CSS Writing Modes Level 3

W3C Proposed Recommendation,

This version:
Latest published version:
Editor's Draft:
Previous Versions:
Test Suite:
Issue Tracking:
GitHub Issues
Elika J. Etemad / fantasai (Invited Expert)
Former Editors:
(Antenna House)
Suggest an Edit for this Spec:
GitHub Editor


CSS Writing Modes Level 3 defines CSS support for various writing modes and their combinations, including left-to-right and right-to-left text ordering as well as horizontal and vertical orientations.

CSS is a language for describing the rendering of structured documents (such as HTML and XML) on screen, on paper, etc.

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 https://www.w3.org/TR/.

This document was produced by the CSS Working Group as a Proposed Recommendation.

This document is intended to become a W3C Recommendation. This document will remain a Proposed Recommendation at least until in order to ensure the opportunity for review from the W3C Membership and other interested parties.

GitHub Issues are preferred for discussion of this specification. When filing an issue, please put the text “css-writing-modes” in the title, preferably like this: “[css-writing-modes] …summary of comment…”. All issues and comments are archived, and there is also a historical archive. Advisory Committee Representatives should consult their WBS questionnaires.

A Disposition of Comments and an Implementation Report are available.

Publication as a Proposed Recommendation 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 March 2019 W3C Process Document.

For changes since the last draft, see the Changes section.

1. Introduction to Writing Modes

CSS Writing Modes Level 3 defines CSS features to support for various international writing modes, such as left-to-right (e.g. Latin or Indic), right-to-left (e.g. Hebrew or Arabic), bidirectional (e.g. mixed Latin and Arabic) and vertical (e.g. Asian scripts).

A writing mode in CSS is determined by the writing-mode, direction, and text-orientation properties. It is defined primarily in terms of its inline base direction and block flow direction:

The inline base direction is the primary direction in which content is ordered on a line and defines on which sides the “start” and “end” of a line are. The direction property specifies the inline base direction of a box and, together with the unicode-bidi property and the inherent directionality of any text content, determines the ordering of inline-level content within a line.

The block flow direction is the direction in which block-level boxes stack and the direction in which line boxes stack within a block container. The writing-mode property determines the block flow direction.

The typographic mode determines if text should apply typographic conventions specific to vertical flow for vertical scripts. This concept distinguishes vertical flow for vertical scripts from rotated horizontal flow.

A horizontal writing mode is one with horizontal lines of text, i.e. a downward or upward block flow. A vertical writing mode is one with vertical lines of text, i.e. a leftward or rightward block flow.

These terms should not be confused with vertical block flow (which is a downward or upward block flow) and horizontal block flow (which is leftward or rightward block flow). To avoid confusion, CSS specifications avoid this latter set of terms.

Writing systems typically have one or two native writing modes. Some examples are:

The text-orientation component of the writing mode controls the glyph orientation.

See Unicode Technical Note #22 [UTN22] (HTML version) for a more in-depth introduction to writing modes and vertical text.

1.1. Module Interactions

This module replaces and extends the unicode-bidi and direction features defined in [CSS2] sections 8.6 and 9.10. The interaction of its features with other text operations in setting lines of text is described in CSS Text 3 § Text Processing Order of Operations.

The computed values of the writing-mode, direction, and text-orientation properties (even on elements to which these properties themselves don’t apply [CSS-CASCADE-4]) are broadly able to influence the computed values of other, unrelated properties through calculations such as the computation of font-relative lengths or the cascade of flow-relative properties which purposefully depend on the computed writing mode or on font metrics that can depend on the writing mode.

1.2. Value Types and Terminology

This specification follows the CSS property definition conventions from [CSS2]. Value types not defined in this specification are defined in CSS Values & Units [CSS-VALUES-3]. Other CSS modules may expand the definitions of these value types.

In addition to the property-specific values listed in their definitions, all properties defined in this specification also accept the CSS-wide keywords keywords as their property value. For readability they have not been repeated explicitly.

Other important terminology and concepts used in this specification are defined in [CSS2] and [CSS-TEXT-3].

2. Inline Direction and Bidirectionality

While the characters in most scripts are written from left to right, certain scripts are written from right to left. In some documents, in particular those written with the Arabic or Hebrew script, and in some mixed-language contexts, text in a single (visually displayed) block may appear with mixed directionality. This phenomenon is called bidirectionality, or "bidi" for short.

An example of bidirectional text is a Latin name in an Arabic
                 sentence. The sentence overall is typeset right-to-left, but
                 the letters in the Latin word in the middle are typeset


The Unicode standard (Unicode Standard Annex #9) defines a complex algorithm for determining the proper ordering of bidirectional text. The algorithm consists of an implicit part based on character properties, as well as explicit controls for embeddings and overrides. CSS relies on this algorithm to achieve proper bidirectional rendering.

Two CSS properties, direction and unicode-bidi, provide explicit embedding, isolation, and override controls in the CSS layer. Because the base directionality of a text depends on the structure and semantics of the document, the direction and unicode-bidi properties should in most cases be used only to map bidi information in the markup to its corresponding CSS styles.

The HTML specifications ([HTML401], section 8.2, and [HTML] section 14.3.5) define bidirectionality behavior for HTML elements.

If a document language provides markup features to control bidi, authors and users should use those features instead and not specify CSS rules to override them.

2.1. Specifying Directionality: the direction property

Name: direction
Value: ltr | rtl
Initial: ltr
Applies to: all elements
Inherited: yes
Percentages: n/a
Computed value: specified value
Canonical order: n/a
Animation type: not animatable

Because HTML UAs can turn off CSS styling, we recommend HTML authors to use the HTML dir attribute and <bdo> element to ensure correct bidirectional layout in the absence of a style sheet. Authors should not use direction in HTML documents.

This property specifies the inline base direction or directionality of any bidi paragraph, embedding, isolate, or override established by the box. (See unicode-bidi.) In addition, it informs the ordering of table column layout, the direction of horizontal overflow, and the default alignment of text within a line, and other layout effects that depend on the box’s inline base direction.

Values for this property have the following meanings:

This value sets inline base direction (bidi directionality) to line-left-to-line-right.
This value sets inline base direction (bidi directionality) to line-right-to-line-left.

The direction property has no effect on bidi reordering when specified on inline boxes whose unicode-bidi value is normal, because the box does not open an additional level of embedding with respect to the bidirectional algorithm.

The direction property, when specified for table column boxes, is not inherited by cells in the column since columns are not the ancestors of the cells in the document tree. Thus, CSS cannot easily capture the "dir" attribute inheritance rules described in [HTML401], section

2.2. Embeddings and Overrides: the unicode-bidi property

Name: unicode-bidi
Value: normal | embed | isolate | bidi-override | isolate-override | plaintext
Initial: normal
Applies to: all elements, but see prose
Inherited: no
Percentages: n/a
Computed value: specified value
Canonical order: per grammar
Animation type: not animatable

Because HTML UAs can turn off CSS styling, we recommend HTML authors to use the HTML dir attribute, <bdo> element, and appropriate distinction of text-level vs. grouping-level HTML element types to ensure correct bidirectional layout in the absence of a style sheet. Authors should not use unicode-bidi in HTML documents.

Normally (i.e. when unicode-bidi is normal) an inline box is transparent to the unicode bidi algorithm; content is ordered as if the box’s boundaries were not there. Other values of the unicode-bidi property cause inline boxes to create scopes within the algorithm, and to override the intrinsic directionality of text.

The following informative table summarizes the box-internal and box-external effects of unicode-bidi:

Effect of non-normal values of unicode-bidi on inline boxes
strong neutral
Inside scoped embed isolate
override bidi-override isolate-override
plaintext plaintext

Values for this property have the following (normative) meanings:

The box does not open an additional level of embedding with respect to the bidirectional algorithm. For inline boxes, implicit reordering works across box boundaries.
If the box is inline, this value creates a directional embedding by opening an additional level of embedding with respect to the bidirectional algorithm. The direction of this embedding level is given by the direction property. Inside the box, reordering is done implicitly.

This value has no effect on boxes that are not inline.

On an inline box, this bidi-isolates its contents. This is similar to a directional embedding (and increases the embedding level accordingly) except that each sequence of inline-level boxes uninterrupted by any block boundary or forced paragraph break is treated as an isolated sequence:
  • the content within the sequence is ordered as if inside an independent paragraph with the base directionality specified by the box’s direction property.
  • for the purpose of bidi resolution in its containing bidi paragraph, the sequence is treated as if it were a single Object Replacement Character (U+FFFC).
In effect, neither is the content inside the box bidi-affected by the content surrounding the box, nor is the content surrounding the box bidi-affected by the content or specified directionality of the box. However, forced paragraph breaks within the box still create a corresponding break in the containing paragraph.

