The use of statistics to define taxonomies of film voices: a practical approach to documentary classification

Abstract

The study of visual content using qualitative methods currently involves much scholarly work in the field of digital humanities. Documentary films are considered among those objects defined as visual content and are therefore not the focus of most attention within Film Studies. One of the bestknown researchers in this area, Bill Nichols, proposed a theoretical classification of six documentary voices that is widely recognized in academic circles. However, the classification is based on induction and intuition which, until now, have not been verified in accordance with the evidence. This study proposes a mixed qualitative and quantitative methodology that allows the authors to surpass Nichols' model and generate a new classification based on empirical criteria. To that end, they study the opinions of users of and experts on Nichols' theory through the card sorting technique borrowed from cognitive science that is based on multivariate statistics. In this way, they generate a hierarchy of formal cinematographic elements grouped into four documentary voices and Nicholas' classification proposal is partly replaced as two of the voices in his model disappear but the four others are confirmed. In addition, this new model allows the authors to determine which formal elements are assigned to each one of the confirmed documentary modes and with what relative weights, thereby opening up the possibility of generating a predictive model. The article therefore demonstrates that cluster analysis and other techniques borrowed from multivariate statistics can be used to empirically and quantitatively verify theories in film studies as well as in other areas such as poetics. This suggests that it is possible to verify empirically and use quantitative models for the study of visual content and complement other methods based on induction and intuition.