P R O ME T H E US P R E S S / P A L A E O N T O L O G I C A L N E T W O R K F O UN D A T I O N (TERUEL) 2005 Journal of Taphonomy VOLUME 3 Available online at www.journaltaphonomy.com Thompson (ISSUE 2) The Impact of Post-Depositional Processes on Bone Surface Modification Frequencies: A Corrective Strategy and its Application to the Loiyangalani Site, Serengeti Plain, Tanzania Jessica C. Thompson* Box 872402, Department of Anthropology, Arizona State University, Tempe, AZ 85287-2402 Journal of Taphonomy 3 (2) (2005), 57-80. Manuscript received 9 October 2004, revised manuscript accepted 8 February 2005. The frequencies of surface modification such as percussion, cut, and tooth marks on experimental faunal assemblages are not always directly comparable to those in fossil assemblages. Extensive postdepositional modification of bone surfaces may render many of these marks unidentifiable, depressing the overall frequencies or affecting some mark classes more than others. An analysis of the fauna from an open-air Middle Stone Age site on the Loiyangalani River in the Serengeti Plain, Tanzania, illustrates this point. A coding system is presented here that allows the elimination of heavily affected fragments from analysis so that the observed mark frequencies can more closely approximate their original ones. Keywords: DIAGENESIS, RECORDING ZOOARCHAEOLOGY, SERENGETI SYSTEM, Introduction they can access the existing body of literature and use it to inform them about archaeological situations. As long as an observable process can be linked to a clear physical trace, middle range theory (e.g. Binford 1981) allows us to take these modern experimental observations and use them to make robust inferences about the taphonomic history of a given fossil assemblage. However, a direct comparison of the relative frequencies of these traces is not always appropriate. Fossil assemblages are likely to have undergone a wider array and longer duration of processes than have Over the last decade, zooarchaeology has witnessed an expansion in research focusing on actualistic and naturalistic experimental studies. Many of these have emphasized the behavioral or ecological information available in frequencies of bone surface modification such as percussion, cut, and tooth marks (e.g. Marean et al. 1992, Lupo 1994, Selvaggio 1994, Blumenschine 1995, Capaldo 1997, Domínguez-Rodrigo 1997, Bunn 2001, Lupo & O’Connell 2002). Although continuing work in this area is vital, researchers are now at the point where Article JTa030. All rights reserved. SURFACE MODIFICATION, * E-mail:Jessica.C.Thompson@asu.edu 57 Bone Surface Modification Frequencies developed a coding system which records, for each specimen, the extent and severity of surface degradation. This allows the elimination of heavily affected fragments from analysis so that the observed mark frequencies can more closely approximate their original ones. When the modified frequencies are compared to those from an experimental assemblage, they will then speak more directly to the relevant research question and less to the vagaries of preservation. Although broadly applicable across time and space, the impetus for developing this system was derived from analysis of the fauna recovered in the 2000 field season from the Loiyangalani river site in the Serengeti Plain, Tanzania (Bower & Gogan-Porter 1981; Bower, 1985; Thompson et al. 2004). The Loiyangalani site is contained in alluvial deposits of the Loiyangalani Valley, where the Loiyangalani River flows westward from the Serengeti Plain to a confluence with the Mbalageti River, a stream that empties into Lake Victoria (Figure 1). The MSA-bearing sediments, designated Unit 1, are interpreted as proximal overbank alluvial floodplain deposits, with an upper limit at about 90100 cm below surface. A lower limit has yet to be determined because the water table was reached, but it appears to exceed 150 cm below surface. Overlying these sediments are about 50 cm of more distal floodplain deposits (Unit 2) with a marked prismatic to blocky structure. These deposits are basically devoid of archaeological content; they are overlain by 40-50 cm of mixed aeolian and fluvial deposits (Unit 3) whose upper limit forms the modern surface. The Unit 3 deposits contain a light, background scatter of LSA artifacts and bone fragments that exhibit no Figure 1. Map of Serengeti National Park showing location of Loiyangalani site. modern experimental ones. Extensive degradation or alteration of fossil bone surfaces by abrasion, elemental transport, or diagenesis may render many of these marks unidentifiable. This may depress the overall frequencies or, more dangerously, differentially affect some mark classes more than others. Researchers often refer to bone surfaces as “well-preserved” or “poorlypreserved” without specifying what this means in terms of which marks are likely to remain identifiable and how their frequencies in the overall assemblage may have been affected. It should be standard practice to take these factors into account before attempting to infer behavior from a fossil assemblage. In response to these issues, I 58 Thompson crocodile. The fauna from Loiyangalani shows a great deal of post-depositional alteration through breakage, transport, and diagenesis. This makes surface