This value has no effect on boxes that are not inline.

This value puts the box’s immediate inline content in a directional override. For an inline, this means that the box acts like a directional embedding in the bidirectional algorithm, except that reordering within it is strictly in sequence according to the direction property; the implicit part of the bidirectional algorithm is ignored. For a block container, the override is applied to an anonymous inline box that surrounds all of its content.
This combines the isolation behavior of isolate with the directional override behavior of bidi-override: to surrounding content, it is equivalent to isolate, but within the box content is ordered as if bidi-override were specified. It effectively nests a directional override inside an isolated sequence.

This value behaves as isolate except that for the purposes of the Unicode bidirectional algorithm, the base directionality of each of the box’s bidi paragraphs (if a block container) or isolated sequences (if an inline) is determined by following the heuristic in rules P2 and P3 of the Unicode bidirectional algorithm (rather than by using the direction property of the box).

Following Unicode Bidirectional Algorithm clause HL3 [UAX9], values other than normal effectively insert the corresponding Unicode bidi control codes into the text stream at the start and end of the inline element before passing the paragraph to the Unicode bidirectional algorithm for reordering. (See § 2.4.2 CSS–Unicode Bidi Control Translation, Text Reordering.)

Bidi control codes injected by unicode-bidi at the start/end of display: inline boxes
unicode-bidi value direction value
ltr rtl
start end start end
embed LRE (U+202A) PDF (U+202C) RLE (U+202B) PDF (U+202C)
isolate LRI (U+2066) PDI (U+2069) RLI (U+2067) PDI (U+2069)
bidi-override* LRO (U+202D) PDF (U+202C) RLO (U+202E) PDF (U+202C)
isolate-override* FSI,LRO (U+2068,U+202D) PDF,PDI (U+202C,U+2069) FSI,RLO (U+2068,U+202E) PDF,PDI (U+202C,U+2069)
plaintext FSI (U+2068) PDI (U+2069) FSI (U+2068) PDI (U+2069)
* The LRO/RLO+PDF pairs are also applied to the root inline box of a block container if these values of unicode-bidi were specified on the block container.

Because the unicode-bidi property does not inherit, setting bidi-override or plaintext on a block box will not affect any descendant blocks. Therefore these values are best used on blocks and inlines that do not contain any block-level structures.

Note that unicode-bidi does not affect the direction property even in the case of plaintext, and thus does not affect direction-dependent layout calculations.

Because the Unicode algorithm has a limit of 125 levels of embedding, care should be taken not to overuse unicode-bidi values other than normal. In particular, a value of inherit should be used with extreme caution in deeply nested inline markup. However, for elements that are, in general, intended to be displayed as blocks, a setting of unicode-bidi: isolate is preferred to keep the element together in case the display is changed to inline (see example below).

2.3. Example of Bidirectional Text

The following example shows an XML document with bidirectional text. It illustrates an important design principle: document language designers should take bidi into account both in the language proper (elements and attributes) and in any accompanying style sheets. The style sheets should be designed so that bidi rules are separate from other style rules, and such rules should not be overridden by other style sheets so that the document language’s bidi behavior is preserved.

In this example, lowercase letters stand for inherently left-to-right characters and uppercase letters represent inherently right-to-left characters. The text stream is shown below in logical backing store order.

<section dir=rtl>
  <para>HEBREW1 HEBREW2 english3 HEBREW4 HEBREW5</para>
  <para>HEBREW6 <emphasis>HEBREW7</emphasis> HEBREW8</para>
<section dir=ltr>
  <para>english9 english10 english11 HEBREW12 HEBREW13</para>
  <para>english14 english15 english16</para>
  <para>english17 <quote dir=rtl>HEBREW18 english19 HEBREW20</quote></para>

Since this is arbitrary XML, the style sheet is responsible for setting the writing direction. This is the style sheet:

/* Rules for bidi */
[dir=rtl] {direction: rtl; unicode-bidi: isolate; }
[dir=ltr] {direction: ltr; unicode-bidi: isolate; }

/* Rules for presentation */
section, para  {display: block;}
emphasis       {font-weight: bold;}
quote          {font-style: italic;}

If the line length is long, the formatting of this text might look like this:

               5WERBEH 4WERBEH english3 2WERBEH 1WERBEH

                                8WERBEH 7WERBEH 6WERBEH

english9 english10 english11 13WERBEH 12WERBEH

english14 english15 english16

english17 20WERBEH english19 18WERBEH

The first <section> element is a block with a right-to-left base direction, the second <section> element is a block with a left-to-right base direction. The <para>s are blocks that inherit the base direction from their parents. Thus, the first two <para>s are read starting at the top right, the final three are read starting at the top left.

The <emphasis> element is inline-level, and since its value for unicode-bidi is normal (the initial value), it has no effect on the ordering of the text.

The <quote> element, on the other hand, creates an isolated sequence with the given internal directionality. Note that this causes HEBREW18 to be to the right of english19.

If lines have to be broken, the same text might format like this:

  -EH 4WERBEH english3


english9 english10 en-
glish11 12WERBEH

english14 english15

english17 18WERBEH
20WERBEH english19

Notice that because HEBREW18 must be read before english19, it is on the line above english19. Just breaking the long line from the earlier formatting would not have worked.

Note also that the first syllable from english19 might have fit on the previous line, but hyphenation of left-to-right words in a right-to-left context, and vice versa, is usually suppressed to avoid having to display a hyphen in the middle of a line.

2.4. Applying the Bidirectional Reordering Algorithm

User agents that support bidirectional text must apply the Unicode bidirectional algorithm to every sequence of inline-level boxes uninterrupted by any block boundary or “bidi type Bforced paragraph break. This sequence forms the paragraph unit in the bidirectional algorithm.

2.4.1. Bidi Paragraph Embedding Levels

In CSS, the paragraph embedding level must be set (following UAX9 clause HL1) according to the direction property of the paragraph’s containing block rather than by the heuristic given in steps P2 and P3 of the Unicode algorithm.

There is, however, one exception: when the computed unicode-bidi of the paragraph’s containing block is plaintext, the Unicode heuristics in P2 and P3 are used as described in [UAX9], without the HL1 override.

2.4.2. CSS–Unicode Bidi Control Translation, Text Reordering

The final order of characters within each bidi paragraph is the same as if the bidi control codes had been added as described for unicode-bidi (above), markup had been stripped, and the resulting character sequence had been passed to an implementation of the Unicode bidirectional algorithm for plain text that produced the same line-breaks as the styled text.

Note that bidi control codes in the source text are still honored, and might not correspond to the document tree structure. This can split inlines or interfere with bidi start/end control pairing in interesting ways.

2.4.3. Bidi Treatment of Atomic Inlines

In this process, replaced elements with display: inline are treated as neutral characters, unless their unicode-bidi property is either embed or bidi-override, in which case they are treated as strong characters in the direction specified for the element. (This is so that, in case the replaced element falls back to rendering inlined text content, its bidi effect on the surrounding text is consistent with its replaced rendering.)

All other atomic inline-level boxes are treated as neutral characters always.

2.4.4. Paragraph Breaks Within Embeddings and Isolates

If an inline box is broken around a bidi paragraph boundary (e.g. if split by a block or forced paragraph break), then the HL3 bidi control codes assigned to the end of the box are also added before the interruption and the codes assigned to the start of the box are also added after it. (In other words, any embedding levels, isolates, or overrides started by the box are closed at the paragraph break and reopened on the other side of it.)

For example, where <BR/> is a forced paragraph break the bidi ordering is identical between




for all values of unicode-bidi on inline elements <i1> and <i2>.

Note that this behavior is applied by CSS for CSS-declared bidi controls applied to the box tree; it does not apply to Unicode’s bidi formatting controls, which are defined to terminate their effect at the end of the bidi paragraph.

2.4.5. Reordering-induced Box Fragmentation

Since bidi reordering can split apart and reorder text that is logically contiguous, bidirectional text can cause an inline box containing such text to be split and its fragments reordered within a line.

For each line box, UAs must take the fragments of each inline box and assign the margins, borders, and padding in visual order (not logical order). The start-most fragment on the first line box in which the box appears has the start edge’s margin, border, and padding; and the end-most fragment on the last line box in which the box appears has the end edge’s margin, border, and padding. For example, in the horizontal-tb writing mode:

Analogous rules hold for vertical writing modes.

