Core Accessibility API Mappings 1.2

W3C Editor's Draft

This version:
Latest published version:
Latest editor's draft:
Latest Recommendation:
Joanmarie Diggs (Igalia, S.L.)
Michael Cooper (W3C)
Former editors:
Richard Schwerdtfeger (Knowbility) (Editor until October 2017)
Joseph Scheuhammer (Inclusive Design Research Centre, OCAD University) (Editor until May 2017)
Andi Snow-Weaver (IBM) (Editor until December 2012)
Aaron Leventhal (IBM) (Editor until January 2009)
Platform Mapping Maintainers:
Bogdan Brinza (Microsoft Corp.) (UIA)
James Craig (Apple, Inc.) (AX API)
Joanmarie Diggs (Igalia, S.L.) (ATK / AT-SPI)
Alexander Surkov (Mozilla Foundation) (MSAA, IAccessible2)


This document describes how user agents should expose semantics of web content languages to accessibility APIs. This helps users with disabilities to obtain and interact with information using assistive technologies. Documenting these mappings promotes interoperable exposure of roles, states, properties, and events implemented by accessibility APIs and helps to ensure that this information appears in a manner consistent with author intent.

This Core Accessibility API Mappings specification defines support that applies across multiple content technologies, including general keyboard navigation support and mapping of general-purpose roles, states, and properties provided in Web content via WAI-ARIA [WAI-ARIA]. Other Accessibility API Mappings specifications depend on and extend this Core specification for specific technologies, including native technology features and WAI-ARIA extensions. This document updates and will eventually supersede the guidance in the WAI-ARIA 1.0 User Agent Implementation Guide [WAI-ARIA-IMPLEMENTATION] W3C Recommendation. It is part of the WAI-ARIA suite described in the WAI-ARIA Overview.

Status of This Document

This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this technical report can be found in the W3C technical reports index at

This document was published by the Accessible Rich Internet Applications Working Group as an Editor's Draft.

Comments regarding this document are welcome. Please send them to (archives).

Publication as an Editor's Draft does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.

This document was produced by a group operating under the W3C Patent Policy. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the group; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.

This document is governed by the 1 February 2018 W3C Process Document.

1. Introduction §

This section is non-normative.

In traditional desktop graphical user interface (GUI) applications, components of the user interface (UI) are displayed when needed and hidden when not needed based on user interactions. Accessibility application programming interfaces (APIs) are used to communicate semantic information about the user interface to assistive technology software used by people with disabilities. These APIs constitute a contract between applications and assistive technologies, such as screen readers, magnifiers, alternate input devices, and speech command and control, to enable them to access the appropriate semantics needed to produce a usable alternative to interactive applications. For example, screen reading software for blind users can determine whether a particular UI component is a menu, button, text field, list box, etc.

In traditional static Web pages, the HTML elements provided the necessary semantic information. The user agent provides keyboard navigation but only to the HTML elements that are known to be interactive, specifically links and form elements. Assistive technologies obtain the semantic information from the Document Object Model (DOM) or, in the case of links and form elements, through the Accessibility API. In both cases, the assistive technology expects that nothing changes until a new page is loaded based on a user action.

Yet technologies such as JavaScript, Ajax, and CSS have enabled Web pages to look and behave more like interactive desktop GUI applications, without the need to reload the page with each user interaction. Developers can now re-purpose HTML elements into UI components not previously defined in HTML. For example, Javascript can be used with CSS to modify a <div> element based on user interactions to make it look and behave like a popup menu. Unfortunately, the <div> element does not provide the author with a vehicle to add semantic metadata that can be exposed through the DOM and mapped to Accessibility APIs. These accessibility deficiencies in traditional markup render rich Internet applications unusable by people who use assistive technologies or who rely on keyboard navigation.

The W3C Web Accessibility Initiative's (WAI) Protocols and Formats working group (PFWG) has addressed these deficiencies through several W3C standards efforts, with a focus on the Accessible Rich Internet Applications [WAI-ARIA] specification.

WAI-ARIA enables rich Internet applications to have the same accessibility features as desktop GUI applications by adding metadata to markup technologies such as (X)HTML. Authors include WAI-ARIA in their markup and user agents translate the WAI-ARIA markup to the platform accessibility APIs.

For an introduction to WAI-ARIA, see the WAI-ARIA Overview. The User Agent Implementation Guide describes how WAI-ARIA roles, states, and properties should be supported in user agents using platform accessibility APIs. It is part of a set of resources that define and support the WAI-ARIA specification which includes the following documents:

The WAI-ARIA User Agent Implementation Guide begins by providing a general overview of accessibility APIs and the accessible object hierarchy known as the accessibility tree. The following sections give guidance on supporting keyboard navigation and mapping WAI-ARIA roles, states, and properties to accessibility APIs. Other sections give guidance on calculating text alternatives, mapping actions to events, event processing, special document handling procedures, and error handling.

This guide assumes that a user agent already exposes static content to assistive technology via the accessibility API on a given platform. Most of the additional work to enable WAI-ARIA can be divided into three parts:

  1. Enabling keyboard navigation on elements that previously were not focusable
  2. Mapping the WAI-ARIA roles and attributes into the roles, states and other property getters in the accessibility API
  3. Computing text alternatives and managing states and events

In general, WAI-ARIA attributes should only affect how content is mapped to platform accessibility APIs. They should not affect the visual rendering of content nor behavior of mainstream desktop browsers, except when style sheets are deliberately keyed off of WAI-ARIA attributes as recommended in the specification. This allows one of the primary principles of WAI-ARIA to be upheld—that content still renders and behaves the same for the majority of users.

This document includes information for user agents specifying how to map WAI-ARIA roles, states, and properties to platform accessibility APIs. It also includes host-language specific requirements where necessary to complete the accessibility model. Examples of host languages include HTML and SVG. However, in order to provide the basic, core mappings of WAI-ARIA, discussion of host-language features are generally avoided. How host languages modify or override the core mappings is specified in separate documents.

1.1 Accessibility APIs  §

To provide access to desktop GUI applications, assistive technologies originally used heuristic techniques to determine the meaning of the user interface and build an alternative off-screen model. For example, a row of labels displayed horizontally near the top of an application window might be a menu. Labels with a border drawn around them might be buttons. Heuristic techniques are not always accurate, however, and require assistive technologies to be updated whenever the software application is updated.

A much better technique is for the software application to provide the necessary information for interoperability with assistive technology. To meet this need, platform owners have developed specialized interfaces, called accessibility APIs, which can be used to communicate accessibility information about user interfaces to assistive technologies. Accessibility APIs allow developers to express the function of controls and text to assistive technologies. Accessibility APIs include a tree of accessible objects (controls and text) and information about each of them:

  1. Descriptive properties (role, name, value, position, etc.),
  2. Transient states (pressed, focusable, etc.),
  3. Events (text changed, button was clicked, checkbox was toggled),
  4. Actions the user might take (click, check/toggle, drag, etc.),
  5. Relationships (parent/child in the tree, controls, flowsto, etc.), and
  6. Textual content.

In the case of Web pages, the Document Object Model (DOM) is used to represent the structure and state of the elements in the document being rendered by a user agent. The elements of the document are organized into a hierarchy of nodes known as the DOM tree. To interact with static Web content, assistive technologies, such as screen readers, have tended to rely on the DOM provided by the user agent. However, platform accessibility APIs provide a quicker and more comprehensive way for assistive technologies to learn about and interact with Web page content. Especially with UI elements that are known to be interactive, such as HTML form elements and desktop applications, accessibility APIs allow the more complex roles, properties, states, and relationships of those elements to be communicated to assistive technology in a way that the DOM cannot provide on its own.

In the case of rich Internet applications, developers use DOM APIs to manipulate objects in the DOM tree to make them behave like interactive desktop GUI applications. In order to make a Web application understandable to assistive technologies, the user agent needs to map accessibility information from the elements in the DOM tree to the Accessibility APIs of the underlying operating system or software platform throughout the lifecycle of the application. The information needed is provided when developers use WAI-ARIA to supply semantic role, state, and property information for elements. The screen reader or other assistive technology uses the semantic information exposed via the accessibility API to provide an alternative rendering of an application that is meaningful to a user.

Accessibility APIs covered by this document are:

The WAI-ARIA 1.0 User Agent Implementation Guide included mappings for UIA Express, also known as IAccessibleEx, which was implemented in Microsoft Internet Explorer 8.0 - 11. New implementations are strongly encouraged to use User Interface Automation instead.

If user agent developers need to expose information using other accessibility APIs, it is recommended that they work closely with the developer of the platform where the API runs, and assistive technology developers on that platform.

1.2 The Accessibility Tree and the DOM Tree §

The accessibility tree and the DOM tree are parallel structures. Roughly speaking the accessibility tree is a subset of the DOM tree. It includes the user interface objects of the user agent and the objects of the document. Accessible objects are created in the accessibility tree for every DOM element that should be exposed to an assistive technology, either because it may fire an accessibility event or because it has a property, relationship or feature which needs to be exposed. Generally if something can be trimmed out it will be, for reasons of performance and simplicity. For example, a <span> with just a style change and no semantics may not get its own accessible object, but the style change will be exposed by other means.

1.3 Comparing Accessibility APIs §

For various technological and historical reasons, accessibility APIs do not all work in the same way. In many cases, there is no simple one-to-one relationship between how each of them names or exposes roles, states, and properties to user agents. The following subsections describe a few of the distinguishing characteristics of some of the APIs.

