Guidelines for Describing STEM Images

1. Brevity

The most frequent recommendation from respondents was for more brevity in description. Simply put, it takes people with visual impairments more time to read books and articles than people without visual impairments and the process should not be further slowed down by unnecessarily long image descriptions.

Many images may be rich in visual details but light on important information. If the gist of an image can be understood in a glance, then the description should be just as brief.

Likewise, if the information in an image also included in the main text and therefore accessible, then the image description should not repeat the information.

When an image does contain important information, the description should provide access to the information in as few words as possible.

2. Data

Description should focus on the data and not extraneous visual elements. Elaborately illustrated diagrams, for example, often contain key data that can be made accessible by presenting the data separate from description of the overall image.

3. Clarity

It is vital that in the quest for providing the most brief and data-filled descriptions, clarity is not overlooked. If the reader needs to listen to a description several times because it is poorly written or is presented in a confusing manner, then the overall goal has not been achieved.

4. Drill-Down Organization

Descriptions should follow a drill-down organization, e.g., a brief summary followed by extended description and/or specific data. Drill-down organization allows the reader to either continue reading for more information or stop when they have read all they want.

For example, if a description of a diagram begins with the title (e.x., "Moon Phases") the reader can decide whether or not they want to read the details of the diagram. The description might then continue with a brief summary of the diagram (e.x., "A diagram shows eight different phases of the moon as it orbits the earth") and then continue on with details and data. This approach is supported by the XML code that underlies digital talking books. Very brief descriptions should be marked as "alt" while the longer descriptions should be marked as "prodnote." Note: the "prodnote" tag serves the same purpose as the "longdesc" tag in HTML.

5. Tables

Tables, pie charts and bar charts should be presented as tables, not as narrative description. Proper coding of tables, including captions, table headers, and table data, provide better access to tables than narrative description. Brief summaries or overviews of the charts should be presented before the tables.

6. Processes

Processes that are presented visually in flow charts, diagrams and illustrated chemical reactions, for example, can be converted into nested lists with good results.

There was general agreement among the survey participants that narrative descriptions of flow charts and other processes could be burdensomely long and, despite best efforts, often failed to provide access to the information. The technique of converting processes into lists, recommended in these guidelines, was developed directly from the feedback of our Delphi survey participants. Because this approach is not widely used, we have included many examples of how it can be done.

7. Mathematics

Math equations should be marked up with MathML and rendered in a way that is preferable to the individual reader. Read a full discussion about providing access to math.

8. Narrative Description

Many STEM images are best described by linear, narrative description. Nonetheless, such "traditional" descriptions of purely visual images (e.g. a drawing of a paramecium or a math diagram) benefit from descriptions that follow these guidelines, especially brevity, drill-down organization, clarity and emphasis on data.

9. Navigation Control

Description presented as text is generally preferred over recorded audio because text readers provide superior navigation control. In addition, when description is presented as properly marked up HTML, especially lists and tables, it provides speedy and independent access to data that is unavailable through traditional linear, narrative description.

Navigation is also aided breaking up individual bits of information by putting them on separate lines, using bullet points, or by creating lists.

Note, however, that text readers often mispronounce technical terms and do not provide emphasis that human readers can use to distinguish key elements.

Funding for this project is provided by the National Science Foundation.