NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Journal Article Tag Suite Conference (JATS-Con) Proceedings 2011 [Internet]. Bethesda (MD): National Center for Biotechnology Information (US); 2011.

Cover of Journal Article Tag Suite Conference (JATS-Con) Proceedings 2011

Journal Article Tag Suite Conference (JATS-Con) Proceedings 2011 [Internet].

Show details

Variations in XML Reference Tagging in Scholarly Publication


Author Information

From the earliest days of markup languages, reference tagging in scholarly publications has presented unique challenges. References to journals are relatively easy to tag, once a publisher has settled on a consistent structure — for example, with or without the loose text and punctuation between elements. References to non-journal content such as books, conference proceedings, and “gray” literature, however, can present unexpected or awkward situations, and even journal references can occasionally prove idiosyncratic. This paper will provide a brief history of reference tagging in SGML and XML and will discuss specific reference markup structures in the Journal Article Tag Suite (JATS), from the common to the arcane. The evolution from <citation> and <nlm-citation> to the newer <mixed-citation> and <element-citation> elements in 3.x will be reviewed, including a discussion of the workflow implications of each model. The author will conclude with some observations about the intersection among reference markup, online reference linking, and the true meaning of references in the world of electronic publishing.


What makes XML markup of scholarly journals and books so different from markup of other types of publications? Like scholarly content, publications such as trade books and instruction manuals have the challenges of tables, math, and special characters (to a greater or lesser degree, depending on the text), but one element sets scholarly materials apart: the consistent requirement to structure references.

References can be the bane of existence for those who work in scholarly publishing. Authors dislike finding them and inserting them into their articles. Editors, too, have little love for references, spending a significant portion of their copy editing time correcting them (Meyer, 2008; Wates and Campbell, 2007). Even with all of the energy spent on copy editing, reference quality is an ongoing problem (Lukić et al., 2004).

Although they present many challenges, references are essential to scholarly literature. The researcher's need for references was well stated by Pam Knox and Sasha Schwarzman at the American Geophysical Union:

What is the purpose of listing the article's sources in the reference list in the first place? It is to give the reader enough information to find the source (without overwhelming him with unnecessary/extraneous metadata) AND to give an indication of how authoritative the source is — in order for the reader to decide whether he should bother to pursue the source in the first place. The objective… is to determine what is the minimal set of metadata that can unambiguously identify the source and give an indication of its authority/importance/reputation, so that the reader can go to a shelf and grab the source or write an ILL request for it, or — in the case we're talking about — to send this sufficient metadata set to CrossRef, so that its matching algorithm could return the source's DOI to the requester. [personal communication, 22 November 2004]

Beyond the scientific need for references, they are important to a scientist's career:

Corrupt or missing references can be a source of minor irritation or major inconvenience — misquoted references increase the probability that a citation index such as Web of Science will not be able to link the citations to the source article. In today’s metric-driven world, not receiving credit, in terms of citations, for the work that one has published can actually make a difference in terms of promotions, tenure, grant funding, etc. (Wates and Campbell, 2007)

We have no choice but to live with references, and in an increasingly electronic world, the ability to find, link to (Hunter, 1998), and follow references is contingent on accurate semantic tagging (Wates and Campbell, 2007). All of this is further complicated by the proliferation of citing methods and reference styles in the online world (Wall Street Journal, 2002).

This paper will focus on issues of reference tagging, from the macro-level to the arcane, and how they are related to the new requirements of electronic publishing.


The following terms will be used in this paper:

Reference: The citation, usually in a reference list or bibliography, of another published work (e.g., journal article, book chapter, web site).

Generated text: Inconsequential, formulaic, or stereotypical text, punctuation, and formatting omitted from an XML file, which is applied to content by a style sheet when an XML file is rendered. The style sheet generates this text and visual formatting based on the structural information provided by the markup elements and attributes.

Boilerplate text: Inconsequential, formulaic, or stereotypical text, punctuation, and formatting that could have been omitted but which the publisher has chosen to keep in the XML file rather than to generate with a style sheet.

