THE STATISTICS OF ARCHAEOLOGICAL DE FORMATION PROCESS ES . A n archae o zoological experiment

1. THE STATISTICS OF ARCHAEOLOGICAL DEFORMATION PROCESSES.An archaeozoological experiment Laura MAMELI, Jordi ESTEVEZ, Juan A.BARCELO Universidad Autonoma Barcelona

2. Archaeologists traditionally have drawn their inferences about past behaviour from dense, spatially discrete aggregations of artefacts, bones, features, debris,assumingthat the main agent responsible for creating such aggregates was only human behaviour. ACCUMULATION AS AN ARCHAEOLOGICAL FORMATION PROCESS

3. Archaeological assemblages are still usually viewed as a deposit or an aggregate of items, which are part of single depositional events, as if the materials of social action through time were only accumulative. ACCUMULATION AS AN ARCHAEOLOGICAL FORMATION PROCESS

4. In the case of faunal processing:this approach assumes that the remains of the animals under consideration have survived since deposition to a more-or-less similar extent, and that the relative abundance of species remains is representative of that originally deposited. THE FORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

5. However, the accumulation of animal remains in the archaeological record is not always the result of purposeful human activity. An aggregation of bones may not reflect past human social action, but rather post depositional processes THE FORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

6. Usually, archaeologists assume that prehistoric social activities, including procurement, butchering, storage, cooking and refuse will produce faunal assemblages different from those generated by natural processes. THE FORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

7. Most post-depositional processes make archaeozoological assemblages more amorphous, lower in elements density, more homogeneous in their internal density, less distinct in their boundaries, and more similar (or at least skewed) in composition. THE FORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

8. A specific statistical pattern may characterize bone assemblages which have been post-depositionally modified:for any given species the frequencies of different skeletal elements show at least some significant biases from the frequencies in which they would be represented in complete skeletons. THE DEFORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

9. Modifications in original skeletal frequencies may appear:as a result of preferential human transport as a result of subtraction by animal scavenging as a result of differential preservation THE DEFORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

10. We are interested in calculating a statistical modelof “deformations” experimented by bone assemblages as a result of scavenging. THE DEFORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

11. Scavenging should be considered as a sequence of modifications that convert an animal carcass into a disintegrated set of bones. After scavenging, the remaining evidence contains just some distorted elements (palimpsest) of the original animal carcass. THE DEFORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

12. We think that the sum of quantitative modifications experimented by a carcass (in content and spatial distribution) produces a significative qualitative change (a bone assemblage). THE DEFORMATION OF ARCHAEOZOOLOGICAL ASSEMBLAGES

13. In this paper we present a case study where the natural formation process of bone assemblages is experimented. Contrary to the usual view, when wild animals (scavengers) are the causal agent of the assemblage, the archaeologically observable consequence is not an accumulation of bones, but a considerable dispersal of them. Tierra del Fuego. An archaeozoological case study

14. We have studied 30 carcasses of “guanaco” (Lama guanicoe) scavenged by foxes in Tierra del Fuego (Argentina). During three years we have taken detailed measurements from animal carcasses produced by a catastrophic natural death in 1995. Tierra del Fuego. An archaeozoological case study

21. We have designed a series of controlled observations in order to be able to calculate the probability of the relationship between the disturbance effect and the composition and spatial pattern of bone remains. Preliminary Hypotheses

22. As a preliminary stage we have produced some qualitative statistics of the general pattern of scavenging, as observed in our data. Preliminary Hypotheses

24. A simple plotting of percentages of scavenged body parts shows that during the first year of observation, softer parts (anus, belly) where scavenged until its total consumption, and the head (cranium, mandibulae, mouth) of the animal is much more ravaged than the rest of the carcass. Qualitative Statistics

25. The less scavenged parts during the first year are limb bones, thorax, neck, lumbar and sternum. Qualitative Statistics

26. In the second year, those neglected parts are scavenged in greater proportion, probably because the skin of the animal preserved edible tissues in those parts. During the third year, many elements of the carcasses are hardly visible, and nearly all the parts remaining are equally scavenged. Qualitative Statistics

27. To go beyond this frequency description we need multivariate techniques to disentangle the multiplicity of effects produced by different post-depositional processes. We have studied through Correspondence Analysis the presence/absence of observed evidences for scavenging. Qualitative Statistics

28. The upper part of the graph shows those carcasses whose belly, mouth or cranium have been ravaged by scavenging (red circles). Blue triangles correspond to carcasses whose belly, mouth or cranium have never observed to be scavenged by red foxes. The upper part of the left graph shows those carcasses whose belly, mouth or head have been ravaged by scavenging (red circles). Blue triangles correspond to carcasses whose belly, mouth or head have never observed to be scavenged by foxes. In the second figure, it is interesting to observe that differential scavenging of forelimb or feet explains also an important part of global variance

29. It is also interesting to remark the strong statistical difference between the contributions of the different variables Qualitative Statistics

30. belly, mouth, cranium and mandible, against limb bones and feet. Bones from the low neck, lumbar and pelvic region are somewhere between the two groups but with more similarities with limb bones. High Neck, Thorax and sternum join the limb bones group.

