This document is a placeholder for sitnificatn sections of content that were removed from the ARIA Authoring Practices Guide. It allows reviewers to more easily see what was removed, and makes it easier to restore content.

Design Patterns and Widgets

Auto Complete

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

A text box and an associated drop-down list of choices where the choices offered are filtered based on the information typed into the box. Typically, an icon associated with the text box triggers the display of the drop-down list of choices. An editable auto-complete accepts text entry of choices that are not in the list. An example of an editable auto-complete is the URL field in the browsers.

Keyboard Interaction

  • With focus in an empty text box, press Down Arrow, Up Arrow, Alt + Down Arrow, or Alt + Up Arrow to display the entire list of choices. Focus remains in the text box and no choice is highlighted.
    • Press the Down Arrow to highlight the first choice in the list.
    • Press the Down Arrow and Up Arrow keys to highlight the desired choice in the list.
    • Note that the arrows will wrap through the text box when the top or bottom of the list is reached. For example, pressing the down arrow when the last choice is highlighted will move focus back to the text box, pressing down again will move focus to the first item in the list. Likewise, with focus in the text box and the list displayed, pressing up arrow will move focus to the last item in the list.
    • When a choice is highlighted using the arrow keys, the highlighted choice is displayed in the text box.
    • Press Enter to select the highlighted choice and close the drop-down list. This mimics the behavior of the HTML select element.
  • With the drop-down list of choices displayed, move the mouse pointer over an item in the list to highlight it. The text box value is not modified when the mouse is used to highlight a choice. Clicking on the highlighted choice will close the drop-down and update the text box with the selected choice. This mimics the behavior of the HTML select element.
  • With focus in an empty text box, type any letter. If any of the available choices begin with the letter typed, those choices are displayed in a drop down. If the letter typed does not match any of the available choices the drop-down list is not displayed.
  • With focus in text box with an existing value type additional letters. As the user types letters the list of choices is filtered so that only those that begin with the typed letters are displayed.
    • Until the user presses the arrow keys to highlight a particular choice, only the typed letters are displayed in the text box.
    • In an editable auto-complete, if no choices match the letter(s) typed, the drop down list closes.
    • In a non-editable auto-complete, any letters that do not result in a match from the list are ignored. The drop down list of choices remains static until the user presses:
      • Escape to clear the text field
      • Backspace to remove some of the letters previously typed
      • an additional letter that results in a valid list of choices.
    • Navigation through the list of choices and display of the highlighted choice in the text box works as described above.
      Optional: When a choice is highlighted via arrow key navigation, the input cursor is left at the end of the typed entry and the highlighted choice is displayed in the text box with the characters after the input cursor selected. Typing an additional character will remove the auto-completed portion and append the newly typed character to the end of the previously typed characters. The list will be filtered based on the additional character(s) typed.
  • With focus in a text box, press Escape
    • If there is no text in the text box, pressing Escape closes the drop-down if it is displayed.
    • For an editable autocomplete that has text in the text box that was both typed by the user and auto-completed by highlighting a choice using the keyboard, the auto-completed portion of the text is cleared and the user typed characters remain in the text box. The drop-down list is closed. To completely clear the text box contents the user must use the backspace key to remove the typed characters. This is how the Google search box in the Firefox UI works. Recommend that pressing the Escape key again completely clears the text box rather than relying on only the backspace key.
    • For a non-editable auto-complete that has text in the text box that was both typed by the user and auto-completed by highlighting a choice using the keyboard, pressing Escape closes the drop-down list and leaves the current choice in the text box.
    • For an editable or non-editable auto complete with text in the text box that was typed by the user and the mouse is highlighting a choice in the drop down (keyboard navigation was NOT used), pressing Escape closes the drop down and leaves the typed text displayed in the text box. Need to consider if pressing Escape again should clear the typed text. The user must press the Down arrow or Alt + Down arrow or click the associated icon to invoke the drop-down list of choices again.
  • Moving focus out of an empty auto complete field where a value is required should either invoke an error or if a default value was initially assigned, reset the value to the default value.
  • Moving focus out of an auto complete field that does not contain a valid entry should either invoke an error or if a default value was initially assigned, reset the value to the default value.

It is good practice to limit the number of matching items in the drop down to a reasonable number. The reasonable number is determined by the task at hand. A list of the 50 US States is probably reasonable, but a list containing all of the office numbers in a building is probably not appropriate.

WAI-ARIA Roles, States, and Properties

  • The widget has a role of combobox.
  • The combobox has an aria-autocomplete property set to one of 'inline', 'list', or 'both'.
  • For more information, see the combobox design pattern.


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Dojo autocomplete

Combo Box

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

A combo box enables the user to type in a field and at the same time chose a predefined value from a list. By using the keyboard the user can select an item from the list. After selection she will be able to type further characters in the field.

Keyboard Interaction

  • Left Arrow or Right Arrow move the caret within the edit field.
  • Alt + Up/Down Arrow opens and closes the list.
  • Up Arrow and Down Arrow moves focus up and down the list. As focus moves inside the dropdown list, the edit field is updated.
  • Page Up / Page Down selects the next/previous pages item depending on the lists size.
  • Escape closes the dropdown list, returns focus to the edit field, and does not change the current selection.
  • Enter selects the current item on the list, updates the edit field, highlights the selected item in the dropdown list, closes the dropdown list, and returns focus to the input field.
  • Typing a letter (printable character) key moves focus to the next instance of a visible node whose title begins with that printable letter.

For combo boxes that implement aria-autocomplete, see the autocomplete design pattern.

WAI-ARIA Roles, States, and Properties

A combobox can be implemented using either of two design patterns:

  1. As a combination of text field, which may be editable, a displayable list of items, and a drop button to toggle the display of that list; all wrapped in the form of a single widget with role of combobox.
  2. As a combobox, which behaves like a textfield and may be editable, with a displayable list of items, and a drop button to toggle the display of that list;

Like text fields a combobox should be labeled to convey the purpose of the widget. Keyboard focus within the widget must be managed by the widget. Comboboxes are used extensively in graphical user interfaces and the design pattern for the widget should be semantically correct.

For the first combobox design pattern:

  • The container element that wraps the combobox has a role of combobox.
  • The first element within the combobox is an input text field and is responsible for managing the keyboard focus between the text field and the list as well as displaying the list. The text field is in the tab order. If you create a text field without using a standard HTML text field form control then ensure that it is in the tab order.
  • If the text field is not editable it must have aria-readonly = true.
  • The combobox must have aria-expanded = true if the list is displayed or aria-expanded = false when it is not.
  • The next element is an html <button>, or another element with a role of button. This element is used to toggle the display of the combobox's drop down list.
  • The next element has a listbox role and represents the drop down list. It manages keyboard navigation among list items and navigating back to the text field if necessary.
  • Each item in the listbox is an option. Options are not in the tab order.
  • Provide a label for the combobox by referencing the text field in the combobox. You can use an aria-label to associate this label with the combobox or you may use the HTML <label> element and its for attribute to reference the text field.

For the second combobox design pattern:

  • The first element for the combobox has a role of combobox and behaves like an input text field and is responsible for managing the keyboard focus between the combobox and the list as well as displaying the list. The text field is in the tab order. If you create a text field without using a standard HTML text field form control then ensure that it is in the tab order.
  • If the combobox is not editable it must have aria-readonly = true.
  • The combobox must have aria-expanded = true if the list is displayed or aria-expanded = false when it is not.
  • The next element is an html <button>, or another element with a role of button. This element is used to toggle the display of the combobox's drop down list.
  • The next element has a listbox role and represents the drop down list. It manages keyboard navigation among list items and navigating back to the text field if necessary.
  • Each item in the listbox is an option. Options are not in the tab order.
  • Provide a label for the combobox by referencing the text field in the combobox. You can use an aria-label to associate this label with the combobox or you may use the HTML <label> element and its for attribute to reference the text field.

For each combobox pattern the button need not be in the tab order if there is an appropriate keystroke associated with the input element such that when focus is on the input , the keystroke triggers display of the associated drop down list. In this case, a Tab to focus the button is unnecessary. This is the same behavior as the select element.


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

First design pattern -- container element with role combobox:

Second pattern -- text field with role combobox:

Date Picker

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

The date picker widget allows the user to select a date or date ranges. The date picker shows one month at least. All navigation that is described below depends on the application. If no range selection is possible, the relevant keystroke interaction can be ignored. Also navigation to the past might be optional. Each week might be labeled with the corresponding calendar week number.

As a general rule the actual calendar portion of the date picker follows a table structure where days of the week and calendar day numbers are laid out in table cells. This provides context so an assistive technology can render the day of the week; its corresponding numeric calendar day, and week number if necessary. Consequently, it is best to start with an HTML table and apply WAI-ARIA semantics for a grid. However, should the author wish to uses a div or span to represent the cells then the DOM structure for a table is duplicated with rows marked with role="row".

The calendar portion can be displayed in a numbers of ways, including as a popup associated with another widget, or as a static region of a page.

Keyboard Interaction

Keyboard navigation on days that are not included the currently displayed month should move to the month automatically and lead to the day in the next or previous month.

  • Tab - Like other widgets, the date picker widget receives focus by tabbing into it. Once focus is received, focus is repositioned on today's date in a grid of days and weeks. A second tab will take the user out of the date picker widget. Focus initially is placed on today's date.
  • Shift + Tab - reverses the direction of the tab order. Once in the widget, a Shift+Tab will take the user to the previous focusable element in the tab order.
  • Up Arrow and Down Arrow - goes to the same day of the week in the previous or next week respectively. If the user advances past the end of the month they continue into the next or previous month as appropriate.
  • Left Arrow and Right Arrow - advances one day to the next, also in a continuum. Visually focus is moved from day to day and wraps from row to row in a grid of days and weeks.
  • Control + Page Up - Moves to the same date in the previous year.
  • Control + Page Down - Moves to the same date in the next year.
  • Space -
    • Singleton Mode: acts as a toggle either selecting or deselecting the date.
    • Contiguous Mode: Similar to selecting a range of text. Space selects the first date. Shift + Arrows add to the selection. Pressing Space again deselects the previous selections and selects the current focused date.
  • Home - Moves to the first day of the current month.
  • End - Moves to the last day of the current month.
  • Page Up - Moves to the same date in the previous month.
  • Page Down - Moves to the same date in the next month.
  • Enter -
    • If the calendar is a popup attached to some other widget (e.g., a text field), then Enter dismisses the popup, and the selected date(s) are shown in the associated widget.
    • If the calendar is a static region on the page, then Enter confirms the selected date(s).
  • Escape - in the case of a popup date picker, closes the widget without any action.

Navigation into the past is optional

Do not implement keyboard navigation schemes that would place more than one calendar day in the tab order at any time as this impacts the usability of keyboard navigation. For example, using HTML anchors for the gridcells places them all in the tab order impacting the usability of keyboard navigation.

WAI-ARIA Roles, States, and Properties

  • The current month has a label representing the month and year. It is advisable to use a role heading but is not essential. This "label" should have a unique ID.
  • If the author would like to ensure that a label is announced by a screen reader, as the label changes, include live region properties with the label element: aria-live="assertive" and aria-atomic="true".
  • The container for the day of week headers and numeric days of the week has a role of grid.
  • The grid has an aria-labelledby property with a value equivalent to the id of the label for the grid.
  • Each name for the day of the week has a role columnheader and is not navigable via the keyboard.
  • Each numeric day of the week has the role gridcell.
  • When a day is selected its aria-selected is set to true, otherwise it is set to false or removed.
  • Changes in aria states, identified here, as well as focus, are clearly styled to show the user where their point of regard is and what days are selected.

When the date picker is active a calendar day of the week always has focus. This can be achieved by setting the tabindex on that day as appropriate and then using script to give it focus. Alternatively, the grid container could set aria-activedescendant to the id of the currently focused gridcell. Keep in mind that older browsers may not support aria-activedescendant.


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Dialog (Non-Modal)

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

A dialog is a small application window that sits above the application and is designed to interrupt the current processing of an application in order to prompt the user to enter information or require a response (dialog).

A non-modal dialog is one which is displayed and focusable at the same time as the application which invoked it. Also like the modal dialog, focus via the tab and shift-tab key must be maintained within the dialog. However, a non-modal dialog should have a keyboard mechanism to return focus to the application while leaving the dialog open.

Keyboard Interaction

  • Escape cancels the dialog without taking any action.
  • Enter submits any data gathered in the dialog.
  • F6 is the recommended key to move focus between the application and an open non-model dialog.


  • The mouse user may click on either the application or the dialog to change focus between the two.
  • In a Web application the non-modal dialog is usually always displayed above the application page, rather than in a separate browser window but that is not a requirement.
  • Authors should take care when using Enter to trigger default actions since Enter might be connected to and trigger some other user interface element, or it might trigger the focused element. Authors should ensure that Enter activates only the widget they intend.

WAI-ARIA Roles, States, and Properties

See Dialog (Modal).


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Experimental dojo floating non-modal pane

Dialog (Tooltip)

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

A tooltip dialog is a modal dialog that is rendered near the invoking element and visually connected via a cartoon bubble-like protrusion. It is displayed when the mouse passes over or rests on that element.

Keyboard Interaction:

  • Escape The tooltip dialog is closed by pressing the escape key when focus is within the dialog, mouse clicking on a close icon, or mouse clicking outside of the dialog onto the application.
  • Tab Focus must be held within the dialog until it is cancelled or submitted. As the user presses tab to move within items in the dialog, pressing tab with focus on the last focusable item in the dialog will move focus back to the first focusable item in the dialog.
  • Shift + Tab Likewise, if the user is shift-tabbing through elements in the dialog, pressing shift-tab with focus on the first focusable item in the dialog will move focus to the last item in the dialog.

It is modal because focus is trapped within the dialog as the user navigates via the Tab and Shift + Tab key.

Unlike a true modal dialog, the user can click outside of the dialog, however in that case the tooltip dialog is immediately closed.

A tooltip dialog can not be moved/dragged.

Other than the close and move behavior, all other behaviors of a modal dialog are implemented by the tooltip dialog.

WAI-ARIA Roles, States, and Properties

See Dialog (Modal).


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Dojo nightly

Landmark Navigation

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

Keyboard Interaction

WAI-ARIA Roles, States, and Properties

Provide the appropriate landmark for the role attribute. See the section describing landmark use.


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Popup Help (aka Bubble Help)

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

Popup Help contains more descriptive, or actionable, help-like text and elements. It may contain, and was designed to handle, interactive elements such as a button, link, or text field. It is essentially a Popup Menu with un-necessary keystrokes turned off. The key sequence for posting Popup Help was to take advantage of F1's tie to the Help paradigm (F1 calls up application Help for example).

Keyboard Interaction

  • Control + F1
    • Posts the Popup Help widget.
    • Input focus is placed on the first interactive element in the Popup Help.
    • Control F1 is used by IE to "Display a help dialog box". In IE7 and IE8 event.cancelBubble=true; and event.returnValue=false; will allow the re-purposing of keys used by the browser. In the case of IE6, you can not stop the bubble up of keys used by the browser, but can stop the bubble up to the OS. In the case of Firefox and other standard compliant browsers, event.stopPropagation(); and event.preventDefault(); will re-purpose the keys.
    • With the exception of Control F1 to bring up Popup Help this widget is very similar to Dialog (Modal) and/or Dialog (Non-Modal) and/or Dialog (tooltip) described elsewhere in this document.
  • Esc
    • Causes no menu action and dismisses Popup Help.
    • Input focus is returned to the element, or widget the Popup Help was invoked from.
    • Pressing Enter when input focus is on the "X" glyph acts the same as pressing Esc.
  • Tab
    • One of the following occurs:
      • Modal behavior
        • Moves input focus between elements in the Popup Help's tabbing order.
          • Input focus stays in the Popup Help until one of the following occurs:
            • Esc is pressed.
            • Enter is pressed when input focus is on an interactive widget/element.
      • Non-Modal behavior
        • Moves input focus to the next tabbable element in the tabbing order if the following applies:
          • Popup Help is posted.
          • It contains no tab navigable elements in it.
  • Shift + Tab
    • As with other keyboard conventions described here, the Shift + Tab has the effect of moving the focus up rather than down and follows the same conventions as described for the modal and non-modal Tab key above.
  • Enter
    • Activates the element in the Popup Help that has input focus, if applicable, then dismisses the popup.
      • Input focus should be placed on the appropriate element after the user presses the Enter key.
        • The appropriate place to move input focus to will not always be the parent element the Popup Help was invoked from.
      • Nothing occurs if input focus is on an element that has no associated action.
  • F6
    • In the non-modal instance, the F6 key can be used to move focus between the application and the open non-modal window. This is also the behavior described in Dialog (Non-Modal) above.

WAI-ARIA Roles, States, and Properties

See Dialog (Modal) and/or Dialog (Non-Modal) and/or Dialog (tooltip).


Add link to example.

Rich Text Editor

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

Input control that accepts free-form text as its value.

