EPUB Accessibility Techniques 1.1.1

EPUB creators not familiar with [wcag2] may find the number of techniques daunting, as they are intended to provide broad coverage of possible solutions.

Assistance applying these techniques to EPUB content documents is available from the following sources:

• DAISY Accessible Publishing Knowledge Base — compendium of best practices by content type, with links to the applicable WCAG techniques.

• BISG Guide to Accessible Publishing & Cheat Sheets — includes a set of top practices for making content accessible.

Success Criterion 1.3.2 [wcag2] specifies that each web page have a meaningful order (i.e., that the visual presentation of the content match the underlying markup).

As EPUB allows two EPUB content documents to be rendered together in a synthetic spread [epub-3], the order of content within a single document cannot always be evaluated in isolation. Content may span visually from one document to the next. For example, a sidebar might span the bottom of two pages.

Ordering each document separately by the visual display will lead to users of assistive technologies encountering gaps between the start and end of the spanned text. If the markup cannot be arranged to provide a more logical reading experience (e.g., the beginning of the spanned content at the end of the first page followed by the conclusion at the start of the next), another means of satisfying this criteria will be necessary to avoid failure (e.g., a hyperlink could be provided to allow a user to jump from the break point on the first page to the continuation on the next).

Success Criterion 2.4.5 [wcag2] requires there be more than one way to locate a web page within a set of web pages. By default, EPUB publications meet this WCAG requirement so long as EPUB creators follow the EPUB requirements to include all EPUB content documents in the spine and ensure access to all non-linear documents [epub-3].

The reason an EPUB publication passes by meeting these requirements has to do with differences in how a user interacts with the set of documents in an EPUB publication. In particular, although an EPUB publication typically consists of many EPUB content documents, reading systems automatically provide the ability for the user to move seamlessly from one document to the next, so long as they are listed in the spine [epub-3]. To the user, an EPUB publication is a single document they have complete access to, not a set of disconnected pages that they need links to move through.

The required table of contents provides a second method to access the major headings of the publication. The user can jump to any heading and continue to navigate from there, regardless of how the publication is chunked.

Following these two requirements therefore satisfies the need for multiple ways to access the content. Reading systems also typically provide search capabilities, something the EPUB creator cannot provide, so users also have a third option available in most cases.

Although EPUB creators only need to follow EPUB requirements to meet this criterion, they are still encouraged to provide additional methods to improve access beyond the minimum. Some suggestions include:

• adding at least one link to every EPUB content document in the spine to the table of contents, when feasible;

• adding an index to locate major topics; and

• adding additional navigation aids to the EPUB navigation document (e.g., lists of figures and tables).

The table of contents provides users more than just links into the content. It is also a means to understand the structure and ordering of an EPUB publication. Consequently, users may have difficulty locating where they are in a publication, where they want to go, and also how to return to previous locations when the order of entries in the table of contents does not match the linear reading order.

EPUB creators should therefore ensure that the entries in the table of contents always match the linear order of the content. Specifically, the order of entries should reflect both:

• the order of EPUB content documents in the spine; and
• the order of each referenced section within its respective EPUB content document.

Only if there is a logical case for an alternative arrangement of entries should the ordering differ. Such scenarios typically only occur when the content does not have to be read linearly or when additional information is included at the end of a table of contents. For example, the table of contents for a magazine might be ordered to list all the major articles first, followed by features, etc.

When the ordering of the table of contents does not match the content, EPUB creators should include an explanation why in the accessibility summary.

EPUB creators should avoid including links to supplementary content at the end of the table of contents. Links to figure, tables, illustrations and similar content is better included as a separate navigation elements (either in the EPUB navigation document or in the spine). EPUB creators can include links to these additional navigation lists in the table of contents.

Web sites are constructed very differently from EPUB publications. A typical web site wraps the content of each page within a repeating template, for example. This template gives each page a consistent look and feel, but users are rarely interested in the wrapper content after visiting the first page. Visual readers can typically skip past the site header, navigation bars, search boxes, and other helpful but seldom-used features to get right to the content.

To provide the same ease of access to readers who would have to navigate sequentially through the repetitive content, success criterion 2.4.1 [wcag2] requires a means of bypassing the repeated content in a set of pages. This success criterion does not apply to typical EPUB publications, however, as EPUB content documents do not repeat content in the same way that web sites do.