31. In most taphonomic analysis of scavenged carcasses, temporal dynamics are only partially taken into account. We think that it is impossible to understand scavenging as a disturbance process, if we do not characterize it as a dialectical process, where a non-linear sum of quantitative changes, beyond a threshold, produces a qualitative transformation. Temporal Dynamics

32. There are important differences between the first year pattern and successive years. But it does not seem possible to separate between the second and the third year.

33. 1996 The main qualitative difference in the first year is the intensity of scavenging. Individuals represented as a red circle and a number 1 are those carcasses with evidences of scavenging in softer tissues (mouth, belly, head). Around the periphery of the main area we found some carcasses represented as a triangle (belly not scavenged), as a blue circle (mouth not scavenged) or with a number 2 (head not scavenged)

34. 1997 There is a strong difference in the scavenged pattern of the head and the neck, against other skeletal parts. Scavenging has been reduced, because the softer parts of the carcass have been already dissapeared. Now the main significative difference is the group of carcasses where the head and the neck (torax, cranium, mandibulae, sternum, high neck, low neck) have not been scavenged.

35. 1998 During the third year, when edible tissues are less frequent, a series of carcasses remain unaltered (blue triangles).

36. It is easy to notice that scavenging has a strong temporal component towards increasing entropy.After three years of scavenging, carcasses loose their integrity.The main significant factor is the intensity of deformation rather than the kind of deformation. Temporal Dynamics

37. We have not identified any regular features to characterize scavenging processes, but a general pattern towards increasing qualitative deformation. This result is in strong opposition with usual approaches trying to identify natural disturbance processes using single identifiers. Temporal Dynamics

38. In our case study, we have discovered evidence for a relationship between the intensity of scavenging, the distance of transported elements and the relative frequency of lost material:although the intensity of scavenging diminishes as the time passes, the spatial disturbance (distance) increases Spatial Dynamics

39. These graphs show the spatial distribution of skeletal parts from some carcasses analysed in our case study. In the image, a 3 meters zoom window from the original location has been selected. It is important to realize that the 3 meters zoom window only includes a subset of bones from the third year.

40. Our analysis pretends to examine if the number and nature of bones in one location have anything to do with characteristics in a neighbour location through the definition of a general model of spatial dependence. Spatial Dynamics

41. What we are looking for is whether the location of individual bones or skeletal parts is homogeneous or heterogeneous in the area defined by the natural disturbance process. Spatial Dynamics

42. This can be easily computed by estimating the probability of the spatial density function associated to each location. Given that locations are defined bidimensionally, we can calculate an interpolated surface representing the form of a probability density distribution for two continuous random variables, Cartesian co-ordinates x and y. Spatial Dynamics

43. The idea is to estimate this bidimensional density function, given a sample of known locations, by estimating the density in that area. In this way, the relative frequency of all observations falling in a given interval is counted. We use Kernel estimation techniques for this task Spatial Dynamics

44. Results for carcass Number 1. Contours in Black correspond to the probability distribution of first year observations, those in red to the second year and blue ones to the third year. That is, when animal bones begin to be integrated into what will become an archaeological record (after trampling) we have a disturbed spatial distribution characterised by its lower density.

45. 1996 1997 1998 Spatial model for all carcasses in the case study. Although coordinates have been taken locally at each carcass, we are looking for a probability model of densities, and ordinal transformations preserve original densities

46. Results are clearer if we use a 3D representation of the same kernel filter

47. We see that differences are not greater between the second and third year, than between the first and second. That means scavenging during the second year is still produced on relatively dense carcasses. Dispersal seems to be a consequence of long scavenging and not a direct consequence of animal action on carcasses. Spatial Dynamics

48. We obtain a new test of the previous hypothesis:as the time goes, the density of locations diminishes and the spread of bones increases. That is, although scavenging is less intense after the first year, occasional disturbance produces much more spatial dispersal than original scavenging Spatial Dynamics

49. This fact implies a negation of Tobler’s Law: near skeletal parts are not more related than distant parts. In some cases distant skeletal parts can even be more related than bones lying closer together. Spatial Dynamics

50. Post depositional transformations of the original patterning are not a simple accumulation process from low entropy (identity between depositional and activity set) to higher entropy patterns (disturbed deposits), but a non-linear sum of quantitative changes, which beyond a threshold, produce a qualitative transformation. CONCLUSIONS