CSL 1.0 Specification

by Rintze M. Zelle, PhD

with contributions from Frank G. Bennett, Jr. and Bruce D’Arcus.



The Citation Style Language (CSL) is an XML format for describing the formatting of in-text citations, notes and bibliographies. CSL offers:

  • An open format that may be used by any application
  • The ability to write compact and robust styles
  • Extensive support for style requirements
  • Automatic style localization
  • Easy distribution and updating of styles
  • A fast growing library with thousands of freely available styles

This document is meant as a complete and accurate specification of CSL 1.0. Additional documentation, such as the CSL schema, CSL styles, and information on how to add CSL support to applications, can be found at the official home of CSL, citationstyles.org.

CSL Styles - Basic Structure


All elements in CSL are namespaced. The recommended prefix cs is attached to element names throughout this specification, but is usually omitted from CSL styles when the default namespace is declared in the root cs:style element.

The CSL namespace

XML Declaration

It is highly recommended to initialize each CSL style with an XML declaration, specifying the version of XML used as well as the character encoding. In most cases, the declaration will be:

<?xml version="1.0" encoding="UTF-8"?>

The Root Element - cs:style

The root element of a CSL style is cs:style. This element carries the following arguments:

Specifies whether the style uses notes (value “note”) or in-text citations (“in-text”).
default-locale (optional)
Fixes style localization to the locale code specified. This is desirable for most journal styles.
Indicates with which version of the CSL schema the style is compatible. Should have a value of “1.0” for CSL 1.0-compatible styles.
xmlns (optional)
The namespace declaration that binds the elements in the style to the given namespace URI. CSL elements in the style don’t need individual namespace declarations if this attribute is set to “http://purl.org/net/xbiblio/csl”.

In addition, cs:style may carry any of the global options, as well as the inheritable name options.

An example of a style preamble:

<?xml version="1.0" encoding="UTF-8"?>
<style xmlns="http://purl.org/net/xbiblio/csl" version="1.0" class="in-text" default-locale="fr-FR">

Child Elements of cs:style

All independent CSL styles share the same basic structure, with five possible types of child elements in the cs:style root element. The roles of each of these elements (which will be described in more detail later) are:

The cs:info element must be included as the first child element of cs:style. It contains the metadata that describes the style (the name of the style, a unique style identifier, the style authors, etc.).
This required element describes the formatting instructions of in-text citations or notes.
cs:bibliography (optional)
This optional element describes the formatting instructions of the bibliography.
cs:macro (optional)
Styles may include one or more cs:macro elements. Macros allow for reuse of formatting instructions, which helps keeping styles compact and maintainable.
cs:locale (optional)
Styles may include one or more cs:locale elements. These elements allow styles to override the default localization data (terms, date formats and formatting options) on a per-locale basis.


The cs:info element contains all the style’s metadata, many elements of which are borrowed from the Atom Syndication Format. Although it has no influence on how citations are formatted, complete and correct metadata is important if styles are made publicly available. Below is an example of a cs:info element, followed by a description of all possible elements.

 <title>Style Title</title>
 <link href="http://www.zotero.org/styles/style-title" rel="self"/>
  <name>Author Name</name>
  <email>[email protected]</email>
 <category citation-format="author-date"/>
 <category field="zoology"/>
 <summary>Style for Some Journal</summary>
 <rights>This work is licensed under a Creative Commons
         Attribution-Share Alike 3.0 Unported License
cs:author and cs:contributor (optional)
One or more of these elements may be used to acknowledge style authors and contributors. Authorship is generally limited to those who have written a new style, or have made significant changes to existing styles, while contributorship can be assigned to those who have made small changes. Both elements require one child element, cs:name, and allow for two others, cs:email, and cs:uri, indicating respectively the name, email address and URI of the author or contributor in question.
cs:category (optional)

Styles may be assigned one or more categories. This information can be used to organize style repositories. Two types of categories exist. The first category type describes how in-text citations are rendered. For this type the citation-format attribute is set to one of the following values:

  • “author-date”: e.g. “… (Doe, 1999)”
  • “author”: e.g. “… (Doe)”
  • “numeric”: e.g. “… [1]”
  • “label”: e.g. “… [doe99]”
  • “note”: the citation appears as a footnote or endnote

The second category type indicates the fields or disciplines for which the style is relevant. For this category type the field attribute is set to one of the discipline categories.

This required element should contain a URI. This identifier establishes the identity of the style. A valid, stable, and unique URL that resolves to the style is desired if the style is made publicly available. Keeping the same URI is crucial for applications that support automatic style updating.
cs:issn/cs:issnl (optional)
Journal-specific styles may include one or more cs:issn elements, containing the journal’s ISSN identifiers (multiple ISSNs can be assigned to a single journal, e.g. for the print and online editions). In addition, the cs:issnl element may be used for the newly established ISSN-L identifier.
cs:link (optional)

The cs:link element is used to specify a URI (usually a URL), which is set on the href attribute. The accompanying rel attribute must be set to indicate the relation of the URI to the style. The possible values of rel:

  • “self”: if the URI is that of the CSL style itself. Needed for automatic style updating.
  • “independent-parent”: if the URI is that of the parent CSL style, the content of which should be used for the citation formatting. Needed for dependent styles.
  • “template”: if the URI is that of the CSL style from which the current independent style is derived. May be used to indicate style parentage.
  • “documentation: if the URI points to the online style documentation.

The cs:link element may contain textual content to describe the link, and may carry the xml:lang attribute to specify the language of either the link description or of the link target (the value should be a xsd:language locale code).

cs:published (optional)
The contents of this element must be a timestamp. This timestamp indicates when the style was initially created or made available.
cs:rights (optional)
This element specifies the license under which the style file is released. See, e.g. the Creative Commons. The element may include a xml:lang attribute to specify the language of the content (the value should be an xsd:language locale code).
cs:summary (optional)
This element gives a summary of the style.The element may include a xml:lang attribute to specify the language of the content (the value should be an xsd:language locale code).
The contents of this required element should be the name of the style as it should be shown to users. The element may include a xml:lang attribute to specify the language of the content (the value should be an xsd:language locale code).
The contents of this required element must be a timestamp. This timestamp is used for automatic updating of styles.


The cs:citation element describes the formatting of citations, which can consist of one or multiple references to bibliographic sources, and may appear in the form of either in-text citations (generally formatted as label [doe99], number [1], author [Doe] or author-date descriptors [Doe 1999]) or notes. The required cs:layout child element describes what, and how, bibliographic data should be included in the citations (see the chapter on the Layout element). The cs:citation element may carry attributes for citation-specific formatting options and inheritable name options (see the Citation-specific Options and Inheritable Name Options sections). Finally, the optional cs:sort element, which should precede the cs:layout element, specifies how citations consisting of multiple references should be sorted (see the chapter on Sorting). An example of a cs:citation element:

<citation option="option-value">
    <!-- sort keys -->
    <!-- rendering elements -->

A note to developers of CSL processors Note styles are unique in that citations may effectively become full sentences. Because of this, the first character of the output should be uppercased when a citation is footnoted without any additional text. By contrast, if the citation occurs within a pre-existing footnote, and is preceded by non-citation text, then it should be printed as is.


The cs:bibliography element describes the formatting of bibliographies, and is used in a similar way as cs:citation: the required cs:layout child element describes how each reference should be formatted in the bibliography, while the optional cs:sort element (which should precede the cs:layout element) specifies the sorting order of the references in the bibliography. In addition, the ``cs:bibliography element may carry attributes for bibliography-specific formatting options and inheritable name options (see the Bibliography-specific Options and Inheritable Name Options sections).