1.3.1 ATK/AT-SPI §

MSAA, IAccessible2, UIA, and AX API each define an API that is shared by both the software application exposing information about its content and interactive components, and the user agent (assistive technology) consuming that information. Conversely, Linux/GNOME separates that shared interface into its two aspects, each represented by a different accessibility API, ATK or AT-SPI.

ATK defines an interface that is implemented by software in order to expose accessibility information, whereas AT-SPI is a desktop service that gathers accessibility information from active applications and relays it to other interested applications, usually assistive technologies.

For example, the GNOME GUI toolkit [GTK], implements the relevant aspects of ATK for each widget (menu, combobox, checkbox, etc.) in order that GTK widgets expose accessibility information about themselves. AT-SPI then acquires the information from applications built with GTK and makes it available to interested parties.

ATK is most relevant to implementors, whereas AT-SPI is relevant to consumers. In the context of mapping WAI-ARIA roles, states and properties, user agents are implementors and use ATK. Assistive Technologies are consumers, and use AT-SPI.

1.3.2 UIA (UI Automation) §

UI Automation expresses every element of the application user interface as an automation element. Automation elements form the nodes of the application accessibility tree, that can be queried, traversed and interacted with by automation clients.

There are several concepts central to UI Automation:

  • Automation element - controls and some application content is presented as automation elements.
  • Element properties - Automation elements have several common properties describing native framework element characteristics in an agnostic way that all automation clients can understand. There are several ways to access element property values, described below.
  • Control Patterns - Some common interactivity in different frameworks is expressed as control patterns in UIA, allowing different automation client to interact with controls using common programmatic interfaces.
  • Events - Similarly to other accessibility APIs, automation elements support various events that allow automation providers notify clients on important state changes.

All automation elements inherit from the IUIAutomationElement interface and all properties that are not specific to a particular control pattern can be queried through that interface. There are several ways to access UI Automation element properties:

  • Direct property accessors to the current values - Current{PropertyName}, e.g. IUIAutomationElement::CurrentName for the Name property
  • Cached property accessors - Cached{PropertyName}, e.g. IUIAutomationElement::CachedName for the Name property. Using cached values is preferred when providers and clients are used in remote environments.
  • GetCurrentPropertyValue and passing the UIA Property ID enumeration value corresponding to that property to get the current value, e.g. IUIAutomationElement::GetCurrentPropertyValue(UIA_NamePropertyId) for the Name property.
  • GetCachedPropertyValue and passing the UIA Property ID enumeration value corresponding to that property to get the cached value, e.g. IUIAutomationElement::GetCachedPropertyValue(UIA_NamePropertyId) for the Name property.

Properties for specific UIA control patterns are queried the same way using relevant control pattern interfaces. Taking Toggle Pattern as an example, to query the ToggleState property clients can use IUIAutomationTogglePattern::CurrentToggleState or IUIAutomationTogglePattern::GetCurrentPropertyValue(UIA_ToggleToggleStatePropertyId) to get the current value.

The property mapping in this specification provide the {PropertyName} and do not specify all specific ways to access the property value. Automation clients can access current or cached values using conventions described above, depending on specific needs and coding style conventions.

1.3.3 Accessible Names and Descriptions §

Each platform accessibility API includes a way to assign and retrieve accessible name and accessible description properties for each accessible object created in the accessibility tree. How these the relevant properties are implemented and what they are called vary depending on the API.

For instance, in MSAA, all accessible objects support the accName property, which stores the object's accessible name. Where the object also supports having an accessible description, MSAA stores this property in the object's accDescription property.

Software using ATK can read and write to an object's accessible-name and accessible-description properties. In turn, AT-SPI can query the values of those properties through its atspi_accessible_get_name and atspi_accessible_get_description functions.

Automation elements in the UIA accessibility tree have a Name property. Where the object also supports having an accessible description, UIA stores this property in the object's FullDescription property.

The approach to accessible names and accessible descriptions in AX API is somewhat different to the other platform APIs. Accessible names are exposed using the AXTitle property when the name is visually rendered, while the AXDescription property is used when the object's name is not rendered visually. An object's accessible description, where provided, should always be exposed in the AXHelp property.

For more detail, see the Accessible Name and Description Computation and API Mappings specification.

2. Conformance §

As well as sections marked as non-normative, all authoring guidelines, diagrams, examples, and notes in this specification are non-normative. Everything else in this specification is normative.

The key words MAY, MUST, MUST NOT, OPTIONAL, SHOULD, and SHOULD NOT are to be interpreted as described in [RFC2119].

2.1 RFC-2119 Keywords §

RFC-2119 keywords are formatted in uppercase and contained in a strong element with class="rfc2119". When the keywords shown above are used, but do not share this format, they do not convey formal information in the RFC 2119 sense, and are merely explanatory, i.e., informative. As much as possible, such usages are avoided in this specification.

2.2 Normative and Informative Sections §

The indication whether a section is normative or non-normative (informative) applies to the entire section including sub-sections.

Informative sections provide information useful to understanding the specification. Such sections may contain examples of recommended practice, but it is not required to follow such recommendations in order to conform to this specification.

2.3 Features Deprecated in WAI-ARIA §

The WAI-ARIA specification lists some features as deprecated. Although this means authors are encouraged not to use such features, it is expected that the features could still be used in legacy content. Therefore, it is important that user agents continue to map these features to accessibility APIs, and doing so is part of conformance to this specification. When future versions of the WAI-ARIA specification change such features from deprecated to removed, they will be removed from the mappings as well and user agents will no longer be asked to continue support for those features.

3. Important Terms §

While some terms are defined in place, the following definitions are used throughout this document.

Accessibility API

Operating systems and other platforms provide a set of interfaces that expose information about objects and events to assistive technologies. Assistive technologies use these interfaces to get information about and interact with those widgets. Examples of accessibility APIs are Microsoft Active Accessibility [MSAA], Microsoft User Interface Automation [UI-AUTOMATION], MSAA with UIA Express [UIA-EXPRESS], the Mac OS X Accessibility Protocol [AXAPI], the Linux/Unix Accessibility Toolkit [ATK] and Assistive Technology Service Provider Interface [AT-SPI], and IAccessible2 [IAccessible2].

Accessibility Subtree

An accessible object in the accessibility tree and its descendants in that tree. It does not include objects which have relationships other than parent-child in that tree. For example, it does not include objects linked via aria-flowto unless those objects are also descendants in the accessibility tree.

Accessibility Tree

Tree of accessible objects that represents the structure of the user interface (UI). Each node in the accessibility tree represents an element in the UI as exposed through the accessibility API; for example, a push button, a check box, or container.

Accessible Description

An accessible description provides additional information, related to an interface element, that complements the accessible name. The accessible description might or might not be visually perceivable.

Accessible Name

The accessible name is the name of a user interface element. Each platform accessibility API provides the accessible name property. The value of the accessible name may be derived from a visible (e.g., the visible text on a button) or invisible (e.g., the text alternative that describes an icon) property of the user interface element. See related accessible description.

A simple use for the accessible name property may be illustrated by an "OK" button. The text "OK" is the accessible name. When the button receives focus, assistive technologies may concatenate the platform's role description with the accessible name. For example, a screen reader may speak "push-button OK" or "OK button". The order of concatenation and specifics of the role description (e.g., "button", "push-button", "clickable button") are determined by platform accessibility APIs or assistive technologies.

Accessible object

A node in the accessibility tree of a platform accessibility API. Accessible objects expose various states, properties, and events for use by assistive technologies. In the context of markup languages (e.g., HTML and SVG) in general, and of WAI-ARIA in particular, markup elements and their attributes are represented as accessible objects.

Assistive Technologies

Hardware and/or software that:

  • relies on services provided by a user agent to retrieve and render Web content
  • works with a user agent or web content itself through the use of APIs, and
  • provides services beyond those offered by the user agent to facilitate user interaction with web content by people with disabilities

This definition may differ from that used in other documents.

Examples of assistive technologies that are important in the context of this document include the following:

  • screen magnifiers, which are used to enlarge and improve the visual readability of rendered text and images;
  • screen readers, which are most-often used to convey information through synthesized speech or a refreshable Braille display;
  • text-to-speech software, which is used to convert text into synthetic speech;
  • speech recognition software, which is used to allow spoken control and dictation;
  • alternate input technologies (including head pointers, on-screen keyboards, single switches, and sip/puff devices), which are used to simulate the keyboard;
  • alternate pointing devices, which are used to simulate mouse pointing and clicking.

In this specification, attribute is used as it is in markup languages. Attributes are structural features added to elements to provide information about the states and properties of the object represented by the element.

Desktop focus event

Event from/to the host operating system via the accessibility API, notifying of a change of input focus.


In this specification, element is used as it is in markup languages. Elements are the structural elements in markup language that contains the data profile for objects.


A programmatic message used to communicate discrete changes in the state of an object to other objects in a computational system. User input to a web page is commonly mediated through abstract events that describe the interaction and can provide notice of changes to the state of a document object. In some programming languages, events are more commonly known as notifications.


Indicates that the element is not visible, perceivable, or interactive to any user. An element is considered hidden if it or any one of its ancestor elements is not rendered or is explicitly hidden.


Content provided for information purposes and not required for conformance. Content required for conformance is referred to as normative.

