Sie sind auf Seite 1von 22

Society for American Archaeology

Trampling the Archaeological Record: An Experimental Study Author(s): Axel E. Nielsen Source: American Antiquity, Vol. 56, No. 3 (Jul., 1991), pp. 483-503 Published by: Society for American Archaeology Stable URL: http://www.jstor.org/stable/280897 Accessed: 06/05/2010 09:56
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=sam. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to American Antiquity.

http://www.jstor.org

TRAMPLING THE ARCHAEOLOGICAL RECORD: AN EXPERIMENTAL STUDY


Axel E. Nielsen
This paper reports on several experiments carried out to explore the transformations of the archaeological record affected by trampling. These transformations include changes in artifact distributions and formal alterations that should be taken into account when carrying out studies of activity areas. The experiments were made on dry, hard-packed surfaces and in the same sediments after a rain. The materials used were bones, obsidian flakes, sherds, and fragments of brick and wood. The analysis focuses on vertical displacement, horizontal displacement, and damage (breakage, microflaking, and abrasion), paying special attention to the response of the trodden substrate and its implications for the whole process. The interaction of trampling with other formation processes (e.g., maintenance) also is considered. The main patterns observed in the trampled materials are vertical and horizontal size sorting, and characteristic size distributions in sherds. These empirical generalizations are then integrated in a model that can help to identify trampled contexts and assess their potentialfor behavioral inference. El presente art?culo describe varios experimentos realizados con el prop6sito de explorar las transformaciones producidas por pisoteo en el registro arqueologico. Tales transformaciones incluyen cambios en la distribucion de artefactos y alteraciones formales que deben tenerse en cuenta al realizar estudios de areas de actividad. Los experimentos fueron efectuados sobre superficies muy compactas, secas y luego de una lluvia. Se utilizaron huesos, lascas de obsidiana, tiestos y fragmentos de ladrillo y madera. Los aspectos que se analizan son desplazamiento vertical, desplazamiento horizontal y danio (fractura, microlascado y abrasion), prestando especial atencion a la respuesta del substrato pisoteado y sus implicancias para el proceso en su conjunto. Tambien se considera la interaccion delpisoteo con otros procesos deformacion (p.e., mantenimiento). Los principales patrones observados en los materialespisoteados incluyen ordenamiento verticaly horizontalpor tamanio y distribuciones caracteristicas en la dimensi6n de los tiestos. Estas generalizaciones empiricas son luego integradas en un modelo que puede contribuir a identificar contextos pisoteados, asi como a evaluar su potencial para establecer inferencias de cardcter conductual.

Since Stockton's (1973) pioneering study, trampling by humans and animals has been recognized as a major process by which archaeological materials and deposits are transformed in their formal and spatial attributes (e.g., Schiffer 1983, 1987). Understanding the potential effects of this process is a prerequisite for many behavioral inferences in situations where treadage is likely to have taken place. During the last two decades, for instance, many studies have attempted fine-grained reconstructions of the spatial organization of living floors. Typically these analyses identify discrete areas devoted to limited groups of activities like food processing and consumption, storage, trash disposal, tool manufacture and maintenance, resting, etc. In order to make these kinds of inferences it is necessary to know minimally: (a) the activities in which the artifacts were used; (b) the circumstances that led to artifact deposition (whether they constitute primary, secondary, or de facto refuse); and (c) if there have been changes in their formal and spatial attributes after deposition. It is in the context of this last problem that trampling, along with other processes of disturbance have to be taken into account. Intensive trampling modifies the horizontal distribution of artifacts, it obscures patterns existing in their original deposition, and eventually introduces new trends in their spatial arrangement. By producing vertical migration of materials it also can move artifacts across stratigraphic units, and mix in the same deposits items originating in different occupations. When trodden, artifacts undergo several types of damage, like breakage, microchipping and abrasion. The resulting traces sometimes mimic the damage produced by use or by other postdepositional processes, and
Axel E. Nielsen, Laboratory of Traditional Technology, Department of Anthropology, University of Arizona, Tucson, AZ 85721, and Cdtedra de Prehistoria y Arqueologia, Escuela de Historia, Universidad Nacional de Cordoba, Argentina

Copyright? 1991 by the Society for AmericanArchaeology

American Antiquity, 56(3), 1991, pp. 483-503.

483

484

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

therefore can unwittingly lead to erroneous functional interpretations. Since trampling is a ubiquitous process on occupation surfaces, its effects cannot be overlooked when assessing the suitability of particular deposits for carrying out spatial studies at the microscale (Clarke 1977). Trampling also can be considered a broad category of human activity in itself, or a common element of various activities. Some models concerning the differential use of space can be characterized in terms of sharp differences in the amount of human traffic. By inferring the presence and relative intensity of trampling in different spatial units rough functional distinctions can be made (e.g., storage rooms vs. habitation rooms; areas of domestic or restricted circulation vs. areas open to public traffic [cf. Whittlesey et al. 1982]). A number of authors have taken into account possible alterations of deposits resulting from trampling based on "reasonable assumptions" of what its effects are likely to be (e.g., Bradley and Fulford 1980; Hughes and Lampert 1977; Rosen 1986, 1989). In addition, there have been several attempts to explore trampling ethnoarchaeologically (De Boer and Lathrap 1979; Gifford 1978; Gifford and Behrensmeyer 1977; Wilk and Schiffer 1979) and experimentally under different degrees of control (Behrensmeyer et al. 1986; Courtin and Villa 1982; Flenniken and Haggerty 1979; GiffordGonzalez et al. 1985; Lindauer and Kisselburg 19n and 1981;Muckle 1985; Olsen and Shipman 1988; Pintar 1987; Pryor 1988; Stockton 1973; Tringham et al. 1974; Villa and Courtin 1983). These studies have focused primarily on two issues: (1) how human trampling disturbs stratigraphic sequences by producing vertical migration of items, and (2) how treadage generates patterns of damage (mainly in lithics and bone) in order to differentiate them from damage produced by use or butchering
activities.

Although several generalizations have begun to emerge as a result of this work, it is surprising that the results of different studies vary widely and are even contradictory in many respects (compare, for example, Tringham et al. [1974] and Flenniken and Haggarty [1979] on edge damage, or GiffordGonzalez et al. [1985] and Pintar [1987] on the relation between size/weight of artifacts and vertical displacement). This situation indicates that these kinds of experiments will have to be repeated many times for reliable generalizations to be drawn, and that considerable work is still needed before we are able to apply them to archaeological inference. The present paper reports on six experiments designed to examine some of the contradictory results achieved by previous studies and to explore aspects of trampling processes that have received little attention in the literature. These include: (a) patterns of ceramic breakage, (b) the influence of the different density of various materials on displacement, and (c) the interaction between trampling and other formation processes. MATERIALS AND PROCEDURES The six experiments are labeled TR-I through TR-VI and were carried out in backyards and in a park in the city of Tucson. A summary description of them is presented in Table 1. The next three sections offer details about the trampled substrate, the materials used, and the design followed in each case. The Substrate Except for TR-III, all experiments were performed on dry, highly consolidated surfaces with no vegetation cover. TR-III was carried out on the same sediments but five hours after a heavy rain in order to assess the effects of trampling on a wet, softer substrate. Two attributes of the substrate are considered to have the most influence on the way trampling impacts the archaeological record: texture and penetrability. A grain-size analysis of the sediments in the trampled areas showed that according to their texture they could be classified as "muddy gravels" (Folk 1980): 79 percent gravel, mostly in the granule fraction; 10 percent sand; 11 percent mud. A pocket penetrometer (Bradford 1986) was used to measure the penetrability of the substrate, with limited success. This is a hand-operated, calibrated-spring tester that measures penetrability in kilograms per square centimeter necessary to stick its tip into the ground. This is the only technique

Table 1. Summary of Features of Each Experiment. TR-I Number of items Materials used Size of original concentration Wet/dry Soil penetrabilitya Number of crossings Variables considered 318 bones, lithics, sherds 1 x 1m dry 2.49 1,500 vertical-movement, damage 173 bones, lithics, sherds 1x Im dry >4.5 800 vertical and horizontal movement TR-II 78 bones, lithics, sherds 1 x 1m wet 1.63 800 vertical and horizontal movement TR-III 88 sherds .5 x .5 m dry >4.5 100, 200, 300, 400, 800 fracture TR-IV

.5

>

a Soil penetrability measured in kg/cm2.

