NONMETRIC MULTIDIMENSIONAL SCALING APPLIED TO PROVENANCE STUDIES: ANCIENT GREEK AND TURKISH MARBLE AS A CASE STUDY

Ancient artifact provenance studies may use multiple variables to characterize unknown samples. A well-designed study will group samples with similarities and separate them from samples of greater dissimilarity. In multivariate studies, it may be difficult to observe the degree of dissimilarity between samples without the use of a statistical method. This study investigated the use of nonmetric multidimensional scaling as applied to 40 ancient Greek and Turkish marble quarry samples. The 40 quarry samples were of known origin. Nonmetric multidimensional scaling is a numerical ordination method that uses a predefined number of axes to reduce dimensionality based on the dissimilarity among samples in the dataset, effectively reducing the number of variables. Unlike discriminant function analysis, a method used previously in marble provenance, no pre-defined groups are set at the beginning of the analysis. Results are based entirely on the variable scores of each sample. Variables in this study included maximum calcite grain size, presence or absence of specific accessory minerals, and stable isotopes δ¹³C and δ¹⁸O. The results of this ordination method provided a relatively simple way to visually note similarities and differences between the 40 marble quarry samples. The method may be used in the future to compare unknown artifact samples to known quarry samples, in effort to better determine the unknown samples’ possible provenance.