The box-decoration-break property can override this behavior to draw box decorations on both sides of each fragment. [CSS3-BREAK]

3. Vertical Writing Modes

In addition to extensions to CSS2.1’s support for bidirectional text, this module introduces the rules and properties needed to support vertical text layout in CSS.

3.1. Introduction to Vertical Writing

This subsection is non-normative.

Unlike languages that use the Latin script which are primarily laid out horizontally, Asian languages such as Chinese and Japanese can be laid out vertically. The Japanese example below shows the same text laid out horizontally and vertically. In the horizontal case, text is read from left to right, top to bottom. For the vertical case, the text is read top to bottom, right to left. Indentation from the left edge in the left-to-right horizontal case translates to indentation from the top edge in the top-to-bottom vertical case.

A comparison of horizontal and vertical Japanese shows that
                 although the lines rotate, the characters remain upright.
                 Some glyphs, however change: a period mark shifts from the
                 bottom left of its glyph box to the top right. Running
                 headers, however, may remain
                 laid out horizontally across the top of the page.

Comparison of vertical and horizontal Japanese: iBunko application (iOS)

For Chinese and Japanese lines are ordered either right to left or top to bottom, while for Mongolian and Manchu lines are ordered left to right.

The change from horizontal to vertical writing can affect not just the layout, but also the typesetting. For example, the position of a punctuation mark within its spacing box can change from the horizontal to the vertical case, and in some cases alternate glyphs are used.

Vertical text that includes Latin script text or text from other scripts normally displayed horizontally can display that text in a number of ways. For example, Latin words can be rotated sideways, or each letter can be oriented upright:

A dictionary definition for ヴィルス
                 might write the English word 'virus' rotated 90° clockwise,
                 but stack the letters of the initialisms 'RNA' and 'DNA' upright.

Examples of Latin in vertical Japanese: Daijirin Viewer 1.4 (iOS)

In some special cases such as two-digit numbers in dates, text is fit compactly into a single vertical character box:

An excerpt from MacFan shows several possible vertical layouts
                 for numbers: the two-digit month and day are written as
                 horizontal-in-vertical blocks; the years are written with
                 each character upright; except in the English phrase
                 “for Mac 2011”, where the date is rotated to
                 match the rotated Latin.

Mac Fan, December 2010, p.49

Layouts often involve a mixture of vertical and horizontal elements:

Magazines often mix horizontal and vertical layout; for
                 example, using one orientation for the main article text
                 and a different one for sidebar or illustrative content.

Mixture of vertical and horizontal elements

Vertical text layouts also need to handle bidirectional text layout; clockwise-rotated Arabic, for example, is laid out bottom-to-top.

3.2. Block Flow Direction: the writing-mode property

Name: writing-mode
Value: horizontal-tb | vertical-rl | vertical-lr
Initial: horizontal-tb
Applies to: All elements except table row groups, table column groups, table rows, table columns, ruby base container, ruby annotation container
Inherited: yes
Percentages: n/a
Computed value: specified value
Canonical order: n/a
Animation type: not animatable

This property specifies whether lines of text are laid out horizontally or vertically and the direction in which blocks progress. Possible values:

Top-to-bottom block flow direction. Both the writing mode and the typographic mode are horizontal.
Right-to-left block flow direction. Both the writing mode and the typographic mode are vertical.
Left-to-right block flow direction. Both the writing mode and the typographic mode are vertical.

The writing-mode property specifies the block flow direction, which determines the ordering direction of block-level boxes in a block formatting context; the ordering direction of line boxes in a block container that contains inlines; the ordering direction of rows in a table; etc. By virtue of determining the stacking direction of line boxes, the writing-mode property also determines whether the line boxes' orientation (and thus the writing mode) is horizontal or vertical. The text-orientation property then determines how text is laid out within the line box.

The content of replaced elements do not rotate due to the writing mode: images and external content such as from <iframe>s, for example, remain upright, and the default object size of 300px×150px does not re-orient. However embedded replaced content involving text (such as MathML content or form elements) should match the replaced element’s writing mode and line orientation if the UA supports such a vertical writing mode for the replaced content.

In the following example, two block elements (1 and 3) separated by an image (2) are presented in various flow writing modes.

Here is a diagram of horizontal writing mode (writing-mode: horizontal-tb):

Diagram of horizontal layout: blocks 1, 2, and 3 are stacked top-to-bottom

Here is a diagram for the right-to-left vertical writing mode commonly used in East Asia (writing-mode: vertical-rl):

Diagram of a right-to-left vertical layout: blocks 1, 2,
                  and 3 are arranged side by side from right to left

And finally, here is a diagram for the left-to-right vertical writing mode used for Manchu and Mongolian (writing-mode: vertical-lr):

Diagram of left-to-right vertical layout: blocks 1, 2,
                  and 3 are arranged side by side from left to right

In the following example, some form controls are rendered inside a block with vertical-rl writing mode. The form controls are rendered to match the writing mode.

  form { writing-mode: vertical-rl; }
<p><label>姓名 <input value="艾俐俐"></label>
<p><label>语言 <select><option>English

Screenshot of vertical layout: the input element is
                laid lengthwise from top to bottom and its contents
                rendered in a vertical typographic mode, matching the
                labels outside it. The drop-down selection control
                after it slides out to the side (towards the after
                edge of the block) rather than downward as it would
                in horizontal writing modes.

If a box has a different writing-mode value than its parent box (i.e. nearest ancestor without display: contents):

As all other inherited CSS properties do, the writing-mode property inherits to SVG elements inlined (rather than linked) into the source document. This could cause unintentional side effects when, for example, an SVG image designed only for horizontal flow was embedded into a vertical flow document.

Authors can prevent this from happening by adding the following rule:

svg { writing-mode: initial; }

3.2.1. Obsolete SVG1.1 writing-mode Values

SVG1.1 [SVG11] defines some additional values: lr, lr-tb, rl, rl-tb, tb, and tb-rl.

These values are obsolete in any context except SVG1 documents and are therefore optional for non-SVG UAs. Supporting SVG1.1 writing-mode values in CSS syntax

UAs that wish to support these values in the context of CSS must compute them as follows:

Mapping of Obsolete SVG1.1 writing-mode values to modern CSS
Specified Computed
lr horizontal-tb
tb vertical-rl

The SVG1.1 values were also present in an older of the CSS writing-mode specification, which is obsoleted by this specification. The additional tb-lr value of that revision is replaced by vertical-lr. Supporting SVG1.1 writing-mode values in presentational attributes

In order to support legacy content with presentational attributes, and to allow authors to create documents that support older clients, SVG UAs must add the following style sheet rules to their default UA stylesheet:

@namespace svg "http://www.w3.org/2000/svg";
svg|*[writing-mode=lr], svg|*[writing-mode=lr-tb],
svg|*[writing-mode=rl], svg|*[writing-mode=rl-tb] {
  writing-mode: horizontal-tb; }
svg|*[writing-mode=tb], svg|*[writing-mode=tb-rl] {
  writing-mode: vertical-rl; }
Authors who wish to create forwards and backwards-compatible SVG content in CSS syntax can use the CSS forwards-compatible parsing rules to do so, e.g.
svg|text { writing-mode: tb; writing-mode: vertical-rl; }

4. Inline-level Alignment

When different kinds of inline-level content are placed together on a line, the baselines of the content and the settings of the vertical-align property control how they are aligned in the transverse direction of the line box. This section discusses what baselines are, how to find them, and how they are used together with the vertical-align property to determine the alignment of inline-level content.

4.1. Introduction to Baselines

This section is non-normative.

A baseline is a line along the inline axis of a line box along which individual glyphs of text are aligned. Baselines guide the design of glyphs in a font (for example, the bottom of most alphabetic glyphs typically align with the alphabetic baseline), and they guide the alignment of glyphs from different fonts or font sizes when typesetting.

Picture of alphabetic text in two font sizes with the baseline and em-boxes

Alphabetic text in two font sizes with the baseline and em-boxes

Different writing systems prefer different baseline tables.

Latin prefers the alphabetic baseline, on top of which most
                 letters rest, though some have descenders that dangle below it.
                 Indic scripts are sometimes typeset with a hanging baseline,
                 since their glyph shapes appear to be hanging from a
                 horizontal line.
                 Han-based systems, whose glyphs are designed to fill a square,
                 tend to align on their bottoms.

Preferred baselines in various writing systems

A well-constructed font contains a baseline table, which indicates the position of one or more baselines within the font’s design coordinate space. (The design coordinate space is scaled with the font size.)

