HOW USERS READ ON THE WEB

1. Measuring the Effect of Improved Web Writing
2. Why Web Users Scan instead of Read

3. Be Succinct! (writing for the web)

They don't.

People rarely read Web pages word by word; instead, they scan the page, picking out individual words and sentences. In a recent study John Morkes and I found that 79 percent of our test users always scanned any new page they came across; only 16 percent read word-by-word.

As a result, Web pages have to employ scannable text, using

• highlighted keywords (hypertext links serve as one form of highlighting; typeface variations and color are others)
• meaningful sub-headings (not "clever" ones)
• bulleted lists
• one idea per paragraph (users will skip over any additional ideas if they are not caught by the first few words in the paragraph)
• the inverted pyramid style, starting with the conclusion.
• half the word count (or less) than conventional writing

We found that credibility is important for Web users, since it is unclear who is behind information on the Web and whether a page can be trusted. Credibility can be increased by high-quality graphics, good writing, and use of outbound hypertext links. Links to other sites show that the authors have done their homework and are not afraid to let readers visit other sites.

Users detested "marketese"; the promotional writing style with boastful subjective claims ("hottest ever") that currently is prevalent on the Web. Web users are busy: they want to get the straight facts. Also, credibility suffers when users clearly see that the site exaggerates.

MEASURING the EFFECT of IMPROVED WEB WRITING

To measure the effect of some of the content guidelines we had identified, we developed five different versions of the same website (same basic information; different wording; same site navigation). We then had users perform the same tasks with the different sites. As shown in the table, measured usability was dramatically higher for the concise version (58% better) and for the scannable version (47% better). And when we combined three ideas for improved writing style into a single site, the result was truly stellar: 124% better usability.

Site Version	Sample Paragraph	Usability Improvement (relative to control condition)
Promotional writing (control condition) using the "marketese" found on many commercial websites	Nebraska is filled with internationally recognized attractions that draw large crowds of people every year, without fail. In 1996, some of the most popular places were Fort Robinson State Park (355,000 visitors), Scotts Bluff National Monument (132,166), Arbor Lodge State Historical Park & Museum (100,000), Carhenge (86,598), Stuhr Museum of the Prairie Pioneer (60,002), and Buffalo Bill Ranch State Historical Park (28,446).	0% (by definition)
Concise text with about half the word count as the control condition	In 1996, six of the best-attended attractions in Nebraska were Fort Robinson State Park, Scotts Bluff National Monument, Arbor Lodge State Historical Park & Museum, Carhenge, Stuhr Museum of the Prairie Pioneer, and Buffalo Bill Ranch State Historical Park.	58%
Scannable layout using the same text as the control condition in a layout that facilitated scanning	Nebraska is filled with internationally recognized attractions that draw large crowds of people every year, without fail. In 1996, some of the most popular places were: Fort Robinson State Park (355,000 visitors) Scotts Bluff National Monument (132,166) Arbor Lodge State Historical Park & Museum (100,000) Carhenge (86,598) Stuhr Museum of the Prairie Pioneer (60,002) Buffalo Bill Ranch State Historical Park (28,446).	47%
Objective language using neutral rather than subjective, boastful, or exaggerated language (otherwise the same as the control condition)	Nebraska has several attractions. In 1996, some of the most-visited places were Fort Robinson State Park (355,000 visitors), Scotts Bluff National Monument (132,166), Arbor Lodge State Historical Park & Museum (100,000), Carhenge (86,598), Stuhr Museum of the Prairie Pioneer (60,002), and Buffalo Bill Ranch State Historical Park (28,446).	27%
Combined version using all three improvements in writing style together: concise, scannable, and objective	In 1996, six of the most-visited places in Nebraska were: Fort Robinson State Park Scotts Bluff National Monument Arbor Lodge State Historical Park & Museum Carhenge Stuhr Museum of the Prairie Pioneer Buffalo Bill Ranch State Historical Park	124%

It was somewhat surprising to us that usability was improved by a good deal in the objective language version (27% better). We had expected that users would like this version better than the promotional site (as indeed they did), but we thought that the performance metrics would have been the same for both kinds of language. As it turned out, our four performance measures (time, errors, memory, and site structure) were also better for the objective version than for the promotional version. Our conjecture to explain this finding is that promotional language imposes a cognitive burden on users who have to spend resources on filtering out the hyperbole to get at the facts. When people read a paragraph that starts "Nebraska is filled with internationally recognized attractions," their first reaction is no, it's not, and this thought slows them down and distracts them from using the site.

