on the ground and did not engage in intensive screening of the ground for small bone fragments. Bones were identified to elements using estimates of the minimum number of elements (MNE). Given the overall completeness or limited fragmentation of most non-axial bones, quantifications of those anatomical areas were very precise. Regarding the axial skeleton, estimates of vertebrae were also very accurate (with minimal modification, and mostly restricted to damage on the arch processes). Estimates of ribs were very conservative, since we did not collect (in those carcasses affected by more intense fragmentation) all the small fragments to piece them together. Although for most of the sample, ribs were fairly complete, despite the high percentage of carnivore damage documented on them, in a couple of instances, fragmentation was more intense. In those cases, MNE estimates were derived using a combination of criteria involving: all ribs that were preserved in more than one third of their size, and the presence of specimens with tubercles and heads. Carnivore impact was documented macroscopically. The size and density of proboscidean bones make it very difficult for hyenas to break them, and when they do so, they leave abundant and conspicuous traces of gnawing and tooth marks (Fig. 1 ). For this reason, we did not quantify the smaller isolated tooth marks that might have required higher magnification. We tallied carnivore damage as the presence of tooth pits, scores and furrowing; most commonly occurring in combination in most of the specimens affected. We also quantified skeletal survival indices, based on the number of preserved bones compared to complete skeletons. The highly skewed profiles were caused by the predominant deletion/disappearance of ribs. It was also due to the more intense spatial scattering of these elements. In order to keep spatial windows in maximum comparable proportion to archaeological sites, we did not collect information from bone scatters occurring on a radius of more than ~ 25 m from the main cluster. All the spatial information of carcasses was collected through the spatial reconstruction of the delimited surfaces containing the bulk of the accumulation and scattering of bones from each carcass (Fig. 6 a-d). The data were obtained with various software tools, employed both during fieldwork and in the laboratory. The first step in obtaining the spatial data consisted of creating initial photogrammetric models using an iPad Pro (Apple Inc.) and the 3d Scanner App™. The use of this device's LiDAR technology allows not only for rapid documentation of the study areas, but also for the real-time scaling of the 3D models. We did at least two 3D models per carcass to ensure that we could retrieve all information later in the laboratory. This was key to obtaining data from bones located at the periphery of the models. Once in the laboratory, the 3D models were exported, including their textures and all the photographs taken by the 3d Scanner App™ used to generate them. Subsequently, the 3D models were imported into Blender 4.2; since they were already scaled, a 1x1 m grid was created and the light sources were adjusted to achieve the best possible view. The models were rendered including the grid to generate orthophotographs in .tiff format. The Cycles rendering engine was used to prioritize maximum image quality. However, as the quality of the orthophotos generated with 3d Scanner App™/Blender 4.2 is not particularly high, higher-quality orthophotographic models were subsequently generated in Agisoft Metashape using the photos exported from 3d Scanner App™. This provided a .tiff format orthophotograph of higher quality, but without scaling. Therefore, both orthophotographs were then imported into ArcMap 10.5, and using the grid generated in Blender 4.2, both orthophoto sets were scaled. Once the high-quality photogrammetry was scaled, .shp files were generated to digitize the polygons of the bones from the different samples. Once we had the polygon layer completed, we generated a point layer from the polygons to obtain the xy coordinates of the centroids of each bone through the ‘Feature to Point’ function (selected features: Inside (optional): checked ). In addition, we also created point layers to generate the spatial windows (i.e., analysis areas) that would be used later. In addition to the experimental/observational carcasses that we documented, we also added three additional modern elephant carcasses. Two were from Hayne´s work in Zimbawe [58: Figs. 10 and 13], and the third one was from White and Diedrich [8: Fig. 5 ] in Zambia. These carcasses were selected due to the high quality of the figures at our disposal, which was key to the accurate digitization of the floor plans and the subsequent extraction of spatial data. The methodology for spatial data collection remained consistent with the protocols established for our documented carcasses. In the three cases, not quantifiable information was provided on carnivore damage, and for this reason, we excluded these three samples from the analysis when the focus was taphonomic specific (analytical stage 2). Archaeological samples For the study of archaeological samples, we selected three previously published sites (Fig. 6 e-g), each containing a single anthropogenically modified elephant carcass. The selection of these sites was based on the feasibility of accurately digitizing the published cartographies, as certain cases lacked sufficient detail to ensure reliable map digitization [ 59 , 60 ]. The fact that well-verified prior information regarding the anthropogenic modification of the carcasses existed was also taken into account. The first site was recently published by our research team [ 1 ], with the cartography already available. The site is named Emiliano Aguirre Korongo (EAK) and is located at the base of Bed II in Olduvai Gorge (Tanzania). It is located at the confluence of the two gorges, between the sites of FLK-N and FLK-NN [ 61 ]. Stratigraphically, the site rests directly upon Tuff 1F (1.78 Ma), which marks the beginning of Bed II [ 62 ]. The sedimentary context of the assemblage corresponds to a low-energy clay environment with seasonal inputs of coarser sediments. The faunal material from EAK consists of a minimum number of 46 elements from a Elephas recki carcass. Although no cut marks were identified in the assemblage, two specimens exhibit green breakage interpreted as anthropogenic. The absence of several highly significant anatomical elements (e.g., many vertebrae and ribs, bones from three feet, and the femoral epiphysis) suggests carnivore activity within the assemblage. Furthermore, 80 stone tools, mostly flakes and flake fragments, were identified in the same area, showing a spatial correlation with the elephant carcass. The Áridos-2 site (Arganda del Rey, Spain), excavated in 1976 [ 10 , 63 ], is located in the complex terrace of the Jarama River, specifically within the Arganda I stratigraphic unit [ 64 , 65 ], dated between MIS 10 − 9 [ 66 ]. The site revealed the presence of a Palaeoloxodon antiquus carcass associated with 34 Acheulean lithic artifacts [ 56 ] within a context of muddy overbank deposits and secondary pebble and sandy low-energy channels on an alluvial plain [ 64 ]. The elephant remains were largely preserved in anatomical connection. The presence of cut marks has been demonstrated [ 48 ], some of which are consistent with those generated by bifacial tools [ 48 , 67 ]. Furthermore, microwear analysis of the lithic tools supports anthropogenic intervention on the carcass, documenting the use of tools for intensive butchery activities [ 68 ]. The digitized map is Fig. 1 from Santonja and Querol [ 56 ]. The other site is Marathousa 1, located in the Megalopolis basin (Greece) [ 57 ] and dated between 0.48 and 0.42 Ma [ 69 – 71 ]. Sedimentologically, the site corresponds to an extensive mudflat surrounding a lake shore [ 72 ]. As with Áridos-2, a single Palaeoloxodon antiquus carcass associated with lithic industry was identified at this site [ 57 ]. The lithic sample is significantly larger than in the previous cases (n = 1,170), with a notable presence of chips and flakes; the use of certain bones as tools has also been proposed [ 73 ]. Another distinction from the other sites is the greater presence of remains from smaller ungulates (omitted in this analysis), which also exhibit evidence of anthropogenic exploitation [ 74 ]. Other animals, such as birds, are present at the site, although their presence appears to be related to natural accumulation in this lacustrine environment rather than human activity [ 75 ]. A previous spatial study [ 76 ] also demonstrated the minimal alteration of the archaeological assemblage and the existence of a spatial association between the lithic tools and the faunal remains. The digitization of the site cartography was performed using Fig. 4 from Panagopoulou et al. [ 57 ], as it offered the highest quality among the publications available to date. The digitization of the cartography for Áridos-2 and Marathousa 1 followed the same process as the previously published modern elephant carcasses [ 8 , 58 ]. In the case of EAK, cartographic data were acquired during excavation; specimen centroids were recorded using a total station, and the polygons for each bone were subsequently digitized using orthophotographs. Spatial statistical analysis The spatial analysis was carried out using the comprehensive R [ 77 ] “spatstat” library [ 78 – 80 ]. This software has previously been used to analyze different archaeological contexts with satisfactory results for exploratory spatial point pattern distribution analyses [ 79 , 81 ], mainly seeking to determine the degree of randomness in point patterns and whether or not a correlation existed among different types of archaeological materials [ 1 , 82 – 94 ]. The first step of the study was to carry out an exploratory analysis of the spatial properties of each of the case studies. The methods used and the results are presented in the Supplementary Files. These initial analyses already showed that the techniques applied in conventional spatial studies are not sufficient to clearly differentiate the spatial patterns of the cases examined. Accordingly, we carried out a more in-depth study and information was collected on several spatial tests, indices and measurements and articulated within a hyperframe format. The use of hyperframes has already been shown to be highly useful for comparing spatial patterns in archaeological contexts [ 88 ]. The hyperframe was subsequently expanded to fit the taphonomic variables (carnivore impact and bone survival), so that exploratory (principal component analyses) and similarity-dissimilarity (phylogenetic hierarchical cluster) analyses could be carried out. The hyperframe used for the present analysis included a set of variables determined by first and second order tests, as well as topographic factorials. These are described in Table 2 . Table 2 Variables used in the hyperframe for the spatial statistical analysis and the subsequent cluster analysis on the PCA-transformed data set. Variable Description Linhom_ml Mean value of the estimated inhomogeneous intensity function evaluated at the observed point locations; reflects the average local density experienced by points. Linhom_mn Mean expected number of neighbors derived from the inhomogeneous L-function; summarizes local clustering relative to the estimated intensity. Linhom_mlg Mean value of the inhomogeneous intensity function over the entire observation window; represents global background density. Linhom_clust Ratio between local mean intensity and global mean intensity; values > 1 indicate concentration of points in high-intensity regions, values ≈ 1 indicate consistency with the background intensity. Finhom_nn Mean expected number of points within radius r based on the inhomogeneous empty-space (F) function; captures spacing relative to local intensity variation. pcf_index Summary deviation of the homogeneous pair correlation function from complete spatial randomness; larger values indicate stronger clustering or regularity. pcf_index2 Summary deviation of the inhomogeneous pair correlation function from randomness after correcting for spatial variation in intensity. hopskel Mean Hopkins–Skellam statistic across simulations; values 0.5 regularity. quadrat_stat Chi-square statistic from quadrat counts; large values indicate strong deviation from spatial randomness due to clustering or regular spacing. NND_index Mean nearest-neighbor distance scaled by the square root of local intensity; integrates spacing and density into a single metric. M_area Area of the union of discs around points at a fixed radius (Minkowski functional); reflects spatial footprint and point aggregation. M_perimeter Perimeter of the same union of discs; sensitive to boundary complexity and fragmentation of point clusters. M_euler Euler characteristic of the union of discs; captures topological structure (number of connected components minus holes), with lower values indicating clustering. scatter_area Area of the convex hull enclosing all points; represents overall spatial spread max_distance Maximum pairwise distance between points; reflects spatial extent of the pattern. mean_pairwise Mean pairwise distance among all point pairs; summarizes overall dispersion. To characterize the spatial intensity of each point pattern and assess the extent to which its distribution deviates from a completely random process, we first estimated the inhomogeneous intensity surface using a kernel-based estimator. Specifically, we computed a continuous density field from the point pattern using an optimal bandwidth selector (bw.ppl), which provides an estimate of intensity λ(x) that accounts for local spatial heterogeneities. This step is essential in contexts where intensity is not constant across the study area, as it prevents misinterpretations that would arise from assuming homogeneity when the underlying process is spatially variable. Based on this intensity estimate, we applied the inhomogeneous L-function, an extension of Ripley's L-function that explicitly corrects for spatial variation in λ(x). This function quantifies spatial dependence between points while controlling for non-uniform background intensity, thereby distinguishing true clustering from patterns produced solely by changes in the underlying intensity surface. For each pattern, we calculated the mean intensity at the exact point locations, providing a direct measure of the “effective density” experienced by the points themselves. We then estimated the expected number of neighbors under the inhomogeneous model by multiplying this mean local intensity by the values of the inhomogeneous L-function. To determine whether points tend to concentrate in areas of higher-than-average intensity, we compared this local intensity measure with the global mean intensity of the entire observation window. The ratio between these values (here referred to as the relative clustering index) quantifies whether the pattern preferentially occupies high-intensity regions (values > 1), whether its distribution is neutral with respect to λ(x) (≈ 1), or whether points tend to occur in lower-intensity regions (< 1). This index thus provides a robust, normalized measure of spatial aggregation, enabling meaningful comparisons between point patterns originating from different spatial contexts or exhibiting widely differing overall densities. In addition to the L-function analysis, we evaluated the spatial structure of each point pattern using the inhomogeneous empty-space F-function. Whereas the L-function describes the spatial relationships between the points themselves, the F-function, instead, characterizes the distribution of empty space relative to the points. This distinction provides a complementary perspective on spatial structure: while clustering of points may indicate aggregation, the empty-space function reveals how the surrounding landscape of voids is organized. To compute the inhomogeneous F-function, we again began by estimating a spatially varying intensity surface using the same kernel-based estimator and bandwidth selection procedure described above. The resulting intensity field λ(x) was used to correct the F-function for non-uniformity of the underlying process, ensuring that the probability of observing empty space at increasing distances from random locations is not confounded by local variations in density. Using this corrected version of the F-function, we derived an estimate of the expected number of points within a disk of radius r by multiplying the mean local intensity (evaluated at the observed point locations) by the area of the search region (πr²). Averaging these values across all distances r provided a single summary statistic representing the overall expected neighbor count in the inhomogeneous context. This measure serves as an intuitive indicator of how “filled” or occupied the space is when local intensity variations are taken into account. To express the degree to which the pattern tends toward high- or low-intensity regions, we again compared the mean intensity at the point locations with the global mean intensity across the observation window. The resulting ratio, analogous to the one used for the inhomogeneous L-function, quantifies whether points preferentially occur in regions of higher intensity (> 1), are distributed in proportion to the overall intensity surface (≈ 1), or disproportionately occupy low-intensity areas (< 1). This index enables direct comparison between patterns and offers insight into whether spatial aggregation is driven by genuine interaction processes or simply reflects the geometry of intensity variation. To further characterize the fine-scale structure of spatial interactions, we computed both the homogeneous and inhomogeneous pair correlation functions. Whereas the L- and F-functions evaluate cumulative departures from randomness across increasing spatial scales, the pair correlation function provides a localized measure of clustering or inhibition at specific distances. Conceptually, pair correlation quantifies the probability of finding a neighboring point at distance r relative to what would be expected under complete spatial randomness. Thus, values greater than 1 indicate clustering at that distance, while values below 1 reflect spatial inhibition or regular spacing. We first computed the standard (homogeneous) pair correlation function for each pattern. Because pair correlation is known to behave unstably at very small spatial lags (where edge effects and sampling noise can produce extreme or undefined values) we removed radii smaller than 2% of the maximum r-range. This trimming step ensured that the resulting summary statistic reflected meaningful spatial structure rather than numerical artifacts. We then summarized the pair correlation curve by computing the mean absolute deviation of pair correlation from the value expected under a Poisson process. This yielded a single index, the PCF clustering index, which quantifies the overall magnitude of deviation from spatial randomness across all evaluated scales. Because spatial patterns often exhibit inhomogeneous intensity, we also calculated the inhomogeneous pair correlation function using isotropic edge correction. This version of pair correlation accounts explicitly for spatially varying intensity λ(x), thereby isolating interaction effects from broader trends in density. As with the homogeneous case, we trimmed small radii here (excluding values below 5% of the radius range) to prevent instability in the estimator. The mean absolute deviation of the inhomogeneous pair correlation curve from 1 provided a second summary statistic reflecting interaction strength in the context of varying intensity. Together, the homogeneous and inhomogeneous pair correlation indices allow us to distinguish whether apparent spatial structure arises from genuine point–point interactions or merely reflects large-scale variation in the underlying intensity surface. The homogeneous index captures deviations without accounting for intensity variation, while the inhomogeneous index corrects for these trends, thus isolating true interaction processes. To complement the above second-order analyses, we implemented the Hopkins–Skellam statistic, a classical test for detecting departures from spatial randomness. Unlike Ripley’s functions or the pair correlation function, which focus on spatial dependence across a range of distances, the Hopkins–Skellam test evaluates the balance between clustering and regularity based on nearest-neighbor distances. The statistic compares the distances from randomly selected locations to the nearest observed point with the distances between observed points themselves. Under complete spatial randomness, these two distributions should be similar; deviations toward smaller observed distances indicate clustering, whereas larger distances reflect regularity. Because the Hopkins–Skellam statistic is inherently stochastic and sensitive to the particular random samples drawn, we estimated a more stable value by averaging across multiple repetitions. For each point pattern, we ran 99 independent iterations of the test and computed the mean statistic as the final index. Lower values (< 0.5) indicate a tendency toward clustering, while values approaching or exceeding 0.5 suggest greater regularity or randomness. This resampling-based implementation reduces the influence of sampling noise and yields a more reliable summary of the overall spatial structure. To further characterize large-scale deviations from spatial randomness, we calculated a quadrat-based chi-square statistic. Quadrat tests partition the observation window into a regular grid and compare the observed number of points in each cell with the numbers expected under a homogeneous Poisson process. Whereas second-order functions capture dependence across continuous distances, quadrat statistics are particularly effective in identifying coarse-grained aggregation or voids that manifest at the scale of the grid cells. Because the sensitivity of quadrat tests depends strongly on the chosen grid resolution, we adopted a data-driven approach in which the number of quadrats scales with the total number of points. Specifically, for each pattern we computed q = nq = \sqrt{n}q = n, where n is the number of points, and rounded this value to obtain both the number of quadrats along the x- and y-axes. This adaptive scheme maintains an approximate balance between the number of quadrats and the number of points in the pattern, avoiding over- or undersplitting the window. We then extracted the chi-square statistic returned by the test, which quantifies the degree to which the observed distribution of point counts deviates from spatial randomness. Larger values indicate stronger heterogeneity (either clustering or regular spacing) whereas values closer to zero are consistent with a homogeneous Poisson process. Although the chi-square approximation may be inaccurate when expected counts are low, using a grid proportional to the density of points helps mitigate this limitation and yields a more interpretable measure of large-scale spatial structure. To complement the quadrat analysis and second-order statistics, we computed an intensity-normalized nearest-neighbor distance index (NNDI). Nearest-neighbor distances are a classical metric for assessing spatial structure, but raw nearest-neighbor values can be misleading in the presence of strong intensity gradients. In regions of high point density, nearest-neighbor distances will naturally be smaller even if the pattern is random, while in sparse regions distances will be larger. To address this, we normalized each nearest-neighbor distance by the square root of the local intensity estimate. This adjustment follows theoretical expectations under an inhomogeneous Poisson process, where the expected nearest-neighbor distance scales inversely with the square root of the intensity. For each point in the pattern, we first computed the Euclidean distance to its nearest neighbor. We then extracted the corresponding local intensity values from the kernel intensity estimator and multiplied each distance by the square root of its associated intensity. The resulting dimensionless NNDI values quantify how “close” or “far” points are relative to what would be expected given the local density of the pattern. We then averaged these normalized values across all points to derive a single index per pattern. Low values indicate clustering relative to local intensity, values near one approximate spatial randomness, and high values suggest local inhibition or regularity. This index thus offers a robust, first-order–adjusted view of local spatial structure that is not confounded by spatially varying point densities. To quantify higher-order spatial structure beyond classical point-pattern measures, we computed a set of Minkowski functionals. These functionals describe the geometry and topology of a point pattern after morphological dilation, and they are widely used in fields such as cosmology, materials science, and ecology because they summarize global shape properties in ways that are sensitive to clustering and void structure. Applied to archaeological and paleoanthropological spatial datasets, Minkowski functionals provide a complementary descriptor of spatial organization, capturing aspects of point aggregation that are not easily detected using local statistics or second-order functions. To compute these metrics, we first established an appropriate range of dilation radii. The choice of radius is critical because the morphological expansion of points must reflect meaningful physical distances relative to the size of the observation window. Very small radii amplify pixel noise and exaggerate Euler characteristic fluctuations, while excessively large radii saturate the window and erase meaningful structure. We therefore determined radius values adaptively based on the spatial dimensions of the study area and on the mean nearest-neighbor distance within each pattern. Following standard