Proscribed order: Placement of reference elements in an order enforced by the DTD. Elements in proscribed order may be ordered very differently in XML than they are in the rendered HTML or PDF.

Presentation order: Placement of reference elements in the visual order they appear for publication. This is a looser model than Proscribed order.

Gray literature: Publications that fall outside the mainstream of published journal and monograph literature, e.g. technical reports or working papers.

NLM DTD: This paper refers to the NLM DTD for all references to versions 3.0 and earlier. JATS (Beck, 2011) is used specifically to reference the tag set currently available as a draft standard for trial use.


The information in this paper is drawn from the author's 15 years of experience developing systems to mark up references and his participation in the NLM DTD/JATS Working Group since 2002. Data with respect to inclusion of boilerplate text in references can be found in Rosenblum and Golfman (2001) and Rosenblum (2010). This paper does not serve as a scientific survey of reference markup styles but rather provides some insight into the elusive corners of reference markup that often evade even the most complete DTD documentation and the effect editorial style can have on correct reference tagging.


A Short History of JATS Citation Elements

In the beginning, the 12083 DTD <citation> element was created to structure references. And publishers saw that it was good, but it was not good enough.

Through the 1990s, publishers created variations of this 12083 structure with alternate names and models — for example, <bb> in the Elsevier (Bernickus, et al., 2005) and Blackwell DTDs, and <bibl> in PMC 1.0. The PMC 1.0 <bibl> element was especially challenging because it was used for both references and article metadata, a dual use that left it dangerously overloaded and not well suited for either purpose.

12083 Roots

The <citation> name of 12083 heritage was revived in the NLM Archiving DTD 1.0 but with a very different model. The first-generation NLM DTD <citation>, designed for the requirements of archiving, allowed for a) markup of journal and non-journal references, b) inclusion of boilerplate text, and c) element placement in presentation order. (Well… almost. Author names in <citation> in version 1.0 did not permit boilerplate text, and author-related elements were in a proscribed order. See more about this below in the discussion of person-group-model.)

NLM DTD 1.x and 2.x Citation Models

When the NLM Archiving (Green) DTD was expanded to Publishing (Blue) and Authoring (Pumpkin) DTDs, a second element, <nlm-citation> was added for PubMed Central's (PMC) needs, and it was defined as follows:

Structured citation model to assist users creating “new” content; the model loosely reflects the NLM’s style in that it allows the tagging of all “legal” NLM citations and enforces the sequence in which content must appear if it is present.

Note: This model does not provide guidance on what information is required for each type of cited content. Moreover, the model assumes that punctuation between the parts of a citation will be generated on display or on export from the XML tagged according to this DTD to XML for another use (National Center for Biotechnology Information, 2003).

Unlike <citation>, <nlm-citation> had a proscribed element order whose model was somewhat based on the editorial style requirements of "NLM Recommended Formats for Bibliographic Citation" (Personal communication with Jeff Beck, Nov 13, 2002; U.S. National Library of Medicine, 2003). The model also did not allow boilerplate text (PCDATA) between elements, and many elements could not be repeated.

<nlm-citation> was problematic when applied to some references. For example, in the reference:

Fukumoto Y (1972b) Study on the behaviour of stabilization piles for landslides. Soil and Foundation 12(2), 61–73 [in Japanese].

there are two pieces of text that are comments: "b" after the year [see discussion below], and "in Japanese". Because of the proscribed order, the XML markup for this reference is

<nlm-citation citation-type="journal"><person-group><name><surname>Fukumoto</surname><given-names>Y</given-names></name></person-group><article-title>Study on the behaviour of stabilization piles for landslides.</article-title><source>Soil and Foundation</source><year>1972</year><volume>12</volume><issue>2</issue><fpage>61</fpage><lpage>73</lpage><comment>b</comment><comment>[in Japanese]</comment></nlm-citation>

The <comment> element did not have any attributes until content-type was added in NLM DTD 2.3. The result is that rendering the reference, as shown above, from the nlm-citation tagging is difficult at best because there is no context provided by element order or attributes in the two comment elements to indicate that one should be after the year and the other at the end of the reference.

