Headroom and human trampling: cave ceiling-height determines the spatial patterning of stone artefacts at Petzkes Cave, northern New South Wales.

Theunissen, Robert ; Balme, Jane ; Beck, Wendy 等

Going into a cave or shelter, one walks where one can stand upright or has to crouch less. That affects which zones objects are trampled on, which zones they may be kicked out of, which zones they may be kicked into. And those effects interact with the usual spatial order - with its activity zones and drop zones - that develops through occupation of the enclosed cave or shelter.

As has been widely recognized (for example, by Schiffer 1983; Hivernal & Hodder 1984; Nash & Petragila 1987; Frankel 1989; Holdaway & Irwin 1993), identification of activity areas from maps of spatial artefact distribution is difficult because materials at archaeological sites are rarely recovered from their original position. Disturbance of the materials through postdepositional processes may alter the original spatial distribution and deliberate site maintenance at the time of occupation may also move artefacts horizontally. Human trampling, animal activity, water and wind disturbance are the main post-depositional processes, with human trampling regarded as a major disturbance agent in sandy rock-shelter deposits in Australia and elsewhere (Stockton 1973; Hughes & Lamperr 1977; Villa & Courtin 1983).

This paper investigates the effects of human trampling on the horizontal distribution of stone artefacts at Petzkes Cave, a large sandstone rockshelter in northern New South Wales. Stone artefacts were chosen for study; less subject to animal activity than organic artefacts, they are resistant to decay. Petzkes Cave [ILLUSTRATION FOR FIGURE 1 OMITTED]) was chosen for excavation because it is a large (25 m long) and relatively deep (9 m) rock-shelter. Its size makes it likely that a variety of activities took place at the site.

In this paper we find that trampling effects on the artefact distribution at Petzkes Cave can be predicted from ceiling-height. This allows us to focus on the artefacts and the cave zones which may be least affected by trampling, when investigating the original areas of activity. We predicted a generalized artefact distribution from previous work, and refined it for Petzkes Cave by a trampling experiment. When this 'hypothetical' distribution was compared with the actual distribution of artefacts in the top 3 cm of the deposit, we found good agreement between the expected and observed distribution of artefacts, suggesting that trampling was a major agent of disturbance.

Predicting the distribution of artefacts after trampling

Existing models for the effects of human trampling

Human trampling has long been recognized as a significant source of post-depositional disturbance in archaeological sites, especially in periodically re-occupied sites such as caves and rock-shelters (Hughes & Lampert 1977). In some caves, particularly sandstone caves such as Petzkes Cave, sediment may only accumulate at a rate of a few centimetres per century; materials near the surface are subjected to the combined trampling disturbance of all the occupations occurring in that time-span (Hughes & Lainpert 1977; 136; Villa & Courtin 1983: 270). Most experimental studies on the effects of human trampling on the distribution of archaeological materials in archaeological sites - for example, Gifford-Gonzalez et al, (1985), Stockton (1973) and Villa & Courtin (1983) - have mainly dealt with vertical displacement of artefacts and its implications for the temporal association of artefacts with archaeological strata.

[TABULAR DATA FOR TABLE 1 OMITTED]

The little previous research on the effects of trampling on the horizontal distribution of artefacts may suggest that larger artefacts should be displaced further than smaller with a positive but statistically insignificant relationship between horizontal movement and artefact size (Nielson 1991: 490-93); the incidental results of Villa & Courtin (1983: 277) show an inverse relationship (also not significant) between lateral movement and artefact size.

Nielson (1991: 492-3) differentiates 'traffic zones', such as paths where people walk (and so scuff artefacts), from 'marginal zones' along walls or in the corners of rooms, where people generally don't walk. Nielson hypothesizes that the larger and medium-sized artefacts displaced by trampling in the traffic zones will eventually collect in the marginal zones where they are protected (in some cases by physical barriers) from further trampling.

The main aspect of rock-shelters which controls traffic flow is the available headroom (ceiling-height). High ceiling areas might be the place where people undertook most activities; perhaps tossing and caching artefacts into the low ceiling zones (e.g. Morwood 1981). Adapting Nielson's hypothesis to a rock-shelter context, areas of the shelter with ceiling above average human height would have been cleared of most larger artefacts by trampling while areas with low ceilings would have collected these artefacts. Smaller artefacts would remain in the traffic areas.

