APPLICATION OF MULTIVARIATE CHEMOMETRIC TECHNIQUES TO THE STUDY OF ROMAN POTTERY (TERRA SIGILLATA)

Supervised and unsupervised pattern recognition techniques were used to classify 48 sherds of Roman pottery (terra sigillata), analysed by inductively coupled plasma atomic emission spectrometry and atomic absorption spectrometry for seven major and minor elements (Al, Fe, Ca, Mg, K, Ti and Mn). Hierarchical agglomerative clustering and principal component analysis were used to classify the studied material into compositional groups which could account for different centres of production; soft independent modelling of class analogy (SIMCA) was used to solve questions regarding doubtful assignments. The results indicate that, in the case study, a throughout statistical treatment can allow one to discriminate wares produced in different geographical areas on the basis of the seven elements accounted for.