morphological analysis practice, radii were automatically set between 5% and 40% (depending on the point pattern properties) of the smallest window dimension but were also constrained to avoid radii that were either far smaller or far larger than typical interpoint distances. This dual-criterion approach ensured that Minkowski functionals remained sensitive to real structural differences rather than artifacts of scale mismatch. We also implemented an automated procedure to recommend radii and image resolution parameters for each point pattern, ensuring consistent spatial resolution across datasets of differing sizes and densities. Once radius and resolution parameters were adaptively defined for each point process, we constructed a high-resolution distance map over the observation window, which computes for every pixel the distance to its nearest point. This raster representation enabled the creation of a binary mask identifying all pixels lying within a distance r of any point, effectively creating a union of disks around the points. Minkowski functionals were then computed directly from this binary mask. The area functional quantifies the total dilated surface occupied by the union of disks, reflecting overall clustering and spread. The perimeter functional, estimated by counting transitions between foreground and background pixels along both horizontal and vertical edges, measured boundary complexity and the fragmentation of occupied space. Finally, the Euler characteristic was computed using a digital topology estimator based on 2×2 pixel blocks. This measure captures the number of connected components minus the number of holes in the dilated region and is particularly sensitive to changes in spatial topology as the radius increases. To mitigate well-known instabilities in Euler characteristic estimation (especially at high resolution or when disk radii approach the pixel size) we implemented two safeguards: (1) moderating resolution to avoid excessive pixel-level fragmentation, and (2) using a 4-connectivity estimation approach that is stable for binary morphological objects. These strategies prevented pathological behavior and ensured that Minkowski values varied appropriately across different spatial patterns. The Minkowski metrics derived for each point pattern (area, perimeter, and Euler characteristic) provided complementary information on spatial structure at a scale explicitly tied to the spatial extent of the underlying archaeological surfaces. These functionals allowed us to capture differences in clustering intensity, boundary roughness, and topological complexity that were not detectable using more traditional spatial statistics. By integrating these geometric descriptors with first- and second-order point-pattern measures, we obtained a comprehensive view of spatial structure across our comparative dataset. To further characterize the spatial extent and overall occupied footprint of each assemblage, we calculated several global measures based on the geometric configuration of the points. First, we computed the convex hull of each pattern, which represents the smallest convex polygon that encloses all points. The area of this hull provides a robust estimate of the total spatial footprint of the assemblage that is minimally influenced by internal voids or irregularities. Unlike density-based measures, which depend on local point concentrations, the convex-hull area captures the overall spread and territorial extent of the distribution, offering an integrative measure of how widely dispersed or spatially constrained a point pattern is within its sampling window. In addition to the hull, we quantified the maximum diameter of each point pattern by computing the full matrix of pairwise interpoint distances and extracting the greatest value. This diameter corresponds to the largest linear separation between any two points and serves as a simple but informative descriptor of the maximum spatial reach of the assemblage. Because this measure is independent of local clustering or density heterogeneity, it provides a complementary indicator of large-scale dispersion compared to area-based metrics. Finally, we calculated the mean pairwise distance among all unique point pairs. This statistic captures the average spatial separation across the entire pattern and is sensitive to both clustering and broad-scale spread. While the maximum diameter highlights extreme separation, and the convex-hull area defines the spatial envelope, the mean pairwise distance describes the typical spacing among points throughout the distribution. Taken together, these three global measures (convex-hull footprint, maximum interpoint distance, and mean pairwise separation) characterize the overall spatial envelope and extent of each assemblage, providing a coarse-grained spatial signature that complements the more detailed first-, second-, and higher-order statistics described above. Once the hyperframe was built, we converted it into a conventional data frame for multivariate statistics (first using only the spatial variables, and then adding the taphonomic variables) excluding the site identifier column so that only numeric variables informed the clustering. We then applied a principal component analysis (PCA) to modulate potential collinearity effects. We computed Euclidean distances between sites based on all standardized spatial descriptors, producing subsequently a distance matrix that quantified how different each site was from all others in multidimensional spatial-pattern space. This matrix allowed us to visualize dissimilarities using a distance heatmap, thereby revealing the relative closeness or divergence of sites across the full spatial-statistics profile. We then applied hierarchical agglomerative clustering to the distance matrix. Average linkage was used for the analysis, which merges clusters based on the mean intercluster distance. Dendrograms were generated to visualize the hierarchical structure of similarities among sites, with both conventional and phylogram-style graphics produced to facilitate interpretation. We also created circular dendrograms with cluster highlighting to illustrate potential grouping structure and to evaluate whether sites naturally segregated into distinct spatial-behavioral categories. Together, the hyperframe construction and the hierarchical clustering workflow served to integrate many independent spatial statistics into a unified inferential framework. This allowed us to assess whether sites shared common spatial structuring processes or instead exhibited distinct spatial signatures suggestive of different depositional, behavioral, or taphonomic dynamics. The overall hypothesis to be tested is that proboscidean carcasses highly impacted by carnivores will have a higher degree of bone dispersal (affecting bone survival percentages), affecting the spatial point patterns (i.e., either decreasing clustering or increasing the spatial window or both). Such a taphonomic factor should have a distinctive spatial point pattern associated. In order to evaluate this hypothesis, a three-stage analytical process was adopted: a) The first approach involved the use of modern and archaeological carcasses together. This was done as an initial exploratory step, in which some cautions were contemplated. The main goal was to understand how modern carcasses documented to exhibit various degrees of carnivore impact generate spatial pattern ranges within which the archaeological samples could be interpreted. b) The second approach involved only those modern carcases for which a taphonomic assessment is available, with the main goal being understanding the post-depositional disturbance processes, regardless of their scale. Archaeological samples may be palimpsestic and affected by processes not documented in modern carcasses. For this reason, the overall phylogenetic cladogram of all the samples may be impacted by the archaeological samples. To remove their impact and better understand the interplay of taphonomic and spatial patterns, we restricted the analysis to the modern carcasses. Thus, overall variance is caused by immediate postdepositional processes; namely, the impact of carnivore disturbance. This second approach aims to capture all the information within a reasonable radius around the main carcass cluster and is, therefore, conditioned by spatial windows of variable sizes. c) The third approach tried to minimize the impact of the wide range of window sizes and, instead of focusing on reconstructing the association of spatial patterns and taphonomic processes, it intends to realistically approach proboscidean archaeological sites, which normally span much smaller windows than all the experimental/observational spatial windows used as analogical referents. The variables that we applied to the modern and archaeological point patterns may be affected by the heterogeneous scales, and diverse expressions of agglomerative/scattering patterns involved in the first approach. This is necessary to understand the modifying taphonomic processes, which operate at differential scales. To standardize spatial comparisons across assemblages and minimize the confounding effects of heterogeneous observation window sizes, we implemented a procedure to isolate the densest portion of each point pattern. Archaeological proboscidean butchery sites are much smaller than the areas surveyed for modern elephant carcass dispersal. At the most fundamental level, the observation window defines the spatial frame within which intensity, distances, and spatial relationships are measured. Changing window size alters global intensity estimates (λ), which directly affects Linhom, Finhom, NND-based indices, and any statistic that rescales distances by intensity. A larger window with the same number of points lowers global intensity and tends to inflate apparent clustering, whereas a smaller or tighter window raises intensity and can make the same point pattern appear more regular or dispersed. This means that absolute values of Linhom-based clustering indices, Finhom-derived neighbor expectations, and NND indices are not window-invariant and must be interpreted comparatively only among assemblages analyzed with comparable window definitions. Window size also strongly conditions second-order statistics such as the pair correlation function and quadrat tests. Quadrat statistics scale with both the number of quadrats and the spatial extent of the window, so larger windows almost inevitably produce larger χ² values even if the underlying pattern structure is similar. Likewise, the maximum meaningful distance for PCF interpretation depends on window extent; expanding the window introduces larger inter-point distances and can dilute short-range clustering signals relative to background space. This is why trimming small radii and focusing on relative deviations from CSR, rather than absolute magnitudes, is essential when comparing sites. The influence of window size is especially critical for Minkowski functionals. Area, perimeter, and Euler characteristic are explicitly geometric and therefore scale with window extent, resolution, and edge effects. Larger windows increase the number of connected components and holes captured at a given radius, potentially exaggerating Euler values and footprint size even when local clustering is unchanged. Your adaptive-resolution strategy mitigates pixel instability, but it does not remove the fundamental dependence of Minkowski measures on the spatial domain. Consequently, these metrics are most reliable for within-study comparisons where window construction reflects real site boundaries (e.g., carcass scatter limits, excavation extents) rather than arbitrary clipping. This is why we emphasized the study of the modern referents over the three archaeological assemblages. The procedure that we implemented to overcome this scale-induced limitation is encapsulated in a function, which identifies and extracts a fixed-area window centered on the highest-density region of a spatial point process. First, a kernel density estimation (KDE) of the point pattern was computed with automatic bandwidth selection. This adapts the smoothing parameter to the spatial configuration of points. This step produces a continuous intensity surface representing the estimated local density of points across the observation window. Second, the location of maximum estimated intensity is identified by finding the pixel with the highest density value on the KDE surface. This location represents the spatial peak of point concentration and is interpreted as the center of the densest cluster within the pattern. The x- and y-coordinates of this peak are extracted to serve as the reference point for window definition. Third, a square observation window of fixed dimensions (10 m × 10 m, corresponding to 100 m²) is defined, centered on the density peak. The window is constructed symmetrically around the peak coordinates by extending half the side length in all directions. This standardized window size ensures that all point patterns are analyzed within an equivalent spatial extent, allowing direct comparison of spatial statistics across datasets. Finally, the original point pattern is cropped to this newly defined window using spatial subsetting. Only points falling within the standardized high-density window are retained, producing a trimmed point pattern that captures the core spatial structure of the assemblage while excluding low-density peripheral areas. This approach allows spatial analyses to focus explicitly on the most informative regions of point concentration while controlling for differences in original window size and shape. By anchoring comparisons to equivalent, density-centered spatial extents, the method improves the robustness of cross-assemblage comparisons of clustering, dispersion, and other spatial characteristics. Declarations Code availability The code used is available as a supplementary file. Data availability statement All data generated or analyzed during this study are included in this published article (and its Supplementary Information files) Author contributions statement MDR and EB designed research and collected data. MDR and AM coded data. AM and JAC used photogrammetric software to transfer original data. MDR wrote the first version of the manuscript. All authors reviewed the manuscript. Additional information Competing interests: The author(s) declare no competing interests. Funding This research was conducted in Botswana with funding from IDEA. Archaeological work at EAK (Olduvai Gorge, Tanzania) was funded by the Spanish Ministry of Science and Innovation (PID2023-146260NB-C2), and the Spanish Ministry of Culture through the program of Archaeology Abroad. Author Contribution MDR and EB designed research and collected data. MDR and AM coded data. AM and JAC used photogrammetric software to transfer original data. MDR wrote the first version of the manuscript. All authors reviewed the manuscript. Acknowledgements We thank the Commission for Science and Technology (COSTECH), the Ngorongoro Conservation Area Authorities (NCAA), the Division of Antiquities, and the Tanzanian Ministry of Natural Resources and Tourism for their permission to conduct research in Tanzania. AM is funded by a postdoctoral grant from the Fyssen Foundation. We are extremely thankful to Eduardo Méndez-Quintas and Elia Organista, for their help during fieldwork. References Domínguez-Rodrigo, M. et al. Earliest Evidence of Elephant Butchery at Olduvai Gorge (Tanzania) Reveals the Evolutionary Impact of Early Human Megafaunal Exploitation. eLife 14, RP108298 (2026). Agam, A. & Barkai, R. Not the brain alone: The nutritional potential of elephant heads in Paleolithic sites. Quatern. Int. 406 , 218–226 (2016). Byers, D. A. & Ugan, A. Should we expect large game specialization in the late Pleistocene? An optimal foraging perspective on early Paleoindian prey choice. J. Archaeol. Sci. 32 , 1624–1640 (2005). Blumenschine, R. J. & Cavallo, J. A. Scavenging and Human Evolution. Sci. Am. 267 , 90–97 (1992). Marean, C. W. Sabertooth cats and their relevance for early hominid diet and evolution. J. Hum. Evol. 18 , 559–582 (1989). Selvaggio, M. M. Concerning the three stage model of carcass processing at FLK Zinjanthropus : A reply to Capaldo. J. Hum. Evol. 35 , 313–315 (1998). Domínguez-Rodrigo, M. & Baquedano, E. The Ecology of an Elephant Kill: Lion Defleshing of an African Elephant in the Okavango Delta (Botswana), and Limited Kleptoparasitic Opportunities for Meat-Eating Scavengers. Afr. J. Ecol. 63 , e70079 (2025). White, P. A. & Diedrich, C. G. Taphonomy story of a modern African elephant Loxodonta africana carcass on a lakeshore in Zambia (Africa). Quatern. Int. 276–277 , 287–296 (2012). Organista, E. et al. Did Homo erectus kill a Pelorovis herd at BK (Olduvai Gorge)? A taphonomic study of BK5. Archaeol. Anthropol. Sci. 8 , 601–624 (2016). Villa, P. & Torralba Aridos: elephant exploitation in Middle Pleistocene Spain. J. Hum. Evol. 19 , 299–309 (1990). Villa, P. et al. New data from Ambrona: closing the hunting versus scavenging debate. Quatern. Int. 126–128 , 223–250 (2005). Villa, P. et al. Nuevos datos sobre Ambrona: cerrando el debate caza versus carroñeo. Zona Arqueológica . 5 , 352–381 (2005). Agam, A. & Barkai, R. Elephant and Mammoth Hunting during the Paleolithic: A Review of the Relevant Archaeological, Ethnographic and Ethno-Historical Records. Quaternary 1 , 3 (2018). Altamura, F., Gaudzinski-Windheuser, S., Melis, R. T. & Mussi, M. Reassessing Hominin Skills at an Early Middle Pleistocene Hippo Butchery Site: Gombore II-2 (Melka Kunture, Upper Awash valley, Ethiopia). J. Paleo Arch. 3 , 1–32 (2020). Barkai, R. W. E. R. & Asia The Significance of Proboscideans in Diet, Culture and Cosmology in Paleolithic Asia. in Animals and Human Society in Asia: Historical, Cultural and Ethical Perspectives (eds (eds Kowner, R., Bar-Oz, G., Biran, M., Shahar, M. & Shelach-Lavi, G.) 33–62 (Springer International Publishing, Cham, doi: 10.1007/978-3-030-24363-0_2 . (2019). Ben-Dor, M. & Barkai, R. The importance of large prey animals during the Pleistocene and the implications of their extinction on the use of dietary ethnographic analogies. J. Anthropol. Archaeol. 59 , 101192 (2020). Berthelet, A. & Chavaillon, J. The early Palaeolithic butchery site of Barogali (Republic of Djibouti). in The World of Elephants. Proceedings of the First International Congress 176–179Consiglio Nazionale delle Ricerche, Rome, (2001). Boschian, G. & Saccà, D. In the elephant, everything is good: Carcass use and re-use at Castel di Guido (Italy). Quatern. Int. 361 , 288–296 (2015). Chavaillon, J. et al. Le site de dépecage pléistocène à Elephas recki de Barogali (République de Djibouti): nouveaux résultats et datation. Comptes Rendus de l’Académie des. Sci. 305 , 1259–1266 (1987). Chavaillon, J. & Berthelet, A. The archaeological sites of Melka Kunture. in Studies on the Early Paleolithic site of Melka Kunture, Ethiopia (eds (eds Chavaillon, J. & Piperno, M.) 25–80 (Instituto Italiano di Preistoria e Protostoria, Florence, (2004). Delagnes, A. et al. Interpreting pachyderm single carcass sites in the African Lower and Early Middle Pleistocene record: A multidisciplinary approach to the site of Nadung’a 4 (Kenya). J. Anthropol. Archaeol. 25 , 448–465 (2006). Espigares, M. P. et al. Homo vs. Pachycrocuta : Earliest evidence of competition for an elephant carcass between scavengers at Fuente Nueva-3 (Orce, Spain). Quaternary International 295, 113–125 (2013). Gaudzinski, S. et al. The use of Proboscidean remains in every-day Palaeolithic life. Quatern. Int. 126–128 , 179–194 (2005). Gaudzinski-Windheuser, S., Kindler, L., MacDonald, K. & Roebroeks, W. Hunting and processing of straight-tusked elephants 125.000 years ago: Implications for Neanderthal behavior. Sci. Adv. 9 , eadd8186 (2023). Gaudzinski-Windheuser, S., Kindler, L. & Roebroeks, W. Widespread evidence for elephant exploitation by Last Interglacial Neanderthals on the North European plain. Proceedings of the National Academy of Sciences 120, e2309427120 (2023). Goren-Inbar, N., Lister, A., Werker, E. & Chech, M. A Butchered elephant skull and associated artifacts from the Acheulian site of Gesher Benot Ya’Aqov. Isr. Paléorient . 20 , 99–112 (1994). Hannus, L. A. Clovis Mammoth Butchery: The Lange/Ferguson Site and Associated Bone Tool Technology (Texas A&M University, 2018). Haynes, G. Late Quaternary Proboscidean Sites in Africa and Eurasia with Possible or Probable Evidence for Hominin Involvement. Quaternary 5 , 18 (2022). Human-Elephant Interactions. From Past to Present (Tübingen University, 2021). Leakey, M. D., Olduvai & Gorge Excavations in Beds I & II 1960–1963 Vol. 3 (Cambridge University Press, 1971). Lemorini, C., Santucci, E., Caricola, I., Nucara, A. & Nunziante-Cesaro, S. Life Around the Elephant in Space and Time: an Integrated Approach to Study the Human-Elephant Interactions at the Late Lower Paleolithic Site of La Polledrara di Cecanibbio (Rome, Italy). J. Archaeol. Method Theory . 30 , 1233–1281 (2023). Lev, M. & Barkai, R. Elephants are people, people are elephants: Human–proboscideans similarities as a case for cross cultural animal humanization in recent and Paleolithic times. Quatern. Int. 406 , 239–245 (2016). Mielgo, C., Venter, S., Bosch, A., Huguet, R. & Venter, J. A. Feeding on giants: Neotaphonomic evidence of hyena exploitation of megafaunal skulls and its implications for Pleistocene archaeology. J. Archaeol. Sci. 186 , 106458 (2026). Mosquera, M. et al. Barranc de la Boella (Catalonia, Spain): an Acheulean elephant butchering site from the European late Early Pleistocene. J. Quat. Sci. 30 , 651–666 (2015). Mussi, M. & Villa, P. Single carcass of Mammuthus primigenius with lithic artifacts in the Upper Pleistocene of northern Italy. J. Archaeol. Sci. 35 , 2606–2613 (2008). Pineda, A. et al. Changing competition dynamics among predators at the late Early Pleistocene site Barranc de la Boella (Tarragona, Spain). Palaeogeogr., Palaeoclimatol. Palaeoecol. 477 , 10–26 (2017). Reshef, H. & Barkai, R. A taste of an elephant: The probable role of elephant meat in Paleolithic diet preferences. Quatern. Int. 379 , 28–34 (2015). Rocca, R., Boschin, F. & Aureli, D. Around an elephant carcass: Cimitero di Atella and Ficoncella in the behavioural variability during the Early Middle Pleistocene in italy. in Human-elephant interactions: from past to present (eds Konidaris, G. E., Barkai, R., Tourloukis, V. & Harvati, K.) (Tübingen University Press, (2021). https://dx.doi.org/10.15496/publikation-55591 Saccà, D. Taphonomy of Palaeloxodon antiquus at Castel di Guido (Rome, Italy): Proboscidean carcass exploitation in the Lower Palaeolithic. Quatern. Int. 276–277 , 27–41 (2012). Santonja Gómez, M., López, N. & Pérez-González, A. Acheulean Occupation Sites in the Jarama Valley (Madrid, Spain). Curr. Anthropol. 19 , 394–395 (1978). Santonja, M., López Martínez, N., Querol, M. Á. & Soto, E. Estudio de las dispersiones de industria lítica y de los restos de vertebrados en Áridos-1 (Arganda, Madrid). in Ocupaciones achelenses en el valle del Jarama (Arganda, Madrid). Arqueología y Paleoecología (eds Santonja, M., López Martínez, N. & Pérez-González, A.) 279–295 (Diputación Provincial de Madrid, Madrid, 1980). Santucci, E. et al. Palaeoloxodon exploitation at the Middle Pleistocene site of La Polledrara di Cecanibbio (Rome, Italy). Quaternary International 406, 169–182 (2016). Solodenko, N. et al. Fat Residue and Use-Wear Found on Acheulian Biface and Scraper Associated with Butchered Elephant Remains at the Site of Revadim, Israel. PLOS ONE . 10 , e0118572 (2015). Yravedra, J. et al. Not seen before. Unveiling depositional context and Mammuthus meridionalis exploitation at Fuente Nueva 3 (Orce, southern Iberia) through taphonomy and microstratigraphy. Q. Sci. Rev. 329 , 108561 (2024). Yravedra, J., Rubio-Jara, S., Panera, J. & Martos, J. A. Hominins and Proboscideans in the Lower and Middle Palaeolithic in the Central Iberian Peninsula. Quatern. Int. 520 , 140–156 (2019). Yravedra, J. et al. Neanderthal and Mammuthus interactions at EDAR Culebro 1 (Madrid, Spain). J. Archaeol. Sci. 42 , 500–508 (2014). Yravedra, J., Rubio-Jara, S., Panera, J., Uribelarrea, D. & Pérez-González, A. Elephants and subsistence. Evidence of the human exploitation of extremely large mammal bones from the Middle Palaeolithic site of PRERESA (Madrid, Spain). J. Archaeol. Sci. 39 , 1063–1071 (2012). Yravedra, J. et al. Cut marks on the Middle Pleistocene elephant carcass of Áridos 2 (Madrid, Spain). J. Archaeol. Sci. 37 , 2469–2476 (2010). Haynes, G. Mammoths, mastodonts, and elephants. Biology, behavior, and the fossil record (Cambridge University Press, 1991). Haynes, G. Longitudinal studies of african elephant death and bone deposits. J. Archaeol. Sci. 15 , 131–157 (1988). Haynes, G., Krasinski, K. & Wojtal, P. Elephant bone breakage and surface marks made by trampling elephants: Implications for interpretations of marked and broken Mammuthus spp. bones. J. Archaeol. Science: Rep. 33 , 102491 (2020). Haynes, G. & Hutson, J. African elephant bones modified by carnivores: Implications for interpreting fossil proboscidean assemblages. J. Archaeol. Science: Rep. 34 , 102596 (2020). Behrensmeyer, A. K. Taphonomic and Ecologic Information from Bone Weathering. Paleobiology 4 , 150–162 (1978). Haynes, G. & Wojtal, P. Weathering Stages of Proboscidean Bones: Relevance for Zooarchaeological Analysis. J Archaeol Method Theory (2022). https://doi.org/10.1007/s10816-022-09569-3 doi:10.1007/s10816-022-09569-3. Domínguez-Rodrigo, M. et al. Fluvial spatial taphonomy: a new method for the study of post-depositional processes. Archaeol. Anthropol. Sci. 10 , 1769–1789 (2018). Santonja, M. & Querol, M. Á. El sitio de ocupación achelense de Áridos-2. Descripción general y estudio de la industria lítica. in Ocupaciones achelenses en el valle del Jarama (Arganda, Madrid). Arqueología y Paleoecología (eds Santonja, M., López Martínez, N. & Pérez-González, A.) 297–306 (Diputación Provincial de Madrid, Madrid, (1980). Panagopoulou, E. et al. The Lower Palaeolithic site of Marathousa 1, Megalopolis, Greece: Overview of the evidence. Quatern. Int. 497 , 33–46 (2018). Haynes, G. Las acumulaciones modernas de huesos de elefante como modelo para interpretar Ambrona y otras áreas con fauna fósil a orillas del agua. Zona Arqueológica . 