As this example shows, when a reference model has a proscribed order, and the presentation order of the reference elements is different from the tagged element order, it places a greater burden on the template developer. Instead of just adding missing punctuation, elements must be reordered for presentation, and this reordering can have more complex requirements based on the reference type (e.g., book versus journal). In some cases, it's only possible to apply a correct template by analyzing the content of one or more elements. For many publishers, the increased template complexity is not worth the investment. The result was that publishers who preferred to use generated text often used <citation>, excluding PCDATA between elements, rather than <nlm-citation> because the proscribed order was burdensome.

NLM DTD 3.0 Citation Models

After much debate, the NLM DTD Working Group decided to clean up a number of issues for version 3.0 and break backward compatibility with earlier versions (NLM DTD Working Group, 2007). In the process of conducting this cleanup, it was decided that the two existing reference models did not serve users as well as they could. In their place, two new models were introduced to replace <citation> and <nlm-citation>:

  • <mixed-citation>: a direct replacement for <citation> in earlier versions, but with a new name and revised attributes; like the earlier <citation> model, it is a mixed content model that permits boilerplate text between elements (hence the "mixed" part of the name) and has no proscribed element order
  • <element-citation>: a replacement for <nlm-citation>, except
    • the element order is no longer proscribed; elements can occur in any order
    • all elements can be repeated as many times as necessary

The <element-citation> model solved the problems of <nlm-citation>, allowing the sample reference from above to be tagged in XML as

<element-citation publication-type="journal"><person-group><name><surname>Fukumoto</surname><given-names>Y</given-names></name></person-group><year>1972</year><comment>b</comment><article-title>Study on the behaviour of stabilization piles for landslides.</article-title><source>Soil and Foundation</source><volume>12</volume><issue>2</issue><fpage>61</fpage><lpage>73</lpage><comment>[in Japanese]</comment></element-citation>

and easily rendered with a simple style sheet. If desired, content-type attributes could be added to the comment elements as a rendering aid, but they are not required.

NLM DTD 3.0 Citation Attributes

NLM DTD 3.0 also changed the attributes available for citations. Before 3.0, <citation> and <nlm-citation> had a single citation-type attribute. However, in some cases this was inadequate. For example, there was no reasonable way to describe an online book published by a government entity, except perhaps citation-type="book|gov|eref", which was not ideal. This problem and others were addressed in DTD 3.0 by replacing the overloaded citation-type attribute with three new attributes:

  • publication-type
  • publisher-type
  • publication-format
permitting an online government book to be tagged as <mixed-citation publication-type="book" publisher-type="gov" publication-format="online">.

Boilerplate versus Generated Text

The decision to keep or discard boilerplate text when using the <citation> or <mixed-citation> models is a matter of choice by the user of NLM/JATS, and we have consistently seen publishers divided almost equally in this selection (Rosenblum, 2010). However, in some unusual cases, the editorial style may make the decision easier.

In 2002, the author spoke with a scientific journal publisher developing a proprietary DTD (before the NLM DTD was released), and their reference model did not allow boilerplate text. This seemed fine until it was learned that the publisher's editorial style required a comma after journal names that did not end with an abbreviated word and a period after those that did. When the additional regular expression development work to create style sheets to support this style in each medium was described to the publisher, they quickly changed to a boilerplate text model, realizing it was easier and cheaper to keep the punctuation than generate it for every rendering.

Most situations are not this unusual, and most references can be rendered reasonably from either a <mixed-citation> or <element-citation> model. However, if content includes a significant number of non-journal references, use of <mixed-citation> will make template development easier because setting up punctuation for all possible cases of missing elements is easier when the generated text need not be created to cover every possible case (e.g., a book reference that has a publisher name but not a publisher location vs. a reference with both elements vs. a reference with neither of these elements).


For those who prefer keeping all boilerplate text, markup of author names has been problematic in the NLM DTD. The earliest versions of the NLM DTD (even Green) did not allow PCDATA in <person-group> When using <citation>, most of the reference could have PCDATA, but not the authors, so it was not possible to keep complete boilerplate text for a reference.