<bibliography option="option-value">
    <!-- sort keys -->
    <!-- rendering elements -->


Macros, which are defined using cs:macro elements, can contain the same set of rendering elements that are available within cs:layout inside cs:citation or cs:bibliography. Macros allow formatting instructions to be reused, both within the same style (e.g. the same macro could be used in both cs:citation and cs:bibliography) as well as between styles. It is therefore recommended to use common macro names as much as possible. Correct use of macros can greatly improve the readability, compactness and maintainability of styles. Ideally, the contents of cs:citation and cs:bibliography should be kept compact and agnostic of resource types (i.e. books, journal articles, etc.), depending mainly on macro calls.

By convention, macros are placed after any cs:locale elements and before the cs:citation element. The cs:macro element must carry the name attribute (the value of which is used to identify the macro), and contain one or more rendering elements. Once defined, macros can be called by rendering elements in cs:citation or cs:bibliography (from within cs:layout), or by the rendering elements in other macros.

The following example shows a style that call the “title” macro. This macro outputs the contents of the title variable, applying italics when the resource type is “book”:

  <macro name="title">
      <if type="book">
        <text variable="title" font-style="italic"/>
        <text variable="title"/>
      <text macro="title"/>


CSL supports localization of terms, date formats and formatting options. Default localization data for several tens of locales is provided through “locales-xx-XX.xml” files (“xx-XX” represents the locale code, e.g. “en-US” for US English). Localization data can also be included in styles, using one or more of the optional cs:locale elements, which by convention are included directly after the cs:info element. For each cs:locale element, the relevant locale can be indicated with the xml:lang attribute (set to an xsd:language locale code). If the attribute is absent, the cs:locale element’s localization data will apply to all locales.

See Localized Terms, Localized Dates and Localized Options in the Locale Files - Basic Structure section for more details on the use of cs:locale.

An example illustrating the use of cs:locale in a CSL style:

  <locale xml:lang="en">
      <term name="editortranslator" form="short">
        <single>ed. &amp; trans.</single>
        <multiple>eds. &amp; trans.</multiple>
Locale Prioritization

Locale codes can indicate either the language (e.g. “en” for English) or the language dialect (e.g. “en-US” for American English and “en-GB” for British English). While the “locales-xx-XX.xml” files are only maintained for language dialects, the (optional) xml:lang attribute on cs:locale in styles may be set to languages as well as language dialects. The “locales-xx-XX.xml” files must contain the full set of localization data. The cs:locale elements are typically only used in styles to redefine the localization data provided via the “locales-xx-XX.xml” files, and these may include only the localization data that should be redefined. The existence of two types of locale codes (languages and language dialects), and the ability to define localization both in “locales-xx-XX.xml” files and in styles, requires prioritization of localization data. This prioritization can best be illustrated with an example. If a CSL processor is asked to localize the style output to “de-AT” (Austrian German), the priority of the locale data is as follows:

Localization data specified in styles using cs:locale

  1. xml:lang set to “de-AT” (Austrian German)
  2. xml:lang set to “de” (German)
  3. xml:lang not set (all locales)

Localization data stored in “locales-xx-XX.xml” files

  1. xml:lang set to “de-AT” (Austrian German)
  2. xml:lang set to “de-DE” (Standard German)
  3. xml:lang set to “en-US” (American English)

Thus, when a style does not include a cs:locale element for the “de-AT” locale, or when it exists but is incomplete, the missing localization data is retrieved from the cs:locale set to “de” (if present). If the set of localization data is still incomplete, the cs:locale element without a xml:lang element is used (if present). The localization data is completed with the data stored in the “locales-xx-XX.xml” files. If the file for “de-AT” does not exist, fallback locales consist of “de-DE” and, as a last resort, “en-US”. Note that locale substitution is only activated when a term is not defined. It does not occur when a term is defined, but consists of an empty string (e.g. <term name=”and”/> or <term name=”and”></term>).

Dependent Styles

In addition to independent styles, which are self-contained, CSL also supports dependent styles, which function like aliases or shortcuts. Dependent styles can be used when multiple journals share the same style format. In such a case, it is sufficient to create a single independent master style for the format (e.g. “Nature Journals”). Dependent styles, which only contain a cs:info element, can then be added for all journals that use this format (e.g. “Nature Biotechnology”, “Nature Nanotechnology”, etc.). With this approach, style repositories can show entries for the individual journals, without the need to duplicate formatting instructions. If the common format has to be modified, it is sufficient to change the independent master style, which makes style maintainance simpler and faster.

The cs:info element of dependent styles should provide the metadata of the individual journals. A cs:link element which points to the independent master style must be included (for this the rel attribute on the relevant cs:link element should be set to “independent-parent”, see also Info).

N.B. Dependent styles cannot be used to indicate changes compared to the independent master style. If there is any difference in formatting between two styles, however small, separate independent styles have to be created.

Locale Files - Basic Structure

CSL ships with a number of “locales-xx-XX.xml” files (the “xx-XX” is the locale code, e.g. “en-US” for US English). While localization data can also be specified in styles (see Locale), locale files conveniently provide complete sets of default localization data (terms, dates and formatting options).

The locale files, which like styles are written in XML, each contain the localization data for a single locale. The cs:locale root element require two attributes: xml:lang, to specify the locale of the data, and version, to indicates with which version of the CSL schema the locale file is compatible (this attribute should have a value of “1.0” for CSL 1.0-compatible styles). The root element also typically carries a xmlns namespace declaration, set to the CSL namespace (“http://purl.org/net/xbiblio/csl”). The cs:locale element has three required child elements, which are described in the sections below: cs:terms, cs:date and cs:style-options. An example of the (incomplete) contents of a locale file:

<?xml version="1.0" encoding="UTF-8"?>
<locale xml:lang="en-US" version="1.0" xmlns="http://purl.org/net/xbiblio/csl">
    <term name="no date">n.d.</term>
    <term name="et-al">et al.</term>
    <term name="page">
    <term name="page" form="short">
  <date form="text">
    <date-part name="month" suffix=" "/>
    <date-part name="day" suffix=", "/>
    <date-part name="year"/>
  <date form="numeric">
    <date-part name="year"/>
    <date-part name="month" form="numeric" prefix="-" range-delimiter="/"/>
    <date-part name="day" prefix="-" range-delimiter="/"/>
  <style-options punctuation-in-quote="true"/>

Localized Terms

Terms are localized strings. For example, if a style specifies that the term “and” should be used, the string that appears in the style output depends on the locale: “and” for English, “und” for German. Terms are defined using cs:term elements, child elements of cs:terms, itself a child element of cs:locale. Terms are identified by the value of the name attribute of cs:term. Two types of terms exist: simple terms, where the content of the cs:term is the localized string, and compound terms, where cs:term includes the two child elements cs:single and cs:multiple, which respectively contain the singular and plural variant of the term (e.g. “page” and “pages”). Some terms are defined for multiple forms. In these cases, multiple cs:term element share the same value of name, but differ in the value of the optional form attribute. The different forms are:

  • “long” - the default, e.g. “editor” and “editors” for the term “editor”
  • “short” - e.g. “ed” and “eds” for the term “editor”
  • “verb” - e.g. “edited by” for the term “editor”
  • “verb-short” - e.g. “ed” for the term “editor”
  • “symbol” - e.g. “§” for the term “section”

Examples of how terms are defined have been given above (Locale Files - Basic Structure). The complete list of terms can be found in Appendix III - Terms.

Localized Dates

Styles can use either localized or non-localized date formats. Localized date formats are defined with the cs:date element as child element of cs:locale. The required form attribute on cs:date must be set to either “numeric” (for numeric date formats, e.g. “12-15-2005”) or to “text” (e.g. “December 15, 2005”). A date format is then defined by the child elements of cs:date, the cs:date-part elements. These must carry the name attribute, set to day, month or year. The order of the cs:date-part elements is also the display order. Additional formatting can be achieved by setting formatting attributes on the cs:date and cs:date-part elements, as well as a number of attributes that are specific to cs:date-part (see Date-part). In addition, a delimiter may be set on cs:date to delimit the cs:date-part elements, and affixes may be applied to the cs:date-part elements.