5 , 154–175 (2005). La Cotte de St. Brelade 1961–1978. Excavations by C. B. M. McBurney (Geo Books, 1986). Rus, I. & Vega Toscano, G. El yacimiento de Arriaga II: problemas de una definición actual de los suelos de ocupación. in Primeras Jornadas de Metodología e Investigación Prehistórica (Soria ,) 387–404 (Ministerio de Cultura, Soria, 1984).) 387–404 (Ministerio de Cultura, Soria, 1984). (1981). Hay, R. L. Geology of the Olduvai Gorge: A Study of Sedimentation in a Semiarid Basin (University of California Press, 1976). Deino, A. L. 40Ar/39Ar dating of Bed I, Olduvai Gorge, Tanzania, and the chronology of early Pleistocene climate change. J. Hum. Evol. 63 , 251–273 (2012). Santonja, M., López Martínez, N. & Pérez-González, A. Ocupaciones achelenses en el valle del Jarama (Arganda-Madrid) (Diputación Provincial de Madrid, 1980). Pérez-González, A. El marco geográfico, geológico y geomorfológico de los yacimientos de Áridos en la cuenca del Tajo. in Ocupaciones achelenses en el valle del Jarama (Arganda, Madrid). Arqueología y Paleoecología (eds Santonja, M., López Martínez, N. & Pérez-González, A.) 15–28 (Diputación Provincial de Madrid, Madrid, (1980). Pérez-González, A. & Uribelarrea, D. Geología del Cuaternario de los valle fluviales del Jarama y Manzanares en las proximidades de Madrid. Zona Arqueológica . 1 , 302–317 (2002). Moreno, D. et al. ESR dating of Middle Pleistocene archaeo-paleontological sites from the Manzanares and Jarama river valleys (Madrid basin, Spain). Quatern. Int. 520 , 23–38 (2019). de Juana, S., Galán, A. B. & Domínguez-Rodrigo, M. Taphonomic identification of cut marks made with lithic handaxes: an experimental study. J. Archaeol. Sci. 37 , 1841–1850 (2010). Ollé, A. Universitat Rovira i Virgili, Tarragona,. Variabilitat i patrons funcionals en els sistemes tècnics de mode 2. Anàlisi de les deformacions d’ús en els conjunts lítics del Riparo Esterno de Grotta Paglicci (Rigano Garganico, Foggia), Aridos (Arganda, Madrid) i Galeria-TN (Atapuerca, Burgos). TDX (Tesis Doctorals en Xarxa) (2003). Blackwell, B. A. B. et al. ESR Dating Ungulate Teeth and Molluscs from the Paleolithic Site Marathousa 1, Megalopolis Basin, Greece. Quaternary 1, (2018). Jacobs, Z. et al. Optical dating of K-feldspar grains from Middle Pleistocene lacustrine sediment at Marathousa 1 (Greece). Quatern. Int. 497 , 170–177 (2018). Tourloukis, V. et al. Magnetostratigraphic and chronostratigraphic constraints on the Marathousa 1 Lower Palaeolithic site and the Middle Pleistocene deposits of the Megalopolis basin, Greece. Quatern. Int. 497 , 154–169 (2018). Karkanas, P. et al. Sedimentology and micromorphology of the Lower Palaeolithic lakeshore site Marathousa 1, Megalopolis basin, Greece. Quatern. Int. 497 , 123–136 (2018). Tourloukis, V. et al. Lithic artifacts and bone tools from the Lower Palaeolithic site Marathousa 1, Megalopolis, Greece: Preliminary results. Quatern. Int. 497 , 47–64 (2018). Konidaris, G. E. et al. The skeleton of a straight-tusked elephant ( Palaeoloxodon antiquus ) and other large mammals from the Middle Pleistocene butchering locality Marathousa 1 (Megalopolis Basin, Greece): preliminary results. Quatern. Int. 497 , 65–84 (2018). Michailidis, D., Konidaris, G. E., Athanassiou, A., Panagopoulou, E. & Harvati, K. The ornithological remains from Marathousa 1 (Middle Pleistocene; Megalopolis Basin, Greece). Quatern. Int. 497 , 85–94 (2018). Giusti, D. et al. Beyond maps: Patterns of formation processes at the Middle Pleistocene open-air site of Marathousa 1, Megalopolis basin, Greece. Quatern. Int. 497 , 137–153 (2018). R Core Team. R: A language and environment for statistical computing (R Foundation for Statistical Computing, 2022). Baddeley, A., Rubak, E. & Turner, R. The Impact of R on Statistical Science for Spatial Point Processes. Australian New. Z. J. Stat. 68 , e70037 (2026). Baddeley, A., Rubak, E. & Turner, R. Spatial Point Patterns: Methodology and Applications with R (CRC, 2016). Baddeley, A. & Turner, R. spatstat: An R Package for Analyzing Spatial Point Patterns. J. Stat. Softw. 12 , 1–42 (2005). Wiegand, T. & Moloney, K. A. Handbook of Spatial Point-Pattern Analysis in Ecology (Chapman and Hall/CRC, 2013). 10.1201/b16195 Arteaga Brieba, A. Universitat Rovira i Virgili,. Among stones and bison. The lithic assemblage of Gran Dolina TD10.2 (Atapuerca). Technological and spatial implications of a specialised Middle Pleistocene Kill-butchering site. TDX (Tesis Doctorals en Xarxa) (2024). Cobo-Sánchez, L., Domínguez-Rodrigo, M., Mabulla, A. & Baquedano, E. Chapter 5 - Reconstructing early human behavior through the in-site spatial statistical analysis of DS. in Reconstructing Olduvai (ed (eds Domínguez-Rodrigo, M. et al.) 197–240 (Academic, doi: 10.1016/B978-0-443-27382-7.00007-1 . (2024). Diez-Martín, F. et al. Tracing the spatial imprint of Oldowan technological behaviors: A view from DS (Bed I, Olduvai Gorge, Tanzania). PLOS ONE . 16 , e0254603 (2021). Domínguez-Rodrigo, M. et al. Spatial simulation and modelling of the early Pleistocene site of DS (Bed I, Olduvai Gorge, Tanzania): a powerful tool for predicting potential archaeological information from unexcavated areas. Boreas 46 , 805–815 (2017). Luzón, C. et al. Taphonomic and spatial analyses from the Early Pleistocene site of Venta Micena 4 (Orce, Guadix-Baza Basin, southern Spain). Sci. Rep. 11 , 13977 (2021). Marín, J. et al. Neanderthal logistic mobility during MIS3: Zooarchaeological perspective of Abric Romaní level P (Spain). Q. Sci. Rev. 225 , 106033 (2019). Merino-Pelaz, A., Cobo-Sánchez, L., Organista, E. & Baquedano, E. Domínguez-Rodrigo, M. Unraveling the spatial imprint of hominin and carnivore accumulations in Early Pleistocene African sites. Archaeol. Anthropol. Sci. 16 , 128 (2024). Mielgo, C. et al. Intra-site spatial approaches based on taphonomic analyses to characterize assemblage formation at Pleistocene sites: a case study from Buena Pinta Cave (Pinilla del Valle, Madrid, Spain). Archaeol. Anthropol. Sci. 16 , 1–28 (2024). Moclán, A. et al. Identifying activity areas in a neanderthal hunting camp (the Navalmaíllo Rock Shelter, Spain) via spatial analysis. Archaeol. Anthropol. Sci. 15 , 44 (2023). Moclán, A. et al. Spatial analysis of an Early Middle Palaeolithic kill/butchering site: the case of the Cuesta de la Bajada (Teruel, Spain). Archaeol. Anthropol. Sci. 15 , 91 (2023). Panera, J. et al. Assessing functionality during the early Acheulean in level TKSF at Thiongo Korongo site (Olduvai Gorge, Tanzania). Quatern. Int. 526 , 77–98 (2019). Saladié, P. et al. Dragged, lagged, or undisturbed: reassessing the autochthony of the hominin-bearing assemblages at Gran Dolina (Atapuerca, Spain). Archaeol. Anthropol. Sci. 13 , 65 (2021). Villaescusa, L. et al. Towards a formation model of the Neanderthal symbolic accumulation of herbivore crania: Spatial patterns shaped by rockfall dynamics in Level 3 of Des-Cubierta Cave (Lozoya valley, Madrid, Spain). Archaeol. Anthropol. Sci. 18 , 16 (2026). Additional Declarations No competing interests reported. Supplementary Files SupplementaryFiles.pdf DominguezRodrigoetalProboscideanspatialanalysis.r 01SpatialWINDOWS.csv 02SpatialDatasets.csv Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8799314","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":593058645,"identity":"f36cb756-f12c-4744-b263-57f98cb15342","order_by":0,"name":"Manuel DOMÍNGUEZ-RODRIGO","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAApklEQVRIiWNgGAWjYFCC/I8PPjBIgJkSRGpJMDacQaoWM2keKJM4LbrtCcnGtnss5PkZmA/e5iGsgYHB7MyDg49znkkYzmxgS7YmTsuNxGbjnAMSCQYHeBAuJKAlmU3aAqjF/gD/N2K1pLFJM4BsYeBhI1LLmTfMhj0HJAxnHGYztpxDlJbjOYwPfhyok+dvb3544w0xWhCAmTTlo2AUjIJRMArwAQC6VS1l+COapQAAAABJRU5ErkJggg==","orcid":"","institution":"Rice University","correspondingAuthor":true,"prefix":"","firstName":"Manuel","middleName":"","lastName":"DOMÍNGUEZ-RODRIGO","suffix":""},{"id":593058646,"identity":"03e88d05-cfd0-45b3-b161-09089db8b31e","order_by":1,"name":"Abel MOCLÁN","email":"","orcid":"","institution":"UMR 7262 CNRS and Université de Poitiers","correspondingAuthor":false,"prefix":"","firstName":"Abel","middleName":"","lastName":"MOCLÁN","suffix":""},{"id":593058648,"identity":"03504542-248a-4a7c-bd21-071f16387b20","order_by":2,"name":"José Ángel CORREA CANO","email":"","orcid":"","institution":"Archaeological and Paleontological Museum of Madrid and Rice University","correspondingAuthor":false,"prefix":"","firstName":"José","middleName":"Ángel CORREA","lastName":"CANO","suffix":""},{"id":593058651,"identity":"07daa9e4-6519-4ad4-845a-ec752b3bca96","order_by":3,"name":"Enrique BAQUEDANO","email":"","orcid":"","institution":"Archaeological and Paleontological Museum of Madrid and Rice University","correspondingAuthor":false,"prefix":"","firstName":"Enrique","middleName":"","lastName":"BAQUEDANO","suffix":""}],"badges":[],"createdAt":"2026-02-05 16:24:41","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8799314/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8799314/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103238072,"identity":"60b695c2-1aa0-46ca-9ed3-ec194dde8d4d","added_by":"auto","created_at":"2026-02-23 13:37:40","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":853406,"visible":true,"origin":"","legend":"\u003cp\u003eDifferent types of furrowing and bone deletion caused by hyenas on modern elephant femora (Ndutu, Tanzania). Damage can be as intense as to make complete epiphyseal sections disappear.\u003c/p\u003e","description":"","filename":"floatimage16.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/ee10ca42fb362f6390ec781b.png"},{"id":103505784,"identity":"8e5b547a-0ead-4419-b903-669dc9bbce27","added_by":"auto","created_at":"2026-02-26 13:33:00","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":88257,"visible":true,"origin":"","legend":"\u003cp\u003e2\u003c/p\u003e","description":"","filename":"floatimage21.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/40d5b03ee4c5a1ce09d73a18.png"},{"id":103238609,"identity":"7ca93c6f-0ad6-4eb9-88f1-80e22e0d9e97","added_by":"auto","created_at":"2026-02-23 13:41:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":88257,"visible":true,"origin":"","legend":"\u003cp\u003eVisual Assessment of Cluster Tendency (VAT) plot, in the form of a heatmap of the distance matrix. The color of each cell represents the distance (dissimilarity) between two points. Deep Red/Darker colors: Usually represent points that cluster very close to each other (low distance). Blue/Lighter colors: Represent points that cluster very far apart from red/darker clusters (high distance).\u003c/p\u003e","description":"","filename":"floatimage21.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/bd5a7cd4c5e50aa8b423c58a.png"},{"id":103238071,"identity":"9ad093b0-4b97-4cf3-b247-1ab014bfa880","added_by":"auto","created_at":"2026-02-23 13:37:40","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":220838,"visible":true,"origin":"","legend":"\u003cp\u003ea) PCA biplot displaying the distribution of the hyperframe variables (in color) and the carcass assemblages (in black). Color in variables represent their relative contribution to each component. Phylogenetic dendrogram on hierarchical clustering of the spatial point patterns, using “average” (b) and “Ward” (c) cluster methods. For keys to variables see Table 2.\u003c/p\u003e","description":"","filename":"floatimage33.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/48169125ffc63d8a18253308.png"},{"id":103238073,"identity":"b9940640-5beb-4fd2-a5a3-2c5c192c6741","added_by":"auto","created_at":"2026-02-23 13:37:40","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":160125,"visible":true,"origin":"","legend":"\u003cp\u003ea) PCA biplot displaying the distribution of the hyperframe variables (in color) and the carcass assemblages (in black), excluding the archaeological samples. Color in variables represent their relative contribution to each component. Phylogenetic dendrogram on hierarchical clustering of the spatial point patterns from the modern neo-taphonomic carcasses, using “average” (b) and “Ward” (c) cluster methods. For keys to variables see Table 2.\u003c/p\u003e","description":"","filename":"floatimage41.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/0056e6d38e6c1b20a45dec80.png"},{"id":103238074,"identity":"6230fa2b-1146-4155-8082-875e7199ec0b","added_by":"auto","created_at":"2026-02-23 13:37:40","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":187119,"visible":true,"origin":"","legend":"\u003cp\u003ea) PCA biplot displaying the distribution of the hyperframe variables (in color) and the carcass assemblages (in black), using equal window sizes for each spatial pattern. Color in variables represent their relative contribution to each component. Phylogenetic dendrogram on hierarchical clustering of the spatial point patterns from the modern neo-taphonomic carcasses, using “average” (b) and “Ward” (c) cluster methods. Equal window sizes have been used for each spatial pattern. For keys to variables see Table 2.\u003c/p\u003e","description":"","filename":"floatimage53.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/ca1cab20feee27cd95017360.png"},{"id":103238075,"identity":"45b9f250-6bcf-48b9-915b-623d43b2731c","added_by":"auto","created_at":"2026-02-23 13:37:40","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":519247,"visible":true,"origin":"","legend":"\u003cp\u003eDifferent stages of the data acquisition process for Elephant 6: a) Screenshot from 3d Scanner App™ showing the photogrammetric model obtained during fieldwork; b) orthophotograph generated in Blender 4.2 after including the 1x1 m grid; c) high-quality orthophotograph obtained with Agisoft Metashape; d) final map featuring the polygons and centroids for each bone. Cartography of the three selected archaeological sites: e) EAK [1], f) Áridos-2 [56], and g) Marathousa 1 [57].\u003c/p\u003e","description":"","filename":"floatimage61.png","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/171b1725bccb72ff518cbfee.png"},{"id":106401421,"identity":"3976360b-55ad-43d7-9047-57dcd7dfd1d6","added_by":"auto","created_at":"2026-04-08 08:49:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3425794,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/cd336af7-a207-4fbf-936c-a44c16fbcb90.pdf"},{"id":103060048,"identity":"9d84cb08-1906-4e13-a540-0972dedddeab","added_by":"auto","created_at":"2026-02-20 09:49:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":13887058,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFiles.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/c4c1b728cc6397fc66ebc3a2.pdf"},{"id":103060047,"identity":"a4400a72-fdcc-4c8a-9003-c6c619dc351e","added_by":"auto","created_at":"2026-02-20 09:49:53","extension":"r","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":46126,"visible":true,"origin":"","legend":"","description":"","filename":"DominguezRodrigoetalProboscideanspatialanalysis.r","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/bd7ddb31dba5f8db9d75890e.r"},{"id":103060045,"identity":"dc5efbe9-2b4a-4586-ae94-5c334ffc1b42","added_by":"auto","created_at":"2026-02-20 09:49:53","extension":"csv","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":1620,"visible":true,"origin":"","legend":"","description":"","filename":"01SpatialWINDOWS.csv","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/ebfb36f6cee9d40a3e121320.csv"},{"id":104808166,"identity":"d85df4b9-d2b1-4b07-8a12-ac1d1159039a","added_by":"auto","created_at":"2026-03-17 12:24:22","extension":"csv","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":61481,"visible":true,"origin":"","legend":"","description":"","filename":"02SpatialDatasets.csv","url":"https://assets-eu.researchsquare.com/files/rs-8799314/v1/ccf03b55aa5dd97f8d2b17d6.csv"}],"financialInterests":"No competing interests reported.","formattedTitle":"Spatial signatures of carnivore-driven bone dispersal: A taphonomic analysis of modern proboscidean carcasses, and implications for Pleistocene elephant butchery sites","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe study of elephant bone accumulations plays a pivotal role in reconstructing early hominin subsistence strategies, particularly in relation to the exploitation of megafauna [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Given their massive size and high nutritional return [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e], proboscideans would have represented a significant resource for hominin populations if early access could be achieved, either through hunting or confrontational/active scavenging in highly competitive ecosystems, or even through opportunistic behaviors in low-competition habitats. Contrary to previous assumptions about prolonged meat availability from these carcasses [\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], recent studies of proboscidean bulk flesh consumption by prides of lions in Zambia and Botswana show that carcasses can be mostly defleshed in five days before lions give way to other scavenging carnivorans [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. So, hominin exploitation of proboscidean flesh may have occurred early in the consumption sequence. Alternatively, access to proboscidean resources may also have occurred later if targeting grease and flesh scraps instead. Determining the timing of access is a taphonomic challenge.\u003c/p\u003e \u003cp\u003eThe mere co-occurrence of elephant bones and stone tools does not inherently indicate a functional relationship between the two. In fact, most Pleistocene sites where these elements are found in spatial association are situated in alluvial contexts -environments that, as observed in modern African settings, foster highly productive taphocoenotic processes. These processes can result in spurious associations, where the proximity of bones and tools is the product of natural or marginal accumulation dynamics rather than intensive hominin proboscidean exploitation activities [\u003cspan additionalcitationids=\"CR10 CR11\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Fossil proboscidean sites occur in the vicinity of prehistoric lakes, river and ponds more abundantly than elsewhere [\u003cspan additionalcitationids=\"CR11 CR12 CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25 CR26 CR27 CR28 CR29 CR30 CR31 CR32 CR33 CR34 CR35 CR36 CR37 CR38 CR39 CR40 CR41 CR42 CR43 CR44 CR45 CR46 CR47\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e], as they are also documented in modern African savanna ecosystems [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. Understanding the spatial, anatomical, and taphonomic characteristics of elephant carcass accumulations and dispersions is thus essential to distinguish between natural processes and anthropogenic behaviors. This information can also be of relevance when reconstructing the timing of hominin access to such carcasses.\u003c/p\u003e \u003cp\u003eOne of the central challenges in this regard is disentangling the respective roles of carnivores and hominins in the access and modification of elephant remains. Elephant carcasses may persist on the landscape for extended periods, attracting a variety of taphonomic agents (including lions, hyenas, vultures, and environmental forces), which can all alter the bone assemblage [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan additionalcitationids=\"CR50 CR51\" citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. Without careful assessment, these natural agents can mimic or obscure patterns that might otherwise suggest hominin involvement, such as tool-inflicted cut marks, percussion damage from marrow extraction, or the presence of lithic artifacts in spatial association with the bones. Recent neo-taphonomic studies of modern elephant death sites provide invaluable comparative data, offering baseline expectations for carcass modification in the absence of hominins [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan additionalcitationids=\"CR50 CR51\" citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. These studies have shown that while carnivores such as hyenas can produce extensive damage on cancellous anatomical portions, their bone modification patterns may appear conspicuously limited to cancellous or trabecular tissue (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Thus, identifying felid and hyena impacts on proboscidean carcasses, and interpreting the subtle signatures of hominin exploitation (particularly early access to flesh) requires a detailed understanding of how elephant carcasses are naturally disarticulated, scattered, and consumed under undisturbed conditions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn modern African savanna ecosystems, the decomposition and transformation of elephant carcasses are shaped primarily by natural processes, especially the activity of durophagous carnivores such as hyenas. These powerful scavengers play a crucial role in the breaking, disarticulating, and scattering of skeletal elements, significantly altering the integrity and spatial configuration of carcasses over time. Other agents (i.e., smaller carnivorans) might have also played a relevant role in transporting bones away, but these agents have not received proper attention until now. Although one could think of elephant death sites as palimpsests spanning a large amount of taphonomic biotic processes through time, enabling the time-averaged overlying of the actions of successive carnivoran agents, controlled observations of such sites indicate a more nuanced reconstruction of this process. Most carnivore intervention at elephant carcasses occurs during the very initial phase of deposition and accumulation. After a few weeks (usually less than a month), carnivores impact the carcass minimally. This is the time in which no flesh is preserved, and some bone grease is accessible only to hyenas, but bones are already on a stage of drying up [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. We have observed that it takes only a few weeks for some bones to start showing some sign of subaerial weathering, which would be expected much later according to observations made on smaller game [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e, \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHyenas are among the few carnivores capable of crushing large bones, including those of elephants in specific anatomical locations (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Their exceptional bite force allows them to consume not only soft tissues but also to break into denser elements, such as long bone epiphyses, and even cranial structures. As a result, elephant skeletons rarely remain intact for long in the presence of active hyena populations. This is especially relevant in highly competitive ecosystems. The process of bone fragmentation, furrowing, and surface modification often begins within days or immediately after a carcass being abandoned by primary consumers like lions. Moreover, the dismemberment and transport of skeletal parts by hyenas contribute to the dispersion of bones across the landscape. Entire limbs or isolated elements may be dragged dozens of meters from the original site, resulting in spatially disaggregated assemblages. This scattering behavior not only affects the visibility of certain anatomical regions during taphonomic surveys, but also challenges efforts to reconstruct carcass use sequences or infer predator-prey dynamics from the fossil record.\u003c/p\u003e \u003cp\u003eThese patterns of carcass modification, particularly in large-bodied taxa like elephants, offer critical insights into the taphonomic signatures that aggregate under natural conditions. Understanding the role of durophagous carnivores, thus, becomes essential when interpreting fossil assemblages involving megafauna, especially in regions where early hominins may have interacted with similar carnivorans or exploited/competed for similar resources. Given the specific timeline of resource availability and carcass scattering/modification introduced during the post-defleshing scavenging phase, finding taphonomic signatures that could be linked to the timing of hominin intervention in elephant carcass exploitation results crucial.\u003c/p\u003e \u003cp\u003eHere, we intend to specifically address this issue by using a taphonomic spatial approach. We will relate specific spatial patterns to degrees of post-depositional durophagous carnivore modification on bones and impact on bone assemblages (through bone survival estimates). We will then relate that to the timing of carcass access by hominins within the analysis of a selection of prehistoric butchery sites. The present work intends to address all these issues emphasizing an analytical approach, and it is innovative by doing so using spatial statistical taphonomy.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eNeo-taphonomic analysis\u003c/h2\u003e \u003cp\u003eFor the purposes of this study, tooth-mark frequencies (TM) and bone survival percentages are used as complementary taphonomic variables to characterize carnivore impact on large-mammal carcasses (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). TM represents the proportion of skeletal elements bearing carnivore tooth marks (or gnawing modifications) and provides a direct proxy for the intensity of carcass modification by carnivores during feeding. Bone survival reflects the proportion of the original skeletal assemblage that remains at the site after carnivore activity, integrating processes of bone removal, transport, consumption, and destruction. Together, these variables allow us to distinguish between assemblages dominated by intensive carnivore exploitation versus those experiencing more limited scavenging or disturbance.