Abstract
Perhaps the best-known prescriptive rule of pictorial composition is the "rule of thirds" (ROT), which posits that: (a) the best positions for the focal object within a rectangular frame lie along the vertical and horizontal lines that divide the frame into thirds, with maxima at the four intersections of these third-lines, and (b) the worst positions lie along the vertical and horizontal axes of symmetry, with the minimum being at the frames center. We tested these predictions by measuring peoples preferences for placement of a single object at the nine points defined by the 3x3 grid of intersections among the horizontal and vertical third-lines and symmetry-axes. We measured forced-choices between two pictures of the same object (fish/dog/eagle) facing in the same direction (forward/leftward/rightward) at all possible pairs of positions in the 3x3 grid. The results strongly contradicted both of the ROTs main claims. The most preferred position for every object in every facing direction was the frames center, consistent with a pronounced center bias (Palmer, Gardner & Wickens, 2008). This bias was strongest when symmetrical objects faced forward. Preference for horizontal placement of left-facing and right-facing objects was biased asymmetrically away from the center, toward the side opposite the facing direction, consistent with an inward bias (Palmer et al., 2008). Vertical biases were much less pronounced, but people tended to prefer the flying eagle to be higher in the frame, consistent with an ecological bias toward its viewer-relative position in the environment (Sammartino & Palmer, 2011). Further results show that these biases can be modulated by a second visual element at other background locations within the frame. We conclude that, although the ROT is empirically incorrect, it may be heuristically useful in encouraging artists to use off-center focal-object locations when those placements "work" in the context of the entire image.
Meeting abstract presented at VSS 2014