Abstract
Composition is an important topic in visual art. The literature suggests a bias for objects on the right side (Levy, 1976) and two additional biases with respect to positioning of objects within a rectangular frame: a Centre bias and an Inward bias (Palmer, Gardner, & Wickens, 2008). We analysed images of animals from three datasets of works of art: two datasets were from artists well known for their portraits of animals (Bewick, Stubbs) and the third was a medieval bestiary. There was no overall displacement of the subject to the right or to the left of the picture. However, we found a bias consisting of more space in front compared to behind the animal, consistent with Palmer at al.'s findings and with their definition of an Inward bias. Because our animals never face towards the centre we use the term Anterior bias. In addition, we found a modulation of this bias on the basis of the facing direction of the animal, consisting of a stronger Anterior bias for left-facing animals. This asymmetry may originate from a combination of an Anterior bias and a Right bias. Finally, with respect to size we found that the size of the animals predicted the proportion of the picture occupied, an effect known as “canonical size”.
Notes
1To carry out the measurements of the animals in A General History of Quadrupeds (1790) we used the 1970 reprint (Ward Lock, London) of the fifth edition (1807). The total of 226 images does not include the tailpieces that fill the space at the end of many chapters and are smaller in size and more diverse in subject. The 1807 edition of the book is available online (http://books.google.com/). More information on Bewick is available from the Bewick society (http://www.bewicksociety.org/).
2For the Stubbs dataset the images were taken from Part I of the Sotheby's Catalogue (reprinted by Stipple Publishing Ltd, 1989). Excluding anatomical drawings and repetitions there were 142 images of animals. Of these we selected 71 because the other 71 did not have a single main character (i.e., there were multiple dominant characters and each of them commanded a similar amount of space within the overall image).