This situation was not adequate for archives such as Portico, so Green was modified in version 2.0 by adding the <x> element (which was modeled after an element of the same name in Blackwell DTD 3.0) to keep boilerplate text between names, and <string-name> to allow retention of boilerplate text between name parts. <string-name> also did not have a proscribed order, so names could be archived exactly as they had appeared in print.

<string-name> was added to Blue with version 3.0, but that served as only a partial solution for boilerplate text. In the move to JATS 0.4, the Working Group decided (after much discussion) to modify Blue by adding PCDATA (not <x>) to person-group-model. With this change, Blue finally allows full retention of boilerplate text in references.

Non-Journal References

Tagging efforts are primarily concerned with journal references. Most references in scientific articles cite journal content, but references to many other content types such as books, conference proceedings (especially in engineering and computer science), and “gray” literature also appear. Non-journal references appear even more frequently in social science and humanities publications.

Today many publishers tag only journal references because

  • non-journal references are a minority of citations in scientific literature
  • automated tagging of non-journal references is more challenging than tagging of journal references because there is often extra information (e.g., reprint information in humanities book references) or missing information (e.g., publishers in book references)
  • today it is not possible to link to most non-journal content in the same manner that one can link journal references through PubMed or CrossRef; PubMed indexes journals almost exclusively, whereas CrossRef indexes other content types but has been less successful in engaging publishers to deposit metadata for non-journal references

Without online linking, there is less business justification for tagging non-journal references.

For those who want to tag non-journal references, the follow subsections have a few suggestions.

Book Chapter Titles

Prior to NLM DTD 3, the chapter title in a book chapter reference had to be tagged with <article-title>. In DTD 3, the <chapter-title> element was introduced.

Incomplete References

As noted above, template development for generated text is more challenging when references have incomplete information (e.g., a missing publisher location). This situation is especially true for references to conference papers. Beware trying to parse and re-render such references when using <element-citation>.


The irony about standards is that there are few standards for the publication of standards. The <std> element is available in the NLM and JATS citation models. The documented use is:

<element-citation publisher-type="stds-body"> <std>International standard ISO 10993-10:2002(E): Biological evaluation of medical devices-Part 10: Tests for irritation and delayed-type hypersensitivity</std> <edition>Second</edition> <year>2002</year> <month>09</month> <day>01</day> </element-citation>

This tagging does not give sufficient semantic markup to find ISO 10993-10:2002 in a semantically driven online search. For this reason, the author believes the JATS Working Group should revisit tagging of standards.

Bad Print Habits

A number of editorial styles have appeared through the years to create shorter references. Although the author has not researched the history of these styles, they may well have arisen as a paper-saving measure; an abbreviated reference list might save a page or two from each article, cutting typesetting, printing, and mailing costs over a large number of issues.

In a world of increasingly online publication, saving paper is becoming less of an issue, and these shorter references can sometimes be difficult to tag. They can also cause problems with online linking by reducing the accuracy of matched links or causing linking to fail altogether. Below are a few examples.

Excluded Article Title

In some disciplines, notably chemistry and physics, the article title is typically not included in the reference, e.g.:

M. G. Banwell, M. D. McLeod, Chem. Commun. 1998, 1991

This practice can save a lot of paper. However, it results in several problems:

  • It does not allow the "reader to decide whether he should bother to pursue the source in the first place" [Schwarzman, personal communication]. If the title isn't present, how can you, a reader of the article, tell if the article is of interest, unless you already know the article (in which case, perhaps the citation isn't even needed)?
  • Linking to CrossRef is less accurate when an article title is not provided. CrossRef's fuzzy matching returns more accurate results for references in which some data is incorrect (e.g., incorrect year, typo in a journal title) when the article title is included in an XML query.

One practical solution to the paper-saving issue was created by the journal Science. Before 2009, Science did not include article titles. Since that time they have updated their workflow to include article titles in only the HTML versions of their articles. This approach gives complete information online, more accurate DOIs, and still saves paper in the print/PDF version.

