This paper discusses a general scheme for determining the quality of scientific visualization systems. It presents psychological tests which have been performed in order to find quantitative relaationships for this scheme, and discusses why simple numerical relationships may be hard to find. Our work is motivated by research into interactive and immerxive scientific visualization which pose two opposing demands: maximum image quality at sufficient frame rates. We believe that these two factors are also crucial for many other applications, e.g., virtual reality. For this scheme, special focus is on the user's perception of the system. Three components of Visualization System Quality are identified: Data quality, image quality, and interaction quality. In order to migrate from a merely qualitative to a mor quantitative model, psychological tests were performed to measure the influence of frame rate and rendering mode on the perception of visualized threedimensional vector data. Results o f the tests are presented and general suggestions for good perceptual tests are made. We believe that the experience we gained will be of benefit to many who are interested in questions of visualization system theory.