486

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

Table 2. Size Distribution of Pieces Used in TR-I, TR-II, and TR-III. Size Categorya 1 TR-I Bones Sherds Lithics Total TR-II Bones Sherds Lithics Total TR-III Bones Sherds Lithics Total
a C,7,t I31L C fLL,rt

2 25 30 41 96 5 12 20 37 5 7 8 20
I

3 23 26 27 76 9 3 15 37 10 7 5 22
mm

4 19 19 7 45 3 11 10 24 6 1 3 10
i=

6 1 1 2 4 2 9 12 0 2 2 4

7 15 0 1 16 4 12 1 1 17 5 2 0 7

Total 100 107 111 318 34 73 66 173 30 22 26 78


4
= -t l_ I-

11 25 28 64 4 2 17 23 1 0 7 8
l_, lll i. lr- C tal =

6 6 5 17 7 14 2 23 3 3 1 7

"
I1-1-Ullllll,.L

m. - j Am = '11I-+killllll,-t m

50 mm; 5 = 51-60 mm; 6 = 61-70 mm; 7 = >70 mm.

for measuring the resistance to penetration of a surface, since others, like bulk density (Black 1965: 381), measure the compaction of the top layer as a whole. Its results, however, are not precise enough to be taken as an absolute measure of penetrability but rather as a relative estimation for broad comparisons. Each locality was tested over an area of 5 m2 (10 points per m2) immediately before and after the experiments in order to assess variations in penetrability. The means of these
measurements Materials The materials used were obsidian flakes, coyote and sheep bones weathered 2-3 months (fragments of mandible, diaphysis and articular parts of long bones, and vertebrae), fragments of oak wood slabs made of and brick, and sherds from the following five types of pottery: (a) High-tempered for each case are displayed in Table 1 and discussed in detail below.

commercial clay (Westwood EM-207) fired at 700?C for 30 minutes (thickness 7 mm); (b) small Mexican low-fired globular vessel (12 cm high, wall thickness 4.3-5.6 mm); (c) large Mexican lowfired globular vessel (40 cm high, wall thickness 4.8-7.4 mm); (d) biglobular (gourd-shaped) Mexican vessel (wall thickness 4.5-7.0 mm); (e) Italian high-fired flower pot (wall thickness 3.9-4.8 mm with an increase to 6.2 mm in a 30-mm band along the rim). These types are presented in order of increasing hardness, determined mainly by differences in firing temperature, and therefore the grades A through E can be considered a rough ordinal measure of the strength of the paste.
The Experiments During the experiments attention was paid to three different aspects of trampling. TR-I through

TR-III focused on horizontal and vertical movement and general damage in artifacts. TR-IV and TR-V were designed to examine patterns of ceramic breakage, and TR-VI focused on the influence of material density and object "bulk"' on horizontal migration. Accordingly, three basic designs
were followed. TR-I, TR-II and TR-III. Before each experiment all items were numbered and weighed, and

their maximum length was recorded (Table 2). The flakes were spray painted to facilitate the

REPORTS

487

identification of damage. All pieces were then placed in a 1-x- 1-m square, mapped, and exposed to different amounts of trampling. The areas were then excavated with paint brushes, recording the location of every object and noting the existence of abrasion, microflaking, or breakage. To facilitate the mapping process in TR-II, instead of triangulating each item (as in TR-I and TR-III), provenience was recorded by 20-x-20-cm grid units. The amount of trampling was measured in terms of the number of crossings over the square where the artifacts were placed. Each crossing represents two steps per square meter. Attention was paid to enter the area evenly from all possible directions. All tramplers weighed 62-75 kg and wore tennis shoes. The number of crossings was 1,500 for TR-I and 800 each for TR-II and TR-III. TR-IVand TR-V. In the first case, a gourd-shaped pot (type D) was dropped from 1.5 m on a cement floor. The resulting fragments were measured (maximum length, width, curvature, and weight), and classified in 1-cm size categories by maximum length. Pieces smaller than 1 cm were not considered, and those bigger than 71 mm were lumped in category "7." Sherds included in each size category were spray painted with different colors. The assemblage was then laid out randomly in a 50-x-50-cm area, trampled 100 times (each crossing one step on the material), collected, and measured again. The size distributions for fragments belonging to each original size category (identified by a distinct color) were recorded separately. The same procedure was repeated four times (100, 200, 300, 400 crossings) and again after 400 crossings, for a total of 800 crossings. TR-V followed the same general design using a flower pot (type E) and a very low-fired Mexican vessel (type B). The size distributions of sherds were recorded after the initial fracture, and after 50, 100, and 400 crossings. TR- VI. Fragments of oak wood and ceramic brick of three different sizes (73.4 cm3, 17.6 cm3, and 6.3 cm3) and 20 sherds were laid on a frequently used dirt path in a park in Tucson. Objects were scattered along the line of most intensive traffic and mapped. Changes in their horizontal distribution were recorded after three and six days. RESULTS In the following sections the results of the experiments are described and discussed considering separately four kinds of transformations that trampling can produce in the archaeological record: changes in the trampled substrate, in the vertical position of items, in their horizontal distribution, and various classes of damage. Trampling/Substrate Interaction The measurements taken before the first experiment showed that soil penetrability is extremely variable, even within small areas (TR-I before: mean = 2.49 kg/cm2, s.d. = .99, range = .75-4.5). It was expected that trampling would increase the compaction of the soil, reducing its penetrability. However, the measurements taken after treadage demonstrated that the area was 14 percent more penetrable (mean = 2.14 kg/cm2) and 17 percent more homogeneous (s.d. = .83). The erosion produced by the feet, enhanced by the presence of the artifacts as abrasive media, resulted in the formation of a loose, more penetrable top layer. The excavation revealed that this top layer was 12 cm thick and rested on an extremely compact and almost impenetrable level. All artifacts that migrated downward were included in the loose layer; none of them penetrated the second one. The same phenomenon was observed in the other experiments performed on dry soils, though it could not be quantified because the surfaces initially were so hard that they exceeded the range of the penetrometer scale (maximum 4.5 kg/cm2). To corroborate this point, observations were made on segments of intensively used paths on different types of sediments (including dirt sidewalks and gardens with high humus content) on the campus of the University of Arizona and across the city of Tucson. In all cases, a loose cover 415 mm thick overlaid an extremely compact layer. In three cases the loose material was brushed away over 1-x- 1-m areas, exposing the hard surface. After one week of treadage the loose cover had begun to develop again. Later, a heavy rain compacted this material, but the same process of loosening was again observed after three or four days.