WHY WEB USERS SCAN INSTEAD of READ

More research is needed to truly know why 79 percent of Web users scan rather than read, but here are four plausible reasons:

• Reading from computer screens is tiring for the eyes and about 25 percent slower than reading from paper. No wonder people attempt to minimize the number of words they read. To the extent this reason explains users' behavior, they should read more when we get high-resolution, high-scanrate monitors in five years since lab studies have shown such screens to have the same readability as paper.
• The Web is a user-driven medium where users feel that they have to move on and click on things. One of our users said: "If I have to sit here and read the whole article, then I'm not productive." People want to feel that they are active when they are on the Web.
• Each page has to compete with hundreds of millions of other pages for the user's attention. Users don't know whether this page is the one they need or whether some other page would be better: they are not willing to commit the investment of reading the page in the hope that it will be good. Most pages are in fact not worth the users' time, so experience encourages them to rely on information foraging. Instead of spending a lot of time on a single page, users move between many pages and try to pick the most tasty segments of each.
• Modern life is hectic and people simply don't have time to work too hard for their information. As one of our test users said, "If this [long page with blocks of text] happened to me at work, where I get 70 emails and 50 voicemails a day, then that would be the end of it. If it doesn't come right out at me, I'm going to give up on it."

BE SUCCINCT! (WRITING FOR THE WEB)

The three main guidelines for writing for the Web are:

• Be succinct: write no more than 50% of the text you would have used in a hardcopy publication
• Write for scannability: don't require users to read long continuous blocks of text
• Use hypertext to split up long information into multiple pages

Short Texts

Reading from computer screens is about 25% slower than reading from paper. Even users who don't know this human factors research usually say that they feel unpleasant when reading online text. As a result, people don't want to read a lot of text from computer screens: you should write 50% less text and not just 25% less since it's not only a matter of reading speed but also a matter of feeling good. We also know that users don't like to scroll: one more reason to keep pages short.

The screen readability problem will be solved in the future, since screens with 300 dpi resolution have been invented and have been found to have as good readability as paper. High-resolution screens are currently too expensive (high-end monitors in commercial use have about 110 dpi), but will be available in a few years and common ten years from now.

Scannability

Because it is so painful to read text on computer screens and because the online experience seems to foster some amount of impatience, users tend not to read streams of text fully. Instead, users scan text and pick out keywords, sentences, and paragraphs of interest while skipping over those parts of the text they care less about.

Skimming instead of reading is a fact of the Web and has been confirmed by countless usability studies. Webwriters have to acknowledge this fact and write for scannability:

• Structure articles with two or even three levels of headlines (a general page heading plus subheads - and sub-sub-heads when appropriate). Nested headings also facilitate access for blind users with screenreaders
• Use meaningful rather than "cute" headings (i.e., reading a heading should tell the user what the page or section is about)
• Use highlighting and emphasis to make important words catch the user's eye. Colored text can also be used for emphasis, and hypertext anchors stand out by virtue of being blue and underlined

Hypertext Structure

Make text short without sacrificing depth of content by splitting the information up into multiple nodes connected by hypertext links. Each page can be brief and yet the full hyperspace can contain much more information than would be feasible in a printed article. Long and detailed background information can be relegated to secondary pages; similarly, information of interest to a minority of readers can be made available through a link without penalizing those readers who don't want it.

Hypertext should not be used to segment a long linear story into multiple pages: having to download several segments slows down reading and makes printing more difficult. Proper hypertext structure is not a single flow "continued on page 2"; instead split the information into coherent chunks that each focus on a certain topic. The guiding principle should be to allow readers to select those topics they care about and only download those pages. In other words, the hypertext structure should be based on an audience analysis.

Each hypertext page should be written according to the "inverse pyramid" principle and start with a short conclusion so that users can get the gist of the page even if they don't read all of it.