A Review of Beyond ALT Text: Making the Web Easy to Use for Users with Disabilities
Authors: Pernice Coyne and Jakob Nielsen
Fremont, California: Nielsen-Norman Group, 2001. PDF document.
148 Pages PDF format
Download site: DigiBuy (7 MB)
$190 for a single report, $390 for the report and the right to make copies within your organization
Summary of Beyond ALT Text
Beyond ALT Text is a 140-page report that details the results of Web usability testing that involved subjects with disabilities. Conducted by the leading Web usability firm in the U.S., the tests involved a total of 104 subjects and 19 Web sites. The subjects included people who are totally blind, people with low vision, and people with impaired mobility. There was also a control group of 20 individuals without disabilities. The subjects with disabilities used a variety of assistive or adaptive technologies. For example, subjects who were blind used either screen readers or refreshable Braille displays or both. Subjects with low vision used screen magnification software (supplemented in some cases by synthetic speech) or enlarged the display fonts on their computers. Subjects with mobility impairments used a number of different devices, including trackballs and footpads and other alternative input devices such as modified keyboards.
Most Valuable Things
Validation of claims made by disabled users and accessibility community about the pervasive inaccessibility of the Web: First, Beyond ALT Text offers vindication for people in the disability and accessibility communities, who have been saying for a long time that the Web is not very accessible for people with disabilities. That's not news, as the report acknowledges. What is news, though, is to have independent confirmation from the leading Web usability expert in the United States, who writes that he and his colleagues "were surprised by how poorly designed the Web is for people with low or no vision" (p. 10). Thank you for that, Dr. Nielsen!
Quantifying the Scale of the Problem
Perhaps even more important, this report gives us a way to begin measuring the scale of the problem. When Congress amended Section 508 of the Rehabilitation Act (more commonly known as Section 508) in 1998, accessibility was defined in operational terms. To paraphrase slightly, we can say that a site is accessible when people with disabilities can access and use it as effectively as people without disabilities. In other words, the goal is parity between the experiences of people with and without disabilities.
Beyond ALT Text gives us, for the first time, a measure of how great the disparity between these two groups is.
Users with disabilities were about three times less likely to succeed in carrying out such routine Web tasks as searching for information and making purchases as users in the control group. (p. 3)
That's a huge discrepancy-and it's not as though the control group of users without disabilities did all that well either. The control group succeeded 78.2% of the time, as against about 26% for the users with disabilities (p. 4). Twenty-six percent is just plain embarrassing. But the figures for people using screen readers and screen magnifiers are even worse: 12.5% and 21.4%, respectively (p. 4). The authors are careful to point out that these figures do not reflect incompetence or inexperience on the users' part: test participants who were blind had been using computers and assistive technology for more than three years (p. 127), and many of them are employed as knowledge workers (p. 124). But even 78% is only a C+ where I come from. Nothing to write home happily about. And 78% is high, Nielsen says-typically, usability studies find that the success rate is between 40 and 60%. That's an F. The good news is that there's really no place to go but up.
Observations about how People use Screen Magnification Software and the Problems they Encounter
The devil's in the details, and the report provides compelling detail about the way people with low vision use screen magnification software and the problems they encounter.
The most severe problems stem from loss of context (p. 23). At 6x magnification, for example, you may see only a few letters at a time, or only part of a button. You might see the button but not the text above it or next to it that tells you what it's for.
There is an excellent description of what happens when someone using a screen magnifier runs into a rollover button (where passing the mouse over an image causes some previously hidden text, or image, to replace it on the screen).
Typically, people using screen magnifiers cannot read rollover text. Often the screen is zoomed in so much that they see only small segments. When they move their mouse over an image and a description pops up, a few confusing things can happen:
- They are not sure what is happening at all,
- or they can only read part of the text, and
- if they move their mouse so as to see more of the zoomed-in screen, they are no longer hovering over the image and the rollover text disappears. (p. 61)
If this isn't reason enough to avoid using rollovers, consider this: the "text" that replaces the rollover button isn't really text. It's an image of text. That means screen readers can't touch it unless the developer provides ALT text that matches what's visible on the screen. And rollovers typically assume that you're using a mouse-not a valid assumption, as Nielsen points out. Hence the guideline: "Do not rely on rollover text to convey any information" (p. 61).
For similar reasons, Nielsen advises against another popular technique: "Avoid using cascading menus," he warns (p. 63).
Observations about Problems that Tight Spacing of Buttons and Link Text Creates for People with Low Vision and/or Limited Hand-Mobility
Beyond ALT Text also notes that busy screens crammed with tightly packed buttons or text links create major headaches both for people with low vision and for people who have difficulty making precise hand-motions.
The headaches are often literal for people with low vision. (Not everyone with low vision uses a screen magnifier: some can't afford the products, which typically cost $500-$800, while others simply prefer not to.) Nielsen's informants talk about headaches, eyestrain, and shoulder and neck fatigue that come from the effort to make out text and graphics while sitting extremely close to the screen (p. 22). Even from such close range, Coyne and Nielsen report, many users with low vision were unable to select the link they wanted from a tightly spaced menu.