modification analysis challenging and raises the distinct possibility that the absolute and relative frequencies of surface modification are not directly comparable to experimental ones. concentrated occurrence. Because the 2000 season was comprised of test excavations, there is little specific information about the spatial distribution of the fauna described here beyond the level of 1 m2 test pits. Given the small sample, the spatial distribution of surface alterations has not been analyzed, and radiometric dating of all three units is also pending. Loiyangalani is unique for many reasons. It is one of only a handful of Middle Stone Age (MSA) sites from East Africa with preserved fauna. It is an openair site, whereas most excavated MSA localities with fauna are caves or rock shelters. Also, materials rare to absent in the MSA (e.g. ostrich eggshell beads, worked ochre and bone, and fish remains) have been recovered in association with MSA lithic technology. As such, it promises to provide important data on the adaptations, technology, and behavioral ecology of Middle Stone Age hominins that go well beyond what one might be able to infer from a site without preserved organic materials. Perhaps in part because it is an open-air site, Loiyangalani has had an extremely complex taphonomic history that must be unraveled before these finds may be used to address behavioral questions. The assemblage under consideration was recovered during the 2000 field season, and is comprised of a NISP of 1795. Only 525 specimens could be confidently identified as larger mammals (falling into size classes 1 – 5 following Brain [1981]). Also included in this assemblage are 76 fragments of turtle carapace or plastron, some of which retain surface modification. A cursory examination of the remaining material reveals the presence of small mammals, fish, freshwater mollusks, birds, and Methods The entire surface of every specimen was examined under a binocular 10-40 power microscope. Percussion marks, cut marks, and tooth marks were identified and coded as having a high or medium confidence level according to the criteria set out by Blumenschine et al. (1996). For the purposes of this analysis, I have discarded all medium confidence marks to reduce uncertainty in the data. Any reference in this paper to “surface modification” refers specifically to these three mark classes and not to any other type of bone surface alteration. Figure 2 illustrates examples of high confidence percussion, cut, and tooth marks from the assemblage, on which surface degradation did not affect their ability to be identified. Five major morphological effects of post-depositional processes were observed in the assemblage. All of these were recorded using descriptive criteria rather than terminology that implies a specific bone-modifying agent (sensu Gifford-Gonzalez 1991). Figure 3a illustrates modification I have termed “dendritic” to describe what some researchers have called “root etching”. Currently there are no published accounts of controlled or experimental root activity that 59 Bone Surface Modification Frequencies Figure 2. Illustrations of (a) a percussion mark; (b) a cut mark; and (c) a tooth mark (arrow) that have not been affected by surface degradation. Photographs on the right show close-ups of the boxed areas in the photographs to the left. grooved the surface of the bone. In some cases the channels are deep and in others shallow. They may occur widely spaced or closely packed together. In Figure 3b a undeniably demonstrate that this is the sole agent that causes microscopic features such as these. Dendritic etching is defined as a series of branching channels that have 60 Thompson Figure 3. (a) Illustration of dendritic etching; (b) Close-up of percussion mark in boxed area of (a) with dendritic channel running through it; (c) Smooth and feathered channels both classified as dendritic etching. 61 Bone Surface Modification Frequencies of a bone have been rounded on a gross level so that formerly sharp-edged breaks may take on a smooth appearance. In addition to this rounding, anatomical features such as foramina, tuberosities, or fossae may become indiscernible. Although the processes behind both “rounding” and “smoothing” are likely similar, I have chosen to use smoothing as a descriptor because affected areas take on a distinctly smooth appearance. Sheen (Figures 7a – 7c)) indicates the degree to which a fragment bears a “polish” or “shine”. It is defined by how brilliantly and continuously the surface of a specimen reflects light. Under a 10 – 40 power light microscope the surface looks bright and smooth. In lightly affected fragments the microtopography of the bone surface is still apparent, with the highest parts bearing a shine. In heavily affected fragments no microtopography remains, and the area is entirely smooth and shiny. Although it is tempting to attribute these forms to abrasive action from water- or airborne particles, this may not always be the case. Because of differences in the reflective properties of different minerals involved in fossilization, the amount of sheen may rely in part on the particulars of the fossilization processes that occurred in an assemblage. At Loiyangalani the worked bone fragment recovered in 2000 displayed both smoothing and