Stanford University Home

Stanford News Archive

Stanford Report, May 10, 2000

Eye movement research points to importance of text over graphics on websites

BY KATHLEEN O'TOOLE

A picture is said to be worth a thousand words, but perhaps not on the Internet, at least as we know it today.

Text plays a more important role than graphics as entry points for online news, according to preliminary analysis of research conducted at Stanford and the Poynter Institute, a Florida-based nonprofit teaching and research institution on journalism. This contrasts with an earlier Poynter study, which found that readers of print newspapers looked first at the lead art element on a newspaper page and then moved their eyes to the biggest headline.

Using advanced eye-movement interpretation technology developed at Stanford's Center for the Study of Language and Information (CSLI), the researchers in the current study tracked the online news reading patterns of 67 people in the Chicago and St. Petersburg, Fla., areas. The participants used their own definitions of news and selected the sites they wanted to read from their own list of web page bookmarks.


Participants in the Stanford-Poynter project were set up with head gear to track their eye movements as they read online news.


"General news providers should be encouraged by our results, because we found readers are paying attention to general news on the web, and not just reading about their particular interests," said Marion Lewenstein, Stanford professor emerita of communication and the principal investigator on the project. "Serendipity still plays a role in what people read. They look around a web news page, and when their eye catches something, they read it."

The Stanford-Poynter study found online readers often fixating first on news briefs (short summaries of stories) or captions. In analyzing the eye-movement patterns, the researchers noticed that most readers' eyes shifted to the photos and illustrations on the screen, but only after clicking on a brief to get to a complete article and then returning to the page with the illustration or photograph.

The difference cannot be totally explained by the longer time it takes to display some graphics on web pages, Lewenstein said, because the readers in the study were using fast computers on fast networks where the delay in graphics loading was about 2 seconds or less.

Preliminary results also indicate that online readers were willing to scroll down screen pages to read a long story. "Of the stories our web users called up, there was vertical reading behavior down to at least 75 percent of the length of the page," Lewenstein said. "In the 10-year-old study of people reading print newspapers, people on average read only about 20 percent of a story."

The conclusions are preliminary because the researchers intend to do more analysis and post additional results on the Poynter website as they become available. Because the human eye rarely fixates on a point for more than a second, it requires an enormous amount of data to analyze for meaningful patterns. The 67 participants, who read news for an average of 34 minutes each, tallied up a total of 608,063 eye fixations and 24,530 mouseclicks.


Related Information:


Eye-aware software

The participants wore a lightweight head-mounted device originally developed by the military for tracking eye movements and later adapted by market researchers. Using software developed under the direction of Gregory Edwards, a researcher at Stanford's CSLI, the research team was able to computer-record the participants' eye movements and keyboard strokes as well as the screen images displayed to them.

In addition to Edwards and Lewenstein, the research team included psychologist Deborah Tatar, a CSLI research associate, and Andrew Devigal, a Poynter fellow and expert in online design.

When people are reading English, their eye fixations follow a "right, right, sudden-left" pattern that can distinguish reading from other types of looking behavior, Edwards said. While the current study is one way to apply the technology, he and other CSLI researchers say they also hope to develop "eye-aware" software that people can use to control technology. As examples, they suggest an elevator in which riders can select a floor simply by looking at the floor number on an information panel, a phone support application that eases the workload of operators, and an "eye-mouse" that would make it possible for a paralyzed person to gain full control of a computer with eye movements.

Study results "positive for democracy"

Participants in the Poynter study were volunteers who responded to online newspaper ads, so the results cannot be generalized to all online readers. Nevertheless, the study enables researchers to draw such general conclusions as the relative importance of text over graphics as entry points, at least among those who are relatively heavy online news seekers.

The results should be encouraging to people who have worried that the new technology might cause people to read only news tailored to their interests, Lewenstein said. Such specialized reading patterns would tend to lead to less well informed voters, a problem for democracies, she said. The readers in this study did look for news on specialized sites, such as those devoted to stock market reports, sporting or entertainment news, and even drugstore sites and corporate press release sites. "But the majority begin their news reading with general news providers, they go back to them more frequently, and spend more total time with them," Lewenstein said. General news providers included local and national newspapers and broadcast providers.

Another indication that people define news as broader than their own interests comes from the attention they gave to opinions expressed on the websites, Lewenstein said. Study participants read at least part of 58 percent of all the brief summaries of editorials, letters to the editor and columnists' views presented on the web pages they looked at. When full opinion articles were presented on a page they called to their screen, 90 percent of the articles had some reading in them.

Not all graphic artists who have heard about the study in seminars believe the findings about text's relative importance to art, Devigal said. Some felt the art is currently less salient because of present screen size and image resolutions.

Lewenstein pointed out that the research team had analyzed the page entry points only for the first page looked at by every fifth subject and some of those pages did not have both art and text for comparison. "We want to check more pages per subject, and perhaps more subjects, to see if this pattern continues to hold up," she said.