Keyboard Interaction

  • The edit control is provided by the browser; it provides the keyboard support for navigating, adding, removing and selecting text, so that behavior is not defined by the rich internet application.
  • The browser should also provide a keyboard mechanism for navigating into and out of the edit control. Within most browsers the edit control is put into the tab order of the page and can be navigated into, out of, and through using the tab and shift-tab keys like any standard form control.
  • A rich text editor widget needs to provide a user interface for interacting with the browser provided edit control. Interaction between the user interface and editor is defined here assuming that a toolbar is used.
  • Tab and Shift + Tab - If not provided by the browser, the rich text editor widget provides a keyboard mechanism to move into and out of the edit control. Tab and shift-tab are the recommended keystrokes. The toolbar or other user interface component associated with the editor is placed in the tab order immediately before the editor. To set an attribute on text within the edit control the user sets focus into the edit control, moves the insertion point, selects text and presses shift-tab to move focus from the editor back to the toolbar. The user navigates through the toolbar (see toolbar behavior) to a desired attribute and invokes that attribute. When an attribute is invoked, that attribute is applied to the selected text in the editor and focus moves back into the editor at the previous insertion point with the selection intact.
  • Options:
    • Rather than using Shift + Tab to move focus from within the editor to the toolbar, another key combination could be used (Alt + Up Arrow, Control + Shift + Letter, etc.). This would eliminate the need to put the user interface control (in this example a toolbar) into the tab order immediately before the editor component. However, there are drawbacks to using a different keystroke to navigate to the user interface:
      1. It is not as "discoverable" as relying on the standard Tab/Shift + Tab behavior;
      2. It is difficult to find key combinations which are not already captured by the browser or assistive technology.
      3. Focus could stay within the toolbar after the user invokes an attribute. The user would then have to press an additional key to move focus back into the editor. This would allow multiple attributes to be set on the current selection without having to return back to the user interface but it would add an extra key sequence after setting just a single attribute. Requiring a keystroke to move focus back into the editor would also require modifying the toolbar behavior to intercept this keystroke and to know how to set focus back to the component (the editor) that the toolbar is associated with.

Optionally, if the developer wishes to provide the ability to insert a tab into the document, it is recommended one of the following methods be used.

  • Provide indent and outdent buttons in the menu. Keyboard shortcuts to the buttons should be Control + M for indent and Control + Shift + M for outdent.
  • Provide a button in the menu to toggle the use of Tab between the two modes. If this button is used, then Control + M is recommended as a keyboard shortcut to toggle the button.

WAI-ARIA Roles, States, and Properties

Authors are advised to not use ARIA for rich text editors, but to rely on native HTML markup. Current rich text editors typically use an iframe element for editable content. As a result, the editable content is implicitly mapped to a document role in accessibility APIs. If using HTML5, it is recommended that authors use the designMode or contentEditable attributes.


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Tree Grid

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

A grid whose rows can be expanded and collapsed in the same manner as for a tree. A Tree Grid is a combination of a Treeview and a Table with rows that are expandable.

Keyboard Interaction

There are two modes of keyboard interaction:

  • Navigation Mode (Read-Only) is the default mode, and allows quick and intuitive navigation around the grid.
    • Tab
      • The initial tab enters the grid with focus on the first cell of the first row, often a header.
      • Once in the grid a second tab moves out of the grid to the next tab stop.
      • Once focus is established in the grid, a Tab into or a Shift + Tab into the grid will return to the cell which last had focus.
    • Left Arrow and Right Arrow keys navigate between columns. If the next cell in the row is empty, focus should not move.
    • Up Arrow and Down Arrow The down arrow moves focus to the first column of a child row, if expanded. Otherwise focus is moved to the same column in the next row. Up arrow performs the same navigation but in reverse.
    • Control + Left and Control + Right Arrows expand or collapse rows.
    • If the cell contains an editable field, the Enter key starts edit mode and the Escape key exits edit mode.
    • Selecting Cells
      • Control + Space selects the current column.
      • Shift + Space selects the current row.
      • Control + A selects the entire grid.
      • Shift + Arrow selects contiguous cells.
      • Shift + F8 Allows additional cells to be added to a previous selection to accomplish non-contiguous selection.

      See Global Recommendations for information on cut, copy and paste.

The author may choose to indent child nodes visually. This should be done with an appropriate number of spacer cells marked as presentation in order to keep the headers aligned.

If cells are used for padding or layout of the hierarchy, navigation to those presentational cells should be prevented.

  • Actionable Mode (Interactive) allows the interaction with other objects that might be found in the grid cells such as edit fields, links, etc.
    • F2 pressed anywhere inside the grid will enter Actionable Mode. Focus will not be moved.
    • Enter pressed while focus is on an actionable item will enter Actionable Mode. Focus will remain on the actionable item that has focus.
    • Optionally, alphanumeric keys pressed while focus is on an actionable item will enter Actionable Mode. Focus will remain on the actionable item that has focus.
    • Tab will move to the next actionable (tabbable) item in the grid and stay within the grid wrapping at the bottom. In this mode each tabbable object in each cell of the grid can be reached with the tab key. If multiple tabbable items are located inside a single grid cell, the tab will stop at each one. When the last tabbable item in a cell is reached the next tab will move to the next tabbable item in the grid, wrapping at the last tabbable item in the grid.
    • Shift + Tab moves to the previous actionable (tabbable) item in the grid and stays within the grid, wrapping at the top.
    • Escape exits Actionable mode (by which the user may enter text or perform an action to complete a operation) and returns to Navigation Mode (where the user is allowed to move focus among elements). If a widget is in the current grid cell that also uses the Escape key, then it should cancel the event propagation. A subsequent press of the Escape key will return focus to the parent widget.

WAI-ARIA Roles, States, and Properties

Uses the WAI-ARIA role treegrid, and requires the child element row. By default a treegrid is considered to be editable, meaning all gridcells are editable.

  • To make a treegrid read-only, set aria-readonly="true" on the document element having a role="treegrid." This will make all gridcells read-only.
  • To override the read-only status on an individual gridcell, set its aria-readonly property to false.


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

  • For a visual (not accessible) example of where padding cells have been implemented see: dojo's TreeGrid example.
  • An example where padding cells have not been used (also not accessible): Oracle's Tree Grid.


This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

A sequence of dialogs or panels guiding the user through performing a task.

Keyboard Interaction

A Wizard can be done in several ways. Either is valid.

  • Method 1: Like a Toolbar
  • Method 2: Controls as Default Actions
    • Escape cancels the wizard.
    • Enter invokes the "next" action; If the last page, it invokes "finish"
  • Method 3: Hot Keys
    • Control + Alt + N next, finish
    • Control + Alt + P previous
    • Escape cancel, exit without saving
    • Control + Alt + R reset current page to default settings
    • Control + Alt + S save and exit
  • Method 4: Like a Dialog

Authors should take care when using Enter to trigger default actions since Enter might be connected to and trigger some other user interface element, or it might trigger the focused element. Authors should ensure that Enter activates only the widget they intend.

WAI-ARIA Roles, States, and Properties


Any examples referenced here that are hosted outside may have changed and may not accurately exemplify the guidance in this section. The APG task force is developing examples for APG version 1.1 that will be directly incorporated into the guide.

Dojo nightly

General Steps for Building an Accessible Widget with WAI-ARIA

This content could be useful when working on issue 73.

At this point you should have a basic understanding of how WAI-ARIA is used to support interoperability with assistive technologies. If you are not reusing a WAI-ARIA-enabled widget library and wish to create your own the following steps will guide you through the thought process for creating an accessible widget using WAI-ARIA.

  1. Pick the widget type (role) from the WAI-ARIA taxonomy

    WAI-ARIA provides a role taxonomy ([[WAI-ARIA]], Section 3.4) constituting the most common UI component types. Choose the role type from the provided table. If your desire was to create a toolbar set the role to toolbar:

    <div role="toolbar">
  2. From the role, get the list of supported states and properties

    Once you have chosen the role of your widget, consult the [[WAI-ARIA]] for an in-depth definition for the role to find the supported states, properties, and other attributes. For example, the toolbar role definition includes:

    superclass role
    In the taxonomy the widget you selected inherits states and properties from this role. In the case of a toolbar you will see that a toolbar is a subclass of a group. This makes sense given that a toolbar is a collection of commonly used functions.
    related concept
    This is really more informative to state what other concepts are similar to this role. These may exist in different host languages outside WAI-ARIA. The keyboard model for the control should emulate that of the related concept control.
    supported states and properties
    These are unique states and properties that this widget supports and that were not inherited from its ancestors in the taxonomy. In the case of a toolbar there are no such states or properties. However, in the case of a listbox, you may choose to set the property of aria-multiselectable to true if you were to have more than one item in the listitem selected at a time. This indicates to the assistive technology that the listbox manages a collection of selectable options.
    inherited states and properties
    These are all the states and properties which are inherited from the roles's ancestors and which you may use.
    global states and properties
    These are states and properties which apply to all host language components regardless of whether a role is set or not. You may use these as well.

    Once you have chosen the states and properties that apply to your widget you must set those properties you will use to set their initial values. Note: You do not need to use all the states and properties available for your role. In our case we shall use:

    <div role="toolbar" id="customToolbar" tabindex="0" aria-activedescendant="button1"
          onkeydown="return optionKeyEvent(event);"
          onkeypress="return optionKeyEvent(event);"
          <img src="img/btn1.gif" title="Home" alt="Home" role="button" id="button1"
               onclick="updateText('Home was invoked');">
          <img src="img/btn2.gif" title="Refresh" alt="Refresh" role="button" id="button2"
               onclick="updateText('Refresh was invoked');">
          <img src="img/btn3.gif" title="Help" alt="Help" role="button" id="button3"
               onclick="updateText('Help was invoked');">

    By setting tabindex=0 on the toolbar, the toolbar will receive focus in the document order. It is necessary then to use script and the aria-activedescendant property to manage virtual focus among the buttons. The details are given in step five, below.

    Important: When embedding WAI-ARIA markup in (X) HTML, all WAI-ARIA states and properties must be preceded with the characters aria- with the exception of the role and tabindex attributes. Otherwise, the user agent will not map the WAI-ARIA information, resulting in it not being recognized by assistive technologies.

    When embedding WAI-ARIA into other host languages, tabindex does not carry over. The WAI-ARIA tabindex extensions are specific to (X)HTML to repair deficiencies in keyboard support.

  3. Establish the widget structure in the markup (parent/child)

    Assistive technologies are very dependent on the structure of widgets as well as general document structure. Structure provides context to the user. A toolbar is a collection of common functions made available to the user. Therefore, all functions you would like in the toolbar must be contained within it. This can be determined by using the [[dom]] tree structure created by the browser when parsing the host language. By using the parent child relationship of the DOM, an assistive technology can determine all the related toolbar widgets associated with the toolbar. The toolbar widgets would be DOM children of the "toolbar" container. For our purposes we will define three image buttons for cut, copy, and paste.

    <div role="toolbar" tabindex="0" aria-activedescendant="button1">
      <img src="buttoncut.png" alt="cut" id="button1">
      <img src="buttoncopy.png" alt="copy" id="button2">
      <img src="buttonpaste.png" alt="paste" id="button3">
  4. Repeat steps 1-3 for the children of the parent

    We now need to assign the roles and states for each of the children. However, we shall save the detailed navigation for step 5.

    <div role="toolbar" tabindex="0" aria-activedescendant="button1">
      <img src="buttoncut.png" alt="cut" role="button" id="button1">
      <img src="buttoncopy.png" alt="copy" role="button" id="button2">
      <img src="buttonpaste.png" alt="paste" role="button" id="button3">

    The process of setting roles and states may be a recursive procedure if the children themselves have children, such as in the case of an expandable/collapsible tree widget.

  5. Establish keyboard navigation of the widget and plan for how it will be navigated to within the document

    It is very important that your widget be keyboard accessible. In fact, there must be a keyboard equivalent for every mouse operation. Where possible you should refer to the WAI-ARIA examples in this guide for tips on how to implement keyboard navigation for your widget. If you find that an example is not provided, you should follow standard keyboard bindings for UI components such as those used for the Java Foundation Classes for Windows 95/NT.

    Commented out link to conference paper, which a) was broken, and b) we should have more canonical references than conference papers for things like this.

    For our toolbar, we have chosen to have the toolbar manage the focus for its children and through the use of the aria-activedescendant property. We have also chosen to have the toolbar receive focus based on the tab order by using tabindex. In order to use aria-activedescendant, each focusable descendant must have an assigned ID.

        function optionKeyEvent(event)
          var tb =;
          var buttonid;
          DOM_VK_ENTER = 13;
          // Partial sample code for processing arrow keys
          if (event.type == "keydown") {
             if (event.altKey) {
               return true;  // Browser should use this, the menu view doesn't need alt-modified keys
             // XXX Implement circular keyboard navigation within the toolbar buttons
             if (event.keyCode == DOM_VK_ENTER) {
               ExecuteButtonAction(getCurrentButtonID()); // This is an author defined function
             else if (event.keyCode == event.DOM_VK_RIGHT) {
               // Change the active toolbar button to the one to the right (circular) by
               var buttonid = getNextButtonID();   // This is an author defined function
               tb.setAttribute("aria-activedescendant", buttonid);
             else if (event.keyCode == event.DOM_VK_LEFT) {
                // Change the active toolbar button to the one to the left (circular) by
                var buttonid = getPrevButtonID();  // This is an author defined function
                tb.setAttribute("aria-activedescendant", buttonid);
             else {
                return true;
             return false;
          else if (event.type == "keypress") {
    <div role="toolbar" tabindex="0" aria-activedescendant="button1" id="tb1"
      onkeydown="return optionKeyEvent(event);"
      onkeypress="return optionKeyEvent(event);">
      <img src="buttoncut.png" alt="cut" role="button" id="button1">
      <img src="buttoncopy.png" alt="copy" role="button" id="button2">
      <img src="buttonpaste.png" alt="paste" role="button" id="button3">

    The details of implementing keyboard navigation are described in Keyboard and Structural Navigation section of this document.

    You must also show the visual focus for each element that has focus.

  6. Apply and manage needed WAI-ARIA states in response to user input events

    Similar to the processing of aria-activedescendant in Step 5, as author you must set any additional WAI-ARIA states and properties on document elements.

  7. Synchronize the visual UI with accessibility states and properties for supporting user agents

    Thomas comment: is confusing that the example switches from toolbar to treeitem. Maybe the best overall sample is to do a tree because it demonstrates each of the points you want to make?

    You should consider binding user interface changes directly to changes in WAI-ARIA states and properties, such as through the use of CSS attribute selectors. For example, the setting of the aria-selected state may change the background of a selected treeitem in a tree. This may also be done with JavaScript.

    .treeitem[role="treeitem"][aria-selected="true"] {color: white; background-color: #222222;}
    .treeitem[role="treeitem"][aria-selected="false"] {color: white; background-color: beige;}        

    Authors should be aware that CSS attribute selectors are not supported in some browsers, such as Internet Explorer 6. A consistent way to apply styling to reflect WAI-ARIA semantics would be to assign an element a class name based on the WAI-ARIA attribute being set using script as shown here:

    function setSelectedOption(menuItem)
            if (menuItem.getAttribute("role") != "menuitem") {
            var menu = getMenu(menuItem);
            var oldMenuItem = getSelectedOption(menu);
            // Set class so that we show selection appearance
            menuItem.className= "selected";
  8. Showing and Hiding Sections in a Widget

    The proper synchronization of showing and hiding sections in a widget with the WAI-ARIA display state is also critical. Some platform accessibility APIs provide events for applications to notify the assistive technology when pop-ups such as menus, alerts, and dialogs come into view or go away. Rich Internet applications can assist browsers which support these conventions by:

    1. Creating an entire section and then insert it into the [[dom]], as a subtree of the parent element activated to show the pop-up, and then removing the section from the inserted area when the pop-up goes away.


    2. Using the following style sheet properties to show and hide document sections being used to represent the pop-up items, menus or dialogs:

      • display:block
      • display:none
      • visibility:visible
      • visibility:hidden

      By monitoring these behaviors a user agent may use this information to notify assistive technology that the pop-up has occurred by generating the appropriate accessibility API event.

    Some assistive technologies may use the DOM directly to determine these when pop-up occurs. In this case, the first mechanism of writing a section to the DOM would work using the DOM events as demonstrated here.

        // create new table row with table cell and div
        var newTr = document.createElement('TR');
        var newTd = document.createElement('TD');
        var newDiv = document.createElement('DIV');
        //insert this new table row before the Node selected
        var container = theNode.parentNode;
        container.insertBefore(newTr, theNode);
        //remove theNode selected

    However, if you are using CSS to show and hide sections of the DOM (2) it is essential that you set the corresponding WAI-ARIA aria-hidden property to indicate that the section is visible or hidden and synchronize it with your CSS styling as shown here:

    [aria-hidden=true] {visibility: hidden;}
    <div role="button" aria-haspopup="true" aria-owns="mypopupmenu">
    <div role="menu" aria-hidden="true" id="mypopupmenu">…</div>
  9. Support basic accessibility, such as alternative text on images

    When an image is used to represent information within a component, such as image buttons, you need to set the alternative text on those images. This is then mapped by the user agent to the accessible name in the platform accessibility API. Using our example:

    <div role="toolbar" tabindex="0" aria-activedescendant="button1" id="tb1"
         onkeydown="return optionKeyEvent(event);"
         onkeypress="return optionKeyEvent(event);">
       <img src="buttoncut" role="button" id="button1" alt="cut">
       <img src="buttoncopy" role="button" id="button2" alt="copy">
       <img src="buttonpaste" role="button" id="button3" alt="paste">
  10. Establish WAI-ARIA relationships between this widget and others

    Once you have made the basic widget accessible you may then need to establish its relationship to other widgets. Examples of this are aria-labelledby, aria-controls, aria-describedby and aria-flowto. The details of using these relationships are described in the Relationships section of this document.

    Other relationships which should be considered are more declarative and provide context to the widget within a set. For these, aria-level, aria-posinset, and aria-setsize are provided. If you structure your Document Object Model appropriately so that the user agent can determine this information from it using the DOM hierarchy directly, you do not need to set these properties. There are, however, situations in rich internet applications where all the elements in a container are not in the DOM at one time, such as when you can only render the ten of fifty items in a subtree. In this case the user agent cannot determine the number of tree items (aria-setsize) for the position in the set (aria-posinset), and potentially the tree depth (aria-level) from the DOM. In these situations you will need to provide these WAI-ARIA properties.

