The purpose of this study was to develop and test a self-rating restoration scale (RS) designed to measure the restorative quality of environments. Both the Kaplan and Kaplan [The Experience of Nature: A Psychological Perspective, Cambridge University Press, New York] and Ulrich [Aesthetic and affective response to natural environment, in: I. Altman, J.F. Wohlwill (Eds.), Behavior and Natural Environments, Plenum Press, New York] hypothesize that restorative environments are settings that facilitate the reduction of stress. Over the past decade, an increasing amount of empirical research has also shown that the restorative influences of environments manifest themselves in emotional, physiological, and cognitive responses of humans [J. Environ. Psychol. 11 (1991) 201]. Thus, the RS should cover, at least, these three dimensions. Moreover, the dimension of intended behavior in environments was also included. This RS was examined and revised through a two-phased experimental design. Forty-eight color slides selected from thousands of slides were used as the stimuli and the surrogates for the actual environments in the experiments. These 48 strictly controlled slides represented a proper and comprehensive sample of the six major terrestrial biomes of the world: desert, tundra, grassland, coniferous forest, deciduous forest, and tropical forest [Ecology and Our Endangered Life-Support Systems, Sinauer Associates Publishers, Sunderland], and varied as a function of high and low levels of three physical variables: complexity, openness, and water features. Five experiments with five groups of undergraduate students at Texas A&M University as subjects (total n=505) were conducted to test the RS. Results of exploratory, confirmatory, analysis of moment structures (AMOS), correlation, principal component, and reliability analyses indicated that internal validity, convergent and discriminant validity, convergent and divergent construct validity, and reliability of the RS were all adequate. Thus, the operational definition and the construct of restorativeness developed in this study can be applied to future research on recovery from stress. Hopefully, this theory-derived and data-oriented RS will be eventually applied to various environments by any concerned individuals to examine the result of planning and design practice in terms of a desired state of recovery from stress.