Keyboard Accessible

Accessible to the user using a keyboard or assistive technologies that mimic keyboard input, such as a sip and puff tube. References in this document relate to WCAG 2.1 Guideline 2.1: Make all functionality available from a keyboard [WCAG21].

Live Region

Live regions are perceivable regions of a web page that are typically updated as a result of an external event when user focus may be elsewhere. These regions are not always updated as a result of a user interaction. This practice has become commonplace with the growing use of Ajax. Examples of live regions include a chat log, stock ticker, or a sport scoring section that updates periodically to reflect game statistics. Since these asynchronous areas are expected to update outside the user's area of focus, assistive technologies such as screen readers have either been unaware of their existence or unable to process them for the user. WAI-ARIA has provided a collection of properties that allow the author to identify these live regions and process them: aria-live, aria-relevant, aria-atomic, and aria-busy.

Managed State

Accessibility API state that is controlled by the user agent, such as focus and selection. These are contrasted with "unmanaged states" that are typically controlled by the author. Nevertheless, authors can override some managed states, such as aria-posinset and aria-setsize. Many managed states have corresponding CSS pseudo-classes, such as :focus, and pseudo-elements, such as ::selection, that are also updated by the user agent.


Basic type of object in the DOM tree or accessibility tree. DOM nodes are further specified as Element or Text nodes, among other types. The nodes of an accessibility tree are accessible objects.


Required for conformance. By contrast, content identified as informative or "non-normative" is not required for conformance.


In the context of user interfaces, an item in the perceptual user experience, represented in markup languages by one or more elements, and rendered by user agents.

In the context of programming, the instantiation of one or more classes and interfaces which define the general characteristics of similar objects. An object in an accessibility API may represent one or more DOM objects. Accessibility APIs have defined interfaces that are distinct from DOM interfaces.
Owned Element

An 'owned element' is any DOM descendant of the element, any element specified as a child via aria-owns, or any DOM descendant of the owned child.


Attributes that are essential to the nature of a given object, or that represent a data value associated with the object. A change of a property may significantly impact the meaning or presentation of an object. Certain properties (for example, aria-multiline) are less likely to change than states, but note that the frequency of change difference is not a rule. A few properties, such as aria-activedescendant, aria-valuenow, and aria-valuetext are expected to change often. See clarification of states versus properties.


A connection between two distinct things. Relationships may be of various types to indicate which object labels another, controls another, etc.


Main indicator of type. This semantic association allows tools to present and support interaction with the object in a manner that is consistent with user expectations about other objects of that type.


The meaning of something as understood by a human, defined in a way that computers can process a representation of an object, such as elements and attributes, and reliably represent the object in a way that various humans will achieve a mutually consistent understanding of the object.


A state is a dynamic property expressing characteristics of an object that may change in response to user action or automated processes. States do not affect the essential nature of the object, but represent data associated with the object or user interaction possibilities. See clarification of states versus properties.


Presentable to users in ways they can construct an appropriate meaning. References in this document relate to WCAG 2.1 Principle 3: Information and the operation of user interface must be understandable [WCAG21].

User Agent

Any software that retrieves, renders and facilitates end user interaction with Web content. This definition may differ from that used in other documents.


A reference to a target element in the same document that has a matching ID


Discrete user interface object with which the user can interact. Widgets range from simple objects that have one value or operation (e.g., check boxes and menu items), to complex objects that contain many managed sub-objects (e.g., trees and grids).

4. Supporting Keyboard Navigation §

Enabling keyboard navigation in web applications is a necessary step toward making them accessible. User agents MUST provide a mechanism for authors to specify that any renderable element may be focusable without placing the element in a pre-defined tabbing order.

User agents MUST also provide programmatic access to all focusable elements. This allows for device-independent access, is needed to conform to the User Agent Accessibility Guidelines [UAAG10], and is vital for a successful implementation of WAI-ARIA.

Usable keyboard navigation in a rich Internet application is different from the tabbing paradigm among interactive elements, such as links and form controls, in a static document. In rich internet applications, the user tabs to significantly complex widgets, such as a menu or spreadsheet, and uses the arrow keys to navigate within the widget. The changes that WAI-ARIA introduces to keyboard navigation make this enhanced accessibility possible. In WAI-ARIA, any element can be keyboard focusable. In addition to host language mechanisms such as tabindex, aria-activedescendant provides another mechanism for keyboard operation. Most other aspects of WAI-ARIA widget development depend on keyboard navigation functioning properly.

Assistive technologies often need to set the focus. For example, voice input software, onscreen keyboards and screen readers supply their own structured navigation modes, providing additional commands for moving to elements in a page. User agents need to allow assistive technologies to set the focus. See the section titled "Handling focus changes from the Assistive Technology" for details.

4.1 Focus States and Events Table §

The following table defines the accessibility API keyboard focus states and events used in later sections of the document.

Table of accessibility APIs for focus states and events
Focusable state STATE_SYSTEM_FOCUSABLE Current state reflected in IUIAutomationElement::CurrentIsKeyboardFocusable, can be retrieved with IUIAutomationElement::GetCurrentPropertyValue method using UIA_IsKeyboardFocusablePropertyId property identifier. STATE_FOCUSABLE boolean AXFocused: the AXUIElementIsAttributeSettable method returns YES.
Focused state STATE_SYSTEM_FOCUSED Current state reflected in IUIAutomationElement::CurrentHasKeyboardFocus, can be retrieved with IUIAutomationElement::GetCurrentPropertyValue method using UIA_HasKeyboardFocusPropertyId property identifier. STATE_FOCUSED boolean AXFocused
Focus event EVENT_OBJECT_FOCUS Clients can subscribe with IUIAutomation::AddFocusChangedEventHandler using callback interface is IUIAutomationFocusChangedEventHandler object:state-changed:focused and: AXFocusedUIElementChanged

4.2 Controlling focus with tabindex §

User agents that support WAI-ARIA expand the usage of tabindex, focus, and blur to allow them on all elements. Authors may add any element such as a div, span or img to the default tab order by setting tabindex="0". In addition, any item with tabindex equal to a negative integer is focusable via script or a mouse click, but is not part of the default tab order. This is not supported in the HTML4 specification but is in compliance with HTML5 and SVG2.

The tabindex system provides one way to develop custom widgets which are keyboard accessible, from simple widgets like a slider to container widgets like a menubar, treeview or grid.


Refer to the Table of accessibility APIs for focus states and events for the rules in this section.

The user agent MUST do the following to enable accessible tabindex usage on all elements:

  1. Where tabindex equals a negative integer, set the focusable state, but do not include the element in the sequential tab order.
  2. Where tabindex="0", set the focusable state and include it in the sequential tab order.
  3. Where tabindex is greater than zero, set the focusable state, and include the element in the sequential tab order according to the value of the tabindex attribute and before any elements with tabindex either omitted or with a value of zero. See Sequential focus navigation [HTML5] for details.
  4. Expose the element.tabIndex property for every element that supports the tabindex attribute.
  5. Support the focus() and blur() methods on element interfaces that support the tabindex attribute, e.g., HTMLElement and SVGElement. This allows scripts to move the focus to the element.
  6. Fire focus, blur, DOMFocusIn, and DOMFocusOut events for any element that can receive focus.
  7. When a keydown event is cancelled, also cancel the keypress event.
  8. Expose the focusable states for any element in the accessibility tree.
  9. When any object has focus, expose the focused state. When it loses focus, remove the focused state.

4.3 Controlling focus with aria-activedescendant §

When implementing aria-activedescendant as described below, the user agent keeps the DOM focus on the container element but communicates desktop focus events and states to the assistive technology as if the active descendant has focus. User agents are not expected to validate that the active descendant is a descendant of the focused container, as noted in the Author Errors section. It is the responsibility of the user agent to ensure that keyboard events are processed at the container element that has DOM focus. Any keyboard events directed at the active descendant bubble up to the DOM element with focus, the container element, for processing.

The aria-activedescendant property may be used to enable keyboard accessibility on WAI-ARIA elements that support this attribute. It is often a more convenient way of creating container widget keyboard navigation (where the entire widget is in the tab order just once, but the user can use other keys, typically arrow keys, to navigate to descendant items of the container).

Typically, the author will use host language semantics to put the container element in the sequential tab order (e.g., tabindex="0" in HTML) and aria-activedescendant to point to the ID of the currently active descendant. The author, not the user agent, is responsible for styling the currently active descendant to show it has keyboard focus. The author cannot use :focus to style the currently active descendant since actual focus is on the container.


Refer to the Table of accessibility APIs for focus states and events for the rules in this section.

The user agent MUST do the following to implement aria-activedescendant:

  1. Implement the host language method for keyboard navigation so that the container widget may be included in the tab order. See Controlling focus with tabindex.
  2. For platforms that expose desktop focus or accessibility API focus separately from DOM focus, do not expose the focused state in the accessibility API for any element when it has DOM focus and also has aria-activedescendant which points to a valid ID.
  3. When the aria-activedescendant attribute changes on an element that currently has DOM focus, remove the focused state from the previously focused object and fire an accessibility API desktop focus event on the new active descendant. If aria-activedescendant is cleared or does not point to an element in the current document, fire a desktop focus event for the container object that had the attribute change.
  4. For any element with an ID attribute, where the element is a descendant of an element with the aria-activedescendant attribute, apply the following accessibility API states to the target to ensure the object is accessible:
    1. Focusable, if the element also has a WAI-ARIA role — because the aria-activedescendant of the container can potentially point to it. It is not absolutely necessary to check this when there is no role, because elements that would be focusable would already have the focusable state.
    2. Focused, whenever the container element sets aria-activedescendant to match the ID of this descendant and the container widget with aria-activedescendant has DOM focus.