N.B. Affixes are not allowed on cs:date when used as a child element of cs:locale. This helps in separating locale-specific affixes (which should be set on the cs:date-part elements) from any style-specific affixes (such as parentheses, which should be set on the cs:date rendering element). E.g. a macro could specify:

<macro name="issued">
 <date variable="issued" form="numeric" prefix="(" suffix=")"/>

Localized Options

CSL 1.0 includes a single localized global option (affecting both citation and bibliography output), punctuation-in-quote (see Locale Options). This option is set as an attribute on cs:style-options, a child element of cs:locale.

Rendering Elements

Rendering elements are used to specify which, and in what order, bibliographic data should be included in citations and bibliographies. Rendering elements also partly control the formatting of this data.


As discussed in the citation and bibliography sections, cs:layout is a required child element of both cs:citation and cs:bibliography. All the rendering elements that should appear in the citations and bibliography should be nested inside the cs:layout element. Itself a rendering element, cs:layout accepts both affixes and formatting attributes. When used in the cs:citation element, a delimiter can be set to separate multiple bibliographic items in a single citation. For example, citations like “(1, 2)” can be produced with:

<layout prefix="(" suffix=")" delimiter=", ">
  <text variable="citation-number"/>


The cs:text element is used to output text, which can originate from different sources. The source-type is indicated with an attribute, and the attribute value acts as an identifier within the source-type. For example,

<text variable="title" form="short" font-style="italic"/>

indicates that the source-type is a variable, and that the variable that should be displayed is the italicized short form of “title”. The different source-types are:

  • variable - the text contents of a variable (see Standard Variables). The optional form attribute can be set to either “long” (the default) or “short” to select the long or short forms of variables, e.g. the full and short title.
  • macro - the text generated by a macro. The value of macro should correspond to the value of the name attribute of the desired cs:macro element.
  • term - the text of a localized term (see Appendix III - Terms and Locale). The plural attribute can be set to choose either the singular (value “false”, the default) or plural variant (value “true”) of a term. In addition, the form attribute can be set to select the desired term form (“long” [default], “short”, “verb”, “verb-short” or “symbol”). If for a given term the desired form does not exist, another form may be used: “verb-short” reverts to “verb”, “symbol” reverts to “short”, and “verb” and “short” both revert to “long”.
  • value - used to output verbatim text, which is set via the value of value (e.g. value=”some text”)

In all cases the attributes for affixes, display, formatting, quotes, strip-periods and text-case may be applied to cs:text.


The cs:date element is used to output dates, in either a localized or a non-localized format. The desired date variable (see Date Variables) is selected with the variable attribute.

Localized date formats are selected with the form attribute. This attribute can be set to “numeric” (for numeric date formats, e.g. “12-15-2005”), or to “text” (for date formats with a non-numeric month, e.g. “December 15, 2005”). Localized dates can be customized in two ways. First, the date-parts attribute may be used to specify which cs:date-part elements are shown. The possible values are:

  • “year-month-day” - default, displays year, month and day
  • “year-month” - displays year and month
  • “year” - displays year only

Secondly, cs:date may include one or more cs:date-part elements (see Date-part). The attributes set on these elements override those originally specified for the localized date formats (e.g. the form attribute of the month-cs:date-part element can be set to “short” to get abbreviated month names in all locales.). Note that the use of cs:date-part elements for localized dates does not affect which, and in what order, the cs:date-part elements are included in the rendered date. Also, the cs:date-part elements may not carry the attributes for affixes, as these are considered to be locale-specific.

Non-localized date formats are self-contained: the date format is entirely controlled by cs:date and its cs:date-part children. In contrast to localized dates, cs:date is used without the form and date-parts attributes. Only the included cs:date-part elements will be rendered, in the order in which they are specified. The cs:date-part elements may carry attributes for both affixes and formatting, while cs:date may carry a delimiter (delimiting the various cs:date-part elements).

For both localized and non-localized dates, affixes, display and formatting attributes may be specified for the cs:date element.


The cs:date-part element is used to control how the different date parts of the date variable specified in the parent cs:date element are rendered. The date parts are identified by the value of the name attribute, which can be:


For day, cs:date-part may carry the form attribute, with values:

  • “numeric” - default, e.g. “1”
  • “numeric-leading-zeros” - e.g. “01”
  • “ordinal” - e.g. “1st”

For month, cs:date-part may carry the strip-periods and form attributes. Abbreviated months (e.g. “Jan.”, “Feb.”) are localized terms and include periods by default (if applicable). These periods are removed when strip-periods is set to “true” (“false” is the default). The form attribute can be set to:

  • “long” - default, e.g. “January”
  • “short” - e.g. “Jan.”
  • “numeric” - e.g. “1”
  • “numeric-leading-zeros” - e.g. “01”

For year, cs:date-part may carry the form attribute, with values:

  • “long” - default, e.g. “2005”
  • “short” - e.g. “05”

All cs:date-part elements may carry the formatting and text-case attributes. Attributes for affixes are also allowed, except when cs:date is used to call a localized date format. Finally, the cs:date-part elements may carry the range-delimiter attribute (see Date Ranges).

Date Ranges

By default, date ranges are delimited by an en-dash (e.g. “May–July 2008”). The range-delimiter attribute can be used to specify custom date range delimiters. The attribute value set on the largest date-part (“day”, “month” or “year”) that differs between the two dates of the date range will then be used instead of the en-dash. For example,

      <date variable="issued">
        <date-part name="month" suffix=" "/>
        <date-part name="year" range-delimiter="/"/>

would result in “May–July 2008” and “May 2008/June 2009”.

AD and BC

The terms ad and bc (Anno Domini and Before Christ) are automatically appended to years: bc is added to negative years (e.g. 2500BC), while ad is added to positive years of less than four digits (79AD).


If a date includes a season instead of a month, a season term (season-01 to season-04, respectively Spring, Summer, Autumn and Winter) will substituted the month date-part. E.g.,

      <date variable="issued">
        <date-part name="month" suffix=" "/>
        <date-part name="year"/>

would result in “May 2008” and “Winter 2009”.

Uncertain Dates

Uncertain dates can receive special formatting by using the is-uncertain-date conditional (see Choose) and the “circa” term. The conditional tests “true” when a date is flagged as uncertain. For example,

    <layout delimiter="; ">
        <if is-uncertain-date="issued">
          <text term="circa" form="short" suffix=" "/>
      <date variable="issued">
        <date-part name="year"/>

would result in “2005” (normal certain date) and “ca. 2003” (uncertain date).


The cs:number element can be used to output any of the following variables (selected with the variable attribute):

  • “edition”
  • “volume”
  • “issue”
  • “number”
  • “number-of-volumes”

Although these variables can also be rendered with cs:text, cs:number has the benefit of offering number-specific formatting via the form attribute, with values:

  • “numeric” (default) - e.g. “1”, “2”, “3”
  • “ordinal” - e.g. “1st”, “2nd”, “3rd”
  • “long-ordinal” - e.g. “first”, “second”, “third”
  • “roman” - e.g. “i”, “ii”, “iii”

If a variable displayed with cs:number contains a mixture of numeric and non-numeric text, only the first number encountered is used for rendering (e.g. “12” when the entire string is “12th edition”). If a variable only contains non-numeric text (e.g. “special edition”), the entire string is rendered, as if cs:text were used instead. Fields can be tested for containing numeric content with the is-numeric conditional, e.g. “12th edition” would test “true” while “third edition” would test “false” (see Choose).

The cs:number element may carry any of the affixes, display, formatting and text-case attributes.


The cs:names element can be used to display the contents of one or more name variables, each of which can contain multiple names (e.g. the “author” variable will contain all the cited item’s author names). The variables to be displayed are set with the variable attribute. If multiple variables are selected (separated by single spaces, see example below), each variable is independently rendered in the order specified, with one exception: if the value of variable consists of “editor” and “translator” (in either order), and if the contents of the two name variables is identical, then the contents of only one name variable is rendered. In addition, the “editor-translator” term is used if the cs:names element contains a cs:label element, replacing the default “editor” and “translator” terms (e.g., this might result in “Doe (editor & translator)”. The delimiter attribute may be set on cs:names to delimit the names of the different name variables (e.g. the semicolon in “Doe (editor); Johnson (translator)”).

