References and attribution

A reference is a detailed description of a source a document was built from, given so that a reader can find the work behind a statement. Referencing well means deciding which statements need a reference and which work to cite, keeping each reference attached to its claim while the draft moves, recording and formatting it in the style the destination requires, checking that it identifies a real work that says what the sentence claims, and quoting and crediting other people's words exactly.

What needs a reference

A reference is a detailed description of an information source used to construct a document: a book or a chapter in a book, an article from a journal or a newspaper, or a website, e-book or e-journal. ¹ One guide to medical writing gives these reasons to reference:

• to acknowledge the work of others
• to give credibility and reliability to the writer's work
• to support the arguments made
• to show that the writer has read widely and considered the relevant literature that already exists on the topic
• to validate the points made and verify any quotations
• to allow the reader to locate the sources independently
• to follow academic writing standards
• to avoid plagiarism, since sources that are not acknowledged may lead to the suspicion that the writer is trying to pass off someone else's work as their own

²

The reference section directs the reader to relevant background literature, acknowledges the work of other researchers and supports the authors' statements. ³ A reference list shows where the author obtained background information and indicates awareness of prior work in the area, and a focused list is more valuable to the reader, and to the author, than a very large collection that includes papers only slightly related to the topic. ⁴ Readers often use a reference list as part of their own research, and although it is already a little out of date by the time it is published, it can be a useful starting point; it is also a sort of merit badge for the cited authors. ⁴ A paper should have the right number of references, neither too many nor too few. ⁵

A reference is needed when quoting another person, when paraphrasing or summarising someone else's ideas or opinions, when using someone else's data, diagrams or images, and when using controversial facts or opinions that may be challenged. ⁶ The sources a document may draw on include books and chapters, journal, newspaper and magazine articles, conference proceedings and papers, government publications, media such as videos, DVDs, TV excerpts and sound recordings, personal communications such as letters and e-mails, electronic sources such as websites, e-books, e-journals and software, and information such as tables, graphs and photographs. ⁷

In a manuscript, each statement should have its reference next to it as the point is written, except established scientific facts, which can be stated without one. ⁸ Writers sometimes go a little overboard with references, but where in doubt, a statement should be referenced. ⁸

A writer who moved from science writing found that the most valuable lesson of her first freelance work for another medical writer was that medical writing has to be thoroughly referenced and annotated to absolute flawlessness: her science-writing audience had taken her articles and features as fact, but in medical writing her words could no longer stand alone without referencing line by line. ⁹ She tries to reference everything that way whether or not a client requires it, so that she can always go back if needed. ⁹

Which work to cite

Since their December 2014 update, the ICMJE recommendations state that references should not be used by authors, editors or peer reviewers to promote self-interests, and a personal communication reported with the change adds that they should instead be chosen according to their relevance and usefulness for the reader. ¹⁰ The amendment was introduced to discourage journal editors, peer reviewers and authors from choosing references so as to increase citations of their own or their journal's papers. ¹¹

Subtle forms of falsification are prevalent, and the most ubiquitous of them is including literature references, for example in the discussion, that are inappropriate, outdated or nonexistent. ¹² Authors are often in a rush to finish a manuscript or thesis, and the temptation to add ill-reviewed references is overwhelming, because arguments supported by citations appear more impressive and credible and lead the reader to believe that the author's interpretation is generally accepted. ¹² Reviewers rarely spot faked reference citations unless they know the literature intimately or the citation is grossly suspect. ¹²

The ideal reference citation is the original research source. ⁴ It is best to cite data from original papers rather than from reviews that repurpose or quote them, because original papers can often be misinterpreted or misquoted in subsequent summaries and reviews. ^{13, 14} Grey literature, such as working papers, white papers and abstracts from scientific congresses, should be avoided as a reference. ^{4, 5} Websites should be avoided as references in a scientific report where possible, because they contain a lot of specious data and outright fiction; a writer who must cite one should print a copy of the online material in case it is requested later, because electronic sites can change or disappear. ⁴ A source should never be cited unless the writer has read it and copied it for their electronic or paper files. ⁴

A personal communication should not be cited as a reference, except in extraordinary circumstances in which vital information is unavailable from a public source. ⁵ A paper known to be accepted but not yet published should be cited only if really necessary, and only after obtaining its authors' written permission. ⁵ Such a paper can be cited as in press or forthcoming, and if it is published before the citing article goes to press, the citation can be updated in the page proofs. ^{4, 5}