  11. Review widget to ensure that you have not hard coded sizes

    The ability for applications to respond to system font settings is a requirement. Most user agents are designed to meet this requirement. This also means your Web application running within your browser is impacted when the user agent changes the font sizes to meet the need. If you have hard coded your font size in pixels an increase in system fonts will not be reflected in your Web application. You must also not hard code the size of your widgets in pixels either. If the fonts are scalable, but the widget they are encapsulated in is not, then the text could flow outside your widget.

    Follow these rules to allow your application to respond to system font settings:

    • Establish a base set of font sizes used in widgets based on percentage of the container element font size.
    • Use CSS width, borders, margin, padding, background, and positioning properties to specify the graphical rendering of widgets and their sub-components, use percentage units or em units to specify widths of widget components (An em is a the font unit of measure between the top and bottom of an upper case letter M.). Border widths, padding, and margins can use PX units.
    • Use scripting for run time CSS positioning of widget sub-components in relation to other sub components.
    • Make sure all widgets use consistent height and width units of measure.

    Percentages are the most reliable way to consistently specify proportional text sizes in widgets. The use of percentages and em should be used for specifying widths of a widget and the widget sub components. The use of percentages for text size and percentages and em units for widths support browser zoom capabilities to make widgets larger or smaller. Pixels can be used for specifying line widths, padding and margins.

    Important: Most browsers today have a zoom feature which allow the user to magnify the entire Web page. Most legislation today requires that your application respond to system font and color settings and therefore you will want to consider the fact the system settings could adversely affect your Web page should you decide to hard code pixel sizes.

  12. Compensate for Background Images when in High Contrast Mode

    Authors use background images when styling their widgets, including situations where the background image is not merely decorative, but informative. An example is a horizontal progress bar that it is filled by gradually revealing more of a background image. This is accomplished by initially setting the width of the element to zero, and then incrementing its width in accordance with the degree of progress.

    High contrast mode is an operating system display modification that makes the screen easier to see for low vision users. Some operating systems (e.g., Windows), do not display background images when in high contrast mode. Consequently, the progress bar described above appears empty regardless of the progress. It is recommended that authors not use background images as the sole method to convey important information, and to compensate with alternative or additional style rules.

    In the case of the progress bar example, a technique that works when in high contrast mode is to style the element with a border. Since the width of the element is updated as progress increases, its border gradually expands horizontally forming an ever wider unfilled rectangle. This provides alternative visual feedback as to the extent of the progress.

    Another technique is to replace a background image with text. Consider a dialog that uses a background image for its close box. To compensate for the missing close box when in high contrast mode, a lower case 'x' is used instead. The compensation technique used depends on the context, particularly the purpose of the background image.

    There are two general approaches with respect to detecting high contrast mode. They are (1) executing a script to determine if the system is in high contrast mode, or (2) providing a preference to use alternative styles. The advantage of automatic detection is that some operating systems simply apply a different color palette when in high contrast mode and do not turn off background images. In that case, authors need not compensate for missing background images. However, detection of high contrast mode by script is relatively expensive compared to a preference that users can set, and, presumably, users can see whether background images are displayed in high contrast mode on their system. It is up to individual authors to decide which method they find acceptable for their particular situation.

    The following code fragment outlines how to detect high contrast mode.

    /* Define styles for the high contrast test element */
    #hiContrastTestEl {
        border: 1px solid;
        border-color:red green;
        position: absolute;
        height: 5px;
        top: -999px;
        background-image: url('resources/blank.gif');
    // An onload event handler that inserts the high contrast test element and
    // then tests its computed styles.
    function detectHiContrast() {
        var div = document.createElement('div');
        div.setAttribute ('id', 'hiContrastTestEl');
        // … append <div#hiContrastTestEl> to the <body> element …
        var cs = window.getComputedStyle(div);
        var bki = cs.backgroundImage;
        var hiContrast = false;
        // The CSS sets the top and right borders of the test element to red and
        // green, respectively.  In high contrast mode either the borders are
        // the same color, or there is no legitimate url to the background image.
        hiContrast =    (cs.borderTopColor == cs.borderRightColor) ||
                        (bki != null && (bki == 'none' || bki == 'url (invalid-url:)'));
        if (hiContrast) {
            // apply hi contrast styles to compensate for missing background images.
        // … remove the test element from the document …
  13. Test with User agent, Assistive Technology, and People with disabilities

    To ensure you have set your WAI-ARIA semantics correctly, test your application with your user agent, an assistive technology, and a person with disability. Example assistive technologies are screen readers and screen magnifiers. Ensure that your user agent is designed to support WAI-ARIA and their your assistive technology is designed to support WAI-ARIA in the selected user agent.

    Finding people with disabilities to do testing may be difficult but many companies employ people with disabilities, including your customers, and you should reach out to them to ask for help. Other places to look are advocacy groups like the National Federation of the Blind or your local schools for the blind, reading and learning centers, and so on. The people you reach out to may someday need to use what you produce.

Providing Navigable Structure within Web Pages

Matt King: Removed on July 10 per discussion in May 23, 2016 meeting. This section is redundant given the new landmark section. The parts that are not redundant are either out-of-date or not necessary. The possible exceptions are content covering the group and directory roles.

This section takes a broader look at the Web page. It is intended to assist you in conveying a logical, usable, and accessible layout to an assistive technology or adaptive system designed to modify the visual layout to meet the users needs.

One of the deficiencies of (X)HTML for disabled users has been the usability of keyboard navigation. Users dependent on a keyboard for navigation have been forced to tab everywhere in the document, as the only keyboard accessible document elements are form and anchor elements. This has forced developers to make most everything a link to make it keyboard accessible, and to get to each link you have to tab to it. With the advent of portals and other content aggregation means Web pages are divided into visible regions and there has been no vehicle to get to them other than perhaps to do things such as:

There are number of problems with this approach:

This remainder of this section provides the author with a playbook for using WAI-ARIA to add semantically rich navigation structure in a Web page so that an assistive technology may provide an effective user experience and avoid these shortcomings.

Steps for Defining a Logical Navigational Structure

Due to the complexity of today's web content, it is imperative that all content reside in a navigation landmark. Sight and mobility impaired users must be able to quickly access a list of landmarks for a web page, to be used as a table of contents. This allows the user to quickly navigate to important sections of the page without "endless" TAB navigation. If this is not done, then content will be orphaned and missed by the user. In fact many of today's accessibility test tools have built-in rules to flag this issue as a compliance error. This section defines how to create these navigable regions of the page:

  1. Identify the logical structure of your page

    Break up your page into perceivable block areas which contain semantically associated information called "regions". You can further break down each region into logical regions as needed. This is a common process undertaken by portal developers who break the page into perceivable regions called portlets. Think about the user wanting to navigate to these logical block areas on your Web page. Write down a description of what you believe each region to be.

    Depending on your Web application you may then need to break it into sub regions depending on the complexity of your Web application. This is a recursive process. A user will look at these perceivable areas like pealing an onion. You might have an outermost collection of perceivable regions or block areas and upon entering each region you may have a collection of regions within it.

  2. Implement the logical structure in markup

    Implementing the block structure in markup often involves wrapping elements contained with a "region" such as a <div> or <iframe> with visible borders so that each region or block is perceivable to the user.

  3. Label each region

    Once you have broken up each region you need to label it. The start of each region must have a perceivable header and it should be used to label the region. For details on how to do this see section Header Levels vs. nesting level to create a header to label each region. If you're finding it difficult to come up with a label for the region, make sure at least to use one of the standard roles defined in the following step. The drawback of not providing a label is that users will not know the purpose of the region. By definition, regions would be included in a summary of a page. If an assistive technology were to summarize it would be unable to provide information about the section without a label.

  4. Assign landmark roles to each region

    Now that you labeled your regions, you need to assign them semantic navigation landmarks. Landmarks are a vast improvement over the rudimentary "skip to main content" technique employed prior to WAI-ARIA. If possible it is best to use these as landmarks. Skip to main content links may impact the placement of web application elements. This may be a consideration for developers that want to try to keep their web application in a specific visible area. Consequently, layout adjustments may need to be made to support skip to main content links. Also, skip to main content links are limited in that they only address only one main content area. WAI-ARIA provides a collection of landmarks which are applied to each of the regions you identified in step 2.

    The presence of common, semantic, navigation landmarks allows each site to support the same standard and allows your assistive technology to provide a consistent navigation experience - an important feature for screen readers and alternate input solutions. For users with cognitive and learning disabilities the landmark information could be used to expand and collapse these regions of your page to aid in simplifying the user experience by allowing the user to manage the amount of information processed at any one time.

    WAI-ARIA landmark roles that are applied to elements having strong native semantics are not mapped consistently to the platform accessibility API. An example is the TABLE element. For this reason, it is recommended that authors apply landmarks to DIV and SPAN containers.

    There are also mainstream benefits of providing navigation landmarks. Your browser may assign key sequences to move focus to these sections as they can be set on every site. Navigation to these landmarks is device independent. A personal digital assistant (PDA) could assign a device key to get to them in your document. The common landmarks in WAI-ARIA include:

    Represents a region of the page representing a unique software unit executing a set of tasks for its users. It is an area where assistive technologies should also return browse navigation keys back over to the web application in this region.

    If the entire web page has a role of application then it should not be treated as a navigational landmark by an assistive technology.

    A region that contains the prime heading or internal title of a page.
    Any section of the document that supports but is separable from the main content, but is meaningful on its own even when separated from it.
    Meta information which applies to the first immediate ancestor whose role is not presentation.
    A region of the document that represents a collection of form-associated elements, some of which can represent editable values that can be submitted to a server for processing.
    Main content in a document.
    A collection of links suitable for use when navigating the document or related documents.
    The search tool of a Web document.

    To set a landmark in your page:

    <div role="navigation" title="Table of Contents"> … </div>
    <div role="main" title="Game Statistics"> … </div>
  5. For application landmarks with static prose

    If you have a regional landmark of type application and static descriptive text is available, then on the application landmark, include an aria-describedby reference to associate the application and the static text as shown here:

    <div role="application" aria-labelledby="calendar" aria-describedby="info">
    <h1 id="calendar">Calendar<h1>
    <p id="info">
    This calendar shows the game schedule for the Boston Red Sox.
    <div role="grid">
  6. For landmarks unsuited to specialized region WAI-ARIA roles

    You can create custom regional landmarks by using a generic region. While it is not essential to label these specialized regions with a header, you should provide a title to ensure that the region name is accessible to all users, just as you would the standard regional landmarks. See Header levels versus Nesting levels for directions on how to label the region.

    <div role="main"> … <div role="region" title="weather">

    Note: the region role is generic and should only be used when a more specific role is not applicable.

  7. Indicate to the assistive technology who controls the keyboard navigation

    Today's screen readers for the blind have been designed to give the user a "browsing" experience meaning the screen reader consumes a number of keys to aid in the browsing experience. For example, the "up" and "down" arrow keys are used to read the next and previous line of text on your Web page. Accessible rich internet applications will use these keys in an application to navigate within "widgets" like a Tree View.

    Assistive technologies must be able to identify widgets which implement WAI-ARIA and allow them use of the keyboard. However, if you have used numerous widgets to form an application you must set the role on the encompassing region to application. Should you have regions which should be browsed as documents within the application region you should mark those regions with a document role as needed. See section 3.4.2 Structural Roles used to facilitate assistive technology navigation.

Structural Roles that Facilitate Navigation with Assistive Technologies

While WAI-ARIA is designed to address many disabilities, this section is best described in terms of screen reader use. In desktop applications, screen readers typically treat widgets as discrete entities, reading and interacting with each widget one at a time. The user moves the point of focus from widget to widget using tab/shift tab, mnemonics, or accelerators, keyboard commands which are usually provided by the application or the operating system. We refer to this mode of interaction as "application mode."

When viewing Web content however, screen readers often gather information about all the widgets in an area and present them in a document-like view which the user navigates using keyboard commands provided and controlled by the screen reader. Think of this mode as a virtual environment that presents Web content in a way that makes it convenient for adaptive technology users to navigate and read. This is sometimes called browse mode, or virtual mode. We refer to this as "document browse mode."

Because many screen readers provide document mode navigation support using single key mnemonics on the alpha-numeric keyboard, they may provide a third mode, called "forms mode," used to interact with form controls that are encountered in document mode. Behavior in forms mode is similar to that of application mode. The key feature of forms mode is that it can be toggled with document mode to make it easy to both interact with a specific widget, and read virtualized content of which the widget is a part. Since, as described above, a screen reader's perception of an area as either a document or an application greatly influences how the user reads and interacts with it, WAI-ARIA provides content authors a way to indicate whether their pages must be viewed as applications or documents by assistive technologies.

To set document or application mode follow these steps:

  1. Set the Application Role

    After you have divided your Web page into regions through the use of role landmarks and custom regions, you must make a decision: Is your Web page an application or not? If the majority of the content is application-like the default interaction mode should be set to application, with document regions embedded within the application if applicable. Otherwise the default interaction mode is document-like and therefore should not be overridden by the use of a global role of application. In this case the application role can be placed on discrete regions within the document if applicable.

    If it is, set the role of application on the body tag as follows:

    <body role="application"> 

    When using application, all static text must be associated with widgets, groups or panes via using the aria-labelledby and aria-describedby properties, otherwise it will not be read by the screen reader when the user navigates to the related widget or group.

    Special Considerations:

    • If you have selected to have a role of application on the body tag, it is recommended that a widget have focus after the page is first loaded. There may be an instance when an application itself needs to receive focus, such as an application consisting solely of a scrollable editable text area.
    • If when your page loads and you should have focus on a widget this is a strong indicator that your Web page should have role of application.
    • If your page has only a few isolated widgets, like pop-up calendars located on a Web page, it is not necessary to expressly set the role of application on the body. Screen readers, based on widget roles, must be able to provide access to these widgets without recognizing the entire page as an application.
    • Also, browsers make use of a feature called the contenteditable attribute, which will be incorporated into HTML 5. Contenteditable allows an author to turn the browser section into a rich text editor, similar to a word processor. Any section which has contenteditable set to "true" is considered a widget.
    • If the body element has been given the role of application, please follow step 3. Otherwise, the Web page is considered a document, and no further action is required in this regard.
  2. Dialogs and alert dialogs - special case application roles

    WAI-ARIA provides dialog and alertdialog roles that are to be treated as special case application roles. Like application, screen readers will leave the main job of keyboard navigation up the dialog.

    When using dialog, all static text must be associated with widgets, groups or panes using the aria-labelledby and aria-describedby properties, otherwise it will not be read by the screen reader when the user navigates to the related widget or group. Unlike an application, your Web page is unlikely to start out as either of these two roles but rather the author is most likely to dynamically create dialogs or alertdialog sections within the Web page.

    Unlike dialog, descriptive text of the alert does not need to be associated via a relationship, as the contents of alert dialogs will be read automatically by the screen reader when the dialog opens.

  3. Set the document role

The default mode for a screen reader to be processing an HTML Web page is document browse mode. This is equivalent to having a role of document on the HTML <body> tag. If you have already specified a role of application on the body tag there may be times in which you tell the screen reader to switch into "document browse mode" and start stealing the necessary keys to browse a document section of your Web page. These keys are the typical keys used by WAI-ARIA widgets and to provide this level of navigation the keys must be stolen from your browser.

To mark areas of your page to tell the assistive technology when to switch into document browse mode, look at the regions/landmarks you have defined and determine which ones must be browsed as a document or navigated as an application. For each region which must be browsed in document browse mode, embed a div element within it with the role of document as follows:

<div role="document"> 

Now, when a screen reader encounters this region, it will change its interaction model to that of document browsing mode.

Implicit Nesting and Headings

This section discusses the use of the heading role and nesting levels.

Header Levels Versus Nesting Levels

The heading role value signifies a heading for a section of the document instance. Use of the heading role indicates that a specific object serves as a header. The region of the document to which the heading pertains to should be marked with the aria-labelledby property containing the value of the id for the header. If you have a heading and there is no element containing the content that it heads, wrap the content in a <div> bearing this aria-labelledby attribute. If headings are organized into a logical outline, the aria-level property can be used to indicate the nesting level. Here is an example with and without ARIA. Note that the preferred way is using native semantics without ARIA, but the approach with ARIA is shown first to demonstrate use of the heading role:

Using ARIA:
<div role="main" aria-labelledby="page_title">
  <p id="page_title" role="heading" aria-level="1">Top News Stories</p>
   … main page contents here …

Using native markup:
<div role="main" aria-labelledby="page_title">
  <h1 id="page_title">Top News Stories</h1>
   … main page contents here …
Owns Repairs Nesting

Assistive technology briefs users on the context where they are. When they arrive at a new page, a page summary may be given. When they move into a new context, some of the labeling from elements containing the new focus or reading location may be rendered by the assistive technology, to give context to the details to be read next.

The syntactic structure of a page provides the default nesting of contexts. If a paragraph is nested in a <div> or table cell, it is assumed that labels for the <div> or headers for the table cell are pertinent to what is in the paragraph. On the other hand, it is not possible to always flow the logical structure one-to-one into the parse structure.

The aria-owns relationship is provided to annotate logical nesting where the logical child is not a syntactic descendant of the logical parent. In a Rich Internet Application, updates may be made to the document without updating the logical syntactic structure, yet elements may appear to be added to the document structure. From a DOM and accessibility hierarchy perspective aria-owns is a fallback mechanism to using the tree hierarchy provided in the DOM. An example of where aria-owns is needed is a treeitem, where children in a folder subtree are added to a visible subtree but not reflected in the actual document subtree of the folder. The aria-owns relationship can be used to associate the folder with the new "adopted" child. For more details on the use of aria-owns see section 4.2 Owning and Controlling. The aria-owns relationship is used to indicate to a user agent that a menu is an adopting parent of a sub menu. Another use for aria-owns is a hierarchical diagram where the child nodes of the diagram are not be adequately represented using the DOM structure.