<names variable="editor translator" delimiter="; ">
  <label prefix=" (" suffix=")"/>

There are four child elements associated with the cs:names element: cs:name, cs:et-al, cs:substitute and cs:label (all discussed below). In addition, the cs:names element may carry the attributes for affixes, display and formatting.


The cs:name element is a required child element of cs:names, and describes both how individual names are formatted, and how names within a name variable are separated from each other. The attributes that may be used on cs:name are:

This attribute specifies the delimiter between the second to last and the last name of the names in a name variable. The value of the attribute may be either “text”, which selects the “and” term, or “symbol”, which selects the ampersand (&).
Specifies the text string to separate names of a name variable. The default value is “, ” (“J. Doe, S. Smith”).

Determines in which cases the delimiter used to delimit names is also used to separate the second to last and the last name in name lists. The possible values are:

  • “contextual” (default): the delimiter is only included for name lists with three or more names
    • 2 names: “J. Doe and T. Williams,”
    • 3 names: “J. Doe, S. Smith, and T. Williams”
  • “always”: the delimiter is always included
    • 2 names: “J. Doe, and T. Williams”
    • 3 names: “J. Doe, S. Smith, and T. Williams”
  • “never”: the delimiter is never included
    • 2 names: “J. Doe and T. Williams,”
    • 3 names: “J. Doe, S. Smith and T. Williams”
et-al-min / et-al-use-first
Together, these attributes control et-al abbreviation. When the number of names in a name variable matches or exceeds the number set on et-al-min, the rendered name list is truncated at the number of names set on et-al-use-first. If truncation occurs, the “et-al” term is appended to the names rendered (see also Et-al). With a single name (et-al-use-first=”1”), the “et-al” term is preceded by a space (e.g. “Doe et al.”). With multiple names, the “et-al” term is preceded by the name delimiter (e.g. “Doe, Smith, et al.”).
et-al-subsequent-min / et-al-subsequent-use-first
The (optional) et-al-min and et-al-use-first attributes take effect for all cites and bibliographic entries. With the et-al-subsequent-min and et-al-subsequent-use-first attributes divergent et-al abbreviation rules can be specified for subsequent cites (cites referencing earlier cited items).

The remaining attributes, discussed below, only affect personal names. Personal names require a “family” name-part, and may also contain “given”, “suffix”, “non-dropping-particle” and “dropping-particle” name-parts. The roles of these name-parts, which are delimited by single spaces in rendered names, are:

  • “family”: the surname minus any particles and suffixes
  • “given”: the given names, which may be either full (“John Edward”) or initialized (“J. E.”)
  • “suffix”: name suffix, e.g. “Jr.” in “John Smith Jr.” and “III” in “Bill Gates III”
  • “non-dropping-particle”: name particles that are not dropped when only the last name is shown (“de” in the Dutch surname “de Koning”) but which may be treated as a separate object from the family name (e.g. for sorting)
  • “dropping-particle”: name particles that are dropped when only the surname is shown (“van” in “Ludwig van Beethoven”, which becomes “Beethoven”)
Specifies whether all the name-parts of personal names should be displayed (value “long”), or only the family name and the non-dropping-particle (value “short”). A third value, “count”, returns the total number of names that would be otherwise displayed by the use of the cs:names element (taking into account the effects of et-al abbreviation and editor/translator collapsing), and may be used for advanced sorting.
If this attribute is set, given names are converted to initials. The attribute value specifies the suffix that is included after each initial (“.” results in “J.J. Doe”). Note that the global initialize-with-hyphen option controls how compound given names (e.g. “Jean-Luc”) are hyphenated when initialized (see Hyphenation of Initialized Names).

Specifies that names should be displayed with the given name following the family name (e.g. “John Doe” becomes “Doe, John”). The attribute may have one of the two values:

  • “first”: name-as-sort-order applies to the first name in each name variable
  • “all”: name-as-sort-order applies to all names

Note that the sort order of names may differ from the display order for names containing particles and suffixes (see Name-part order). Also, this attribute only affects names written in the latin or Cyrillic alphabet. Names written in other alphabets (e.g. Asian scripts) are always shown with the family name preceding the given name.

Sets the delimiter for name-parts that have switched positions as a result of name-as-sort-order. The default value is “, ” (“Doe, John”). As is the case for name-as-sort-order, this attribute only affects names written in the latin or Cyrillic alphabet.

The cs:name element may also carry any of the attributes for affixes and formatting.

Name-part Order

The order of name-parts depends on the values of the form and name-as-sort-order attributes on cs:name, the value of the demote-non-dropping-particle attribute on cs:style (one of the global options), and the alphabet of the individual name. Note that the display and sorting order of name-parts often differs. An overview of the different orders:

Display order of latin/Cyrillic names


form set to “long”

  1. given
  2. dropping-particle
  3. non-dropping-particle
  4. family
  5. suffix

[Gérard] [de] [la] [Martinière] [III]


form set to “long”, name-as-sort-order active, demote-non-dropping-particle set to “never” or “sort-only”

  1. non-dropping-particle
  2. family
  3. given
  4. dropping-particle
  5. suffix

[la] [Martinière], [Gérard] [de], [III]


form set to “long”, name-as-sort-order active, demote-non-dropping-particle set to “display-and-sort”

  1. family
  2. given
  3. dropping-particle
  4. non-dropping-particle
  5. suffix

[Martinière], [Gérard] [de] [la], [III]


form set to “short”

  1. non-dropping-particles
  2. family

[la] [Martinière]

Sorting order of latin/Cyrillic names

N.B. The sort keys are listed in descending order of importance.


demote-non-dropping-particle set to “never”

  1. non-dropping-particle + family
  2. dropping-particle
  3. given
  4. suffix

[la Martinière] [de] [Gérard] [III]


demote-non-dropping-particle set to “sort-only” or “display-and-sort”

  1. family
  2. dropping-particle + non-dropping-particle
  3. given
  4. suffix

[Martinière] [de la] [Gérard] [III]

Display and sorting order of non-latin/Cyrillic names


form set to “long”

  1. family
  2. given

毛泽东 [Mao Zedong]


form set to “short”

  1. family

毛 [Mao]

Non-personal names lack name-parts and are sorted as is, although English articles (“a”, “an” and “the”) at the start of the name are stripped. For example, “The New York Times” sorts as “New York Times”.

Name-part Formatting

The cs:name element may include one or two cs:name-part child elements. These child elements accept the formatting and text-case attributes, which allows for separate formatting of the different name parts (e.g. “Jane DOE”, see example below). The required name attribute on cs:name-part specifies which name-parts are affected: when set to “given”, the formatting only acts on the “given” name-part. When set to “family”, the formatting acts on the “family”, “dropping-particle” and “non-dropping-particle” name-parts (the “suffix” name-part is not subject to any name-part formatting). The order of the cs:name-part elements does not affect which, and in what order, the name-parts are rendered.