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eRepresentation of skeletal parts (MNE: Minimum Number of elements) -except for ribs- in the Chobe and Okavango carcasses studied. Numbers in parentheses are for carnivore tooth-marked bones (numerator) and percentages per element (denominator). Tooth mark and bone survival frequencies are also included.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c6\" namest=\"c3\"\u003e \u003cp\u003eChobe National Park\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c9\" namest=\"c7\"\u003e \u003cp\u003eOkavango Delta\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eElephant_1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eElephant_2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eElephant_3\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eGiraffe\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eElephant_4\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eElephant_5\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eElephant_6\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCranium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMandible\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRibs*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30 (6/20)*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e35 (25/70)**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e23 (10/43.4)***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e21 (9/39.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e37 (24/64.8)****\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e34 (8/23.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e40 (6/15)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSternum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVertebrae\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAtlas\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAxis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1(1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCervical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6 (6/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThoracic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14 (3/21.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15 (14/93.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7 (4/57.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e9 (6/66.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e4 (3/75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLumbar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSacrum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCaudal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInnominates\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eScapula\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHumerus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadius\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUlna\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemur\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 (2/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTibia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2 (1/50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (1/100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFibula\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (1/100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCalcaneum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAstragalus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetacarpals/Metatarsals\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCarpals/Tarsals\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhalanges\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTooth mark (TM) frequencies\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e61.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e5.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBone survival percentages\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e14.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e18.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e36.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"9\" nameend=\"c9\" namest=\"c1\"\u003e \u003cp\u003eEstimates in MNE (Minimal Number of Elements). Parentheses: Numerator is number of toothmarked elements; denominator is the percentage of MNE bearing TM (pits/scores/furrowing and/or green breaks)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"9\" nameend=\"c9\" namest=\"c1\"\u003e \u003cp\u003e* Out of 53 fragments\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"9\" nameend=\"c9\" namest=\"c1\"\u003e \u003cp\u003e**Out of 37 fragments\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"9\" nameend=\"c9\" namest=\"c1\"\u003e \u003cp\u003e***Out of 34 fragments\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003e****Out of 40 fragments\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe modern assemblages analyzed here show marked variability in both TM and survival, indicating substantial differences in carnivore behavior and access to carcasses. Elephant_2 exhibits extremely high TM values (84.3%) combined with low survival (21.4%), a pattern consistent with intense carnivore utilization leading to extensive bone modification and removal. Elephant_4 also shows high TM (61.6%) and low survival (18.4%), suggesting similarly strong carnivore pressure. In contrast, Elephant_5 and Elephant_6 present very low TM values (6.6% and 5.3%, respectively) coupled with higher survival (36.8% and 40%), indicating limited carnivore access, resulting in minimal bone modification and greater skeletal retention at the site. Intermediate cases, such as Elephant_1 (TM\u0026thinsp;=\u0026thinsp;26.5%, survival\u0026thinsp;=\u0026thinsp;19.6%) and Elephant_3 (TM\u0026thinsp;=\u0026thinsp;36.9%, survival\u0026thinsp;=\u0026thinsp;14.1%), reflect moderate carnivore involvement, with noticeable but not overwhelming impacts on carcass integrity.\u003c/p\u003e \u003cp\u003eThe giraffe assemblage occupies an intermediate position, with TM values of 40% and relatively low survival (15%), suggesting substantial carnivore interaction comparable to heavily exploited elephant carcasses despite taxonomic differences. Overall, the inverse relationship observed between TM and survival across assemblages underscores the role of carnivores as primary agents structuring bone modification and assemblage composition. These modern patterns provide a quantitative baseline for interpreting fossil elephant assemblages, allowing inferences about the intensity and nature of carnivore involvement based on comparable combinations of tooth-mark frequencies and skeletal survival.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eNeo-taphonomic analysis\u003c/h3\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eStage 1: Modern and archaeological samples\u003c/h2\u003e \u003cp\u003eThe hyperframe that we used summarizes a rich and multiscale description of spatial organization across sites, combining local intensity structure (Linhom, Finhom), second-order interactions (PCF), neighborhood statistics (NND, Hopskel), global dispersion (quadrat test), and geometric\u0026ndash;topological descriptors (Minkowski functionals, convex hull, distances) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e; Supplementary Table\u0026nbsp;2; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The PCA distills this high-dimensional structure into a smaller number of orthogonal gradients, allowing us to understand which sites are truly similar in \u003cem\u003eoverall spatial behavior\u003c/em\u003e rather than in isolated metrics (Supplementary Tables\u0026nbsp;3\u0026ndash;4).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSince PCA was used to avoid the effects of collinearity, it was required that the resulting number of retained components should include a minimum of 90% of the sample variance. In order to do so, four components were retained, comprising 93.47% of the total variance. Across the dataset, PC1 clearly separates sites by overall spatial scale, density, and geometric footprint, while PC2 primarily captures differences in aggregation versus dispersion at local and intermediate scales, and PC3\u0026ndash;PC4 reflect finer contrasts in topology (Euler, perimeter) and anisotropy/heterogeneity.\u003c/p\u003e \u003cp\u003ePC1 summarizes scale and extent (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). It primarily captures a gradient in overall spatial scale and footprint expansion versus compact clustering. This component loads positively on inhomogeneous intensity measures (Linhom_ml, Linhom_mn, Linhom_mlg) and negatively on geometric descriptors such as scatter area, maximum distance, mean pairwise distance, and Minkowski area and perimeter. Sites with strongly positive PC1 (40% of variance) scores (notably EAK, \u0026Aacute;ridos-2, Haynes_2, Giraffe, Marathousa 1, and White-Diedrich) are characterized in the hyperframe by high Linhom_ml values, elevated clustering indices, limited convex hull areas, and small maximum distances. These sites represent spatial systems that occupy limited windows and show substantial spatial structure beyond simple CSR. \u0026Aacute;ridos-2 and Haynes_2 cluster close to one another on PC1 because they share elevated Linhom values, strong clustering indices, and similar Minkowski Euler values, indicating fragmented but spatially restrictive point patterns. In contrast, negative PC1 scores (Elephant_2, Elephant_4, Elephant_5, Haynes_1) correspond to sites with larger spatial extent, lower intensity, and more compact footprints, despite sometimes having strong clustering at local scales. These sites are spatially constrained and dominated by short-range interactions rather than broad spatial organization. This axis thus separates archaeological assemblages formed within restricted depositional contexts from modern carcasses where bone displacement and carcass spread are more extensive.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003ePC2 (30% of variance) introduces a second, orthogonal gradient (capturing local aggregation intensity) that is largely independent of overall size. Strong negative loadings on Linhom_clust, Finhom_nn, and both pair-correlation indices (pcf_index and pcf_index2) contrast with positive loadings on the Hopkins\u0026ndash;Skellam statistic. Negative PC2 values correspond to pronounced clustering and spatial dependence, while positive values indicate weaker interaction and more regular or diffuse patterns. Sites with strongly negative PC2 scores (Elephant_1, Elephant_3, EAK) are characterized by very high PCF indices, strong local clustering, and low Hopskel values, indicating intense aggregation at short distances. Elephant_1 and Elephant_3 are particularly informative here. Despite moderate PC1 values, they cluster tightly on PC2 due to their very similar hyperframe profiles: extremely high PCF indices, high quadrat statistics, and strong Euler values. This suggests intense, fine-scale clustering within relatively constrained spatial domains. EAK shares this strong short-range clustering signature, explaining its negative PC2 position and highlighting its multiscale complexity. Conversely, sites with positive PC2 scores (Giraffe, Marathousa 1, Haynes_1) tend to show lower PCF values, higher Hopskel values, and weaker local clustering, consistent with more dispersed, less dense or mixed spatial patterns.\u003c/p\u003e \u003cp\u003ePC3 (16.5% of variance) reflects a contrast between intensity-driven aggregation and spatial fragmentation linked to carcass modification and redistribution. Negative loadings on Linhom metrics and quadrat statistics oppose positive contributions from pair-correlation functions and NND_index. High positive PC3 scores indicate patterns with strong local aggregation but reduced large-scale coherence, whereas negative scores correspond to assemblages structured by intensity gradients and broader-scale density fields. In the scores, \u0026Aacute;ridos-2 and Marathousa 1 have positive PC3 values, suggesting aggregation driven by localized processes rather than simple density scaling. EAK, however, plots strongly negative on PC3, indicating a spatial pattern tightly structured by inhomogeneous intensity, consistent with a strongly constrained depositional environment. Modern elephant carcasses show mixed PC3 behavior, with Elephant_3 and Elephant_1 positive (localized aggregation) and Elephant_2 negative (more fragmented, intensity-driven distribution).\u003c/p\u003e \u003cp\u003ePC4 (6.7% of variance) (which highlights secondary geometric contrasts) captures subtler variation, largely decoupled from size and clustering intensity. This component loads positively on nearest-neighbor structure (NND_index), Hopskel, and some interaction metrics, while loading negatively on the Euler characteristic, indicating sensitivity to the connectivity and perforation of occupied space. High positive PC4 scores indicate spatial configurations with many small clusters or disconnected small patches, whereas negative values reflect more topologically cohesive distributions. PCA4 also reflects differences in internal organization, such as whether clustering produces elongated structures versus compact ones, and how perimeter scales with area. Elephant_1 and Haynes_1, which load positively on PC4, show disproportionately large perimeters relative to area, suggesting elongated or irregular clustering. Negative PC4 values (e.g., Elephant_2, Elephant_6, Giraffe) correspond to more compact, isotropic spatial arrangements.\u003c/p\u003e \u003cp\u003eTaken together, the PCA scores show that archaeological elephant assemblages (EAK, \u0026Aacute;ridos-2, Marathousa 1) occupy a distinct region of multivariate space compared to modern elephant carcasses. When the raw hyperframe data and the PCA are considered jointly, several robust site groupings emerge: Elephant_1 and Elephant_3 are among the most similar overall. They share: Very high PCF indices, strong quadrat deviation, and high Euler values.PCA confirms this similarity by placing them close together, especially along PC2 and PC3. They represent intensely clustered, fine-scale spatial systems, likely reflecting constrained accumulation processes. Elephant_4, Elephant_5, Elephant_6 form a loose cluster characterized by: Moderate clustering, longer scatter areas, and lower Euler fragmentation. PCA places them near one another on PC1 and PC2, indicating broadly similar spatial organization with moderate internal variation.\u003c/p\u003e \u003cp\u003e\u0026Aacute;ridos-2 and Haynes_2 align on PC1 and PC3, driven by: High intensity, strong clustering indices, and large Euler values. Giraffe \u0026amp; Marathousa 1 cluster on PC2 and PC3, sharing: Lower PCF values, higher Hopskel values, and moderate Minkowski measures. Their spatial structure is less clustered and more dispersed, suggesting different formation dynamics from the elephant-dominated sites. EAK occupies an extreme position: high PC1, negative PC2, strongly negative PC3. This reflects a site that is simultaneously strongly clustered, and topologically cohesive, setting it apart from all others. Modern elephant carcasses span a broader range of PCA space, reflecting variability in carcass spread, scavenger-driven bone displacement, and landscape openness. Overall, the joint interpretation of loadings and scores indicates that spatial patterning in the archaeological sites cannot be explained simply as scaled-down versions of modern carcass dispersal, but instead reflects distinct formation dynamics involving spatial confinement, aggregation, and limited bone transport.\u003c/p\u003e \u003cp\u003eA look at the distribution of the spatial point patterns shows that at least two (Elephants_1 and 3) of the three least conspicuously impacted by carnivores cluster together and separately (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb-c). We also noticed that time of carcass exposure is unrelated to either carnivore ravaging intensity, spatial point pattern or spread of the scattering distance. It is remarkable that the intensity of carnivore damage on carcasses shows a strong negative correlation with the skeletal element survival index (r = \u0026minus;\u0026thinsp;0.63, p-value\u0026thinsp;=\u0026thinsp;0.1258). This is expected in the face of observational studies showing that the most intense damage to elephant carcasses occurs in the first few days of exposure, and that depending on the degree of food availability and carnivore competition in the ecosystem, proboscidean carcasses may remain unscattered and minimally impacted by durophagous carnivores, thus indicating that carcass damage and survival is ecologically-determined beyond the ecosystem predator carrying load [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. To analyze this better, a second spatial analysis was carried out using only the experimentally documented modern carcasses, for which taphonomic information is available.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStage 2: Spatial and neo-taphonomic analysis of modern samples\u003c/h3\u003e\n\u003cp\u003eFor this part of the analysis, only the modern carcasses that we studied were included in the analysis (Supplementary Tables\u0026nbsp;5\u0026ndash;7). Taken together, the modern-elephant spatial results and the PCA provide a coherent and internally consistent picture of how different carcasses are distributed in space, how strongly they cluster, and how those patterns relate to movement scale, intensity structure, and inferred behavioral regimes such as exposure time and bone survival. What is especially powerful here is that the PCA does not contradict the raw spatial statistics; instead, it compresses them into a small number of interpretable spatial syndromes that map cleanly back onto the original measures.\u003c/p\u003e \u003cp\u003eAt the level of the raw spatial metrics (Supplementary Table\u0026nbsp;5), the modern elephants separate into two broad behavioral\u0026ndash;spatial modes. Elephant_1 and Elephant_3 show extremely strong clustering signatures. Both exhibit very high Linhom_clust values (above 13), very large Finhom_nn values, and extreme deviations in pair-correlation indices (pcf_index and pcf_index2 both above 6). Their Hopskel values are very low, especially Elephant_3, indicating strong departure from CSR toward aggregation. Quadrat statistics are also very large, reinforcing the interpretation of highly clustered point patterns. Despite this strong clustering, their movement envelopes are not the largest in absolute terms: their scatter areas and convex-hull areas are moderate compared to some other elephants. This combination suggests limited transport by carnivores rather than wide-ranging dispersion. This seems to be unrelated to bone survival, which is reflected in low values for both carcasses.\u003c/p\u003e \u003cp\u003eIn contrast, Elephant_2, Elephant_4, and Elephant_5 form a second group characterized by weaker clustering and much larger spatial footprints. Their Linhom_clust values are substantially lower (around 4\u0026ndash;8), Finhom_nn values are roughly half those of Elephant_1 and Elephant_3, and their pcf indices sit in the moderate range (roughly 2\u0026ndash;3). At the same time, these individuals show very large M_area, scatter_area, and max_distance values, particularly Elephant_2 and Elephant_4, which have the largest convex hulls and pairwise distances in the dataset. This indicates extensive movement across space with less intense local reuse by carnivores, a pattern that aligns well with medium to short bone survival values and higher TM values, suggesting more transient or carnivore-impacted spatial behavior.\u003c/p\u003e \u003cp\u003eElephant_6 stands somewhat apart from both groups. It shows moderate clustering indices and relatively low pcf values compared to Elephant_1 and Elephant_3, but its spatial footprint is small, with the lowest scatter area, shortest max_distance, and lowest mean_pairwise distance among all elephants. This suggests a spatially constrained individual that nonetheless does not exhibit extreme point aggregation. Its very short TM and higher bone survival are consistent with a brief time exposure, spatially limited post-depositional disturbance, and more limited carnivore impact.\u003c/p\u003e \u003cp\u003eThe giraffe provides a useful external anchor. Its Linhom_clust is the lowest of all patterns, its pcf indices are well below 1, and its Finhom_nn is the smallest in the dataset. Quadrat statistics are also much lower than those of the elephants. At the same time, its spatial footprint is very small, with minimal scatter area and short pairwise distances. This combination reflects near-random or weakly regular spacing over a small area, fundamentally different from the elephant patterns and validating that the analytical framework is sensitive to genuine ecological or behavioral contrasts rather than producing uniform outputs.\u003c/p\u003e \u003cp\u003eWhen these same variables are projected into PCA space, the underlying structure becomes clearer (Supplementary Tables\u0026nbsp;6\u0026ndash;7) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). PC1 primarily captures a scale-versus-dispersion gradient. Variables associated with large spatial extent (scatter_area, max_distance, mean_pairwise distance, M_area, and M_perimeter) load strongly and negatively on PC1, whereas intensity-related measures such as Linhom_ml, Linhom_mn, Linhom_mlg, NND_index, and Survival load positively. As a result, Elephant_2, Elephant_4, and Elephant_5, which have very large spatial footprints, plot strongly negative on PC1, while Elephant_6 and especially the giraffe plot strongly positive. Elephant_1 and Elephant_3 occupy intermediate PC1 positions, reflecting their moderate spatial extent but high internal clustering.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003ePC2 separates patterns according to clustering structure rather than scale. Linhom_clust, Finhom_nn, and both pcf indices load strongly and positively on PC2, while Hopskel loads negatively. High PC2 scores therefore correspond to intense clustering across multiple spatial statistics. Elephant_1 and Elephant_3 sit very high on PC2, matching their extreme pcf and Linhom_clust values. Elephant_4 and Elephant_5 are negative on PC2, consistent with weaker clustering, while Elephant_2 occupies an intermediate position. The giraffe again lies apart, with low clustering metrics translating into a negative PC2 score.\u003c/p\u003e \u003cp\u003ePC3 introduces a contrast between quadrat-scale aggregation and topological structure. Quadrat_stat and Hopskel load strongly and positively, while M_euler loads strongly and negatively. This component differentiates patterns where aggregation manifests as large-scale count heterogeneity versus patterns where the occupied space becomes highly perforated or fragmented. Elephant_3, which shows extreme quadrat values and low Hopskel, is strongly negative on PC3, whereas Elephant_1 is positive, reflecting subtle but meaningful differences in how clustering is organized spatially despite similar pcf behavior.\u003c/p\u003e \u003cp\u003ePC4 and higher components refine these interpretations by incorporating carnivore damage (TM), survival, and finer-scale geometric properties. PC4 shows strong positive loadings for TM, mean_pairwise distance, and Linhom measures, separating Elephant_1 (positive PC4, long TM and long survival) from Elephant_2 (negative PC4, short TM and short survival). PC5 and PC6 further isolate Elephant_6, whose unusual combination of limited carnivore damage, constrained space use, and moderate clustering gives it extreme loadings on these axes. The giraffe consistently remains isolated across components, reinforcing its role as a structurally distinct spatial pattern rather than simply an endpoint along an elephant gradient.\u003c/p\u003e \u003cp\u003eViewed holistically, the fusion of raw spatial statistics and PCA reveals that modern elephant space use is not defined by a single continuum but by at least two interacting dimensions: the scale of movement across the landscape and the intensity of local clustering within that space. Elephant_1 and Elephant_3 converge in PCA space because they share extreme clustering signatures, even though they differ in bone survival and carnivore damage. Elephant_2, Elephant_4, and Elephant_5 cluster together because their dominant trait is large-scale dispersion with relatively weaker aggregation. Elephant_6 emerges as a distinct spatial strategy characterized by confinement without strong clustering. The giraffe stands apart as a fundamentally different spatial process.\u003c/p\u003e \u003cp\u003eThis integrated perspective strengthens confidence in the results (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb-c). The PCA clarifies how multiple spatial statistics co-vary and how individual animal carcasses impacted by different degrees of carnivore impact and preservation lead to consistent spatial patterns. In this sense, the modern-elephant dataset provides a well-resolved reference framework that can be directly compared to fossil or archaeological assemblages, allowing differences in clustering, dispersion, and space-use intensity to be interpreted, not as isolated metrics, but as coherent spatial models with taphonomic implications.