Directories, Groups, and Regions

There are several WAI-ARIA roles used to indicate a collection of user interface objects, and each has a distinct purpose:

Contains a static table of contents
Small set of related user interface objects that would not be included in a page summary or table of contents by an assistive technology
Section of related user interface objects that should be included in a page summary or table of contents.

The use of each is described below.

Directories and Their Applicability

The WAI-ARIA role, directory, allows authors to mark static table of content sections of a document. Prior to WAI-ARIA, the user would need to guess if an area of a document actually pertained to the table of contents. Authors should mark these sections within a document with a role of directory.

<div role="directory">
      <li>Global Warming Causes
             <li>CO2 Buildup</li>
             <li>Auto emissions<li>
             <li>Factory emissions</li>
             <li>Destruction of rainforests</li>


      <li>Tailoring CO2 buildup
             <li>Elimination of the incandescent light bulb</li>
             <li>Hydrogen fuel cells</li>
             <li>Solar energy</li>
             <li>Wind power</li>


Groups and Their Applicability

Authors should use a role of group to form logical collections of items with related functionality in a widget. A group should not be considered a major perceivable section on a page. A group neither has a heading nor appears in the "Table of Contents." A group may delineate a set of treeitems having the same level in the tree or it may be used to group a set of related checkboxes in a form. Other examples include:

  • a row in a grid (a row is a specialized group representing a row in a grid);
  • a group of children in a tree widget forming a collection of siblings in a hierarchy; or
  • a group of items having the same container in a directory

Proper handling of a group by assistive technologies, therefore, is determined by the context in which it is provided. Group members that are outside the DOM subtree of the group need to have explicit relationships assigned for them in order to participate in the group. Groups may also be nested.

If an author believes that a section is significant enough in terms of the entire document instance, then the author must assign the section a role of region or one of the designated standard landmarks. The designated landmark roles are listed in the section Regions and XHTML Role landmarks.

Regions and Their Use

When defining a region for a section of a document, authors must consider using standard document landmark roles defined in the XHTML Vocabulary. This makes it possible for user agents and assistive technologies to treat roles as standard navigation landmarks. If the definition of these regions is inadequate, authors must use the WAI-ARIA region role and provide the appropriate title text.

For more information on the use of region see Regions and XHTML Role landmarks.


Remaining Structural Roles

Tabular Structure-Related Roles Supporting Tabular Widgets

A number of structural roles support the tabular widgets, grid and treegrid. These structural roles indicate additional keyboard navigation as well as the ability to select rows and/or columns. Typically, you would apply these roles to an underlying table in the base markup as shown here:

<table role="grid">    

However, that may not work for the developer in all instances, such as when the developer has need for a <div> or <span> or when additional semantics are needed. To assist, the following roles are provided to support tabular widgets:

When constructing a grid or treegrid the author must use gridcells for the actual cells:

<table role="grid">
      <td role= "columnheader">Apples</td><td role= "columnheader">Oranges</td>
      <td role="gridcell">Macintosh</td><td role="gridcell">Valencia</td>

Unlike table cells within a table, authors should ensure a grid's gridcells are focusable through the use of tabindex (in HTML), or aria-activedescendant on the grid.They may also be editable, as is shown in the above example.

Treegrid's may require expanding and collapsing rows which may not be performed using a <tr>. In these instances authors will use an HTML <div>. WAI-ARIA provides a role of row which may be assigned to the <div> to convey to the assistive technology that this is still a row. If you decide to not use the native HTML <table> elements and choose more flexible elements, such as <div> and <span>s, with applied WAI-ARIA roles in this section, you should structurally lay out your elements in line with what you would expect from HTML. All of your rowheaders should be in a row and your gridcells should be children of each subsequent row in the same format as HTML for <td>s and <th>s within <tr>s.

Marking Descriptive Sections

A new feature of WAI-ARIA is the ability to associate descriptive text with a section, drawing, form element, picture, and so on using the aria-describedby property. The use case for aria-describedby is to reference the local id of the long description. The most common use case for longdesc is for the author to create a separate file to describe a picture. It is preferable to have the descriptive text in prose as well so that it is readily available to all users. This is useful for authors who do not want to create a separate document that contains the description, and for cognitively impaired users who can be confused by context changes when having to go to another page for the description. Yet, like longdesc, descriptive text is treated separately from the short name provided by the title or alt attributes in HTML.

The aria-describedby property allows for the text of hidden elements to be used for accessible descriptions. Some screen reader vendors have used non-standard attributes to provide hidden help text for form elements. In order to standardize this feature, hidden text is exposed in the description property of Accessibility API objects when referenced by aria-describedby.

<img src="foo" alt="abbreviated name" aria-describedby="prose1">

<div id="prose1">
   The prose in this div describes the image foo in detail.

This is the preferred vehicle for providing long descriptions for elements in your document.

When the author does not desire the entire descriptive text to be located on the main page, aria-describedby can also be used to point to a link to another page.

<div id="figuretitle">Figure 1-1: Entity Relationship Diagram showing EMP and DEPT</div>
  <img src="foo" aria-labelledby="figuretitle" aria-describedby="link1">
  <a href="descriptionLocation.html" id="link1">Description of Figure 1-1: Entity Relationship Diagram showing EMP and DEPT</a>

It is not good practice to use the above pattern when the describing element—the <a> tag with @id='link1'—is hidden, since there is no way for a user to navigate to and activate the link. Use the technique only when the description is not hidden.

Miscellaneous XHTML Section Roles

In order to synchronize with the XHTML Role Attribute module, WAI-ARIA includes two XHTML roles which are neither landmarks nor widgets of any kind. These roles were incorporated to replace standard elements found in host languages. These roles are definition and note. If either role has a corresponding element in the host language, it is recommended that authors use the corresponding element in the host language.

The definition role indicates a definition statement in your document. For HTML developers implementing lists of definitions, we recommend you using the DL, DT, and DD elements rather than marking an arbitrary element with a role of definition. An arbitrary element would be appropriate for inline definitions used in conjunction with the DFN element.

Example of an inline definition with an explicit labelledby relationship:

<p>The doctor explained it had been a <dfn id="placebo">placebo</dfn>, <span role="definition" aria-labelledby="placebo">
an inert preparation prescribed more for the mental relief of the patient than for its actual effect on a disorder.</span>

Example of an inline definition with an implicit labelledby relationship determined by nesting:

<p>The doctor explained it had been a <span role="definition">
<dfn>placebo</dfn>, an inert preparation prescribed more for the mental relief of the patient than for its actual effect on a disorder.</span>

Note: In the case where host language semantics do not allow for implicit nesting of definitions terms inside definition roles, authors should explicitly use the aria-labelledby attribute, even when the definition term is nested in the definition as shown here:

<p>The doctor explained it had been a <span role="definition" aria-labelledby="placebo">
<dfn id="placebo">placebo</dfn>, an inert preparation prescribed more for the mental relief of the patient than for its actual effect on a disorder.</span>

The following is an example using a definition list:

 <dt id="anathema">anthema</dt>
 <dd role="definition" aria-labelledby="anathema">a ban or curse solemnly pronounced by ecclesiastical authority and accompanied by
 <dd role="definition" aria-labelledby="anathema">a vigorous denunciation : cursor</dd>
<dt id="homily">homily</dt>
<dd role="definition" aria-labelledby="homily">a usually short sermon</dd>

<dd role="definition" aria-labelledby="homily">a lecture or discourse on or of a moral theme</dd>


The note role defines a parenthetic or ancillary remark to the main content of the resource. It also allows assistive technologies to skip over this section of the document unless more information is requested about the main content.

<div role="main" aria-labelledby="foo">
   <h1 id="foo">Wild fires spread across the San Diego Hills</h1>
   Strong winds expand fires ignited by high temperatures …

<div role="note">
   This article was last updated July 30, 2008 at 6:05PM.

Other Widget Authoring Practices

MCK: Setting this content aside to determine whether we will remove or rewrite it.

The Common Widget Design Patterns section of this document contains best practices, such as keyboard navigation, for creating common widgets found on the Web. This section contains miscellaneous authoring practices that you should also consider.

Managing Focus with Scroll

Not yet edited ... considering what should be written/included. We don't want to write a javascript book.

In some browsers, a JavaScript call to scrollIntoView() on this element should suffice, but in browsers where this is unreliable, authors should explicitly set the scrollTop and scrollLeft properties of the "cell23" element and its ancestors to scroll the element into view. scrollTop and scrollLeft adjust the node positions by the amounts(pixels) needed to scroll a node into view. Scrolling values are adjusted by the amounts(pixels) needed to scroll a node into view. This is done by comparing the sizes of the nodes using available measurements such as scroll+offset+clientWidth/Height/Left/Top. It's important to note that you have to adjust a node so that it's viewable within the context of its parent node. Then you have to move up the DOM tree and make each parent node visible.

For example, create a custom scrollIntoView() method that is called at various times, including coincidence with the setting of the aria-activedescendant property. The method takes a DOM node argument, say "n". Here is the high level algorithm:

  1. If n is already in view, stop; otherwise, continue.
  2. adjust n.scrollTop and n.scrollLeft such that it is in view.
  3. adjust scrollTop and scrollLeft for the ancestor nodes of n such that that the region of the ancestors which n consumes is visible.

This is a minimum-position-change algorithm.

Understanding how the DOM scrollIntoView works across browsers is important. Browsers (including Firefox3) force the node either to the top or the bottom of the screen (as defined by the single Boolean parameter) even if its already in view. This is problematic when you only need to scroll horizontally to see the element. It is also problematic when the node is partially off the bottom of the screen and the parameter is (true) which forces the node all the way to the top, and vice versa with the top going to the bottom on (false). IE forces the node to the left of the client area (or right in right-to-left mode) even if it's horizontally in view already.

The scrollTop and scrollLeft functions create some challenges. scrollTop is always accurate but misleading with respect to inner (nested) scrollbars. scrollLeft cannot be relied on in right-to-left languages because it is sometimes negative and sometimes positive especially with respect to inner (nested) scrollbars. Different browsers handle right-to-left completely differently.

Other Keyboard Interface Development Tips

MCK: this section is a collection of keyboard related old content with a yet undecided fate.

Miscelaneous stuff we might or might not keep

  • Use focus and blur events (or event delegation) to monitor changes to the current focus - focus and blur events can now be used with every element. Don't assume that all focus changes will come via key and mouse events, because assistive technologies such as screen readers can set the focus to any focusable element, and that needs to be handled elegantly by the JavaScript widget. Techniques such as "event delegation" (for example, intercepting events on a list rather than on every listitem) can improve performance.
  • Follow keydowns to move focus - A keydown event handler can determine the next object to receive focus and call that element's focus() method.
  • Use onkeydown to trap key events, not onkeypress - Key press events do not fire for all keys and they vary across browsers.
  • Use element.focus() to set focus - Do not use createEvent(), initEvent() and dispatchEvent() to send focus to an element, because these functions do not change the focus. DOM focus events are informational only, generated by the user agent after an element has acquired focus, but not used themselves to set focus.
  • The use of :focus pseudo-class selectors to style the keyboard focus is not supported in many versions of Internet Explorer. Authors should use the :active pseudo-class (which older versions of IE treat like :focus) in conjunction with the :focus pseudo-class. Example: a:focus, a:active { text-decoration: underline; } If the related CSS pseudo-classes are not appropriate or not supported in all browsers, authors can use JavaScript techniques to indicate an appropriate focus alternative, such as using focus and blur events to toggle a classname on an element.
  • Always draw the focus for tabindex="-1" items and elements that receive focus programmatically when supporting versions of Internet Explorer older than 8 - Choose between changing the background color via something like = "gray"; or add a dotted border via = "1px dotted invert". In the dotted border case, you will need to make sure those elements have an invisible 1px border to start with, so that the element doesn't grow when the border style is applied (borders take up space, and IE doesn't implement CSS outlines).
  • Prevent used key events from performing browser functions - If a key such as an arrow key is used, prevent the browser from using the key to do something (such as scrolling) by using code like the following:

    <span tabindex="-1" onkeypress="return  handleKeyPress();"> 

    If handleKeyDown() returns false, the event will be consumed, preventing the browser from performing any action based on the keystroke. In addition to the return value the browser must call the event methods that will prevent the default action, for IE this is setting the event property "returnValue=false", and for other browsers supporting the W3C event model this means calling the "preventDefault" method of the event object.

  • Use key event handlers to enable activation of the element - For every mouse event handler, a keyboard event handler is required. For example, if you have an onclick="doSomething()" you may also need onkeydown="return event.keyCode != 13 || doSomething();" in order to allow the Enter key to activate that element. There are user agent-specific considerations for key event handling.

Supporting Tooltips with the Keyboard

A tooltip is a popup messages typically triggered by moving a mouse over a control or widget causing a small popup window to appear with additional information about the control. To provide simple text tooltips, the HTML title attribute should more than suffice because the user agent will render it for tooltips. When creating a tooltip, it is essential that the user be able to activate it using the keyboard. When a form control or widget receives keyboard focus, the tooltip must display. When the form control or widget loses focus, the tooltip must disappear. Browsers do not currently support this functionality.

The following code snippet from the iCITA site shows the use of a onfocus="tooltipShow();" function to display the tooltip when focus is placed on an element.

<html lang="en-us"">
   <title>inline: Tooltip Example 1</title>
   <link rel="stylesheet" href="css/tooltip1_inline.css"  type="text/css">
   <script type="text/javascript" src="js/tooltip1_inline.js"></script>
   <script type="text/javascript" src="../js/widgets_inline.js"></script>
   <script type="text/javascript" src="../js/globals.js"></script>
   <link rel="icon" href="" type="image/x-icon">
   <link rel="shortcut icon" href="" type="image/x-icon">

   <body onload="initApp()">

   <div id="container">

   <h1>Tooltip Example 1</h1>
     <h2>Create Account</h2>
   <div class="text">
   <label for="first">First Name:</label>

   <input type="text"  id="first" name="first" size="20"
          onmouseover="tooltipShow(event, this, 'tp1');"
          onfocus="tooltipShow(event, this,  'tp1');"

   <div id="tp1" role="tooltip" aria-hidden="true">Your first name is optional. </div>
Maintain a valid format for aria-valuenow

It is essential that application vendors maintain a valid format for aria-valuenow, and aria-valuenow should accurately represent the value of the widget.The value must be within range of aria-valuemax and aria-valuemin, where the value of aria-valuemin is less than or equal to the value of aria-valuemax, throughout the life cycle of your widget. If aria-valuemin is not less than or equal to the value of aria-valuemax, or if the aria-valuemax is indeterminate, this creates an error condition that will be handled by the assistive technology, resulting in undesirable results. Should an alternative text equivalent be needed to properly represent the aria-valuenow, such as a day of the week, then you should accompany the aria-valuenow with an appropriate aria-valuetext equivalent such as in this example:

<div role="slider"

Here the values 1 through 7 represent the days of the week and aria-valuenow of 1 is equivalent to the first day of the week or Sunday. When aria-valuetext is made available aria-valuenow should also be treated as an index which is 1 based.


This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

Labeling and Describing

Marked up content or widgets will often need additional context to make clear what the meaning or purpose is. It is also reasonable that some content media types will need additional descriptions in another format to give clarity to those who are unable to consume the origin format. These additional meta-content sections are linked together by tagging them as labels or descriptions. WAI-ARIA provides the aria-labelledby and aria-describedby attributes to signal the browser to feed these relationships into the accessibility layer. This layer is then used by screen readers and other accessibility technology (AT) to gain awareness of how disparate regions are actually contextually connected to each other. With this awareness the AT can then present a meaningful navigation model for discovery and presentation of these additional content sections. The user agent itself can also choose to present these insights in a meaningful way. Generally you should always add these attributes to any widgets on your site as they are often merely a construct of HTML and JavaScript which provides no obvious insight as to what the widget's behavior or interactivity is.


When using one element to label another use the aria-labelledby by attribute. A label should provide the user with the essence of what the object does. For example, you could have a dialog box erected from HTML <div> and you need to associate a label for the dialog. With a WAI-ARIA role of dialog, you can indicate its widget type and define a label using an HTML header and then associate that label with the dialog using the aria-labelledby relationship.
  <div role="dialog" aria-labelledby="dialogheader">
  <h2 id="dialogheader">Choose a File</h2>
  … Dialog contents

The section doing the labeling might be referenced by multiple widgets or objects as these need only reference the same id, so contextual description may not always be viable. The labelledby attribute can have multiple ids specified as a space separated list much like applying multiple classes to a single DOM object.

The aria-labelledby property can be used to label all visual objects, however it should be noted that (X)HTML provides a @for attribute on the label element which is used to label form controls. Use this attribute where possible and valid. Because the aria-labelledby attribute accepts multiple IDREFs, it is recommended that authors use aria-labelledby for labeling elements that require more than one label string.

Some elements receive their label for embedded text content, but that is the exception.

Often user agents will operate with images turned off for network performance reasons. In these situations, alt text is provided in the place of the image. When the host language provides alternative text capability it is recommended that authors use the alternative text to support these situations and not use separate labeling as a replacement.


Using aria-describedby

When one element describes another, use the aria-describedby attribute. An aria-describedby section provides further information about a given object or widget, which may not be intuitively obvious from the context, content or other attributes. For example, a fake window is a common widget used in Web applications and is often constructed using a div absolute positioned in a layer above other content. To simulate common window behavior look and feel there is often an X box in the corner which, when activated, dismisses the window widget. One common way to make this X box is to simply make a link whose content is an X. This doesn't give a non-visual user much to go on and belies the real purpose of the X link. To help we add more descriptive text stored elsewhere in the page like this:

<button aria-label="Close" aria-describedby="descriptionClose"
and then elsewhere in the HTML
<div id="descriptionClose">Closing this window will discard any information entered and
return you back to the main page</div>
Like labelledby, describedby can accept multiple ids to point to other regions of the page using a space separated list. It is also limited to ids for defining these sets. In our contrived example we would also want a good labelledby section to fully explain what the window does and how closing will effect the task being worked on. If an object or widget lacks describedby the user agent or AT may try to extract information from the label or th tags, if present. The label and th tags have limited use in that they can only be applied to forms or tables, respectively.

