The impact of hyperlinks, skim reading and perceived importance when reading on the Web

Abstract

It has previously been shown that readers spend a great deal of time skim reading on the Web and that this type of reading can affect comprehension of text. Across two experiments, we examine how hyperlinks influence perceived importance of sentences and how perceived importance in turn affects reading behaviour. In Experiment 1, participants rated the importance of sentences across passages of Wikipedia text. In Experiment 2, a different set of participants read these passages while their eye movements were tracked, with the task being either reading for comprehension or skim reading. Reading times of sentences were analysed in relation to the type of task and the importance ratings from Experiment 1. Results from Experiment 1 show readers rated sentences without hyperlinks as being of less importance than sentences that did feature hyperlinks, and this effect is larger when sentences are lower on the page. It was also found that short sentences with more links were rated as more important, but only when they were presented at the top of the page. Long sentences with more links were rated as more important regardless of their position on the page. In Experiment 2, higher importance scores resulted in longer sentence reading times, measured as fixation durations. When skim reading, however, importance ratings had a lesser impact on online reading behaviour than when reading for comprehension. We suggest readers are less able to establish the importance of a sentence when skim reading, even though importance could have been assessed by information that would be fairly easy to extract (i.e. presence of hyperlinks, length of sentences, and position on the screen).

Fig 1. Passage Type x Length of Sentence x Position on Page interaction for the importance ratings of the unlinked sentences in Experiment 1.

Each box on the graph represents the different positions a sentence can sit on the page, from top to bottom (left to right on the graph). A 95-percent confidence interval (the grey shaded region) is drawn around the estimated effect.

Fig 2. Length of Sentence x Position on Page interaction for the linked passages in Experiment 1.

The lines on the graph represent the different positions a sentence can sit on the page. A 95-percent confidence interval (the grey shaded region) is drawn around the estimated effect.

Fig 3. Sentence Contains Links x Position on Page interaction for the linked passages in Experiment 1.

The lines on the graph whether the sentence contains links (dashed line) or not (solid line). A 95-percent confidence interval (the grey shaded region) is drawn around the estimated effect.

Fig 4. Length of Sentence x Position on Page interaction for all sentences that contain links in Experiment 1.

The lines on the graph represent the different positions a sentence can sit on the page. A 95-percent confidence interval (the grey shaded region) is drawn around the estimated effect.

Fig 5. Length of Sentence x Position on Page x Number of Links interaction for all sentences that contain links in Experiment 1.

The lines on the graph represent the number of links a sentence contains. Each box on the graph represents the different lengths of sentence from shortest to longest (left to right on the graph). A 95-percent confidence interval (the grey shaded region) is drawn around the estimated effect.

Fig 6. Number of Links x Position on Page x Sentence Length interaction for all sentences that contain links in Experiment 1.

The lines on the graph represent the different positions a sentence can sit on the page. A 95-percent confidence interval (the grey shaded region) is drawn around the estimated effect.

Fig 7. Task Type x Importance interaction for total sentence reading time in Experiment 2.

A 95-percent confidence interval (the shaded region) is drawn around the estimated effect.

Fig 8. Task Type x Importance interaction for total sentence reading time per character in Experiment 2.

A 95-percent confidence interval (the shaded region) is drawn around the estimated effect.

Fig 9. Task Type x Importance interactions for first pass wrap up reading time in Experiment 2.

The lines on the graph represent the different tasks. A 95-percent confidence interval (the shaded region) is drawn around the estimated effect.

Fig 10. Task Type x Importance interaction for wrap up total time in Experiment 2.

The lines on the graph represent the different tasks. A 95-percent confidence interval (the shaded region) is drawn around the estimated effect.