A methods section can refer to a paper where a detailed description of a method is found instead of describing the method in detail, and a reference to the methods of a former study must be full and accurate. ¹⁵

Referencing while writing

Referencing should not be left until the end of document production: it is part of the writing, and referencing while writing lets the writer develop their ideas, so referencing while writing should be one of the writer's golden rules. ^{1, 16}

One clinician author's method for articles with fewer than 20 references keeps the full reference source on each page or saved document of notes. ¹⁷ When he writes a sentence that calls for a citation, he types the full source in parentheses at the end of the sentence, in bold so that he can find it later, and when the article is almost done and beyond any major changes, he substitutes sequential numbers for the references and moves the citations to the reference list by cutting and pasting. ¹⁷

Referencing styles

Every source used appears in two places: in the text, as brief identifying information directly following any quotation or summary of the source, and in a reference list at the end of the text, which gives detailed information about each source. ¹⁸ Each referencing style has rules for how a reference is written, including punctuation, font and capitalisation, for both the in-text reference and the reference list, and each university faculty, academic journal or book publisher has its own preferred style. ¹⁸ So the writer should check the reference style required before starting to write, and follow its rules. ^{18, 16} Reference formats should conform to the instructions for authors, and an easy way to handle this is to look at the reference list of a recent paper from the journal the writer intends to submit to. ¹⁹

The two main methods are numeric and author–date. ¹ The main referencing styles in medical writing are Vancouver, Harvard and AMA. ^{18, 16} Other styles include MLA, used in literature, arts and humanities, Chicago, used by books, magazines, newspapers and other non-academic publications, and Turabian, used by students in all subjects. ²⁰

The most common format in biomedical journals is the Vancouver style, which numbers references in the text in the order they appear and lists them in numerical order in the reference list. ¹⁹ Its journal reference gives the authors' family names and initials, the article title, the abbreviated journal title, the publication year, month and day (month and day only if available), the volume and issue, and the pages, as in Snowdon J. Severe depression in old age. Medicine Today. 2002 Dec;3(12):40-47. ¹⁹ Another common style, the Harvard system, cites the author's name and the year of publication in the text and lists the complete references alphabetically by the surname of the first author. ¹⁹

The Vancouver style is commonly used in the sciences, medicine, health sciences, health promotion and public health; its numbers let the text flow without intrusions by in-text citations, and let the reader check the reference list easily. ²¹ In it, each reference is assigned an Arabic number when it is first cited, and the number identifies that source each time it is cited again in the document. ²² The numbers can be in round brackets, square brackets or superscript, and they stand outside commas and full stops and inside colons and semicolons. ²² An en rule joins the first and last of inclusive numbers, as in (5-8), commas without spaces separate non-inclusive numbers, as in (3,7,11,14), and the number is still included when the author's name is written in the sentence, as in Brown and colleagues (7) proposed. ²² The reference list is numbered in the order the references are cited, and both its punctuation and the order of references must be correct. ²³ Titles of books and journals are not italicised or placed in quotation marks, journal titles are abbreviated in the format published in Index Medicus with no punctuation, only spaces, and an article or book with six or fewer authors lists them all, while one with seven or more lists the first six followed by et al. ²³ The AMA style, used for medicine, health and the biological sciences, follows the same rules as Vancouver except that book and journal titles are italicised. ²⁴

The Harvard style, also known as the citation method, the APA system or the author/date system, is the most widespread form of referencing and is common in the humanities and social sciences; it cites the author, the date and sometimes the page number. ²⁵ An alternative in the Harvard style uses footnotes: a superscript number in the text corresponds to a note at the bottom of the page giving the details of the source, and the footnotes are numbered consecutively through the document and not repeated when a new page begins. ²⁵ In the text, only the author's surname and the date are needed, in brackets at the end of the sentence before the full stop, as in (Smith, 2000), or the author's name can be part of the sentence with the date in brackets. ²⁶ Page numbers are unnecessary when referring to the general theme of a book but must be included when quoting or referring to data, and are included for journal articles, as in (Smith, 2000, p. 33). ²⁶ Two or three authors are all listed in the text, more than three are given as the first author followed by et al., several works by one author in one year are distinguished by a, b, c after the date, and two or more works cited at one point are given alphabetically, separated by semicolons. ²⁶ The reference list is arranged alphabetically by the authors' surnames, with several works by one author listed from earliest to most recent. ²⁷ Book and journal titles are italicised or underlined, journal titles are not abbreviated, second and subsequent lines are indented to highlight the alphabetical order, all elements after the date are separated by commas with a full stop at the end, and titles of articles, chapters and conference presentations are placed in single quotation marks. ²⁷