Using a tooltip as a description

WAI-ARIA also defines the tooltip role to which aria-describedby may reference an author defined tooltip. The assistive technology can tell from the type of object describing the document element what the purpose of that element is. For example, a screen reader could announce the tooltip without the user having to wave the mouse over the element by following the describedby relationship to a document area with a tooltip role. The aria-describedby property is also useful for describing groups.

Here is a code snippet showing a set of the tooltip:

  <div class="text">
    <label for="first">First Name:</label>
      <input type="text"
  <div id="tp1" class="tooltip" role="tooltip">Your first name is optional</div>

Descriptions on external pages

The aria-describedby property is not designed to reference descriptions on an external resource—since it is an IDREF, it must reference an element in the same DOM document. This is different from the HTML longdesc attribute, which accepts any URI. In general, the preferred pattern for WAI-ARIA applications is to include all relevant resources in the same DOM, But if you wish to reference an external resource with aria-describedby, you can reference a link that in turn references the external resource. This requires the user to follow two steps, first following the aria-describedby arc, then following the link, but does address the use case. The following example demonstrates this:

       alt="Histogram of Blackberry tree heights"

    <a id="longdesc1" href="blackberry-description.html" target="_description">Data for Blackberry Histogram</a>

It is not good practice to use the above pattern when the describing element—the <a> tag with @id='longdesc1'—is hidden, since there is no way for a user to navigate to and activate the link. Use the technique only when the description is not hidden.

Owning and Controlling

Two relationships expand the logical structure of a WAI-ARIA Web application. These are aria-owns and aria-controls .  The aria-owns relationship completes the parent/child relationship when it cannot be completely determined from the DOM created from the parsing of the markup. The aria-controls relationship defines a cause-and-effect relationship so that assistive technologies may navigate to content effected by and changes to the content where the user is operating.

The Owns Relationship

In addition to WAI-ARIA role and state information, for a document element, an assistive technology needs to convey its context. In the case of a treeitem, it is important to know the parent (container), which may provide a folder depth and the number of siblings in the folder. This containment hierarchy can often be determined by the DOM tree, as it provides parent, siblings, and children for a given document element. That said, the DOM hierarchy is rigid and non cyclical in that each node may only have one parent. In some situations, a child is reused by multiple parents or a child is separated from its siblings or parent in the DOM tree. One example is when a radio button appears in a table but it is not a DOM child of the radiogroup, due to the authors use of a table for formatting and the placement of the radio buttons placing them outside the radiogroup container. To solve this problem WAI-ARIA provides the aria-owns property.

The aria-owns property is set on a document element, and its values are the unique IDs of all the adopted children. These elements may appear anywhere in the DOM, yet they are treated as siblings of each owning parent. This example illustrates a radiogroup that first uses the DOM hierarchy to convey context and then aria-owns:

First, consider the preferred method for using the W3C DOM to describe the relationship between radiogroup and radio roles for HTML:

<div id="radio_label">My radio label</div>
<ul role="radiogroup" aria-labelledby="radio_label">
<li role="radio">Item #1</li>
<li role="radio">Item #2</li>
<li role="radio">Item #3</li>

In this example, the elements with role="radio" are child nodes of the parent role="radiogroup" element node.

Now, consider an alternative method using the aria-owns property to describe the parent-child role relationship between radiogroup and radio roles for HTML:

<td role="radiogroup" aria-owns="myradio1 myradio2 myradio3" rowspan="3" >
My Radio Label
<td id="myradio1" role="radio">Item #1</td>
<td id="myradio2" role="radio">Item #2</td>
<td id="myradio3" role="radio">Item #3</td>

The aria-owns property should be used sparingly, since it increases the computational load on assistive technology to provide alternative renderings. Also, when accessing the DOM and enumerating children of a DOM node, an AT should then enumerate the adopted children using the aria-owns property. At that instance of walking the DOM, the parent of the adopted children is the adopted parent and not the DOM parent. This will avoid recursion problems.

Each child, adopted or natural, should have the appropriate aria-posinset and aria-setsize properties set consistent with their rendering, if these cannot be determined from the DOM from a direct parsing of the host language. Places where direct parsing does not allow the user agent to determine aria-posinset and aria-setsize are long lists where only the currently visible items are loaded (via Ajax). If the children are re-sorted then the aria-posinset and aria-setsize values should be updated consistent with their visual rendering. Rather than setting size only on a container, content authors should specify aria-setsize on each member of a set to avoid performance issues.  If this property is not set, the user agent must compute and set the property on supporting roles.

Platform accessibility API mappings must invert this relationship for assistive technologies, so that they may determine the owning parent from a child and couple it with aria-posinset and aria-setsize information to provide context information to the user.

 Using Owns with Reusable Content

If you are re-using content across different, transient sections of content by restyling it to render it in association with a different object, you need to maintain the aria-owns property as well to match the current use and apparent ancestry of the reused sub-section. A good example of this is a context help menu that resides at the end of the document. When the user wishes to launch the context help menu in association with different visual elements, styling is used to render the menu in context with that object. Prior to rendering the visual submenu you should ensure the object to which you have requested help assigns its aria-owns property value to the submenu ID. When the menu closes you can remove the aria-owns property.

The Controls Relationship

In rich internet applications document elements may control the behavior on another part of Web page outside themselves. The aria-controls attribute indicates cause-and-effect relationships between document elements.

An example might be a group of radio buttons that "control" contents of a listbox in another part of the page. Here, you would want the radio group to assign a aria-controls relationship to the listbox which will be updating without a page reload. The user, through their assistive technology, can then navigate to the listbox by following the aria-controls relationship when a different radio is selected, to see how the contents changed in the listbox. The ability to update parts of a page without a page reload is a common practice of applications making use of Ajax. Without the aria-controls attribute, a user would be unable to effectively use these types of Web pages as assistive technologies often will not make the user aware of what is happening outside the context of the element the user is currently operating.

With the aria-controls attribute the user may use the assistive technology to follow the relationship to the object it is controlling. It is extremely important for an assistive technology to present changes in the document in response to user input. Therefore, an assistive technology immediately presents changes to a live region when the controlling widget is the one which has user keyboard focus. For example, if a tree view controls a help document pane, each time
the tree item changes the new tree item and then the new help contents should also be read. In the case of a screen reader, the amount of information read in the target live region is dependent on how the live region is configured.

The aria-controls attribute takes one or more ids which refer to the document element. If this relationship is not implied by the host language semantics, then the controlling element must be given a controls attribute with the IDs of the other elements where the changes will show up listed in the attribute value.

Changing the Reading Flow

(X)HTML suffers from a number of disadvantages in keyboard navigation today. One such example is the restriction of navigation to the tabbing order. This is a common problem with presentations in office suites where the logical, perceivable, navigation order of a slide may not match the tabbing order. Sighted users may see a logical navigation process (such as visual steps in the process for assembling a lawn mower). This "flow" is not conveyed by the tab order. The user might tab from step 1 and land on step 4. Another problem is the construction of model-based authoring tools on a Web page. In a model-based authoring tool, a visual object may provide a number of paths that the user can take, such as a multiplexor, which may have output that logically flows to a number of optional electronic components in a drawing. In Web 2.0, developers are actually creating drawings like this to perform tasks such as visually model a work flow. In this scenario, the user will want to decide which object they will navigate their screen reader or alternate input device to next.

Although it is recommended that Tab order follow the reading flow, there may be instances where this is not possible. For these reasons, WAI-ARIA provides a relationship property, called aria-flowto. This allows the author to provide an assistive technology with alternative navigation flows through the document that best represents the author's intent and which is more logical for people with disabilities. aria-flowto establishes the recommended reading order of content, so that the an assistive may overriding the default of reading in document order to its user. aria-flowto does not change the keyboard navigation order of the browser.

Consider the first case of changing a basic reading flow to circumvent (X)HTML tabbing. A good example of this is a logical reading flow in a portal with landmarks. In the future, the user may wish to change the reading flow based on the order of priority with which they navigate a personalized Web application. In the following example, the navigation would follow the order of "Top News Stories", "television listings", "stock quotes", and "messages from friends" by following (X)HTML document reading order. However, the author or end user may determine that the main content is most important, followed by "stock quotes", "messages from friends", and then "TV listings." The end user can change the order if the web page or assistive technology provides an interface for such personalization.

<div role="main" title="Top News Stories" id="main" aria-flowto="stock"></div>
<div role="complementary" title="television listings" id="tv"></div>
<div role="complementary" title="stock quotes" id="stock" aria-flowto="messages"></div>
<div role="complementary" title="messages from friends" id="messages" aria-flowto="tv"></div>  

The second use case is such that the Web developer may wish to circumvent the flow by branching to multiple paths in the Web page, requiring the assistive technology to present a collection of options where the user could go next. This is important for work flows or business process management applications. More of these applications are becoming Web based, and a vehicle is required to tell the user how to get through the work flow. The flowto property takes multiple idrefs, allowing the author to define each object the user could flow to. To implement this technique do the following.

Popups and drop-downs

In order for menus, menubars, and menuitems to indicate that it opens a menu, set its aria-haspopup property to "true." The type of menu being launched (submenu, context help, etc.) is not explicitly indicated with WAI-ARIA but is based on the operational characteristics (keyboard and mouse commands).

Combo boxes, or drop down lists, work differently. Controls with the role combobox must contain a list of items that can be opened, usually as a drop-down. Because this is intrinsic to the functionality of a combo box, it does not need to be explicitly indicated with aria-haspopup.

The following html fragment shows the use of aria-haspopup with a menubar, its menus, and submenus. All of the menuitems associated with the menubar have aria-haspopup set to 'true'. Also, the "Zoom" menuitem of the "View" menu has an aria-haspopup property as it leads to a submenu.

<div role="menubar">
    File menu: File menuitem has an aria-haspopup attribute set to 'true'.
    That popup div follows immediately below.
  <div role="menuitem" aria-haspopup="true" id="fileMenu">File</div>
  <div role="menu" aria-labelledby="fileMenu">
    <div role="menuitem">Open</div>
    <div role="menuitem">Save</div>
    <div role="menuitem">Save as …</div>
    View menu:
  <div role="menuitem" aria-haspopup="true" id="viewMenu">View</div>
  <div role="menu" aria-labelledby="viewMenu">
    <div role="menuitem">Normal</div>
    <div role="menuitem">Outline</div>
      The View's Zoom menuitem has aria-haspopup='true' as it leads to a
    <div role="menuitem" aria-haspopup="true" id="zoomSubMenu">Zoom</div>
    <div role="menu" aria-labelledby="zoomSubMenu">
      <div role="menuitem">50%</div>
      <div role="menuitem">75%</div>
      <div role="menuitem">100%</div>
      <div role="menuitem">150%</div>
      <div role="menuitem">200%</div>

Managing Dynamic Changes

This section has had only minor edits since it was integrated from APG version 1.0 -- a complete APG task force review is pending.

Managing Content and Presentational Changes

General rules for managing content and displaying information

If you are refreshing areas asynchronously, then you need to designate live regions. The following sections explain how to implement live regions and when to use roles that are considered "live" sections by default, including alert, status, or log.

Implementing Live Regions

Live regions are parts of a Web page that the author expects to change. Examples of live regions include dynamically updated content (sports stats, stock information), logs where new information is being added (chat transcript logs), notification areas (status, alerts), etc.

Live regions enable assistive technologies, such as screen readers, to be informed of updates without losing the users' place in the content. Live region settings provide hints to assistive technologies about how to process updates. Note that the assistive technology is responsible for handling these updates and enabling users to override these hints.

The section on Live Region Properties and how to use them gives the details of how to apply live region properties. This process will help rich internet application (RIA) developers to set live region settings that will provide a good user experience for most assistive technology users with little configuration on their part. When applying these live region properties, it is recommended that you conduct user testing. If the AT supports scripting of the response to live regions you may wish to customize the response, such as through a screen reader script for your Web page.

  1. Identify the live regions

    Live regions are any region on a page that receives dynamic updates through client-side scripting. Note the regions of your page that will be live.

  2. See if any of the special case live regions meet your needs

    WAI-ARIA provides a number of special case live region roles whose live region properties are pre-defined and which you may use. If one of these live region roles meet your needs just apply the specific role to the region of the Web page.

  3. Decide the priority of each live region

    When a live region changes, should the user be notified of the change? Notifications could include a sound for a screen reader user. (For simplicity, we will use the case of an audio notification in this discussion.) The shorter the interval between changes and the less important the information, the less likely that the user needs to hear every change. A simple example of changes that should not be heard are changes to time; a sound for every second would be very annoying.

    If the user should hear the change, should the change be announced immediately, as soon as possible, or only when the user is idle? Announcing a change immediately can be disorienting for users, so that should be done sparingly. Most updates should probably only be announced when the user is idle.

    After you have decided the priority for each live region, then decide the live property value:

  4. Decide how much context is needed for each update

    When part of a live region changes, how much context is needed to understand the change. Does the user need to hear the entire live region or just the change by itself?

    If the user needs to hear the entire live region, then mark the entire live region with aria-atomic="true".

  5. Decide what types of changes are relevant for each live region

    Three possible types of changes are: additions, removals, and text. Additions means new nodes are added to the DOM; removals means nodes are removed from the DOM; and text means changes are solely to the textual content. Should the user hear all types of changes or only certain types?

    By default, the user will hear additions and text type changes. If you wish to explicitly define the types of changes, you need to set relevant="THE_TYPES_OF_CHANGES". If more than one type of change is relevant, the types are separated by a space. For example, to define additions and removals as relevant but not text, set relevant="additions removals".

Live Region Properties and How to Use Them

One of the most important concepts behind live regions is politeness. Politeness indicates how much priority a live region has. The following politeness values are available for aria-live: off, polite, and assertive.

This is the default. Any updates made to this region must not be announced to the user. live="off" would be a sensible setting for things that update very frequently such as GPS coordinates of a moving vehicle.
Any updates made to this region should only be announced if the user is not currently doing anything. live="polite" should be used in most situations involving live regions that present new information to users, such as updating news headlines.
Any updates made to this region are important enough to be announced to the user as soon as possible, but it is not necessary to immediately interrupt the user. live="assertive" must be used if there is information that a user must know about right away, for example, warning messages in a form that does validation on the fly.

There are times to suppress AT presentation changes while a region is updating. For that you can use the aria-busy property.


To suppress presentation of changes until a region is finished updating or until a number of rapid-fire changes are finished, set aria-busy="true" and then clear the attribute when the region is finished. While it is busy, the AT will track and collate the changes. It will finally speak the changes once the region is no longer busy.

When a live region is updated, the update can often be announced on its own and still make sense. For example, if a news headline is added, it would make sense to simply announce the headline. However, sometimes the entire live region must be read in order to give the user enough context to make sense of the update. For example, it would not make sense to only give the current value of a stock without saying the name of the stock. The atomic setting gives assistive technologies a hint about which of these cases an update falls into.

This is the default. It means that when there is a change in the region, that change can be presented on its own; the AT should not present the entire region. atomic="false" is generally a good idea as it presents users with only changes and does not cause them to hear repetitive information that has not changed. However, Web developers should take care that the changed information, when presented by itself, can still be understood and contextualized by the user.
If atomic is set to "true", it means that the region must be presented as a whole; when there is a change, the AT should present the entire region, not just the change. atomic="true" can make it harder for users to understand changes as the changed areas are not presented independently. atomic="true" can also be annoying as it can force users to listen to repetitive information that has not changed. However, atomic="true" is necessary in cases where the change, when presented by itself, cannot be understood and contextualized by the user. Note that when aria-atomic="true", the AT will attempt to speak the atomic region only once when multiple changes occur in the same region and it hasn't been spoken yet.
Not all updates to a live region are relevant. For example, if the oldest headline in a list of headlines is removed and a new headline is added, the removal of the oldest headline is probably not important enough to announce to the user. However, in a chat application, when a user leaves the chat and their username is removed from the list of users, that removal may be important enough to announce. The relevant setting gives a hint about what types of changes are relevant and should be announced. Any change which is not relevant will be treated as if the region had live="off" and will not be announced. Multiple types of changes can be listed as relevant; the types are separated by a space. The default is relevant="additions text" and this is the most common use case. If the default is applicable to your application, you do not need to provide the relevant property explicitly.
Insertion of nodes to the live region should be considered relevant.
Removal of nodes to the live region should be considered relevant. Often, removals are not relevant because nodes are removed to make space for new information - e.g. a log implemented as a table where items are taken off the top. However, in the case of something like a buddy list, it is relevant if a buddy is removed. It doesn't require the screen reader to speak the removal, but it notifies the screen reader that it could be useful to do so. Use of aria-relevant="removals" or aria-relevant="all" should be used sparingly. Notification of an assistive technology when content is removed may cause an unwarranted number of changes to be notified to the user.
Changes to the textual content of nodes that already exist in the live region should be considered relevant. Textual content includes text equivalents, such as the alt attribute of images.

This example shows two live regions. If both regions update simultaneously, liveRegionA should be spoken first because its message has a higher priority than liveRegionB.