4.4 Handling focus changes from the Assistive Technology §

Assistive technologies, such as screen readers, voice input software and on-screen keyboards, might request that the keyboard focus be moved using the following accessibility APIs:


Refer to the Table of accessibility APIs for focus states and events for the rules in this section.

When an assistive technology requests a change of focus using one of the above APIs, user agents MUST do the following:


The inability to set DOM focus to the containing element indicates an author error.

5. Mapping WAI-ARIA to Accessibility APIs §

5.1 General rules for exposing WAI-ARIA semantics §

Where supported by the platform Accessibility API, user agents expose WAI-ARIA semantics through the standard mechanisms of the desktop accessibility API. For example, for WAI-ARIA widgets, compare how the widget is exposed in a similar desktop widget. In general most WAI-ARIA widget capabilities are exposed through the role, value, Boolean states, and relations of the accessibility API.

With respect to WAI-ARIA 1.0 and 1.1, accessibility APIs operate in one direction only. User agents publish WAI-ARIA information (roles, states, and properties) via an accessibility API, and an AT can acquire that information using the same API. However, the other direction is not supported. WAI-ARIA 1.0 and 1.1 do not define mechanisms for assistive technologies to directly modify WAI-ARIA information.

The terms "exposing", "mapping", and "including" refer to the creation of accessible object nodes within the accessibility tree, and populating these objects with Accessibility API specific states and properties. The accessibility tree is a sub-set of the DOM since not every DOM element is exposed as an accessible object. Whether or not a DOM element is exposed depends on whether it has any semantic import. If it has none, no accessible object is needed, and it is excluded from the accessibility tree. For example, authors might create a nested set of DOM elements solely for visual layout. Those DOM elements are not included as they contain no semantic information. In addition, some host markup language elements have no semantics and are excluded from the accessibility tree. For example, the HTML <base> element is used to define a base URL for all relative URLs within the document. It is important to the browser, which uses the information to resolve relative URLs. But, it has no meaning to end users, is not rendered on screen by the browser, and is excluded from the accessibility tree.

A DOM element is included if it has semantic significance. As a first approximation, it is significant if it has an explicit or implicit WAI-ARIA role, state, or property, is focusable or interactive, or bears a relationship with some other element (e.g., controller-for). It is not semantically significant if it is purely presentational, or is not rendered in any medium. The details and exceptions are described below, but a procedure that provides the gist of how to build the accessibility tree is as follows:

  1. Include elements based on host language rules. These rules are documented in the host language mapping specification for those elements.
  2. Include elements with WAI-ARIA roles and other WAI-ARIA attributes.
  3. Exclude hidden elements and their descendants.
  4. Exclude presentational elements, but include their descendant elements where presentation is not inherited, and include their textual content.
    • In some cases, the presentation/none role is inherited and the descendant elements are excluded.
  5. Include, and possibly add back focusable or otherwise interactive elements.
    • For example, an element marked with both role="presentation" and tabindex="0" is focusable. The presentation role is ignored, and the element is included.
    • In instances where the host language elements support CSS display:none, the element is not focusable even if it has tabindex="0" since there is nothing rendered on screen to focus, and users cannot interact with the element. The tabindex attribute is ignored, and the element is excluded.

5.1.1 Notification of State Changes §

User agents notify assistive technologies of state and property changes as defined in Events.

5.2 Conflicts between native markup semantics and WAI-ARIA §

WAI-ARIA roles, states, and properties are intended to add semantic information when native host language elements with these semantics are not available, and are generally used on elements that have no native semantics of their own. They can also be used on elements that have similar but not identical semantics to the intended object (for instance, a nested list could be used to represent a tree structure). This method can be part of a fallback strategy for older browsers that have no WAI-ARIA implementation, or because native presentation of the repurposed element reduces the amount of style and/or script needed. Except for the cases outlined below, user agents MUST always use the WAI-ARIA semantics to define how it exposes the element to accessibility APIs, rather than using the host language semantics.

Host languages can have features that have implicit WAI-ARIA semantics corresponding to roles. When a WAI-ARIA role is provided that has a corresponding role in the accessibility API, user agents MUST use the semantic of the WAI-ARIA role for processing, not the native semantic, unless the role requires WAI-ARIA states and properties whose attributes are explicitly forbidden on the native element by the host language. Values for roles do not conflict in the same way as values for states and properties, and because authors are expected to have a valid reason to provide a WAI-ARIA role even on elements that would not normally be repurposed. For example, spin buttons are typically constructed from text fields (<input type="text">) in order to get most of the default keyboard support. But, the native role, "text field", is not correct because it does not properly communicate the additional features of a spin button. The author adds the WAI-ARIA role of spinbutton (<input type="text" role="spinbutton" ...>) so that the control is properly mapped in the accessibility API. When a WAI-ARIA role is provided that does not have a corresponding role in the accessibility API, user agents MAY expose the native semantic in addition to the WAI-ARIA role. If the host language element is overridden by a WAI-ARIA role whose semantics or structure is not equivalent to the native host language semantics or to a subclass of those semantics, then treat any required owned elements of the native role as having role presentation or none.


The above text differs slightly from the WAI-ARIA specification. The requirement for user agents to expose the WAI-ARIA role instead of the native role was intended to only apply in cases where there is a direct mapping from the WAI-ARIA role to a corresponding role in the accessibility API. The wording of the requirement is not clear in the WAI-ARIA specification, however, and has been interpreted differently by implementers. The requirement has been clarified here and an additional statement added to indicate that user agents may expose native semantics if there is not a direct mapping to a role in the accessibility API. Because there are differing implementations, authors will be advised against adding such WAI-ARIA roles to native elements that have their own semantics in the WAI-ARIA Authoring Practices Guide.

When WAI-ARIA states and properties correspond to host language features that have the same implicit WAI-ARIA semantic, it can be problematic if the values become out of sync. For example, the HTML checked attribute and the aria-checked attribute could have conflicting values. Therefore to prevent providing conflicting states and properties to assistive technologies, host languages will explicitly declare where the use of WAI-ARIA attributes on a host language element conflict with native attributes for that element. When a host language declares a WAI-ARIA attribute to be in direct semantic conflict with a native attribute for a given element, user agents MUST ignore the WAI-ARIA attribute and instead use the host language attribute with the same implicit semantic.

Host languages might also document features that cannot be overridden with WAI-ARIA (these are called "strong native semantics"). These can be features that have implicit WAI-ARIA semantics as well as features where the processing would be uncertain if the semantics were changed with WAI-ARIA. While conformance checkers might signal an error or warning when a WAI-ARIA role is used on elements with strong native semantics, user agents MUST still use the value of the semantic of the WAI-ARIA role when exposing the element to accessibility APIs.

5.3 Exposing attributes that do not directly map to accessibility API properties §

Platform accessibility APIs might have features that are not in WAI-ARIA. Likewise, WAI-ARIA exposes capabilities that are not supported by accessibility APIs at the time of publication. There typically is not a one to one relationship between all WAI-ARIA attributes and platform accessibility APIs. When WAI-ARIA roles, states and properties do not directly map to an accessibility API, and there is a mechanism in the API to expose the WAI-ARIA role, states, and properties and their values, user agents MUST expose the WAI-ARIA data using that mechanism as follows:


MSAA does not provide a mechanism for exposing attributes that do not map directly to the API and among implementers, there is no agreement on how to do it.

User agents MUST also expose the entire role string through this mechanism and MAY also expose WAI-ARIA attributes and values through this mechanism even when there is a direct mapping to an accessibility API.

Browser implementers are advised to publicly document their API methods for exposing any relevant information, so that assistive technology developers can use the API to support user features.

5.4 Role mapping §

Platform accessibility APIs traditionally have had a finite set of predefined roles that are expected by assistive technologies on that platform and only one or two roles may be exposed. In contrast, WAI-ARIA allows multiple roles to be specified as an ordered set of space-separated valid role tokens. The additional roles are fallback roles similar to the concept of specifying multiple fonts in case the first choice font type is not supported.

5.4.1 General rules §

The following rules describe how to expose WAI-ARIA roles using the accessibility API:

  1. The user agent MUST use the first token in the sequence of tokens in the role attribute value which matches the name of any non-abstract WAI-ARIA role according to rules that are specified in the Role Mapping Table below. See Conflicts between native markup semantics and WAI-ARIA for additional information. Note that when WAI-ARIA roles override host language semantics, there are no changes in the DOM, only in the accessibility tree. In the absence of author-supplied scripts, the presence of WAI-ARIA roles might not make sense. But user agents MUST map WAI-ARIA roles even in the absence of author-supplied scripts.