In a well-designed mixed-script font, the glyphs are positioned in the coordinate space to harmonize with one another when typeset together. The baseline table is then constructed to match the shape of the glyphs, each baseline positioned to match the glyphs from its preferred scripts.

The baseline table is a property of the font, and the positions of the various baselines apply to all glyphs in the font.

Different baseline tables can be provided for alignment in horizontal and vertical text. UAs should use the vertical tables in vertical typographic modes and the horizontal tables otherwise.

4.2. Text Baselines

In this specification, only the following baselines are considered:

The alphabetic baseline, which typically aligns with the bottom of uppercase Latin glyphs.
The central baseline, which typically crosses the center of the em box. If the font is missing this baseline, it is assumed to be halfway between the ascender (over) and descender (under) edges of the em box.

In vertical typographic mode, the central baseline is used as the dominant baseline when text-orientation is mixed or upright. Otherwise the alphabetic baseline is used.

A future CSS module will deal with baselines in more detail and allow the choice of other dominant baselines and alignment options.

4.3. Atomic Inline Baselines

If an atomic inline (such as an inline-block, inline-table, or replaced inline element) does not have a baseline, then the UA synthesizes a baseline table thus:

The alphabetic baseline is assumed to be at the under margin edge.
The central baseline is assumed to be halfway between the under and over margin edges of the box.

The vertical-align property in [CSS2] defines the baseline of inline-table and inline-block boxes with some exceptions.

4.4. Baseline Alignment

The dominant baseline (which can change based on the typographic mode) is used in CSS for alignment in two cases:

5. Introduction to Vertical Text Layout

Each writing system has one or more native orientations. Modern scripts can therefore be classified into three orientational categories:

Scripts that have horizontal, but not vertical, native orientation. Includes: Latin, Arabic, Hebrew, Devanagari
Scripts that have vertical, but not horizontal, native orientation. Includes: Mongolian, Phags Pa
Scripts that have both vertical and horizontal native orientation. Includes: Han, Hangul, Japanese Kana

A vertical script is one that has a native vertical orientation: i.e. one that is either vertical-only or that is bi-orientational. A horizontal script is one that has a native horizontal orientation: i.e. one that is either horizontal-only or that is bi-orientational. (See Appendix A for a categorization of scripts by native orientation.)

A Venn diagram of these distinctions would show two circles:
                     one labelled 'vertical', the other 'horizontal'. The overlapped
                     region would represent the bi-orientational scripts, while
                     horizontal-only and vertical-only scripts would occupy their
                     respective circles' exclusive regions.

In modern typographic systems, all glyphs are assigned a horizontal orientation, which is used when laying out text horizontally. To lay out vertical text, the UA needs to transform the text from its horizontal orientation. This transformation is the bi-orientational transform, and there are two types:

Rotate the glyph from horizontal to vertical Rotate the glyph from horizontal to vertical
Translate the glyph from horizontal to vertical Translate the glyph from horizontal to vertical

Scripts with a native vertical orientation have an intrinsic bi-orientational transform, which orients them correctly in vertical text: most CJK (Chinese/Japanese/Korean) characters translate, that is, they are always upright. Characters from other scripts, such as Mongolian, rotate.

Scripts without a native vertical orientation can be either rotated (set sideways) or translated (set upright): the transform used is a stylistic preference depending on the text’s usage, rather than a matter of correctness. The text-orientation property’s mixed and upright values are provided to specify rotation vs. translation of horizontal-only text.

5.1. Orienting Text: the text-orientation property

Name: text-orientation
Value: mixed | upright | sideways
Initial: mixed
Applies to: all elements except table row groups, rows, column groups, and columns
Inherited: yes
Percentages: n/a
Computed value: specified value
Canonical order: n/a
Animation type: not animatable

This property specifies the orientation of text within a line. Current values only have an effect in vertical typographic modes: the property has no effect on boxes in horizontal typographic modes.

Values have the following meanings:


In vertical writing modes, typographic character units from horizontal-only scripts are typeset sideways, i.e. 90° clockwise from their standard orientation in horizontal text. Typographic character units from vertical scripts are typeset with their intrinsic orientation. See Vertical Orientations for further details.

This value is typical for layout of dominantly vertical-script text.


In vertical writing modes, typographic character units from horizontal-only scripts are typeset upright, i.e. in their standard horizontal orientation. Typographic character units from vertical scripts are typeset with their intrinsic orientation and shaped normally. See Vertical Orientations for further details.

This value causes the used value of direction to be ltr, and for the purposes of bidi reordering, causes all characters to be treated as strong LTR.

Note: The used value, rather than the computed value, of direction is influenced so that rtl can inherit properly into any descendants (such as the contents of a horizontal inline-block) where this directional override does not apply.


In vertical writing modes, this causes all text to be typeset sideways, as if in a horizontal layout, but rotated 90° clockwise.

text-orientation: mixed text-orientation: upright text-orientation: sideways
mixed upright sideways

text-orientation values (writing-mode is vertical-rl)

Changing the value of this property may affect inline-level alignment. Refer to Text Baselines for more details.

UAs may accept sideways-right as a value that computes to sideways if needed for backward compatibility reasons.

As of writing, major implementations do not support the automatic LTR treatment of RTL characters for upright typesetting. In such cases, authors may need to explicitly specify unicode-bidi and direction as in the following example:
.vertical-upright-hebrew {
    writing-mode: vertical-rl;
    text-orientation: upright;
    unicode-bidi: bidi-override;
    direction: ltr;

5.1.1. Vertical Typesetting and Font Features

When typesetting text in vertical-rl and vertical-lr modes, text is typeset either “upright” or “sideways” as defined below:

upright typesetting
Typographic character units are individually typeset upright in vertical lines with vertical font metrics. The UA must synthesize vertical font metrics for fonts that lack them. (This specification does not define heuristics for synthesizing such metrics.) Additionally, font features (such as alternate glyphs and other transformation) intended for use in vertical typesetting must be used. (E.g. the OpenType vert feature must be enabled.) Furthermore, characters from horizontal cursive scripts (such as Arabic) are shaped in their isolated forms when typeset upright.

Note that even when typeset “upright”, some glyphs should appear rotated. For example, dashes and enclosing punctuation should be oriented relative to the inline axis. In OpenType, this is typically handled by glyph substitution, although not all fonts have alternate glyphs for all relevant codepoints. (East Asian fonts usually provide alternates for East Asian codepoints, but Western fonts typically lack any vertical typesetting features and East Asian fonts typically lack vertical substitutions for Western codepoints.) Unicode published draft data on which characters should appear sideways as the SVO property in this data file; however, this property has been abandoned for the current revision of [UAX50].

Typographic character units which are classified as Tr or Tu in [UAX50] are expected to have alternate glyphs or positioning for typesetting upright in vertical text. In the case of Tr characters, if such vertical alternate glyphs are missing from the font, the UA may wish to [RFC6919] (but is not expected to) synthesize the missing glyphs by typesetting them sideways etc.

sideways typesetting
Typographic character units typeset as a run rotated 90° clockwise from their upright orientation, using horizontal metrics and composition, and vertical typesetting features are not used. However, if the font has features meant to be enabled for sideways text that is typeset in vertical lines (e.g. to adjust brush stroke angles or alignment), those features are used. (An example of such a feature would be the proposed vrtr OpenType font feature.)

5.1.2. Mixed Vertical Orientations

[UAX50] defines the Vertical_Orientation property for the default glyph orientation of mixed-orientation vertical text. When text-orientation is mixed, the UA must determine the orientation of each typographic character unit by its Vertical_Orientation property: typeseting it upright if its orientation property is U, Tu, or Tr; or typesetting it sideways (90° clockwise from horizontal) if its orientation property is R.

Note that UAX50 does not handle scripts that rotate -90° in vertical contexts, so they will not be typeset correctly with mixed orientation. The sideways-lr value in Level 4, however, can correctly display such scripts.

The OpenType vrt2 feature, which is intended for mixed-orientation typesetting, is not used by CSS. It delegates the responsibility for orienting glyphs to the font designer. CSS instead dictates the orientation through [UAX50] and orients glyphs by typesetting them sideways or upright as appropriate.