<div id="liveRegionA" aria-live="assertive">
<div id="liveRegionB" aria-live="polite>

Choosing Between Special Case Live Regions

You may wish to use a special live region role instead of applying live region properties. WAI-ARIA contains a number of standard roles which are by default considered "live" sections of your Web page. It is important to know when to use these and when to create a custom live region on your known. Here are some rules of thumb:

alert - You may use the alert role for a one-time notification which shows for a period of time and then goes away. It is intended to alert the user that something has happened. The assistive technology should be notified by the user agent that an alert has occurred, if your operating system supports this type of event notification. When choosing to use alert you should use the alertdialog role instead if something inside the alert is to receive focus. Both alert and alertdialog usually appear to pop-up to the user to get their attention. Unless specified otherwise an alert region has an implicit aria-live of "assertive" and aria-atomic of "true".

status - You may use the status role when you want to mark an area which is updated periodically and provides a general status of your Web application. Many applications provide status widgets and they are often found, visibly, at the bottom of the application and contain a variety of widgets within it to convey status. Unless specified otherwise a status region has an implicit aria-live of "polite" and aria-atomic of "true".

timer - You may use a timer role when you want to mark an area which indicates an amount of elapsed time from a start point, or the time remaining until an end point. The text encapsulated within the timer indicates the current time measurement, and is updated as that amount changes. However, the timer value is not necessarily machine parsable. The text contents MUST be updated at fixed intervals, except when the timer is paused or reaches an end-point. The timer role has an impicit aria-live of "off".

marquee- You may use a marquee role when you need to mark an area with scrolling text such as a stock ticker. The latest text of the scrolled area must be available in the DOM. A marquee behaves like a live region, with an implicit aria-live property value of "off."

log - You may use log if you have a live area where new information is added, like a scrolling chat log of text. Unlike other regions, implied semantics indicate the arrival of new items and the reading order. The log contains a meaningful sequence and new information is added only to the end of the log, not at arbitrary points. Log has an implicit aria-live of "polite"If you have a chat text entry area you should indicate that it also controls the aria log aria like so:

<div contenteditable="true" role="log" id="chatlog">

<label id="gettext">Send Text</label>
<div aria-controls="chatlog"

live region - If you have some other live area use case, WAI-ARIA allows you to mark an area using the aria-live attribute. This, accompanied by the collection of attributes which support the live property, allows you to create your own custom live area on your Web page. For more details regarding live regions refer to the live region section of this guide.

Live region roles that are applied to elements having strong native semantics are not mapped consistently to the platform accessibility API. An example is the TABLE element. It is recommended that authors apply live regions to DIV and SPAN containers. For example:

<!-- Live region 'log' role used with TABLE element:  the 'log' role is not consistently mapped to platform AAPI. -->
<!-- Do not use. -->
<table role="log" … >

<!-- Wrap the TABLE element in a DIV with role 'log' instead: -->
<div role="log" … >
  <table … >

Presentation Role

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

This section describes the presentation role, including the conditions under which it is inherited by an element's children.


Authors devote a good deal of effort to the appearance of their web pages, and this is especially true in the case of scripted web applications. To this end authors employ various elements purely for visual presentation. For example, <img> elements are used for spacing and decoration; and <table>s are used to create a column based layout. Elements used strictly for presentation are semantically neutral and irrelevant in terms of accessibility. It is necessary to mark such elements as presentational so that they do not appear in the accessible tree created by the user agent. For example, a current technique used with spacer images is to provide blank alt text to indicate that the image is not meaningful. User agents will not publish any information about images with blank alt text to the platform accessibility API.

There are elements other than <img> and <table> that are used solely for visual presentation. Any element is a potential candidate for presentation, and, if so used, these elements also need to be marked as semantically neutral.

It is important to separate content and presentation. CSS 3 has introduced new table layout functionality to allow a user agent to layout content using CSS. There are many advantages to using this feature of CSS such as browser style sheet caching and improved adaptability to mobile devices with limited screen real estate. There are, however, cases where presentational markup cannot be avoided. In such instances, authors are counseled to consult WCAG 2.0 for more detailed guidance.

WAI-ARIA introduces the presentation role as a general device for marking elements as presentational. The presentation role overrides the element's native implicit role, and informs the user agent to not publish the element to the accessibility API. In fact, the preferred way to mark <img> elements as decorative is to use a role="presentation" attribute instead of (or in addition to) an empty alt attribute. Here is a code fragment sample:

<!-- [role="presentation"] informs the user agent that the spacer images are for layout only. -->
<h2>Other Histories of Architecture</h1>
  <a href="">Ancient Roman Architecture</a>
  <img src="spacer.png" role="presentation">
  <a href="">19th Century Italian Architecture</a>
  <img src="spacer.png" role="presentation">
  <a href="">Modern Buildings more than 100 Years Old</a>

<h2>Modern Building Design</h1>

The resulting accessible tree is shown below. Note that none of the spacer <img> elements are present:

Inheritance of Presentation by Parent Element's Children

In general, the presentation role is not inherited by an element's children. The exceptions are elements whose role entails that the element has required owned children. Examples include the <table> element and list role, and these exceptions are discussed further below. For all other elements, only the element marked presentational is eliminated from the accessible tree. Note, however, that its contents are published; in particular, the element's textual content is published, but any attributes of the element that may form a text-equivalent are not. For example, a header element with a presentation role would not appear in the accessible tree, but its text does. An image with a role of presentation is not exposed in the accessible tree, nor is the contents of its alt attribute. This is shown in the following code fragment:

<!-- 1. [role="presentation"] negates the implicit 'heading' role semantics but does not affect the contents. -->

<h1 role="presentation">Presentation role overrides Header role</h1>
<h1>Another header</h1>

<!-- 2. [role="presentation"] hides the image from the accessibility API and does not publish the alt attribute contents. -->

<img role="presentation" alt="This text will not appear in the Accessibility API" src="…">

The first header element is absent from the accessible tree for the above example, but its text appears. The second header element is present as a heading. The img element is not exposed at all:

Be aware that the markup <img role="presentation" alt="non-empty alt text"…> is in fact contradictory: Declaring a role of presentation says that the image is for layout, is decorative and meaningless, whereas the non-empty alt text implies that the image is meaningful. It is recommended that authors instead use empty alt text (alt="") where they use role="presentation".

Earlier it was noted that the presentation role is not generally inherited by an element's children. The exception are roles that have required owned elements. Examples include the <table> element, and the list and tree roles. A list is required to have listitem children; a tree, treeitem children. The <table> element's child components are <tbody>, <thead>, <tfoot>, <th>, <tr>, <td>, and <caption>. These component elements would not appear in the markup without the enclosing <table> root element. Thus, when a table is used for layout, it follows that all of its component elements are present in the markup for layout as well. Since annotating all the required child elements with role="presentation" is error prone and difficult to maintain, it is sufficient to mark the table root element as presentational. The presentation role propagates to all the table component elements. However, as before, the contents of the table elements do appear in the accessible tree. Here is an example:

<!-- Layout table marked with [role="presentation"]. -->

<table role="presentation">
    <tr> <td>Some text</td> <td><p>Paragraph</p></td> </tr>
    <tr> <td><a href="">Link text</a></td> <td><img src="meaningful.jpg" alt="meaningful image"></td> </tr>

All table specific elements (table, tr, td, etc.) are eliminated from the tree by inheriting the presentation role, but other elements and textual content in the table cells are exposed:

The same logic applies to other roles that have required owned children, such as a list. If the list's root element is declared presentational using role="presentation", all of its listitem elements inherit the presentation role, and none of the list item elements appear in the accessible tree. However, the contents of each list item, that is, other elements and textual content, are included in the tree.

Overriding Presentation

The presentation role is overridden in some circumstances. Recall that the presentation role informs user agents that certain elements are semantically neutral, and are irrelevant for accessibility. If, however, there is an aspect of an element that makes it meaningful, it is no longer neutral, and cannot simultaneously be presentational. The Core Accessibility API Mappings describes the cases where this occurs:

These situations entail that the given element is semantically relevant and will appear in the accessible tree. Marking the element with a role="presentation" is an error, and user agents will ignore the presentation role in these cases. Authors are advised to not mark an element with a role of presentation when it has any of the above properties; rather, to use a role that reflects the element's purpose.

Form Properties

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

This section identifies authoring practices for elements used as form elements.

Use aria-invalid and aria-required To Improve Access to Forms

Until the introduction of WAI-ARIA's aria-invalid state and aria-required property, only presentational strategies have been available to Web content authors to indicate that certain form fields and controls are required or invalid. In applications, these states are only available through styling or varying markup, which is inconsistent, and therefore is inconclusive. In Web-based forms, fields required may be marked by an asterisk. Forms submitted with required data missing or improperly specified may be redisplayed with the unacceptable elements highlighted in red. The assistive technology user is poorly served by such imprecise methods, and all users confront inconsistent indicators for consistent situations.

The WAI-ARIA invalid state and required property provide:

Alert the User When Maximum Length Value Is Reached

When a text input field that has a maximum length value (or the host markup language's equivalent) receives focus, the value defined for "maximum length" should be communicated to the user. When text entry reaches that maximum length (or the markup language's equivalent), an alert (expressed in accordance with user preferences and capabilities) should inform the user that the maximum length for a given field has been reached. Such an alert can be expressed programmatically or, using as an aural icon, by using a WAI-ARIA alert; the user agent may alert the user through a system beep and by triggering the operating systems' "show sounds" facility. When maximum length (or the markup language's equivalent) is reached, the user must then be able to move to another form field in a manner consistent with tab-navigation for that document.

Automatic Focus Changes

Having a user agent automatically change focus in a form in response to user input can be advantageous in situations where that change saves the user keystrokes or on-screen keyboard interaction in order to manually move the focus from one field to another. However, as with form auto-completion, this type of text input event must be firmly under user control because this may not have been the user's intention and some users with disabilities may become disoriented such as those with sight impairments. Consider these cases:

Form Auto-submit

Use caution when using automatic submission of a form without explicit user command or in response to a user-defined setting that permits such behavior, as expressed by the Priority 1 UAAG 1.0 Checkpoints 7.1, 7.2 and 11.1. Unless the user has specifically chosen to set the user agent to allow auto-submission, authors are advised not to set onChange or onFocus events either to trigger submission of a form or to provide an auto-submission event upon completion of all or all required fields of a form or input field.


This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

Editors' note: This section was added as part of disposition of comment 4, but is very incomplete. The section needs to be reordered, so that instructions on how to use the math role properly appear before considerations of legacy content and negative examples (such as the use of generic HTML to approximate a visual representation of a mathematical expression). It needs more introductory text about how to use math. The examples need better introductions, and the positive examples should preceded the negative examples, which need to be explained more fully.

There exists significant amounts of legacy content that uses images and/or textual approximations to represent mathematical expressions. Examples of this include the use of ASCII art and/or the misuse of HTML tags -- in particular, SUB and SUP -- to achieve a visual approximation of a mathematical expression, one which is void of the structure needed to render mathematical expressions inherently meaningful.

Use of generic HTML to approximate a visual rendering of a mathematical expression:

<i>a</i><i>x</i><sup>2</sup> + <i>b</i><i>x</i> + <i>c</i> = 0

Accessible example of the same function, using the math role, appropriate label, and MathML rendering:

<div role="math" aria-label="a times x squared plus b times x plus c equals 0">
  <math xmlns=''>
            <mo>&#x2062; <!-- invisible times --></mo>
            <mo>&#x2062; <!-- invisible times --></mo>


<i>f</i>(<i>x</i>) = <i>x</i><sup>2</sup> + <i>x</i> - 2

Accessible example of the same function, using the math role, appropriate label, and MathML rendering:

Todo: add aria-label here also

<div role="math">
  <math xmlns=''>

Drag-and-Drop Support

aria-grabbed and aria-dropeffect have been deprecated in ARIA 1.1 - as such this section has been removed. Advice for implementing drag and drop will be added to a future version of the Authoring Practices Guide.

Reusable Component Libraries

This section has not been updated since it was integrated from APG version 1.0 -- an APG task force review is pending.

Rich internet applications are complex to author. To save time, it is often faster to use existing widget libraries that implement WAI-ARIA and that have already gone through:

Some publicly available UI component libraries have already implemented WAI-ARIA. Authors can reuse such libraries to start developing accessible rich internet applications.


Background on WAI-ARIA

This section has not been updated since it was integrated from the ARIA 1.0 Primer -- an APG task force review is pending.

According to the SecuritySpace Technology Penetration Report, more than 55% of all Web sites today contain JavaScript, dramatically affecting the ability for persons with disabilities to access Web content. New Rich Internet Applications (RIAs) render custom widgets, modeling rich desktop components to perform UI updates without having to reload the entire page - much like a graphical user interface (GUI). Legacy GUI accessibility frameworks address these issues through a comprehensive accessibility application programming interface (API) and infrastructure to foster interoperability with assistive technologies. These APIs constitute a contract between applications and assistive technologies, such as screen readers, to enable them to access rich dynamic content with the appropriate semantics needed to produce a usable alternative. No such contract exists between modern RIAs and assistive technologies, creating an accessibility gap for persons with disabilities.

Unfortunately, HTML and other markup languages do not provide adequate support for accessible dynamic content. Until now, the W3C WAI has discouraged the use of JavaScript per [[WAI-WEBCONTENT]], Checkpoint 6.1). A number of W3C initiatives underway address this problem using a declarative markup approach. This primer describes a means to bridge the interoperability problem with assistive technologies now by incorporating the appropriate metadata into current HTML and XHTML markup to support today's accessibility APIs. Moreover, the primer provides web developers with a conceptual understanding of WAI-ARIA as a prelude to using the [[WAI-ARIA]]. WAI-ARIA brings advanced accessibility features of the desktop to the web through the creation of cross-cutting technologies that (X)HTML authors can incorporate in their markup. WAI-ARIA defines roles, states, and properties, where those roles reflect standard GUI elements that are recognized by accessibility Application Programming Interfaces (APIs). These metadata that describes these GUI widgets and document structures aides assistive technology vendors in providing accessible, usable solutions. The W3C WAI PF working group is working with User Agent manufacturers, assistive technology vendors, and accessibility tool providers, to ensure an end-to-end working solution.

For an introduction to WAI-ARIA, see the Accessible Rich Internet Applications Suite (WAI-ARIA) Overview. The WAI-ARIA Primer is part of a set of resources that support the WAI-ARIA specification. The Primer provides a basic introduction to the concepts behind and reason for WAI-ARIA, and the WAI-ARIA Authoring Practices describe recommended usage patterns for Web content developers. The WAI-ARIA Suite fills gaps identified by the [[WAI-ARIA-ROADMAP]]. These documents serve as important places of clarification where topics appear to be unclear.

The Problem

Aspects of traditional Hypertext Markup Language (HTML) make accessible support of dynamic content difficult:

  1. Accessibility of dynamic content relies on abstracting semantics from both content and presentational information. Extracting semantic cues from current HTML content is typically unreliable as the cues are limited to tag elements names.
  2. While HTML allows content to be repurposed for presentational formatting, it lacks the ability to attach meaningful metadata about document structure and to convey semantic information. A common example of this is content formatted with tables rather than style sheets.
  3. When combined with script and cascading style sheets (CSS), HTML can be repurposed to create dynamic custom components without providing a means to convey semantic information to native accessibility architectures designed to support dynamic graphical user interface (GUI) content.
  4. Custom components built from common HTML elements often are not keyboard accessible.

Authors of JavaScript-generated content do not want to limit themselves to using standard tag elements that define the actual user interface element such as tables, ordered lists, etc. Rather, they make extensive use of elements such as DIV tags in which they dynamically apply a user interface (UI) through the use of style sheets and dynamic content changes. HTML DIV tags provide no semantic information. For example, authors may define a DIV as the start of a pop-up menu or even an ordered list. However, no HTML mechanism exists to:

  • Identify the role of the DIV as a pop-up menu
  • Alert assistive technology when these elements have focus
  • Convey accessibility property information, such as whether the pop-up menu is collapsed or expanded
  • Define what actions can be formed on the element other than through a device-dependent means through the event handler type (onmouseover, onclick, etc.)

In short, JavaScript needs an accessibility architecture to write to such that a solution can be mapped to the accessibility frameworks on the native platform by the user agent.

The following diagram illustrates a typical document object model (DOM) node in a Model-View-Controller architecture. Accessibility information surfaced to assistive technologies is provided only by the HTML element's tag name, with only the accessibility attributes that tag can provide.

shows that on the node, data, or the "Model", which should include semantic information, is separated from the user interface presentation, the "View." Here, the document element is managed by the user agent based on the default behavior of the element. The user agent's default behavior at the document element forms the controller.

Accessibility information mapped to a DOM element in the Document Object Model
Accessibility Interoperability at a DOM Node without JavaScript

The box between the DOM node and the assistive technology contains the contract provided by the user agent to the assistive technology. This data includes typical accessibility information found in the accessibility API for many accessible platforms for GUIs (role, state, caret, selection, text, hypertext, name description, parent/child information, parent/child information, and relations). For HTML and other W3C markup, the accessibility information provided solely depends upon what the element's tag name and any accessibility attributes that map to that tag provides. For example, the accessible role of a table is table. The author provides an accessible description by assigning a title attribute.

In contrast, with JavaScript, user actions can trigger updates in the data and presentation, but the default accessibility information available in the element tags is no longer valid.

DOM Element with JavaScript controller
Accessibility Interoperability at a DOM Node with JavaScript

shows the same DOM node provided in Figure 1.0 but with JavaScript acting as the new controller. JavaScript overrides the default user agent behavior at the DOM node. It manipulates the data, content, and style in response to events caused by user interaction to produce custom widgets. In this situation, the default accessibility information is no longer valid and, therefore, the contract is now invalid. Figure 2.0 shows the contract with asterisks in front of role, state, actions, value, event changes, and relations. These asterisks represent potential accessibility errors and gaps in the base markup. These gaps result from the author's inability to provide the new semantic data needed to support the contract.


