Towards computerized typology and classification of ceramics Ayelet Gilboa a , Avshalom Karasik b,c *, Ilan Sharon c , Uzy Smilansky b a The Zinman Institute of Archaeology, University of Haifa, Mount Carmel, Haifa 31905, Israel b Department of Physics of Complex Systems, The Weizmann Institute of Science, Rehovot 76100, Israel c Institute of Archaeology, The Hebrew University, Jerusalem 91904, Israel Received 28 July 2003; received in revised form 26 October 2003; accepted 29 October 2003 Abstract We report on newly developed mathematical and computational tools for morphological description, classification and analysis of archaeological artifacts. The need for such tools is most acutely felt, due to two main factors: (1) The lack of objective, quantitative criteria for shape analysis, classification and comparanda; (2) The overwhelming abundance of data, which renders impossible any extensive comparative typological analysis using traditional methods. We shall describe the main ideas which distinguish our method, and demonstrate its applicability by presenting the analysis of two assemblages of Iron Age ceramics from sites in Israel. 2003 Elsevier Ltd. All rights reserved. Keywords: Pottery; Typology and classification; Ceramic profiles; Curvature analysis; Near Eastern Iron Age archaeology; Phoenician pottery 1. Introduction Artifacts are by definition the direct products of human action. As such, they convey significant infor- mation regarding various cognitive and behavioral as- pects of both their producers and users. Archaeologists who study assemblages of such artifacts seek to identify in them distinctive patterns that can be used for drawing inferences regarding a plethora of issues, from straight- forward spatial and temporal settings, to higher-range issues such as technological and cognitive capabilities, motor skills, symbolic assertions and negotiations, social, ethnic and other identities, economic activities, and more. There are two basic techniques by which archaeologi- cal research approaches artifacts analysis: 1. Typological classification, where the numerous arti- facts are clustered into discrete ‘types’, and the assemblage is sorted according to these pre-defined types. Underlying this approach is the assumption that artifacts were produced according to, and there- fore can be classified into, discrete templates. Such classification raises serious issues—what is a ‘type’? Is there an objective way of classifying artifacts to ‘types’? (For a pioneering concern with these ques- tions, see for example [22]) Is the choice of types optimal (keeping the number of types to a minimum, while losing minimal information). 2. An alternative approach is attribute analysis, in which artifacts are described according to a set of selected attributes, and which seeks to establish a direct correlation between one (or arrays) of these attributes and temporal sequences, spatial associ- ations, or cultural patterns. The main problem of this approach is how, among the myriad of at- tributes usually observed in a large assemblage of artifacts, can one isolate a subset which varies intelligibly with the facets under study. In practice, most archaeologists bring both of these approaches to bear, depending on the problem at hand. In both approaches, shape attributes are among the most fundamental properties by which artifacts are characterized and studied. These include the description of the general shape of the artifact, defined by its contour (the line which marks its boundary or cross section), as * Corresponding author. Tel.: +972-8-93-44-987; fax: +972-8-93- 44-109 E-mail address: fekarasi@wisemail.weizmann.ac.il (A. Karasik). Journal of Archaeological Science 31 (2004) 681–694 SCIENCE Journal of Archaeological http://www.elsevier.com/locate/jas SCIENCE Journal of Archaeological http://www.elsevier.com/locate/jas 0305-4403/04/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2003.10.013