Each new content document may begin with similar content, such as learning objectives or key terms, but this content is part of the body of the publication and not identical to what came before. Consequently, it is not required to add a link to skip it. (Secondary content should be identified in accordance with success criterion 1.3.1, however.)

Although the role attribute may seem similar in nature to the type attribute [epub-3], their target uses in EPUB content documents do not overlap.

The key difference between these attributes is that the role attribute bridges accessibility in content while the type attribute provides hooks to enable reading system behaviors. Omitting roles lessens the accessibility for users of assistive technologies, in other words, while omitting types diminishes certain functionality in reading systems (e.g., pop-up footnotes or special presentations of the content).

Since each attribute offers different advantages, it is not necessary that they be used together. Due to the lack of restrictions on where EPUB creators can use the type attribute, pairing the attributes may cause accessibility issues (e.g., putting roles on the [html] body element).

In particular, the use of the type attribute is not a means of satisfying requirements for ARIA roles in WCAG.

For EPUB creators looking to move from the type attribute to using ARIA roles, the EPUB Type to ARIA Role Authoring Guide guide details notable authoring differences between the two attributes. It also includes a mapping table of semantics in the EPUB Structural Semantics Vocabulary to equivalent ARIA roles in [dpub-aria-1.1] and [wai-aria].

Although EPUB publications appear as single contiguous documents to users when read, they are typically composed of many individual EPUB content documents. This practice keeps the amount of markup that has to be rendered small to reduce the load time in reading systems (i.e., to minimize the time the user has to wait for a document to appear). It is rare, at least for books, for an EPUB publication to contain only one EPUB content document with all the content in it.

When content is chunked in this way, it often requires the EPUB creator to make decisions about how best to restructure the information. A part, for example, will typically not include all the chapters that belong to it. The EPUB creator will instead separate the part heading from each chapter, putting each into a separate document.

Although visually these restructuring decisions can be hidden from readers, they impact the functionality of assistive technologies. In the case of [wai-aria] roles, the result is that only the subset present in the currently-loaded EPUB content document are exposed to users. An assistive technology cannot provide a list of landmarks for the whole publication, as it cannot see outside the current document.

To counteract this destructuring effect, EPUB creators sometimes think to re-add or re-identify structures in the belief that having this information in every document will be helpful to users (e.g., adding an extra [html] section element around a chapter to indicate it belongs to a part). All this practice does, however, is add repetition that is not only disruptive when reading but can make the structure of the publication harder to follow. EPUB creators are therefore advised not to attempt to rebuild structures in these ways.

For example, consider a book that has five parts and each part contains five chapters. Structurally, each chapter belongs to its part (i.e., is grouped with it), as in the following markup:

<html … >
   …
   <body>
      <section
          role="doc-part"
          aria-labelledby="p1">
         
         <h1 id="p1">Part 1</h1>
         
         <section
             role="doc-chapter"
             aria-labelledby="c1">
            
            <h2 id="c1">Chapter 1</h2>
            …
         </section>
         …
      </section>
      …
   </body>
</html>

Since this would lead to a large content file, the part heading is typically split out into its own EPUB content document so that it will appear on its own page:

<html … >
   …
   <body>
      <section 
          role="doc-part">
         <h1 id="p1">Part 1</h1>
      </section>
   </body>
</html>

Each chapter is then separated into a separate EPUB content document:

<html … >
   …
   <body>
      <section
          role="doc-chapter">
         
         <h2>Chapter 1</h2>
         …
      </section>
   </body>
</html>

If another section tag were added around the chapter in the preceding example to indicate it is in part one (e.g., using the aria-label attribute so the heading is not visible) users would hear "Part 1 Part 1 Chapter 1" because "Part 1" is also the heading in the document that precedes the first chapter.

[wai-aria] landmarks are similar in nature to EPUB landmarks [epub-3]: both are designed to provide users with quick access to the major structures of a document, such as chapters, glossaries and indexes. ARIA landmarks are compiled automatically by assistive technologies from the roles that have been applied to the markup, so EPUB creators only need to follow the requirement to include roles for the landmarks to be made available to users.

Although automatic generation of ARIA landmarks simplifies authoring, it also means that ARIA landmarks are limited to how the EPUB publication has been chunked up into EPUB content documents. An assistive technology can only present the landmarks available in the currently-loaded document; it cannot provide a complete picture of all the landmarks in a multi-document publication (see the previous section for more discussion about content chunking).