488

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

Table 3. Number of Items of Each Kind of MaterialBuriedand on the Surfacein TR-I and TR-II.
Surface TR-I Bones Sherds Lithics TR-II Bones Sherds Lithics
a

Subsurface 45 109 96 250

Totala 93 184 123 400 33 214 70 317

48 75 27 150 22 179 36 237

(51.6%) (40.8%) (21.9%) (37.5%) (66.7%) (83.6%) (51.4%) (74.8%)

(48.4%) (59.2%) (78%) (62.5%)

11 (33.3%) 35 (16.4%) 34 (48.6%) 80 (25.2%)

The differencesbetweenthe numbersof items recoveredand those in the

original assemblages as described in Table 2 are due to the combined effects of loss and breakage.

The same phenomenon was recorded by Gifford-Gonzalez et al. (1985:808) in their experiment on loamy soil, and is familiar to ecologists (e.g., Liddle 1975; Weaver and Dale 1978) who conceive trampling both as an erosive process that increases the depth of the paths and as a compacting process that increases the bulk density of the soil near the surface. For a given trampling agent there exists a maximum stable compaction value that is a function of the microstructure of the soil. The loose cover is a more dynamic element that is likely to vary in thickness depending not only on the soil, but also on the intensity of treadage, slope, and patterns of rainfall. Trampling after rain (TR-III) had different effects. The muddy surface was initially very penetrable (mean = 1.63 kg/cm2; s.d. = .55), but doubled its compaction after treadage (mean = 3.25 kg/cm2; s.d. = .58). No loose layer developed, and very few artifacts were buried completely. Most of them were stuck in the soft substrate during the first few crossings, and remained in the same position throughout the experiment. At the end they still were visible from the surface. These various responses of the substrate are important for understanding many effects of trampling. They will be referred to while discussing particular aspects in the following sections. Vertical Displacement This dimension of trampling processes has received the most attention because it has implications for the interpretation of stratigraphic sequences and chronology. Most of the studies have been carried out on loose sandy soils, where artifacts from the same original assemblage have been recovered in levels separated up to 16 cm (Stockton 1973). In a more compact soil (loam) GiffordGonzalez et al. (1985) recorded 3 cm as the maximum downward movement, with 94 percent of the items found within the first centimeter. Existing studies are contradictory regarding the presence of a correlation between the size, weight, or density of the artifacts and their vertical migration. Villa and Courtin (1983:277) worked with different kinds of material and found no correlation between this variable and vertical migration. They only generalize that pieces lighter than 50 g may move, while heavier ones will tend to remain near the level where they were placed. Gifford-Gonzalez et al. (1985:811) report that "none of the attributes indexing size or volume yielded a significant correlation with depth below surface." On the other hand, Pintar (1987) obtained a significant correlation value (Spearman's rank coefficient = -.8) for size/vertical displacement, suggesting that smaller items tend to be more displaced downward. A similar correlation is apparent in Muckle's (1985:Table 16) trampling experiments with shell on a loam substrate. In an ethnoarchaeological context, Gifford (1978:82) previously had observed a tendency of smaller objects to be trapped in loose sand surfaces.

REPORTS

489

and Vertical Table 4. TR-II: T Tests for Length/Weight


Migration. Surface Mean All items Length Weight Bones Length Weight Sherds Length Weight Lithics Length Weight 34.6 4.9 54.7 3.9 30.9 5.2 34.6 4.3 s.d. 13.5 4.6 21.5 3.6 12.3 4.4 13.5 6.1 Subsurface Mean 23.5 1.9 26.3 .7 23.5 3.0 23.5 1.1 s.d. 9.4 2.5 9.6 .5 9.2 3.2 9.4 1.3 t value 3.99 7.48 5.24 4.02 4.09 3.50 3.99 3.05 p .000 .000 .000 .001 .000 .001 .000 .004

Note: All lengths are in millimeters and all weights are in grams.

The maximum vertical migration recorded during the experiments here reported was 1.5 cm.I Under dry conditions, this corresponds to the thickness of the loose top layer discussed in the previous section. No artifact penetrated into the hard-packed bottom one. It follows from this that (1) the proportion of buried items will covary with the thickness of the loose top layer, and (2) size sorting will occur, since only objects no thicker than the thickness of the top stratum can be buried. In TR-I, which was carried out on a more permeable substrate (2.49 kg/cm2) and trampled 1,500 times, the top level averaged 1.5 cm and contained 62.5 percent of the artifacts. In TR-II (>4.5 kg/cm2, 800 crossings), the loose layer did not exceed 1 cm and included only 25.2 percent of the recovered assemblage (Table 3). Size sorting is apparent when the proportions of buried and unburied pieces of each kind of material are considered. Lithics, which included smaller (Table 2) and flatter items than sherds and bones, were consistently buried in higher proportions. T tests comparing two indicators of sizelength and weight-for surface and subsurface sets from TR-II show that these differences are very significant (Table 4). No such sorting occurred after treadage on wet ground. The proportions of surface/subsurface artifacts seem to vary randomly across material type (Table 5). Moreover, neither length nor weight render statistically significant differences between objects recovered in various levels (Table 6). No loose cover is developed in this situation. The materials, rather, are pushed down by the feet and stuck in the permeable substrate. If the surface is penetrable enough (ca. 2 kg/cm2 or less), no sorting occurs. Trampling under these conditions will tend to fix in their initial horizontal location objects of all sizes except the very large ones. Eventually, once the soil dries and hardens, erosion will develop the loose top layer releasing some of the artifacts, and sorting will start again. These contrasting observations call attention to the different mechanisms of vertical displacement of artifacts trampled on wet and drysubstrates. Under dry conditions the artifacts tend to act as passive elements (Pryor 1988) that are covered by the loose dirt scuffed onto them by treadage. Since this loose top layer is very thin, size becomes a critical factor for the materials to be covered. Since in hard-packed surfaces vertical displacement does not exceed 1.5-2 cm, no serious stratigraphic disturbance or archaeologically recognizable sorting by size will occur. These patterns of vertical migration, however, limit the impact of other forms of disturbance on parts of the assemblage, since burial will drastically reduce the horizontal movement of the small items and will protect them from being removed during maintenance.

490

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

Table 5. Number of Items of Each Kind of MaterialBuriedand


on the Surface in TR-III. Surface Bones Sherds Lithics Total 7 (25%) 4 (17%) 8 (31%) 19 (25%) Semiburied3 13 (46%) 17 (74%) 16 (61%) 46 (60%) Subsurface 8 (29%) 2 (8%) 2 (8%) 12 (15%) Total 28 23 26 77

a Items that penetrated completely into the substrate but still were partially visible from the surface.

From an archaeological point of view two situations can be expected when dealing with hardpacked sediments as those analyzed in the present study: 1. If the surface was buried after a period of dry trampling (as can be assumed, for instance, in the case of roofed areas), the less-disturbed evidence will be found in a thin (20 mm at most), loose level overlaying a hard, compact, and probably sterile one (unless previous occupations exist in the site). Holding constant other factors, the artifacts recovered in that upper layer should be very small and could be considered primary refuse. 2. If the last trampling period took place under wet conditions, items of all sizes will be found embedded in a relatively hard layer. The previous discussion also illustrates the complexity of the formation of "living floors." The widely shared notion that intensively occupied surfaces are hard and highly compacted needs to be treated with caution. For instance, if a period of "dry trampling" preceded the burial of the surface, once the excavation reaches the hard "occupation floor," quite probably the most relevant behavioral evidence in the form of small remains and microartifacts already has been retrieved. Special techniques, like microarchaeological analysis (Hassan 1978; Rosen 1986; Stein and Teltser 1989), should be employed to recover this information. Horizontal Displacement Only two experimental studies have searched for patterns in the horizontal displacement of trampled artifacts. Villa and Courtin (1983:277) observed that "the most displaced pieces are light while the heavy pieces moved little" but "there is no obvious linear correlation between horizontal displacement and weight. . .thus weight is not a good predictor of displacement." Pintar (1987:1618) arrived at a similar conclusion, obtaining an inverse but nonsignificant correlation between length and horizontal migration of flakes.

