The Rules of Certainty: A History of Rational Connoisseurship
thesisposted on 10.12.2012, 00:00 by Andrew P. Belongea
This study examines several computer systems developed during the past decade that analyze paintings and drawings through statistical measurement to attribute artistic authorship. The developers of these systems highlight the practical applications of their technology, such as detecting forgeries, and furthermore propose that it can identify stylistic schools and reveal potential influences between artists. The present work contextualizes these recent algorithms by tracing the history of connoisseurs’ attempts to quantify artistic performance and evaluate artworks according to a set of standard and explicit rules. Although methods seeking to quantitatively evaluate art date from the late seventeenth century, the concept of employing a statistical analysis only began during the latter half of the twentieth century. Influenced by the fields of information theory and cybernetics, a number of art historians and theorists started to propose the idea of identifying stylistic differences between artists through a statistical analysis of features in paintings and drawings. Ernst Gombrich, a figure to whom this study will give particular emphasis, even formulated techniques in the late 1950s, outside of a computational framework, which underlie the principles of these computer systems. This history demonstrates how the application of computer analysis for artistic evaluation is more than an implementation of a new form of technology; it is also a strategy for pursuing certain goals and methods of connoisseurship formed throughout several centuries. Moreover, these systems signal a new role for objective and scientific examination in attributing artworks. Previously, scientific procedures – pigment sampling for example – were only able to disqualify a work from a particular artist’s oeuvre by demonstrating inconsistencies between the material construction of a piece and the purported time of its creation. In contrast, these computer systems can provide a connoisseur with positive evidence in favor of an attribution by quantifying particular stylistic continuities between an artwork and an artist’s oeuvre.