EPUB landmarks, on the other hand, are compiled by the EPUB creator prior to distribution, and are not directly linked to the use of the type attribute [epub-3] in the content. They are designed to simplify linking to major sections of the publication in a machine-readable way, as reading systems do not scan the entire publication for landmarks, either. EPUB landmarks are typically not as numerous as ARIA landmarks, as reading systems only expose so many of these navigation aids.

Given these differences in application, however, it is important to include EPUB landmarks and not rely only on the presence of ARIA roles to facilitate navigation, and vice versa. Each aids navigation in its own way.

The EPUB specification does not require that EPUB creators include a specific set of landmarks, but it is recommended to include a link to the start of the body matter as well as to any major reference sections (e.g., table of contents, endnotes, bibliography, glossary, index).

Success Criterion 2.4.2 [wcag2] requires that each web page include a title. EPUB has a similar requirement for EPUB publications: publications require a [dcterms] title element in the package document metadata. The [wcag2] requirement is not satisfied by the EPUB requirement, however.

When authoring an EPUB publication each EPUB content document also requires a descriptive title that describes its content. If not provided, assistive technologies often will announce the name of the file to users.

If the title includes structural context (e.g., the part heading a chapter belongs to, or the name of the publication), order the title such that the most precise description of the current document comes first.

For more information about titles, see Technique H25.

To a user, an EPUB publication appears as a single document that they read from beginning to end, even though the content is often split across numerous EPUB content documents. As a result, their natural expectation is that the headings reflect their position in the overall hierarchy of the publication, despite the publication not actually being a single document (e.g., if a part heading is expressed in an [html] h1 element, each chapter that belongs to the part will have an h2 heading).

Technique G141: Organizing a page using headings instructs EPUB creators on correctly using numbered headings within a document, but with EPUB publications the numbered headings also need to remain consistent across documents. Practically, this means that each EPUB content document does not have to begin with an h1 heading unless the first heading is a top-level heading — the first heading needs to have a numbered heading element that reflects its actual position in the publication.

EPUB creators also need to chunk their content so that the first heading in a document always has the highest number. For example, if a document starts with an h3 heading, there should not be an h2 heading later in the document (e.g., do not include the start of a new section with the trailing subsections of the previous). It is acceptable for there to be subsequent headings at the same level as the first (e.g., multiple subsections in one document could all have h3 headings).

Success Criterion 2.4.6 [wcag2] currently states that all headings must describe their topic or purpose. The implication of this wording is that all chapters in a novel, for example, have a topic or purpose and that the topic or purpose is always clearly reflected by the title of the chapter. Not only is this not always the case, but this success criterion also complicates the use of chapter numbers as headings since these do not establish a topic.

After discussion in the Accessibility Guidelines Working Group's issue tracker, it is clear that the understanding document for 2.4.6 captures the intent of the success criterion better than the current wording — namely, that the requirement for headings is only that they establish a unique context for the content.

By this interpretation, the headings in publications should always pass provided they are unique.

It is expected that the wording of the success criterion will be updated to better reflect the uniqueness requirement, likely in the future WCAG 3 due the complexities of changing the wording for WCAG 2.

The first version of these techniques only required alternative text for images regardless of their complexity. This exception is no longer valid.

EPUB creators must now ensure that their image-based content meets [wcag2] requirements for alternative text and extended descriptions to conform with [epub-a11y-111].

Success Criterions 3.1.1 and 3.1.2 [wcag2] deal with the language of a page and changes of language with in, respectively.

For EPUB publications, the package document is also an important source of metadata information about the publication. For example, reading systems expose details of the publications to users in their bookshelves using this information.

Consequently, it is necessary to provide the language of all text content in the package document to conform with these WCAG success criteria. The easiest way to meet this requirement is to add an xml:lang attribute on the root package element [epub-3].

If individual metadata fields within the package document are expressed in a different language, it is similarly required that the language change be identified by an xml:lang attribute on the element for the field.

Providing this information enables reading systems to correctly render the text content in the proper language for users.

In addition to being able to express the language of text content, the package document also allows EPUB creators to identify the languages of the EPUB publication in dc:language elements [epub-3].

Although it is not strictly required to set this information to meet Success Criterion 3.1.1 [wcag2], as it is non-normative, it should be considered a best practice to always set this field with the proper language information. (Note that EPUB3 requires the language always be specified, so omitting will fail validation requirements.)