sheen (Figure 7b and 7c), as do gastrically etched bones I have observed in other assemblages. In all of these cases, the specimens were recorded using the same criteria as rest of the assemblage because regardless of the agent the effects on surface modification identification will be the same. All of these post-depositional processes were recorded for that portion of channel has cut directly across a percussion mark. The mark remains intact on either side of the channel, but only because the channels are discontinuous and not widely spaced. For these reasons, I have predicted that mark frequencies will be less affected by this particular form of surface degradation than by the other four types of degradation. At Loiyangalani, dendritic etching took two forms (Figure 3c). As in the illustrated specimen some channels were smooth while others had tiny fingerlike projections radiating from them. Identifying the agencies behind these different modifications would prove useful for taphonomists seeking to understand these common processes at open-air sites. A second morphological category is pocking (Figures 4a – 4b). This is when the bone surface has been obliterated in small, semicircular patches. As with dendritic etching, pocking may occur sparsely or densely, deep or shallow. Figure 4c illustrates a heavily pocked area with a tooth mark preserved next to this area. Had the pocking covered the entire bone surface or a different area of the bone, the mark would have been obscured. Exfoliation (Figures 5a – 5b) is as described when the bone surface is peeling away or being removed in sheets. Of the three degradation types described so far, I predicted that this would have the greatest affect on the probability of identifying marks because it is continuous and marks will be obliterated in affected areas. Smoothing and sheen were also recorded as morphological types of postdepositional modification. These two types of modification are most likely caused by similar processes operating at different scales. Smoothing (Figures 6a and 6b) indicates the degree to which the contours 62 Thompson Figure 5. (a) Illustration of exfoliation; (b) Close-up of exfoliation in the boxed area of (a). was recorded separately. All fragments were coded using a 0 – 3 system for two variables: severity and extent. Extent indicates if the bone has none of that particular type of alteration (0); 1 – 50% of it is covered (1); greater than 50% but less than 100% is covered (2); or 100% is covered (3). This system could be modified to any scale to fit the resolution required by the analysis. For example, recording the extent in 10 percent increments would be nearly as easy and more precise. Severity indicates the amount of impact that each form of modification has had on the integrity of the bone surface. It is a dependent variable, in that at least a 1 must be recorded for severity if any extent has been noted at all. Conversely, if a 0 is recorded for extent, then severity must also be 0. If more than one level of severity Figure 4. (a) Illustration of pocking; (b) Extensively degraded area of bone near the top with tooth mark near pocking; (c) Close-up of tooth mark (arrow) near pocking from boxed area in (b). Note the grooved cross-section with no associated microstriations, and the small fragments of bone that have been pressured up around the mark. the bone that is not covered with matrix, including the interior of medullary cavities. The proportion of the surface of each fragment that was not covered by matrix 63 Bone Surface Modification Frequencies Figure 6. (a) Illustration of smoothing; (b) Close-up of smoothed edge of the bone in boxed area of (a). The specimen was turned slightly for the second photograph to capture the edge. 64 Thompson Figure 7. (a) Illustration of sheen; (b) Worked bone with both smoothing and sheen on edges; (c) Close-up of tip of the worked bone from boxed area of (b) showing sheen. 65 Bone Surface Modification Frequencies the variables are present, a single click enters zero for every field (Figure 8). The major benefit of a sheet such as this linked to a database is that the parameters may be entered very rapidly, costing the analyst little time while potentially providing a wealth of additional information about the taphonomic history of the assemblage, the geological history of the site, and bone surface preservation. occurs on the bone surface, the greater value is recorded. Severity is a much more qualitative measure than extent. In order to standardize my evaluations, I selected several specimens to act as references for each level of severity and then coded the entire assemblage according to those criteria. The most severe, stage 3, is illustrated in Figure 3a (dendritic), Figure 4a (pocking), Figure 5a and 5b (exfoliation), Figure 6a and 6b (smoothing), and Figure 7a (sheen). I have designed a data entry form that is inserted as a page in a larger data entry system for bone fragments written for the Paradox 9 database program. I am currently designing a revision of this sheet for the more widely used program Microsoft Access. If none of Analysis of effects of surface degradation I constructed a series of graphs to illustrate the effects these five types of postdepositional processes have on surface modification frequencies (Figures 9a – 9e). Figure 8. Datasheet for Paradox 9 currently used to record the five types of surface degradation described here. This sheet is being revised for use with Microsoft Access. As each button is pushed the data is entered directly into a database. If no surface alterations are present, a single click on “No Diagenetic Weathering” and “No Sheen or Smoothing” will enter zeros into each field. 