Most medical references follow the model of the ICMJE, and the instructions of JAMA, NEJM and the ICMJE differ only very slightly, notably in how they cite books, usually in punctuation or how terminal page numbers are listed. ⁵ In one experienced author's opinion the information needed is virtually the same for all, the copy editor will make the small changes in punctuation if a style is not exactly the journal's, and such differences will not cause the rejection of a good paper. ⁵ He advises numbering references in the order they appear in the text and not alphabetising the list unless that is the journal's style, giving text citation numbers in the journal's style, whether superscript, parentheses or square brackets, and making sure that all the information is included and presented clearly and consistently. ⁵

How many authors to list varies between styles. ^{5, 28, 23} One guide's example article citation lists all authors where there are six or fewer, and the first three followed by et al. where there are seven or more. ⁵ The tenth edition of the AMA Manual of Style names all authors unless there are more than six, in which case the first three are followed by et al., and notes that the National Library of Medicine's guidelines do not limit the number of authors listed but that the manual departs from them for space considerations; so a reference to a paper with 12 authors need not name them all. ²⁸

The elements of a reference

For a journal article, the identifying information includes authors, article title, journal, publication date, volume and issue, and pages. ^{19, 29} A journal name in doubt should be written out, because, for example, Psych could mean psychiatry or psychology; PubMed provides a guide to the standard abbreviation of journal titles, and by custom a journal with a single-word title, such as Nature or Science, is written in full. ⁴

One guide gives an article citation, a chapter in an edited book, which names the chapter author as well as the book's editors, and an electronic source, which gives the address and the date of access, as the three most common types of reference. ⁵

1. Kaddourah A, Basu RK, Bagshaw SM, Goldstein SL. Epidemiology of acute kidney injury in critically ill children and young adults. N Engl J Med. 2017;376(1):11-20.
2. Massart MB. Genetic disorders. In: Paulman P, Taylor RB, editors. Family medicine: principles and practice. 7th ed. New York: Springer; 2017. p. 205-15.
3. Journal of the American Medical Association. Instructions for authors. Available at: http://jamanetwork.com/journals/jama/pages/instructions-forauthors#. Accessed 5 Jan 2017.

⁵

Each kind of source has its own fields, in both a numeric and an author–date form. A chapter gives the chapter author and title as well as the book's editors, title, publisher, place and pages; a book edited or compiled gives the editors as authors and the edition where it is the second or later; conference proceedings give the editors, the title and the conference, its date and location, while a single conference presentation gives its own authors and title within the proceedings; a newspaper or magazine article gives the date and, where applicable, the section and page; multimedia material gives the form of the item, such as DVD, and its director, producer or publisher; an e-journal article gives its address and the date it was cited; and a website or web document gives its address and the date it was accessed or retrieved. ^{30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50} A web document whose date of publication is not known is given n.d. in the author–date form, and an article with no author given begins with its title. ^{41, 51, 52}

Personal communications are not included in the reference list, because the reader cannot trace them; their details, such as the author and the date, are given in the text, as in was confirmed (Higgins S 2000, personal communication, June 6). ^{53, 22, 26}

Passim is a Latin adverb meaning here and there, used in bibliographic references to indicate that the writer has drawn on material scattered throughout the source cited, and it is not an abbreviation. ^{54, 55} One medical editor sees no reason to include it in a list of references, because the definition attaches it to bibliographic references and reference lists in medical literature are not ordinarily bibliographic; professional indexers avoid it because a good index would give each page anyway, and she doubted that any phrase or idea would appear throughout the article cited in the case she was asked about, which was only 15 pages long. ⁵⁵ The Chicago Manual of Style says to use passim sparingly and only after inclusive page numbers indicating a reasonable stretch of text or after a reference to a whole section such as a chapter or part, to follow it with no period unless it ends the citation, since it is a complete word, and no longer to italicise it. ^{55, 56}

