Rapid Visual Flow: How Fast Is Too Fast? Andrew Wallace Joshua Savage Andy Cockburn Department of Computer Science, University of Canterbury, Christchurch, New Zealand {ajw125, jps42}@student.canterbury.ac.nz andy@cosc.canterbury.ac.nz Abstract It is becoming increasingly common for user interfaces to use zooming visual effects that automatically adapt to user actions. The MacOs X ‘dock’ icon panel, for instance, uses a fisheye dis- tortion to assist users in targeting items. Another example is ‘speed-dependent automatic zooming’, which has been shown to improve scrolling by automatically varying zoom level with scroll speed—when scrolling fast the document is zoomed out, but when scrolling slowly the document is fully zoomed in. When implementing automatic zooming interfaces, designers must calibrate the behaviour of their zooming systems so that the visual effects allow rapid navigation without stressing the human visual system. At present, these calibrations are de- rived from trial and error. This paper describes an attempt to determine metrics of visual flow to answer the question “how fast is too fast”? Our main focus is on automatic zooming in document scrolling tasks. We performed an experiment to mea- sure participants’ preferred and maximum-tolerable scrolling speeds at two different magnifications. We found that magnifi- cation affected the length of time that data needed to remain on screen. We also used the data to provide estimations regard- ing the appropriate calibration of threshold values in speed- dependent automatic zooming systems. Keywords: Visual flow, scrolling, zooming, magnifica- tion, speed. 1 Introduction Scrolling is one of the most fundamental activities in computer use. In an analysis of five-hours of web use, Byrne, John, Wehrle & Crow (1999) observed that users spent 40 minutes scrolling. They commented that “An obvious case where widget design could make a difference is scrolling”. Igarashi & Hinck- ley (2000) identified the problem that scrolling too fast results in the information blurring (see Figure 1). Motion blur can cause disorientation, and reduces the users ability to determine whether they have reached their target. Igarashi & Hinckley proposed speed-dependent au- tomatic zooming (SDAZ) as a solution to the prob- lem of motion blur when scrolling rapidly. SDAZ au- tomatically alters the zoom level, based on scrolling speed. The document is smoothly zoomed out when the user’s scrolling speed exceeds a predefined thresh- old. The resulting lower magnification means that information need not be moved so rapidly across the screen to achieve the same comparative rate of naviga- Copyright c 2004, Australian Computer Society, Inc. This pa- per appeared at the 5th Australasian User Interface Conference (AUIC2004), Dunedin. Conferences in Research and Practice in Information Technology, Vol. 28. A. Cockburn, Ed. Reproduc- tion for academic, not-for profit purposes permitted provided this text is included. Figure 1: Scrolling too fast causes intolerable blur- ring. (Image generated artificially using The Gimp.) tion through the document, thereby reducing the mo- tion blur. A video demonstrating several SDAZ sys- tems is available at www.cosc.canterbury.ac.nz/ ~andy/sdaz.mov In order to calibrate an SDAZ system, a value needs to be set for the threshold. Igarashi & Hinck- ley refer to this threshold value as one of several “predefined constants” used in their equations, but do not provide the values of these constants. Cock- burn & Savage (2003) in their analysis of SDAZ also mention that they use a threshold, but similarly say nothing about the specific value used. Anecdotal ev- idence from the designers and implementers of SDAZ systems suggests that a large proportion of the im- plementation effort in constructing SDAZ systems is spent in fine-tuning the zooming behaviour through an informal look-and-feel approach. The experiments reported in this paper aim to aid SDAZ designers and implementers by exploring the relationship between zoom level, rate of display move- ment, and user comfort. They also investigate the fundamental question of whether lowering magnifica- tion actually reduces motion blur. The following section briefly describes some of the human-factors research on motion blur and uses these to derive theoretical estimates for the amount of time that scrolling data should be displayed while moving. The evaluation is described in Section 3, and the re- sults are presented in Section 4 prior to a discussion of further work.