3D scanning of stone tools : Quantity versus Quality
At the ArcheObjets3D workshop (29-30 Juni 2016, Maison de l’Orient, Lyon) we were treated to some fine examples of high resolution scanning and the beautiful results that they can produce. As in many things though, there is a trade-off between quality and quantity. Unless you have a dedicated scanning lab like the Israel Antiquities Authority, it is necessary to come to some compromise between the two. My own work has always taken a statistical approach to the analysis of stone artefacts in order that the patterns which we observe may be tested in an ostensibly objective way. With this in mind my 3D scanning operates at the lower end of the quality spectrum so that I can scan larger samples of each assemblage and compare them statistically.
I use a NextEngine laser scanner, scanning at a resolution of 1,000 to 10,000 points per inch. The objects which I scan are typically stone axes between 5 and 30 cm long, and scanning at this resolution means the quality is just enough to use them as illustrations (Fig. 1). For smaller artefacts like flakes it is difficult to get the extremely acute edges and detail on the dorsal surfaces with a NextEngine. An advantage of the NextEngine when scanning elongate objects like axes is that the turntable and clamp allows the object to be scanned in a vertical position so that almost the entire object is able to be scanned without moving it. Indeed for bifacially worked axes where the edges converge around most of the circumference it is only the very small parts of the axe in contact with the clamp that are not visible to the scanner. The automatic hole filling function is usually sufficient to make a complete and reasonably accurate scan without repositioning the object.
The NextEngine software is very easy to use and total scanning time for a typical stone axe including removing noise (parts of the clamp that are scanned), fusing the scans and filling holes, and orienting the object (done manually using the NextEngine Software), takes around 25 minutes. About 15 minutes of that time the NextEngine doing the automated scanning during time which I am able tocount the flake scars and record qualitative attributes on other axes.
The best thing about having numerous scans is being able to go back and look at a broad sample of an assemblage without having to leave your desk. I have just finished a paper on the Indian Acheulean for which I got the scans back in 2011. This paper evolved into something quite different from what I was originally intending with the scans, but I was able to keep looking at the assemblages and analysing them in different ways. Figure 2 shows a pattern which I only observed after recording the actual objects with the benefit of being able to accurately examine them in section: the side-struck cleavers on the left have thick biconvex sections, whereas the end-struck cleavers on the right have thin flat sections.
|Fig. 2 : Cleavers from the classic Indian Acheulean sites of Chirki and Morgaon (right) and from the transitional Acheulean to Middle Palaeolithic site of Bhimbetka (From Shipton in press).|
The scans not only allowed me to observe the pattern shown in Figure 2, but also to quickly illustrate it on several different artefacts. Scans allow surface topography to be viewed without the distraction of colour variation, which is ideal for reading knapped stone. Making such illustrations from the scans is very easy and allows for accurate characterization of sections and profiles that are often key variables for understanding shaped stone tools, as well as oblique perspectives which would otherwise be tricky to illustrate (Fig. 3-5).
Another advantage of the scans is being able to measure variables with greater accuracy. Axial length, width, and thickness measurements are actually quite difficult to obtain with callipers on larger objects if the points of maximum distance are not directly opposite each other. These variables can be measured with absolute accuracy on the scans once they have been correctly oriented. Edge angle is one of the most notoriously unreliable measurements in lithic analysis, yet we know it to be a very important influence on flake propagation. Scans allow angles to be measured according to precise protocols, for example by choosing three well defined points. Millan Mozota showed us another very interesting way of measuring angles using scans by defining planes on the surface of the object. In my recent paper (Shipton in press) I measured the edge angles on cleaver bits and was able to show statistical differences between those at a transitional site versus a classic Acheulean site, which I argued was related to the control of convexities on core surfaces.
Two other variables that scans allow for the accurate measurement of, are volume and surface area. Chris Clarkson and I devised an index of reduction intensity by dividing the number of flake scars by the surface area of an object (Clarkson 2013; Shipton and Clarkson 2015a). Using controlled knapping experiments where we scanned stone artefacts at various stages of reduction, we were able to demonstrate that flake scar density is a good predictor of the mass lost to a piece of stone during knapping (Fig. 6). Thus we are now able to control for reduction intensity : a critical and elusive variable in lithic analysis.