5.1.3. Obsolete: the SVG1.1 glyph-orientation-vertical property

Name: glyph-orientation-vertical
Value: auto | 0deg | 90deg | 0 | 90
Initial: n/a
Applies to: n/a
Inherited: n/a
Percentages: n/a
Computed value: n/a
Canonical order: n/a
Animatable: n/a

Some SVG user agents will need to process documents containing the obsolete SVG glyph-orientation-vertical property, which was defined to accept an auto keyword as well as <angle> and <integer> values representing multiples of 90°. While supporting this property is optional, UAs that do so must alias glyph-orientation-vertical as a shorthand of text-orientation as follows:

Shorthand glyph-orientation-vertical value Longhand text-orientation value
auto mixed
0deg upright
0 upright
90deg sideways
90 sideways

UAs must ignore and treat as invalid any other values for the glyph-orientation-vertical property; and treat as invalid the glyph-orientation-horizontal property in its entirety.

Note: The 180deg and 270deg values, the radian and gradian values, and the glyph-orientation-horizontal property are not mapped because they have no known use cases nor significant amounts of dependent content, and are therefore not part of CSS, and have been likewise dropped from SVG.

6. Abstract Box Terminology

CSS2.1 [CSS2] defines the box layout model of CSS in detail, but only for the horizontal-tb writing mode. Layout is analogous in writing modes other than horizontal-tb; however directional and dimensional terms in CSS2.1 must be abstracted and remapped appropriately.

This section defines abstract directional and dimensional terms and their mappings in order to define box layout for other writing modes, and to provide terminology for future specs to define their layout concepts abstractly. (The next section explains how to apply them to CSS2.1 layout calculations and how to handle orthogonal flows.) Although they derive from the behavior of text, these abstract mappings exist even for boxes that do not contain any line boxes: they are calculated directly from the values of the writing-mode and direction properties.

There are three sets of directional terms in CSS:

Interpreted relative to the page, independent of writing mode. The physical directions are left, right, top, and bottom.
Interpreted relative to the flow of content. The flow-relative directions are start and end, or block-start, block-end, inline-start, and inline-end if the dimension is also ambiguous.
Interpreted relative to the orientation of the line box. The line-relative directions are line-left, line-right, line-over, and line-under.

The physical dimensions are width and height, which correspond to measurements along the x-axis (horizontal dimension) and y-axis (vertical dimension), respectively. Abstract dimensions are identical in both flow-relative and line-relative terms, so there is only one set of these terms.

Note: [CSS3-FLEXBOX] also defines flex-relative terms, which are used in describing flex layout.

6.1. Abstract Dimensions

The abstract dimensions are defined below:

block dimension
The dimension perpendicular to the flow of text within a line, i.e. the vertical dimension in horizontal writing modes, and the horizontal dimension in vertical writing modes.
inline dimension
The dimension parallel to the flow of text within a line, i.e. the horizontal dimension in horizontal writing modes, and the vertical dimension in vertical writing modes.
block axis
The axis in the block dimension, i.e. the vertical axis in horizontal writing modes and the horizontal axis in vertical writing modes.
inline axis
The axis in the inline dimension, i.e. the horizontal axis in horizontal writing modes and the vertical axis in vertical writing modes.
block size
logical height
A measurement in the block dimension: refers to the physical height (vertical dimension) in horizontal writing modes, and to the physical width (horizontal dimension) in vertical writing modes.
inline size
logical width
A measurement in the inline dimension: refers to the physical width (horizontal dimension) in horizontal writing modes, and to the physical height (vertical dimension) in vertical writing modes.

6.2. Flow-relative Directions

The flow-relative directions, block-start, block-end, inline-start, and inline-end, are defined relative to the flow of content on the page. In an LTR horizontal-tb writing mode, they correspond to the top, bottom, left, and right directions, respectively. They are defined as follows:

The side that comes earlier in the block flow direction, as determined by the writing-mode property: the physical top in horizontal-tb mode, the right in vertical-rl, and the left in vertical-lr.
The side opposite block-start.
The side from which text of the inline base direction would start. For boxes with a used direction value of ltr, this means the line-left side. For boxes with a used direction value of rtl, this means the line-right side.
The side opposite start.

Where contextually unambiguous or encompassing both meanings, the terms start and end are used in place of block-start/inline-start and block-end/inline-end, respectively.

Note that while determining the block-start and block-end sides of a box depends only on the writing-mode property, determining the inline-start and inline-end sides of a box depends not only on the writing-mode property but also the direction property.

Physical/logical terms as applicable to typical English text layout
Physical/logical terms as applicable to typical Chinese text layout

6.3. Line-relative Directions

The line orientation determines which side of a line box is the logical “top” (ascender side). It is given by the writing-mode property. Usually the line-relative “top” corresponds to the block-start side, but this is not always the case: in Mongolian typesetting (and thus by default in vertical-lr writing modes), the line-relative “top” corresponds to the block-end side. Hence the need for distinct terminology.

Mongolian mixed with English

A primarily Mongolian document, such as the one above, is written in vertical lines stacking left to right, but lays its Latin text with the tops of the glyphs towards the right. This makes the text run in the same inline direction as Mongolian (top-to-bottom) and face the same direction it does in other East Asian layouts (which have vertical lines stacking right to left), but the glyphs' tops are facing the bottom of the line stack rather than the top, which in an English paragraph would be upside-down. (See this Diagram of Mongolian Text Layout.)

In addition to a line-relative “top” and “bottom” to map things like 'vertical-align: top', CSS also needs to refer to a line-relative “left” and “right” in order to map things like text-align: left. Thus there are four line-relative directions, which are defined relative to the line orientation as follows:

over or line-over
Nominally the side that corresponds to the ascender side or “top” side of a line box. (The side overlines are typically drawn on.)
under or line-under
Opposite of over: the line-relative “bottom” or descender side. (The side underlines are typically drawn on.)
The line-relative "left" side of a line box, which is nominally the side from which LTR text would start.
The line-relative "right" side of a line box, which is nominally the side from which RTL text would start. (Opposite of line-left.)

See the table below for the exact mappings between physical and line-relative directions.

Line orientation compass

Line orientation in horizontal-tb

Typical orientation in vertical

Line orientation in vertical-rl and vertical-lr

Baseline of an upright glyph is drawn vertically from the top center

Vertical baseline of an upright glyph

When text-orientation: upright, the baseline is still vertical, and the vertical baseline in the font is used, or the vertical baseline is synthesized if the font does not provide.

Since the baseline is vertical, the definitions for mixed or sideways above still apply; i.e., line-over is on right, and line-under is on left.

This is in line with font systems such as OpenType which defines the ascender on right and the descender on left in their vertical metrics.

6.4. Abstract-to-Physical Mappings

The following table summarizes the abstract-to-physical mappings (based on the used direction and writing-mode):

Abstract-Physical Mapping
writing-mode horizontal-tb vertical-rl vertical-lr
direction ltr rtl ltr rtl ltr rtl
block-size height width
inline-size width height
block-start top right left
block-end bottom left right
inline-start left right top bottom top bottom
inline-end right left bottom top bottom top
over top right
under bottom left
line-left left top
line-right right bottom

Note: The used direction depends on the computed writing-mode and text-orientation: in vertical writing modes, a text-orientation value of upright forces the used direction to ltr.

7. Abstract Box Layout

7.1. Principles of Layout in Vertical Writing Modes

CSS box layout in vertical writing modes is analogous to layout in the horizontal writing modes, following the principles outlined below:

Layout calculation rules (such as those in CSS2.1, Section 10.3) that apply to the horizontal dimension in horizontal writing modes instead apply to the vertical dimension in vertical writing modes. Likewise, layout calculation rules (such as those in CSS2.1, Section 10.6) that apply to the vertical dimension in horizontal writing modes instead apply to the horizontal dimension in vertical writing modes. Thus:

For example, in vertical writing modes, table rows are vertical and table columns are horizontal. In a vertical-rl mixed rtl table, the first column would be on the bottom (the inline-start side), and the first row on the right (the block-start side). The table’s margin-right and margin-left would collapse with margins before (on the right) and after (on the left) the table, respectively, and if the table had auto values for margin-top and margin-bottom it would be centered vertically within its block flow.

Diagram of a vertical-rl mixed rtl table in a
        vertical block formatting context, showing the ordering of rows,
        cells, and columns as described above.

Table in vertical-rl RTL writing mode

For features such as text alignment, floating, and list marker positioning, that primarily reference the left or right sides of the line box or its longitudinal parallels and therefore have no top or bottom equivalent, the line-left and line-right sides are used as the reference for the left and right sides respectively.

Likewise for features such as underlining, overlining, and baseline alignment (the unfortunately-named vertical-align), that primarily reference the top or bottom sides of the linebox or its transversal parallels and therefore have no left or right equivalent, the line-over and line-under sides are used as the reference for the top and bottom sides respectively.

