The central research issue addressed by this paper is how we can design computer interfaces that better support human attention and better maintain the fluency of work. To accomplish this we propose to use semi-transparent user interface objects. This paper reports on an experimental evaluation which provides both valuable insights into design parameters and suggests a systematic evaluation methodology. For this study, we used a variably-transparent tool palette superimposed over different background content, combining text, wire-frame or line art images, and solid images. The experiment explores the issue of focused attention and interference, by varying both visual distinctiveness and levels of transparency.