    The following steps will correctly identify the applicable WAI-ARIA role:

    1. Use the rules of the host language to detect that an element has a role attribute and to identify the attribute value string for it.
    2. Separate the attribute value string for that attribute into a sequence of whitespace-free substrings by separating on whitespace.
    3. Do a comparison of the substrings to all the names of the non-abstract WAI-ARIA roles. Case-sensitivity of the comparison inherits from the case-sensitivity of the host language.
    4. Use the first such substring in textual order that matches the name of a non-abstract WAI-ARIA role for the API role mapping. See the Role Mapping Table below for details.
  2. User agents MUST NOT map roles defined in the WAI-ARIA specification as "abstract" via the standard role mechanism of the accessibility API. Use the fallback procedure specified in the next rule if only an abstract role is provided. The abstract roles are:
  3. If the element does not have a role attribute, or if the role attribute contains no tokens matching the name of a non-abstract WAI-ARIA role, the user agent MUST fall back on normal processing of the base markup for the element. For example, for <table role="foo"> use the HTML table element to determine the platform accessibility API role mapping. For <input type="text" role="bar">, use the mapping for an HTML text input.
  4. When an explicit or inherited role of none or presentation is applied to an element, the user agent MUST implement the rules for the none or the presentation role defined in Accessible Rich Internet Applications (WAI-ARIA) 1.0 [WAI-ARIA].
  5. User agents MUST expose the WAI-ARIA role string if the API supports a mechanism to do so. This allows assistive technologies to do their own additional processing of roles.
    • MSAA: not supported. User agents SHOULD NOT expose a custom role in MSAA's accRole property.
    • IAccessible2: expose as an object attribute pair (xml-roles:"string")
    • UIA: expose as AriaRole property. The AriaRole property can also support secondary roles using a space as a separator.
    • ATK/AT-SPI: expose as an object attribute pair (xml-roles:"string")
  6. Platform accessibility APIs typically do not provide a vehicle to notify assistive technologies that a role has changed. Due to this and document caching, assistive technologies are unlikely to process a change in role attribute value. Authors who wish to change a role are advised by the WAI-ARIA specification to delete the associated element and its children and replace it with a new element having the appropriate role. If a role is changed, however, user agents SHOULD update the mapping in order to reflect the content in the DOM. Since assistive technologies will not know that the role has changed, user agents MAY address this error condition by treating it as removing a subtree item and inserting a new one as described in Changes to document content or node visibility.

5.4.2 Role Mapping Table §


Translators: For label text associated with the following table and its toggle buttons, see the mappingTableLabels object in the <head> section of this document.


[Note 1] User agent should return a user-presentable, localized string value for the AXRoleDescription.


[Note 2] This specification does not currently contain guidance for when user agents should fire system alert events. Some guidance may be added to the specification at a later date but it will be a recommendation (SHOULD), not a requirement (MUST).

5.5 State and Property Mapping §

This section describes how to expose WAI-ARIA states and object properties.

5.5.1 General rules §

  1. User agents MUST compute managed states VISIBLE/INVISIBLE, SHOWING/OFFSCREEN, etc. This typically is done in the same way as for ordinary elements that do not have WAI-ARIA attributes present. The FOCUSABLE/FOCUSED states may be affected by aria-activedescendant. See the rules in Controlling focus with aria-activedescendant.
  2. User agents MUST continue to expose native semantics in addition to WAI-ARIA state and property semantics except where an explicit WAI-ARIA override is allowed by the host language. For example, an HTML checkbox may have an aria-labelledby attribute but the native HTML semantics must still be exposed.
  3. User agents MUST expose additional states for certain roles as defined in the Role Mapping Table.
  4. User agents MUST compute states for the relevant WAI-ARIA attributes and map to the accessibility API as specified in the State and Property Mapping Table. To determine the relevant WAI-ARIA attributes, refer to the Definition of Roles [WAI-ARIA]]. Where the author has not provided values for required attributes, user agents SHOULD process as if the default value was provided.
  5. Some WAI-ARIA properties are not global, and are only supported on certain roles. If a non-global WAI-ARIA state or property is used where it is not supported, user agents SHOULD NOT map the given WAI-ARIA property to the platform accessibility API. For example, if aria-checked="true" is specified on <div role="grid">, it should not be exposed in MSAA implementations as STATE_SYSTEM_CHECKED.
  6. When an explicit or inherited role of none or presentation is applied to an element, the user agent MUST implement the rules for the none or the presentation role defined in Accessible Rich Internet Applications (WAI-ARIA) 1.0 [WAI-ARIA]].

5.5.2 State and Property Mapping Table § Not Mapped §

There are a number of occurrences in the table where a given state or property is declared "Not mapped". In some cases, this occurs for the default value of the state/property, and is equivalent to its absence. User agents might find it quicker to map the value than check to see if it is the default. For computational efficiency, user agents MAY expose the state or property value if doing so is equivalent to not mapping it. These cases are marked with an asterisk.

In other cases, it is mandatory that the state/property not be mapped, since exposing it implies a related affordance. An example is aria-grabbed. Its absence not only indicates that the accessible object is not grabbed, but further defines it as not grab-able. These cases are marked as "Not mapped" without an asterisk.


Translators: For label text associated with the following table and its toggle buttons, see the mappingTableLabels object in the <head> section of this document.

5.6 Special Processing Requiring Additional Computation §

5.6.1 Name and Description §

For information on how to compute an accessible name or accessible description, see the section titled Text Alternative Computation of the Accessible Name and Description Computation and API Mappings specification.

5.6.2 Widget Values §

Some WAI-ARIA roles are widgets that have a particular value within a range of values. For example progressbar, spinbutton, and slider use aria-valuemin and aria-valuemax to specify the range of valid values, aria-valuenow to specify the current value, and optionally aria-valuetext to specify a text string equivalent for the current value.

If the value is not set on a control that requires value, then user agents SHOULD return an error when the current value is requested.

When assistive technology requests the current value, user agents are not required to ensure that aria-valuenow is greater than aria-valuemin and less than aria-valuemax.

If an element has the aria-valuetext property set, but not aria-valuenow, user agents MUST expose the string value of aria-valuetext as specified in State and Property Mapping.

5.6.3 Relations §

Often in a GUI, there are relationships between the widgets that can be exposed programmatically to assistive technology. WAI-ARIA provides several relationship properties which are globally applicable to any element: aria-controls, aria-describedby, aria-flowto, aria-labelledby, aria-owns, aria-posinset, and aria-setsize. Therefore, it is not important to check the role before computing them. User agents can simply map these relations to accessibility APIs as defined in the section titled State and Property Mapping. ID Reference Error Processing §

User agents SHOULD ignore ID references that do not match the ID of another element in the same document.

It is the web author's responsibility to ensure that IDs are unique. If more than one element has the same ID, the user agent SHOULD use the first element found with the given ID. The behavior will be the same as getElementById.

If the same element is specified multiple times in a single WAI-ARIA relation, user agents SHOULD return multiple pointers to the same object.

aria-activedescendant is defined as referencing only a single ID reference. Any aria-activedescendant value that does not match an existing IDREF exactly is an author error and will not match any element in the DOM. Reverse Relations §

A reverse relation exists when an element's ID is referenced by a property in another element. For APIs that support reverse relations, user agents MUST use the mapping defined in the following table when an element's ID is referenced by a relation property of another element and the referenced element is in the accessibility tree according to the rules in General rules for exposing WAI-ARIA semantics. All WAI-ARIA references must point to an element that is exposed as an accessible object in the accessibility tree. When the referenced object is not exposed in the accessibility tree (e.g. because it is hidden), the reference is null. aria-labelledby and aria-described by have an additional feature, which allows them to pull a flattened string from the referenced element to populate the name or description fields of the accessibility API. This feature is described in the Name and Description section.

Reverse relation mapping table
WAI-ARIA Relation IAccessible2 ATK/AT-SPI
aria-controls references the element's ID IA2_RELATION_CONTROLLED_BY RELATION_CONTROLLED_BY
aria-describedby references the element's ID IA2_RELATION_DESCRIPTION_FOR RELATION_DESCRIPTION_FOR
aria-details references the element's ID IA2_RELATION_DETAILS_FOR RELATION_DETAILS_FOR
aria-errormessage references the element's ID IA2_RELATION_ERROR_FOR RELATION_ERROR_FOR
aria-flowto references the element's ID IA2_RELATION_FLOW_FROM RELATION_FLOWS_FROM
aria-labelledby references the element's ID IA2_RELATION_LABEL_FOR RELATION_LABEL_FOR

Special case: If both aria-labelledby and HTML <label for= … > are used, the user agent MUST use the WAI-ARIA relation and MUST ignore the HTML label relation.

Note that aria-describedby may reference structured or interactive information where users would want to be able to navigate to different sections of content. User agents MAY provide a way for the user to navigate to structured information referenced by aria-describedby and assistive technology SHOULD provide such a method. Implied reverse relations §

In addition to the explicit relations defined by WAI-ARIA properties, reverse relations are implied in two other situations: elements with role="treeitem" where the ancestor does not have an aria-owns property and descendants of elements with aria-atomic property.

In the case of role="treeitem", when aria-owns is not used, user agents SHOULD do the following where reverse relations are supported by the API:

  • If the current treeitem uses aria-level, then walk backwards in the tree until a treeitem is found with a lower aria-level, then set RELATION_NODE_CHILD_OF to that element. If the top of the tree is reached, then set RELATION_NODE_CHILD_OF to the tree element itself.
  • If the parent of the treeitem has a role of group, then walk backwards from the group until an element with a role of treeitem is found, then set RELATION_NODE_CHILD_OF to that element.

In the case of aria-atomic, where reverse relations are supported by the API:

  • User agents SHOULD check the chain of ancestor elements for aria-atomic="true". If found, user agents SHOULD set the RELATION_MEMBER_OF relation to point to the ancestor that sets aria-atomic="true".

