Web Access by Persons with Visual Disabilities

Irvin R. Katz [1]

Educational Testing Service

Introduction

The Educational Testing Service (ETS) web site (http://www.ets.org) contains information on a variety of examinations administered by ETS. The purpose of the site is primarily to inform test-takers about the exams by providing sample tests (with feedback), background materials, registration information, and so forth. As with any web site, it is important to evaluate its usability: how well can users locate the information and services that they need? However, rather than evaluating the overall usability of the web site, we are currently focusing on its usability by a particular population: persons with visual disabilities (e.g., blind, low vision). These persons often use screen reading software (or magnification software) to access web pages. Screen reading software uses synthesized speech to read selected portions of a web page. For example, the user can search through the links and buttons on a page, hearing the text associated with each.

To evaluate the web site, we used the Heuristic Evaluation methodology, developed by Jakob Nielsen (Nielsen & Mack, 199x). In this method, evaluators review an interface (or, here, a web site) and report usability problems. The evaluators are cued to the types of errors to look for by a set of usability principles, or heuristics. As pointed out by Levi & Conrad (1996), these heuristics often must be tailored for the particular type of interface under evaluation. In the case of this project, we not only needed to develop heuristics for use of web sites, but specifically needed to identify those heuristics that apply to web access by persons with visual disabilities.

The Heuristics

Our heuristics were based on Levi & Conrad's (1996) heuristics for evaluating a web site. From these heuristics, we selected those that would be particularly important for persons accessing the web site using screen-reading or screen-magnification software. Further heuristics were based on access recommendations found in the literature, as well as on our own knowledge of how people use screen reading software.

We developed two documents consisting of our heuristics: (1) a summary of the heuristics, which lists each heuristic along with a short explanatory paragraph. The purpose of this document was to “cue” the evaluators when locating usability errors in the web site. (2) An “education” document, which expands on the summary and provides motivations for some of the heuristics with respect to access for persons with visual disabilities. The education document is given below:

Web Accessibility Project:

Usability Heuristics for Evaluating Web Structure and Content

for People with Visual Disabilities

(c) 1996, 1997 Educational Testing Service

Introduction

The following usability principles or "heuristics" are to help Web site designers to ensure that persons with visual disabilities -- especially those who use magnification or screen reading software -- can find, gather, remember, and use the information they need from a Web site.

While the heuristics are proposed as being relevant to individuals with a wide range of visual disabilities, the first of these heuristics ("Enable all essential information to be accessed auditorially") is especially important for persons who are blind. The full population of individuals with visual disabilities includes many with low vision as well as individuals who are blind. Individuals with low vision often make use of magnification, large print, and modification of screen contrast or color.

Enable all essential information to be accessed auditorially.

a. Provide text-based means of conveying all essential information. Provide essential information via text because text can readily be output to a speech synthesizer, other sound output device, or a Braille output device. Use text to indicate presence of meaningful nontext objects (text boxes, radio buttons, pop-up menus). For example, use default in text boxes to indicate information to be entered.

b. Provide alternative access to critical information. Provide non-Web access (e.g., email, phone, fax) to critical information or graphics that cannot be easily described in text. For example, because screen readers tend to improperly read multicolumn content, provide other means of accessing the information if that format is essential to the web site.

c. Provide prerecorded audio and parallel text. Make judicious use of prerecorded audio and accompany it with parallel text. Certain types of text (e.g., mathematical equations, complex charts, poetry) may be difficult to understand using standard screen reading software and hardware. Prerecorded audio may contain specialized inflections or additional descriptions to facilitate comprehension. Audio clips should be accompanied by a parallel text descriptions for speech synthesis or visual display. Make prerecorded audio a supplementary rather than primary source of information.

Use explicit labeling of content and structural elements.

a. Label something for what it is. The user should know the purpose of each significant element on the screen via its descriptive text. Distinguish between content and different structural elements (e.g., headings, alternative text for graphics and video, links, bulleted lists). Distinguish among different structural elements, using distinct words/phrases for each. For example, use descriptive links that suggest the content of the page being linked to. Distinguish a link that stays within a site from one that jumps to another site.

Minimize Visual Strain.

a. Use simple visual designs. Use simple designs of text and graphics that are amenable to magnification and black-and-white presentation.

b. Avoid patterned or complex backgrounds. Most individuals with low vision prefer simple dark background colors and bright foreground colors.

Enable the user to locate information quickly.