While these high and low ceiling-height zones - corresponding to Nielson's traffic and marginal zone respectively - are roughly defined around average adult human height, we do not know at just what height the expected changes in artefact size and frequency should occur, or exactly how much clearance is needed for comfortable walking. A trampling experiment was conducted (by RT) to clarify this. When studies in post-depositional movement of artefacts have gone straight from a generalized model to testing in the archaeological data, the temptation is to keep varying the model's parameters until a fit with the archaeology is found. Meaningful patterns can be obscured and/or weak patterns emphasized by uncritically selecting divisions between continuous variables (Frankel 1989; Blankholm 1991). Our experiment was designed to define precisely ceilingheight zones relevant to human trampling in Petzkes Cave and to identify the distances artefacts move horizontally.

An experiment

For this experiment a 3-m wide section of the cave was selected [ILLUSTRATION FOR FIGURE 2 OMITTED]) with the greatest variety of ceiling-heights; 147 flakes, flaked pieces and cores made on basalt were selected for the experiment, divided into size classes by weight - when all artefacts are of the same material their weights are directly proportional to their size. We attempted to match the size distribution and range with that perceived during excavation.

Seven groups, each containing 21 experimental artefacts with a similar size distribution were created (TABLE 1). One group was placed in each of the seven excavation squares (shown in [ILLUSTRATION FOR FIGURE 3 OMITTED]).

The seven groups of artefacts were laid on a back-filled surface which covered the full range of ceiling-heights recorded at the site. Each group was allocated to a 1-m strip [ILLUSTRATION FOR FIGURE 3 OMITTED]) within which individual artefacts were spaced at 5-cm intervals along the measured line, alternating each side of the line also by 5 cm. The position of each artefact was then recorded.

Once placed, the artefacts were left to be trampled incidentally by the 15 excavation team members, wearing fiat-soled soft shoes or bare feet, for a period of three weeks. Team members were not aware of the full details of the experiment or of its expectations and for the most part were unaware of the presence of the experimental artefacts. They went through the study area for a number of reasons, to access equipment and the theodolite station at the far end of the cave, and to excavate adjacent squares. Although these activities are not those of the original occupants of the cave, human movement through the cave should have been similarly constrained; the shape of cave topography has not changed much since the surface sediments were deposited.

After three weeks the positions of the experimental artefacts were individually recorded with the Electronic Distance Measurer. Eventually, all but one (from size class 2) of the experimental artefacts were recovered, yielding a final experimental sample of 146 artefacts moved as a result of trampling.

The effects of ceiling-height and of barriers on the extent of movement of stone artefacts of different sizes was then assessed. The outline of the cave's walls and ceiling-height was recorded. The GIS program IDRISI was used to interpolate a contour map of the ceiling-heights from these point measurements and then to overlay trajectories of the experimental artefacts on to the map for direct comparison. The GIS was then used to quantify the movement, frequency and size distribution of artefacts.

Experiment results

FIGURE 4 provides striking evidence that ceiling-height influences the horizontal movement artefacts experience through human trampling. Artefacts in higher ceiling zones moved further than artefacts in low ceiling zones. The largest change occurs at a ceiling-height of [approximately] 2 metres (clearly related to adult human height), with average movement of artefacts (9.3 cm) from areas with ceilig [greater than] [greater than] 2.0 metres about two and a half times ]hat of artefacts from areas with ceilings [less than] 2.0 metres (3.9 cm).

The experiment also shows that at Petzkes Cave larger stone artefacts are horizontally displaced longer distances by trampling than smaller artefacts. FIGURE 5 shows the average movement of artefacts in each of the six size classes.

How likely is it that these measured differences result by chance? The significance was tested using the Monte Carlo technique (Kintigh 1990: 172), with average movements for each size class compared with four randomly generated sets of artefact movement data. Each random set contains the same quantity of artefacts and the same length trajectories, with the size classes randomly assigned in correct proportion (for example, four size class 1 artefacts) among the artefact trajectories. The experimental observations are significant if the real experimental artefact movements show stronger or more consistent trends than the randomly generated movement files.

