5. Conclusions

This experiment has revealed the following significant findings:
  • For large overviews, browser type affects search time
  • For large overviews, browser type affects subjective difficulty
  • For all overviews, browser type affects user satisfaction
  • For all overviews, browser type affects subjective predictability
We believe each of these findings is relevant except for subjectivedifficulty. Because of the ambiguous nature of the question asked, itsresults are not valid and will not be included in this discussion.

Because the browser type affects search time on large overviews, one caneasily see that current theory must be refined. Because each image usedin this study was closely related, parallel processing was notpossible. Current theory states that there is an optimum overview size;however, our results, along with those of previous experiments, suggestthat an optimum overview size is dependent on the similarity of theimages in a given database. The exact size has not been determined bythis study and therefore warrants further research. One should alsonote that this has an impact for current browser design. At a minimum,browser designers must allow for control of thumbnail size because inmost circumstances, it is unknown what type of database users willbrowse.

The velocity slider, when compared with the other two browsers, wasliked the least. However, the mean task performance time and errorrates on the velocity slider was certainly within the ranges of thewidely used scrollbar and button interfaces; in fact, both then meantask performance times and error rates of the velocity slider were belowthose of the scrollbar interface. Since the tested subjects had noprevious experience the velocity slider, it is reasonable to contendthat upon more frequent usage, users might learn the time benefits ofthe velocity slider. This might compel practitioners to consider usingthe velocity slider in the regular commercial market.

The results for the velocity slider calls for further research into thisparticular tool. Similar experiments could be done whereby subjectsundergo longer training times of more than a week in using the velocityslider. It is contended that as users become more familiar with thetool their performance times will decrease and their subjectivesatisfaction will increase.

It can be concluded that the scrollbar interface used in this experimentshould be avoided given reasonable alternatives. This is because themean task time and mean error rates for this interface proved to be thehighest among the tested interfaces; also, users generally expressedhigher dissatisfaction with this interface compared with the buttoninterface. This is not intended to imply that scrollbars areunfavorable, but that the scrolling interface itself is. This studybroke the scrollbar interface into its two components, a previous/nextwidget, and a scrolling widget; and, the statistics imply that thescrolling widget provides longest mean search time. Therefore, thefollowing statement can be made: the scrollbar provides the user withthe ability to browse the content of an image database instantly,however, due to the concentration required to manipulate the scrollingwidget, it proves to be an inferior tool if scrolling is used as theprinciple means of navigation.

This experiment also confirmed that the button-based browser interfaceis the most efficient and user friendly. Subjects generally preferredthe button-based interface to the others. Subjects also displayed thelowest mean error rates and fastest task completion times when using thebutton interface; therefore, a general case theory can once again bedefined. It is contended that the most efficient and well-likedinterface is one that requires the least amount of user manipulation andconcentration. Such a design frees the user from the burden of thecontrol implementation, and allows him/her to concentrate on the task athand.