Material found online is a vast resource for citation in scholarly articles, and a publication writer can be trained to cite non-print materials associated with journals, accepted publications ahead of print, e-pages and e-publications, e-letters, post-publication peer reviews and data supplements. ⁵⁷

The second edition of the National Library of Medicine's style guide, Citing Medicine, was free online in 2007, and one of the principal aims of that revision was to provide authoritative guidance for citing internet resources, including blogs, wikis and multiple versions of online journal articles. ⁵⁸ It also addresses how to cite audiocassettes, videocassettes, slides, photographs, CD-ROMs and DVDs, as well as print documents. ⁵⁸

In 2012 the Modern Language Association recommended giving the tweeter's real name followed by the user name in parentheses, or the user name without parentheses where the real name is not used on Twitter, then the full text of the tweet in inverted commas, then the date and time of the tweet as read where it was received, then the medium. ⁵⁹ Its example is Athar, Solhaib (ReallyVirtual). 'Helicopter hovering above Abbottabad at 1AM (is rare event)'. 1 May 2011, 3:58. Tweet. ⁵⁹ In the body of the text, the tweet can be quoted by naming the author, as in Solhaib Athar noted that the presence of a helicopter at that hour was 'a rare event', or by putting the surname in parentheses after the quotation. ⁵⁹

Rethinking the reference format

One long-serving copyeditor of research papers argued in 2022 that librarians no longer paste abstracts on index cards, yet the bibliographic details under the heading "References" seem to belong to the Stone Age. ^{60, 61} His premise is that reference lists are used by people, not processed by machines alone, and not only to locate one source but also to peruse the list as a whole, as one would scan a table of contents. ⁶¹ Each element of a reference informs that reader: the year tells how recent or old the source is, the title what the paper is about, and the inclusive page numbers its length. ⁶¹ Some signs of change were already there, such as the latest AMA manual no longer recommending the place of publication for books, and many journals now requiring DOIs. ⁶¹

His proposals concern which elements a reference carries, in what sequence and in what typography. ⁶¹ Authors' initials can be dropped, because they hardly ever distinguish one source from another, tie writers in knots over punctuation, and make the reference differ from an author–date citation, which seldom carries them; and journal titles can be given in full, because abbreviation saved space that mattered in print and does not in the digital world. ⁶¹ The year can be placed straight after the authors, because it then mirrors the author–date citation, makes it easy to skim how current the references are, and shortens the error-prone run of volume, issue and page numbers at the end. ⁶² And typographic coding can serve the reader: italic journal titles and bold volume numbers are being lost to minimalism, a trend he would reconsider, and he suggested bold for the titles of articles and chapters, the actual sources, where journals and books are mere containers, so that a reader can scan the list for what was used, toned down to 60% black if the page looks spotty. ⁶³ He offered these as suggestions whose details a reader may disagree with, and as a step towards references that help readers. ⁶³

Exact quotation and attribution

One referencing guide requires single quotation marks around a direct quotation and double quotation marks around a quotation within a quotation. ²²

Sic, from the Latin sic erat scriptum, thus it had been written, indicates the origin of a mistake in quoted material: the person quoted, not the person quoting, is responsible for the error. ⁶⁴ Typed after a misspelt or incorrect word in quoted material, it shows that this is the original source and that it has not been changed. ²² Perceived inappropriate use or overuse of sic can meet with criticism. ⁶⁴ Sic written at the end of a longish piece of quoted text without indicating where the errors are may be read as criticism of the content itself; marking a trivial deviation, such as a missing capital on a company name so ubiquitous as both noun and verb that it may cease to be capitalised, invites the question whether it warrants a sic; and marking something that is not an error, such as a US spelling, makes the quoter look foolish. ⁶⁴

A copyeditor of a long monograph found in 2010 that several direct quotations contained improper usage and grammatical or syntactical mistakes, and that many of the references could not be found on PubMed or elsewhere, and asked whether correcting them would be overstepping. ⁶⁵ A usage columnist's answer was to adhere as closely as possible to the original wording of a direct quotation and mark what is patently incorrect with "[sic]", because correcting incorrect quotations would take all of an editor's time; the copyeditor could also open the paragraph by saying that the original wording is being followed as closely as possible. ⁶⁶