The results of this comparison are shown in FIGURE 6. In the experiment the smallest size classes i and 2 moved less and the largest size classes 5 and 6 moved more than the same size classes in any of the randomly generated data sets. We conclude that the experimentally observed trend of increasing artefact movement with increasing artefact size is significant. The experiment helps to define for a rock-shelter site the two distributional zones of artefacts generally predicted by Nielson (1991): a 'traffic' zone in areas with high ceilings [greater than]2.0 m) will have fewer and smaller artefacts than the non-traffic or 'marginal' zone [les than] 2 m). However, our results suggest a further breakdown of these zones may be useful.

At 9.3 cm the average horizontal movement of experimental artefacts in the traffic zone, though greater than elsewhere, is quite small. Rather than collecting against the cave walls, this suggests artefacts will mainly collect in a narrow band occurring at a ceiling-height similar to, and just below, average adult human standing height. This comparatively rich 'transitional' zone is made up of the lower ceiling-height areas of the traffic zone and the higher ceiling-height areas of the 'marginal' zone suggested above.

The remaining 'marginal' zone can be further divided into the area immediately adjacent to cave walls and the area which is not, to separate out effects on the horizontal distribution of artefacts from behaviours other than trampling. Tossing (Binford 1983: 152-4), sweeping and dumping are likely to move larger artefacts well away from activity areas (O'Connell 1987: 91-2), far into low ceiling zones against cave walls. Larger artefacts may also be stored for later use in deep, low ceiling-height recesses (Morwood 1981: 35).

To summarise, this experiment, in conjunction with previous studies, predicts four artefact distributional zones from human trampling:

* The traffic zone, with a ceiling-height greater than 2.0 m, (nominally [greater than]2.15 m)

* The transitional zone 1.25 m-2.15 m which covers a range of human heights including older children

* The final two categories are marginal zones: 0.75-1.25 m and [less than] ]0.75 m

The predicted distribution of artefacts is modelled in FIGURE 7.

This simple model assumes that most knapping activities were carried out in high ceiling areas and that tossing and caching behaviour placed some artefacts in low ceiling areas. The original distribution of the artefacts is unknown, but assuming that there were originally more artefacts within the high-ceiling zones than low-ceiling zones, subsequent trampling should affect the distribution pattern of the artefacts in the way in which the model describes.

Testing the model at Petzkes Cave

To test this model we used the sample of artefacts excavated from the surface 3 cm from the same part of the shelter in which the experiment was carried out. As artefacts were found during excavation, their position was recorded using an electronic distance measurer. The analysis of the artefacts addressed firstly how closely artefact frequency and secondly artefact size fit our predictions.

Distribution of artefacts across ceiling-height zones

A plan [ILLUSTRATION FOR FIGURE 8 OMITTED]) shows the distribution of 1998 artefacts from this area whose co-ordinates were recorded during excavation as well as contours of the ceiling-height; the patterns appear to match predictions.

The most striking feature of the distribution is an extremely dense cluster of artefacts in the centre of the study area. Its almost certain explanation is the effect of the cave's drip-line which at this point particularly concentrates water during rain. At the site today water from the cave's drip-line washes fine sediments away down the slope at the front of the cave, removing all but the largest particles and exposing the artefacts. The tell-tale evidence is the lack of fine sediment around the exposed artefacts and their location immediately below the line of the cave's overhang (Morwood 1981: 34).

This 'drip-line' disturbance complicates the identification of the pattern resulting from human trampling under the control of cave ceiling-height. On a reasonably level surface it is not likely to affect the average weight of artefacts present (as even larger particles of sediment remain), but it does create a local peak in spatial artefact frequency. The small drip-line cluster has been removed from the following visual and quantitative analysis.

This drip-line cluster apart, the spatial patterning of artefacts closely matches the modelled pattern [ILLUSTRATION FOR FIGURE 7 OMITTED]): the high ceiling 'traffic' zone contains far fewer artefacts than are present in the low ceiling areas (transitional and marginal zones) to either side. Artefact frequency (in general) declines gradually on moving deeper into the lowest ceiling areas (marginal zone) on the extreme left and right.