Table 6. TR-IIl: T Tests for Length/Weight and Vertical Migration. Surface Mean Bones Length Weight Sherds Length Weight Lithics Length Weight 37.4 2.7 54.7 21.6 29.( 4.8 s.d. 8.8 1.8 29.1 21.7 14.9 11.2 Subsurface Mean 35.9 1.8 26.0 1.2 27.5 1.5 s.d. 19.0 2.0 7.1 1.2 16.3 1.9 t value .21 .98 1.87 .51 .12 .78 p .84 .34 .14 .15 .92 .46

REPORTS

491

Table 7. Mean Horizontal Migration of Materials in TR-II and TR-III. TR-II Bone Ceramics Lithics 78.1 cm (range 0-314) 41.0 cm (range 0-336) 23.9 cm (range 0-126) TR-III 52.7 cm (range 0-228) 11.9 cm (range 0-83) 19.2 cm (range 0-122)

The results of TR-II and TR-III concur in general terms with those of Villa and Courtin and Pintar. Although the correlations length/horizontal movement and weight/horizontal movement are positive in all the cases, they are not at all significant (range of r values = .0884-.5545). TR-III produced similar results. The only observed difference is that materials moved less than in the experiments performed on dry surfaces because they were "trapped" in the substrate from the beginning of the process. However, when the different materials are compared in terms of their mean horizontal migration some trends arise (Table 7). Bones moved more than lithics, whereas, at least in the dry-trampling case, ceramics had an intermediate response. Three factors could account for these results: density, size, and shape. Denser materials-like lithics-may have moved less because, holding size constant, they weigh more. It also could be argued that although length is not a predictor of horizontal migration, there is still a weak positive correlation between both variables. Since bone assemblages included more large pieces than lithic and ceramic ones, differences in size still could be responsible for the differences in mean displacement. Shape is a third variable that may be reflected in these figures. Three of the more displaced bones were vertebrae which, by their very shape, are more likely to be kicked away than flat elements like sherds or flakes. In fact, size and shape are better conceived as a single attribute, that can be referred to as "bulk," which determines the probability of an object being kicked or scuffed by human traffic. TR-VI was designed to examine the relative influence of these variables on horizontal migration. An assemblage with equal numbers of prismatic fragments of oak wood (.59 g/cm3) and brick (1.84 g/cm3) distributed in three size categories was used. The sizes were "large" (.57 x .46 x .29 cm = 73.4 cm3), "medium" (.45 x .28 x .14 cm = 17.6 cm3), and "small" (.30 x .19 x .11 cm = 6.3 cm3). Each one included eight pieces of each material. Materials and sizes were chosen to maximize contrasts in density and bulk. Sixteen small sherds (2.6-4.6 cm3, maximum length = 32 mm) were included to facilitate comparisons with the assemblages used in previous cases. All pieces were scattered along a heavily used dirt path in a park in Tucson. The mean displacement recorded after three days is shown in Table 8. As noticed previously, denser materials tend to move less when size and shape are held constant. A t test run between wood and brick fragments of all sizes indicates that the differences in horizontal movement between both materials are significant (t = 1.90; df = 32; one-tail p < .05).

Table 8. TR-VI: Mean Horizontal Displacement of Items after Three Days of Trampling. Wood Large (N = 8) Medium (N = 16) Small (N= 16) All sizes s.d. Note: All measurements in cm. 259 129 111 158 192 Brick 69 47 69 59 101

492

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

12

6 days

10-

D 3 days

E
0

Q)

8-

E
:3

4.6 cm3 or less

6.3 cm3

17.6 cm3

73.4 cm3

Size
Figure 1. TR-VI: Number of pieces of various sizes cleaned up after three and six days.

Size, on the other hand, shows no statisticallysignificantcorrelationwith the amount of displacement, whether wood and brick are considered together or separately. Sherds were not included in the calculations because their mean movement was only 7 cm and therefore, they would have skewed the results. There is another way, however, of approaching the relation between bulk and horizontal migration. Wilk and Schiffer (1979:533), for instance, based on vacant-lot data, have observed that "[l]arge objects (ca. 50 cm3) do not stay for long on paths. They are kicked or moved aside." The results of TR-VI support this statement. After three days only one large piece was found in the path, whereas nine medium and 11 small ones were still in the line of most-intense traffic. It follows that the relation between size and displacement is better represented by a model consisting of a series of size thresholds that determine qualitative changes in the response of objects, rather than by a linear model. In other words, when trampled on unpenetrable surfaces (>2 kg/m2), objects follow at least three different patterns of horizontal migration according to their size. Very small items (<2 cm) are trodden readily into the loose sediments that cover hard-packed substrates when they are exposed to trampling erosion. As a result, they drastically reduce their horizontal movement. Therefore, if other factors are held constant, very small items will be found close to their original place of deposition. Small and medium size objects (up to approximately 50 cm3) remain on the surface where treadage displaces them at random. Bigger and less dense items (e.g., bones) move more, but no sorting along this variables is likely to arise. Any spatial pattern existing at the moment of their deposition will be obscured rapidly. Eventually, their random movement will place these objects outside the zone of most intense traffic, where they will remain stationary if they are not transported by other processes. If trampling continues long enough these items will accumulate in these marginal zones that can be sides of paths, corners, and sectors along the walls in enclosed spaces, or around large objects or features like storage vessels, benches, looms, etc. (a phenomenon that has been called "fringe effect" [Wilk and Schiffer 1979:533]). Large, bulky objects (>50 cm3) are kicked and scuffed rather than trodden and therefore will move faster and more systematically to stable positions in the "marginal zone." Less dense elements will tend to move farther, but again, a horizontal sorting by density is not likely to occur. TR-VI also serves to illustrate how maintenance, by acting selectively upon size, can modify

REPORTS

493

Table 9. BreakageIndex (bx) for Nine CeramicAssemblagesAfter DifferentAmounts of


Trampling. Number of Crossings 50 Type Type Type Type Type Type Type Type Type A TR-I A TR-II B TR-V C TR-I C TR-II D TR-II D TR-III D TR-IV E TR-V 100 200 300 400 800 3.46 4.00 1.22 1.23 1.36 2.05 2.89 1.45 1.11 1,500 1.84 Cumulative Index 1.84 3.46 6.00 1.36 2.05 2.89 1.45 3.18 2.53

