Fuzzy Logic Application to Artifact Surface Survey Data

Survey archaeologists, studying human activities over space during time, need to assign a chronological framework to their field data and consequently produce maps showing sites and their chronology, determined mainly by the cultural material. Specialists who assign chronological attributions, have difficulties with rigid chronological categories, and tend to create additional ones, in order to match their data. Moreover, they face materials with uncertain, or multiple chronology, and end up with large chronological ranges. Survey archaeologists need to extract meanings out of those data and, using traditional classification methods, are forced to reduce those classes into fewer, and often not representative of the surface data. We propose to use a fuzzy logic approach in order to give data more transparency and to present a more realistic map, according to the real nature of the data. 1. Fuzziness in Artifact Surface Surveys Datasets In the methodological framework of intensive systematic surveys, the distribution of material over the space allows us to interpret, at several degrees, the distribution of human activities in the landscape, during time. Once recognized an activity focus – or site – in the landscape, by using criteria related to density of surface material (corrected by several variables) and geomorphological characteristics of the landscape, we then go much deeper into the ‘site’, recording and collecting material at a higher level of intensity. Result of the process are assemblages of material, spatially linked to a survey unit (geometric-regular or irregular according to cases), that have to be processed and studied in order to produce meaningful results on the occupation of the landscape on that site-area. Of different nature and entity are the problems that a landscape archaeologist has to deal with while managing data from a systematic intensive artifact surface survey. Data representativity, as well as data reliability and period visibility are, for instance, crucial issues to be considered while interpreting data and extracting meanings from different landscape datasets. In this paper, we would like to examine one of the many critical issues linked to the methodology of data retrieving from survey data. We will concentrate especially on the fuzziness of chronological attribution, crucial factor in order to recognize the presence of an activity focus in a certain period, as well as to assess the continuity of occupation of a certain site. Time, as well as space, are distinctive factors in the archaeological research and in the interpretation of the human landscape. Time, seen as the chronological attribution to material culture, is always characterized by a fuzziness, that can be named ‘temporal fuzziness’. This has to do with our imprecise knowledge of the material culture in the different periods of the past, but also with the need, for research purposes, to look at chronological ranges, since we cannot go back to the individual moment in which, for instance, a pot was made. We need, therefore, to classify the chronological attributes. This process ends up with the creation of a huge variety of classifications, that become even more if one takes into account the terminology in use for the diverse chronological classifications. Often the chronological systems in use in different areas differ quite sensibly, and many examples of either slight or substantial diversion between chronological classification systems can be found. This has to do with the fuzzy nature of period terminology: in Greece, for instance, discrepancies can be found in the ways of defining the Hellenistic period. Some scholars use the chronological range HL as corresponding to the Hellenistic period according to historical sources (which means from the death of Alexander the Great, dated 323 BC ...). On the other hand, in terms of material culture/archaeological material, and in terms of settlement pattern, EHL is differentiated from LHL: EHL goes often together with LC and is differentiated from LHL and then from the ER period. The LC is a period of peak in the settlement pattern and high densities on the landscape, while LHL and ER periods are characterized by a decrease in site density and appearance of larger estates in the landscape. We must therefore devise – as stated recently by Van Leusen (2002: 238) – a system of fuzzy dates that would allow us not to loose the fuzzy nature of period terminology. Using fuzzy logic we would try to assess a better categorization of real data that better matches the actual reality of occupation, and the degree of occupation in the landscape for each period. In the following section, we will focus on the problem of chronological attribution to sherd assemblages. 2. Fuzziness in Chronological Attribution