The details of these mappings are provided below.

7.2. Dimensional Mapping

Certain properties behave logically as follows:

The height properties (height, min-height, and max-height) refer to the physical height, and the width properties (width, min-width, and max-width) refer to the physical width. However, the rules used to calculate box dimensions and positions are logical.

For example, the calculation rules in CSS2.1 Section 10.3 are used for the inline dimension measurements: they apply to the inline size (which could be either the physical width or physical height) and to the inline-start and inline-end margins, padding, and border. Likewise the calculation rules in CSS2.1 Section 10.6 are used in the block dimension: they apply to the block size and to the block-start and block-end margins, padding, and border. [CSS2]

As a corollary, percentages on the margin and padding properties, which are always calculated with respect to the containing block width in CSS2.1, are calculated with respect to the inline size of the containing block in CSS3.

7.3. Orthogonal Flows

We appreciate feedback in general, but we are particularly interested in feedback on this particularly complicated section.

When a box has a different writing-mode from its containing block two cases are possible:

When a box has a writing mode that is perpendicular to its containing block it is said to be in, or establish, an orthogonal flow.

To handle this case, CSS layout calculations are divided into two phases: sizing a box, and positioning the box within its flow.

Since auto margins are resolved consistent with the containing block’s writing mode, a box establishing an orthogonal flow can, once sized, be aligned or centered within its containing block just like other block-level boxes by using auto margins.

Diagram of a vertical flow box appearing in between two horizontal flow boxes.

An example of orthogonal flow

For example, if a vertical block is placed inside a horizontal block, then when calculating the physical height (which is the inline size) of the child block the physical height of the parent block is used as the child’s containing block inline size, even though the physical height is the block size, not the inline size, of the parent block.

On the other hand, because the containing block is in a horizontal writing mode, the vertical margins on the child participate in margin-collapsing, even though they are in the inline-axis of the child, and horizontal auto margins will expand to fill the containing block, even though they are in the block-axis of the child.

Note that this section requires that when a child box auto-sized in its block axis establishes an orthogonal flow, the used block size of the child is calculated to fit its content; and this resulting content-based size is used as input to the inline-axis min-content size and max-content size of the parent.

This means that when applying shrink-to-fit formula to a box such as an inline-block, float, or table-cell, if its child establishes an orthogonal flow, the calculation dependency must be changed so that the sizing phase of the child runs first and its used block size becomes an input to the inline-size shrink-to-fit formula of the parent.

7.3.1. Available Space of Orthogonal Flows

It is common in CSS for a containing block to have a definite inline size, but not a definite block size. This typically happens in CSS2.1 when a containing block has an auto height, for example: its width is given by the calculations in 10.3.3, but its block size depends on its contents. In such cases the available inline space is defined as the inline size of the containing block; but the available block space, which would otherwise be the block size of the containing block, is infinite.

Putting a box in an orthogonal flow can result in the opposite: for the box’s available block space to be definite, but its available inline space to be indefinite. In such cases a percentage of the containing block’s inline size cannot be defined, and inline axis computations cannot be resolved. In these cases, an additional fallback size is used in place of the available inline space for calculations that require a definite available inline space: this size is the smallest of

See [css-sizing-3] for further details on CSS sizing terminology and concepts.

7.3.2. Auto-sizing Orthogonal Flow Roots

The inline-axis automatic size of a block-level or block container orthogonal flow (i.e. the size used when its preferred size property is auto) is calculating as its fit-content size, i.e. min(max-content inline size, max(min-content inline size, stretch-fit inline size), where the available space used to calculate the stretch-fit inline size is either the size of the containing block if that is definite, or else the fallback size as defined above.

The automatic sizing of orthogonal multi-column containers (in both axes) and of other display types not mentioned above is not defined in this specification.

Note: See also CSS Writing Modes Level 4.

7.3.3. Fragmenting Orthogonal Flows

This section is informative.

With regards to fragmentation, the rules in CSS2.1 still hold in vertical writing modes and orthogonal flows: break opportunities do not occur inside line boxes, only between them. UAs that support [CSS3COL] may break in the (potentially zero-width) gap between columns, however.

Note that if content spills outside the pagination stream established by the root element, the UA is not required to print such content. Authors wishing to mix writing modes with long streams of text are thus encouraged to use CSS columns to keep all content flowing in the document’s pagination direction.

In other words, if your document would require two scrollbars on the screen it probably won’t all print. Fix your layout, e.g. by using columns so that it all scrolls (and therefore paginates) in one direction if you want to make sure it’ll all print. T-shaped documents tend not to print well.

7.4. Flow-Relative Mappings

Flow-relative directions are calculated with respect to the writing mode of the containing block of the box and used to abstract layout rules related to the box properties (margins, borders, padding) and any properties related to positioning the box within its containing block (float, clear, top, bottom, left, right, caption-side). For inline-level boxes, the writing mode of the parent box is used instead. (The left/right/top/bottom-named properties and values themselves are still mapped physically; with a special exception made for caption-side, whose top/top-outside and bottom/bottom-outside values are associated to the block-start and block-end sides of the table, respectively.)

For example, the margin that is dropped when a box’s inline dimension is over-constrained is the end margin as determined by the writing mode of the containing block.

The margin collapsing rules apply exactly with the block-start margin substituted for the top margin and the block-end margin substituted for the bottom margin. Similarly the block-start padding and border are substituted for the top padding and border, and the block-end padding and border substituted for the bottom padding and border. Note this means only block-start and block-end margins ever collapse.

Flow-relative directions are calculated with respect to the writing mode of the box and used to abstract layout related to the box’s contents:

7.5. Line-Relative Mappings

The line-relative directions are over, under, line-left, and line-right. In an LTR horizontal-tb writing mode, they correspond to the top, bottom, left, and right directions, respectively.

The line-right and line-left directions are calculated with respect to the writing mode of the box and used to interpret the left and right values of the following properties:

The line-right and line-left directions are calculated with respect to the writing mode of the containing block of the box and used to interpret the left and right values of the following properties:

The over and under directions are calculated with respect to the writing mode of the box and used to define the interpretation of the "top" (over) and "bottom" (under) sides of the line box as follows:

7.6. Purely Physical Mappings

The following values are purely physical in their definitions and do not respond to changes in writing mode:

8. The Principal Writing Mode

The principal writing mode of the document is determined by the used writing-mode, direction, and text-orientation values of the root element. This writing mode is used, for example, to determine the direction of scrolling and the default page progression direction.

As a special case for handling HTML documents, if the root element has a body child element [HTML], the used value of the of writing-mode and direction properties on root element are taken from the computed writing-mode and direction of the first such child element instead of from the root element’s own values. The UA may also propagate the value of text-orientation in this manner. Note that this does not affect the computed values of writing-mode, direction, or text-orientation of the root element itself.

Note: Propagation is done on used values rather than computed values to avoid disrupting other aspects of style computation, such as inheritance, logical property mapping logic, or length value computation.

8.1. Propagation to the Initial Containing Block

The principal writing mode is propagated to the initial containing block and to the viewport, thereby affecting the layout of the root element and the scrolling direction of the viewport.

8.2. Page Flow: the page progression direction

In paged media CSS classifies all pages as either left or right pages. The page progression direction (see [CSS3PAGE]), which determines whether the left or right page in a spread is first in the flow and whether the first page is by default a left or right page, depends on the principal writing mode as follows:

principal writing mode page progression
horizontal-tb and ltr left-to-right
horizontal-tb and rtl right-to-left
vertical-rl right-to-left
vertical-lr left-to-right

Note: Unless otherwise overridden, the first page of a document begins on the second half of a spread, e.g. on the right page in a left-to-right page progression.

9. Glyph Composition

9.1. Horizontal-in-Vertical Composition: the text-combine-upright property

Name: text-combine-upright
Value: none | all
Initial: none
Applies to: non-replaced inline elements
Inherited: yes
Percentages: n/a
Computed value: specified keyword
Canonical order: n/a
Animation type: not animatable

This property specifies the combination of multiple typographic character units into the space of a single typographic character unit. If the combined text is wider than 1em, the UA must fit the contents within 1em, see below. The resulting composition is treated as a single upright glyph for the purposes of layout and decoration. This property only has an effect in vertical writing modes. Values have the following meanings:

No special processing.
Attempt to typeset horizontally all consecutive typographic character units within the box such that they take up the space of a single typographic character unit within the vertical line box.