1.71

2.01 1.10

1.17

1.08

1.24 1.34

trampling patterns. After three days, seven of the large pieces (both wood and brick) had been removed. Three days later fragments of all sizes had been cleaned up, but still the larger ones were Thremoved. the most affected. On the other hand, none of the sherds (all of them smaller than 4.6 cm3) were missing (Figure 1). In accordance with the McKellar Principle (Schiffer 1987:62), these observations show that smaller items are left behind in regularly maintained areas. From the point of view of spatial analysis, they imply that maintenance eliminates part of the "noise" that trampling introduces in depositional patterns, since bigger, probably more displaced objects, are more likely to be removed. Damage Different sorts of damage affect each material according to its physical properties. The present section focuses on ceramics and lithics. Bones were only abraded and will not be considered here. For a discussion of trampling marks on bone see Behrensmeyer et al. (1986) and Olsen and Shipman (1988). Ceramics. Sherds showed various abrasion traces (Schiffer and Skibo 1989), such as a slight rounding of edges, and in few cases microchipping and delamination, especially along the edges of polished surfaces. But breakage is certainly the most obvious kind of damage. To facilitate comparisons among assemblages, a breakage index (bx = number of fragments after selected trampling episode divided by number of sherds before that episode) was calculated for each type of pottery in each experiment. The results are displayed in Table 9. The differences in fragmentation for the same type in different experiments indicates the critical role of surface hardness in the process of fracture. The three assemblages trampled on relatively penetrable soils (A and C TR-I, D TR-III) had a breakage index lower than two, even though the sherds of TR-I were trampled twice as much; the rest of the assemblages (that were trodden on
surfaces harder than 4.5 kg/cm2) exceeded this value.

Another trend reflected in these figures is the decreased fracture rate after the first few crossings, showing how the reduction in size increases the strength of the sherds (cf. Kirkby and Kirkby 1976: 237).2 Eventually a stable size where no further breakage occurs would be reached. This value would be a function of the microstructure of the paste, sherd thickness and curvature, and the nature (weight and contact surface) of the trampling agent. It follows that after a few crossings sherd size should be unimodal, distributed around a value that, when reached, would effect a significant increase in breakage resistance. Untrampled assemblages, on the other hand, would present a random distribution of sizes produced by the original fracture of the vessels. Further trampling would result in a slow reduction of the modal value and an increasing positive skewness of the whole distribution. When the modal value reaches the smallest size category, the whole curve would approximate a Poisson distribution. If this proposition is correct, it could be a useful device for recognizing archaeologically trampled assemblages and perhaps even for determining the relative amount of treadage that occurred.

494

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

50 *

TYPEC TR-I N=28 TYPEA TR-I N=80

40 -

30-

....
//

/ -

.......

-,

\N=

TYPE

TR-ii

20-

//

..?' A-Y^

"

,,~

.....

'

,.

"

/-

--

"*<

- /v^^^< -

...****

TYPE A TR-i

^\

*%..:.===..=

N=35
'
|

I0 |L-.---......X-.-..-

TYPED TR-11 N=22

size
2 = 21-30 mm; 3 = 31-40 mm; 4
=

categories

of Figure2. Size distribution six ceramicassemblagesbeforetrampling.Size categoriesare: 1 = 11-20 mm;


41-50 mm; 5 = 51-60 mm; 6 = 61-70 mm; 7 = 71 mm or more.

In order to test this hypothesis, the size frequency distributions for six ceramic assemblages before and after trampling were calculated and put in graphic form. Figure 2 shows the random distribution of size expected for untrampled assemblages, with frequent bimodality and large sherds consistently represented in most cases. It should be noted that none of these curves reflects the distribution of a "naturally broken" pot, the sherds of these six assemblages were chosen arbitrarily.
1 60 ............ ..\ 2... 12 TYPE C TR-I N=38 TYPEA TR-1

50 -

10

\N=41

Zo ~\ /t~~~~~~// \~

"

~TYPE

D TR-II

< \ 0 1 ' i 2 l 3

\i

= L a,^--.^ "^***^^.f--7?'"r> 5 4

< \^ 6

. ^*-sr:

^^7

TYPED TR-lii l N=33

size categories caotegories


Figure 3. Size distribution of six ceramic assemblages after trampling. Size categories are: 1 = 11-20 mm; 2 = 21-30 mm; 3 = 31-40 mm; 4 = 41-50 mm; 5 = 51-60 mm; 6 = 61-70 mm; 7 = 71 mm or more.

REPORTS

495

50

0 cross 100 cross

40
-,

---_-

.......

;;*,
'" "^"\

200 cross
300 cross

3..0

400 cross
20-

800 cross

size

categories

Figure 4. TR-IV: Progressive reduction in size of one ceramic assemblage (type D) as a result of trampling. Size categories are: 1 = 11-20 mm; 2 = 21-30 mm; 3 = 31-40 mm; 4 = 41-50 mm; 5 = 51-60 mm; 6 = 61-70 mm; 7 = 71 mm or more. After trampling (Figure 3) the least fractured assemblages (Type C TR-II bx = 2.050; Type D TR-II bx = 2.889; Type D TR-III bx = 1.454) show an unimodal curve with the mode in category 3, while the most reduced ones approximate a Poisson distribution. The "abnormally skewed" curves of types A TR-I and C TR-I, considering their relatively small breakage index (1.837 and 1.357), are explained readily by their originally skewed distribution which, if types B, E (TR-V) and

D (TR-IV) are assumed to be representative cases, are very unlikely to occur in "naturally" broken
pots. Furthermore, if the areas under analyss contain secondary refuse, their original size distribution

will tend to be skewed negatively, provided that smaller objects are more likely to escape maintenance activities and be left behind in original activity areas. This would provide an even stronger contrast between trampled and untrampled assemblages in secondary refuse.
To determine how much trampling is necessary for the size distribution to adopt each shape, Figures 4-6 wre constructed. In these cases, the curves drawn as solid lines do reflect the size of sherds produced by the initial breakage of whole pots. In TR-IV, after just 100 crossings only one

sherd larger than 71 mm is left (which remains unbroken throughout the process), and the general curve already shows the characteristic configuration. Changes after this first stage are much more gradual. This process is even clearer for type B TR-V (Figure 5) in which no sherd exceeds 45 mm in length after 50 crossings. As noted previously, type E was a high-fired flower pot that had the hardest paste among the vessels used. Consequently, it experienced the slowest size reduction (Figure 6). Several fragments corresponding to the reinforced rim of the vessel remained relatively large. Unlike type B, which was exposed to the same trampling conditions, the mode of the distribution for type E did not reach category 1 by the end of the process. The response of this type suggests that particularly strong wares-beyond the range considered in this study-probably will depart from the trends described thus far. Fragments of storage vessels with extremely thick walls or ceramics fired at exceptionally high temperatures can be so strong that the stress of human trampling may not be enough to effect a significant amount of fracture. The preceding observations can be summarized as follows: 1. Holding constant other factors, it can be inferred that a sherd sample has been trampled if its size distribution is found to be unimodal, with the mode lower than 30 mm, and with no fragments

496

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

70 60 50 40 30 20 10 0 1 2 3 4 5 6 7

0 cross 50 cross 100 cross 400 cross


-

size categories
Figure 5. TR-V: Progressive reduction in size of one ceramic assemblage (type B) as a result of trampling. Size categories are: 1 = 11-20 mm; 2 = 21-30 mm; 3 = 31-40 mm; 4 = 41-50 mm; 5 = 51-60 mm; 6 = 61-70 mm; 7 = 71 mm or more.

larger than 50 mm in length or just very few corresponding to especially strong parts of the vessels (like the articulation of the body and the base). 2. If the penetrability of the soil, the nature of the trampling agents, and the strength of the ceramic material can be assumed approximately constant across samples (i.e., they show a similar range of internal variability), and no other cultural formation processes are acting upon this dimension of

50

0 cross 50 cross

40 100 cross 30 400 cross


-

20

10

size categories
Figure 6. TR-V: Progressive reduction in size of one ceramic assemblage (type E) as a result of trampling. Size categories are: 1 = 11-20 mm; 2 = 21-30 mm; 3 = 31-40 mm; 4 = 41-50 mm; 5 = 51-60 mm; 6 = 61-70 mm; 7 = 71 mm or more.