Any solution to facilitate the creation of accessible rich Internet applications (RIAs) must:

Allow for discovery of custom UI components through the use of Semantic Web technologies
Web ontologies allow for storage of semantic information about objects and how they relate to others in the ontology.
Support today's accessibility architectures
Accessibility architecture today is centered around object technology. Each object in an application or document exposes its accessible properties to an assistive technology.
Allow for separation of content and presentation
Dynamic content authors must be able to store the accessible meta data in the document independent of how it is rendered.
Allow for passive monitoring of the application by an assistive technology
Assistive technology vendors should not be required to poll an application for changes in accessibility information.
Leverage new W3C efforts to solve the problem
This problem needs to be solved quickly. A number of efforts are underway, such that minimal changes may be required to bring them to the level needed.
Be light weight
The solution needs to be light-weight in order to promote adoption by Web authors.
The solution needs to be scalable; it must make simple things easy while making complex things possible.
Like other Web solutions, our solutions must be internationalizable.
Guarantee user agent support up front
User agent manufacturers must be involved up front to ensure support for the solution when the specification is complete.
Involve assistive technology vendors up front
To ensure interoperability, assistive technology vendors need to be involved from day one. The effort must leverage support by AT vendors to ensure that a total solution is available when the specification is complete.
Produce Developer Guidelines during the process
This is a critical mistake made by people creating a new accessibility model. Developers must be engaged early on so that they can contribute feedback and start producing workable solutions early.


What is clear from the problem statement is that developers of dynamic web content cannot provide the appropriate accessibility information in the markup to support the accessibility APIs on the target platform. This problem is not limited to HTML. It extends to other markup, including Scaleable Vector Graphics [SVG]. This primer addresses the problem for web content delivered to desktop browsers and introduces you to common extensions to both HTML and XHTML called [[WAI-ARIA]]. The goal is to make these standard features in HTML 5.

HTML Accessibility API Support Analysis

Using as a template for addressing the problem and U.S. Section 508 accessibility standards, Table 1.0 illustrates gaps in the infrastructure and identifies W3C standards that should be used to address the problem. In the right column, table cells that are empty or that indicate a limitation represent accessibility gaps in HTML and XHTML.

Table 1.0 Platform Gap Analysis for Accessible Dynamic Web Content for HTML and XHTML
Required Components Who does what today? (HTML)
FocusChange DOM 2, 3 events
Activation User Agent API
Caret Change User Agent API
Value Change  
State Change  
Selection Change User Agent API
Mutation DOM Events
Accessible Actions:  
Event Handler functional information to label the actions  
Access to the available event handlers for enumerating the actions  
State Information:  
Role Information: Limited to standard HTML tag names. (Mix Content/presentation)
Relationships: Parent/child Limited DOM (affected by style) (Mix Content/presentation)
Relationships: (Label, MemberOf - Group, ControllerFor) Limited to HTML (Title, alt, label)
Text Core DOM from parsed HTML
Content selection: Browser-dependent (incomplete)
Font/Font Style Information: Can set but can't get final format
Description/Help: Limited to HTML 4.0 - Alt Text, title text
Accessible value: Limited to form elements
Coordinates (Bounding rectangle, etc.): User Agents. platform accessibility API
Accessible Name:  
Respond Desktop Font/Color Changes: Partial (conflicts with CSS and JavaScript)
Device independent navigation:  
Accessibility API Mapping: Partial - User Agents
Provide focus to all active elements (important for equivalent keyboard access on desktops) Limited to forms and anchors

Filling the Gaps for Content Delivered to Desktop Browsers

This section has not been updated since it was integrated from the ARIA 1.0 Primer -- an APG task force review is pending.

At this time,W3C WAI Protocols and Formats working group's primary focus is on extensions to HTML 4.01 and XHTML 1.X by extending the host language to include WAI-ARIA with a migration path to HTML 5. This will require the creation of new hybrid document type definitions (DTDs) that incorporate the extensions. This work will be in the [[WAI-ARIA]] specification. WAI-ARIA will constitute extensions to fill most of the gaps needed to support accessibility API infrastructures and dynamic (X)HTML content. The comprehensive gap analysis of (X)HTML, used to form WAI-ARIA is found in Table 1.0 and how WAI-ARIA fills those gaps. In the future we hope to incorporate WAI-ARIA into many host languages to improve their accessibility. The critical extensions needed to make accessible dynamic Web content accessible, through rich interoperability with assistive technologies, are summarized here:

States, and Property attributes - This is the set of attribute modifications to (X)HTML necessary to provide full keyboard focus and states and properties that may be mapped directly or indirectly to platform accessibility APIs to ensure full interoperability with assistive technologies for WAI-ARIA.

Role attribute - The role attribute, borrowed from the, [[ROLE-ATTRIBUTE]], allows the author to annotate host languages with machine-extractable semantic information about the purpose of an element. It is targeted for accessibility, device adaptation, server-side processing, and complex data description. WAI-ARIA uses the role attribute to provides the role information, in to an assistive technology.

Role document landmark values - These values, borrowed from the [[ROLE-ATTRIBUTE]] provides a standard set of role attribute values designed to define pertinent parts of a document (landmarks) for the purpose of accessibility. User agents may incorporate device equivalents, such as key mappings in the case of a desktop user agent, to navigate to these sections of a document.

Taxonomy of WAI-ARIA role values - The necessary core roles found in Accessibility API sets for Windows and Linux as well as roles representative of document structure, such as banner or treegrid. Use of document structure is necessary for assistive technologies to navigate complex documents and to know when they have entered active areas of a Web page such as in the case of a dynamic scripted Web application. The taxonomy is designed to help user agents or authoring tools determine what properties a given role supports and to assist with accessibility API mapping of these properties. The taxonomy will is like a class hierarchy used to convey semantics and structure and includes knowledge about each role. At this time, that taxonomy is modeled using [[RDF-CONCEPTS]] and [[OWL-FEATURES]].

In short, WAI-ARIA will be used to fix the dynamic accessibility of scripted Web content, in particular the use of JavaScript with (X)HTML markup. They are meant to be cross-cutting and should apply to other markup like SVG. Less critical for (X)HTML but helpful for accessibility will be the descriptive extensions to XML events and the new [[XHTML Access]]. Web Content Accessibility Guidelines 2.0 calls for the WAI-ARIA properties in guideline 4.1 Maximize compatibility with current and future agents, including assistive technologies (roles, states, properties, and values) and section guideline 1.3 Create content that can be presented in different ways (for example spoken aloud, simpler layout, etc.) without losing information or structure (relationships).

The next section describes how the specifications are used together as well as how they will be implemented in HTML 4.

Use of New Provisions for Keyboard Focus and Semantics to Support Platform Accessibility APIs

Adaptive technologies, which need to provide alternative access to complex user interfaces authored via HTML, are often left guessing at the semantics behind specific portions of HTML document. As an example, an XHTML document might use a certain HTML construct, such as a collection of DIV tags, to create navigation bars, a site-navigation menu, and other GUI-like user interface widgets. To fix the problem, we incorporate the role attribute, assign the accessible state properties, and give the object focus using the new TABINDEX feature. Addition of this information helps authors to provide the necessary information to enable the user agent to support the accessibility API accessed by the adaptive technology.

Provision of the Role Attribute: "What is the Object?"

Each platform accessibility API has the notion of a "role" for a GUI object. This is the case for [[JAPI]], [[MSAA]]], [[AXAPI]], and the [[ATK]], or [[UI-AUTOMATION]] (called content type in UI Automation). The WAI-ARIA specifications are based on XHTML 1.X and include the role attribute. The "role" attribute takes a qname, enabling authors to reference the role attribute from the WAI-ARIA Roles. In the following example, we use qname to reference the menu role in the WAI-ARIA specification.

Example: Use of WAI-ARIA to incorporate role information into XHTML 1.x

<?xml version="1.1" encoding="us-ascii"?>
<!DOCTYPE html PUBLIC "Accessible Adaptive Applications//EN"

<html xmlns=""
        <div role="menu">

WAI used RDF/OWL to model our taxonomy for WAI-ARIA. In fact, if a host language sought to use namespaces or qnames, they could do so to reference the WAI-ARIA role taxonomy. The WAI-ARIA role taxonomy could be used by authoring tool developers to ensure that states and properties assigned to a given role are accurate.

Provision of the Accessibility State Information: "What meaningful properties does this object have at this time?"

Since this is dynamic content, the state of these new repurposed objects will change. The WAI-ARIA specification shall provide the common accessible properties needed to support the accessible state or property information provided by the platform accessibility API defined previously. This specification was created based on an analysis of the accessibility properties defined in MSAA and ATK. The following example extends the previous approach by adding the aria-haspopup accessibility property.

Example: Use of WAI-ARIA to incorporate accessible state information into XHTML 1.x

<?xml version="1.1" encoding="us-ascii"?>
<!DOCTYPE html PUBLIC "Accessible Adaptive Applications//EN"">

            <div role="menu" aria-haspopup="true">

A Windows user agent may now map this property to the Microsoft Active Accessibility state of STATE_SYSTEM_HASPOPUP. Adding or removing this state would result in the Windows user agent sending an EVENT_OBJECT_STATECHANGE event to the assistive technology. The task of the JavaScript page author would be to maintain this state attribute, which can easily be done through the use of Document Object Model calls.

Provision of the Keyboard or Input Focus: "What object am I working on?"

Virtually all adaptive technology solutions, such as screen readers and onscreen keyboards, must know which object currently has focus. For example, an author might want to insert text into the current object with focus or to announce information about the object that has focus. With today's HTML 4.01 and XHTML 1.x, script authors can only provide focus to form and anchor elements yet the Document Object Model Specification allows all elements to receive events including keyboard events. This means that HTML, by design prohibits script authors from making all HTML elements keyboard accessible. This single problem effects usability of Web pages where one gains access to elements by using the Tab key on desktop browsers. This slow, unproductive, approach makes portal navigation difficult because all active elements must be tabbed through to put focus on an active element in the last portlet in a document. To solve this problem in XHTML 1.x, we are incorporating a feature in Firefox and Internet Explorer to define the tabindex for -1. This allows a script author to give an element focus without placing it in the tab order: The following table describes these changes that will be incorporated into the new Accessible Adaptive Application specification.

Accessible Adaptive Application Changes to Support Use of tabindex to give Element Focus
tabindex attribute Focusable with mouse or JavaScript via element.focus() Tab navigable
not present Follows default behavior of element (yes for form controls, links, etc.) Follows default behavior of element
Negative, e.g. tabindex="-1" Yes No, author must focus it with element.focus() as a result of arrow or other key press
Zero, e.g. tabindex="0" Yes In tab order relative to element's position in document
Positive, e.g. tabindex="33" Yes Tabindex value directly specifies where this element is positioned in the tab order. These elements will be positioned in the tab order before elements that have tabindex="0" or that are naturally included in the tab order (form elements and links)

The following example shows the introduction of TABINDEX to provide focus to a DIV having the new accessibility meta data:

Example: Use of tabindex to give non-form and anchor elements focus in XHTML 1.x

<?xml version="1.1" encoding="us-ascii"?>
   <!DOCTYPE html PUBLIC "Accessible Adaptive Applications//EN"">

        <div role="menu" aria-haspopup="true" tabindex=-1>

Supporting WAI-ARIA in XHTML and HTML 4.01

Unlike XHTML, HTML 4.01 is non-extensible in that it is not possible to extend HTML 4 through the use of namespaces. That said, members of the working group have worked with the HTML working group to agree on a vehicle that does not use namespaces, which at this time is supported by XHTML and HTML which will be supported in HTML 5 when it becomes recommendation. Firefox 3 is leading the way to implement this, and other browser manufacturers are working to support it as well. The technique allows all role values specified in WAI-ARIA (including those specified by the XHTML Role attribute module) to be specified without a namespace prefix. Additionally, WAI-ARIA states and properties shall be represented as aria- followed by the concatenated WAI-ARIA state or property.

Example: Browser supported HTML technique for the tabindex example in section 5.1.3

<html lang="en">
<meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
</head> <body> <div role="menu" aria-haspopup="true" tabindex=-1> File </div> </body> </html>

In order to validate these extended attributes for HTML and XHTML the WAI-PF working group will investigate the creation of an enhanced schema or DTD for each host language.

Use of XHTML Role Landmarks to Improve Document Navigation

In addition to the common roles which will reside in WAI-ARIA Roles, both XHTML 2.0, and the XHTML Role attribute module ([[ROLE-ATTRIBUTE]], Section 4) defines a collection of common role, regional, landmarks that define pertinent parts of a document for the purpose of accessibility. User agents may incorporate device equivalents, such as key mappings in the case of a desktop user agent, to navigate to these sections of a document independent of the Web site. The addition of these semantics allows the user agent to provide standardized navigation to common document sections. This is especially important for portals to improve the usability. These may be used as attributes in XHTML 1.x by applying them to sections of the document as shown in this example. Note: since these roles are a part of XHTML they do not need to be namespace qualified.

Example: Use of XHTML navigation role to define a landmark for the navigation section in an XHTML 1.X document

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "">
<html xml:lang="en" xmlns="">
  <title>application/xhtml+xml: Navigation Roles Example 1</title>
  <link rel="stylesheet" href="css/nav1_xhtml.css"  type="text/css" ></link>
  <script type="text/javascript" src="../js/globals.js"></script>
  <script type="text/javascript" src="../js/widgets_xhtml.js"></script>  <link rel="icon" href="" type="image/x-icon" />
  <link rel="shortcut icon" href="" type="image/x-icon" />
<body onload="widgets.init()">
<div id="leftnav">
<h2 class="nav" id="leftnav_label">Career Center Services</h2>
 <ul title="Career Center Services" role="navigation" aria-labelledby="leftnav_label">
   <li><a href="">Career Home</a></li>
   <li><a href="">Career Counseling</a></li>
   <li><a href="">Pre-Health Advising</a></li>
   <li><a href="">Services</a></li>
   <li><a href="">Workshops</a></li>
   <li><a href="">Resource Center</a></li>
   <li><a href="">Question Board/FAQ</a></li>



The example above was taken from the header from the Career Center Web page at the University of Illinois at Urbana-Champaign. Students from this university, under Jon Gunderson's guidance created Accessibility extensions for Mozilla/Firefox, in part, to allow a page author or user to view a list of navigation landmarks. This tool, shown in , lists the navigation sections on the page. Keyboard navigation of the list of navigation bars causes the corresponding document section to be highlighted. The title for each navigation region displays in the list.

Table of Contents from Landmarks
Table of Contents generated from navigation landmarks in the header

shows the accessibility extensions for Mozilla/Firefox from the University of Illinois at Urbana-Champaign to render the document landmarks. This picture shows the Firefox browser rendering the University of Illinois Career Center home page. In this example. The "List of Navigation Bars" viewer is shown, launched from the extension on the toolbar. The viewer shows that the secondary "Career Center Services" is selected resulting in that section of the document being highlighted in yellow. The Navigation Bar Lists Viewer lists the following list of titles corresponding to the navigation sections:

  • Career Counseling Resources
  • Resources by Audience
  • Career Center Services
  • Quick Links

WAI-ARIA Role Taxonomy - Extensible Semantic Role Model, using RDF/OWL

The WAI-ARIA role taxonomy was modeled using semantic web technology, in the form of [[RDF-CONCEPTS]] and the [[OWL-FEATURES]], as a vehicle to define a knowledge-based class hierarchy for roles. This model shows what states and properties are supported, by each role in the taxonomy. The role in the class hierarchy inherits properties from its ancestors and defines its own properties. Where applicable, semantic web technology is used to define related concepts within other namespaces. This information is critical in determining how to choose a role and how to interact with it. The role taxonomy uses RDF as a way for using data to describe data and provides a W3C standards-based approach to represent this information.

Sample Semantic Map for Taxonomy
Example, Partial RDF Map for a possible ButtonUndo role as an extended role to WAI-ARIA

shows a basic RDF mapping that defines a set of terms and relationships defining an object. At the center is a Widget object that defines common states and properties for all GUI widgets. The Button object extends Widget and inherits defined accessibility properties from the superclass Widget. It also defines a relatedConcept property to a Link object. The ButtonUndo role extends Button. It has a relatedConcept of an HTML input object. ButtonUndo will introduce Dublin Core meta data such as the description of the object. The terms relatedConcept and requiredState are terms that will be defined as part of the corresponding RDF schema. Each role instance will represent standard Roles found in the platform accessibility APIs of platforms like Windows and Gnome as well as content structure. These roles will form the taxonomy. Although host language browser implementations may reference WAI-ARIA roles without namespaces, the RDF representation for a given role may be referenced using a qname from a Host XML markup language. This examples shows an XHTML reference to a grid role in the RDF representation of the WAI-ARIA taxonomy:

<div role="grid"> whereby grid expands to: in the button markup.

The power of this design is that it enables a web authoring tool to go back into the corresponding RDF/OWL markup and determine what properties it supports for Accessibility API mapping. Additional, middleware solutions can now make intelligent transformations of Web Content by processing the semantics behind rich browser and rich structured frameworks to adapt accessible solutions for a broader user base. Our immediate goal is to fix the accessibility problem with scripted Web content. Assistive technologies will use the standard roles to determine how to render most content.

Interoperability Example: Grid Role

To understand the power of this approach, consider the case of a Grid Role, analogous to a table. shows a DHTML example using XHTML, JavaScript, and CSS to produce a GUI-like application. This example developed in IBM shows a notebook tab with a data grid that behaves like a traditional desktop GUI. The user uses arrow keys to navigates the data grid and among the page tabs. Using the Tab key, a user navigates between the notebook tab, the edit fields, buttons, and the data grid.

DHTML example of GUI-like notebook tab with a data grid
DHTML Example

Accessible role and state meta data from the WAI-WAI-ARIA Roles, States, and Properties specification, are added as attributes to each of the XHTML elements repurposed as GUI widgets dynamically. The user agent, in this case Firefox, maps this information to the platform accessibility API. shows the Microsoft Active Accessibility rendering of the new accessibility markup provided on the DataGrid page tab which has focus.