Excluded Last Pages

In some publications, the last page number is typically not included in the reference. If we look at the same reference as shown above, but without the bold and italic:

M. G. Banwell, M. D. McLeod, Chem. Commun. 1998, 1991

it becomes much harder to tell the year from the page, especially when one knows that the year may be in bold and the page number is never bold in any reference style this author has seen. Why is this important? Well, in many machine-automated applications to process references, bold may not be present, especially if the reference is being tagged from an Optical Character Recognition (OCR) stream. In these situations, accurate tagging can be challenging if not impossible.

One might suggest that the page number always comes last, but that's an unsafe assumption because some journals (e.g., Nature) use an editorial style in which the year comes at the end of the reference.

Note also that this particular publication does not use volume numbers (the year doubles as the volume number), so this abbreviated format presents numerous problems to the naïve tagger. Without knowing the specific journal and editorial style, tagging may not be accurate, resulting in semantically incorrect XML and potentially a lost online link resulting from the incorrect XML.

Excluded Repeated Authors

In some editorial styles, mostly in humanities, dashes are used in place of repeated author names, e.g:

Bornstein, Eli. “The Crystal in the Rock,” The Structurist no. 2 (1961-62): 5-18.

---, “Creation/Destruction/Creation in Art and Nature,” The Structurist no. 7 (1967): 55-63.

This practice can save a little paper, but it has several problems:

  • There is no good way to tag the dashes with the <element-citation> model. Tagging the dashes as <surname>---</surname> is not a good option. Perhaps the least-bad approach with <element-citation> is <name content-type="repeated-author"><surname>Bornstein</surname><given-name>Eli</given-name></name> where the repeated-author attribute can be used to emit dashes when rendering. Note that the attribute must be applied to the name element, not person-group, because of cases like:

    Newman, William R. 1991. The Summa perfectionis of pseudo-Geber. Leiden: Brill.

    ---, and Lawrence M. Principe. 1998. "Alchemy vs. Chemistry: The Etymological Origins of a Historiographic Mistake." Early Science and Medicine, 3:32-65.

    in which only some authors are replaced with dashes.
  • Linking to CrossRef is not as accurate when an author name is not provided in the query. CrossRef's fuzzy matching will not return a result when an author name is missing and the first page number is incorrect. To avoid this problem, it's best to explicitly include all authors in the XML and then use dashes, if desired, only when rendering the PDF.

Markup Miscellanea


Crediting authorship of articles correctly is of vital importance to researchers. However, the distinctions can be more challenging from a tagging perspective when comparing the <collab>, <role>, and <on-behalf-of> elements.

<collab> is meant to tag organizational authors. For example, "Coudray C, Roussel AM, Arnaud J, Favier A, and the EVA Study Group" would be tagged as:

<collab>EVA Study Group</collab>

However if the authorship is "Coudray C, Roussel AM, Arnaud J, Favier A, for the EVA Study Group" with "for the" instead of "and the", the tagging should be

<on-behalf-of>for the EVA Study Group</on-behalf-of>

because in this case the named individuals are writing in the role of study group members. The element <on-behalf-of> is a specialized form of <role>.

Name–Date Years

When a name–date citation style is used, an additional letter is commonly used to disambiguate two or more references by the same author(s) published in the same year (e.g., "Smith, 2000a" and "Smith, 2000b"). The question is sometimes asked about how to tag these cases, because the "b" after "2000" is not semantically a year but is part of the citation information for maintaining a cross-reference with the body of the article.

When using the <element-citation> model, the only choice is to tag it as <year>2000</year><comment>b</comment>. With <mixed-citation>, the <comment> element can be dropped, and it can simply be tagged as <year>2000</year>b.

Why shouldn't it be tagged as <year>2000b</year>? This model is not appropriate because the letter is not semantically part of the year, and it must be stripped with a regular expression to complete reference linking to a service like PubMed or CrossRef, or deposit of metadata to another organization.

