Surface lightness influences perceived room height

Abstract

Surprisingly little scientific research has been conducted on the effects of colour and lightness on the perception of spaciousness. Practitioners and architects typically suggest that a room's ceiling appears higher when it is painted lighter than the walls, while darker ceilings appear lower. Employing a virtual reality setting, we studied the effects of the lightness of different room surfaces on perceived height in two psychophysical experiments. Observers judged the height of rooms varying in physical height as well as in the lightness of ceiling, floor, and walls. Experiment 1 showed the expected increase of perceived height with increases in ceiling lightness. Unexpectedly, the perceived height additionally increased with wall lightness, and the effects of wall lightness and ceiling lightness were roughly additive, incompatible with a simple effect of the lightness contrast between the ceiling and the walls. Experiment 2 demonstrated that the floor lightness has no significant effect on perceived height, and that the total brightness of the room is not the critical factor influencing the perceived height. Neither can the results be explained by previously reported effects of brightness on apparent depth or perceived distance.

Keywords:

• Room perception
• Architecture
• Architectural psychology
• Interior design
• Colour
• Brightness
• Contrast
• Depth
• Psychophysics
• Visual perception
• Illumination
• Interior space
• Virtual reality

Acknowledgments

We thank two anonymous reviewers for helpful comments on an earlier version of this article. We are grateful to Agnes Münch for constructing the virtual rooms. The work was supported by the Interdisciplinary Centre for Neuroscience (IFSN) at the University of Mainz.

Notes

¹ Due to the repeated measures structure of the data, a subject-specific, random-effects model approach was used (SAS PROC MIXED; cf. Burton, Gurrin, & Sly, 1998; Liang & Zeger, 1993). Subject-specific models assume regression parameters (i.e., intercept and slope) to vary from subject to subject. Random-effects models belong to the class of subject-specific models and model the correlation structure by treating the subjects as a random sample from a population of all such subjects. In the analysis, the variance–covariance matrix was specified as being of type “unstructured” (UN)—that is, the procedure placed no constraints on the correlations across observations within one subject.

² We are grateful to an anonymous reviewer for pointing out this possibility.