REPORTS

497

40 - k ~ /' *\-

0+100 0+300 0+800

cr. cr. cr.

10

'I

size
Figure 7.
Size categories are: 1 = 11-20 mm; 2 = 21-30

categories
mm; 3 = 31-40 mm; 4 = 41-50 mm; 5 = 51-60 mm; 6 = 61-70

Size distribution of three hypothetical assemblages combining trampled and untrampled sherds.

mm; 7 = 71 mm or more.
the material, then the degree of positive skewness of the distribution is a relative indicator of the

amount of trampling undergone by different assemblages. Two objections could be raised against these generalizations.
are mixed refuse

First, if "freshly broken" assemblages

if a path crosses a secondary with previously ones, as may happen, for instance, trampled to detect Would it be possible above still be recognizable? area, would the patterns described

the presence of both kinds of artifacts in the mixed assemblage? To answer these questions, the figures for some trampled and untrampled
were combined in three

sets chosen at random

crossings; type A TR-II 0 + type D TR-IV 300 ways (type D TR-IV 0+100 in D TR-IV and type D TR-II 0+type 800 crossings) and the distributions represented crossings; form for the resulting "mixed graphic assemblages" (Figure 7). As can be seen, the "trampling still are perfectly visible. The slightly "high" proportion of large pieces shown in the curve patterns" as a solid line could serve as an indicator of the presence of the untrampled set, if these sherds do

not have special attributes that would give them higher breakage resistance. Another process that could produce a similarly skewed size distribution
maintenance (Schiffer fragments 1987:64), corresponding since the biggest artifacts parts would

in the assemblages
and discarded

is
in in

be retrieved that should

secondary
a few large

refuse areas. Two alternatives

exist to distinguish Second,

both processes.

First, after trampling,


not be present

to stronger

of the vessels

rdual refusesidual primary primarprobably

will remain.

in trampled

assemblages

not subjected

to

one should find a consistent of big pieces of other kinds of materials more maintenance, proportion resistant to fracture (e.g., bone, lithics). It should be recalled, that these large items may however, have migrated to the margins of the areas of most intense traffic (see section on horizontal displaceabove). A second possible objection is random, it is possible-though for trampled point. would ment stems from the fact that, if the size distribution for untrampled sherds such distributions fall within the specifications represented by the solid line in Figure 2, would be a case in if such a distribution was found it

improbable-that

material. Type C (TR-I), it does not represent any "natural breakage," Although be interpreted as a trampled erroneously assemblage.

To provide an additional control, it was postulated that there should be a correlation between the mean size of the sherds and their size variability (as measured by the standard deviation of the

498

ANTIQUITY AMERICAN

[Vol. 56, No. 3, 1991]

30

25 0 ._ 20-

25A

before trampling

after trampling

> q)

10-

5-

10

20

30

40

50

60

70

Mean Size (maximum

length)

Figure 8. Size and size variability for eight ceramic assemblages before and after different amounts of trampling. Before trampling, r2 = .2438; after trampling, r2 = .8576; a = -3.2830; and b = .4686.

set) through successive stages of reduction. Figure 8 displays a scatter plot for these two variables in all the assemblages before (nine samples) and after different amounts of treadage (16 samples). the As predicted, the points representing the sets before reduction are dispersed through diagram (Pearson's r = .4938), while those corresponding to trampled assemblages are aligned in a regular pattern (r = .9271) indicated by the regression line in the graph. The trajectory of individual assemblages through increasing treadage show even higher correlations (r = .99 for types D TR-IV and B TR-V). Certainly, the general validity of this pattern should be tested and, eventually, readjusted using a larger sample. However, considering that the materials used were extremely heterogenous, it is likely that the pattern will hold in any archaeological situation regardless of the internal variability of the ceramic material. Therefore, a second procedure for differentiating trampled and untrampled samples can be provisionally postulated. If a given ceramic assemblage has been trampled, the mathematical expression of the regression shown in Figure 8 (y = bx + a) should predict its standard deviation from its mean size. In other words, the expression S = .4686 X - 3.283, where 5S= standard deviation of length, and X = mean length, should hold with a margin of error of ? 1.19 (measurements taken in mm), corresponding to the standard error of the regression line. The range of values predicted

Table 10. Predictedand Actual StandardDeviations of Length


for Nine Ceramic Assemblages Before Trampling. Assemblage A TR-I A TR-II B TR-V C TR-I C TR-II D TR-II D TR-III D TR-IV E TR-V Mean (mm) 29.59 41.63 42.29 33.86 54.00 59.33 41.36 40.86 45.73 Predicted S 9.39-11.77 15.03-17.41 15.34-17.72 11.39-13.77 20.83-23.21 23.33-25.71 14.91-17.29 14.67-17.05 16.96-19.34 Actual S 12.28 18.79 25.60 9.08 13.76 23.85 18.58 21.16 24.09

REPORTS

499

25

3 ass. trampled
1 ass. trmp. + 1 untrmp.
A

20 ?
C
-

qa 15C)
^ -U *e

'>

A *....o ---'""

o ,,-o

3 ass. untrampled

-~0

5-^

-6^

15

20

25

30

35

40

45

Mean Size

(maximum

length)

Figure 9. Size and size variability for three hypothetical ceramic assemblages combining trampled and untrampled sherds. The line represents the pattern obtained from the eight assemblages with various amounts of trampling (small circles) (see Figure 8). by this procedure and the actual ones for the standard deviation of the nine assemblages used in

this study before trampling are displayed in Table 10. Most values of S fall, as expected, out of the predicted range for trampled assemblages, including C TR-I that could have been mistaken as
trampled following the size-distribution procedure. The only exception is D TR-II, that, in any

event, can never be mistaken as trodden if its size distribution-35 71 mm-is considered (Figure 2, long dashed line).

percent of the sherds larger than

procedure for detecting assemblages Finally, Figure 9 shows the ability of the size-variability containing various types of untrampled material and for classifying correctly other hypothetical "mixed" samples. The line represents the pattern obtained on the basis of all trampled sets (small

circles). As expected, the solid circle (three trampled sets added) falls close to the line, within the range of variability predicted for trodden materials. The combination of one trampled and one
untrampled assemblages (solid triangle) is situated together fall even farther away from the line. out of this range, and three untrampled sets

Consequently, it is suggested that the application of both procedures, size distribution and size variability, by using two different dimensions of the data, can discriminate with a high degree of confidence between trampled and untrampled sherd samples, and even establish relative amounts of treadage (or activity) on the material if certain conditions are met. Lithics. The three kinds of damage considered here (breakage, microflaking, and abrasion) were observed on lithics. Table 11 shows the number of pieces representing the first two alterations in

Table 11. Number of Flakes Showing Breakage and Microflaking. TR-I Broken Microflaked Broken and microflaked Undamaged Total 29 (24.8%) 31 (26.5%) 11 (9.4%) 46 (39.3%) 117 TR-II 17 (19.8%) 32 (37.2%) 25 (29.7%) 12 (13.9%) 86 TR-III 5 (19.0%) 6 (23.0%) 2 (8.0%) 13 (50.0%) 26