66 Thompson Figure 9. Graphical illustration of the differences in the proportions of high confidence percussion, cut, and tooth marks between little affected fragments and heavily affected fragments for (a) Dendritic etching; (b) Pocking; (c) Exfoliation; (d) Smoothing; and (e) Sheen. The number of marked fragments (n) is given to the right of each bar. worsens. In the case of dendritic etching only, a pattern opposite of that expected is apparent: as surface preservation worsens, the proportion of fragments exhibiting a percussion or cut mark actually increases, though the proportions of tooth marked fragments behave as expected. The increase in all other surface alterations causes a decrease in the frequency of surface modifications, as expected. Pocking and exfoliation both show some decrease in the proportion of marked fragments as severity worsens, and both smoothing and sheen show a dramatic drop. For all types of alteration except dendritic etching, the relative frequencies of identifiable marks decrease, as expected. This supports the hypothesis that as surface preservation worsens, marks become more difficult to identify. It also supports the For each type of alteration, the entire assemblage was broken down into two subassemblages termed “little affected” and “heavily affected”. I defined heavily affected fragments as those with an extent of two or three and a severity of two or three for any of the diagenetic types. Fragments with an extent of 0 or 1 were included with the little affected category because regardless of the severity bones with less than 50% of their surface affected would likely still have a high chance of retaining surface modification regardless of how severe it is. Within each subassemblage, the percentage of fragments displaying a high confidence percussion mark, cut mark, or tooth mark was calculated. For example, Figure 9a shows how the proportions of percussion, cut, and tooth marks changes as dendritic etching 67 Bone Surface Modification Frequencies 68 Table 1. Two-way tables used in Fisher’s exact test for independence. H0 stated that the amount of surface degradation is independent of the presence or absence of percussion, cut, and tooth marks. For the two-tailed tests, a lower p-value indicates a greater probability that the two variables are not independent. H1 stated that a higher amount of surface degradation resulted in fewer marked specimens. For the one-tailed tests, a lower p-value indicates a greater probability that higher frequencies of marked bones are associated with a higher degree of surface degradation. Thompson Table 2. Data underlying graphs in Figure 10. Percentage change was calculated by taking the difference between the proportion of marked bones in the assemblage before and after removal of heavily affected fragments, then dividing this difference by the “before” proportion. HC Percussion Mark HC Cut Mark HC Tooth Mark Total Fragments Entire assemblage Number of marked fragments Before removal After removal 25 20 28 22 17 14 1795 Percentage of marked fragments Before removal After removal 1.4% 1.8% 1.6% 2.0% 0.9% 1.3% 1099 % change overall % change relative to previous HC Percussion Mark HC Cut Mark HC Tooth Mark HC Percussion Mark HC Cut Mark HC Tooth Mark Total Fragments 0.4% 0.4% 0.3% 30.7% 28.3% 34.5% Midshafts Number of marked fragments Before removal After removal 9 7 13 11 3 3 202 Percentage of marked fragments Before removal After removal 4.5% 5.2% 6.4% 8.1% 1.5% 2.2% 135 % change overall % change relative to previous HC Percussion Mark HC Cut Mark HC Tooth Mark 0.7% 1.7% 0.7% 15.6% 26.6% 46.7% hypothesis that some types of alteration will have a greater effect than others. In order to assess this statistically, three two-way tables were assembled within each of the five types of alteration (Table 1). Each table provides a count of the number of little affected and heavily affected fragments that preserve and do not preserve surface modification. For each table I performed Fisher’s exact test which, unlike a Chi- 69 Bone Surface Modification Frequencies may be more quickly erased than the main body of the mark in some cases and not others (Shipman & Rose 1988). If this is true, then raw comparisons of the relative frequencies of hominin and carnivore modification would be particularly suspect in assemblages with a substantial proportion of poorly preserved bone surfaces, where hominin damage might be rendered unidentifiable more often than carnivore damage. Most of the tables failed to provide statistically significant results at the 0.05 or even 0.1 level. This could be attributable to several factors. First, there may simply be no statistically significant relationship between the degree of surface degradation and the probability of identifying a given mark type. However, 33% (5/15) of the one-tailed results were statistically significant in the expected direction at