Computer aided techniques for the attribution of Attic black-figure vase-paintings using the Princeton painter as a model.

Abstract

Because of their abundance and because of the insight into the ancient world offered by the depictions on their decorated surfaces, Attic painted ceramics are an extremely valuable source of material evidence. Knowing the identities and personalities of the artists who painted them not only helps us understand the paintings, but also helps in the process of dating them and, in the case of sherds, reconstructing them. However, few of the artists signed their wares, and the identities of the artists have to be revealed through a close analysis of the style in a process called attribution. The vast majority of the attributions of archaic Attic vases are due to John Beazley whose monumental works set the stage for the dominance of attribution studies in the scholarship of Greek ceramics for most of the 20th century. However, the number of new scholars trained in this arcane art is dwindling as new avenues of archaeological research have gained ascendency. A computer-aided technique for attribution may preserve the benefits of the art while allowing new scholars to explore previously ignored areas of research. To this end, the present study provides a theoretical framework for computer-aided attribution, and using the corpus of the Princeton Painter - a painter active in the 6th century BCE - demonstrates the principal that, by employing pattern recognition techniques, computers may be trained to serve as an aid in the attribution process. Three different techniques are presented that are capable of distinguishing between paintings of the Princeton Painter and some of his contemporaries with reasonable accuracy. The first uses shape descriptors to distinguish between the methods employed by respective artists to render minor anatomical details. The second shows that the relative positions of cranial features of the male figures on black-figure paintings is an indicator of style and may also be used as part of the attribution process. Finally a novel technique is presented that can distinguish between pots constructed by different potters based on their shape profiles. This technique may offer valuable clues for attribution when artists are known to work mostly with a single potter.