5.6.4 Group Position §

aria-level, aria-posinset, and aria-setsize are all 1-based. When the property is not present or is "0", it indicates the property is not computed or not supported. If any of these properties are specified by the author as either "0" or a negative number, user agents SHOULD use "1" instead.

If aria-level is not provided or inherited for an element of role treeitem, user agents implementing IAccessible2 or ATK/AT-SPI MUST compute it by following the explicit or computed RELATION_NODE_CHILD_OF relations.

If aria-posinset and aria-setsize are not provided, user agents MUST compute them as follows:

  • for role="treeitem", walk the tree backward and forward until the explicit or computed level becomes less than the current item's level. Count items only if they are at the same level as the current item.
  • Otherwise, if the role supports aria-posinset and aria-setsize, process the parent (DOM parent or parent defined by aria-owns), counting items that have the same role.
  • Because these value are 1-based, include the current item in the computation. For aria-posinset, include the current item and other group items if they are before the current item in the DOM. For aria-setsize, add to that the number of items in the same group after the current item in the DOM.

If the author provides one or more of aria-setsize and aria-posinset, it is the author's responsibility to supply them for all elements in the set. User agent correction of missing values in this case is not defined.

MSAA/IAccessible2 API mappings involve an additional function, groupPosition() [IAccessible2], when aria-level, aria-posinset, and/or aria-setsize are present on an element, or are computed by the user agent. When this occurs:

  • aria-level is exposed in the groupLevel parameter of groupPosition(),
  • aria-setsize is exposed in the similarItemsInGroup parameter, and
  • aria-posinset is exposed in the positionInGroup parameter.

5.7 Actions §

As part of mapping roles to accessible objects as defined in Role Mapping, users agents expose a default action on the object.


Authors will need to create handlers for those click events that update WAI-ARIA states and properties in the DOM accordingly, so that those updated states can be populated by the user agent in the Accessibility API.

5.8 Events §

User agents fire events for user actions, WAI-ARIA state changes, changes to document content or node visibility, changes in selection and operation of menus as defined in the following sections.

5.8.1 State and Property Change Events §

User agents MUST notify assistive technology of state changes as defined in the table below, SHOULD notify assistive technology of property changes if the accessibility API defines a change event for the property, and SHOULD NOT notify assistive technology of property changes if the accessibility API does not define a change event for the property. For example, IAccessible2 defines an event to be used when aria-activedescendant changes. WAI-ARIA properties that are expected to change include aria-activedescendant, aria-valuenow, and aria-valuetext.


In some APIs, AT will only be notified of events to which it has subscribed.

For simplicity and performance the user agent MAY trim out change events for state or property changes that assistive technologies typically ignore, such as events that are happening in a window that does not currently have focus.


Translators: For label text associated with the following table and its toggle buttons, see the mappingTableLabels object in the <head> section of this document.

5.8.2 Changes to document content or node visibility §

Processing document changes is important regardless of WAI-ARIA. The events described in the table below are used by user agents to inform AT of changes to the DOM via the accessibility tree. For the purposes of conformance with this standard, user agents MUST implement the behavior described in this section whenever WAI-ARIA attributes are applied to dynamic content on a Web page.

Table of document change scenarios and events to be fired in each API
Scenario MSAA + IAccessible2 event UIA event ATK/AT-SPI event AX API Notification
When text is removed IA2_EVENT_TEXT_REMOVED EVENT_OBJECT_LIVEREGIONCHANGED text_changed::delete If in a live region, AXLiveRegionChanged.
If in aria-errormessage, AXValidationErrorChanged.
When text is inserted IA2_EVENT_TEXT_INSERTED EVENT_OBJECT_LIVEREGIONCHANGED text_changed::insert If in a live region, AXLiveRegionChanged.
If in aria-errormessage, AXValidationErrorChanged.
When text is changed IA2_EVENT_TEXT_REMOVE and IA2_EVENT_TEXT_INSERTED EVENT_OBJECT_LIVEREGIONCHANGED text_changed::delete and text_changed::insert If in a live region, AXLiveRegionChanged.
If in aria-errormessage, AXValidationErrorChanged.

Fire these events for node changes where the node in question is an element and has an accessible object:

Table of document change scenarios and events to be fired in each API
Scenario MSAA Microsoft UIA event ATK/AT-SPI event AX API Notification
When an accessibility subtree is hidden EVENT_OBJECT_HIDE
The MSAA event called EVENT_OBJECT_DESTROY is not used because this has a history of stability issues and assistive technology avoids it. In any case, from the user's point of view, there is no difference between something that is hidden or destroyed.
AutomationElement..::.StructureChangedEvent children_changed::remove


If in a live region, AXLiveRegionChanged

When an accessibility subtree is removed EVENT_OBJECT_REORDER
The MSAA event called EVENT_OBJECT_DESTROY is not used because this has a history of stability issues and assistive technology avoids it. In any case, from the user's point of view, there is no difference between something that is hidden or destroyed.
AutomationElement..::.StructureChangedEvent children_changed::remove


If in a live region, AXLiveRegionChanged

When an accessibility subtree is shown EVENT_OBJECT_SHOW   children_changed::add


If in a live region, AXLiveRegionChanged

When an accessibility subtree is inserted EVENT_OBJECT_REORDER   children_changed::add


If in a live region, AXLiveRegionChanged

When an accessibility subtree is moved Treat it as a removal from one place and insertion in another Treat it as a removal from one place and insertion in another Treat it as a removal from one place and insertion in another

AXUIElementDestroyed/ AXUIElementCreated

If in a live region, AXLiveRegionChanged

When an accessibility subtree is changed (e.g. replaceNode) Treat it as a removal and insertion Treat it as a removal and insertion Treat it as a removal and insertion

AXUIElementDestroyed/ AXUIElementCreated

If in a live region, AXLiveRegionChanged

In some cases, node changes may occur where the node is not an element or has no accessible object. For example, a numbered list bullet ("12.") may have a node in the accessibility tree but not in the DOM tree. For text within a paragraph marked in HTML as <strong>, the <strong> element has a node in the DOM tree but may not have one in the accessibility tree. The text itself will of course be in the accessibility tree along with the identification of the range of text that is formatted as strong. If any of the changes described in the table above occur on such a node, user agents SHOULD compute and fire relevant text change events as described above.

User agents SHOULD ensure that an assistive technology, running in process can receive notification of a node being removed prior to removal. This allows an assistive technology, such as a screen reader, to refer back to the corresponding DOM node being deleted. This is important for live regions where removals are important. For example, a screen reader would want to notify a user that another user has left a chat room. The event in MSAA would be EVENT_OBJECT_HIDE. For ATK/AT-SPI this would be children_changed::remove. And in macOS, the event is AXLiveRegionChanged. This also requires the user agent to provide a unique ID in the accessibility API notification identifying the unique node being removed.

When firing any of the above-mentioned change events, it is very useful to provide information about whether the change was caused by user input (as opposed to a timeout initiated from the page load, etc.). This allows the assistive technology to have different rules for presenting changes from the real world as opposed to from user action. Mouse hovers are not considered explicit user input because they can occur from accidental bumps of the mouse.

To expose whether a change occurred from user input:

  • In ATK/AT-SPI this can be provided by appending the string ":system" to the event name when the user did not cause the change.
  • In IAccessible2, which screen readers typically access in process on the same thread, the best practice is to expose the object attribute event-from-user-input:true on the accessible object for the event, if the user caused the change.

Exposing additional useful information about the context of the change:

  • In ATK/AT-SPI and IAccessible2, the RELATION_MEMBER_OF relation on the accessible event's target accessible object SHOULD point to any ancestor with aria-atomic="true" (if any).
  • In ATK/AT-SPI and IAccessible2, the container-live, container-relevant, container-busy, container-atomic object attributes SHOULD be exposed on the accessible event object, providing the computed value for the related WAI-ARIA properties. The computed value is the value of the closest ancestor. It is recommended to not expose the object attribute if the default value is used.

Additional MSAA events may be necessary:

  • If something changes in an ancestor with a mapped MSAA role of ROLE_SYSTEM_ALERT, then an EVENT_SYSTEM_ALERT event SHOULD be fired for the alert. The alert role has an implied value of "assertive" for the aria-live property.
  • Menu events may need to be fired. See Special Events for Menus.

5.8.3 Selection §

There are two cases for selection:

  • Single selection
  • Multiple selection

In the single selection case, selection follows focus (see the section "Focus States and Events Table" for information about focus events). User agents MUST fire the following events when aria-selected changes:

Single selection events
Scenario MSAA Microsoft UIA ATK/AT-SPI AX API
Focus change EVENT_OBJECT_SELECTION and EVENT_OBJECT_STATECHANGE on newly focused item. UIA_SelectionItem_ElementSelectedEventId on the newly focused element.

If on a gridcell, row, option, or tab, fire UIA_SelectionItem_ElementSelectedEventId.

  • object:selection-changed on the current container,
  • object:state-changed:selected on the descendant accessible object whose selection has changed:
    • detail1 = 1 for the descendant which just became selected.
    • detail1 = 0 for the descendant which just became unselected.

