Hypotheses We had two hypotheses: (H1) Participants would be able to perceive and memorize changes better in the phosphor con- dition. (H2) We expected the increased performance to cause participants to subjectively prefer the phosphor inter- face. We had no particular hypothesis about the impact of playback speed.
We analyzed the data for this experiment at the summary level, taking the median of the completion times and the mean of the errors over the 8 trials in each condition.. The dependent variables were error rate and task completion time. We analyzed each variable using a 2 (interface) × 3 (playback speed) repeated measures analysis of variance (RM-ANOVA).
For task completion time, we observed a significant main effect of interface (F(1,11)=20.07, p=.001), with users completing the task faster with the phosphor interface (Med=7.17 s) than with the control interface (Med=9.35 s).
We also observed a significant main effect of playback speed (F(2,22)=4.06, p=.031). Paired comparisons using Bonferroni corrections showed that this difference was driven by significant differences between the fastest condi- tion (Med=7.34 s) with the slowest condition (Med=9.01 s, p=.02). We observed no significant interactions between the two factors.
10 8 6 4 2 0
slow medium fast
Figure 15: Task completion times of the first user study. Participants performed the undo task faster when widgets were provided with a phosphor effect (times in seconds, +/- standard error of mean).
We observed no significant effects when we examined error rates. Users made on average about 3.1 errors, i.e., they were able to undo about half of the observed changes, with no significant differences across conditions. Hence, while Phosphor seemed to help with the speed at which users were able to retrieve answers from memory, our task and error rate metric was not sensitive enough to pick up differ- ences when using this interface.
At the end of the experiment, participants answered a ques- tionnaire. All 12 participants preferred the phosphor inter- face to the control interface. On a 7-point Likert scale, par- ticipants agreed that the phosphor interface helped them remember more changes (mean value: 5.9). Participants
also expressed that it did not take long to get used to the phosphor interface (mean value: 2.6).
Discussion The timing data supports our first hypothesis. We believe that the difference in completion times represented the level of confidence that users had in their responses. When they used the Phosphor interface, they were more certain of the answers and spent less time thinking when undoing changes. This performance difference also manifested itself in participants’ subjective preference.
Interestingly, participants performed faster in the faster playback conditions. One possible interpretation is that the reduced time spent watching the actions helped participants keep the actions in their working memory.
USER STUDY 2: COMPARISON WITH ANIMATION The purpose of the second user study was to compare phos- phor with animated transitions. While phosphor effects have the benefit of not introducing lag, we were wondering if this benefit would come at the expense of reduced task performance when compared with animation.
We also used this experiment to learn more about how mul- tiple simultaneous transition effects and distractors impact user performance and user preference.
Tasks This study simulated the situation of a user who has copied and pasted one or more files into a folder window and who now tries to visually verify whether the expected action has taken place.
Each trial proceeded as follows. (1) A simulated windows desktop screen was displayed. The screen contained 11 small windows and 25 icons of the same file type. (2) For 1200ms, the interface highlighted three icons and one target window as shown in Figure 16a. (3) The highlights were removed to prevent users from completing the task by sim- ply tracking the highlights. There was a 500ms pause. (4) Three icons moved across the screen. (5) Participants pressed the “Y” key if they felt that the shown transition corresponded to the highlighting showed earlier; otherwise they pressed the “N” key. The answer keys could be pressed as soon as the transition started and participants were encouraged to press the correct key as soon as they knew the answer.
There were four possible outcomes for each trial, each of which occurred equally often. Correct: All three icons suc- cessfully reached the target window (Figure 16b). Er- rorNeighbor: An incorrect icon moved; it was located closely to the expected one. ErrorOther: An incorrect icon moved; it was located far away from the expected one. Er- rorUndershoot: The right icons moved, but one of them did not reach the target folder because it over or undershot.
There were three versions of this task called singleIcon, tripleIcon, and distractor. The tripleIcon task was de- scribed above and shown in Figure 16. The singleIcon task was identical to tripleIcon, but only a single icon was high- lighted, moved, and needed to be verified. The distractor