the 0.1 level, where only 10% are expected to be significant by random chance. Likewise, 20% (3/15) were statistically significant in the expected direction at the 0.05 level, where only 5% are expected to be significant by chance. This suggests that other variables may be affecting the significance tests. One distinct possibility is the overall small sample size of marked fragments in the Loiyangalani assemblage. Another is that there may be some causal relationship between the agent of surface degradation and the agent of surface modification. The potential underlying causes of both are discussed below. To explore the effects that removing heavily affected fragments would have on bone surface modification frequencies, I plotted the proportion of marked bones in the assemblage before and after removal of these fragments (Figures 10a and 10b). I chose to remove only square test, can be performed on two-way tables that contain very small values or values of zero. Fisher’s exact test is used to determine if there are nonrandom associations between two categorical variables, in this case severity of bone surface degradation and the ability of an analyst to identify surface modification (as measured by the frequency of these marks). The null hypothesis H0 states that there is no association between the two variables, and is tested using a two-tailed version of Fisher’s exact test. A lower p-value indicates a higher chance of rejecting the null hypothesis and concluding that, for a given type of surface alteration and mark type, the amount of surface degradation does influence the number of observed surface modifications. The alternative hypothesis H1 states that there is a negative association between the two variables: as surface degradation increases mark frequency decreases. A one-tailed version of Fisher’s exact test calculates the exact probability of obtaining the observed cell frequencies and all cell frequencies with a more extreme deviation from the expected values for all two-way tables with the same fixed marginal totals. Lower p-values for the one-tailed test indicate a higher chance of rejecting the alternative hypothesis and concluding that, for a given type of surface alteration and mark type, heavily affected fragments will exhibit fewer identifiable marks. The results of this procedure show that the results with the lowest p-values are observed with the frequencies of percussion marks. This suggests that some mark classes are differentially susceptible to surface degradation. For example, the very fine microstriations required for cut and percussion marks to be reliably identified 70 Thompson Figure 10. Graphical illustration of the differences in the overall proportions in the assemblage of high confidence percussion, cut, and tooth marks before removal of the heavily affected fragments and after removal of these fragments for (a) The entire assemblage; and (b) Midshafts only. The number of marked fragments (n) is given to the right of each bar. Data used in these graphs are available in Table 2. diagnostic of carnivore or hominid behavior because these portions are most likely to survive both ravaging by carnivores and other density-mediated processes (e.g. Bunn & Kroll 1986, Lyman 1984, Lam et al. 1998, Marean et al. 1992). If one were to reconstruct the timing and nature of hominin or carnivore interaction with the assemblage, it is important to know how the surface modification frequencies for these heavily affected fragments because in the total assemblage from Loiyangalani, only 19% (337 of 1795) of the fossils had surfaces that were completely unaffected by any of the alterations described above. I performed this operation both for the entire assemblage and for midshafts only. The second graph is included because experimental data have shown that surface modification on the midshaft is most 71 Bone Surface Modification Frequencies rejected on the basis of these tests alone at the either the 0.01 or 0.05 level as is the customary target in many statistical reports. Clearly, an increased sample size would result in more robust statistical inferences. This paper deals only with data available from the 2000 season at Loiyangalani, but fauna has also been recovered from excavations in 2003 (n = 1517) and 2004 (n = 1615). When these additional assemblages have been studied and recorded using the same system described here, the statistics can be run again on a sample that has been tripled. Future statistical work that might prove fruitful would be the application of multivariate techniques. Changes in the relative frequencies of cut, tooth, and percussion marks could then all be taken into account simultaneously, rather than analyzed individually. diagnostic portions relate to the results from the larger assemblage. Figures 10a and 10b both show an increase in the relative frequencies of all mark types. Relative to the previous percentage in the total assemblage, percussion marks increase by 30.7% , cut marks by 28.3% , and tooth marks by 34.5%. For midshafts, percussion marks increase by 15.6%, cut marks by 26.6%, and tooth marks by 46.7% (Table 2). I tested for the statistical significance of this increase through a procedure that takes advantage of the binomial distribution of these data. First, I took the number of