The <caption> Element

Technical Summary

Definition / Content model

Definition and conformance conditions for caption

Item	Specification	Condition / scope
Meaning	Represents the title of its parent when that parent is a table	Participates in the table model
Placement	First child element of table	Conforming markup has at most one caption per table; caption is optional in the table content model
Content model	Flow content	No descendant table elements
Content attributes	Global attributes	The obsolete align attribute is non-conforming; use CSS for position and alignment
DOM interface	HTMLTableCaptionElement	Inherits from HTMLElement; the obsolete align IDL attribute also remains defined

A title helps readers understand what the table covers. When a table is the only content in a figure other than its figcaption, the HTML Standard advises omitting caption in favor of figcaption. This recommendation is scoped to that stated figure structure; it should not be generalized to figures containing other content.

Legacy HTML includes the caption align attribute and HTMLTableCaptionElement.align IDL attribute, but both are obsolete. New markup should use CSS caption-side and text-align.

Table association and DOM interface

caption is a child of its table, rather than text that is merely nearby. The table.caption getter returns the first caption child, or null when there is none. The setter sets a caption on the table. createCaption() creates one if needed, and deleteCaption() removes it.

const table = document.querySelector("table");
const caption = table.caption; // first caption, or null

const ensuredCaption = table.createCaption();
ensuredCaption.textContent = "Visitors by month";

table.deleteCaption();

A caption placed elsewhere, such as inside a tr, is not returned by that table's caption getter. DOM-setter repair and non-conforming markup with multiple captions are separate from ordinary conforming markup.

Rendered position and table model

HTML placement and visual placement are separate. CSS Tables defines caption-side to position the caption box above or below the table grid. Setting caption-side: bottom does not change the caption's position as the first child in the HTML or its DOM relationship with that table.

caption {
  caption-side: bottom;
  text-align: start;
}

Do not assume that visual order, DOM order, and assistive-technology speech order are always identical. Measure each surface separately.

Accessible role, relation, and table name

Caption and table relationships defined by HTML-AAM

Surface	Mapping	Condition / scope
Caption role	Maps to the WAI-ARIA caption role	Platform role values are then resolved through each API's WAI-ARIA mapping
Parent relation	Defines a label relation from caption to its parent table for platform APIs	IAccessible2, ATK, and AX use a label-for relation; UIA exposes the caption through the parent table's LabeledBy property
Table accessible name	After computing aria-label / aria-labelledby, uses the text subtree of the first caption child	If the name is still empty, falls back to the title attribute. A caption hidden from the accessibility tree does not provide the name

An accessible name identifies the table to assistive technology. The visible caption is often the source, but an author-provided ARIA name has priority. HTML-AAM's mapping does not by itself specify exactly when or how a screen reader announces that name on table entry.

Fact / Evidence

Track the HTML definition, accessibility mapping, and CSS position as separate claims. Browser observations, WPT, and assistive-technology results are recorded under Implementation Evidence.

Evidence

1. HTML Standard: The caption element — meaning, context, content model, HTMLTableCaptionElement
2. HTML Standard: The table element — table content model and DOM API including caption
3. HTML Accessibility API Mappings: caption — caption role and platform relations to the table
4. HTML-AAM: table accessible name — ARIA, caption, and title order; hidden-caption condition
5. CSS Table Module Level 3: caption-side — visual caption placement
6. HTML Standard: Obsolete features — obsolete caption align content and IDL attributes
7. WPT: caption_001.html — caption getter / setter, null, and child relationship
8. W3C WAI: H39 — associating a table caption using the caption element

Implementation Evidence

These are observations from the selected fixture, WPTs, Chrome, and NVDA. They do not establish conformance or identical output in other environments. AAM platform mappings, the browser accessibility tree, Windows UI Automation, and NVDA speech output are recorded as separate surfaces.

Only selected WPT cases are run. Browser accessibility-tree and Windows UI Automation values are not a cross-platform AAM conformance verdict. NVDA speech is recorded only for combinations where its actual output was captured.

Coverage / Open Issues

• ReviewedHTML meaning, first-child condition, content model, DOM API, CSS caption-side, HTML-AAM role and table-name computation
• OpenHTML parsing and DOM-setter behavior in all browsers, invalid markup with multiple captions, and caption descendants in the table model
• OpenWPTs beyond the selected cases, other browsers, macOS AX / Linux ATK, and all Windows UIA properties
• OpenNVDA reading order, table navigation, JA / EN voice variation, other assistive technologies, and Expert-gap Review

This is initial coverage. The observations recorded here do not generalize to every table structure, browser, platform API, or assistive technology.

Related surfaces

For a beginner-friendly example and placement guidance, see the caption element page in Yugien. The table grid and parser recovery are covered by the table Topic; header and data-cell relationships are covered by the th Topic and td Topic.