Physical and affective correlates to perceived order in open-plan architectural space Presented at the Dresden International Symposium of Architecture

v1.2, June 21, 2004 While experiential qualities of rectangular architectural spaces can be effectively predicted from properties like room proportions or area (Franz, von der Heyde, & Bülthoff, 2003), these factors from normative architectural knowledge obviously cannot be directly transfered on open-plan indoor spaces. Phenomenologically as well as in empirical aesthetics, nontrivial forms are often compared by collative variables (e.g., complexity, regularity, cf. Berlyne, 1960, 1972) that are introspective assessment criteria of structural properties of a stimulus array (Wohlwill, 1976) and have proven to allow predictions mainly on the arousal dimension of affective experience (Stamps, 2000). In the following, we introduce a novel approach that relates affective and collative qualities of arbitrarily shaped architectural spaces to directly measurable parameters. The concept of isovists (i. e. viewshed polygons, cf. Benedict, 1979) is used to generically describe spatial properties of architectural spaces from a perceptually-oriented viewer-centered perspective. In a psychological experiment, we compared numerical factors describing isovists and their derivatives (Turner, Doxa, O’Sullivan, & Penn, 2001; Psarra & Grajewski, 2001) of virtual reality (VR) simulated rooms with their experiential qualities, quantified with a semantic differential scaling in six principal categories (pleasure, interestingness, beauty, spaciousness, complexity, and clarity). Since VR offers a high degree of perceptual realism, control, and flexibility, properties of spaces can be individually varied, allowing to ascribe observed variances in the measured experience to particular differences between the stimuli. The study consisted of two stages comprising 18 and 16 scenes from a fictive art gallery and was conducted with 2 groups of eight participants. The stimuli were presented as radiosity-rendered spherical panorama images on a 130x90 degree wide-angle projection system. The original analysis (Franz, von der Heyde, & Bülthoff, 2004), correlating averaged ratings with 28 characteristic values derived from the isovists, was focused on automatically detectable measurands mainly related to complexity. Isovist area, density of wall edges, and enclosure ratio turned out to be the most effective predictor variables. Extending the initial scope, an additional exploratory correlation analysis of the second scene set concentrated on manually evaluated aspects commonly related to order, as symmetry or self-similarity. Particularly the number of symmetries turned out to be a relevant factor explaining additional variance in the ratings (e.g., correlation coefficient with rated pleasure r=-0.75, p<0.001). Beside symmetry, from the tested factors ∗Max Planck Institute for Biological Cybernetics, Spemannstrasse 38, 72076 Tübingen, Germany. E-mail [email protected] , internet http://www.kyb.tuebingen.mpg.de/ ̃gf.