article review - 'When Users Do and Don't Rely on Icon Shape'

  • Article

Article title: When Users Do and Don't Rely on Icon Shape

Author: Jackie Moyes

Publication: Conference companion of the SIGCHI Conference on Human Factors in Computing Systems

Year of publication: 1994

In this CHI '94 short paper, the author explores when users rely on icon shape. The test combines two earlier works. Kaptelinin showed that, after training on textual menus, users had problems finding menu entries if the positions were changed; but if the positions were the same and the letters were masked but the menu entries retained their length, users performed as
well as they had with the familiar commands. Blankenberger and Hahn conducted an experiment wherein they found that their users learned the position of icons if the icons were abstract, but that representational icons were easier to find if the icon positions were randomised.

Moyes designed a test that consisted of 8 blocks of 17 trials (one for each of the 17 commands represented by the icons), and four groups of users. The four groups completed the following tests:

  1. Exposed to the representational set of icons, which changed to abstract icons on the 6th block; positions were static.
  2. Exposed to the abstract set of icons, which changed to representational icons on the 6th block; positions were static.
  3. Exposed to representational icons throughout; positions changed on the 6th block.
  4. Exposed to abstract icons throughout; positions changed on the 6th block.

In the test, Moyes found that users who used representational icons relied on the shape, whereas users who used abstract icons relied on the position. Users did not appear to use both shape and position to associate an icon with its command. The author does note that users might use both over longer use, but that is outside the scope of this paper.

Another item that the paper did not take into account is the effect of icon grouping -- does that affect the user's speed and accuracy when finding icons? Does icon grouping have a different effect on representational or abstract icons?

Related links:
ACM Digital Library page
for this paper