<names variable="author">
    <name-part name="family" text-case="uppercase">


Et-al abbreviation, controlled via the et-al attributes on cs:name (see Name), can be further customized with the optional cs:et-al element, which should be included directly after the cs:name element. The term attribute of this element can be set to either “et-al” (default) or to “and others” to use either term (with this different et-al terms can be used for citations and the bibliography). In addition, attributes for affixes and formatting can be used, for example to italicize the et-al term:

<names variable="author">
  <et-al term="and others" font-style="italic"/>


The optional cs:substitute element, which should be included as the last child element of cs:names, controls substitution in case the name variables specified in the parent cs:names element are empty. The substitutions are specified as child elements of cs:substitute, and can consist of any of the standard rendering elements (with the exception of cs:layout). It is also possible to use a shorthand version of cs:names, which doesn’t allow for any child elements, and uses the attributes values set on the cs:name and cs:et-al child elements of the original cs:names element. If cs:substitute contains multiple child elements, the first element to return a non-empty result is used for substitution. Substituted variables are repressed in the rest of the output to prevent duplication. An example, where an empty “author” name variable is substituted by the “editor” name variable, or, when no editors exist, by the “title” macro:

<macro name="author">
  <names variable="author">
      <names variable="editor"/>
      <text macro="title"/>

Label in cs:names

The cs:label element, used to output text terms whose pluralization depends on the contents of another variable (e.g. “(editors)” in “Doe and Smith (editors)”), is discussed in detail in the label section. It should be included after the cs:name and cs:et-al elements, but before the cs:substitute element. When used within cs:names, the variable attribute should be omitted, as the value set on the parent cs:names element is used.


The Citation Style Language includes several variables that have matching terms. The cs:label element can be used to render one of these terms, while matching the term plurality with that of the corresponding variable. The variable/term combination is selected with the variable attribute, which can be set to either “page” or “locator”. When cs:label is used as a child element of cs:names, the value of the variable attribute is automatically inherited from the parent cs:names element. The example below displays the “page” variable, using the singular form of the “page” term for a single page (“page 5”), or the plural form for a page range (“pages 5-7”).

<group delimiter=" ">
  <label variable="page" form="long"/>
  <text variable="page"/>

The cs:label element may carry attributes for affixes, formatting, text-case and strip-periods, as well as:


Selects the form of the term, with possible values:

  • “long”: the default, e.g. “editor”/”editors” for the “editor” term
  • “verb”: e.g. “edited by” for the “editor” term
  • “short”: e.g. “ed”/”eds” for the “editor” term
  • “verb-short”: e.g. “ed” for the “editor” term
  • “symbol”: e.g. “§” for the singular “section” term

Sets pluralization of the term, with values:

  • “contextual”: the default, pluralization is dependent on the variable contents, e.g. “page 1” and “pages 1-3”
  • “always”: always use the plural form, e.g. “pages 1” and “pages 1-3”
  • “never”: always use the singular form, e.g. “page 1” and “page 1-3”


The cs:group element may contain one or more rendering elements (not cs:layout). cs:group itself may carry the delimiter attribute (to delimit the enclosed elements) and the attributes for affixes (applied to the group output as a whole), display and formatting (formatting settings are transmitted to the enclosed elements). Note that cs:group implicitly acts as a conditional: cs:group and its child elements are suppressed if a) at least one rendering element in cs:group calls a variable (either directly or via a macro), and b) all variables that are called are empty. This behavior exists to accommodate descriptive cs:text elements. For example,

  <group prefix="(" suffix=")">
    <text value="Published by: "/>
    <text variable="publisher"/>

results in “(Published by: Company A)” when the “publisher” variable is set to “Company A”, but doesn’t generate output when the “publisher” variable is empty.


Similarly to the conditional statements encountered in programming languages, the cs:choose element allows for the conditional rendering of rendering elements. An example is shown below:

  <if type="book thesis" match="any">
    <text variable="title" font-style="italic">
  <else-if type="chapter">
    <text variable="title" quotes="true">
    <text variable="title">

cs:choose requires a cs:if child element, which may be followed by one or more cs:else-if child elements, and an optional closing cs:else child element. The cs:if and cs:else-if elements may contain any number of rendering elements (except for cs:layout). As an empty cs:else element would be superfluous, cs:else must contain at least one rendering element. cs:if and cs:else-if elements must each hold at least one condition, which are expressed as attributes. The different types of conditions available are:

The contents of an <if disambiguate=”true”> block is only rendered if it disambiguates two otherwise identical citations. This attempt at disambiguation will only be made when all other disambiguation methods have failed to uniquely identify the target source.
Tests whether the given variables (Appendix I - Variables) contain numeric data.
Tests whether the given date variables contain uncertain dates.
Tests whether the locator matches the given locator variable subtype (see Locators).

Tests whether the position of the item cite matches the given positions (when called within cs:bibliography, this condition will always test “false”). The different positions are (note on terminology: a citation refers to a citation group, which contains one or more cites to individual items):

  • “first”: the position of a cite that is the first to reference an item

  • “ibid”/”ibid-with-locator”/”subsequent”: a cite that references an earlier cited item always has the “subsequent” position. In special cases cites may have the “ibid” or “ibid-with-locator” position. These positions are only assigned when:

    1. the current cite immediately follows on another cite, within the same citation, that references the same item


    1. the current cite is the first cite in the citation, and the previous citation includes a single cite that references the same item

    If either requirement is met, the presence of locators determines which position is assigned:

    • Preceding cite does not have a locator: if the current cite has a locator, the position of the current cite is “ibid-with-locator”. Otherwise the position is “ibid”.
    • Preceding cite does have a locator: if the current cite has the same locator, the position of the current cite is “ibid”. If the locator differs the position is “ibid-with-locator”. If the current cite lacks a locator the position is “subsequent”.
  • “near-note”: the position of a cite following another cite that references the same item. Both cites have to be located in foot or endnotes, and the distance between both cites may not exceed the maximum distance (measured in number of foot or endnotes) set with the near-note-distance option (see Note Distance).

Note that each cite can have multiple position values. Whenever position=”ibid-with-locator” is true, position=”ibid” is also true. And whenever position=”ibid” or position=”near-note” is true, position=”subsequent” is also true.

Tests whether the item matches the given types (Appendix II - Types).
Tests whether the given variables (Appendix I - Variables) contain non-empty values.

With the exception of disambiguate, all conditions allow for multiple test values (separated with spaces, e.g. “book thesis”).

The cs:if and cs:else-if elements may include the match attribute to control the testing logic, with possible values:

  • “all” (default): the element only tests “true” when all conditions test “true” for all given test values
  • “any”: the element tests “true” when any condition tests “true” for any given test value
  • “none”: the element only tests “true” when none of the conditions test “true” for any given test value

Style Behavior


Styles can be extensively configured with (optional) options, which are set as attributes. Citation-specific options are set on cs:citation, while bibliography-specific options are set on cs:bibliography. Global options, which affect both citations and the bibliography, are set on cs:style. Inheritable name options may be set on cs:style, cs:citation and cs:bibliography. Finally, Locale Options may be set on cs:locale elements.

Citation-specific Options


Disambiguation can be achieved in five ways:

  1. The number of names shown can be increased.
  2. A given name can be added.
  3. Initialized given names can be expanded.
  4. A year-suffix can be included.
  5. The cite can be rendered with the disambiguate attribute of cs:choose conditions testing “true”.

Note that the term “disambiguation” in the statement above is itself ambiguous. Steps (1), (4) and (5) aim solely to disambiguate cites that otherwise would be the same. Steps (2) and (3), however, are different. In addition to the strict purpose of disambiguating cites, the adding or expansion of given names may be used for the broader purpose of disambiguating names throughout the document. In the description below, this difference is referred to as “the scope of names transformation”.

The five potential steps to disambiguation are activated with the attributes described in this section, and are always performed, if at all, in the order listed below.

disambiguate-add-names [Step (1)]
If set to “true” (“false” is the default), names that would otherwise be hidden as a result of et-al abbreviation are added one by one, until either the target reference is uniquely identified, or all names are shown.
disambiguate-add-givenname [Steps (2) & (3)]

If set to “true” (“false” is the default), given names are added or expanded. For example:

Original form Disambiguated form
(Simpson 2005; Simpson 2005) (H. Simpson 2005; B. Simpson 2005)
(Doe 1950; Doe 1950) (John Doe 1950; Jane Doe 1950)

Note that the value of the givenname-disambiguation-rule attribute (the default is “all-names”) determines a) the precise method of name expansion, and b) whether or not cites that are not themselves ambiguous but do contain the ambiguous name(s) are affected by this type of disambiguation.

givenname-disambiguation-rule [Steps (2) & (3) supplemental]

This attribute accepts one of five possible values, which vary in three respects: the scope of names transformation within the document; the steps included in the disambiguation attempt; and the names within a cite that are affected.