In East Asian documents, the text-combine-upright effect is often used to display Latin-based strings such as components of a date or letters of an initialism, always in a horizontal writing mode regardless of the writing mode of the line:

Diagram of tate-chu-yoko, showing the two digits of a date
                   set halfwidth side-by-side in a vertical column of text

Example of horizontal-in-vertical tate-chu-yoko

The figure is the result of the rules

date span { text-combine-upright: all; }

and the following markup:


In Japanese, this effect is known as tate-chu-yoko.

Future levels of CSS Writing Modes will introduce values to automatically detect commonly-affected sequences. For example, CSS Writing Modes Level 4 introduces the digits value to combine sequences of digits.

9.1.1. Text Run Rules

To avoid complexity in the rendering and layout, text-combine-upright can only combine plain text: consecutive typographic character units that are not interrupted by a box boundary.

However, because the property inherits, the UA should ensure that the contents of the box effecting the combination are not part of an otherwise-combinable sequence that happens to begin or end outside the box; if so, then the text is laid out normally, as if text-combine-upright were none.

For example, given the rule

tcy { text-combine-upright: all; }

if the following markup were given:


no text would combine.

9.1.2. Layout Rules

When combining text as for text-combine-upright: all, the glyphs of the combined text are bidi-isolated and composed horizontally (ignoring letter-spacing and any forced line breaks, but using the specified font settings), similar to the contents of an inline-block box with a horizontal writing mode and a line-height of 1em. Processing of document white space included in the combined text is not defined in this level. The effective size of the composition is assumed to be 1em square; anything outside the square is not measured for layout purposes. The UA should center the glyphs horizontally and vertically within the measured 1em square.

The baseline of the resulting composition must be chosen such that the square is centered between the text-over and text-under baselines of its parent inline box prior to any baseline alignment shift (vertical-align). For bidi reordering, the composition is treated the same as a typographic character unit with text-orientation: upright. For line breaking before and after the composition, it is treated as a regular inline with its actual contents. For other text layout purposes, e.g. emphasis marks, text-decoration, spacing, etc. the resulting composition is treated as a single glyph representing the Object Replacement Character U+FFFC.

9.1.3. Compression Rules

The UA must ensure that the combined advance width of the composition fits within 1em by compressing the combined text if necessary. (This does not necessarily mean that the glyphs will fit within 1em, as some glyphs are designed to draw outside their geometric boundaries.) OpenType implementations must use width-specific variants (OpenType features hwid/twid/qwid; other glyph-width features such as fwid or pwid are not included) to compress text in cases where those variants are available for all typographic character units in the composition. Otherwise, the UA may use any means to compress the text, including substituting half-width, third-width, and/or quarter-width glyphs provided by the font, using other font features designed to compress text horizontally, scaling the text geometrically, or any combination thereof.

For example, a simple OpenType-based implementation might compress the text as follows:

  1. Enable 1/n-width glyphs for combined text of n typographic character units (i.e. use OpenType hwid for 2 typographic character units, twid for 3 typographic character units, etc.) if the number of typographic character units > 1. Note that the number of typographic character units ≠ number of Unicode codepoints!
  2. If the result is wider than 1em, horizontally scale the result to 1em.

A different implementation that utilizes OpenType layout features might compose the text first with normal glyphs to see if that fits, then substitute in half-width or third-width forms as available and necessary, possibly adjusting its approach or combining it with scaling operations depending on the available glyph substitutions.

In some fonts, the ideographic glyphs are given a compressed design such that they are 1em wide but shorter than 1em tall. To accommodate such fonts, the UA may vertically scale the composition to match the advance height of 水 U+6C34 as rendered according to the specified font settings. In such a case the resulting composition assumes the advance height of 水 U+6C34 rather than 1em. Full-width Characters

In order to preserve typographic color when compressing the text to 1em, when the combined text consists of more than one typographic character unit, then any full-width typographic character units should first be converted to their non-full-width equivalents by reversing the algorithm defined for text-transform: full-width in [CSS-TEXT-3] before applying other compression techniques.

Properties that affect glyph selection, such as the font-variant and font-feature-settings properties defined in [CSS3-FONTS], can potentially affect the selection of variants for characters included in combined text runs. Authors are advised to use these properties with care when text-combine-upright is also used.

10. Privacy and Security Considerations

This specification introduces no new privacy leaks, or security considerations beyond "implement it correctly".


Changes since the September 2019 CSS Writing Modes Module Level 3 Candidate Recommendation

No substantive change; minor editorial fixes (see issue 4293, 4272, and 4273).

Changes since the July 2019 CSS Writing Modes Module Level 3 Candidate Recommendation

Changes since the May 2018 CSS Writing Modes Module Level 3 Candidate Recommendation

Changes since the December 2017 CSS Writing Modes Module Level 3 Candidate Recommendation

Changes since the December 2015 CSS Writing Modes Module Level 3 Candidate Recommendation

A Disposition of Comments is available.

Changes since the March 2014 CSS Writing Modes Module Level 3 Candidate Recommendation


L. David Baron, Brian Birtles, James Clark, John Daggett, Nami Fujii, Daisaku Hataoka, Martin Heijdra, Laurentiu Iancu, Richard Ishida, Jonathan Kew, Yasuo Kida, Tatsuo Kobayashi, Toshi Kobayashi, Ken Lunde, Shunsuke Matsuki, Nat McCully, Eric Muller, Paul Nelson, Kenzou Onozawa, Chris Pratley, Xidorn Quan, Florian Rivoal, Dwayne Robinson, Simon Sapin, Marcin Sawicki, Dirk Schulze, Hajime Shiozawa, Alan Stearns, Michel Suignard, Takao Suzuki, Gérard Talbot, Masataka Yakura, Taro Yamamoto, Steve Zilles

Appendix A: Vertical Scripts in Unicode

This section is informative.

This appendix lists the vertical-only and bi-orientational scripts in Unicode 6.0 [UNICODE] and their transformation from horizontal to vertical orientation. Any script not listed explicitly is assumed to be horizontal-only. The script classification of Unicode characters is given by [UAX24].

Vertical Scripts in Unicode
Code Name Transform (Clockwise) Vertical Intrinsic Direction
Bopo Bopomofo ttb
Egyp Egyptian Hieroglyphs ttb
Hira Hiragana ttb
Kana Katakana ttb
Hani Han ttb
Hang Hangul ttb
Merc Meroitic Cursive ttb
Mero Meroitic Hieroglyphs ttb
Mong Mongolian 90° ttb
Ogam Ogham -90° btt
Orkh Old Turkic -90° ttb
Phag Phags Pa 90° ttb
Yiii Yi ttb

Exceptions: For the purposes of this specification, all fullwidth (F) and wide (W) characters are treated as belonging to a vertical script, and halfwidth characters (H) are treated as belonging to a horizontal script. [UAX11]

Note that for vertical-only characters (such as Mongolian and Phags Pa letters), the glyphs in the Unicode code charts are shown in their vertical orientation. In horizontal text, they are typeset in a 90° counter-clockwise rotation from this orientation.

Due to limitations in the current featureset of Unicode Technical Report 50 and CSS Writing Modes, vertical mixed typesetting cannot automatically handle either Ogham or Old Turkic. For these scripts, sideways-lr (in CSS Writing Modes Level 4) can be used to typeset passages.


Document conventions

Conformance requirements are expressed with a combination of descriptive assertions and RFC 2119 terminology. The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in the normative parts of this document are to be interpreted as described in RFC 2119. However, for readability, these words do not appear in all uppercase letters in this specification.

All of the text of this specification is normative except sections explicitly marked as non-normative, examples, and notes. [RFC2119]

Examples in this specification are introduced with the words “for example” or are set apart from the normative text with class="example", like this:

This is an example of an informative example.

Informative notes begin with the word “Note” and are set apart from the normative text with class="note", like this:

Note, this is an informative note.

Advisements are normative sections styled to evoke special attention and are set apart from other normative text with <strong class="advisement">, like this: UAs MUST provide an accessible alternative.

Conformance classes

Conformance to this specification is defined for three conformance classes:

style sheet
A CSS style sheet.
A UA that interprets the semantics of a style sheet and renders documents that use them.
authoring tool
A UA that writes a style sheet.

A style sheet is conformant to this specification if all of its statements that use syntax defined in this module are valid according to the generic CSS grammar and the individual grammars of each feature defined in this module.

A renderer is conformant to this specification if, in addition to interpreting the style sheet as defined by the appropriate specifications, it supports all the features defined by this specification by parsing them correctly and rendering the document accordingly. However, the inability of a UA to correctly render a document due to limitations of the device does not make the UA non-conformant. (For example, a UA is not required to render color on a monochrome monitor.)