Although reading systems do not use this language information to render the text content of the EPUB publication, they do use it to optimize the reading experience for users (e.g., to preload text-to-speech engines so users do not have a delay when synthesizing the text).

Success Criterion 1.1.1 [wcag2] requires that text equivalents be provided for all non-text content to meet Level A. In some regions (e.g., Asia), it is not uncommon to find images of individual text characters, despite the availability of Unicode character equivalents. This practice occurs for various reasons, such as ease of translation of older documents and for compatibility across reading systems. The use of images in most instances leads to the text not being accessible to non-visual users, however.

When individual characters are replaced by images, there is invariably a negative effect on text-to-speech playback, even when alternative text is provided (e.g., if single characters within a word are replaced by images, the word will not be read as a single unit of text). It is also problematic for visual users, as the images often scale poorly and the characters cannot be changed to different font families to meet user preferences.

The use of Unicode characters for all text content avoids this problem, allowing content to successfully meet the minimum requirement for Level A.

For compliance with Level AA, EPUB creators are directed to Success Criterion 1.4.5 which further restricts the use of images of text to only a set of essential cases.

Both the EPUB Structural Semantics Vocabulary [epub-ssv] and Digital Publishing WAI-ARIA 1.0 Module [dpub-aria-1.1] include a semantic for static page breaks: pagebreak and doc-pagebreak, respectively.

For accessibility purposes, it is most important to specify the role attribute value as this is what assistive technologies will process. The epub:type attribute is not recognized by these devices, but it may be used by epub reading systems to enhance the user experience. For compatibility with both types of devices, it is recommended to apply both attributes, although only the role attribute is strictly required for accessibility conformance.

A title or aria-label attribute is required on the element, as it provides the value that is announced to the user.

The page number may be inserted as text content within the element, but assistive technologies may not provide a way to disable announcing of the numbers. As a result, users may hear the numbers announced wherever they occur, which could cause confusion (e.g., the number may seem to be part of the sentence it occurs within).

EPUB 2 does not include markup to identify static page break marks in the content. Page break destinations can be included to enable hyperlinking, but the page list is the only way a user can jump to the locations.

Do not use the [html] a element to identify page break locations in EPUB 3 publications. Although this element was previously defined as the anchor for a hyperlink destination, its purpose has been changed in [html] for use solely as a link.

Readers rarely stop reading to review each new page number, so when page numbers are read aloud in the audio playback of a publication it is not only distracting, but can be confusing, as well (e.g., the number could be read out in the middle of a sentence).

To mitigate this potential annoyance to readers, EPUB creators need to identify page announcements in media overlay documents when they are included. Identification allows a reading system to provide a playback experience where the numbers are automatically skipped.

To identify page numbers in media overlay documents, attach an epub:type attribute with the value "pagebreak" [epub-ssv] to each par element [epub-3] that identifies a page number.

Note that as EPUB 2 does not provide text and audio synchronization, this technique does not apply.

A page list — a list of hyperlinks to the static page break locations — is the most effective way for users to find static page locations. Without a page list, the user would have to navigate each page marker in the text, provided they are available and the reading system provides such functionality.

When a page list is included, reading systems can provide users direct access to the list or use it to provide automatic page jump functionality.

The EPUB navigation document allows the inclusion of a page-list nav [epub-3], while the EPUB 2 NCX file provides the same functionality through the pageList element [opf-201].

Users typically want to know the source of the page break markers included in an EPUB publication when they are derived from a static media. Considerations like which printing, by which publisher or imprint, and whether the pagination comes from the hard or soft cover edition will affect decisions about its usefulness (e.g., does it exactly match the pagination of a print book used in a classroom).

To allow users to determine the suitability of the pagination, identify the ISBN of the source work in the package document metadata using the pageBreakSource property [page-source-id].

If an ISBN is not available, include as much information as possible about the source publication (e.g., the publisher, date, edition, and binding).

If the page break markers are unique to the EPUB publication, use the value none with the pageBreakSource property.

Ensuring the complete text of an EPUB publication is synchronized with audio is key to allowing users who require full synchronized playback, or even audio-only playback, have access to the same information as users who do not require synchronized playback.

EPUB 3's media overlays feature [epub-3] allows audio to be synchronized with any element in an EPUB content document, so there is no technical barrier to providing synchronized playback.