The problem was even more severe for people with conditions such as Cerebral Palsy, who often use a trackball or even a foot-operated trackball and find it extremely difficult to make the small, precise movements required to select a link or button from a closely grouped set. Some users had to try between three and 10 times before they succeeded in selecting the link or button they wanted (p. 27).
It may not be easy to find effective solutions. More "padding" between buttons and links would be useful for people with motor control issues and people with low vision who don't use screen magnifiers. But it might create problems for people who use screen magnifiers at high power, who may not realize that a button (or part of a button) belongs to a group. And spacing alone won't help people who use screen readers, which typically ignore white space. (So the white space that this report recommends using to separate adjacent links should be supplemented by a printable, speakable character such as the vertical bar or pipe (|) often used for this purpose.)
Observations about Color Problems Caused by Screen Magnifiers
Coyne and Nielsen note that subjects who used screen magnifiers changed color combinations "all the time," switching back and forth between color and black and white or inverting colors for better contrast. They compare this practice to the way sighted users scroll (p. 23). Of course this practice plays havoc with the color-schemes Web designers are at such pains to select. But it doesn't stop there: screen magnifiers can also affect the ability to recognize link text because the default royal blue doesn't show up when magnified (p. 95).
Observations about the Way People Using Screen Readers use Browser Search, Search Engines, and Link Lists
There are also some good observations about the way people use screen readers. Although I've been using a screen reader daily for the past five and a half years, for example, it hadn't occurred to me that I might use "FindÖ" more often than other users do. But the test subjects who used screen readers and screen magnifiers used the browser "FindÖ" feature more frequently than users in the control group (p. 17). The report states, too, that more experienced screen reader users also tended to skip around more, relying on the links-list features of their screen readers or reading only the first line of a paragraph before moving on (p. 19).
A useful guideline emerges from these observations: the authors advise putting a search box at the top of the screen-preferably in the upper left or upper right corner, where users are accustomed to finding such things. And, after observing that screen reader users sometimes don't know how to spell words or phrases they've only heard spoken by the screen reader's robotic voice, Beyond ALT Text advises using search engines that support spell-checking and are generally "forgiving of error" (p. 99). This would be useful for people with dyslexia, of course-and for a very large part of the general population!
Guidelines about Forms
Beyond ALT Text makes important points about Web-based forms, too. Subjects who used screen magnifiers were often unable to see input fields on the right side of the screen, resulting in submission of incomplete forms, which in turn led to error messages, backtracking, frustration, and fatigue. As a result, Beyond ALT Text makes what will probably be a controversial recommendation to stack form controls vertically (p. 86) rather than using the conventional horizontal layout. This recommendation is likely to increase the amount of scrolling required to complete a form, however, so it may be difficult to follow it while also heeding the advice to minimize the need for scrolling (p. 34).
Demonstration that it's Possible to Create an Accessible PDF Document
On an altogether different note, Beyond ALT Text provides an object lesson on the possibility of creating reasonably accessible PDF documents. The report contains text, graphics, tables, and internal and external links. The graphics aren't readable, of course, and the tables are difficult to follow (though no more difficult than they would be if I had had my Kurzweil 1000 scan and read a printed book). But JAWS 4.00 reads the text (it seems to have some trouble with quotation marks, which it reports as question marks instead), and the links appear to work.
This is important: PDF has been (and continues to be) a huge accessibility problem, especially for people who rely on screen readers-only the most recent versions of the Acrobat Reader (5.0 and later) are accessible at all, and they work only with recent screen readers. There have been few examples of accessible documents to point to, and this one is a welcome addition to the library.
Most Problematic Things
But there are some trouble spots, too. The report jumps to conclusions too quickly in some places. The authors have not taken the trouble to correlate their guidelines with published accessibility guidelines and standards. There are some quality control problems. And the price is far too high.
Jumping to Conclusions
"The best thing for all users," write Nielsen and Coyne, "whether they use assistive technology or not, is to minimize the use of graphics, increasing the speed of page load-time and decreasing superfluous noise." (p. 42). This might sound plausible-it would certainly save a lot of trouble for people who are blind, not to mention that it would eliminate the need to write ALT text for all those images! But it would make for a bland, text-only world for many users, and it might not be the best thing for everyone after all.
There are Web users for whom graphics are essential aids to understanding, including people with certain cognitive disabilities and people with limited proficiency in the native natural language of the document. There are also cases where graphics might be used effectively to help people with low vision locate critical information.
In any case, the content of Beyond ALT Text makes it difficult to take the advice to "minimize graphics" all that seriously. The report, which as a PDF document is meant to be read online if not on the Web, is well stocked with graphics, including many screenshots and many charts and graphs. As a visually impaired reader, I would have liked more detailed descriptions of some of those screenshots-it wasn't always easy to tell how the features were arranged on the page, so it wasn't as easy as I'd have liked to understand why the page had been so difficult for people using screen readers.
