Web page design, as well as web ad design, is still in its infant stage, more ruled by fashion and popular belief than by principles based on investigations on how users interact with the web. This is no less true of web advertisements. This study, conducted in co-operation with a web ad company, explores the use of eye tracking in web page usability testing. In this paper, we focus on the extent to which Internet users fixate Internet advertisements.

Method: Eight subjects, experience web users, were given the task to find answers to ten questions within 30 min, using the web. They were given seven bookmarked pages as starting points and were allowed to find their own way to the answers. They were not allowed to use the keyboard (excluding the use of search engines). Their eye movements, superimposed onto the computer screen, were recorded on videotape for analysis.

Result: Subjects did not look at web advertisements, not even during download when the colourful, flickering ad was the only thing on the page.

The flickering, colourful ads did not attract the overt attention of our subjects. Ad designers seem to expect that any moving visual object that appears on the screen during the search for relevant information will force the subjects to fixate the ad and then preferably to click it.

It is little doubt that the flickering ads disturbed the subjects. Most of them employed avoidance strategies against ads to achieve top-down control over their visual attention (strategic control). They rotated the mouse cursor in the middle of the empty screen during slow downloads, so as to generate a competing motion to the ad at the top of the page.

Subjects thus wanted to exercise strategic control over both visual selection and visual attention on web pages. To their additional help, we hypothesise, our subjects had experience-based expectations (semantic frames) on the structure of web pages that told them were ads were located, so they could avoid them without fixating them.

Several observations showing the usefulness of eye tracking in web design evaluation were also made, among them:

• All subjects exhibited basically the same search pattern towards the different pages, agreeing on which parts of the page were relevant to them.
• They rather searched along vertical than along horizontal lists.
• Subjects often missed out on information inside longer texts, since they did not read, only scan, it (with the exception of a dyslectic subject who therefore solved the tasks faster than anyone else).
• Picture links were only chosen if they were clear and unambiguous, otherwise subjects looked for an unambiguous text link.
• The mouse cursor was often placed (anchored) at one of the latest points of attended task relevant information, like a flag to mark a point possibly useful to return to, while continuing visual search elsewhere.