The primary consideration for this objective is what constitutes the text content of an EPUB publication. The minimal candidates for synchronization with audio are all the elements with visible text content.

In HTML, the class of elements called palpable content [html] can typically be synchronized with audio or have their own built-in audio. Embedded text in SVG documents, on the other hand, is found in the text element [svg] and its descendants.

The media overlays text element is used to reference these elements, either to play back the pre-recorded audio in a sibling audio element [epub-3] or to initiate playback of an audio or video element if the audio element is missing (e.g., for embedded audio and video in the document).

In addition to synchronizing the visible text, synchronized text-audio playback must also address text alternatives for image-based content. Images may have alternative text and descriptions that are not visible to all users. As synchronization is also meant to aid users who cannot see the images, including these text alternatives and descriptions in the playback is essential to providing the user all the information in the EPUB publication.

Text alternatives and descriptions in HTML may be represented in the alt attribute [html] and linked by ARIA attributes (e.g., aria-describedby and aria-details [wai-aria]). Descriptions for image elements in SVG are typically represented in a desc element but ARIA attributes may also be used.

The default reading order should typically represent the order in which reading systems render content to users during synchronized text-audio playback. For EPUB publications, this is a combination of the sequence of EPUB content documents in the spine and the order of elements within each EPUB content document.

If there are cases where the logical reading order (how a reader would naturally read the content) diverges from the default reading order, EPUB creators can order the playback sequence of seq and par elements in a media overlays document [epub-3] to match the logical order.

EPUB creators need to use caution when making alterations, however, as other accessibility issues can arise when the logical order does not match the default order. For example, the content may not be accessible to users of assistive technologies when the order in the markup does not match how the assistive technology reads the content. In these cases, using playback to create a logical order can make the EPUB publication fail WCAG conformance requirements.

One case where the logical may diverge from the reading order and remain accessible is in tables, as assistive technologies typically allow users to choose whether to read by row or by column.

Some content elements are not critical to read when following the primary narrative of a work, and that would interrupt a user's concentration if they had to stop and listen to. Footnotes and endnotes are examples of such content, as users may only want to come back and read this content after finishing the EPUB publication. The announcement of page break numbers can be similarly annoying to readers.

EPUB 3's media overlays feature [epub-3] does not allow reading systems to determine if playback sequences are skippable unless EPUB creators add additional semantics to the markup, however. EPUB creators must use the epub:type attribute [epub-3] to add semantics to seq and par elements [epub-3], thereby allowing reading systems to provide users the option to skip their playback sequences.

The recommended structures to identify for skippability are:

• Endnotes — use the endnotes and endnote semantics [epub-ssv-11] for groups of notes and individual endnotes, respectively.
• Footnotes — use the footnotes and footnote semantics [epub-ssv-11] for groups of footnotes and individual footnotes, respectively.
• Page breaks — use the pagebreak semantic [epub-ssv-11] to identify each.

EPUB creators may identify other structures but it is not necessary to meet this requirement.

Some content elements are containers for expressing complex information. A table, for example, has data arranged in rows and cells. Lists similarly may contain many items. While users may be interested in some of the information in these structures, they also often want to escape from them to keep reading, not have to listen to the entire content before being able to move on. These are called escapable elements, because the user needs to be able to escape from them whenever they choose to simplify the reading experience.

EPUB 3's media overlays feature [epub-3] only supports escapability if EPUB creators add structural semantics to the markup. EPUB creators must use the epub:type attribute [epub-3] to add semantics to seq and par elements [epub-3] to allow reading systems to provide users the option to escape their playback sequences.

The recommended structures to identify for escapability are:

• Figures — use the figure semantic [epub-ssv-11] to identify each.
• Lists — using the list semantic [epub-ssv-11] to identify each.
• Sidebars — use the aside semantic [epub-ssv-11] to identify each.
• Tables — use the table semantic [epub-ssv-11] to identify each.

EPUB creators may identify other structures but it is not necessary to meet this requirement.

EPUB creators can add a media overlay document for the EPUB navigation document even when it is not included in the spine. Doing so allow reading systems to announce the link labels regardless of how they present the navigation elements to users (e.g., many reading systems applications create custom table of contents panels by extracting the data from the EPUB navigation document).

The process for adding a media overlay document is no different than one for any other document.

See the following resources for more information about including accessibility metadata in distribution records:

• Schema.org Accessibility Properties Crosswalk

• ONIX Code List 196