Trailing Punctuation

Trailing punctuation is always PCDATA. Right? Wrong. Consider these cases:

  • Periods at the end of an article title are not part of the title (unless the last word of the title is an abbreviation). But a question mark or exclamation point is part of the title. So when using <element-citation>, you should consider:
    • never stripping a period automatically from the end of an article title when tagging the reference because only a human brain can tell for sure if the last word is an abbreviation; the decision must always be made manually
    • building your rendering transform to add a period after the title only if the reference does not end with a period, question mark, or exclamation point
    This can be a lot of work to do correctly and is perhaps not worth the cost in human time for tagging and transform construction. The easiest road is to a) always include the period inside <article-title>, or b) use <mixed-citation> and always place the period outside of article-title.
  • When the editorial style uses abbreviated journal titles, and periods are included in the abbreviations, the last period of the title may be either part of an abbreviation or the period after the journal title. If the journal title is only one word, it's easy; the period is always outside of <source> because one-word titles are never abbreviated. If it's more than one word, it can be hard to tell. The easiest road may be to either a) include the period inside <source> for a multi-word journal title, or b) use <mixed-citation> and always place the period outside of <source>.

Empty Elements

A few Blue reference elements are empty, notably <anonymous> (for papers authored by "anonymous") and <etal>. In Green, these elements are not empty. Why? When archiving content, it's important to preserve the content exactly as it appears. An anonymous author might be "Anonyme" for an article written in French, or "et al" might have been written in a variety of styles such as "and other authors" or ". . . " (APA sixth edition style).

Special Page Markup

Markup of page ranges in references is usually quite simple. However, special kinds of page indications are sometimes encountered that can be best tagged with alternative markup.


Article identifiers have replaced page numbers in some online-only publications. Article IDs are best designed editorially to distinguish themselves from page numbers by being a minimum of six characters (National Federation of Advanced Information Services, 2009). When article IDs have been found, they should be tagged with <elocation-id>, not <fpage>, as in: <elocation-id>053032</elocation-id>. The challenge with article IDs is largest when publishers have used IDs that do not conform to NFAIS recommendations (i.e. they have fewer than six characters). In this situation, if in doubt, use <fpage> instead.


Sometimes a reference will cite two or more page ranges or discrete pages, e.g. 8–11, 14–19, 40. In such cases, it's tempting to tag the pages as <fpage>8</fpage>-<lpage>11, 14-19, 40</lpage>. However, this markup does not capture the correct last page number. The recommended practice for tagging this type of range is <fpage>8</fpage><lpage>40</lpage><page-range>8-11, 14-19, 40</page-range>.

Reference Linking

Reference tagging has numerous challenges. So why all the effort? It's important to remember that reference tagging provides sufficient semantic information to allow automatic linking of references, and reference linking is of tremendous value (Hunter, 1998, Meyer, 2008). Beyond tagging the references, there are some simple, but important points to improving reference linking quality:

  • Include only semantic data in tags, because boilerplate text can interfere with the matching process of engines like PubMed or CrossRef. For example, don't include a letter from a name–date citation in a year (illustrated above), or when a reference style is verbose — like "Vol. 12, No. 2, pp. 342–356" — don't include "Vol.", etc., in your tagging. Tag this as either
    • element-citation: <volume>12</volume><issue>2</issue><fpage>342</fpage><lpage>356</lpage>
    • mixed-citation: Vol. <volume>12</volume>, No. <issue>2</issue>, pp. <fpage>342</fpage>-<lpage>356</lpage>
  • Be sure the journal title is correct. PubMed, in particular, is unforgiving when journal names are not correct.
  • If possible, add the ISSN to the reference even though it doesn't appear in print. Including the ISSN when linking to CrossRef will improve the link rate when a journal name doesn't match correctly. For example, the journal title "Transportation Research Part F: Traffic Psychology and Behaviour" (ISSN 1369-8478) links just fine in a CrossRef query. However using just the title without the subtitle, i.e. "Transportation Research Part F" (as is often written in references by authors) will fail in a CrossRef query because the short title isn't in the CrossRef database and CrossRef fuzzy matching is unable to make a match in this case without including the ISSN.