An authoring tool is conformant to this specification if it writes style sheets that are syntactically correct according to the generic CSS grammar and the individual grammars of each feature in this module, and meet all other conformance requirements of style sheets as described in this module.

Requirements for Responsible Implementation of CSS

The following sections define several conformance requirements for implementing CSS responsibly, in a way that promotes interoperability in the present and future.

Partial Implementations

So that authors can exploit the forward-compatible parsing rules to assign fallback values, CSS renderers must treat as invalid (and ignore as appropriate) any at-rules, properties, property values, keywords, and other syntactic constructs for which they have no usable level of support. In particular, user agents must not selectively ignore unsupported property values and honor supported values in a single multi-value property declaration: if any value is considered invalid (as unsupported values must be), CSS requires that the entire declaration be ignored.

Implementations of Unstable and Proprietary Features

To avoid clashes with future stable CSS features, the CSSWG recommends following best practices for the implementation of unstable features and proprietary extensions to CSS.

Implementations of CR-level Features

Once a specification reaches the Candidate Recommendation stage, implementers should release an unprefixed implementation of any CR-level feature they can demonstrate to be correctly implemented according to spec, and should avoid exposing a prefixed variant of that feature.

To establish and maintain the interoperability of CSS across implementations, the CSS Working Group requests that non-experimental CSS renderers submit an implementation report (and, if necessary, the testcases used for that implementation report) to the W3C before releasing an unprefixed implementation of any CSS features. Testcases submitted to W3C are subject to review and correction by the CSS Working Group.

Further information on submitting testcases and implementation reports can be found from on the CSS Working Group’s website at https://www.w3.org/Style/CSS/Test/. Questions should be directed to the public-css-testsuite@w3.org mailing list.


Terms defined by this specification

Terms defined by reference


Normative References

Elika Etemad. CSS Box Model Module Level 3. 18 December 2018. WD. URL: https://www.w3.org/TR/css-box-3/
Elika Etemad; Tab Atkins Jr.. CSS Cascading and Inheritance Level 4. 28 August 2018. CR. URL: https://www.w3.org/TR/css-cascade-4/
Tab Atkins Jr.; Elika Etemad. CSS Display Module Level 3. 11 July 2019. CR. URL: https://www.w3.org/TR/css-display-3/
Dave Cramer; Elika Etemad; Steve Zilles. CSS Inline Layout Module Level 3. 8 August 2018. WD. URL: https://www.w3.org/TR/css-inline-3/
Dirk Schulze; Brian Birtles; Tab Atkins Jr.. CSS Masking Module Level 1. 26 August 2014. CR. URL: https://www.w3.org/TR/css-masking-1/
David Baron; Elika Etemad; Florian Rivoal. CSS Overflow Module Level 3. 31 July 2018. WD. URL: https://www.w3.org/TR/css-overflow-3/
Elika Etemad; Koji Ishii. CSS Ruby Layout Module Level 1. 5 August 2014. WD. URL: https://www.w3.org/TR/css-ruby-1/
Tab Atkins Jr.; Elika Etemad. CSS Intrinsic & Extrinsic Sizing Module Level 3. 22 May 2019. WD. URL: https://www.w3.org/TR/css-sizing-3/
Elika Etemad; Koji Ishii; Florian Rivoal. CSS Text Module Level 3. 12 December 2018. WD. URL: https://www.w3.org/TR/css-text-3/
Tab Atkins Jr.; Elika Etemad. CSS Values and Units Module Level 3. 6 June 2019. CR. URL: https://www.w3.org/TR/css-values-3/
Tab Atkins Jr.; Elika Etemad. CSS Values and Units Module Level 4. 31 January 2019. WD. URL: https://www.w3.org/TR/css-values-4/
Elika Etemad; Koji Ishii. CSS Writing Modes Level 4. 30 July 2019. CR. URL: https://www.w3.org/TR/css-writing-modes-4/
Bert Bos; et al. Cascading Style Sheets Level 2 Revision 1 (CSS 2.1) Specification. 7 June 2011. REC. URL: https://www.w3.org/TR/CSS2/
Rossen Atanassov; Elika Etemad. CSS Fragmentation Module Level 3. 4 December 2018. CR. URL: https://www.w3.org/TR/css-break-3/
Elika Etemad; Tab Atkins Jr.. CSS Image Values and Replaced Content Module Level 3. 17 April 2012. CR. URL: https://www.w3.org/TR/css3-images/
Elika Etemad; Koji Ishii. CSS Text Decoration Module Level 3. 13 August 2019. CR. URL: https://www.w3.org/TR/css-text-decor-3/
Bert Bos; Elika Etemad; Brad Kemper. CSS Backgrounds and Borders Module Level 3. 17 October 2017. CR. URL: https://www.w3.org/TR/css-backgrounds-3/
Florian Rivoal; Rachel Andrew. CSS Multi-column Layout Module Level 1. 28 May 2018. WD. URL: https://www.w3.org/TR/css-multicol-1/
Elika Etemad; Simon Sapin. CSS Paged Media Module Level 3. 18 October 2018. WD. URL: https://www.w3.org/TR/css-page-3/
Anne van Kesteren; et al. HTML Standard. Living Standard. URL: https://html.spec.whatwg.org/multipage/
S. Bradner. Key words for use in RFCs to Indicate Requirement Levels. March 1997. Best Current Practice. URL: https://tools.ietf.org/html/rfc2119
R. Barnes; S. Kent; E. Rescorla. Further Key Words for Use in RFCs to Indicate Requirement Levels. 1 April 2013. Experimental. URL: https://tools.ietf.org/html/rfc6919
Erik Dahlström; et al. Scalable Vector Graphics (SVG) 1.1 (Second Edition). 16 August 2011. REC. URL: https://www.w3.org/TR/SVG11/
Ken Lunde 小林劍󠄁. East Asian Width. 25 January 2019. Unicode Standard Annex #11. URL: https://www.unicode.org/reports/tr11/tr11-36.html
Ken Whistler. Unicode Script Property. 6 February 2019. Unicode Standard Annex #24. URL: https://www.unicode.org/reports/tr24/tr24-29.html
Koji Ishii 石井宏治; Ken Lunde 小林劍󠄁. Unicode Vertical Text Layout. 4 February 2019. Unicode Standard Annex #50. URL: https://www.unicode.org/reports/tr50/tr50-22.html
Mark Davis; Aharon Lanin; Andrew Glass. Unicode Bidirectional Algorithm. 4 February 2019. Unicode Standard Annex #9. URL: https://www.unicode.org/reports/tr9/tr9-41.html
The Unicode Standard. URL: https://www.unicode.org/versions/latest/

Informative References

Rossen Atanassov; Elika Etemad. CSS Fragmentation Module Level 4. 18 December 2018. WD. URL: https://www.w3.org/TR/css-break-4/
Tab Atkins Jr.; et al. CSS Flexible Box Layout Module Level 1. 19 November 2018. CR. URL: https://www.w3.org/TR/css-flexbox-1/
John Daggett; Myles Maxfield; Chris Lilley. CSS Fonts Module Level 3. 20 September 2018. REC. URL: https://www.w3.org/TR/css-fonts-3/
Dave Raggett; Arnaud Le Hors; Ian Jacobs. HTML 4.01 Specification. 27 March 2018. REC. URL: https://www.w3.org/TR/html401/
Elika J. Etemad. Robust Vertical Text Layout. 25 April 2005. Unicode Technical Note #22. URL: https://unicode.org/notes/tn22/

Property Index

Name Value Initial Applies to Inh. %ages Ani­mat­able Anim­ation type Canonical order Com­puted value
direction ltr | rtl ltr all elements yes n/a not animatable n/a specified value
glyph-orientation-vertical auto | 0deg | 90deg | 0 | 90 n/a n/a n/a n/a n/a n/a n/a
text-combine-upright none | all none non-replaced inline elements yes n/a not animatable n/a specified keyword
text-orientation mixed | upright | sideways mixed all elements except table row groups, rows, column groups, and columns yes n/a not animatable n/a specified value
unicode-bidi normal | embed | isolate | bidi-override | isolate-override | plaintext normal all elements, but see prose no n/a not animatable per grammar specified value
writing-mode horizontal-tb | vertical-rl | vertical-lr horizontal-tb All elements except table row groups, table column groups, table rows, table columns, ruby base container, ruby annotation container yes n/a not animatable n/a specified value