The exception is a concentration of artefacts against the rear wall of the cave to the lower left of FIGURE 8. This area, well shielded from wind, is out of the main thoroughfare of the cave yet has sufficient ceiling-height for comfortable sitting; it may in the past have been preferred for activities which allow people to sit.

To quantify these observations, the numbers of artefacts per square metre in separate ceilingheight zones were counted using the IDRISI program [ILLUSTRATION FOR FIGURE 9 OMITTED]). A 0.75 m diameter area around the drip-line cluster was not included in the counts.

The trends displayed in FIGURE 9 closely match those predicted in the model. To test whether or not this distribution of artefacts in space could have occurred by chance we again used Monte Carlo principles, but indirectly, through a K-means cluster analysis (Kintigh 1993). The clustering recorded in the archaeological distribution was compared with that of four false random distributions [ILLUSTRATION FOR FIGURE 10 OMITTED]). The per cent SSE value of the trace (y axis) is well below the traces for random distributions; the archaeological sample's distribution is clustered to an extent unlikely to have arisen from random variation (Kintigh 1990: 172). Coupled with the doubling of artefact frequency across the crucial traffic/transitional zone boundary, this result suggests that the match between the frequency distribution of artefacts at Petzkes Cave and the model is significant.

Distribution of different-sized artefacts in different ceiling-height zones

Information on the distribution of sizes in each of these four ceiling-height zones was obtained by applying IDRISI to a sub-sample consisting of 857 point-plotted stone artefacts with attached size data collected during the excavation of the two northern metre strips across the study area (that is, not including the southern 1-m strip).

The artefacts were divided into the same six size classes used in the experiment (artefacts in the drip-line disturbance area included as that disturbance should not affect the size distribution of artefacts), and the average size class in each of the four ceiling-height zones determined [ILLUSTRATION FOR FIGURE 11 OMITTED]). As predicted average artefact size increases across the traffic/transitional zone boundary (around 2 m ceiling-height).

The Monte Carlo style significance test for these results involves a comparison of average artefact size in different ceiling-height zones between the excavated sample and four false data sets where size classes were randomly allocated (in correct proportions) to real artefact positions in space. Variation from the mean artefact size, between the low [less than]2 m) and high [greater than]2 m) ceiling-height zones for both original and random distributions, is presented graphically in FIGURE 12.

The difference in average artefact size between high and low ceiling-height zones [ILLUSTRATION FOR FIGURE 12 OMITTED])is far greater for the original distribution than for any of the random size distributions. This confirms the significance of the observed size-distribution trend against ceiling-height in FIGURE 11 and its correspondence with the trend predicted in the model.

Conclusions and discussion

The horizontal distribution of excavated archaeological artefacts and their size distribution at Petzkes Cave closely conforms to the predicted distribution derived from experimental results, and is shown as unlikely to result from chance. We conclude that horizontal patterning at Petzkes Cave has been strongly affected by human trampling operating under the control of cave ceiling-height, with average horizontal displacement of artefacts from human trampling increasing with increasing artefact size. It matches our prediction (derived from Nielson 1991) that cave ceiling-height influences the horizontal distribution of artefacts by restricting human trampling disturbance to areas where humans can comfortably walk. The distribution patterns formed by human trampling in caves will be distinctive and relatively easy to discern archaeologically as changes in the frequency and average size of artefacts should correspond in space with areas where cave ceiling-height is equal to average human walking height.

The least disturbed artefacts are the very smallest; although there may be fewer stone artefacts remaining in the high ceiling sections of the cave, they have probably not suffered excessive post-depositional horizontal movement. Clusters of artefacts in these areas may represent original concentrations of activities. Areas deep in the marginal zones will have the most direct connection with purposeful human activity, particularly storing or tossing.

Trampling is a major disturbance factor; but, because the horizontal distribution of artefacts resulting from such disturbance correlates to cave topography, it can be predicted. Although the chances are that larger artefacts will move further than smaller it cannot be inferred that all artefacts have moved. If clusters are found that cannot be explained by trampling, it is possible that the clusters are part of the original deposition pattern. Our future research intends to test this by comparing the distribution of stone artefacts to that of other kinds of archaeological evidence - hearths and materials such as charcoal and starch which, from the principles identified here, are less likely to be affected by trampling.