\u003c/p\u003e\n\u003ch3\u003eStage 3: Spatial analysis with homogeneous windows\u003c/h3\u003e\n\u003cp\u003eAfter controlling window size by restricting all point patterns to the same 10 \u0026times; 10 m area, the PCA isolates true differences in spatial organization rather than differences driven by observation scale (Supplementary Tables\u0026nbsp;8\u0026ndash;10). This normalization is critical: previously, large scatter areas and long inter-point distances strongly structured the PCA, but here those effects are intentionally minimized. As a result, the first PCA primarily reflects local clustering intensity, spatial heterogeneity, and topological structure, making the comparison between modern and archaeological assemblages more meaningful at a different level; that is, excluding the overall effect of dispersion captured by the heterogeneous windows and concentrating at agglomerative and segregation processes at the dense cluster local level (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003ePC1, which separates sites most strongly along the horizontal axis, is dominated by intensity-related variables (Linhom_ml, Linhom_mn, Linhom_mlg) loading positively and by Minkowski and distance-related measures (M_area, scatter_area, max_distance, mean_pairwise) loading negatively. High positive PC1 scores, therefore, characterize assemblages with very dense local clustering within the standardized window, whereas negative PC1 values indicate more locally spatially dispersed or fragmented clusters even within the same area. \u0026Aacute;ridos-2 and EAK occupy the extreme positive end of PC1, indicating that their densest 100 m\u0026sup2; zones are characterized by exceptionally high local intensity relative to background, closely resembling tightly packed accumulations. This is due to carcasses at these localities being in articulated state (\u0026Aacute;ridos-2) or semi-articulated and additionally tectonically compressed, further increasing the clustering intensity (EAK). In contrast, Elephant_4, Elephant_5, Haynes_1, and Elephant_6 fall on the negative side of PC1, reflecting more diffuse or broken spatial structure even at peak density, as would be expected from disarticulated carcasses and assemblages modified by dispersing carnivore behavior.\u003c/p\u003e \u003cp\u003ePC2 is shaped almost entirely by second-order interaction statistics, with strong negative loadings from Linhom_clust and Finhom_nn and similarly strong contributions from both pair correlation indices. This axis captures the degree of spatial interaction beyond simple intensity, distinguishing patterns dominated by strong clustering and aggregation at short distances (negative PC2) from those closer to randomness or weak interaction (positive PC2). EAK is again extreme here, with a very strong negative PC2 score, indicating pronounced short-range clustering within its densest cluster. \u0026Aacute;ridos-2 also shows positive clustering structure but less extreme. Among modern assemblages, Elephant_1, Elephant_2, Elephant_3, and Giraffe occupy positive PC2 values, suggesting more moderate or heterogeneous short-range interactions once window size is fixed.\u003c/p\u003e \u003cp\u003ePC3 is primarily structured by quadratic and topological contrasts, with quadrat_stat and Linhom_mn loading negatively and hopskel loading weakly. This axis separates assemblages with high internal heterogeneity and patchiness from those with smoother internal organization. \u0026Aacute;ridos-2 stands out again with a strong negative PC3 score, suggesting that even within its densest cluster, bone distribution is internally irregular and segmented, consistent with repeated disturbance or selective removal. In contrast, Giraffe, Elephant_2, and Haynes_2 load positively on PC3, indicating more internally coherent spatial clusters.\u003c/p\u003e \u003cp\u003ePC4 highlights nearest-neighbor structure and Euler topology, with strong positive loadings from NND_index and Hopskel and strong negative loadings from M_euler. This component differentiates patterns dominated by many small disconnected voids and holes (highly negative Euler values) from those with more continuous occupied space. Elephant_2 and Giraffe score high on PC4, reflecting compact clustering, whereas Elephant_1 and White-Diedrich score negatively, indicating fragmented spatial structure even at high density.\u003c/p\u003e \u003cp\u003eHigher components (PC5\u0026ndash;PC8) capture more subtle contrasts, including trade-offs between nearest-neighbor spacing and clustering strength (PC5), smoothing versus fragmentation effects (PC6), and quadrat-based regularity versus Euler complexity (PC7). These axes refine differences among modern elephant assemblages but do not drive the main archaeological\u0026ndash;modern separation.\u003c/p\u003e \u003cp\u003eWhen viewed together, the PCA reveals a clear and consistent pattern: after standardizing window size, \u0026Aacute;ridos-2and EAK remain distinct from modern elephant carcasses, not because of scale but because of intrinsic spatial organization. Their densest clusters are simultaneously more intense, more tightly aggregated, and more topologically complex than any modern carcass pattern, because they are not completely disarticulated as the modern samples are. Marathousa 1, by contrast, falls closer to the modern elephant cluster, especially Elephant_1 and Elephant_3, suggesting that its spatial organization is more consistent with limited disturbance and localized accumulation, after initial disarticulation.\u003c/p\u003e \u003cp\u003eThe main result of the Stage 3 analysis is that window size was not driving earlier interpretations. Even when controlling spatial extent and focusing only on the densest 100 m\u0026sup2; areas, archaeological assemblages (especially \u0026Aacute;ridos-2 and EAK) retain spatial signatures indicative of non-random accumulation, limited modification, and controlled carnivore-mediated disturbance, whereas modern elephant carcasses exhibit spatial structures consistent with a more disarticulation (probably reflecting time-averaging) and probably more carnivore impact. The first argument can be defended in sites like \u0026Aacute;ridos-2 and EAK exhibiting barely any bone weathering caused by subaerial exposure (in contrast with several of the carcasses studied). The second argument is more speculative, since it can only be effectively assessed over broader areas than those unearthed at the three archaeological sites. This certainly limits behavioral interpretations about the depositional history of those elephant carcasses, their impact by carnivores and additional taphonomic information potentially crucial to unravel the timing of hominin access to them.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn our philosophy of trying to implement an analytical taphonomy of proboscideans (to counter the more traditional descriptive taphonomy), we adopted a taphonomic spatial approach, which was previously successful at differentiating degrees of post-depositional disturbance of experimental assemblages subjected to hydraulic flows [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. Taken together, the PCA structure and the original spatial statistics reveal a coherent and taphonomically meaningful relationship between point-process structure and both carnivore tooth‐mark frequencies and bone survival, with scattering processes also possibly reflecting time averaging intervention of dispersing agents. Rather than acting independently, these taphonomic variables align consistently with gradients of spatial clustering, scale, and geometric complexity captured by the point‐pattern descriptors.\u003c/p\u003e \u003cp\u003eAt one extreme of the ordination lie point processes characterized by large spatial extent, long interpoint distances, and expansive convex hulls; patterns with high scatter area, large maximum distance, and high mean pairwise distances. In the modern dataset, Elephant_2 and Elephant_4 fall clearly into this domain, loading strongly in the negative direction of PC1 and PC2. These processes are spatially diffuse: points are spread over wide areas, nearest-neighbor distances are large, and Minkowski area and perimeter values are high. Importantly, these same cases are associated with relatively high TM values and reduced bone survival. This association suggests that when carcass remains are spatially dispersed, carnivores have had more opportunity to transport, disarticulate, and selectively remove skeletal elements. Spatial diffusion therefore appears to be a strong spatial signature of intensive carnivore intervention.\u003c/p\u003e \u003cp\u003eBy contrast, point processes clustering on the opposite side of PC1 (most clearly Elephant_6 and Giraffe) exhibit compact spatial configurations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). These assemblages show small scatter areas, short maximum distances, and low mean pairwise distances, coupled with lower Minkowski area and perimeter. Their point patterns are dense but spatially constrained, indicating limited lateral transport. Correspondingly, these cases show low TM values and higher bone survival. In practical terms, this implies that when bones remain spatially concentrated, carnivores either had limited access or engaged primarily in localized consumption rather than prolonged transport. The PCA, thus, captures a consistent spatial\u0026ndash;taphonomic signal: compact point processes are associated with better skeletal survival and fewer carnivore modifications.\u003c/p\u003e \u003cp\u003eA second important gradient emerges along PC2 and PC3, which are strongly influenced by measures of clustering intensity and spatial heterogeneity, including Linhom_clust, Finhom_nn, pcf_index, and pcf_index2. Assemblages such as Elephant_1 and Elephant_3 score high on PC2, reflecting strong small-scale clustering and pronounced departures from spatial randomness (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). These sites also exhibit elevated TM values but intermediate bone survival. This combination suggests a scenario in which carcasses remained spatially aggregated, yet were repeatedly revisited by carnivores. Rather than extensive transport, carnivore activity here appears to have been focused on intensive modification within a restricted area, producing high tooth‐mark frequencies without complete removal of remains.\u003c/p\u003e \u003cp\u003eThe role of geometric complexity, captured by Minkowski functionals (especially Euler characteristic) adds another layer of interpretation. High absolute Euler values, indicating fragmented or topologically complex spatial structures, tend to align with higher TM and lower survival. This is particularly evident for Elephant_2 and Elephant_3, where complex spatial topology coincides with strong evidence of carnivore processing. In contrast, assemblages with simpler spatial topology (lower Euler magnitudes) tend to preserve skeletal material more effectively. This suggests that spatial fragmentation itself may be a downstream consequence of carnivore behavior, reflecting repeated disturbance, displacement, and reorganization of remains across the landscape.\u003c/p\u003e \u003cp\u003eImportantly, TM and Survival do not map onto a single spatial metric but instead emerge from the combined effect of scale, clustering, and geometry. TM aligns most strongly with dimensions capturing local clustering intensity and interaction (PC2 and PC3), whereas survival aligns more closely with global spatial extent and dispersion (PC1). This distinction is biologically intuitive: tooth marks accumulate through repeated localized interactions between carnivores and bones, while bone loss is driven primarily by transport beyond the observation window.\u003c/p\u003e \u003cp\u003eAll this spatial information shows that the PCA reveals that spatial point-process structure is not merely descriptive but directly informative about underlying taphonomic processes. Assemblages with compact, tightly clustered point patterns tend to reflect limited carnivore impact and high survival, whereas spatially extensive, fragmented patterns are strongly associated with intensive carnivore modification and bone removal. Intermediate configurations capture mixed behavioral regimes, where carnivores repeatedly modify remains without fully dispersing them. Together, these results demonstrate that spatial statistics provide a powerful framework for disentangling carnivore behavior from the spatial organization of archaeological and paleontological assemblages.\u003c/p\u003e \u003cp\u003eWhen the spatial signatures derived from the modern elephant carcasses are taken as a comparative baseline, they provide a powerful interpretive framework for evaluating carnivore impact and additional taphonomic processes in archaeological elephant assemblages. In modern cases, the point patterns represent well-understood ecological processes: carcasses are initially deposited by natural mortality or predation, and then progressively modified by carnivore feeding, transport, and bone removal. The PCA shows that these processes leave consistent spatial fingerprints. Assemblages characterized by high tooth-mark frequencies (TM) and low survival percentages tend to occupy regions of PCA space associated with dispersed point patterns, large scatter areas, long maximum distances, and elevated mean pairwise distances. Conversely, assemblages with lower carnivore impact cluster in PCA regions dominated by higher local intensity contrasts, stronger clustering indices, and tighter spatial footprints. In this sense, the modern elephant carcasses define a continuum from spatially compact, minimally resedimented bone concentrations to highly dispersed, carnivore-modified assemblages. Unfortunately, all of this cannot be captured through the small windows of the archaeological sites. To interpret archaeological proboscidean sites adequately, we need therefore to exclude the larger picture obtained through the analyses of the extended windows and focus on the small windows represented by the sites themselves. Since no proboscidean site has been excavated beyond 100 m2 (approximately EAK\u0026acute;s size) [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. This requires us to limit our referents to their densest smallest windows (Stage 3 analysis).\u003c/p\u003e \u003cp\u003eWhen the modern spatial\u0026ndash;taphonomic gradient derived from the small-window approach of modern elephant carcasses is used to contextualize the archaeological assemblages, \u0026Aacute;ridos-2 emerges as an extreme but internally dense spatial configuration, rather than a highly dispersed one. Its strongly positive scores along PC1 (driven by high Linhom intensity measures and reduced spatial extent) indicate a concentration of remains within a relatively compact area characterized by elevated local point density. At the same time, the combination of a high quadrat statistic and strong deviation from complete spatial randomness suggests that this compactness is not simply the product of homogeneous accumulation, but instead reflects structured aggregation within the observation window. This fits well with the almost fully articulated nature of the preserved remains, despite the carnivore damage documented on some of the bones [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn PCA space, \u0026Aacute;ridos-2 occupies a position that partially overlaps with modern carcass configurations associated with substantial carnivore modification, though through a different spatial pathway than simple dispersal. Rather than exhibiting extensive spatial spread, \u0026Aacute;ridos-2 shows evidence for intensive impact within a confined area, consistent with repeated carnivore access, trampling, and localized bone rearrangement. Its proximity to modern carcasses with elevated tooth-mark frequencies and reduced skeletal survival supports an interpretation of significant carnivore involvement, likely involving bone consumption and short-distance redistribution rather than wholesale removal across a broad landscape. This pattern is compatible with a scenario in which carnivores acted repeatedly on an initially concentrated accumulation, generating spatial complexity and fragmentation without producing a highly dispersed carcass field.\u003c/p\u003e \u003cp\u003eMarathousa 1, by contrast, occupies a very different position relative to the modern baseline. Its point pattern is comparatively compact, with low pcf indices, modest scatter area, short maximum distances, and low Finhom-based neighborhood estimates. In PCA space, Marathousa 1 aligns more closely with assemblages showing limited dispersion and relatively stable spatial structure. This configuration resembles modern scenarios with lower carnivore transport intensity, where bones remain near the original deposition area and spatial patterning is dominated by primary accumulation processes rather than extensive secondary modification. Under this light, Marathousa 1 appears consistent with a scenario in which carnivores were present but did not extensively dismantle or disperse the assemblage, allowing spatial coherence to persist.\u003c/p\u003e \u003cp\u003eEAK represents a third and more complex case. Its exceptionally high Linhom values, strong clustering index, and large quadrat statistics point to intense local aggregation of remains, while at the same time its Finhom and pair-correlation indices suggest non-trivial spatial structuring across multiple scales. In PCA space, EAK falls near modern assemblages that combine high carnivore interaction with strong spatial anchoring (patterns seen when carcasses are heavily exploited but bones are repeatedly revisited rather than widely transported). This dual signal is consistent with scenarios involving prolonged carnivore access at a relatively fixed location, such as denning or repeated scavenging locales. Compared to \u0026Aacute;ridos-2, EAK shows less evidence for extreme dispersal, but more evidence for sustained carnivore-driven modification than Marathousa 1. The tectonic deformation of the assemblage might also have impacted the original clustering, making it extraordinarily and artificially compact, as detected in the spatial statistical analysis.\u003c/p\u003e \u003cp\u003eThe interpretation of these archaeological assemblages is just preliminary, and must be carried out with extreme caution, since they are dimensionally different from the analogical modern carcasses studied in this work, despite our interpretation being limited to the small-window analysis (Stage 3).\u003c/p\u003e \u003cp\u003eCrucially, however, the interpretive patterns that were derived in this study -especially the contrasts between modern elephant carcasses and archaeological assemblages- remain robust in a relative sense. While window size affects absolute values, it is unlikely to reverse the observed gradients: modern carcasses consistently show various degrees of clustering, spatial spread, and carnivore-driven removal signatures, whereas \u0026Aacute;ridos-2, Marathousa 1, and EAK exhibit within-window limited dispersion, some spatial fragmentation, and signatures consistent with prolonged carnivore modification. As long as windows are taphonomically meaningful, and comparisons are made cautiously across similarly defined spatial frames, window size modulates magnitude but does not invalidate the taphonomic interpretations. Therefore, window size matters greatly for scale, magnitude, and comparability, but when handled consistently, it does not undermine the core conclusions about spatial organization, site formation processes, or carnivore impact.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe main hypothesis of carnivore impact and its related scattering point processes was tested successfully. This is reflected in all carcasses having undergone high carnivore impact (with tooth marked bones in frequencies\u0026thinsp;\u0026gt;\u0026thinsp;20%), showing low bone survival values (\u0026lt;\u0026thinsp;25%), and displaying a higher degree of scattering as well as lower spatial clustering (Elephant_2, 4 and 5). However, the opposite is not true; carcasses with high degree of clustering can also display intense carnivore impact (reflected both in tooth marking frequencies and low bone survival (e.g., Elephant_1 and 3). Even carcasses with low carnivore impact (as reflected in high bone survival rates and low tooth mark frequencies) can be locally scattered to the point of not creating significant clustering assemblages (e.g., Elephant_6 and Giraffe). An alternative way of interpreting the documented point patterns is that (regardless of how much \u003cem\u003ein locus\u003c/em\u003e gnawing tooth place) all of them have undergone variable but intensive carnivore impact, since \u0026gt;\u0026thinsp;50% (in some cases up to \u0026gt;\u0026thinsp;85%) of the original carcass has been removed from its depositional spot.\u003c/p\u003e \u003cp\u003eUnder the modern observational/experimental referential point, patterns generated by the study of the modern elephant (and giraffe) carcasses, and relating them to the taphonomic evidence of carnivore impact, under two different (but complementary) processes (bone transport signaled by the carcass bone survival rates) and in locus bone consumption indicated by gnawing intensity and the resulting bone modification and tooth-marking), it can be argued that the documented point patterns find matching with higher or lower degrees of impact of carnivores on elephant carcasses. When used as a reference for the archaeological elephant butchery assemblages of EAK, \u0026Aacute;ridos-2, and Marathousa 1, as a preliminary testing ground, the archaeological assemblages link to specific referents showing different degrees of taphonomic modification by carnivores. The intense clustering of \u0026Aacute;ridos-2 and EAK indicates a limited time of exposure, and probably, an early access by hominins in the early stages of carcass resource exploitation targeting bulk flesh extraction. This is convincingly shown in the presence of cut marks on the pelvic and axial portion of the elephant skeleton at \u0026Aacute;ridos-2 [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis first testing spatial taphonomic experiment for archaeological interpretation is promising. It warrants further and more detailed analysis of modern and archaeological elephant carcasses, so that we can understand how non-primate carnivores and hominins interacted in the consumption of elephants. Given that most of the historically-documented elephant butchery leaves very few traces on bones, the lack of direct evidence underscores the need of developing alternative and complementary methods to detect in which moment hominins had access to elephant carcasses and what resources they extracted from them. Ultimately, interpreting elephant bone accumulations through a taphonomic lens enables researchers to evaluate key hypotheses about the economic capabilities and ecological roles of early humans, including their capacity for timely accessing key proboscidean flesh and fat anatomical sections, extent of their cooperative behavior, and the use of technology to access high-yield resources. Given the energetic potential of a single elephant carcass, even rare instances of access could have had profound effects on hominin diet, social organization, and territoriality. The integration of controlled taphonomic observations with fossil assemblages is therefore indispensable for advancing our understanding of hominin-megafauna interactions during the Pleistocene.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n\u003ch2\u003eNeo-taphonomic samples\u003c/h2\u003e\n\u003cp\u003eThe present study was conducted in the Chobe National Park and in the Okavango delta (Botswana) during the month of June 2024. We sampled seven carcasses (six elephants and one giraffe for control as to what the taphonomic differences could be when compared to a smaller megafaunal taxon) (Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). All the carcasses that were examined and whose information was collected had been either discovered by local rangers or game drivers and approximate dates of death were provided. The exposure time of the carcasses ranges from three and a half months to more than two years (Supplementary Table\u0026nbsp;3). Given that all the carcasses spotted were inside both National parks and reserves, we limited our study to the documentation of the conspicuous elements that were visually observed on the ground and did not engage in intensive screening of the ground for small bone fragments. Bones were identified to elements using estimates of the minimum number of elements (MNE). Given the overall completeness or limited fragmentation of most non-axial bones, quantifications of those anatomical areas were very precise. Regarding the axial skeleton, estimates of vertebrae were also very accurate (with minimal modification, and mostly restricted to damage on the arch processes). Estimates of ribs were very conservative, since we did not collect (in those carcasses affected by more intense fragmentation) all the small fragments to piece them together. Although for most of the sample, ribs were fairly complete, despite the high percentage of carnivore damage documented on them, in a couple of instances, fragmentation was more intense. In those cases, MNE estimates were derived using a combination of criteria involving: all ribs that were preserved in more than one third of their size, and the presence of specimens with tubercles and heads.\u003c/p\u003e\n\u003cp\u003eCarnivore impact was documented macroscopically. The size and density of proboscidean bones make it very difficult for hyenas to break them, and when they do so, they leave abundant and conspicuous traces of gnawing and tooth marks (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). For this reason, we did not quantify the smaller isolated tooth marks that might have required higher magnification. We tallied carnivore damage as the presence of tooth pits, scores and furrowing; most commonly occurring in combination in most of the specimens affected.\u003c/p\u003e\n\u003cp\u003eWe also quantified skeletal survival indices, based on the number of preserved bones compared to complete skeletons. The highly skewed profiles were caused by the predominant deletion/disappearance of ribs. It was also due to the more intense spatial scattering of these elements. In order to keep spatial windows in maximum comparable proportion to archaeological sites, we did not collect information from bone scatters occurring on a radius of more than ~\u0026thinsp;25 m from the main cluster.