successes for a given table to be the number of fragments that retain a percussion, cut, or tooth mark. Second, I assumed that the relative frequency of successes before removal of the heavily affected fragments was the true probability of obtaining the observed number of successes. I then calculated the probability of obtaining the number of successes observed after removal if the true probability was the same as that observed before removal. A lower p-value indicates a higher chance of rejecting the null hypothesis H0 that there is no statistically significant difference between the relative frequencies of a given mark type before and after removal of heavily affected fragments. The data and results of this procedure are provided in Table 3. The results for the entire assemblage consistently resulted in lower p-values than for midshafts only, which likely reflects the small sample size of midshafts. Again, the fact that all the relative frequencies of surface modifications increased when heavily affected fragments were removed does provide some support for rejection of the null hypothesis. However, none were individually significant at below the 0.15 level and the null hypothesis cannot be Discussion The small absolute and relative frequencies of percussion, cut, and tooth marks at Loiyangalani are not consistent with published frequencies from either naturalistic or experimental conditions (Marean et al. 1992, Selvaggio 1994, Blumenschine 1995, Capaldo 1997, Domínguez-Rodrigo 1997). This is the case even when the effects of poor surface preservation are taken into account. This may indicate that the fauna from Loiyangalani was subjected to less modification from both humans and carnivores than were the fresh assemblages. Alternatively, other taphonomic factors have acted to further depress these frequencies. In general, the more heavily fragmented an assemblage is, the lower the frequencies of surface modification will be 72 Table 3. Data used to test for a statistically significant increase in the proportion of marked fragments after removal of heavily affected fragments for the entire assemblage and for midshafts only. The null hypothesis in each case is that there will be no difference between the proportions of marked fragments before and after removal. Assuming a binomial distribution, the number of fragments displaying a mark is taken to be the number of successes, and the true probability of success is assumed to be the proportion before removal. Given these parameters, the likelihood of obtaining the number of successes observed after removal is reflected in the p-value. 73 Thompson Bone Surface Modification Frequencies Table 4. (a) Numbers of fragments and (b) mean size of fragments exhibiting each stage of extent and severity for dendritic etching, pocking, exfoliation, smoothing, and sheen. Each category of surface alteration should be read independently of the others as some fragments display more than one type. experimental assemblages, the Loiyangalani fauna has undergone a very high degree of post-depositional breakage. The average fragment size is small (162 mm2), and several conjoining pieces of long bones were recovered in a highly fragmented state in situ. Post-depositional breakage on this scale has almost certainly led to depressed relative frequencies of all classes of surface modification at Loiyangalani, and even corrected proportions should be taken as minimum estimates. Table 4 provides the numbers and mean sizes of fragments displaying each type, extent, and severity of surface alteration. (Abe et al. 2002). This is certainly a problem at Loiyangalani, where there seems to have been a great deal of postdepositional fragmentation. As with surface modification, actualistic studies provide a useful starting point. The angle and outline of each ancient long bone break was recorded for the mammal long bones according to the criteria set forth by Villa & Maheiu (1991). Figure 11 shows the frequencies of break angles and outlines associated with experimental assemblages broken by different agents while the bones were in a fresh state (Marean et al. 2000). It is apparent that when compared to the 74 Thompson Figure 11. Proportions of ‘green bone’ breaks (curved/V-shaped outline or oblique angle) and ‘dry bone’ breaks (transverse outline or right angle) on long bones at Loiyangalani compared to experimental assemblages broken by different agents while bones were fresh (from Marean et al. 2000). mark can still be identified. However, dendritic etching may be shown through experimental work to be caused by heavy vegetation. The activities of both hominins and carnivores would differ in this microenvironment from those practiced in open areas, which would be reflected in the relative proportions of carnivore and hominin bone surface modifications. Alternatively, the amount of dendritic etching may relate to the amount of flesh that was left on the bones, perhaps promoting increased biological activity in that area (Behrensmeyer, pers. comm.). The current analysis cannot address either of these scenarios because the cause of dendritic etching is yet to be established. However, it does suggest that valuable taphonomic data are available in the types of surface degradation in a fossil assemblage, and that these processes