Acknowledgements. We would like to thank the Australian Research Council for funding the Petzkes Cave project, the volunteers who helped excavate the site and Doug Hobbs who helped set up our recording system and drew FIGURES 1, 9, 10 & 11.

References

BINFORD, L.R. 1983. In pursuit of the past: decoding the archaeological record. London: Thames & Hudson.

BISKOWSKI, M. 1994. GIS & the study of grinding tools in the Teotihuacan Valley, Mexico, in I. Johnson (ed.), Methods in the mountains: proceedings of UISPP Commission IV Meeting Mt Victoria, Australia, August 1993: 115-34. Sydney: Sydney University. Archaeological Methods Series 2.

BLANKHOLM, H.P. 1991. Intra-site spatial analysis in theory and practice. Arhus: Aarhus University Press.

CARR, C. 1991. Left in the dust: contextual information in modelfocused archaeology, in Kroll & Price (ed.): 221-56.

FRANKEL, D. 1989. Koongine Cave excavations 1986-7: investigating spatial patterning, Australian Archaeology 28: 3-13.

GIFFORD-GONZALEZ, D.P., D.B DAMROSCH, D.R. DAMROSCH, J. PRYOR & R.L. THUNIN. 1985. The third dimension in site structure: an experiment in trampling and vertical dispersal, American Antiquity 50(4): 803-18.

HIVERNAL, F. & I. HODDER. 1984. Analysis of artefact distribution at Ngena (Kenya) depositlanai and post-depositional effects, in H.J. Hietala (ed.), Intra-site spatial analysis in archaeology: 97-115. Cambridge: Cambridge University Press.

HOLDAWAY, S. & G. IRWIN. 1993. Electronic recording systems used during the excavation of Sll/20 Ponuii Island, Hauraki Gulf, Archaeology in New Zealand 36:27-37.

HUGHES, P.J. & R.J. LAMPERT. 1977. Occupational disturbance and types of archaeological deposit, Journal of Archaeological Science 4: 135-40.

KINTIGH, K. 1990. Intrasite spatial analysis: a commentary on major methods, in A. Voorrips (ed.), Mathematics and information science in archaeology: a flexible framework: 165-200. Bonn: Halos. Studies in Modern Archaeology 3.

1993. Tools for quantitative archaeology. Unpublished brochure describing the software package of the same name. Tempe (AZ): Anthropology Department, Arizona State University.

KROLL, E.M & T.D. PRICE (ed.) 1991. The interpretation of archaeological spatial patterning. New York (NY): Plenum Press.

MORWOOD, M.J. 1981. Archaeology in the Central Queensland Highlands: the stone component, Archaeology in Oceania 16(1): 1-53.

NASH, D. & M.D. PETRAGLIA, 1987a. Natural formation processes and the archaeological record: present problems and future requisites, in Nash & Petraglia (ed.): 186-203.

(ed.) 1987b. Natural formation processes and the archaeological record. Oxford: British Archaeological Reports. International series 352.

NIELSON, A.E. 1991. Trampling the archaeological record: an experimental study, American Antiquity 56(3): 483-503.

O'CONNELL, J. 1987. Alyawara site structure and its archaeological implications, American Antiquity 51(1): 74-108.

RIGAUD, J. & J.F. SIMEK. 1991. Interpreting spatial patterns at the Grotte XV: a multiple-method approach, in Kroll & Price (ed.): 199-220.

SCHIFFER, M.B. 1983. Toward the identification of formation processes, American Antiquity, 48: 675-706.

SIMEK, J. 1984. Integrating pattern and context in spatial archaeology, Journal of Archaeological Science 11: 405-20.

SIMEK, J & R.R. LARICK. 1983. The recognition of multiple spatial patterns: a case study from the French Upper Palaeolithic, Journal of Archaeological Science 10: 165-80.

STOCKTON, E. 1973. Shaw's Creek Shelter: human displacement of artefacts and its significance, Mankind 9: 112-17.

VILLA, P. & J. COURTIN. 1983. The interpretation of stratified sites: a view from underground, Journal of Archaeological Science 10: 267-81.

WANDSNIDER, L. 1987. Natural formation process experimentation and archaeological analysis, in Nash & Petraglia (ed.): 151-85.