Usability Testing

The first observation was that all readers started by reading the whole abstract, that is, all the abstract nodes before they expanded the text in any of the nodes.

What they then had to do was to press "next chapter" until the last abstract node, and then they needed to use the navigation menu or press the "back" button of the browser several times to get back to the first page.

To help users to follow this preferred reading pattern, I put in a link at the last of the abstract level notes, labelled "Start over", placed at the same position where the "Next chapter" links had been in previous nodes.

Initially I had made the menu of navigation links in the left column stretch with the text. On the abstract-level pages, there were only links to other pages on that level, while the short-paper nodes had links to all the pages in the hypertext.

The tests showed very quickly that this was a bad design decision. When expanding a chapter, that is, moving from an abstract-level node to the corresponding short paper-level node, users got confused. They did not recognise the new, longer menu as a stretched version of the previous, as they did not recognise the abstract-level pages in the longer menu.

This confusion was made worse by the fact that most users did not use the "next page" and "previous page" links to read, but instead used the navigation links to the left, clicking their way down the table of contents. This habit was disturbed when all of a sudden the menu was collapsed again when moving into a new chapter.

Nielsen and Hoa Loranger repeatedly state that users will bring with them habits formed at other web sites, so designers should stay with conventions. My experience is an example of this logic.

Several users also browsed the longer menu to decide whether to read a chapter or not. When asked after the reading session, they believed that a stretched-out, full menu on the abstract-level pages would make it easier to decide whether to read the chapter, and also which parts of the chapter that seemed interesting.

The easy solution was to put the full list of links to all pages in the hypertext on every page.

[Illustrasjon: Testversjon 1.png, Testversjon 2.png]

When moving from the abstract-length level to the short paper level, the user moves from one HTML page to another. When expanding select paragraphs of the shortpaper level to full paper length, however, new paragraphs appear in the middle of the text, using a simple JavaScript and Cascading Style Sheets (CSS).

To make readers understand that the section was an inset resulting from the stretch, the paragraphs were set in smaller type, and inset from the margins. Several of the testers did not understand this, however. Although they often saw how the text "jumped" on the screen, more than one tester thought the inset meant that the text was a lengthy quote from another text. Often, they did not realize the topical connection with the preceding paragraph, and thus did not understand the stretching/shrinking structure.

To remedy this problem, the full paper length paragraphs were given a distinct background color, an the expand/collapse link was decorated as a button with the same color, making the text visibly connected with the button. After this change, the testers had no misunderstandings of what level they were reading.

[Testversjon 2.png, Testversjon 3.png]

[ferdigversjon1.png, ferdigversjon 2.png]

Barbara Minto argues that there are two basic ways of relating points to a larger category: a formal syllogistic argument with premisses and conclusion; or a simple list, a numbering of parts. The latter seems by far to be the most common, even in scholarly writing. When offereing a list or numbering of parts, Minto recommends to begin by summing up the list, and advice that was confirmed by my test users.

in pages where the full paper-length paragraphs were summed up in bulleted lists in the short paper version, users not only understood the structure of the page, they also had a very clear understanding of what the hidden paragraphs would contain. In other pages, where the maint points of the hidden paragraphs were not summed up in advance, readers were more confused, and more often misunderstood what a hidden paragraph might contain.

In fact, the technique proved so effective that it turned into an expectation. In one or two places, the short paper version had bulleted lists that were not repeated and expanded on in the longer version. Readers were confused by this, they were looking for the bulleted points in the longer version, and lost the flow of the text.