Simple journal references that follow a consistent editorial style are relatively easy to tag with the mature JATS models. However, there are always exception cases that can be more challenging even with a robust XML reference model. Tagging of non-journal references presents additional difficulties that can be resolved through careful understanding of the tag suite and implementation of best practices to ensure correct use. The effort to achieve more accurate tagging will pay off with higher quality publications and great success in online linking, which will give researchers the best experience when reading an article.


The author extends his thanks to his colleagues at Inera, especially reference "wizard" Igor Kleschevich for his expertise in reference parsing and tagging and Elizabeth Blake for her editorial insights and patient editing.


  1. Beck J. (2011). NISO Z39.96 The Journal Article Tag Suite (JATS): What Happened to the NLM DTDs? Journal of Electronic Publishing, 14(1). 10​.3998/3336451.0014.106. [PMC free article: PMC3227009] [PubMed: 22140303]
  2. Bernickus, B., Migchielsen, J., Pepping, S., & Schrauwen, R. (2005) Tag by Tag: The Elsevier DTD 5 Family of XML DTDs. http://www​​/framework_authors​/DTDs/ja50_tagbytag5-v1.1.pdf. Accessed on October 1, 2010.
  3. Hunter K. (1998). Adding Value by Adding Links. Journal of Electronic Publishing, 3(3). 10​.3998/3336451.0003.307.
  4. Lukić I. K., Lukić A., Gluncić V., Katavić V., Vucenik V., Marusić A. (2004). Citation and quotation accuracy in three anatomy journals. Clinical Anatomy (New York, N.Y.), 17(7), 534–539. 10​.1002/ca.10255. [PubMed: 15376293]
  5. Meyer C. A. (2008). Reference Accuracy: Best Practices for Making the Links. Journal of Electronic Publishing, 11(2). 10​.3998/3336451.0011.206.
  6. National Center for Biotechnology Information. (2003). NLM Citation Model. http://dtd​​/publishing/tag-library/1​.0/n-3ua0.html. Accessed on August 31, 2011.
  7. National Federation of Advanced Information Services. (2009). Best Practices for Publishing Journal Articles. http://www​​/files/file/Best_Practices_Final_Public​.pdf. Accessed on August 31, 2011.
  8. NLM DTD Working Group. (2007). Meeting Minutes 2003-2007. http://dtd​​.html. Accessed on August 31, 2011. See especially minutes from 2007 meetings starting in February 2007.
  9. Rosenblum, B., & Golfman, I. (2001). E-Journal Archive DTD Feasibility Study. http://www​​.pdf. Accessed on October 1, 2010.
  10. Rosenblum, B. (2010). NLM Journal Publishing DTD Flexibility: How and Why Applications of the NLM DTD Vary Based on Publisher-Specific Requirements. Proceedings of the Journal Article Tag Suite Conference 2010. November 1, 2010. http://www​.ncbi.nlm.nih​.gov/books/NBK47101/. Accessed on August 31, 2011.
  11. U.S. National Library of Medicine (2003). Citing Medicine: The NLM Style Guide for Authors, Editors, and Publishers. http://www​​/pubs/formats/recommendedformats.html. Accessed on August 31, 2011. This publication has been superseded by http://www​.ncbi.nlm.nih​.gov/books/NBK7256/.
  12. Wall Street Journal. (May, 2 2002). Bibliography Mess: The Internet Wreaks Havoc With the Form --- How Do You Cite a Web Page? Wall Street Journal. p. A1.
  13. Wates E., Campbell R. (2007). Author’s version vs. publisher’s version: an analysis of the copy-editing function. Learned Publishing, 20(2), 121–129. 10​.1087/174148507X185090.
Copyright © 2011 by Bruce D. Rosenblum.

The copyright holder grants the U.S. National Library of Medicine permission to archive and post a copy of this paper on the Journal Article Tag Suite Conference proceedings website.

This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License

Bookshelf ID: NBK62602


  • PubReader
  • Print View
  • Cite this Page

Related information

  • PMC
    PubMed Central citations
  • PubMed
    Links to PubMed

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...