The scope of names transformation

With a value of “all-names”, “all-names-with-initials”, “primary-name”, or “primary-name-with-initials”, disambiguation is performed for all relevant names, without regard to ambiguity in individual cites. Transformations governed by these rules apply to all cites throughout the document. Disambiguation of cites is in this case incidental to the disambiguation of names.

With a value of “by-cite”, only the names within ambiguous cites are transformed, as required to discriminate between references. Cites that are not ambiguous are not affected.

Transformation steps

All five types of given name disambiguation follow the same general transformation steps (the specific steps applied depend on the value of givenname-disambiguation-rule).

  1. If initialize-with is set, then:

    (a) A form value of “short” can be incremented to “long” (e.g. “Doe” becomes “J. Doe”).

    (b) initialize-with can be ignored (e.g. “J. Doe” becomes “John Doe”).

  2. If initialize-with is not set, then the form value of “short” can be immediately incremented to “long” (e.g. “Doe” becomes “John Doe”).

Given name disambiguation rules

The effect of each given name disambiguation rule is described below. In all cases, transformations that do not contribute to disambiguation are omitted, and any names added by disambiguate-add-names that follow the name that results in success are discarded.

The default value. If a name is rendered the same in different cites (e.g. “Doe 2000” and “Doe 2001”), the name is progressively transformed until it can be distinguished from the others (e.g. “A. Doe 2000” and “B. Doe 2001”), or until the transformation steps are exhausted.
Same as “all-names”, but limited to step 1(a). If initialize-with is not set, no disambiguation attempt is made.
Same as “all-names”, but ambiguity is only checked for the first-listed name, and only first-listed names are affected by the transformation.
Same as “primary-name”, but limited to step 1(a). If initialize-with is not set, no disambiguation attempt is made.
Same as “all-names”, but the transformation is limited to ambiguous cites. The appearance of the names transformed will not be affected in other cites.
disambiguate-add-year-suffix [Step (4)]
If set to “true” (“false” is the default), a year-suffix is added to cites that are otherwise identical (e.g. “Doe 2007, Doe 2007” becomes “Doe 2007a, Doe 2007b”). The placement of the year-suffix, which by default is appended to each cite, can be controlled by explictly rendering the “year-suffix” variable using cs:text.

If ambiguous cites remain after the above steps have been exhausted, a final attempt at disambiguation is performed with the disambiguate test value on any cs:choose conditions testing “true” [Step (5)]. If this results in successful disambiguation, any names added by disambiguate-add-names are discarded.

Citation Collapsing

The collapse option activates citation collapsing. Note that “year-suffix” and “year-suffix-ranged” both fall back to “year” when the disambiguate-add-year-suffix attribute is not set to “true” (see Disambiguation). Its possible values are:

  • “citation-number”: collapses numeric citation ranges (e.g. from “[1, 2, 3, 5]” to “[1-3, 5]”). Note that only increasing ranges are collapsed, e.g. “[3, 2, 1]” will not collapse (to see how numeric styles can sort citations by the citation-number variable, see Sorting).
  • “year”: when the names stored in the rendered name variables are the same for two subsequent cites, the latter cite is collapsed to only the year, e.g. from “(Doe 2000, Doe 2001)” to “(Doe 2000, 2001)”.
  • “year-suffix”: collapses as “year”, but also collapses identical years, e.g. “(Doe 2000a, b)” instead of “(Doe 2000a, 2000b)”.
  • “year-suffix-ranged”: collapses as “year-suffix”, but also collapses ranges of year-suffix markers, e.g. “(Doe 2000a-c,e)” instead of “(Doe 2000a, b, c, e)”.
Specifies the delimiter for year-suffix elements. For example, citations like “(Smith 1999a,b; 2000; Jones 2001)” are obtained when the collapse attribute is set to “year-suffix”, the delimiter on cs:layout in cs:citation is set to ”; ”, and the year-suffix-delimiter is set to “,”. When the year-suffix-delimiter attribute is not set, year-suffixes are delimited with the delimiter set on cs:layout in cs:citation.
Specifies the cite delimiter that should be used after a group of collapsed cites. For example, citations like “(Smith 1999a, b, 2000; Jones 2001, Brown 2007)” are obtained when the collapse attribute is set to “year-suffix”, the delimiter on cs:layout in cs:citation is set to “, ” and after-collapse-delimiter is set to ”; ”.
Note Distance
The “near-note” position (see Choose) tests “true” if a preceding note exists that: a) refers to the same item and b) has a distance (measured in footnotes or endnotes) to the current item that does not exceed the value of near-note-distance. This attribute defaults to 5.

Bibliography-specific Options

If set to “true” (“false” is the default), bibliographic entries are rendered with hanging-indents.

If set to “flush”, subsequent lines of each bibliography entry are aligned with the beginning of the second field. If set to “margin”, the first field is put in the margin and all subsequent lines of text are aligned with the margin (as in the IEEE style). An example showing the alignment, if the first field is <text variable="citation-number" suffix=". "/>:

1. Adams, D. (2002). The Ultimate Hitchhiker's Guide to the
   Galaxy (1st ed.).
Specifies a formatting hint for vertical line distance. Defaults to “1” (single-spacing), and can be set to any non-negative integer to specify a multiple of the standard unit of line height (e.g. “2” for double-spacing).
Specifies a formatting hint for vertical distance between bibliographic entries. By default (with a value of “1”), entries are separated by a single additional line-height (as set by the line-spacing attribute). Can be set to any non-negative integer to specify a multiple of this amount.
Reference Grouping

The value of the subsequent-author-substitute attribute (which may be any string) is used to replace the names in a bibliographic entry, when it shares these names with the preceding bibliographic entry. Note that only the first cs:names element rendered is affected. E.g., with subsequent-author-substitute set to “—”:

Asimov. Foundation, 1951.
---. Foundation and Empire, 1952.
---. Second Foundation, 1953.

Global Options

Hyphenation of Initialized Names
Specifies whether compound given names (e.g. “Jean-Luc”) should be initialized with a hyphen (“J.-L.”, value “true”) or without (“J.L.”, value “false”). Defaults to “true”.
Page Ranges
The value of this attribute determines how page ranges are formatted. Available values: “expanded” (e.g. “321-328”), “minimal” (“321-8”), and “chicago” (“321-28”) (see Appendix IV - Page Abbreviation Rules for the Chicago Manual of Style page range collapsing rules). If the attribute is not specified, the content of the page-field is left unchanged.
Name Particles

Many Western names include one or more name particles (e.g. “de” in the Dutch name “W. de Koning”). However, not all particles are equal: name particles can be either maintained or dropped when only the surname is shown (from now on we will refer to these two types as non-dropping-particle and dropping-particle, respectively). A single name can contain particles of both types (with the non-dropping-particle always following the dropping-particle). For example, the French name “Gérard de la Martinière” can be deconstructed into:

    "author": {
        "given": "Gérard",
        "dropping-particle": "de",
        "non-dropping-particle": "la",
        "family": "Martinière"
        "given": "W.",
        "non-dropping-particle": "de",
        "family": "Koning"

When just the surname is shown, only the non-dropping-particle is kept: “La Martinière”.

Whereas the dropping-particle is always treated the same, styles vary in how the non-dropping-particle is handled in case of inverted names, where the family name precedes the given name. First, the non-dropping-particle can be either prepended to the family name (e.g. “de Koning, W.”) or appended (after initials or given names, e.g. “Koning, W. de”). Note that the dropping-particle is always appended in inverted names. Secondly, if the choice has been made to prepend the non-dropping-particle to the family name for inverted names, the author sort order can differ. Either the non-dropping-particle remains part of the family name (as part of the primary sort key; sort order A), or it may be separated from the family name and become (part of) a secondary sort key, joining the dropping-particle, if available (sort order B).