The multiple selection case occurs when aria-multiselectable="true" on an element with a role that supports that property. User agents MUST fire the following events when aria-selected changes on a descendant, as follows:

The multiple selection case occurs when aria-multiselectable="true" on an element with a role that supports that property. There are several important aspects:

  1. In Microsoft UIA, the Selection and SelectionItem Control Patterns expose the selection availability, state, and methods.
  2. User agents MUST fire the following events when aria-selected changes on a descendant, as follows:
Multiple selection events
Scenario MSAA Microsoft UIA ATK/AT-SPI AX API
Toggle aria-selected EVENT_OBJECT_SELECTIONADD/EVENT_OBJECT_SELECTIONREMOVE on the item. SelectionItem Control Pattern:UIA_SelectionItem_ElementAddedToSelectionEventId or UIA_SelectionItem_ElementRemovedFromSelectionEventId on the current container.
  • object:selection-changed on the current container,
  • object:state-changed:selected on any descendant accessible object whose selection has changed:
    • detail1 = 1 for any descendant which just became selected.
    • detail1 = 0 for any descendant which just became unselected.
Selection follows focus EVENT_OBJECT_SELECTION and EVENT_OBJECT_STATECHANGE on newly focused item. FocusChangedEvent should be fired but individual selection event may not happen, to avoid redundancy.
  • object:selection-changed on the current container,
  • object:state-changed:selected on any descendant accessible object whose selection has changed:
Select or deselect many items at once User agent MAY fire an EVENT_OBJECT_SELECTIONWITHIN. If this event is fired the other events noted above MAY be trimmed out for performance. For each element selected or deselected, fire SelectionItem Control Pattern: UIA_SelectionItem_ElementAddedToSelectionEventId or UIA_SelectionItem_ElementRemovedFromSelectionEventId on the current container. User agents MAY choose to fire the Selection Control Pattern Invalidated event, which indicates that the selection in a container has changed significantly and requires sending more addition and removal events than the InvalidateLimit constant permits.
  • the user agent MAY fire a single object:selection-changed event on the container, vs. multiple events, for performance,
  • object:state-changed:selected on any descendant accessible object whose selection has changed:

5.8.4 Special Events for Menus §

Some APIs, provide special events whenever a menu is opened or closed. User agents SHOULD provide the events as described in the table below. If provided, because menus can be made visible or hidden using a variety of techniques, a user agent MUST ensure that the events are nested and symmetrical.

Frequently, a menubar is used to organize a hierarchy of menus. In those cases, the menubar MUST be a DOM parent of the associated menuitems, or one defined by aria-owns. In other cases, no menubar is involved; for example, when the menu is associated with a toolbar button, or is a context menu. Nonetheless the relevant menu events are provided as described in the following table.

Menu events
Scenario MSAA Microsoft UIA AX API

Menubar is currently not active, and user moves focus to the menubar from elsewhere thereby activating it. As a result, a menuitem in the menubar is focused.

Activate the menubar and fire EVENT_SYSTEM_MENUSTART on the accessible object for the menubar. MenuModeStartEvent on the accessible object for the menu. AXMenuOpenedNotification

Focus a menuitem while menubar is activated, or focus a menuitem in a menu.

EVENT_OBJECT_FOCUS AutomationFocusChangedEvent AXMenuItemSelectedNotification

Menu popup made visible (menu is opened).

Should only be fired once until the menu is closed and opened again.

EVENT_SYSTEM_MENUPOPUPSTART MenuOpenedEvent, then a focus event on a menuitem. AXMenuOpenedNotification
Menu popup hidden (menu is closed). EVENT_SYSTEM_MENUPOPUPEND once only for accessible menu object and only if EVENT_SYSTEM_MENUPOPUPSTART was fired for it. MenuClosedEvent AXMenuClosedNotification
Any open menus are closed including sub-menus, and user moves focus away from the menubar; menubar is deactivated. EVENT_SYSTEM_MENUEND on the menubar and deactivate the menubar. MenuClosedEvent, then MenuModeEndEvent AXMenuClosedNotification

6. Special Document Handling Procedures §

6.1 CSS Selectors §


This section might be removed in a future version.

Support for attribute selectors MUST include WAI-ARIA attributes. For example, .fooMenuItem[aria-haspop="true"] would select all elements with class fooMenuItem, and WAI-ARIA property aria-haspopup with value of true. The presentation MUST be updated for dynamic changes to WAI-ARIA attributes. This allows authors to match styling with WAI-ARIA semantics.

6.2 Author Errors §

In general, user agents do not do much validation of WAI-ARIA properties. User agents MAY do some minor validation on request, such as making sure valid IDs are specified for WAI-ARIA relations, and enforcing things like aria-posinset being within 1 and aria-setsize, inclusive. User agents are not responsible for logical validation, such as the following:

  1. Circular references created by relations, such as specifying that two elements own each other.
  2. Correct usage with regard to DOM tree structure, such as an aria-activedescendant being a DOM-descendant of the element with the relation.
  3. Elements with WAI-ARIA roles correctly implement the behavior of the specified role. For example, user agents do not verify that an element with a role of checkbox actually behaves like a checkbox.
  4. Elements that do not correctly observe required child / parent role relationships or that appear elsewhere than in their required parent.
  5. Determining whether aria-activedescendant actually points to a descendant or another owned element.
  6. Determining implicit values of aria-setsize and aria-posinset when they are specified on some but not all the elements of the set.

If the author specifies a non-numeric value for a decimal or integer value type, the user agent SHOULD do the following:

If a WAI-ARIA property contains an unknown or disallowed value, the user agent SHOULD expose to platform accessibility APIs as follows:


In UIA, the user agent might leave the corresponding property set to "unsupported."

User agents MUST NOT expose WAI-ARIA attributes that reference unresolved IDs. For example:

User Agents MUST NOT expose aria-roledescription when:

If a required WAI-ARIA attribute for a given role is missing, user agents SHOULD process the attribute as if the values given in the following table were provided.

Default values for missing required attributes
WAI-ARIA role Required Attribute Default value
checkbox aria-checked See aria-checked="false" in State and Property mapping table.
menuitem aria-haspopup See aria-haspopup="false" in State and Property mapping table.
menuitemcheckbox aria-checked See aria-checked="false" in State and Property mapping table.
menuitemradio aria-checked See aria-checked="false" in State and Property mapping table.
radio aria-checked See aria-checked="false" in State and Property mapping table.
switch aria-checked See aria-checked="false" in State and Property mapping table.
combobox aria-expanded See aria-expanded="false" in State and Property mapping table.
combobox aria-haspopup See aria-haspopup="listbox" in State and Property mapping table.
heading aria-level 2
scrollbar aria-controls no mapping
scrollbar aria-orientation See aria-orientation="vertical" in State and Property mapping table.
scrollbar aria-valuemax 100
scrollbar aria-valuemin 0
scrollbar aria-valuenow (aria-valuemax - aria-valuemin) / 2
slider aria-valuemax 100
slider aria-valuemin 0
slider aria-valuenow (aria-valuemax - aria-valuemin) / 2
spinbutton aria-valuemax A value indicating that the spinbutton has no upper bound (Accessibility API dependent).
spinbutton aria-valuemin A value indicating that the spinbutton has no lower bound (Accessibility API dependent).
spinbutton aria-valuenow 0

A. Appendices §

A.1 Change Log §

A.1.1 Substantive changes since the last public working draft §

  • 06-Jun-2018: Update UIA mappings for aria-placeholder.
  • 04-Jun-2018: Add mappings for blockquote, caption, and paragraph roles.
  • 05-Mar-2018: Add mention of AXTitle for exposing rendered labels for AXAPI.
  • 05-Mar-2018: Add events for aria-label, aria-labelledby, and aria-describedby.

A.1.2 Substantive changes since the Core Accessibility API Mappings 1.1 Recommendation §

A.1.3 Acknowledgments §

This section is non-normative.

The following people contributed to the development of this document.

