Putting Icons in Context: The Influence of Contextual Information on the Usability of Icons

Moyes, Jackie (1995) Putting Icons in Context: The Influence of Contextual Information on the Usability of Icons. PhD thesis, University of Glasgow.

Abstract

Previous research has been driven by the idea that a good icon is one which consists of a form that is instantly recognisable as representing the underlying referent. As a result many of the design decisions suggested have been based round this premiss. However, the empirical method used to support these design decisions has often tended to ignore two factors of everyday interface usage that have an important influence on the ability of an icon to communicate its meaning. The first factor that most researchers tended to ignore was that an icon is usually displayed simultaneously with a number of other icons. Rarely will you find icons displayed in isolation, yet many methodologies ask subjects to judge an icon's ability to communicate its meaning without any additional information from the interface to support it. A second issue is that, in everyday usage, most users interact with icons over an extended period of time. Therefore it is not necessarily essential that the icon's meaning should be easy to guess, only that it be particularly easy to learn. The ultimate aim of this thesis, therefore, is to suggest that icon research should no longer be concerned with measuring the degree to which a particular representational form of a solitary icon can increase performance on a subject's initial exposure to the icon. Instead research should be widened to consider how attributes in the interface interact, and how this interaction may vary over time as user knowledge of the interface increases. A total of six experiments were performed. Each experiment had four conditions, manipulating the abstractness of icon shape and consistency of icon position. After a training period conditions changed without warning and the conclusions depended on whether or not performance was disrupted by the change. Results from the experiments suggested overwhelmingly that issues such as what attributes users relied upon, and when they relied upon them, were far more complicated than initially predicted. The conclusions of the thesis question the generalisability of many long held assumptions about icon design. Findings show that once an icon is placed into an interface it is difficult to predict which of its attributes the user will rely on. It seems that this will not be determined by how representational the attribute is, but rather how discriminate it is in comparison to other attributes, once the icon is placed within a particular interface. Results also show that users are aware of multiple attributes apparent in the interface and may switch the attribute that they rely on several times within the entire user learning curve. However, it is impossible to say how much users learn, since the experimental data has shown that learning appears to continue after the classic performance measures of reaction time and error rate have reached asymptote. Finally, the results highlight the flaws in current evaluation paradigms; namely that most are guilty of testing icons in isolation and only in the initial stages of the learning curve. The experimental results, combined with the extensive literature review in Chapter 2, elicit a number of interesting future research questions, as well as providing designers with some general suggestions as rules-of-thumb for interface design. The thesis concludes by elaborating on these points.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Adviser: Steve Draper
• Keywords: Computer science
• Date of Award: 1995
• Depositing User: Enlighten Team
• Unique ID: glathesis:1995-75488
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Nov 2019 19:39
• Last Modified: 19 Nov 2019 19:39
• URI: https://theses.gla.ac.uk/id/eprint/75488