\u003c/p\u003e\n\u003cp\u003eAll the spatial information of carcasses was collected through the spatial reconstruction of the delimited surfaces containing the bulk of the accumulation and scattering of bones from each carcass (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ea-d). The data were obtained with various software tools, employed both during fieldwork and in the laboratory. The first step in obtaining the spatial data consisted of creating initial photogrammetric models using an iPad Pro (Apple Inc.) and the 3d Scanner App\u0026trade;. The use of this device's LiDAR technology allows not only for rapid documentation of the study areas, but also for the real-time scaling of the 3D models. We did at least two 3D models per carcass to ensure that we could retrieve all information later in the laboratory. This was key to obtaining data from bones located at the periphery of the models.\u003c/p\u003e\n\u003cp\u003eOnce in the laboratory, the 3D models were exported, including their textures and all the photographs taken by the 3d Scanner App\u0026trade; used to generate them. Subsequently, the 3D models were imported into Blender 4.2; since they were already scaled, a 1x1 m grid was created and the light sources were adjusted to achieve the best possible view. The models were rendered including the grid to generate orthophotographs in .tiff format. The Cycles rendering engine was used to prioritize maximum image quality. However, as the quality of the orthophotos generated with 3d Scanner App\u0026trade;/Blender 4.2 is not particularly high, higher-quality orthophotographic models were subsequently generated in Agisoft Metashape using the photos exported from 3d Scanner App\u0026trade;. This provided a .tiff format orthophotograph of higher quality, but without scaling.\u003c/p\u003e\n\u003cp\u003eTherefore, both orthophotographs were then imported into ArcMap 10.5, and using the grid generated in Blender 4.2, both orthophoto sets were scaled. Once the high-quality photogrammetry was scaled, .shp files were generated to digitize the polygons of the bones from the different samples. Once we had the polygon layer completed, we generated a point layer from the polygons to obtain the \u003cem\u003exy\u003c/em\u003e coordinates of the centroids of each bone through the \u0026lsquo;Feature to Point\u0026rsquo; function (selected features: \u003cem\u003eInside (optional): checked\u003c/em\u003e). In addition, we also created point layers to generate the spatial windows (i.e., analysis areas) that would be used later.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn addition to the experimental/observational carcasses that we documented, we also added three additional modern elephant carcasses. Two were from Hayne\u0026acute;s work in Zimbawe [58: Figs.\u0026nbsp;10 and 13], and the third one was from White and Diedrich [8: Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e] in Zambia. These carcasses were selected due to the high quality of the figures at our disposal, which was key to the accurate digitization of the floor plans and the subsequent extraction of spatial data. The methodology for spatial data collection remained consistent with the protocols established for our documented carcasses. In the three cases, not quantifiable information was provided on carnivore damage, and for this reason, we excluded these three samples from the analysis when the focus was taphonomic specific (analytical stage 2).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n\u003ch2\u003eArchaeological samples\u003c/h2\u003e\n\u003cp\u003eFor the study of archaeological samples, we selected three previously published sites (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ee-g), each containing a single anthropogenically modified elephant carcass. The selection of these sites was based on the feasibility of accurately digitizing the published cartographies, as certain cases lacked sufficient detail to ensure reliable map digitization [\u003cspan class=\"CitationRef\"\u003e59\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e60\u003c/span\u003e]. The fact that well-verified prior information regarding the anthropogenic modification of the carcasses existed was also taken into account.\u003c/p\u003e\n\u003cp\u003eThe first site was recently published by our research team [\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e], with the cartography already available. The site is named Emiliano Aguirre Korongo (EAK) and is located at the base of Bed II in Olduvai Gorge (Tanzania). It is located at the confluence of the two gorges, between the sites of FLK-N and FLK-NN [\u003cspan class=\"CitationRef\"\u003e61\u003c/span\u003e]. Stratigraphically, the site rests directly upon Tuff 1F (1.78 Ma), which marks the beginning of Bed II [\u003cspan class=\"CitationRef\"\u003e62\u003c/span\u003e]. The sedimentary context of the assemblage corresponds to a low-energy clay environment with seasonal inputs of coarser sediments. The faunal material from EAK consists of a minimum number of 46 elements from a Elephas recki carcass. Although no cut marks were identified in the assemblage, two specimens exhibit green breakage interpreted as anthropogenic. The absence of several highly significant anatomical elements (e.g., many vertebrae and ribs, bones from three feet, and the femoral epiphysis) suggests carnivore activity within the assemblage. Furthermore, 80 stone tools, mostly flakes and flake fragments, were identified in the same area, showing a spatial correlation with the elephant carcass.\u003c/p\u003e\n\u003cp\u003eThe \u0026Aacute;ridos-2 site (Arganda del Rey, Spain), excavated in 1976 [\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e63\u003c/span\u003e], is located in the complex terrace of the Jarama River, specifically within the Arganda I stratigraphic unit [\u003cspan class=\"CitationRef\"\u003e64\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e65\u003c/span\u003e], dated between MIS 10\u0026thinsp;\u0026minus;\u0026thinsp;9 [\u003cspan class=\"CitationRef\"\u003e66\u003c/span\u003e]. The site revealed the presence of a Palaeoloxodon antiquus carcass associated with 34 Acheulean lithic artifacts [\u003cspan class=\"CitationRef\"\u003e56\u003c/span\u003e] within a context of muddy overbank deposits and secondary pebble and sandy low-energy channels on an alluvial plain [\u003cspan class=\"CitationRef\"\u003e64\u003c/span\u003e]. The elephant remains were largely preserved in anatomical connection. The presence of cut marks has been demonstrated [\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e], some of which are consistent with those generated by bifacial tools [\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e67\u003c/span\u003e]. Furthermore, microwear analysis of the lithic tools supports anthropogenic intervention on the carcass, documenting the use of tools for intensive butchery activities [\u003cspan class=\"CitationRef\"\u003e68\u003c/span\u003e]. The digitized map is Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e from Santonja and Querol [\u003cspan class=\"CitationRef\"\u003e56\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe other site is Marathousa 1, located in the Megalopolis basin (Greece) [\u003cspan class=\"CitationRef\"\u003e57\u003c/span\u003e] and dated between 0.48 and 0.42 Ma [\u003cspan class=\"CitationRef\"\u003e69\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e71\u003c/span\u003e]. Sedimentologically, the site corresponds to an extensive mudflat surrounding a lake shore [\u003cspan class=\"CitationRef\"\u003e72\u003c/span\u003e]. As with \u0026Aacute;ridos-2, a single Palaeoloxodon antiquus carcass associated with lithic industry was identified at this site [\u003cspan class=\"CitationRef\"\u003e57\u003c/span\u003e]. The lithic sample is significantly larger than in the previous cases (n\u0026thinsp;=\u0026thinsp;1,170), with a notable presence of chips and flakes; the use of certain bones as tools has also been proposed [\u003cspan class=\"CitationRef\"\u003e73\u003c/span\u003e]. Another distinction from the other sites is the greater presence of remains from smaller ungulates (omitted in this analysis), which also exhibit evidence of anthropogenic exploitation [\u003cspan class=\"CitationRef\"\u003e74\u003c/span\u003e]. Other animals, such as birds, are present at the site, although their presence appears to be related to natural accumulation in this lacustrine environment rather than human activity [\u003cspan class=\"CitationRef\"\u003e75\u003c/span\u003e]. A previous spatial study [\u003cspan class=\"CitationRef\"\u003e76\u003c/span\u003e] also demonstrated the minimal alteration of the archaeological assemblage and the existence of a spatial association between the lithic tools and the faunal remains. The digitization of the site cartography was performed using Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e from Panagopoulou et al. [\u003cspan class=\"CitationRef\"\u003e57\u003c/span\u003e], as it offered the highest quality among the publications available to date.\u003c/p\u003e\n\u003cp\u003eThe digitization of the cartography for \u0026Aacute;ridos-2 and Marathousa 1 followed the same process as the previously published modern elephant carcasses [\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e58\u003c/span\u003e]. In the case of EAK, cartographic data were acquired during excavation; specimen centroids were recorded using a total station, and the polygons for each bone were subsequently digitized using orthophotographs.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n\u003ch2\u003eSpatial statistical analysis\u003c/h2\u003e\n\u003cp\u003eThe spatial analysis was carried out using the comprehensive R [\u003cspan class=\"CitationRef\"\u003e77\u003c/span\u003e] \u0026ldquo;spatstat\u0026rdquo; library [\u003cspan class=\"CitationRef\"\u003e78\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e80\u003c/span\u003e]. This software has previously been used to analyze different archaeological contexts with satisfactory results for exploratory spatial point pattern distribution analyses [\u003cspan class=\"CitationRef\"\u003e79\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e81\u003c/span\u003e], mainly seeking to determine the degree of randomness in point patterns and whether or not a correlation existed among different types of archaeological materials [\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e82\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e94\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe first step of the study was to carry out an exploratory analysis of the spatial properties of each of the case studies. The methods used and the results are presented in the Supplementary Files. These initial analyses already showed that the techniques applied in conventional spatial studies are not sufficient to clearly differentiate the spatial patterns of the cases examined. Accordingly, we carried out a more in-depth study and information was collected on several spatial tests, indices and measurements and articulated within a hyperframe format. The use of hyperframes has already been shown to be highly useful for comparing spatial patterns in archaeological contexts [\u003cspan class=\"CitationRef\"\u003e88\u003c/span\u003e]. The hyperframe was subsequently expanded to fit the taphonomic variables (carnivore impact and bone survival), so that exploratory (principal component analyses) and similarity-dissimilarity (phylogenetic hierarchical cluster) analyses could be carried out. The hyperframe used for the present analysis included a set of variables determined by first and second order tests, as well as topographic factorials. These are described in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab2\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eVariables used in the hyperframe for the spatial statistical analysis and the subsequent cluster analysis on the PCA-transformed data set.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eVariable\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eDescription\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eLinhom_ml\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean value of the estimated inhomogeneous intensity function evaluated at the observed point locations; reflects the average local density experienced by points.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eLinhom_mn\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean expected number of neighbors derived from the inhomogeneous L-function; summarizes local clustering relative to the estimated intensity.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eLinhom_mlg\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean value of the inhomogeneous intensity function over the entire observation window; represents global background density.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eLinhom_clust\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eRatio between local mean intensity and global mean intensity; values\u0026thinsp;\u0026gt;\u0026thinsp;1 indicate concentration of points in high-intensity regions, values\u0026thinsp;\u0026asymp;\u0026thinsp;1 indicate consistency with the background intensity.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eFinhom_nn\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean expected number of points within radius \u003cem\u003er\u003c/em\u003e based on the inhomogeneous empty-space (F) function; captures spacing relative to local intensity variation.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003epcf_index\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSummary deviation of the homogeneous pair correlation function from complete spatial randomness; larger values indicate stronger clustering or regularity.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003epcf_index2\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSummary deviation of the inhomogeneous pair correlation function from randomness after correcting for spatial variation in intensity.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003ehopskel\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean Hopkins\u0026ndash;Skellam statistic across simulations; values\u0026thinsp;\u0026lt;\u0026thinsp;0.5 indicate clustering, values\u0026thinsp;\u0026asymp;\u0026thinsp;0.5 randomness, and values\u0026thinsp;\u0026gt;\u0026thinsp;0.5 regularity.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003equadrat_stat\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eChi-square statistic from quadrat counts; large values indicate strong deviation from spatial randomness due to clustering or regular spacing.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eNND_index\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean nearest-neighbor distance scaled by the square root of local intensity; integrates spacing and density into a single metric.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eM_area\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eArea of the union of discs around points at a fixed radius (Minkowski functional); reflects spatial footprint and point aggregation.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eM_perimeter\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003ePerimeter of the same union of discs; sensitive to boundary complexity and fragmentation of point clusters.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003eM_euler\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eEuler characteristic of the union of discs; captures topological structure (number of connected components minus holes), with lower values indicating clustering.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003escatter_area\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eArea of the convex hull enclosing all points; represents overall spatial spread\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003emax_distance\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMaximum pairwise distance between points; reflects spatial extent of the pattern.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u003cstrong\u003emean_pairwise\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eMean pairwise distance among all point pairs; summarizes overall dispersion.\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eTo characterize the spatial intensity of each point pattern and assess the extent to which its distribution deviates from a completely random process, we first estimated the inhomogeneous intensity surface using a kernel-based estimator. Specifically, we computed a continuous density field from the point pattern using an optimal bandwidth selector (bw.ppl), which provides an estimate of intensity \u0026lambda;(x) that accounts for local spatial heterogeneities. This step is essential in contexts where intensity is not constant across the study area, as it prevents misinterpretations that would arise from assuming homogeneity when the underlying process is spatially variable.\u003c/p\u003e\n\u003cp\u003eBased on this intensity estimate, we applied the inhomogeneous L-function, an extension of Ripley's L-function that explicitly corrects for spatial variation in \u0026lambda;(x). This function quantifies spatial dependence between points while controlling for non-uniform background intensity, thereby distinguishing true clustering from patterns produced solely by changes in the underlying intensity surface. For each pattern, we calculated the mean intensity at the exact point locations, providing a direct measure of the \u0026ldquo;effective density\u0026rdquo; experienced by the points themselves. We then estimated the expected number of neighbors under the inhomogeneous model by multiplying this mean local intensity by the values of the inhomogeneous L-function.\u003c/p\u003e\n\u003cp\u003eTo determine whether points tend to concentrate in areas of higher-than-average intensity, we compared this local intensity measure with the global mean intensity of the entire observation window. The ratio between these values (here referred to as the relative clustering index) quantifies whether the pattern preferentially occupies high-intensity regions (values\u0026thinsp;\u0026gt;\u0026thinsp;1), whether its distribution is neutral with respect to \u0026lambda;(x) (\u0026asymp;\u0026thinsp;1), or whether points tend to occur in lower-intensity regions (\u0026lt;\u0026thinsp;1). This index thus provides a robust, normalized measure of spatial aggregation, enabling meaningful comparisons between point patterns originating from different spatial contexts or exhibiting widely differing overall densities.\u003c/p\u003e\n\u003cp\u003eIn addition to the L-function analysis, we evaluated the spatial structure of each point pattern using the inhomogeneous empty-space F-function. Whereas the L-function describes the spatial relationships between the points themselves, the F-function, instead, characterizes the distribution of empty space relative to the points. This distinction provides a complementary perspective on spatial structure: while clustering of points may indicate aggregation, the empty-space function reveals how the surrounding landscape of voids is organized. To compute the inhomogeneous F-function, we again began by estimating a spatially varying intensity surface using the same kernel-based estimator and bandwidth selection procedure described above. The resulting intensity field \u0026lambda;(x) was used to correct the F-function for non-uniformity of the underlying process, ensuring that the probability of observing empty space at increasing distances from random locations is not confounded by local variations in density.\u003c/p\u003e\n\u003cp\u003eUsing this corrected version of the F-function, we derived an estimate of the expected number of points within a disk of radius r by multiplying the mean local intensity (evaluated at the observed point locations) by the area of the search region (\u0026pi;r\u0026sup2;). Averaging these values across all distances r provided a single summary statistic representing the overall expected neighbor count in the inhomogeneous context. This measure serves as an intuitive indicator of how \u0026ldquo;filled\u0026rdquo; or occupied the space is when local intensity variations are taken into account. To express the degree to which the pattern tends toward high- or low-intensity regions, we again compared the mean intensity at the point locations with the global mean intensity across the observation window. The resulting ratio, analogous to the one used for the inhomogeneous L-function, quantifies whether points preferentially occur in regions of higher intensity (\u0026gt;\u0026thinsp;1), are distributed in proportion to the overall intensity surface (\u0026asymp;\u0026thinsp;1), or disproportionately occupy low-intensity areas (\u0026lt;\u0026thinsp;1). This index enables direct comparison between patterns and offers insight into whether spatial aggregation is driven by genuine interaction processes or simply reflects the geometry of intensity variation.\u003c/p\u003e\n\u003cp\u003eTo further characterize the fine-scale structure of spatial interactions, we computed both the homogeneous and inhomogeneous pair correlation functions. Whereas the L- and F-functions evaluate cumulative departures from randomness across increasing spatial scales, the pair correlation function provides a \u003cem\u003elocalized\u003c/em\u003e measure of clustering or inhibition at specific distances. Conceptually, pair correlation quantifies the probability of finding a neighboring point at distance r relative to what would be expected under complete spatial randomness. Thus, values greater than 1 indicate clustering at that distance, while values below 1 reflect spatial inhibition or regular spacing.\u003c/p\u003e\n\u003cp\u003eWe first computed the standard (homogeneous) pair correlation function for each pattern. Because pair correlation is known to behave unstably at very small spatial lags (where edge effects and sampling noise can produce extreme or undefined values) we removed radii smaller than 2% of the maximum r-range. This trimming step ensured that the resulting summary statistic reflected meaningful spatial structure rather than numerical artifacts. We then summarized the pair correlation curve by computing the mean absolute deviation of pair correlation from the value expected under a Poisson process. This yielded a single index, the PCF clustering index, which quantifies the overall magnitude of deviation from spatial randomness across all evaluated scales.\u003c/p\u003e\n\u003cp\u003eBecause spatial patterns often exhibit inhomogeneous intensity, we also calculated the inhomogeneous pair correlation function using isotropic edge correction. This version of pair correlation accounts explicitly for spatially varying intensity \u0026lambda;(x), thereby isolating interaction effects from broader trends in density. As with the homogeneous case, we trimmed small radii here (excluding values below 5% of the radius range) to prevent instability in the estimator. The mean absolute deviation of the inhomogeneous pair correlation curve from 1 provided a second summary statistic reflecting interaction strength in the context of varying intensity. Together, the homogeneous and inhomogeneous pair correlation indices allow us to distinguish whether apparent spatial structure arises from genuine point\u0026ndash;point interactions or merely reflects large-scale variation in the underlying intensity surface. The homogeneous index captures deviations without accounting for intensity variation, while the inhomogeneous index corrects for these trends, thus isolating true interaction processes.\u003c/p\u003e\n\u003cp\u003eTo complement the above second-order analyses, we implemented the Hopkins\u0026ndash;Skellam statistic, a classical test for detecting departures from spatial randomness. Unlike Ripley\u0026rsquo;s functions or the pair correlation function, which focus on spatial dependence across a range of distances, the Hopkins\u0026ndash;Skellam test evaluates the \u003cem\u003ebalance between clustering and regularity\u003c/em\u003e based on nearest-neighbor distances. The statistic compares the distances from randomly selected locations to the nearest observed point with the distances between observed points themselves. Under complete spatial randomness, these two distributions should be similar; deviations toward smaller observed distances indicate clustering, whereas larger distances reflect regularity.\u003c/p\u003e\n\u003cp\u003eBecause the Hopkins\u0026ndash;Skellam statistic is inherently stochastic and sensitive to the particular random samples drawn, we estimated a more stable value by averaging across multiple repetitions. For each point pattern, we ran 99 independent iterations of the test and computed the mean statistic as the final index. Lower values (\u0026lt;\u0026thinsp;0.5) indicate a tendency toward clustering, while values approaching or exceeding 0.5 suggest greater regularity or randomness. This resampling-based implementation reduces the influence of sampling noise and yields a more reliable summary of the overall spatial structure.\u003c/p\u003e\n\u003cp\u003eTo further characterize large-scale deviations from spatial randomness, we calculated a quadrat-based chi-square statistic. Quadrat tests partition the observation window into a regular grid and compare the observed number of points in each cell with the numbers expected under a homogeneous Poisson process. Whereas second-order functions capture dependence across continuous distances, quadrat statistics are particularly effective in identifying coarse-grained aggregation or voids that manifest at the scale of the grid cells. Because the sensitivity of quadrat tests depends strongly on the chosen grid resolution, we adopted a data-driven approach in which the number of quadrats scales with the total number of points. Specifically, for each pattern we computed q\u0026thinsp;=\u0026thinsp;nq = \\sqrt{n}q\u0026thinsp;=\u0026thinsp;n, where n is the number of points, and rounded this value to obtain both the number of quadrats along the x- and y-axes. This adaptive scheme maintains an approximate balance between the number of quadrats and the number of points in the pattern, avoiding over- or undersplitting the window.\u003c/p\u003e\n\u003cp\u003eWe then extracted the chi-square statistic returned by the test, which quantifies the degree to which the observed distribution of point counts deviates from spatial randomness. Larger values indicate stronger heterogeneity (either clustering or regular spacing) whereas values closer to zero are consistent with a homogeneous Poisson process. Although the chi-square approximation may be inaccurate when expected counts are low, using a grid proportional to the density of points helps mitigate this limitation and yields a more interpretable measure of large-scale spatial structure.\u003c/p\u003e\n\u003cp\u003eTo complement the quadrat analysis and second-order statistics, we computed an intensity-normalized nearest-neighbor distance index (NNDI). Nearest-neighbor distances are a classical metric for assessing spatial structure, but raw nearest-neighbor values can be misleading in the presence of strong intensity gradients. In regions of high point density, nearest-neighbor distances will naturally be smaller even if the pattern is random, while in sparse regions distances will be larger. To address this, we normalized each nearest-neighbor distance by the square root of the local intensity estimate. This adjustment follows theoretical expectations under an inhomogeneous Poisson process, where the expected nearest-neighbor distance scales inversely with the square root of the intensity.\u003c/p\u003e\n\u003cp\u003eFor each point in the pattern, we first computed the Euclidean distance to its nearest neighbor. We then extracted the corresponding local intensity values from the kernel intensity estimator and multiplied each distance by the square root of its associated intensity. The resulting dimensionless NNDI values quantify how \u0026ldquo;close\u0026rdquo; or \u0026ldquo;far\u0026rdquo; points are relative to what would be expected given the local density of the pattern. We then averaged these normalized values across all points to derive a single index per pattern. Low values indicate clustering relative to local intensity, values near one approximate spatial randomness, and high values suggest local inhibition or regularity. This index thus offers a robust, first-order\u0026ndash;adjusted view of local spatial structure that is not confounded by spatially varying point densities.\u003c/p\u003e\n\u003cp\u003eTo quantify higher-order spatial structure beyond classical point-pattern measures, we computed a set of Minkowski functionals. These functionals describe the geometry and topology of a point pattern after morphological dilation, and they are widely used in fields such as cosmology, materials science, and ecology because they summarize global shape properties in ways that are sensitive to clustering and void structure. Applied to archaeological and paleoanthropological spatial datasets, Minkowski functionals provide a complementary descriptor of spatial organization, capturing aspects of point aggregation that are not easily detected using local statistics or second-order functions.\u003c/p\u003e\n\u003cp\u003eTo compute these metrics, we first established an appropriate range of dilation radii. The choice of radius is critical because the morphological expansion of points must reflect meaningful physical distances relative to the size of the observation window. Very small radii amplify pixel noise and exaggerate Euler characteristic fluctuations, while excessively large radii saturate the window and erase meaningful structure. We therefore determined radius values adaptively based on the spatial dimensions of the study area and on the mean nearest-neighbor distance within each pattern. Following standard morphological analysis practice, radii were automatically set between 5% and 40% (depending on the point pattern properties) of the smallest window dimension but were also constrained to avoid radii that were either far smaller or far larger than typical interpoint distances. This dual-criterion approach ensured that Minkowski functionals remained sensitive to real structural differences rather than artifacts of scale mismatch. We also implemented an automated procedure to recommend radii and image resolution parameters for each point pattern, ensuring consistent spatial resolution across datasets of differing sizes and densities.\u003c/p\u003e\n\u003cp\u003eOnce radius and resolution parameters were adaptively defined for each point process, we constructed a high-resolution distance map over the observation window, which computes for every pixel the distance to its nearest point. This raster representation enabled the creation of a binary mask identifying all pixels lying within a distance \u003cem\u003er\u003c/em\u003e of any point, effectively creating a union of disks around the points. Minkowski functionals were then computed directly from this binary mask. The area functional quantifies the total dilated surface occupied by the union of disks, reflecting overall clustering and spread. The perimeter functional, estimated by counting transitions between foreground and background pixels along both horizontal and vertical edges, measured boundary complexity and the fragmentation of occupied space. Finally, the Euler characteristic was computed using a digital topology estimator based on 2\u0026times;2 pixel blocks. This measure captures the number of connected components minus the number of holes in the dilated region and is particularly sensitive to changes in spatial topology as the radius increases.\u003c/p\u003e\n\u003cp\u003eTo mitigate well-known instabilities in Euler characteristic estimation (especially at high resolution or when disk radii approach the pixel size) we implemented two safeguards: (1) moderating resolution to avoid excessive pixel-level fragmentation, and (2) using a 4-connectivity estimation approach that is stable for binary morphological objects. These strategies prevented pathological behavior and ensured that Minkowski values varied appropriately across different spatial patterns. The Minkowski metrics derived for each point pattern (area, perimeter, and Euler characteristic) provided complementary information on spatial structure at a scale explicitly tied to the spatial extent of the underlying archaeological surfaces. These functionals allowed us to capture differences in clustering intensity, boundary roughness, and topological complexity that were not detectable using more traditional spatial statistics. By integrating these geometric descriptors with first- and second-order point-pattern measures, we obtained a comprehensive view of spatial structure across our comparative dataset.\u003c/p\u003e\n\u003cp\u003eTo further characterize the spatial extent and overall occupied footprint of each assemblage, we calculated several global measures based on the geometric configuration of the points. First, we computed the convex hull of each pattern, which represents the smallest convex polygon that encloses all points. The area of this hull provides a robust estimate of the total spatial footprint of the assemblage that is minimally influenced by internal voids or irregularities. Unlike density-based measures, which depend on local point concentrations, the convex-hull area captures the overall spread and territorial extent of the distribution, offering an integrative measure of how widely dispersed or spatially constrained a point pattern is within its sampling window.\u003c/p\u003e\n\u003cp\u003eIn addition to the hull, we quantified the maximum diameter of each point pattern by computing the full matrix of pairwise interpoint distances and extracting the greatest value. This diameter corresponds to the largest linear separation between any two points and serves as a simple but informative descriptor of the maximum spatial reach of the assemblage. Because this measure is independent of local clustering or density heterogeneity, it provides a complementary indicator of large-scale dispersion compared to area-based metrics.\u003c/p\u003e\n\u003cp\u003eFinally, we calculated the mean pairwise distance among all unique point pairs. This statistic captures the average spatial separation across the entire pattern and is sensitive to both clustering and broad-scale spread. While the maximum diameter highlights extreme separation, and the convex-hull area defines the spatial envelope, the mean pairwise distance describes the typical spacing among points throughout the distribution. Taken together, these three global measures (convex-hull footprint, maximum interpoint distance, and mean pairwise separation) characterize the overall spatial envelope and extent of each assemblage, providing a coarse-grained spatial signature that complements the more detailed first-, second-, and higher-order statistics described above.\u003c/p\u003e\n\u003cp\u003eOnce the hyperframe was built, we converted it into a conventional data frame for multivariate statistics (first using only the spatial variables, and then adding the taphonomic variables) excluding the site identifier column so that only numeric variables informed the clustering. We then applied a principal component analysis (PCA) to modulate potential collinearity effects. We computed Euclidean distances between sites based on all standardized spatial descriptors, producing subsequently a distance matrix that quantified how different each site was from all others in multidimensional spatial-pattern space. This matrix allowed us to visualize dissimilarities using a distance heatmap, thereby revealing the relative closeness or divergence of sites across the full spatial-statistics profile.\u003c/p\u003e\n\u003cp\u003eWe then applied hierarchical agglomerative clustering to the distance matrix. Average linkage was used for the analysis, which merges clusters based on the mean intercluster distance. Dendrograms were generated to visualize the hierarchical structure of similarities among sites, with both conventional and phylogram-style graphics produced to facilitate interpretation. We also created circular dendrograms with cluster highlighting to illustrate potential grouping structure and to evaluate whether sites naturally segregated into distinct spatial-behavioral categories. Together, the hyperframe construction and the hierarchical clustering workflow served to integrate many independent spatial statistics into a unified inferential framework. This allowed us to assess whether sites shared common spatial structuring processes or instead exhibited distinct spatial signatures suggestive of different depositional, behavioral, or taphonomic dynamics.\u003c/p\u003e\n\u003cp\u003eThe overall hypothesis to be tested is that proboscidean carcasses highly impacted by carnivores will have a higher degree of bone dispersal (affecting bone survival percentages), affecting the spatial point patterns (i.e., either decreasing clustering or increasing the spatial window or both). Such a taphonomic factor should have a distinctive spatial point pattern associated. In order to evaluate this hypothesis, a three-stage analytical process was adopted:\u003c/p\u003e\na) The first approach involved the use of modern and archaeological carcasses together. This was done as an initial exploratory step, in which some cautions were contemplated. The main goal was to understand how modern carcasses documented to exhibit various degrees of carnivore impact generate spatial pattern ranges within which the archaeological samples could be interpreted.\u003cbr /\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eb) The second approach involved only those modern carcases for which a taphonomic assessment is available, with the main goal being understanding the post-depositional disturbance processes, regardless of their scale. Archaeological samples may be palimpsestic and affected by processes not documented in modern carcasses. For this reason, the overall phylogenetic cladogram of all the samples may be impacted by the archaeological samples. To remove their impact and better understand the interplay of taphonomic and spatial patterns, we restricted the analysis to the modern carcasses. Thus, overall variance is caused by immediate postdepositional processes; namely, the impact of carnivore disturbance. This second approach aims to capture all the information within a reasonable radius around the main carcass cluster and is, therefore, conditioned by spatial windows of variable sizes.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ec) The third approach tried to minimize the impact of the wide range of window sizes and, instead of focusing on reconstructing the association of spatial patterns and taphonomic processes, it intends to realistically approach proboscidean archaeological sites, which normally span much smaller windows than all the experimental/observational spatial windows used as analogical referents. The variables that we applied to the modern and archaeological point patterns may be affected by the heterogeneous scales, and diverse expressions of agglomerative/scattering patterns involved in the first approach. This is necessary to understand the modifying taphonomic processes, which operate at differential scales. To standardize spatial comparisons across assemblages and minimize the confounding effects of heterogeneous observation window sizes, we implemented a procedure to isolate the densest portion of each point pattern.\u003c/p\u003e\n\u003cp\u003eArchaeological proboscidean butchery sites are much smaller than the areas surveyed for modern elephant carcass dispersal. At the most fundamental level, the observation window defines the spatial frame within which intensity, distances, and spatial relationships are measured. Changing window size alters global intensity estimates (\u0026lambda;), which directly affects Linhom, Finhom, NND-based indices, and any statistic that rescales distances by intensity. A larger window with the same number of points lowers global intensity and tends to inflate apparent clustering, whereas a smaller or tighter window raises intensity and can make the same point pattern appear more regular or dispersed. This means that absolute values of Linhom-based clustering indices, Finhom-derived neighbor expectations, and NND indices are not window-invariant and must be interpreted comparatively only among assemblages analyzed with comparable window definitions.\u003c/p\u003e\n\u003cp\u003eWindow size also strongly conditions second-order statistics such as the pair correlation function and quadrat tests. Quadrat statistics scale with both the number of quadrats and the spatial extent of the window, so larger windows almost inevitably produce larger \u0026chi;\u0026sup2; values even if the underlying pattern structure is similar. Likewise, the maximum meaningful distance for PCF interpretation depends on window extent; expanding the window introduces larger inter-point distances and can dilute short-range clustering signals relative to background space. This is why trimming small radii and focusing on relative deviations from CSR, rather than absolute magnitudes, is essential when comparing sites.\u003c/p\u003e\n\u003cp\u003eThe influence of window size is especially critical for Minkowski functionals. Area, perimeter, and Euler characteristic are explicitly geometric and therefore scale with window extent, resolution, and edge effects. Larger windows increase the number of connected components and holes captured at a given radius, potentially exaggerating Euler values and footprint size even when local clustering is unchanged. Your adaptive-resolution strategy mitigates pixel instability, but it does not remove the fundamental dependence of Minkowski measures on the spatial domain. Consequently, these metrics are most reliable for within-study comparisons where window construction reflects real site boundaries (e.g., carcass scatter limits, excavation extents) rather than arbitrary clipping. This is why we emphasized the study of the modern referents over the three archaeological assemblages.\u003c/p\u003e\n\u003cp\u003eThe procedure that we implemented to overcome this scale-induced limitation is encapsulated in a function, which identifies and extracts a fixed-area window centered on the highest-density region of a spatial point process. First, a kernel density estimation (KDE) of the point pattern was computed with automatic bandwidth selection. This adapts the smoothing parameter to the spatial configuration of points. This step produces a continuous intensity surface representing the estimated local density of points across the observation window. Second, the location of maximum estimated intensity is identified by finding the pixel with the highest density value on the KDE surface. This location represents the spatial peak of point concentration and is interpreted as the center of the densest cluster within the pattern. The x- and y-coordinates of this peak are extracted to serve as the reference point for window definition. Third, a square observation window of fixed dimensions (10 m \u0026times; 10 m, corresponding to 100 m\u0026sup2;) is defined, centered on the density peak. The window is constructed symmetrically around the peak coordinates by extending half the side length in all directions. This standardized window size ensures that all point patterns are analyzed within an equivalent spatial extent, allowing direct comparison of spatial statistics across datasets.\u003c/p\u003e\n\u003cp\u003eFinally, the original point pattern is cropped to this newly defined window using spatial subsetting. Only points falling within the standardized high-density window are retained, producing a trimmed point pattern that captures the core spatial structure of the assemblage while excluding low-density peripheral areas. This approach allows spatial analyses to focus explicitly on the most informative regions of point concentration while controlling for differences in original window size and shape. By anchoring comparisons to equivalent, density-centered spatial extents, the method improves the robustness of cross-assemblage comparisons of clustering, dispersion, and other spatial characteristics.\u003c/p\u003e\n\u003c/div\u003e\n"},{"header":"Declarations","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n\u003ch2\u003eCode availability\u003c/h2\u003e\n\u003cp\u003eThe code used is available as a supplementary file.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n\u003ch2\u003eData availability statement\u003c/h2\u003e\n\u003cp\u003eAll data generated or analyzed during this study are included in this published article (and its Supplementary Information files)\u003c/p\u003e\n\u003c/div\u003e\u003cp\u003e \u003ch2\u003eAuthor contributions statement\u003c/h2\u003e \u003cp\u003eMDR and EB designed research and collected data. MDR and AM coded data. AM and JAC used photogrammetric software to transfer original data. MDR wrote the first version of the manuscript. All authors reviewed the manuscript.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003e \u003cb\u003eAdditional information\u003c/b\u003e \u003c/h2\u003e \u003cp\u003e \u003cstrong\u003eCompeting interests:\u003c/strong\u003e \u003cp\u003eThe author(s) declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis research was conducted in Botswana with funding from IDEA. Archaeological work at EAK (Olduvai Gorge, Tanzania) was funded by the Spanish Ministry of Science and Innovation (PID2023-146260NB-C2), and the Spanish Ministry of Culture through the program of Archaeology Abroad.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eMDR and EB designed research and collected data. MDR and AM coded data. AM and JAC used photogrammetric software to transfer original data. MDR wrote the first version of the manuscript. All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eWe thank the Commission for Science and Technology (COSTECH), the Ngorongoro Conservation Area Authorities (NCAA), the Division of Antiquities, and the Tanzanian Ministry of Natural Resources and Tourism for their permission to conduct research in Tanzania. AM is funded by a postdoctoral grant from the Fyssen Foundation. We are extremely thankful to Eduardo M\u0026eacute;ndez-Quintas and Elia Organista, for their help during fieldwork.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eDom\u0026iacute;nguez-Rodrigo, M. et al. Earliest Evidence of Elephant Butchery at Olduvai Gorge (Tanzania) Reveals the Evolutionary Impact of Early Human Megafaunal Exploitation. \u003cem\u003eeLife\u003c/em\u003e 14, RP108298 (2026).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAgam, A. \u0026amp; Barkai, R. Not the brain alone: The nutritional potential of elephant heads in Paleolithic sites. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e406\u003c/b\u003e, 218\u0026ndash;226 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eByers, D. A. \u0026amp; Ugan, A. Should we expect large game specialization in the late Pleistocene? An optimal foraging perspective on early Paleoindian prey choice. \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e32\u003c/b\u003e, 1624\u0026ndash;1640 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlumenschine, R. J. \u0026amp; Cavallo, J. A. Scavenging and Human Evolution. \u003cem\u003eSci. Am.\u003c/em\u003e \u003cb\u003e267\u003c/b\u003e, 90\u0026ndash;97 (1992).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarean, C. W. Sabertooth cats and their relevance for early hominid diet and evolution. \u003cem\u003eJ. Hum. Evol.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 559\u0026ndash;582 (1989).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSelvaggio, M. M. Concerning the three stage model of carcass processing at FLK\u003cem\u003eZinjanthropus\u003c/em\u003e: A reply to Capaldo. \u003cem\u003eJ. Hum. Evol.\u003c/em\u003e \u003cb\u003e35\u003c/b\u003e, 313\u0026ndash;315 (1998).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDom\u0026iacute;nguez-Rodrigo, M. \u0026amp; Baquedano, E. The Ecology of an Elephant Kill: Lion Defleshing of an African Elephant in the Okavango Delta (Botswana), and Limited Kleptoparasitic Opportunities for Meat-Eating Scavengers. \u003cem\u003eAfr. J. Ecol.\u003c/em\u003e \u003cb\u003e63\u003c/b\u003e, e70079 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhite, P. A. \u0026amp; Diedrich, C. G. Taphonomy story of a modern African elephant \u003cem\u003eLoxodonta africana\u003c/em\u003e carcass on a lakeshore in Zambia (Africa). \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e276\u0026ndash;277\u003c/b\u003e, 287\u0026ndash;296 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrganista, E. et al. Did \u003cem\u003eHomo erectus\u003c/em\u003e kill a \u003cem\u003ePelorovis\u003c/em\u003e herd at BK (Olduvai Gorge)? A taphonomic study of BK5. \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 601\u0026ndash;624 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVilla, P. \u0026amp; Torralba Aridos: elephant exploitation in Middle Pleistocene Spain. \u003cem\u003eJ. Hum. Evol.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, 299\u0026ndash;309 (1990).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVilla, P. et al. New data from Ambrona: closing the hunting versus scavenging debate. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e126\u0026ndash;128\u003c/b\u003e, 223\u0026ndash;250 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVilla, P. et al. Nuevos datos sobre Ambrona: cerrando el debate caza versus carro\u0026ntilde;eo. \u003cem\u003eZona Arqueol\u0026oacute;gica\u003c/em\u003e. \u003cb\u003e5\u003c/b\u003e, 352\u0026ndash;381 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAgam, A. \u0026amp; Barkai, R. Elephant and Mammoth Hunting during the Paleolithic: A Review of the Relevant Archaeological, Ethnographic and Ethno-Historical Records. \u003cem\u003eQuaternary\u003c/em\u003e \u003cb\u003e1\u003c/b\u003e, 3 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAltamura, F., Gaudzinski-Windheuser, S., Melis, R. T. \u0026amp; Mussi, M. Reassessing Hominin Skills at an Early Middle Pleistocene Hippo Butchery Site: Gombore II-2 (Melka Kunture, Upper Awash valley, Ethiopia). \u003cem\u003eJ. Paleo Arch.\u003c/em\u003e \u003cb\u003e3\u003c/b\u003e, 1\u0026ndash;32 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarkai, R. W. E. R. \u0026amp; Asia The Significance of Proboscideans in Diet, Culture and Cosmology in Paleolithic Asia. in Animals and Human Society in Asia: Historical, Cultural and Ethical Perspectives (eds (eds Kowner, R., Bar-Oz, G., Biran, M., Shahar, M. \u0026amp; Shelach-Lavi, G.) 33\u0026ndash;62 (Springer International Publishing, Cham, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-3-030-24363-0_2\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-24363-0_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBen-Dor, M. \u0026amp; Barkai, R. The importance of large prey animals during the Pleistocene and the implications of their extinction on the use of dietary ethnographic analogies. \u003cem\u003eJ. Anthropol. Archaeol.\u003c/em\u003e \u003cb\u003e59\u003c/b\u003e, 101192 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerthelet, A. \u0026amp; Chavaillon, J. The early Palaeolithic butchery site of Barogali (Republic of Djibouti). in The World of Elephants. Proceedings of the First International Congress 176\u0026ndash;179Consiglio Nazionale delle Ricerche, Rome, (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoschian, G. \u0026amp; Sacc\u0026agrave;, D. In the elephant, everything is good: Carcass use and re-use at Castel di Guido (Italy). \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e361\u003c/b\u003e, 288\u0026ndash;296 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChavaillon, J. et al. Le site de d\u0026eacute;pecage pl\u0026eacute;istoc\u0026egrave;ne \u0026agrave; \u003cem\u003eElephas recki\u003c/em\u003e de Barogali (R\u0026eacute;publique de Djibouti): nouveaux r\u0026eacute;sultats et datation. \u003cem\u003eComptes Rendus de l\u0026rsquo;Acad\u0026eacute;mie des. Sci.\u003c/em\u003e \u003cb\u003e305\u003c/b\u003e, 1259\u0026ndash;1266 (1987).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChavaillon, J. \u0026amp; Berthelet, A. The archaeological sites of Melka Kunture. in Studies on the Early Paleolithic site of Melka Kunture, Ethiopia (eds (eds Chavaillon, J. \u0026amp; Piperno, M.) 25\u0026ndash;80 (Instituto Italiano di Preistoria e Protostoria, Florence, (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDelagnes, A. et al. Interpreting pachyderm single carcass sites in the African Lower and Early Middle Pleistocene record: A multidisciplinary approach to the site of Nadung\u0026rsquo;a 4 (Kenya). \u003cem\u003eJ. Anthropol. Archaeol.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 448\u0026ndash;465 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEspigares, M. P. et al. Homo vs. \u003cem\u003ePachycrocuta\u003c/em\u003e: Earliest evidence of competition for an elephant carcass between scavengers at Fuente Nueva-3 (Orce, Spain). \u003cem\u003eQuaternary International\u003c/em\u003e 295, 113\u0026ndash;125 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaudzinski, S. et al. The use of Proboscidean remains in every-day Palaeolithic life. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e126\u0026ndash;128\u003c/b\u003e, 179\u0026ndash;194 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaudzinski-Windheuser, S., Kindler, L., MacDonald, K. \u0026amp; Roebroeks, W. Hunting and processing of straight-tusked elephants 125.000 years ago: Implications for Neanderthal behavior. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, eadd8186 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaudzinski-Windheuser, S., Kindler, L. \u0026amp; Roebroeks, W. Widespread evidence for elephant exploitation by Last Interglacial Neanderthals on the North European plain. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 120, e2309427120 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoren-Inbar, N., Lister, A., Werker, E. \u0026amp; Chech, M. A Butchered elephant skull and associated artifacts from the Acheulian site of Gesher Benot Ya\u0026rsquo;Aqov. \u003cem\u003eIsr. Pal\u0026eacute;orient\u003c/em\u003e. \u003cb\u003e20\u003c/b\u003e, 99\u0026ndash;112 (1994).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHannus, L. A. \u003cem\u003eClovis Mammoth Butchery: The Lange/Ferguson Site and Associated Bone Tool Technology\u003c/em\u003e (Texas A\u0026amp;M University, 2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G. Late Quaternary Proboscidean Sites in Africa and Eurasia with Possible or Probable Evidence for Hominin Involvement. \u003cem\u003eQuaternary\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, 18 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuman-Elephant Interactions. \u003cem\u003eFrom Past to Present\u003c/em\u003e (T\u0026uuml;bingen University, 2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeakey, M. D., Olduvai \u0026amp; Gorge \u003cem\u003eExcavations in Beds I \u0026amp; II 1960\u0026ndash;1963\u003c/em\u003e Vol. 3 (Cambridge University Press, 1971).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLemorini, C., Santucci, E., Caricola, I., Nucara, A. \u0026amp; Nunziante-Cesaro, S. Life Around the Elephant in Space and Time: an Integrated Approach to Study the Human-Elephant Interactions at the Late Lower Paleolithic Site of La Polledrara di Cecanibbio (Rome, Italy). \u003cem\u003eJ. Archaeol. Method Theory\u003c/em\u003e. \u003cb\u003e30\u003c/b\u003e, 1233\u0026ndash;1281 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLev, M. \u0026amp; Barkai, R. Elephants are people, people are elephants: Human\u0026ndash;proboscideans similarities as a case for cross cultural animal humanization in recent and Paleolithic times. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e406\u003c/b\u003e, 239\u0026ndash;245 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMielgo, C., Venter, S., Bosch, A., Huguet, R. \u0026amp; Venter, J. A. Feeding on giants: Neotaphonomic evidence of hyena exploitation of megafaunal skulls and its implications for Pleistocene archaeology. \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e186\u003c/b\u003e, 106458 (2026).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMosquera, M. et al. Barranc de la Boella (Catalonia, Spain): an Acheulean elephant butchering site from the European late Early Pleistocene. \u003cem\u003eJ. Quat. Sci.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, 651\u0026ndash;666 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMussi, M. \u0026amp; Villa, P. Single carcass of \u003cem\u003eMammuthus primigenius\u003c/em\u003e with lithic artifacts in the Upper Pleistocene of northern Italy. \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e35\u003c/b\u003e, 2606\u0026ndash;2613 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePineda, A. et al. Changing competition dynamics among predators at the late Early Pleistocene site Barranc de la Boella (Tarragona, Spain). \u003cem\u003ePalaeogeogr., Palaeoclimatol. Palaeoecol.\u003c/em\u003e \u003cb\u003e477\u003c/b\u003e, 10\u0026ndash;26 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReshef, H. \u0026amp; Barkai, R. A taste of an elephant: The probable role of elephant meat in Paleolithic diet preferences. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e379\u003c/b\u003e, 28\u0026ndash;34 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRocca, R., Boschin, F. \u0026amp; Aureli, D. Around an elephant carcass: Cimitero di Atella and Ficoncella in the behavioural variability during the Early Middle Pleistocene in italy. in Human-elephant interactions: from past to present (eds Konidaris, G. E., Barkai, R., Tourloukis, V. \u0026amp; Harvati, K.) (T\u0026uuml;bingen University Press, (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://dx.doi.org/10.15496/publikation-55591\u003c/span\u003e\u003cspan address=\"10.15496/publikation-55591\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSacc\u0026agrave;, D. Taphonomy of \u003cem\u003ePalaeloxodon antiquus\u003c/em\u003e at Castel di Guido (Rome, Italy): Proboscidean carcass exploitation in the Lower Palaeolithic. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e276\u0026ndash;277\u003c/b\u003e, 27\u0026ndash;41 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantonja G\u0026oacute;mez, M., L\u0026oacute;pez, N. \u0026amp; P\u0026eacute;rez-Gonz\u0026aacute;lez, A. Acheulean Occupation Sites in the Jarama Valley (Madrid, Spain). \u003cem\u003eCurr. Anthropol.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, 394\u0026ndash;395 (1978).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantonja, M., L\u0026oacute;pez Mart\u0026iacute;nez, N., Querol, M. \u0026Aacute;. \u0026amp; Soto, E. Estudio de las dispersiones de industria l\u0026iacute;tica y de los restos de vertebrados en \u0026Aacute;ridos-1 (Arganda, Madrid). in \u003cem\u003eOcupaciones achelenses en el valle del Jarama (Arganda, Madrid). Arqueolog\u0026iacute;a y Paleoecolog\u0026iacute;a\u003c/em\u003e (eds Santonja, M., L\u0026oacute;pez Mart\u0026iacute;nez, N. \u0026amp; P\u0026eacute;rez-Gonz\u0026aacute;lez, A.) 279\u0026ndash;295 (Diputaci\u0026oacute;n Provincial de Madrid, Madrid, 1980).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantucci, E. et al. Palaeoloxodon exploitation at the Middle Pleistocene site of La Polledrara di Cecanibbio (Rome, Italy). \u003cem\u003eQuaternary International\u003c/em\u003e 406, 169\u0026ndash;182 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSolodenko, N. et al. Fat Residue and Use-Wear Found on Acheulian Biface and Scraper Associated with Butchered Elephant Remains at the Site of Revadim, Israel. \u003cem\u003ePLOS ONE\u003c/em\u003e. \u003cb\u003e10\u003c/b\u003e, e0118572 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYravedra, J. et al. Not seen before. Unveiling depositional context and \u003cem\u003eMammuthus meridionalis\u003c/em\u003e exploitation at Fuente Nueva 3 (Orce, southern Iberia) through taphonomy and microstratigraphy. \u003cem\u003eQ. Sci. Rev.\u003c/em\u003e \u003cb\u003e329\u003c/b\u003e, 108561 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYravedra, J., Rubio-Jara, S., Panera, J. \u0026amp; Martos, J. A. Hominins and Proboscideans in the Lower and Middle Palaeolithic in the Central Iberian Peninsula. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e520\u003c/b\u003e, 140\u0026ndash;156 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYravedra, J. et al. Neanderthal and \u003cem\u003eMammuthus\u003c/em\u003e interactions at EDAR Culebro 1 (Madrid, Spain). \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e42\u003c/b\u003e, 500\u0026ndash;508 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYravedra, J., Rubio-Jara, S., Panera, J., Uribelarrea, D. \u0026amp; P\u0026eacute;rez-Gonz\u0026aacute;lez, A. Elephants and subsistence. Evidence of the human exploitation of extremely large mammal bones from the Middle Palaeolithic site of PRERESA (Madrid, Spain). \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e39\u003c/b\u003e, 1063\u0026ndash;1071 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYravedra, J. et al. Cut marks on the Middle Pleistocene elephant carcass of \u0026Aacute;ridos 2 (Madrid, Spain). \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e37\u003c/b\u003e, 2469\u0026ndash;2476 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G. \u003cem\u003eMammoths, mastodonts, and elephants. Biology, behavior, and the fossil record\u003c/em\u003e (Cambridge University Press, 1991).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G. Longitudinal studies of african elephant death and bone deposits. \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 131\u0026ndash;157 (1988).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G., Krasinski, K. \u0026amp; Wojtal, P. Elephant bone breakage and surface marks made by trampling elephants: Implications for interpretations of marked and broken \u003cem\u003eMammuthus\u003c/em\u003e spp. bones. \u003cem\u003eJ. Archaeol. Science: Rep.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e, 102491 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G. \u0026amp; Hutson, J. African elephant bones modified by carnivores: Implications for interpreting fossil proboscidean assemblages. \u003cem\u003eJ. Archaeol. Science: Rep.\u003c/em\u003e \u003cb\u003e34\u003c/b\u003e, 102596 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBehrensmeyer, A. K. Taphonomic and Ecologic Information from Bone Weathering. \u003cem\u003ePaleobiology\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e, 150\u0026ndash;162 (1978).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G. \u0026amp; Wojtal, P. Weathering Stages of Proboscidean Bones: Relevance for Zooarchaeological Analysis. J Archaeol Method Theory (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10816-022-09569-3\u003c/span\u003e\u003cspan address=\"10.1007/s10816-022-09569-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e doi:10.1007/s10816-022-09569-3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDom\u0026iacute;nguez-Rodrigo, M. et al. Fluvial spatial taphonomy: a new method for the study of post-depositional processes. \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, 1769\u0026ndash;1789 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantonja, M. \u0026amp; Querol, M. \u0026Aacute;. El sitio de ocupaci\u0026oacute;n achelense de \u0026Aacute;ridos-2. Descripci\u0026oacute;n general y estudio de la industria l\u0026iacute;tica. in \u003cem\u003eOcupaciones achelenses en el valle del Jarama (Arganda, Madrid). Arqueolog\u0026iacute;a y Paleoecolog\u0026iacute;a\u003c/em\u003e (eds Santonja, M., L\u0026oacute;pez Mart\u0026iacute;nez, N. \u0026amp; P\u0026eacute;rez-Gonz\u0026aacute;lez, A.) 297\u0026ndash;306 (Diputaci\u0026oacute;n Provincial de Madrid, Madrid, (1980).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePanagopoulou, E. et al. The Lower Palaeolithic site of Marathousa 1, Megalopolis, Greece: Overview of the evidence. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 33\u0026ndash;46 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynes, G. Las acumulaciones modernas de huesos de elefante como modelo para interpretar Ambrona y otras \u0026aacute;reas con fauna f\u0026oacute;sil a orillas del agua. \u003cem\u003eZona Arqueol\u0026oacute;gica\u003c/em\u003e. \u003cb\u003e5\u003c/b\u003e, 154\u0026ndash;175 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLa Cotte de St. \u003cem\u003eBrelade 1961\u0026ndash;1978. Excavations by C. B. M. McBurney\u003c/em\u003e (Geo Books, 1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRus, I. \u0026amp; Vega Toscano, G. El yacimiento de Arriaga II: problemas de una definici\u0026oacute;n actual de los suelos de ocupaci\u0026oacute;n. in \u003cem\u003ePrimeras Jornadas de Metodolog\u0026iacute;a e Investigaci\u0026oacute;n Prehist\u0026oacute;rica (Soria\u003c/em\u003e,) 387\u0026ndash;404 (Ministerio de Cultura, Soria, 1984).) 387\u0026ndash;404 (Ministerio de Cultura, Soria, 1984). (1981).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHay, R. L. \u003cem\u003eGeology of the Olduvai Gorge: A Study of Sedimentation in a Semiarid Basin\u003c/em\u003e (University of California Press, 1976).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeino, A. L. 40Ar/39Ar dating of Bed I, Olduvai Gorge, Tanzania, and the chronology of early Pleistocene climate change. \u003cem\u003eJ. Hum. Evol.\u003c/em\u003e \u003cb\u003e63\u003c/b\u003e, 251\u0026ndash;273 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantonja, M., L\u0026oacute;pez Mart\u0026iacute;nez, N. \u0026amp; P\u0026eacute;rez-Gonz\u0026aacute;lez, A. \u003cem\u003eOcupaciones achelenses en el valle del Jarama (Arganda-Madrid)\u003c/em\u003e (Diputaci\u0026oacute;n Provincial de Madrid, 1980).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eP\u0026eacute;rez-Gonz\u0026aacute;lez, A. El marco geogr\u0026aacute;fico, geol\u0026oacute;gico y geomorfol\u0026oacute;gico de los yacimientos de \u0026Aacute;ridos en la cuenca del Tajo. in \u003cem\u003eOcupaciones achelenses en el valle del Jarama (Arganda, Madrid). Arqueolog\u0026iacute;a y Paleoecolog\u0026iacute;a\u003c/em\u003e (eds Santonja, M., L\u0026oacute;pez Mart\u0026iacute;nez, N. \u0026amp; P\u0026eacute;rez-Gonz\u0026aacute;lez, A.) 15\u0026ndash;28 (Diputaci\u0026oacute;n Provincial de Madrid, Madrid, (1980).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eP\u0026eacute;rez-Gonz\u0026aacute;lez, A. \u0026amp; Uribelarrea, D. Geolog\u0026iacute;a del Cuaternario de los valle fluviales del Jarama y Manzanares en las proximidades de Madrid. \u003cem\u003eZona Arqueol\u0026oacute;gica\u003c/em\u003e. \u003cb\u003e1\u003c/b\u003e, 302\u0026ndash;317 (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoreno, D. et al. ESR dating of Middle Pleistocene archaeo-paleontological sites from the Manzanares and Jarama river valleys (Madrid basin, Spain). \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e520\u003c/b\u003e, 23\u0026ndash;38 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Juana, S., Gal\u0026aacute;n, A. B. \u0026amp; Dom\u0026iacute;nguez-Rodrigo, M. Taphonomic identification of cut marks made with lithic handaxes: an experimental study. \u003cem\u003eJ. Archaeol. Sci.\u003c/em\u003e \u003cb\u003e37\u003c/b\u003e, 1841\u0026ndash;1850 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOll\u0026eacute;, A. Universitat Rovira i Virgili, Tarragona,. Variabilitat i patrons funcionals en els sistemes t\u0026egrave;cnics de mode 2. An\u0026agrave;lisi de les deformacions d\u0026rsquo;\u0026uacute;s en els conjunts l\u0026iacute;tics del Riparo Esterno de Grotta Paglicci (Rigano Garganico, Foggia), Aridos (Arganda, Madrid) i Galeria-TN (Atapuerca, Burgos). \u003cem\u003eTDX (Tesis Doctorals en Xarxa)\u003c/em\u003e (2003).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlackwell, B. A. B. et al. ESR Dating Ungulate Teeth and Molluscs from the Paleolithic Site Marathousa 1, Megalopolis Basin, Greece. \u003cem\u003eQuaternary\u003c/em\u003e 1, (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobs, Z. et al. Optical dating of K-feldspar grains from Middle Pleistocene lacustrine sediment at Marathousa 1 (Greece). \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 170\u0026ndash;177 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTourloukis, V. et al. Magnetostratigraphic and chronostratigraphic constraints on the Marathousa 1 Lower Palaeolithic site and the Middle Pleistocene deposits of the Megalopolis basin, Greece. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 154\u0026ndash;169 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarkanas, P. et al. Sedimentology and micromorphology of the Lower Palaeolithic lakeshore site Marathousa 1, Megalopolis basin, Greece. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 123\u0026ndash;136 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTourloukis, V. et al. Lithic artifacts and bone tools from the Lower Palaeolithic site Marathousa 1, Megalopolis, Greece: Preliminary results. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 47\u0026ndash;64 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKonidaris, G. E. et al. The skeleton of a straight-tusked elephant (\u003cem\u003ePalaeoloxodon antiquus\u003c/em\u003e) and other large mammals from the Middle Pleistocene butchering locality Marathousa 1 (Megalopolis Basin, Greece): preliminary results. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 65\u0026ndash;84 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMichailidis, D., Konidaris, G. E., Athanassiou, A., Panagopoulou, E. \u0026amp; Harvati, K. The ornithological remains from Marathousa 1 (Middle Pleistocene; Megalopolis Basin, Greece). \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 85\u0026ndash;94 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGiusti, D. et al. Beyond maps: Patterns of formation processes at the Middle Pleistocene open-air site of Marathousa 1, Megalopolis basin, Greece. \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 137\u0026ndash;153 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR Core Team. \u003cem\u003eR: A language and environment for statistical computing\u003c/em\u003e (R Foundation for Statistical Computing, 2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaddeley, A., Rubak, E. \u0026amp; Turner, R. The Impact of R on Statistical Science for Spatial Point Processes. \u003cem\u003eAustralian New. Z. J. Stat.\u003c/em\u003e \u003cb\u003e68\u003c/b\u003e, e70037 (2026).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaddeley, A., Rubak, E. \u0026amp; Turner, R. \u003cem\u003eSpatial Point Patterns: Methodology and Applications with R\u003c/em\u003e (CRC, 2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaddeley, A. \u0026amp; Turner, R. spatstat: An R Package for Analyzing Spatial Point Patterns. \u003cem\u003eJ. Stat. Softw.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 1\u0026ndash;42 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiegand, T. \u0026amp; Moloney, K. A. \u003cem\u003eHandbook of Spatial Point-Pattern Analysis in Ecology\u003c/em\u003e (Chapman and Hall/CRC, 2013). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1201/b16195\u003c/span\u003e\u003cspan address=\"10.1201/b16195\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArteaga Brieba, A. Universitat Rovira i Virgili,. Among stones and bison. The lithic assemblage of Gran Dolina TD10.2 (Atapuerca). Technological and spatial implications of a specialised Middle Pleistocene Kill-butchering site. TDX (Tesis Doctorals en Xarxa) (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCobo-S\u0026aacute;nchez, L., Dom\u0026iacute;nguez-Rodrigo, M., Mabulla, A. \u0026amp; Baquedano, E. Chapter 5 - Reconstructing early human behavior through the in-site spatial statistical analysis of DS. in Reconstructing Olduvai (ed (eds Dom\u0026iacute;nguez-Rodrigo, M. et al.) 197\u0026ndash;240 (Academic, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/B978-0-443-27382-7.00007-1\u003c/span\u003e\u003cspan address=\"10.1016/B978-0-443-27382-7.00007-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiez-Mart\u0026iacute;n, F. et al. Tracing the spatial imprint of Oldowan technological behaviors: A view from DS (Bed I, Olduvai Gorge, Tanzania). \u003cem\u003ePLOS ONE\u003c/em\u003e. \u003cb\u003e16\u003c/b\u003e, e0254603 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDom\u0026iacute;nguez-Rodrigo, M. et al. Spatial simulation and modelling of the early Pleistocene site of DS (Bed I, Olduvai Gorge, Tanzania): a powerful tool for predicting potential archaeological information from unexcavated areas. \u003cem\u003eBoreas\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e, 805\u0026ndash;815 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuz\u0026oacute;n, C. et al. Taphonomic and spatial analyses from the Early Pleistocene site of Venta Micena 4 (Orce, Guadix-Baza Basin, southern Spain). \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 13977 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMar\u0026iacute;n, J. et al. Neanderthal logistic mobility during MIS3: Zooarchaeological perspective of Abric Roman\u0026iacute; level P (Spain). \u003cem\u003eQ. Sci. Rev.\u003c/em\u003e \u003cb\u003e225\u003c/b\u003e, 106033 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMerino-Pelaz, A., Cobo-S\u0026aacute;nchez, L., Organista, E. \u0026amp; Baquedano, E. Dom\u0026iacute;nguez-Rodrigo, M. Unraveling the spatial imprint of hominin and carnivore accumulations in Early Pleistocene African sites. \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 128 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMielgo, C. et al. Intra-site spatial approaches based on taphonomic analyses to characterize assemblage formation at Pleistocene sites: a case study from Buena Pinta Cave (Pinilla del Valle, Madrid, Spain). \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 1\u0026ndash;28 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMocl\u0026aacute;n, A. et al. Identifying activity areas in a neanderthal hunting camp (the Navalma\u0026iacute;llo Rock Shelter, Spain) via spatial analysis. \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 44 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMocl\u0026aacute;n, A. et al. Spatial analysis of an Early Middle Palaeolithic kill/butchering site: the case of the Cuesta de la Bajada (Teruel, Spain). \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 91 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePanera, J. et al. Assessing functionality during the early Acheulean in level TKSF at Thiongo Korongo site (Olduvai Gorge, Tanzania). \u003cem\u003eQuatern. Int.\u003c/em\u003e \u003cb\u003e526\u003c/b\u003e, 77\u0026ndash;98 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaladi\u0026eacute;, P. et al. Dragged, lagged, or undisturbed: reassessing the autochthony of the hominin-bearing assemblages at Gran Dolina (Atapuerca, Spain). \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 65 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVillaescusa, L. et al. Towards a formation model of the Neanderthal symbolic accumulation of herbivore crania: Spatial patterns shaped by rockfall dynamics in Level 3 of Des-Cubierta Cave (Lozoya valley, Madrid, Spain). \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 16 (2026).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Spatial statistics, carnivore modification, elephant carcass assemblages, taphonomic processes, Inhomogeneous spatial statistics","lastPublishedDoi":"10.21203/rs.3.rs-8799314/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8799314/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study integrates spatial point-process statistics and multivariate ordination to evaluate carcass modification and carnivore impact in modern elephant assemblages and to contextualize three archaeological sites (\u0026Aacute;ridos-2, Marathousa 1 and EAK). Using a suite of inhomogeneous spatial metrics and Minkowski functionals, modern elephant carcasses reveal a structured gradient from highly clustered, carnivore-impacted spatial patterns to more diffuse configurations associated with higher skeletal survival and various degrees of carnivore post-depositional disturbance. Principal component analysis demonstrates that spatial clustering, short inter-point distances, and fragmented carcass footprints covary with increased tooth-mark frequencies and reduced survival, whereas dispersed point patterns correlate with greater skeletal preservation. When archaeological assemblages are projected into this multivariate space, \u0026Aacute;ridos-2 and EAK (with caution, given its complex sedimentary nature) align closely with modern clustered carcass patterns indicative of either intense carnivore intervention with limited cluster disturbance, or early depositional stages (preserving a high degree of carcass articulation) or both, while Marathousa 1 occupies an intermediate position consistent with more limited carnivore access and greater structural integrity of carcass remains after disarticulation. Together, these results show that spatial point-pattern structure captures taphonomic processes at the carcass scale and provides a robust framework for inferring carnivore impacts in deep-time proboscidean assemblages.\u003c/p\u003e","manuscriptTitle":"Spatial signatures of carnivore-driven bone dispersal: A taphonomic analysis of modern proboscidean carcasses, and implications for Pleistocene elephant butchery sites","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-20 09:49:49","doi":"10.21203/rs.3.rs-8799314/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"dcd3f45f-21e7-4304-a661-80f1fc1ea417","owner":[],"postedDate":"February 20th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63103980,"name":"Biological sciences/Ecology"},{"id":63103981,"name":"Earth and environmental sciences/Ecology"},{"id":63103982,"name":"Biological sciences/Evolution"}],"tags":[],"updatedAt":"2026-03-17T10:27:32+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-20 09:49:49","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8799314","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8799314","identity":"rs-8799314","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}