this study, we conducted evaluation experiments by semantic differential method to examine the validity of utilizing three-dimensional (3D) computer graphic images made with virtual reality modeling language (VRML) as a stimulus for landscape assessment both in the laboratory and on the Internet. The results of the laboratory experiment and the Internet experiment were compared. In both experiments, we simulated a garden landscape at Chiba University by using VRML. With three media representing the landscape (Photographs, Static images of VRML and Walk-Through images by VRML), we recorded respondents' impressions of each scene. Their responses to the three stimuli were analyzed statistically with profile analysis, ANOVA and T-test. We also asked respondents which media best represented the actual landscape.The results of the profile analysis in the laboratory experiment were similar to those of the Internet experiment. The level of reality was the order of the Photographs, Walk-Through images and Static images in both experiments. The results showed that the VRML images could be used as an effective stimulus for landscape assessment both on the Internet and in the laboratory. The level of reality of Walk-Through images in the Internet experiment was almost the same as that of photographs. The results also indicated that respondents realized garden landscapes easily by Walk-Through simulation using VRML. (c) 2005 Elsevier B.V. All rights reserved.