A.1.3.1 Participants active in the ARIA WG at the time of publication §
  • Ann Abbott (Invited Expert)
  • Irfan Ali (Educational Testing Service)
  • Amelia Bellamy-Royds (Invited Expert)
  • Michiel Bijl (Invited Expert)
  • David Bolter (Mozilla Foundation)
  • Bogdan Brinza (Microsoft Corporation)
  • Shari Butler (Pearson plc)
  • Thaddeus Cambron (Invited Expert)
  • Michael Cooper (W3C Staff)
  • James Craig (Apple Inc.)
  • Joanmarie Diggs (Igalia)
  • John Foliot (Deque Systems, Inc.)
  • Bryan Garaventa (SSB BART Group)
  • Matt Garrish (DAISY Consortium)
  • Becky Gibson (Invited Expert)
  • Glen Gordon (The Paciello Group, LLC)
  • Jon Gunderson (University of Illinois at Urbana-Champaign)
  • Matthew King (Facebook)
  • JaEun Jemma Ku (University of Illinois at Urbana-Champaign)
  • Charles LaPierre (Benetech)
  • Aaron Leventhal (Google, Inc.)
  • Dominic Mazzoni (Google, Inc.)
  • Shane McCarron (Invited Expert, Aptest)
  • Jan McSorley (Pearson plc)
  • James Nurthen (Oracle Corporation)
  • Ian Pouncey (The Paciello Group, LLC)
  • Ruoxi Ran (W3C Staff)
  • Janina Sajka (Invited Expert, The Linux Foundation)
  • Stefan Schnabel (SAP SE)
  • Lisa Seeman-Kestenbaum (Invited Expert)
  • Tzviya Siegman (Wiley)
  • Alexander Surkov (Mozilla Foundation)
  • Job van Achterberg (Invited Expert)
  • Evan Yamanishi (W. W. Norton)
  • Jason White (Educational Testing Service)
A.1.3.2 Other ARIA contributors, commenters, and previously active participants §
  • Shadi Abou-Zahra (W3C)
  • Jim Allan (TSB)
  • Jonny Axelsson (Opera Software)
  • David Baron (Mozilla Foundation)
  • Art Barstow (Nokia Corporation)
  • Simon Bates
  • Christy Blew (University of Illinois at Urbana-Champaign)
  • Chris Blouch (AOL)
  • Judy Brewer (W3C/MIT)
  • Mark Birbeck (Sidewinder Labs)
  • Sally Cain (Royal National Institute of Blind People (RNIB))
  • Gerardo Capiel (Benetech)
  • Ben Caldwell (Trace)
  • Sofia Celic-Li
  • Jaesik Chang (Samsung Electronics Co., Ltd.)
  • Alex Qiang Chen (University of Manchester)
  • Charles Chen (Google, Inc.)
  • Christian Cohrs
  • Deborah Dahl
  • Erik Dahlström (Opera Software)
  • Dimitar Denev (Frauenhofer Gesellschaft)
  • Micah Dubinko (Invited Expert)
  • Mandana Eibegger
  • Beth Epperson (Websense)
  • Fred Esch (IBM Corporation)
  • Donald Evans (AOL)
  • Steve Faulkner (The Paciello Group, LLC)
  • Chris Fleizach (Apple Inc.)
  • Kelly Ford (Microsoft Corporation)
  • Geoff Freed (Invited Expert, NCAM)
  • Christopher Gallelo (Microsoft Corporation)
  • Billy Gregory (The Paciello Group, LLC)
  • Karl Groves (The Paciello Group, LLC)
  • Birkir Gunnarsson (Deque Systems, Inc.)
  • Kentarou Fukuda (IBM Corporation)
  • Bryan Garaventa
  • Guido Geloso
  • Ali Ghassemi
  • Alfred S. Gilman
  • Andres Gonzalez (Adobe Systems Inc.)
  • Scott González (JQuery Foundation)
  • James Graham
  • Georgios Grigoriadis (SAP AG)
  • Jeff Grimes (Oracle)
  • Loretta Guarino Reid (Google, Inc.)
  • Markus Gylling (DAISY Consortium)
  • Markku Hakkinen (Educational Testing Service)
  • Katie Haritos-Shea (Knowbility)
  • Barbara Hartel
  • James Hawkins (Google, Inc.)
  • Benjamin Hawkes-Lewis
  • Sean Hayes (Microsoft Corporation)
  • Mona Heath (University of Illinois at Urbana-Champaign)
  • Jan Heck
  • Shawn Henry
  • Tina Homboe
  • Nicholas Hoyt
  • John Hrvatin (Microsoft Corporation)
  • Takahiro Inada
  • Masayasu Ishikawa (W3C)
  • Jim Jewitt
  • Kenny Johar (Microsoft Corporation)
  • Shilpi Kapoor (BarrierBreak Technologies)
  • Masahiko Kaneko (Microsoft Corporation)
  • Marjolein Katsma
  • Susann Keohane (IBM Corporation)
  • George Kerscher (International Digital Publishing Forum)
  • Jason Kiss (Department of Internal Affairs, New Zealand Government)
  • Todd Kloots
  • Jamie Knight (British Broadcasting Corporation)
  • Johannes Koch
  • Gerard K. Cohen
  • Sam Kuper
  • Earl Johnson (Sun)
  • Jael Kurz
  • Rajesh Lal (Nokia Corporation)
  • Diego La Monica (International Webmasters Association / HTML Writers Guild (IWA-HWG))
  • Gez Lemon (International Webmasters Association / HTML Writers Guild (IWA-HWG))
  • Alex Li (SAP)
  • Chris Lilley
  • Thomas Logan (HiSoftware Inc.)
  • Brian Loh
  • William Loughborough (Invited Expert)
  • Linda Mao (Microsoft)
  • David MacDonald (Invited Expert, CanAdapt Solutions Inc.)
  • Carolyn MacLeod
  • Anders Markussen (Opera Software)
  • Krzysztof Maczyński
  • Matthew May (Adobe Systems Inc.)
  • Mark McCarthy
  • Charles McCathie Nevile (Yandex)
  • Heather Migliorisi (Invited Expert)
  • Mary Jo Mueller (IBM Corporation)
  • Alexandre Morgaut (4D)
  • Ann Navarro (Invited Expert)
  • Joshue O Connor (Invited Expert, CFIT)
  • Artur Ortega (Microsoft Corporation)
  • Sailesh Panchang (Deque)
  • Lisa Pappas (Society for Technical Communication (STC))
  • Marta Pawlowlska (Samsung Electronics Co., Ltd.)
  • Dave Pawson (RNIB)
  • Steven Pemberton (CWI Amsterdam)
  • Simon Pieters (Opera Software)
  • Jean-Bernard Piot (4D)
  • David Poehlman, Simon Pieters (Opera Software)
  • Sarah Pulis (Media Access Australia)
  • T.V. Raman (Google, Inc.)
  • Jan Richards
  • Gregory Rosmaita (Invited Expert)
  • Tony Ross (Microsoft Corporation)
  • Alex Russell (Dojo Foundation) (
  • Mark Sadecki (Invited Expert)
  • Mario Sánchez Prada (Samsung Electronics Co., Ltd. and Gnome Foundation)
  • Martin Schaus (SAP AG)
  • Doug Schepers (W3C)
  • Cynthia Shelly (Microsoft Corporation)
  • Joseph Scheuhammer (Invited Expert, Inclusive Design Research Centre, OCAD University)
  • Matthias Schmitt
  • Richard Schwerdtfeger (IBM, Knowbility)
  • Marc Silbey (Microsoft Corporation)
  • Leif Halvard Sili
  • Henri Sivonen (Mozilla)
  • Michael Smith (W3C)
  • Andi Snow-Weaver (IBM Corporation)
  • Ville Skyttä
  • Henny Swan (BBC)
  • Neil Soiffer (Design Science)
  • Vitaly Sourikov
  • Mike Squillace (IBM)
  • Maciej Stachowiak (Apple Inc.)
  • Christophe Strobbe
  • Suzanne Taylor (Pearson plc)
  • Terrill Thompson
  • David Todd
  • Gregg Vanderheiden (Invited Expert, Trace)
  • Anne van Kesteren
  • Léonie Watson (The Paciello Group, LLC)
  • Wen He (Tencent)
  • Wu Wei (W3C / RITT)
  • Ryan Williams (Oracle)
  • Tom Wlodkowski
  • Sam White (Apple Inc.)
  • Marco Zehe (Mozilla Foundation)
  • Gottfried Zimmermann (Invited Expert, Access Technologies Group)
A.1.3.3 Enabling funders §

This publication has been funded in part with U.S. Federal funds from the Department of Education, National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR), initially under contract number ED-OSE-10-C-0067 and currently under contract number HHSP23301500054C. The content of this publication does not necessarily reflect the views or policies of the U.S. Department of Education, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

B. References §

B.1 Normative references §

Key words for use in RFCs to Indicate Requirement Levels. S. Bradner. IETF. March 1997. Best Current Practice. URL:
Accessible Rich Internet Applications (WAI-ARIA) 1.1. Joanmarie Diggs; Shane McCarron; Michael Cooper; Richard Schwerdtfeger; James Craig. W3C. 14 December 2017. W3C Recommendation. URL:
WAI-ARIA 1.0 User Agent Implementation Guide. Joseph Scheuhammer; Michael Cooper. W3C. 20 March 2014. W3C Recommendation. URL:

B.2 Informative references §

Assistive Technology Service Provider Interface. The GNOME Project. URL:
ATK - Accessibility Toolkit. The GNOME Project. URL:
The NSAccessibility Protocol for macOS. Apple, Inc. URL:
Core Accessibility API Mappings 1.1. Joanmarie Diggs; Joseph Scheuhammer; Richard Schwerdtfeger; Michael Cooper; Andi Snow-Weaver; Aaron Leventhal. W3C. 14 December 2017. W3C Recommendation. URL:
HTML5. Ian Hickson; Robin Berjon; Steve Faulkner; Travis Leithead; Erika Doyle Navara; Theresa O'Connor; Silvia Pfeiffer. W3C. 27 March 2018. W3C Recommendation. URL:
IAccessible2. Linux Foundation. URL:
Microsoft Active Accessibility (MSAA) 2.0. Microsoft Corporation. URL:
User Agent Accessibility Guidelines 1.0. Ian Jacobs; Jon Gunderson; Eric Hansen. W3C. 17 December 2002. W3C Recommendation. URL:
UI Automation. Microsoft Corporation. URL:
The IAccessibleEx Interface. Microsoft Corporation. URL:
WAI-ARIA Authoring Practices 1.1. Matthew King; James Nurthen; Michiel Bijl; Michael Cooper; Joseph Scheuhammer; Lisa Pappas; Richard Schwerdtfeger. W3C. 26 July 2018. W3C Note. URL:
Roadmap for Accessible Rich Internet Applications (WAI-ARIA Roadmap). Richard Schwerdtfeger. W3C. 4 February 2008. W3C Working Draft. URL:
Web Content Accessibility Guidelines (WCAG) 2.1. Andrew Kirkpatrick; Joshue O Connor; Alastair Campbell; Michael Cooper. W3C. 5 June 2018. W3C Recommendation. URL: