1. Introduction
Understanding the different processes that lead to the formation of fossiliferous assemblages can be problematic as the observed taphonomic signatures are often ambiguous. This since the distinction between damage ensuing from taphonomic agents is frequently subtle and subject to equifinality, i.e., that different taphonomic agents can lead to a common end result [
1]. Experiments under controlled conditions or long-term naturalistic monitoring (e.g., [
2,
3,
4]) have been highlighted as the best ways to distinguish between agents since they enable us to track the more refined signatures of modification in fossil assemblages. Such experiments allow us to gain qualitative insights as to how different processes might occur or influence how fossils accumulate during site formation.
Trampling can be defined as agents treading heavily so as to beat down or crush an object. The effect of trampling would probably not have invited special interest in taphonomic studies, but three decades ago researchers reported that trampling marks and cut marks mimicked each other (e.g., [
5,
6]). The main surface modifications produced by trampling on bone are striations and microstriations [
7,
8,
9,
10], shiny and polished surfaces [
11,
12] and even notches on oblique fracture angles [
13]. In order to obtain key traits to distinguish trampling damage from cut marks, and so facilitate assessment of human involvement in a given assemblage, bones of large vertebrates have been extensively studied [
14,
15]. However, even after these three decades, the importance of distinguishing cut marks from trampling marks is still debated and identification continues to face difficulties even by experts [
14]. Machine learning has recently been proposed as a potentially successful procedure to resolve this issue [
16]. Bone breakage can be an additional modification derived from trampling and in order to distinguish it from the effects of other types of compression, researchers have characterised bone breakage (e.g., [
17]). They have analysed the resistance of bones from different vertebrate species to breakage [
18,
19,
20], or characterised breakage from trampling versus breakage derived from other taphonomic agents [
12,
21,
22,
23,
24].
Compression testing is one of the simplest methods available that have been extensively used to characterise the mechanical properties of different materials. It has been used in ecology and taphonomy to study the response of bite force of carnivores such as hyaenas [
25,
26,
27], to distinguish trampling marks obtained when bones are wet and compressed against gravels [
19], to evaluate bone resistance of large mammals to sudden or slow movements when bones are wet or dry [
28] and to characterise traits of compression in vertebrae of fish to distinguish them from abrasion and digestion [
24,
29]. Apart from digestion and skeletal element representation, breakage can be used to identify the predator that produced small mammal assemblages [
11]. The effects of trampling or compression on such assemblages can thus be linked to the predator that previously modified these bones during ingestion. In such cases, broken edges may be rounded by gastric juice corrosion [
11]. Thus, post-depositional processes such as trampling may be superimposed on the taphonomic pattern which characterises each predator, causing distortive effects that can be identified through experimental work.
Our experiment started with modern barn owl (
Tyto alba) pellets, a raptor known to cause minimal modifications by digestive effects and breakage by ingestion [
11]. In the case of small mammals, the fragility of their skeletal elements may give the a priori impression that they will not resist compression. This has, however, never been tested aside from observations made on two nest sites of barn owl and European eagle owl (
Bubo bubo), and a quick trampling experiment that was described by Andrews (p.10, [
11]). In the later experiment, the author stepped on owl pellets and concluded that “the pattern that emerges from these results on trampling is one of breakage of skulls, reduction in numbers of maxillae, considerable loss of teeth from the jaws leading to large numbers of isolated teeth, considerable breakage of larger postcranial elements and some degree of loss, but no loss or breakage of smaller elements (calcanei, talli))”. In this study we repeated these experiments under high controlled conditions to explore the patterns of breakage caused to small mammal bones under compression. Results of the compression experiments were then compared to small mammal taphonomic analyses of fossil samples from the Earlier Stone Age levels in Wonderwerk Cave in South Africa [
30] that showed extensive breakage exceeding excavation procedures or digestion by the predator.
2. Material and Methods
The experiment presented here tried to reproduce the results obtained by Andrews (1990) on pellets and on isolated small mammal skeletal elements using compression under controlled conditions. Uniaxial compression experiment was conducted using a Zwick/Roell Z5.0TN machine with testXpert II software, held at the Natural Sciences Museum of Madrid (MNCN-CSIC) at the Laboratory of Environmental Analyses and Experimental Taphonomy (
https://www.mncn.csic.es/en/investigacion/servicios-cientifico-tecnicos/laboratory-environmental-analyses (accessed on 27 January 2022)). This uniaxial compression machine has a test load standard of 5 kN (~500 Kg), although for the present experimental study we applied a special load cell to restrict the compression forces to 500 N (~51 Kg), this being an intermediate force between previous experimental and observational samples described by Andrews (1990). This restriction was needed to limit any extra-force exerted by mistake when programming the experiment that could completely destroy these specimens. The equipment to compress the small mammal bones used the same force in the identical direction to prevent the influence of other external parameters which could influence the results. Results of damage were obtained from the compression experiments in the form of curves through the testXpert
® II software of the equipment (
Figure 1).
Small mammal materials used in the experiment derive from a modern collection of pellets of barn owls kept in captivity fed upon laboratory mice (small sized: House mouse,
Mus musculus; large sized: Brown rat,
Rattus norvegicus; see
Table 1). Barn owl pellets are characterised by a high relative abundance of skeletal elements and low bone and tooth breakage together with a low proportion and degree of digestion on diagnostic elements (e.g., [
11,
31]. Despite the potential integrity of elements and abundance of skeletal remains, initial compressions performed on four pellets (to test the experiment and observations described by Andrews [
11]) resulted in two of them providing only tibiae, ulnae, vertebrae and ribs, without cranial elements, while several of the skeletal elements had been affected by digestion. Eleven pellets were opened, and only five of them provided skeletal elements from cranial (jaws and skulls) and postcranial (long, square and flat bones) elements in sufficient quantity to systematically repeat the experiment several times.
Each selected bone to be compressed was photographed and examined under a binocular light microscope (Leica MZ 7.5). Digested and incomplete bones were discarded to avoid fractures and weaker areas caused by digestion process. Half of selected pellets and skeletal elements were kept dry, whilst the rest were immersed in water for an arbitrary time period of fifteen days in order to examine compression under different environmental conditions. Immersion in water was considered because it could provide seasonal indications due to damp substrates. A total of four complete pellets (two dry and two wet, compressed individually) and 60 isolated skeletal elements (compressed in different sets from 5 opened owl pellets) were used in the experiment. The 60 skeletal elements were: 8 skulls, 8 mandibles, 12 femora, 8 humeri, 8 pelves, 9 astragali and 7 calcanei. The specimens selected were adult and young individuals (<1 kg, both large and small sizes) although not always a large and a small size of the same skeletal element was obtained. Two types of sediment, coarse sands (0.7 mm) and silts (0.065 mm), were chosen for testing. Each set of bones is shown in
Table 1.
Pellets, both wet and dry, were pressed directly on the metal surface of the compression plates, in contrast to the isolated anatomical elements that were placed on a sediment tray.
Each set of bones was compressed in three steps as follows:
Step 1: Compression on sand
Anatomical elements selected were placed on a substratum of coarse-medium sand (0.7 mm) to show the protection this sediment type provides to small skeletal elements.
Step 2: Compression on silts
Anatomical elements were placed on silts (0.065 mm) to observe compression on a fine silty sediment typical of caves, as in Wonderwerk Cave which has served as our fossil case study (see below).
Step 3: Compression on silts without skulls
While in nature skulls are associated with other skeletal elements when trampled, the height of the skulls prevented the compression of flatter bones during the experiment. Therefore, the skulls were removed in order to ensure direct compression of the other bones and the whole SET compressed again.
Each step was carried out on four sets of dry skeletal elements and four sets of wet samples of specimens placed in different positions: Dorsal or ventral, lingual or labial, anterior or posterior, plantar or medial (see
Table 1 and
Figure 2).
For long bones, the resulting breakage was classified following the categories identified by Andrews [
11]: Complete, proximal end, shaft and distal end. Classification of long bones was extended to facilitate the identification of breakage types using the categories described by Villa and Mahieu [
17] with documentation of incipient fissures or cracks. The breakage categories proposed by Villa and Mahieu [
17] were developed for long bones of large mammals, so this nomenclature has been used here without any typological meaning as it is applied to flat and small bones, such as pelves and astragali of microfauna. We have consequently named fracture angles as: Right (R), oblique (O), mixed (M); fracture outline as transverse (T), curved (C), intermediate (I) and fracture edge as smooth (S) or jagged (J). Completeness of the shaft circumference was also evaluated considering the classification proposed by Villa and Mahieu [
17]. Circumference completeness includes three categories: Bone circumference less than half of the original ①; circumference more than half of the original ② and complete circumference in at least a portion of the bone length ③ (
Figure 3).
All specimens were photographed before and at the end of each step of the experiment using an automated stereomicroscope (Leica M 205A) provided with a digital camera (Leica DFC450) and image software LAS V4.4. This procedure and the control of individual bones allowed us to discard the effects of sifting and sorting of the resulting trampled assemblage or pellet, that potentially could add to detachment of bone fragments or increase breakage.
A Fisher’s Exact Test of Independence for numeric data in small samples was applied using R [
32] and the library “rcompanion” [
33]. This test uses a contingency table and runs an exact procedure especially for small-sized samples [
34]. Fisher’s Exact Test was considered to be more suitable than other tests, such chi-squared which is better for large samples. Therefore, the “simulate.
p.value” option was used, which automatically applied a Monte Carlo simulation of the
p-values based on 2000 replicates. The Monte Carlo approach randomly generates tables to satisfy the null hypothesis for the test and evaluate the test statistic on those tables. A significant
p-value is considered as
p ≤ 0.05.
Case Study of Wonderwerk Cave
Wonderwerk Cave is a large phreatic cavity (140 m long, 11–26 m wide and 3 to 7 m high) located in the eastern flank of the Kuruman Hills, near the town of Kuruman in central South Africa, (27°50 S, 23°33 E,
Figure 4). Archaeological investigations at the site started in the 1940s [
35,
36], but the cave was most extensively excavated by P.B. Beaumont, A.I. Thackeray and J.F. Thackeray in the 1980s [
37,
38,
39]. Since 2004, it is being investigated by M. Chazan, L.K. Horwitz and their research team [
40,
41]. The sediment is mainly silty, sometimes sandy with some dolomite blocks that fell from the roof that are dispersed in the deposit and on the present-day cave surface.
The cave is an exceptional site yielding signs of hominin occupation spanning ca. 2 million years, from the Oldowan through to historic times [
41,
42,
43]. Small mammal remains are extremely abundant in most of the layers, and in many they formed a background deposit that covered the paleo-cave floor. The physical size of the microfaunal taxa identified at this site can be considered as equivalent to the specimens used in this experiment (for taxon lists see Avery [
44,
45]). Taphonomic studies of these microfaunal remains have been undertaken on Oldowan and Earlier Stone Age (ESA) strata derived from Excavation 1, which is located ~30 metres from the cave entrance [
30,
44,
45,
46,
47]. These publications have demonstrated the value of research on small mammal fossils at Wonderwork in identifying the predator responsible for their introduction into the cave, and for palaeoenvironmental and palaeoecological information. They have also served as sources of information on past human behaviour.
The small mammal assemblages from the oldest cave strata (St. 12 to St. 10; Oldowan through early ESA) come from the old excavations undertaken by P. B. Beaumont and were recovered by dry sieving using 1 mm mesh. Edentulous jaws and large numbers of loose teeth were the first features that caught the attention to Avery [
44] who studied the taxonomy of the P. B. Beaumont collections. She concluded that the extreme rates of breakage may have been influenced by aggressive recovery and preparation procedures. Subsequently, broken edges of these fossils were evaluated by Fernández-Jalvo and Avery [
46], and only 20–30% could be attributed to recent breakage. In order to contrast the incidence of recovery procedures, these old samples were compared to new samples from overlying ESA layers, Strata 6 and 7, recovered during the 2018 excavation undertaken by the Chazan and Horwitz team (
Table 2). During these excavations, all sediment was collected and processed using a mechanized floatation machine. The recent sampling from Strata 6/7, whose data are reported here (
Table 2), combined due to the absence of any significant differences between them, has substantially reduced breakage and loss of small mammal skeletal remains [
30], but the main traits observed by Avery [
44] are still present in the sample despite processing by floatation.
4. Discussion
Frequently, the study of taphonomy is considered to be the study of bias. Probably the reason for such an assumption is due to the inherent comparison with what should be in a site if the faunas and floras represented were complete. The loss of biological information in fossil assemblages is self-evident when observing dead bodies during processes of putrefaction. The lack of muscles and ligaments facilitates the disarticulation of the skeleton, providing important information related to the amount of time passed and weather conditions. A further source of invaluable information is the study of bone breakage and deformation. The “incompleteness” and “imperfection” of fossils, when compared to the original bones, is a source of taphonomic information that is codified in the fossils and needs to be “extracted” in order to identify the past history of assemblages and learn about site formation processes. At all sites, each fossil found can provide information on contemporaneous biotic and abiotic agents, acting either simultaneously or asynchronously on the bone remains, causing the loss of part of their original palaeobiological information and increasing the taphonomic information to be obtained [
48].
The current study was initiated as result of uncharacteristically heavy breakage observed in the small mammal remains from Wonderwerk Cave. Most jaws were recurrently edentulous with a high frequency of tooth loss (the latter could easily pass through the mesh of the screens used in the original excavations), and quite extensive modern damage that was incurred due to sifting, resulting in remains that could not be taxonomically identified [
44]. Despite the extensive tooth loss [
44], other small-sized skeletal elements (calcanei and astragali) appeared in relative high proportions (
Figure 12), suggesting that size was not the sole factor determining representation. We tested the possible involvement of highly destructive predators that could cause intense breakage to small mammals, but the predator identified at Wonderwerk Cave is
Tyto alba (which causes minor destruction according to Andrews, [
11]). Thus, this raptor could not explain the extensive destruction evidenced in these small mammal assemblages. Post-depositional abiotic agents (flowing water or falling rocks) could increase breakage in a raptor prey assemblage. However, cave roof spall does not occur in all of the cave strata that were examined here. Moreover, there is no evidence from the sedimentary record of the cave [
49] that either of these factors were of any consistent intensity.
The new excavations in the Wonderwerk Cave site provided samples from St. 6/7 that exhibit greater completeness than samples from the old excavations (
Table 2), but the frequency of edentulous jaws is still high. The proportion of isolated teeth (detached from jaws) and relative abundance of small-sized elements (astragali and calcanei) are even higher in the new floatation sample. In general, the traits of breakage and destruction observed in the Wonderwerk Cave samples are still higher than is expected in a barn owl assemblage.
Marin-Monfort et al. [
30] observed that the destructive traits that characterised the microfaunal samples from Wonderwerk Cave were constant between layers and were associated with the persistent occupation of barn owls in the site during the ESA strata. In fact, barn owls still inhabit the cave today. Thus, the breakage patterns to the small mammal bones were incongruous with the signatures of both biotic and abiotic taphonomic agents that had been identified. We took into account the possibility of an additional factor that had not been considered before, namely that the assemblage may had been subjected to trampling, an overprint on the barn owl signature. To do that, we performed the systematic experiment described in detail in this paper examining breakage in different skeletal elements. Half of these bones were immersed in water and show a more plastic response, in contrast to dry bones that are brittle and so break more easily under compression [
19,
20,
24].
We have confirmed the initial breakage pattern for small mammals described by Andrews [
11] but added more detail to the model which can now be defined in three stages:
- (A)
Breakage of skulls, reduction in numbers of maxillae. In our experiment, we have proved the pneumatic behaviour of the skull which deformed leaving a mass of bones and teeth. The bones of the skull detached into small unidentifiable bone fragments, especially when subjected to water or gravitational movements such as sifting, leaving the zygomatic arches apart from the maxillae which frequently bear the M1 in situ. Incisors were also frequently found in the interior of the alveoli, and both traits have been observed in the Wonderwerk Cave fossil assemblage. Consequently, both observations can be added as taphonomic criteria of trampling in a fossil site.
- (B)
Considerable loss of teeth from the jaws leading to large numbers of isolated teeth. As seen in Wonderwerk Cave and in the experiment, both mandibles and maxillae showed detachment of molars although in mandibles the M1 remained in situ retained in the diastema portion. The incisors, as observed in maxillae, are also retained in situ (in the alveolar socket of the mandible) and in this type of assemblage it is also common to see the ascending ramus detached from the dental row and the rest of the mandible.
- (C)
Considerable breakage of larger postcranial elements and some degree of loss, but no loss or breakage of smaller elements (calcanei, talli). High frequency of calcanei and astragali is one of the most representative features of microfaunal assemblages of Wonderwerk Cave site, as well as modern assemblages exposed to trampling. In addition, a high frequency of complete bones was observed when compression was undertaken on wet bones together with jagged edges, fissures and cracks, which were frequently transversal to the length of the bone.
We have obtained clear criteria with which to distinguish trampling on small mammals, by subjecting skeletal elements and pellets to compression. However, in this study our compression of pellets (
Supplementary Information) may have been too mild such that the results were less destructive than found by Andrews [
11]. Results of compression directly on bones, however, are highly similar. Both experiments respond to the mechanical properties of bone which behaves similar to engineering materials (e.g., ceramic, metals, rocks, building materials) and, therefore, reacts to loading and fracture following the basic principles of mechanics (e.g., [
50,
51,
52,
53,
54,
55]).
In general, the relevance of detecting trampling in fossil small mammal assemblages facilitates interpretation of the full history of raptor assemblages since, as at Wonderwerk, the barn owl’s ingestion and digestion does not produce high bone breakage rates. Identification of trampling may then allow us to better understand how the assemblage formed. Thus, the reconstructed scenario for Wonderwerk Cave is that small mammal bones were probably trampled by the owls themselves while they nested in the cave’s interior, although bones still inside the pellets could not be substantially destroyed. The owl pellets could also have been trampled by other cave visitors, including humans, thereby increasing breakage but not totally destroying the small mammal bones unless they were directly exposed to trampling (i.e., not inside a pellet). Bones from disaggregated pellets will be more exposed to post-depositional processes such as weathering and trampling than those bones in the interior of a pellet [
11,
56]. This scenario was recently documented in the archaeological rock shelter Álvarez 4 (late Holocene) located in the arid and cold environment of northwestern Patagonia of Argentina, where the bones from the pellets were less fractured than bone assemblages isolated in the sedimentary matrix [
57].
This paper validates breakage traits obtained from experiments in a compression material testing machine against those from more actualistic experimentsof trampling by humans (e.g., [
11]). There are still many experiments that need to be carried out to test other situations of trampling (e.g., pellets of other raptors and excrements of mammalian carnivores, larger and smaller prey, compression by forces around 50 N to simulate breakage by trampling in nests or latrines by predators) and so complement these patterns. There are also modern bones that were collected by one of us (PA) from monitored natural nests that potentially could have been trampled by predators. These were briefly described in Andrews [
11] page 8 but as they are of interest, it is planned to study them in greater detail. In addition, it would be useful to increase the size of the experimental sample to better characterise fissures and cracks in postcranial bones and also to process these samples using sieving and floatation methods. Complementary research could extend to monitoring nesting behaviour and studying remains derived from old raptor nests.