should A second possible confounding factor may be an association between the agencies that leave percussion, cut, and tooth marks and the agencies that degrade bone surfaces. The microenvironments in which particular types of surface degradation are likely to occur might also be the same microenvironments in which hominin or carnivore activity was concentrated or avoided (Blumenschine, pers. comm.). For example, it is striking that the proportions of dendritically etched fragments show the reverse pattern than what was expected, and that the two-tailed result of Fisher’s exact test has the lowest pvalues for percussion-marked fragments displaying dendritic etching. This may be simply because it leaves open areas where marks may be preserved and therefore has very little effect on the probability that a 75 Bone Surface Modification Frequencies Figure 12. Proportions of (a) ‘green’; and (b) ‘dry’ bone breaks for long bones within each stage of smoothing. Stage 0 shows no smoothing and Stage 3 is the most affected declines accordingly (Figures 12a and 12b). These data indicate that bones that were broken while fresh have undergone relatively more smoothing than have bones that were broken while dry. Controlled experimental work aimed toward understanding the agencies behind the bone surface alterations described here would help address why this is the case, and be invaluable to future taphonomic research. Such work would enable researchers to link patterning such as that observed in long bone breakage at Loiyangalani to actual taphonomic processes, thus allowing for the advantages of the coding system presented here to be fully realized. be recorded for reasons other than simply to eliminate heavily affected fragments from analysis. Recording the gross effects of these post-depositional processes using the methods outlined above has another critical use. Breakage and other taphonomic data can be incorporated and used to develop hypotheses about the depositional history of a site. At Loiyangalani, most of the mammal long bones that have been heavily smoothed and/or polished more frequently retain oblique angle and curved/v-shaped fractures than those with breaks more often associated with “dry” bone fragmentation. The frequencies of curved and oblique breaks steadily increase in frequency as they become more smoothed and the frequencies of transverse and right angled breaks 76 Thompson correct for it, and use it in conjunction with other taphonomic lines of evidence. With experimental work designed to understand the agents that cause various surface alterations, the taphonomic and geological history of site formation can be further elucidated. To date, I have used this system to record samples from three separate fossil assemblages including both cave and openair sites. Although the underlying taphonomic processes operating at each of these different localities have likely been very different, I have still found the system to be effective at quantifying and describing the range of surface preservation. Comparing the results from these widely separated sites will provide additional data to test the hypotheses presented here. The issues discussed here are pertinent to all archaeological fossil assemblages, and should be explicitly addressed before higher-level inferences involving comparisons to modern experimental datasets may proceed. Conclusion The Loiyangalani case illustrates that bone preservation should be taken into account with more rigor and standardization. Despite a small sample and the potential influence of confounding factors, the results of this study indicate that an appropriate first step in correcting for the effects of post-depositional processes is to remove heavily affected fragments from calculations of surface modification frequencies. The result will be a general increase in the relative frequencies of all mark classes and a higher level of confidence in comparisions to experimental frequencies of surface modification. A precise quantification of what percentages of surface modification are actually lost would require actualistic research, where the number and type of mark is known prior to degradation of the bone surface. The processes behind long-term diagenesis could be simulated for a variety of depositional (and post-depositional) environments. It would also be useful to determine the extent to which subaerial weathering (Behrensmeyer 1978) and surface coverage by matrix influence mark frequencies. Given that some types of diagenesis impact bone surfaces to a greater degree than others, it would be useful to employ a larger sample size to statistically explore whether or not mark classes are also affected differentially. Zooarchaeologists must become more explicit when they refer to an assemblage as “nicely preserved” or “poorly preserved”. At Loiyangalani, as at most sites, some bones have surfaces that were preserved better than others. Using a system similar to the one presented here, one can begin to sort out this variability, Acknowledgements I would like to thank Curtis Marean, Robert Blumenschine, A. Kay Behrensmeyer, and two anonymous reviewers for their comments on earlier drafts. Briana Pobiner and David Braun organized the symposium that was the forum for the original presentation of this paper, seeing