One of the main questions that I wanted to address with the reduction index is whether or not Acheulean bifaces were resharpened and, if so, how it affected their shape. The test I devised to determine if resharpening had a significant influence on an assemblage was whether or not there was a relationship between size and reduction intensity – such that small bifaces are either small because they were more reduced (resharpening), or because they were made on small clasts (no resharpening). Here again the scans proved useful as volume allowed for a measure of size that was independent of the surface area used in the scar density index. Volume also does not vary according to rock type unlike weight, which is typically used as a measure of size. We found that at the Acheulean site of Kariandusi there was no relationship between size and scar density in local materials, but there was among exotic ones, suggesting that exotic bifaces were resharpened whereas local ones were not. Likewise for the Acheulean site of Isenya where we found a relationship for handaxes, but not for cleavers. This was particularly interesting as cleavers with their bits formed by the original flake blank were demonstrably not resharpened (Fig. 7).
|Fig. 7 : Handaxes (left) and cleavers (right) from the Acheulean site of Isenya. Figure from Shipton and Clarkson 2015a). Note that the cleavers cannot not have been resharpened because, like most cleavers, their bit is formed by the original flake blank.|
To test the influence of resharpening on artefact shape, we resharpened several handaxes according to different protocols and scanned them after each bout of resharpening. The scans come into their own here as they preserve an archive of the earlier stages in the life history of experimental objects that would otherwise be lost. We took a series of 18 3D geometric morphometric landmarks on these scans in the program Rapidworks, and statistically analysed how their shape changed over the course of the resharpening iterations. We then measured the scar density and took the same series of landmarks on samples of British bifaces, which are known for their distinctive shapes (Fig. 8). Our results showed that there was some evidence for resharpening on the British bifaces, but this was not extensive and it did not systematically alter their shape, suggesting the shape variation is more likely due to cultural factors (Shipton and Clarkson 2015b).
Scanning is transforming the way in which we conduct lithic analysis. However there is a trade-off between quality and quantity of scans. I have opted for quantity because my work involves the statistical analysis of lithic variation and thus requires decent sample sizes. In the near future I hope that scanning will have advanced enough so that a compromise between quality and quantity is not necessary. I am excited by the new methods of scanning that were displayed at the workshop. And I am keen to apply some of the ways of analysing scans that we discussed, such as automatically extracting geometric morphometric landmarks, quantifying stone clast reducibility, and using elliptic fourier analysis to assess outline shapes;to my own research.
Shipton, Ceri, Ash Parton, Paul Breeze, Richard Jennings, Huw S Groucutt, Tom White, Nicholas Drake, Remy Crassard, Abdullah Alsharekh, and Michael Petraglia. 2014. Large Flake Acheulean in the Nefud Desert of Northern Arabia. Paleoanthropology 2014: 446-462.
Shipton, Ceri, and Christopher Clarkson. 2015a. Flake scar density and handaxe reduction intensity. Journal of Archaeological Science: Reports 2:169-175.
Shipton, Ceri, and Christopher Clarkson. 2015b. Handaxe reduction and its influence on shape: An experimental test and archaeological case study. Journal of Archaeological Science: Reports 3:408-419.
Shipton, Ceri, Marshall Weisler, Chris Jacomb, Chris Clarkson, and Richard Walter. 2016. A morphometric reassessment of Roger Duff’s Polynesian adze typology. Journal of Archaeological Science: Reports 6: 361-375.
Shipton, Ceri. (2016). Hierarchical organization in the Acheulean to Middle Palaeolithic transition at Bhimbetka, India. Cambridge Archaeological Journal 26 : 601-618
Ceri Shipton is Fellow in East African Archaeology at the British Institute in Eastern Africa and the University of Cambridge. He is a specialist in the technological analysis of knapped stone artefacts.