The report also goes overboard about the use of layout tables. The first recommendation about tables is to "avoid using tables for aesthetic design" (p. 110). It's true that layout tables can cause serious problems for people using screen readers. But many of these problems can be avoided by designing those tables carefully-which means thinking through the order in which the elements on the page will be spoken by screen readers. There are some excellent explanations of how screen readers "linearize" the contents of tables and what happens when they run into nested tables (see, for example, Jim Thatcher's "Web Accessibility for Section 508" at http://www.jimthatcher.com/webcourse1.htm). And there are tools such as The WAVE and Lynx Viewer to help developers make sure that the reading-order makes sense. (You can try The WAVE at http://www.temple.edu/inst_disabilities/piat/wave/. Lynx Viewer is available at http://www.delorie.com/web/lynxview.html.)
In these instances, Beyond ALT Text seems to confuse bad design with a problem inherent to screen readers or other assistive technologies. But the solution to bad design isn't (necessarily) to ban certain practices because some people handle them badly. The solution is good design. And good design is accessible design.
Failure to Cross-Reference Guidelines and Comments with WCAG 1.0 and Section 508
Beyond ALT Text is an important document. But its value is compromised by a consistent failure to cross-reference its findings and guidelines with the existing accessibility guidelines and standards. The Web Content Accessibility Guidelines 1.0 (WCAG) were published in May 1999 by the World Wide Web Consortium's Web Accessibility Initiative. WCAG is the basis for accessibility standards now being set by nations around the world. WCAG is also at the heart of the federal Web accessibility standards established under Section 508 of the Rehabilitation Act.
There are a number of instances where Coyne and Nielsen's recommended guidelines coincide with either WCAG or Section 508 or both. For example, Beyond ALT Text recommends positioning labels for form elements as close as possible to the controls they identify, making it easier for people using screen magnifiers to see the relationship. The recommendation reads as follows: "> PUT TEXT FOR FIELD LABELS VERY CLOSE TO THE ACTUAL FIELD" (p. 83). Here the report misses a valuable opportunity to draw strength from and to reinforce WCAG Checkpoint 10.2, which is quoted below:
Until user agents support explicit associations between labels and form controls, for all form controls with implicitly associated labels, ensure that the label is properly positioned. [Priority 2]
The label must immediately precede its control on the same line (allowing more than one control/label per line) or be in the line preceding the control (with only one label and one control per line).
But positioning isn't always good enough. There are well-documented techniques for ensuring that screen readers and talking browsers can recognize the connection between form elements and their labels. WCAG Checkpoint 12.4 addresses this concern:
"Associate labels explicitly with their controls. [Priority 2]
For example, in HTML use LABEL and its "for" attribute."
There are also instances where Coyne and Nielsen's guidelines actually go beyond WCAG and 508; it's worth bearing in mind that Section 508 establishes minimum accessibility standards, not the be-all and end-all. Two of his other recommendations about forms are cases in point: "LIMIT THE AMOUNT OF INFORMATION THAT FORMS REQUIRE; COLLECT ONLY THE MINIMUM NEEDED" (p. 83). Coyne and Nielsen also suggest grouping all optional items at the bottom of the form to reduce the amount of typing and tabbing that users with limited use of their hands are required to do (p. 82).
These are excellent points. Noting correlations with and amplifications of WCAG checkpoints and 508 standards would have been a valuable service for the entire Web community. Simply referring readers to the published guidelines and standards and saying we ought to follow them isn't the same thing at all.
Sloppy Copyediting and Proofreading
Beyond ALT Text also suffers from sloppy copyediting and proofreading. There are lots of minor instances, but I'll cite just one deliciously ironic example. Noting that people who use screen readers sometimes "flinch" or replay text if they encounter unusual punctuation, the report says: "Screen reader users hear sentence punctuation such as coma [sic] and period." (98) Indeed we do! And misspellings, too: coma for comma is absolutely wonderful. I didn't just flinch, I replayed the text to make sure I wasn't hallucinating, and then I hooted. I almost hate to point it out.
I might not have pointed it out, either, were it not for two things:
- Screen readers are amazing tools for proofreading. As a screen reader user (and English professor) I can't help but hear typos, punctuation problems, misspellings, and misused words (intermittent for intermediate on p. 80, for example). And
- then there's the price of this report. The Nielsen-Norman group charges $190.00 for a single electronic copy, and $390.00 for an electronic copy that can be reproduced digitally or on paper. (We paid $390.00.) Given that this publishing strategy eliminates all the production and distribution costs associated with paper-based reports, customers have a right to expect quality workmanship, especially from the leader in the usability industry.
A final word about price. The price for this report isn't just high. At $190.00 for a single copy and $390.00 for the right to make copies within your organization, it's outrageous. It is in fact prohibitively expensive.
The report includes statistics about the dismal economic and employment status of people with disabilities in the United States. More than 70% of people with severe disabilities, and 47% of those with less severe disabilities, are unemployed. Only one-third of people with visual impairments are employed (p. 124). What those figures mean, of course, is that Beyond ALT Text is economically inaccessible to members of the accessibility and disability communities-individuals and struggling nonprofits, the very people and groups with the greatest interest in the issues that this important report raises. So who can afford these prices? Who might be willing to pay them? Big development firms. Large corporations with large Web departments. The odd university-based research group. Not individual Web developers who contract out their services. And certainly not most Web users with disabilities.