Sort order A: non-dropping-particle not demoted

  • primary sort key: “la Martinière”
  • secondary sort key: “de”
  • tertiary sort key: “Gérard”

Sort order B: non-dropping-particle demoted

  • primary sort key: “Martinière”
  • secondary sort key: “de la”
  • tertiary sort key: “Gérard”

Some names include a particle that should never be demoted. For these cases the particle should just be included in the family name field, for example for the French general Charles de Gaulle:

    "author": {
        "family": "de Gaulle",
        "given": "Charles"

The handling of particles for inverted names is set with the demote-non-dropping-particle option:


Sets the display and sorting behavior of the non-dropping-particle in inverted names (e.g. “Koning, W. de”). The possible values are:

  • “never”: the non-dropping-particle is treated as part of the family name, whereas the dropping-particle is appended (e.g. “de Koning, W.”, “la Martinière, Gérard de”). The non-dropping-particle is part of the primary sort key (sort order A, e.g. “de Koning, W.” appears under “D”).
  • “sort-only”: same display behavior as “never”, but the non-dropping-particle is demoted to a secondary sort key (see sort order B, e.g. “de Koning, W.” appears under “K”).
  • “display-and-sort” (default): the dropping and non-dropping-particle are appended to the rest of the name (e.g. “Koning, W. de” and “Martinière, Gérard de la”). When names are sorted, all particles are part of the secondary sort key (see sort order B, e.g. “Koning, W. de” appears under “K”).

Inheritable Name Options

Attributes for the cs:names and cs:name elements may also be set on cs:style, cs:citation and cs:bibliography. This eliminates the need to repeat the same attributes and attribute values for every occurrence of the cs:names and cs:name elements.

The available inheritable attributes for cs:name are and, delimiter-precedes-last, et-al-min, et-al-use-first, et-al-subsequent-min, et-al-subsequent-use-first, initialize-with, name-as-sort-order and sort-separator. The attributes name-form and name-delimiter accompany the form and delimiter attributes on cs:name. Similarly, names-delimiter, the only inheritable attribute available for cs:names, accompanies the delimiter attribute on cs:names.

When an inheritable name attribute is set on cs:style, cs:citation or cs:bibliography, its value is used for all cs:names elements within the element carrying the attribute. When an element lower in the hierarchy includes the same attribute with a different value, this latter value will override the value(s) specified higher in the hierarchy.

Locale Options

Determines whether punctuation (commas and periods) is placed inside (value “true”) or outside (default, value “false”) quotation marks added with the quotes attribute (see Formatting).


The sort order for citations and the bibliography can be set with the cs:sort element in cs:citation and cs:bibliography. If a style does not include sorting instructions, references are listed in the order cited.

The cs:sort element must contain one or more cs:key child elements. The sort key, set as an attribute on cs:key, can be a variable (see Appendix I - Variables) or a macro. The cs:key element may carry the sort attribute, with possible values of “ascending” (default) or “descending”, to indicate the sort order. The names-min and names-use-first attributes (which affect all names generated by macros called by cs:key) can be used to (further) constrain the number of names used in the sort, overriding the values of the corresponding et-al-min and et-al-use-first and et-al-subsequent options.

Sort keys are evaluated one by one. The primary sort is performed using the first sort key. A secondary sort (using the second sort key) is performed on those items which share the first sort key. A tertiary sort (using the third sort key) is performed on those items which share the first and second sort key. This process continues until either the order of all items is fixed, or until the sort keys are exhausted. Items for which a sort key is empty are placed at the end of the sort (both for ascending and descending sorts).

An example, where citations are first sorted by the output of the author macro with overriding settings for et-al abbreviation. Entries that share the same author macro output are further sorted in reverse order by date of issue.

    <key macro="author" names-min="3" names-use-first="3"/>
    <key variable="issued" sort="descending"/>
    <!-- rendering elements -->

The values returned for variables and macros called in cs:sort may differ from the “ordinary” rendered values. These differences are detailed below.

Sorting Variables

When variables are referenced in cs:key via the variable attribute, the string value is returned, without formatting decorations. Exceptions are name, date and numeric variables, which are returned as follows:

names: Name variables can be set directly on cs:key using the variable attribute (e.g. <key variable="author"/>). In this case, the name-list from the variable will be returned as a string in the “long” form of cs:name, formatted with name-as-sort-order set to “all”.

dates: Date variables that are set directly on cs:key using the variable attribute are returned to cs:key in the YYYYMMDD format, with zeros substituted for any missing date-parts (e.g. 20001200 for December 2000). As a result, dates with more date-parts will come after those with fewer date-parts, e.g. (2000, May 2000, May 1st 2000). Note that negative years are sorted inversely, e.g. (100BC, 50BC, 50AD, 100AD). Seasons are ignored for sorting, as the chronological order of the seasons differs between the northern and southern hemispheres. Date ranges consist of a start date and an end date. The start date is used for comparison with single dates. However, for items with the same (start) date, the items with date ranges are placed after those with single dates, e.g. (1999, 2000, 2000-2002, 2001, 2001-2003). In addition, date ranges are subjected to a secondary sort based on the end date, e.g. (2000, 2000-2001, 2000-2004, 2000-2005, 2001).

numbers: If the variable attribute is used, numeric values are returned as integers (form is “numeric”). If the original variable value only consists of non-numeric text, the value is returned as a text string.

Sorting Macros

A macro called via cs:key returns whatever string value the macro would ordinarily generate, with a few exceptions. In all cases, rich text markup is removed from the sort key.

For name sorting, it is generally preferable to use the same macro that is used to render the names in the context (cs:citation or cs:bibliography) to which the sort applies. The first benefit of using macros is that substitution logic becomes available (e.g. the editor variable might substitute for an empty author variable). Secondly, et-al abbreviation can be used (using either the et-al-min and et-al-use-first or et-al-subsequent options defined within the macro, or the overriding names-min and names-use-first attributes set on cs:key). Note that the “et-al” and “and others” terms are not included in the sort key when et-al abbreviation occurs. The third benefit is that names can be sorted by just the family name and name particles, using a macro for which the form attribute on cs:name is set to “short”. Finally, it is possible to sort by the number of names in a names-list, by calling a macro in which the form attribute of cs:name is set to “count”. In this case a count value of “3” would be obtained for a name variable that would otherwise return “Jones, Smith, Doe”. For name sorting, the name-as-sort-order attribute on cs:name elements is set to “all”.

Number variables (rendered with cs:number) and date variables are treated the same as when they were called via variable. The only exception is that if a date variable is called by the variable attribute, the complete date is returned. In contrast, macros return only those date-parts that would otherwise be rendered (respecting the value of the date-parts attribute for localized dates, or the listing of cs:date-part elements for non-localized dates).


The following formatting attributes may be set on cs:date, cs:date-part, cs:et-al, cs:group, cs:label, cs:layout, cs:name, cs:name-part, cs:names, cs:number and cs:text:


Sets the font style, with values:

  • “normal” (default)
  • “italic”
  • “oblique” (i.e. slanted)

Allows for the use of small capitals, with values:

  • “normal” (default)
  • “small-caps”

Sets the font weight, with values:

  • “normal” (default)
  • “bold”
  • “light”

Allows for the use of underlining, with values:

  • “none” (default)
  • “underline”

Sets the vertical alignment, with values:

  • “baseline” (default)
  • “sup” (superscript)
  • “sub” (subscript)


The affixes attributes prefix and suffix may be set on cs:date (except when cs:date is used within cs:locale), cs:date-part (except when the parent cs:date element calls a localized date format), cs:et-al, cs:group, cs:label, cs:layout, cs:name, cs:names, cs:number and cs:text. The attribute value is included either before (prefix) or after (suffix) the displayed text. Affixes are generally insensitive to the formatting attributes acting on the calling element: the only exception to this rule are affixes set on cs:layout. In cases where formatting of affixes is desired, separate cs:text elements can be used instead, with a value attribute to output verbatim text.