MSAA Inspect Tool diagnostics for Notebook page tab
Microsoft Inspect Tool rendering of the page tab DataGrid

show a Microsoft Inspect 32 rendering of the DataGrid Page page tab in Figure 5.0. Inspect32 provides Microsoft Active Accessibility information; here it shows the accessible role of "page tab" and accessible state information of focused, focusable, and linked. There are no page tab elements in XHTML. Here, an XHTML DIV element is repurposed by a JavaScript controller to look like a notebook tab. It is now able to receive focus, unlike in standard XHTML 1.X, and it does so without requiring tabbing. With these specifications, the script author can now add the accessibility properties to support platform accessibility API. Accessible state properties for the DataGrid page tab are shown as focused, focusable, and linked. Unlike a GUI application, authors need only enable their applications once for multiple operating system platforms.

Beyond scripted Web content, the working group intends to extend the use of roles to enable other user cases. These may include:

  • Structured Textual Markup - enhancing structure of the markup of a document, including Data Tables , or translating the structure of an XML document to a markup structure that user agents are used to dealing with (e.g. myXML to XHTML) Binding sections of a document to a common role. This allows for different navigation techniques though a document
  • Knowledge representation of Web content - As a secondary benefit, roles improve compatibility with Knowledge-Based Services and the Semantic Web. By integrating accessibility and the semantic Web, accessibility can be moved forward, paving the way for customized accessible searches and intelligent user agents with additional applications.
  • Adding concepts in the type of content for adaptation to the user scenario - The more that is understood about content, the better it can be adapted for the end user. For example:
    1. If it is known that a page hyperlink has the role of taking the user to the site's home page, then that knowledge can be used to create enhanced accessibility in different ways in many different scenarios, such as icons or access keys.
    2. If it is known that a text box is for the user email address, then the user agent can support users filling in the form by labeling it with an icon or symbol, automatically validating it, or even form filling.
    3. If it is known that a paragraph is complex, then a simple equivalent can be shown in its place
    4. If a word is ambiguous, then a role of a concept can be given, providing clarity An example of this may be : <span role="role:nonliteral" aria:hasAlternate="no">

Accessibility Events and Event Handling

Interoperability between applications and assistive technologies requires event notification for accessibility. The events will be fired via the user agent. The accessible value and state property changes will be generated in response to changes in the DOM attributes as defined by the WCAG 1.0 AAA specification. User agents supporting the platform accessibility API, will support event notification such as the state change or value change events.

Building Accessible Applications with WAI-ARIA

This section has not been updated since it was integrated from the ARIA 1.0 Primer -- an APG task force review is pending.

This section provides a brief introduction to the process of making applications accessible using WAI-ARIA. The choice and usage of roles can be complex and context dependent. It is beyond the scope of this document to explain implementations for all the possible WAI-ARIA use cases. WAI-ARIA Authoring Practices provides detailed guidance on WAI-ARIA implementation methodology as well as references to sample code.

First steps to making an application accessible:

  1. Each element or widget has correct and complete semantics that fully describe its behavior (using element names or roles);
  2. The relationships between elements and groups are defined;
  3. States, properties, and container relationships are valid for each element's behavior and are accessible via the [[dom]] and the platform accessibility API; and
  4. Keyboard focus should be maintained for the duration of the user's interaction with the application.
  5. All interactive components should be keyboard operable.

WAI-ARIA provides authors with the means to make the different elements in a web application semantically rich. User agents use the role semantics to understand how to handle each element. Roles convey additional information that the assistive technologies need to anticipate the behavior of the elements inside the application such as how to present the corresponding WAI-ARIA states and properties to the user. The user agent will use the accessibility semantics from the host language and WAI-ARIA accessibility semantics (which may augment or override those of the host language) and present them to assistive technologies through the Document Object Model or the platform accessibility API. The user agent will create an accessible representation of each element in the web page, and will use the appropriate accessibility API to notify assistive technologies of changes to the semantics of those elements.

The following steps are recommended when WAI-ARIA is applied to content:

  1. Use native markup when possible.

    Use the semantic elements that are defined in the host markup language. For example, with HTML or XHTML, it is better to use the native checkbox than to use a div element with role checkbox as these should already be accessible through your browser. There may also be cases where WAI-ARIA can augment an existing element in the host language. For example, a grid and gridcell elements can reuse the functionality of a table when overlaying it. WAI-ARIA roles, states, and properties are best used when the markup language does not support all the semantics required. When a role attribute is added to an element, the semantics and behavior of the element are augmented or overridden by the role behavior.

  2. Apply the appropriate roles.

    Set roles to make sure elements behave predictably and correctly describe the behavior of each element within the application, unless element behaviors are fully described by the native markup language. Roles for interactive elements should support all the attributes that the element could use. Once a role attribute is set it should not be changed as this may confuse the end user. This does not preclude an element being removed which has the role attribute set, but only states and properties (aria-* attributes) should be changed for a given element.

  3. Preserve semantic structure.

    Structural information is critical to providing context to persons with disabilities. Preserve DOM hierarchy within structural elements and widgets:

    • Form logical groups within user interface widgets, such as treeitem elements in a group.
    • Identify large perceivable regions and apply a landmark role to those areas. This will facilitate keyboard navigation. It will also facilitate content management by assistive technologies by providing semantics to manage how much information is rendered at a given time. The use of WAI-ARIA landmarks helps everyone, including vision-impaired users, dexterity-impaired users, and even users with cognitive or learning impairments.
    • For areas of the page that contain a group of elements that are likely to change through an Ajax application it may be specified as a live region, through the use of the aria-live attribute or it may be marked with pre-defined roles, such as log, which has assumed behavior of a live region.
  4. Build relationships.

    Look for relationships between elements, and mark them using the most appropriate property or attribute. For example, if a page contains both a search form and search results region, mark each container as a region and set the aria-controls attribute of the search form to reference the search results. See relationships in WAI-ARIA.

    Some relationships are determined automatically from the host language, like label elements associated with input elements in HTML.

  5. Set states and properties in response to events.

    Once the role for an element has been set, change states and properties as appropriate during the element's life cycle, usually in response to user input events. Only use attributes supported for the chosen role or element.

    User agents should notify assistive technologies of state changes. Conversely, assistive technologies' notification of property changes depends on the method by which assistive technologies communicate with the user agent. For example, the aria-multiline attribute (a property) is not something that changes frequently, whereas the aria-checked attribute (a state) changes frequently in response to user input.

  6. Support full, usable keyboard navigation.

    Usable keyboard navigation in a rich internet application is different from the tabbing paradigm in a static document. Rich internet applications behave more like desktop applications where the user tabs to significant widgets and uses the arrow keys to navigate within a complex widget, such as a menu or spreadsheet. The changes that WAI-ARIA introduces in keyboard navigation make this enhanced accessibility possible. Tabbing in the document should follow a logical navigation structure. Authors implementing arrow key navigation should, where possible, follow the design patterns in the WAI-ARIA Authoring Practices Guide. When using these features, it is important as always to ensure keyboard navigation is logical.

  7. Synchronize the visual interface with the accessible interface.

    This will allow the state of your UI to be perceivable to the user as well as assistive technologies. For example, the author should use the appropriate WAI-ARIA attribute on a form element that is visually styled to appear required (aria-required) or a gridcell that is visually styled to appear selected (aria-selected). Authors may choose to achieve visual synchronization of these interface elements by using a script or by using CSS attribute selectors.

Example: Building a Tree Widget

Graphic of an example tree view.

A basic tree view allows the user to select different list items and expand and collapse embedded lists. Arrow keys are used to navigate through a tree, including left/right to collapse/expand sub trees. Clicking the visible expander button with the mouse also toggles expansion. Further keyboard implementation details for tree widgets may found in the .

To make this feature accessible we need to:

  • Inform assistive technologies about the role of each element;
  • Inform assistive technologies about the relationships between tree items;
  • Give a clear keyboard focus that will not confuse users with disabilities; and
  • Expose the changing states (expanded and collapsed) of the tree items.

Following the steps below:

  1. Look at the native markup language

    Although standard list behavior is supported by the native ul and li elements in HTML, there is no element that natively supports list expansion and selection. Since there is not, we will need to use roles.

  2. Finding the right roles

    Since there is no native tree element in HTML, we need to add roles from the taxonomy that support the additional states and properties needed. The roles that support tree behavior are:

    • tree: A tree is the main container element for our tree. It is a form of a select where sub-level groups of treeitems may be collapsed and expanded.
    • treeitem: A treeitem is an option item of a tree. This is an element within a tree; sub-level groups of treeitems may be expanded or collapsed.
    • group: A group encloses a set of sub-level treeitems.
  3. Look for groups and build relationships

    Tree relationships can be made simply via the DOM and logical structure of the page. A tree element will be the main container encompassing all other elements in the tree. Each selectable item in the tree will be a treeitem.

    When a treeitem contains an embedded list of treeitems they will be all be embedded in a group. A group should be contained inside the tree item that is the parent item.

    <h1 id="treelabel">WAI-ARIA Tree Example</h1>
    <ul role="tree" aria-labelledby="treelabel" aria-activedescendant="example_id" tabindex="0">
      <li role="treeitem" aria-expanded="true">Animals
        <ul role="group">
          <li role="treeitem">Birds</li>
          <li role="treeitem" aria-expanded="false">Cats
            <ul role="group">
              <li role="treeitem">Siamese</li>
              <li role="treeitem">Tabby</li>
          <li role="treeitem" aria-expanded="true">Dogs
            <ul role="group">
              <li role="treeitem" aria-expanded="true">Small Breeds
                <ul role="group">
                  <li role="treeitem">Chihuahua</li>
                  <li role="treeitem" id="example_id">Italian Greyhound</li>
                  <li role="treeitem">Japanese Chin</li>
              <li role="treeitem" aria-expanded="false">Medium Breeds
                <ul role="group">
                  <li role="treeitem">Beagle</li>
                  <li role="treeitem">Cocker Spaniel</li>
                  <li role="treeitem">Pit Bull</li>
              <li role="treeitem" aria-expanded="false">Large Breeds
                <ul role="group">
                  <li role="treeitem">Afghan</li>
                  <li role="treeitem">Great Dane</li>
                  <li role="treeitem">Mastiff</li>
      <li role="treeitem" aria-expanded="true">Minerals
        <ul role="group">
          <li role="treeitem">Zinc</li>
          <li role="treeitem" aria-expanded="false">Gold
            <ul role="group">
              <li role="treeitem">Yellow Gold</li>
              <li role="treeitem">White Gold</li>
          <li role="treeitem">Silver</li>
      <li role="treeitem" aria-expanded="true">Vegetables
        <ul role="group">
          <li role="treeitem">Carrot</li>
          <li role="treeitem">Tomato</li>
          <li role="treeitem">Lettuce</li>

    The use of aria-expanded should mirror that which is visibly expanded on screen, so authors may wish to use CSS attribute selectors to toggle visibility or style of an item based on the value of an WAI-ARIA state or property. The following example would hide the sub-level groups of a collapsed tree node.

    [aria-expanded="false"] [role="group"] { display: none; }

    At the time of this writing, this CSS example, while technically correct, will not redraw styles properly in some browsers if the attribute's value is changed dynamically. It may be necessary to toggle a class name, or otherwise force the browser to redraw the styles properly.

    Sometimes a tree structure is not explicit via the DOM and logical structure of a page. In such cases the relationships must still be made explicit using the states and properties. In the following example, the aria-owns attribute indicates that the div with id "external_treeitem" should be considered a child of the ul element with the attribute, even though it is not a child in the DOM.

    <li role="treeitem" aria-expanded="true" aria-owns="external_group">Vegetables</li>
    <ul role="group" id="external_group">
      <li role="treeitem">Carrot</li>
      <li role="treeitem">Tomato</li>
      <li role="treeitem">Lettuce</li>

    Sometimes trees (and other lists or grids) cannot be completed represented in the DOM due to performance limitations of the browser. For example, an application interface may only need to display 100 items out of a set of 100,000. Including all 100,000 items in the DOM could cause performance problems on both the client and server machines.

    If items in a managed widget are loaded, for example, via the XMLHttpRequest object, and not present in the DOM at all times, authors should use aria-level, aria-posinset and aria-setsize, and ensure that aria-owns is not required to convey membership with the widget.

    <li role="treeitem" aria-level="1" aria-posinset="3" aria-expanded="true" aria-setsize="3">
      <ul role="group">
        <li role="treeitem" aria-level="2" aria-posinset="6" aria-setsize="8">Carrot</li>
        <li role="treeitem" aria-level="2" aria-posinset="7" aria-setsize="8">Tomato</li>
        <li role="treeitem" aria-level="2" aria-posinset="8" aria-setsize="8">Lettuce</li>
  4. Use states and properties in response to events

    Control the behavior of the element in response to user input events such as from the keyboard and the mouse, which includes maintaining the current states and properties for that element. For example, a tree control will need to respond to click events on the expand/collapse triggers, and key events on the currently active descendant. Use device-independent events with supporting JavaScript to handle user interaction. For detailed examples of this please refer to the Design Patterns section.

Reasons for Adopting WAI-ARIA

This section has not been updated since it was integrated from the ARIA 1.0 Primer -- an APG task force review is pending.

By adopting WAI-ARIA, both developers of static web sites and of dynamic Internet applications can improve the usability, accessibility, and scalability of their products. Developers of static web content can continue to follow the 1999 WCAG 1.0 standards, while improving usability and accessibility through the use of WAI-ARIA landmark roles, aria-labelledby relationships, and properties like aria-invalid and aria-required that can apply to HTML form controls. In richer, dynamic content, developers create custom widgets like calendars and spreadsheets based on technologies such as Ajax and CSS; to achieve accessibility, they need to use WCAG 2.0. Previously, when delivering rich Internet applications to users, to comply with WCAG 1.0, developers resorted to providing alternative, static content. While such measures met the WCAG 1.0 requirements, people using assistive technologies were not provided the richer user experience. When tables are used for layout, rather than CSS absolute positioning, historically, they have been problematic for assistive technologies to interpret. When the WAI-ARIA role of presentation is used on such tables, the assistive technology ignores the table structure, providing a more accessible experience without requiring major recoding.

Technical Benefits

Consider a rich Internet application where developers attempt to achieve keyboard accessibility using markup language. Without WAI-ARIA, results may be keyboard accessible but not highly usable; consider a user having to press Tab thirty times to put focus on a checkbox. To achieve keyboard accessibility in HTML without WAI-ARIA, developers must code active elements either as a link or as a form element. Accordingly, this means that no additional semantics are required or provided, other than that provided by the host language. In addition, WCAG 1.0 requires that content be renderable with Cascading Style Sheets turned off in the browser. This approach creates the following general usability problems:

  • All keyboard-accessible controls must be either forms or anchors, forcing the user to tab through all focusable elements on the web page to navigate. If you need to navigate from the first link on the page to the last link on the page, that could be a very long trip and takes usability a step back.
  • Without WAI-ARIA semantics, you cannot provide contextual information to the user.
  • If you repurpose HTML elements you cannot provide the appropriate role and context information for the new widget. Lack of context is a serious usability problem. WAI-ARIA semantics results in providing contextual information to the user.
  • CSS is used for absolute positioning. If you remove that capability, usability features of widgets like pop-up menus disappear. Imagine activating the file menu and the menu showing up at the bottom of the page.

WAI-ARIA and WCAG 2.0 coding techniques are useful for developing content and applications that can scale across a variety of user agents, including those on mobile devices.

For all these reasons, adopting WAI-ARIA makes good technical as well as business sense. For a further illustration, compare how accessibility is achieved with WCAG techniques without and with WAI-ARIA, as shown in .

Editor's Note: Figure 7, described as WAI-ARIA tree widget usability comparison, refers to a resource that has not yet been found.

Usability of Tree Widget Using WAI-ARIA Semantics to Implement WCAG 2.0 Guidelines Compared to WCAG 1.0 Without WAI-ARIA

shows an "accessible" widget for a tree item, within a tree widget, using WCAG 1.0 without WAI-ARIA, which ,when supplied to a screen reader, may say "link folder Year." There is no information to indicate that the folder is closed (aria-expanded = "false"). There is no information to indicate its depth (aria-level="2"), position in the set and the set size at the level, all of which provides the user with context something sighted users may take for granted. The role is used to indicate that it is a treeitem which indicates that the item should behave and navigate with the keyboard like a tree item. A screen reader using the WAI-ARIA version might say "Closed Folder Year, Closed Folder one of two, Depth two, unchecked." Furthermore, the WAI-ARIA version can allow the tree items to be navigated with arrow keys and without requiring they be navigated as part of a global web page tabbing scheme. This is not what any user would expect of a tree widget. Finally, if you were told all widgets were links on the web page, consider how usable -- or not -- that would be.

Business Benefits

If, as described above, coding techniques to achieve accessibility compliance do not promote overall usability, then business strategists must ask "Why invest in accessibility?" With WAI-ARIA, businesses can invest in accessible development and reap benefits for all users, not just those with disabilities or using assistive technologies. Some benefits include:

  • Because WAI-ARIA is being developed through the PFWG with cooperation from browser and assistive technology vendors, accessibility and interoperability with those technologies will be easier to achieve, reducing or eliminating the need for per-browser and per-screenreader coding to achieve accessibility.
  • In addition to people with disabilities, all users benefit from WAI-ARIA because it establishes a Web-standard for keyboard interaction, easing the learning curve for users moving among applications, sites, and browsers.
  • Implementing WAI-ARIA can facilitate test automation. Test engines require semantic information about user interface elements, unique names, exposure of state, specific properties in order to run automated test scripts. WAI-ARIA provides that semantic information needed for efficient test automation.