Enable the user to locate information quickly by providing efficient methods for search and navigation. Navigation even within a single document can be laborious if the document is not organized effectively. Because of the linear nature of sound, search and navigation can be tedious and time-consuming for individuals with visual disabilities, especially individuals who are blind.

a. Give navigational feedback. Facilitate jumping between related topics. Allow the user to determine her or his current position in the document structure and in the current document. Make it easy to return to an earlier state or an initial state. For example, provide a top-of-page link at the bottom of pages. Another possibility is to provide speech or nonspeech audio cues during navigation.

b. Provide progressive levels of detail. Organize information hierarchically, with more general information appearing before more specific detail. This heuristic can apply to an entire site, a set of pages, or a single web page. By providing this organization, the document helps people remember what they need to remember and to know their location in the document at all times.

c. Minimize display time and screen refreshes. All information should display quickly. The screen should be refreshed only when new information appears. While a sighted user might perceive only a quick “blink” if a screen refreshes, unnecessary screen refreshes are tedious to persons using screen reading software because the screen reader will reread the entire screen each time it is redrawn.

d. Enable easy searches. Enable easy searches for topics, words, or phrases. This facility is especially important for users of screen reading software, for whom searching through web pages for needed information is laborious.

Minimize the users' memory load.

For persons using either screen magnification or screen reading software, the effective size of the computer screen can be very small. Persons who use screen magnification software can view only a small portion of a Web page at a time. (For example an individual using 2X magnification can see only about one-quarter of the screen.) Persons who use screen reading software can "view" (hear) only one word at a time and must rely on their own memory to integrate portions of a document.

a. Use chunking. Write material so that documents are short and contain exactly one topic, idea, or concept. Do not force users to remember key information across documents or between diverse locations within the same document. Organize documents so related information is physically close to one another. Lay out screens so that frequently accessed information is easily found. Avoid forcing the user to reread.

b. Design minimalist systems. Eliminate information that is irrelevant or distracting. While a sighted person might skip over unnecessary information after a brief glance, for a person using magnification or screen reading software, it can be tedious and/or laborious to process unnecessary information before discarding it. Make each piece of information count.

Method

Two evaluators have participated thus far. Both evaluators are consultants that help corporations learn about making their products accessible by people with visual disabilities. The evaluators themselves have low visual ability. Each used his or her preferred screen-reading software when evaluating the ETS web site. Evaluators were partially guided in the portions of the web site to evaluate, but were free to explore pages as their interest directed. Persons knowledgeable about the ETS web site also attended each session in addition to someone who recorded all errors detected by the evaluators.

Results

Overall, the evaluators showed some consistency in their performance. The first evaluator spent 2 hours viewing the web site, finding 15 unique errors among 7 web pages. The second evaluator spent twice as much time (two 4-hour sessions), finding 31 unique errors among 14 web pages. The table below shows the distribution of the errors among the heuristics.

Percentage of Errors Categorized under each Heuristic

Overall, the evaluators were able to categorize 70% of the errors using the heuristics. This result suggests some room for improvement in our heuristics. Most of the errors in the “other” category fall into three groups: (1) problems with the screen reader being used (i.e., not the fault of the web developer), (2) inconsistencies from one page within a web site to another, and (3) poor HTML technique.

Lack of consistency can be annoying for a sighted user, but becomes a burden for someone using (for example) screen reading software. Each web site has its own conventions for how links are arranged and how information is found. When accessing a site using a screen reader, the user must figure out the site's conventions. Once understood, a person using screen reading software can negotiate a site well. However, if a portion of a site uses different conventions (e.g., the rest of the site uses just links, but a portion uses both buttons and links), the user must (a) discover the inconsistency and (b) learn the new conventions. Links and buttons are visually obvious, but are handled differently by many screen readers and require different commands for a user to search among them.

An example of poor HTML technique includes using HTML codes as visual formatting commands rather than as tags on the structure of the page. That is, the HTML tags are used so that information appears correctly on the screen -- even if a tag is not appropriate. For example, on one page in the site evaluated, the developers used a table to display a bulleted list, rather than using the HTML tag for bulleted lists. The appearance of the page is identical using either method, but any screen reading software that reads the HTML source might perform incorrectly.

Conclusions

Our results so far suggest that web sites can be made accessible to people with visual disabilities. Accessibility need not be an onerous task for developers and need not compromise access by sighted users. In fact, many of the recommendations implied by the heuristics would improve a web site for sighted users as well (e.g., more thoughtful organization of information, better descriptive text for links). Other changes to a improve the accessibility of a site (e.g., alternative text for all graphics) would typically not affect sighted users.

Biographical Sketch

Irvin R. Katz is a Research Scientist in the Division of Cognitive and Instructional Science at Educational Testing Service. He received a B.S. in Computer Science from Rensselaer Polytechnic Institute (1984) and Ph.D. in Cognitive Psychology from Carnegie Mellon University (1988). Before joining ETS in 1990 he was a Visiting Research Fellow at the Computer Science Department of Keio University in Japan. His research interests include problem solving and learning in complex domains. His current work includes investigations into the psychology of design skill and the development and evaluation of systems that support design tasks.

Footnotes

[1] The work reported in this document represents a collaborative effort of the members of the Web Access project: Eric Hansen (PI), Doug Forer, Dan Jacquamin, Mike Narcowich, Mary Michaels, Andy Baird, and Charles Davis.