The delimiter attribute can be used to specify a delimiting string for cs:date (delimiting the date-parts; not allowed when cs:date calls a localized date format), cs:names (delimiting multiple name variables), cs:name (delimiting names in name lists), cs:group and cs:layout (both delimiting the direct child elements).


Many of the anticipated output formats for CSL 1.0 (RTF, LaTeX, XML dialects such as XHTML) allow styling to be applied to individual blocks of text in order to control their appearance and position. The display attribute can be used in cs:bibliography to allocate styling to particular text blocks for this purpose [1]. The attribute may be set on cs:date, cs:group, cs:names, cs:number and cs:text elements, and accepts one of the following values:

  • “block”: A block stretching from margin to margin.
  • “left-margin”: A block of fixed width starting at the left margin (all “left-margin” blocks in a bibliography share the same width, set according to the maximum number of characters appearing in any one such block).
  • “right-inline”: A block directly to the right of any immediately preceding “left-margin” block, and extending to the right margin.
  • “indent”: Block indented to the right by a standard amount.
[1]N.B. if display attributes are used, make sure all rendering elements are under the control of exactly one display attribute.


  1. A similar effect as with second-field-align can be achieved with [2]:

        <text display="left-margin" variable="citation-number"
            prefix="[" suffix="]"/>
        <group display="right-inline">
          <!-- citation rendering elements -->
[2]The styling definitions in the target application (CSS for HTML, styles for Word etc.) can be adjusted to achieve special effects, such as floating the labels into the margin.

  1. A per-author publication listing can be formatted as follows [3]:

    <bibliography subsequent-author-substitute="">
        <key variable="author"/>
        <key variable="issued"/>
        <group display="block">
          <names variable="author"/>
        <group display="left-margin">
          <date variable="issued">
            <date-part name="year" />
        <group display="right-inline">
          <text variable="title"/>

    which would result in

    year-publication1 title-publication1
    year-publication2 title-publication2
    year-publication3 title-publication3
    year-publication4 title-publication4
[3]The effect of the empty subsequent-author-substitute attribute is to render the author name only once, at the top of the list of each author’s publications.

  1. An annotated bibliography with the annotation block-indented below the reference can be formatted as follows:

        <group display="block">
          <!-- citation rendering elements -->
        <text display="indent" variable="abstract" />


The quotes attribute may set on cs:text. When set to “true” (“false” is default), the rendered text is wrapped in quotes. The quote-symbols are defined as (localized) terms. The localized punctuation-in-quote option controls whether punctuation appears inside or outside the quotes (see Locale Options).


The strip-periods attribute may be set on cs:date-part (but only if name is set to “month”), cs:label and cs:text. When set to “true” (“false” is the default), any periods in the rendered text are removed.


The text-case attribute may be set on cs:date-part, cs:label, cs:name-part, cs:number and cs:text and can be used to control the text case of the rendered text. The possible values are:

  • “lowercase”: displays all text in lowercase
  • “uppercase”: displays all text in uppercase
  • “capitalize-first”: capitalizes the first character; the case of other characters is not affected
  • “capitalize-all”: capitalizes the first character of every word; other characters are displayed lowercase
  • “title”: displays text in title case (the Chicago Manual of Style calls this “headline style”)
  • “sentence”: displays text in sentence case (“sentence style”)


Appendix I - Variables

Standard Variables

  • abstract
  • annote
  • archive
  • archive_location
  • archive-place
  • authority
  • call-number
  • chapter-number
  • citation-label
  • citation-number
  • collection-title
  • container-title
  • DOI
  • edition
  • event
  • event-place
  • first-reference-note-number
  • genre
  • ISBN
  • issue
  • jurisdiction
  • keyword
  • locator
  • medium
  • note
  • number
  • number-of-pages
  • number-of-volumes
  • original-publisher
  • original-publisher-place
  • original-title
  • page
  • page-first
  • publisher
  • publisher-place
  • references
  • section
  • status
  • title
  • URL
  • version
  • volume
  • year-suffix

Date Variables

  • accessed
  • container
  • event-date
  • issued
  • original-date

Name Variables

  • author
  • editor
  • translator
  • recipient
  • interviewer
  • publisher
  • composer
  • original-publisher
  • original-author
  • container-author (to be used when citing a section of a book, for example, to distinguish the author proper from the author of the containing work)
  • collection-editor (use for series editor)

Appendix II - Types

These are the different item types available within CSL:

  • article
  • article-magazine
  • article-newspaper
  • article-journal
  • bill
  • book
  • broadcast
  • chapter
  • entry
  • entry-dictionary
  • entry-encyclopedia
  • figure
  • graphic
  • interview
  • legislation
  • legal_case
  • manuscript
  • map
  • motion_picture
  • musical_score
  • pamphlet
  • paper-conference
  • patent
  • post
  • post-weblog
  • personal_communication
  • report
  • review
  • review-book
  • song
  • speech
  • thesis
  • treaty
  • webpage

Appendix III - Terms


  • anthropology
  • astronomy
  • biology
  • botany
  • chemistry
  • communications
  • engineering
  • generic-base - used for generic styles like Harvard and APA
  • geography
  • geology
  • history
  • humanities
  • law
  • linguistics
  • literature
  • math
  • medicine
  • philosophy
  • physics
  • political_science
  • psychology
  • science
  • social_science
  • sociology
  • theology
  • zoology


  • book
  • chapter
  • column
  • figure
  • folio
  • issue
  • line
  • note
  • opus
  • page
  • paragraph
  • part
  • section
  • sub verbo
  • verse
  • volume


  • month-01
  • month-02
  • month-03
  • month-04
  • month-05
  • month-06
  • month-07
  • month-08
  • month-09
  • month-10
  • month-11
  • month-12


  • ordinal-01
  • ordinal-02
  • ordinal-03
  • ordinal-04
  • long-ordinal-01
  • long-ordinal-02
  • long-ordinal-03
  • long-ordinal-04
  • long-ordinal-05
  • long-ordinal-06
  • long-ordinal-07
  • long-ordinal-08
  • long-ordinal-09
  • long-ordinal-10

Quotation marks

  • open-quote
  • close-quote
  • open-inner-quote
  • close-inner-quote


  • author
  • collection-editor
  • composer
  • container-author
  • editor
  • editorial-director
  • editortranslator
  • interviewer
  • original-author
  • recipient
  • translator


  • season-01
  • season-02
  • season-03
  • season-04


  • accessed
  • ad
  • and
  • and others
  • anonymous
  • at
  • bc
  • by
  • circa
  • cited
  • edition
  • et-al
  • forthcoming
  • from
  • ibid
  • in
  • in press
  • internet
  • interview
  • letter
  • no date
  • online
  • presented at
  • reference
  • retrieved

Appendix IV - Page Abbreviation Rules

The page abbreviation rules for the different values of the page-range-format attribute on cs:style are:

All digits repeated in the second number are left out: 42-5, 321-8, 2787-816
Abbreviated page ranges are expanded to their non-abbreviated form: 42-45, 321-328, 2787-2816
Page ranges are abbreviated according to the Chicago Manual of Style-rules:

Table: Chicago Manual of Style page range abbreviation rules

First number Second number Examples
Less than 100 Use all digits 3-10; 71-72
100 or multiple of 100 Use all digits 100-104; 600-613; 1100-1123
101 through 109 (in multiples of 100) Use changed part only, omitting unneeded zeros 107-8; 505-17; 1002-6
110 through 199 (in multiples of 100) Use two digits, or more as needed 321-25; 415-532; 11564-68; 13792-803
4 digits If numbers are four digits long and three digits change, use all digits 1496-1504; 2787-2816