500

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

TR-I, TR-II, and TR-III. As noted by Gifford-Gonzalez et al. (1985:813), breakage is more frequent on harder surfaces (TR-II). Artifacts in TR-II were more damaged even though artifacts in TR-I were trampled twice as much. Lithics trampled on a wet substrate (the most permeable) were the least damaged. Abrasion was especially severe on prominent parts of the flakes, such as percussion bulbs and dorsal ridges. There are considerable differences in the literature concerning the type of edge damage produced by trampling. Working with obsidian flakes, Tringham et al. (1974:113), who performed the first experiment on this subject, established three criteria for differentiating trampling from use damage. The scars are randomly distributed around the perimeter of the flakes; they occur only on the surface opposite to the trampler; and they lack fixed orientation or size, but are characterized by marked elongation. A later experiment carried out by Flenniken and Haggerty (1979) contradicts these criteria. They found that out of 157 flakes (37 percent) that underwent edge modification during treadage, 56 (13 percent) could be classified as "tools," and their edges were remarkably similar to used ones. They pointed to the absence of polish as the only definitive indicator of trampling damage as opposed to use. On the other hand, Gifford-Gonzalez et al.'s (1985) and Pryor's (1988) studies agree with Tringham et al. concerning the sparseness of the scars along the edges, but these scars were not elongated and originated on both surfaces of the artifacts. The results of the experiments reported here are in general agreement with the conclusions of the latter authors. Most pieces show one to three isolated scars randomly distributed along the edges, regardless of their angle. They originate on either surface and no distinctive shape or size could be identified, except for a trend of larger scars to occur on steeper edges. However, six or seven pieces from the dry-trampled assemblages depart from this general trend. They show rows of continuous parallel scars along one or more edges that could be mistaken easily for intentional retouch. CONCLUSIONS Several transformations that occur in spatial and formal attributes of materials exposed to trampling have been discussed separately in previous sections. These observations are now integrated to outline sets of traces in the archaeological record that can help to identify trampled contexts. Of course, the applicability of these generalizations is restricted to situations where the relevant conditions are comparable to those considered in the present study. Minimally, these conditions are: similar materials in terms of size, density, and fracture properties, and trampling by humans on hard-packed substrates (ca. 2 kg/m2 penetrability or more when dry). A small amount of trampling will cause the migration of bulky items to the margins of the trampled area where they will stay stationary unless affected by other factors. Small and medium size objects will move randomly within the traffic zone, blurring previous patterns in their horizontal area. Only very small items will remain trampled arrangement that might have existed within the In other words, even moderately substrate. the in their original location by being "absorbed" in trampled areas will be composed of a "marginal zone" characterized by a high proportion of bulky artifacts, and a "traffic zone" with small- and medium-size items randomly scattered and very small ones buried close to their original spot of deposition. In this initial stage the sherds already will exhibit the typical relation between mean and standard deviation of size and will adopt a size frequency distribution that resembles a bell-shaped curve or a Poisson distribution, depending on the strength of the paste, the thickness of the sherds, and their curvature. Few lithics will break and present isolated and randomly distributed flake scars along their edges. A few damaged edges may mimic retouching. If trampling continues and the area is not cleaned, the original pattern, preserved by the small pieces initially trapped in the substrate, will be obscured increasingly by the "absorption" of new small pieces produced by the fracture of objects after having been displaced horizontally. On the other hand, medium-size items that are unlikely to be trapped in the substrate will reach gradually
stable positions in the "marginal zone."

REPORTS

501

Thus, while original "fine-grained" horizontal configurations within the traffic zone will be no longer recognizable, the contrast between traffic and marginal zones will be stronger. The former will be characterized by a high frequency of small artifacts and microartifacts randomly scattered, low proportion of medium-size items, and virtually no bulky ones (cf. O'Connell 1987:95). The latter will have high frequencies of artifacts in large size categories and very few in the small ones. All of them will be displaced far from their original locations. The substrate of the traffic zone will consist of a top loose layer (5-20 mm thick) containing many small artifacts and microartifacts and an extremely compacted, sterile bottom layer. The latter usually is identified during excavation and can orient the application of microarchaeological techniques to recover the former. Differences in penetrability between traffic and marginal zones also can be recognized in the archaeological record (see Koike 1987). In addition, the whole assemblage should show severe damage--randomly scarred edges and abraded ridges in most flakes, and rounded, microchipped, and delaminated edges in sherds. The size frequency distribution of sherds should be extremely skewed as well. When applying these generalizations to archaeological cases it should be kept in mind that every deposit is the result of multiple formation processes, including human activities and the action of environmental factors. Sometimes the material effects of these processes overlap, this is, different processes can produce similar traces. For instance, ethnoarchaeological studies have demonstrated that horizontal size sorting also can result from the spatial organization of activities themselves and corresponding disposal modes (Binford 1978; O'Connell 1987), or cleaning (DeBoer and Lathrap 1979; Simms 1988). When inferring trampling in archaeological cases, therefore, one should consider as many traces as possible, as well as relevant independent data, in order to differentiate it from other processes that may have acted upon the materials and generated similar patterns. On the other hand, the interaction with other formation processes can modify the effects of trampling itself. For example, as has repeatedly been demonstrated, maintenance operates selectively on larger items. If an intensively trampled area is frequently cleaned, clear depositional patterns are likely to be preserved in the distribution of small artifacts in the traffic zone. Bigger objects, therefore more displaced, will be cleaned up systematically and will not remain long enough to be randomly displaced and fractured in different locations, contributing additional small artifacts that would obscure existing patterns reflected in this size fraction. On the other hand, in rarely maintained areas the contrast between marginal and traffic zones will be stronger, and damage will be more severe because artifacts will be exposed longer to treadage. Differences in environmental conditions, or in exposure to such conditions, can effect variations in these patterns as well. Rain, for instance, generates a muddy and very penetrable surface that will "trap" all artifacts regardless of their size. This will prevent the objects from moving horizontally and will reduce damage. However, once the surface dries and treadage erosion generates the loose cover again, larger items will be released, start moving horizontally, and be exposed to retrieval during cleaning activities, while smaller ones will remain embedded in the top layer. If rains are very frequent and evenly distributed throughout the year, very little horizontal movement (and postdepositional patterning) will take place in unroofed trampled areas. Even in spaces regularly cleaned, compact occupation floors will be found with artifacts of various sizes embedded and damage will be less severe. On the other hand, if the amount of rainfall is small and unevenly distributed during the year (as is often the case in semiarid environments), trampling patterns will be much clearer, more so if the surfaces were buried or abandoned after a long period of "dry treadage." Future studies can define variations in the patterns discussed above when treadage occurs under different conditions. This "basic" understanding gained through experimentation, together with relevant ethnoarchaeological observations, can provide criteria applicable to the archaeological identification of trampling, and hence contribute to assess the potential of particular contexts for various kinds of behavioral inference. Acknowledgments. The researchreported on here has been supportedby the Laboratoryof Traditional
Technology, Department of Anthropology, University of Arizona, Tucson. I want to express my gratitude to

502

AMERICANANTIQUITY

[Vol. 56, No. 3, 1991]

several persons that help me during the work. Michael Schiffer gave me access to the lab and made insightful comments on drafts of the paper. William Walker, James Spicer, and Nieves Zedeniiolent their feet and shared generously their time and ideas. James Skibo, Chuck Bollong, William Walker, and the editorial staff ofAmerican Antiquity helped to make several points clearer in the text. I also benefited from the comments of Steven Simms, Duncan Metcalfe, Harold Hietala, and an anonymous reviewer. Finally, I am indebted to the Fulbright Commission for their support.