many papers given there through to publication. Erich Fisher provided a more detailed version of the map of the Serengeti for Figure 1. I would also like to thank John Bower and Audax Mabulla, co-directors of the Loiyangalani Research Project for access to the faunal material. Carl Vondra provided a preliminary assessment of the 77 Bone Surface Modification Frequencies Brain, C. K. (1981). The Hunters or the Hunted? An Introduction to African Cave Taphonomy. University of Chicago Press, Chicago. Bunn, H. T. and Kroll, E. M. (1986). Systematic butchery by Plio/Pleistocene hominids at Olduvai Gorge, Tanzania. Current Anthropology 27(5): 431-452. Bunn, H. T. (2001). Hunting, power scavenging, and butchering by Hadza foragers and by PlioPleistocene Homo. In Meat Eating and Human Evolution, edited by Stanford, C. B. and Bunn, H. T., pp. 199-218. Oxford University Press, Oxford. Capaldo, S. D. (1997). Experimental determinations of carcass processing by Plio-Pleistocene hominids and carnivores at FLK 22 (Zinjanthropus), Olduvai Gorge, Tanzania. Journal of Human Evolution 33: 555-597. Domínguez-Rodrigo, M. (1997). Meat-eating by early hominids at the FLK 22 Zinjanthropus site, Olduvai Gorge (Tanzania): An experimental approach using cut-mark data. Journal of Human Evolution 33: 669-690. Gifford-Gonzalez, D. (1991). Bones are not enough: analogues, knowledge, and interpretive strategies in zooarchaeology. Journal of Anthropological Archaeology 10: 215-254. Lam, Y. M., Chen, X., Marean, C. W., and Frey, C. J. (1998). Bone density and long bone representation in archaeological faunas: Comparing results from CT and photon densitometry. Journal of Archaeological Science 25: 559-570. Lupo, K. D. and O'Connell, J. F. (2002). Cut and tooth mark distributions on large animal bones: Ethnoarchaeological data from the Hadza and their implications for current ideas about early human carnivory. Journal of Archaeological Science 29: 85-109. Lyman, R. L. (1984). Bone density and differential survivorship of fossil classes. Journal of Anthropological Archaeology 3: 259-299. Marean, C. W., Abe, Y., Frey, C. J., and Randall, R. C. (2000). Zooarchaeological and taphonomic analysis of the Die Kelders Cave 1 layers 10 and 11 Middle Stone Age larger mammal fauna. Journal of Human Evolution 38: 197-233. Marean, C. W., Spencer, LM., Blumenschine, R. J., and Capaldo, S. D. (1992). Captive hyaena bone choice and destruction, the Schelpp effect and Olduvai archaeofaunas. Journal of Archaeological Science 19: 101-121. Selvaggio, M. M. (1994). Carnivore tooth marks and stone tool butchery marks on scavenged bones: Archaeological implications. Journal of Human Evolution 27: 215-228. geological context of the materials. Excavations at the Loiyanalgani site were made possible by Serengeti Genesis, Tanzania Antiquities, Tanzania National Parks, Serengeti National Park, the University of Dar es Salaam, the Leakey Foundation, and the National Geographic Society. References Abe, Y., Marean, C. W., Nilssen, P. J., Assefa, Z., and Stone, E. C. (2002). The analysis of cutmarks on archaeofauna: A review and critique of quantification prodecures, and a new imageanalysis GIS approach. American Antiquity 67(4): 643-664. Behrensmeyer, A. K. (1978). Taphonomic and ecologic information from bone weathering. Paleobiology 4: 150 – 162. Binford, L. R. (1981). Middle Range Research and the Role of Actualistic Studies. In Bones, Ancient Mean and Modern Myths, by L. R. Binford, pp. 2130. Academic Press, New York. Blumenschine, R. J. (1988). An experimental model of the timing of hominid and carnivore influence on archaeological bone assemblages. Journal of Archaeological Science 15: 483-502. Blumenschine, R. J. (1995). Percussion marks, tooth marks, and experimental determinations of the timing of hominid and carnivore access to long bones at FLK Zinjanthropus, Olduvai Gorge, Tanzania. Journal of Human Evolution 29: 21-51. Blumenschine, R. J., Marean, C. W., and Capaldo, S. D. (1996). Blind tests of inter-analyst correspondence and accuracy in the identification of cut marks, percussion marks, and carnivore tooth marks on bone surfaces. Journal of Archaeological Science 23: 493-507. Bower, J. R. F. and Gogan-Porter, P. (1981). Prehistoric Cultures of the Serengeti National Park: Initial archaeological studies of an undisturbed African ecosystem. Papers in Anthropology 3. Iowa State University. Bower, J. R. F. (1985). Excavations at the Loyangalani site , Serengeti National Park, Tanzania (with contributions by D. P. Gifford and D. Livingstone). National Geographic Society Research Reports (1979 projects) 20: 41-59. 78 Thompson Shipman, P. and Rose, J. (1988). Bone tools: An experimental approach. In Scanning Electron Microscopy in Archaeology, edited by Olsen, S. L., pp. 303-335. British Archaeological Reports International Series 452. Thompson, J. C., Bower, J. R. F., Fisher, E. C., Mabulla, A. Z. P., Marean, C. W., Stewart, K., and Vondra, C. F. (2004). Loiyangalani: Behavioral and Taphonomic Aspects of a Middle Stone Age site in the Serengeti Plain, Tanzania. Paper given at the Paleoanthropology Society meeting, Montreal, Québec, Canada. Villa, P. and Mahieu, E.(1991). Breakage patterns of human long bones. Journal of Human Evolution 21: 27-48. 79 Bone Surface Modification Frequencies 80