Articles by Amilcare Bietti

In this paper we describe an application of image processing for the texture study of the use wear polishes on flint artifacts, experimentally obtained on two types of materials: wood and hide. The quantification of the textural properties of the images, obtained with a high magnification (200X) metallographic microscope strictly follows the rules indicated by HARALICK et al. in 1973. Twenty-two textural features were obtained from the image and statistical analysis then allowed the discrimination between the two classes of materials. The techniques employed, T-test and linear stepwise discriminant analysis, show that the discrimination between tools working wood and working hide is quite good, in the case of distance between the pixels = 1. Possibilities of future and more refined analyses are then discussed at the end of the paper.

This paper provides a short summary of some recent results in intrasite spatial analysis in Prehistoric archaeology and, in particular, in Paleolithic archaeology where, as it is well known, intrasite spatial analyses have been applied for at least twenty years. Starting with some “classic” examples of 80’s (the Magdalenian site of Pincevent, the ethnoarchaeological analysis of the Mask site in Alaska), and a short illustration of the techniques applied (correspondence analysis, k-means cluster analysis, unconstrained clustering, factor analysis, etc.), some of the most recent applications are described: the Mesolithic Barmose I site in Denmark, studied by H. Blankholm, and the Middle Pleistocene site of Grotte Vaufrey in South West France (Dordogne), where the intrasite spatial analysis has been performed by J. Simek. We then discuss the relationship between the “quality” of the quantitative procedures used in the data analysis, which is in general of an outstanding level, and the “reliability” of the archaeological interpretation. A major conclusion is that the problems arising in the archaeological interpretation (a good example of which is Grotte Vaufrey) are much more connected with the choice of variables in the quantitative analysis, than with the different algorithms employed: we thus recommend to the archaeologists (and not to the computer scientists involved in the research) the basic selection and choice of the essential variables derived from the data.

As an example of pattern recognition problems in prehistory, the authors present two different kinds of classification (unsupervised and supervised) applied to a sample of common Upper Paleolithic tools: the end-scrapers of Grotta Polesini, coming from layer 1 and layer C12. The results, obtained through the use of statistical techniques, led to a general conclusion: in building a typology more attention should be given to functional considerations and to a deep understanding of behavioural patterns on a more local scale. The traditional prehistoric typological lists are therefore criticised, as they are often too general and too much oriented towards large scale comparisons in space and time.