REFERENCES

CITED

Behrensmeyer, A., K. Gordon, and G. Yanagi 1986 Trampling as a Cause of Bone Surface Damage and Pseudo-cutmarks. Nature 319:768-771. Binford, L. R. 1978 Dimensional Analysis of Behavior and Site Structure: Learning from an Eskimo Hunting Stand. American Antiquity 43:330-361. Black, C. A. (editor) 1965 Methods of Soil Analysis. Part 1:.Physical and Mineralogical Properties, Including Statistics of Measurement and Sampling. American Society of Agronomy, Madison. Bradford, J. 1986 Penetrability. In Methods of Soil Analysis. Part 1.:Physical and Mineralogical Methods. 2nd ed., edited by A. Klute, pp. 463-477. American Society of Agronomy, Madison. Bradley, R., and M. Fulford 1980 Sherd Size in the Analysis of Occupation Debris. Institute of Archaeology, University of London, Bulletin 17:85-94. Clarke, D. 1977 Spatial Archaeology. Academic Press, New York. Courtin, J., and P. Villa 1982 Une experience de pietinement. Bulletin de la Societe Prehistorique Francaise 79:117-123. De Boer, W., and D. Lathrap 1979 The Making and Breaking of Shipibo-Conibo Ceramics. In Ethnoarchaeology: Implications of Ethnographyfor Archaeology, edited by C. Kramer, pp. 103-138. Columbia University Press, New York. Flenniken, J., and J. Haggerty 1979 Trampling as an Agency in the Formation of Edge Damage: An Experiment in Lithic Technology. Northwest Anthropological Research Notes 13:208-214. Folk, R. 1980 Petrology of Sedimentary Rocks. Hemphill, Austin. Gifford, D. 1978 Ethnoarchaeological Observations of Natural Processes Affecting Cultural Materials. In Explorations in Ethnoarchaeology, edited by R. Gould, pp. 77-101. University of New Mexico Press, Albuquerque. Gifford, D., and A. Behrensmeyer 1977 Observed Formation and Burial of a Recent Human Occupation Site in Kenya. Quaternary Research 8:245-266. Gifford-Gonzalez, D., D. Damrosch, J. Pryor, and R. Thunen 1985 The Third Dimension in Site Structure: An Experiment in Trampling and Vertical Dispersal. American Antiquity 50:803-818. Hassan, F. 1978 Sediments in Archaeology: Methods and Implications for Paleoenvironmental and Cultural Analysis. Journal of Field Archaeology 5:197-213. Hughes, P., and R. Lampert 1977 Occupational Disturbance and Types of Archaeological Deposits. Journal of Archaeological Science 4: 135-140. Kirkby, A., and M. Kirkby 1976 Geomorphic Processes and the Surface Survey of Archaeological Sites in Semi-arid Areas. InGeoarchaeology, edited by D. Davison and M. Shackley, pp. 229-253. Duckworth, London. Koike, H. 1987 Measurement of Soil Hardness of Floor Surface for a Reconstruction of Activity Patterns in a Prehistoric Dwelling. Asian Perspectives XXVII:5-13. Liddle, M. 1975 A Selective Review of the Ecological Effects of Human Trampling on Natural Ecosystems. Biological Conservation 7:17-36. Lindauer, O., and J. Kisselburg 1981 Primary and Secondary Breakage. Paper presented at the 46th Annual Meeting of the Society for American Archaeology, San Diego.

REPORTS

503

Muckle, R. 1985 Archaeological Considerations of Bivalve Shell Taphonomy. Unpublished Master's thesis, Department of Anthropology, Simon Fraser University, Vancouver. O'Connell, J. F. 1987 Alyawara Site Structure and its Archaeological Implications. American Antiquity 52:74-108. Olsen, S., and P. Shipman 1988 Surface Modification on Bone: Trampling Versus Butchery. Journal of Archaeological Science 15:535553. Pintar, E. 1987 Controles experimentales de desplazamientos y alteraciones de artefactos liticos en sedimentos arenosos: Aplicaciones arqueol6gicas. Unpublished Tesis de Licenciatura, Facultad de Filosofia y Letras, Universidad Nacional de Buenos Aires, Buenos Aires. Pryor, J. 1988 The Effects of Human Trample [sic]. Damage on Lithics: A Consideration of Crucial Variables. Lithic Technology 17:45-50. Rosen, A. 1986 Cities of Clay. The Geoarchaeology of Tells. University of Chicago Press, Chicago. 1989 Ancient Towns and City Sites: A View from the Microscope. American Antiquity 54:564-578. Schiffer, M. 1983 Toward the Identification of Formation Processes. American Antiquity 48:675-706. 1987 Formation Processes of the Archaeological Record. University of New Mexico Press, Albuquerque. Schiffer, M., and J. Skibo 1989 A Provisional Theory of Ceramic Abrasion. American Anthropologist 91:101-115. Simms, S. R. 1988 The Archaeological Structure of a Bedouin Camp. Journal of Archaeological Science 15:197-211. Stein, J., and P. Teltser 1989 Size Distributions of Artifact Classes: Combining Macro- and Micro-Fractions. Geoarchaeology: An International Journal 4:1-39. Stockton, E. 1973 Shaw's Creek Shelter: Human Displacement of Artifacts and Its Significance. Mankind 9:112-117. Tringham, R., G. Cooper, G. Odell, B. Voytek, and A. Whitman 1974 Experimentation in the Formation of Edge Damage: A New Approach to Lithic Analysis. Journal of Field Archaeology 1:171-196. Villa, P., and J. Courtin 1983 The Interpretation of Stratified Sites: A View From Underground. Journal of Archaeological Science 10:267-281. Weaver, T., and D. Dale 1978 Trampling Effects of Hikers, Motorcycles and Horses in Meadows and Forests. Journal of Applied Ecology 15:451-457. Whittlesey, S. M., E. J. Arnould, and W. E. Reynolds 1982 Archaeological Sediments: Discourse, Experiment, and Application. In Multidisciplinary Research at Grasshopper Pueblo, Arizona, edited by W. A. Longacre, S. J. Holbrook, and M. W. Graves, pp. 28-35. Anthropological Papers No. 40. University of Arizona, Tucson. Wilk, R., and M. Schiffer 1979 The Archaeology of Vacant Lots in Tucson, Arizona. American Antiquity 44:530-536. NOTES was not considered appropriate to assess relative amounts of vertical displacement within such a small range. As a consequence, when analyzing its relation with artifact attributes in the following discussion, vertical displacement is treated as a categorical variable (surface, semiburied, subsurface). 2 The evolution of type D in TR-IV illustrates the influence of the development of the loose top layer previously described on fragmentation rates. From the second round on, the breakage index declines regularly throughout the process except in the third round (200-300 crossings) when the value falls below the trend (bx = 1.082, as compared to 1.175 in the second round and 1.124 in the fourth). During this experiment each trampling round was carried out on a different spot to eliminate variations in breakage resulting from alterations in the penetrability of the soil. The only exception was the third round, in which the artifacts were placed on the same spot as in the second one. The more penetrable character of the previously disturbed surface clearly was reflected in the drop of the breakage index.
1 It

Received July 17, 1990; accepted March 5, 1991

Das könnte Ihnen auch gefallen