Two types of error can occur in references that writers take for granted. ⁶⁷ A verified citation can still carry a wrong quotation, because the reference quoted may itself be wrong. ⁶⁸ An article in the Journal of the American Osteopathic Association gives a case from the writings of Andrew Taylor Still, the founder of osteopathy, who wrote a great deal about his experiences and philosophy and has been quoted widely as the discipline's father. ⁶⁸ Questionable copyright dates and misquotes of the original by later writers had led to errors in quoting him, and the article found a startling number of variations on one of his repeated warnings, which in the original reads "To find health should be the object of the doctor. Anyone can find disease" and circulates also as "A doctor's job is to find health, anyone can find disease". ⁶⁸ Changing "should be" to "is" alters the statement's conditional nature, and even a minor change in wording may forever modify the original meaning. ⁶⁸

The columnist who reported the case drew four precautions from it. ⁶⁸

• When quoting someone, be sure to use the original quotation.
• Do not accept someone's paraphrasing as a reference.
• If a paraphrase is unavoidable, be sure it is accurate.
• Make clear that a paraphrase is the writer's own interpretation, and do not attribute it to the original author.

⁶⁸

Changing the punctuation of a quotation can alter its meaning and intent. ⁶⁹ Adding one comma changes the meaning of "Let me call you sweetheart" to "Let me call you, sweetheart", and a tender love note can become a hate note purely by moving its commas and full stops. ⁶⁹ So a word in a quotation should never be changed, nor a comma, a period or any other punctuation, because quotation marks tell the reader that this is exactly what the author wrote, or said. ⁶⁹

A story can circulate without a source. ⁷⁰ A journal columnist had seen an anecdote about a supposedly Roman inscription found in the north of Scotland in several magazines over the years, possibly once in Nature or the BMJ; when he searched for it before reprinting it, he found the exact text, word for word, on several internet sites, always without a source, and asked in print whether the story was true or was itself an attempt to fool the reader. ⁷⁰

Many of the quoted lines people claim to know are wrong in either the statement or its originator; Tarzan, for example, never said "Me Tarzan, you Jane". ⁷¹ Winston Churchill said "I have nothing to offer but blood, toil, tears, and sweat", which most people remember as "blood, sweat, and tears", and the phrase had been uttered much earlier in slightly different form by Cicero and Livy and in a 1611 poem by John Donne. ⁷¹ The remark that academic politics are so vicious because the stakes are so small is often attributed to Henry Kissinger, but several people had voiced the sentiment years before him; it was probably an old maxim and possibly originated with Woodrow Wilson. ⁷¹ The observation that there are more old drunkards than old doctors is usually credited to Poor Richard, but François Rabelais seems to have said it first. ⁷¹

Quotations about writing and editing are misattributed too. ^{71, 72} The line that an editor separates the wheat from the chaff and then prints the chaff is often attributed to Adlai Stevenson, who passed it off as his own, but records seem to show that Elbert Hubbard had written it years before; Hubbard was known to borrow extensively, so he can be credited only with bringing an old cliché to the fore. ⁷¹ The line about becoming a newspaperman for want of honest employment is attributed to Mark Twain although there is no record that he said it. ⁷² The judgement that the only way for a newspaperman to look at a politician is down is disputed and sometimes attributed to Twain or H. L. Mencken, and is more reliably traced to the journalist Frank Simonds. ⁷² "Nobody likes to write, but everybody likes to have written" is a chapter title in Joel Saltzman's If You Can Talk, You Can Write, while a similar line is usually credited to Gloria Steinem and Saltzman rarely is. ⁷² The remark that writers are always selling somebody out is often attributed incorrectly to Nora Ephron or Janet Malcolm and probably belongs to Joan Didion. ⁷² The judgement that a manuscript's good part is not original and its original part is not good is frequently attributed to Samuel Johnson but is apparently not found in any of his writings. ⁷² The line that an output is not writing at all but typing seems to have originated with Truman Capote, though Gore Vidal often used it. ⁷² And the observation that freedom of the press belongs to those who own one, attributed to several people including Mencken, was apparently introduced by A. J. Liebling in the New Yorker in 1960, who hinted that it might come from a previous source. ⁷² Ralph Keyes's The Quote Verifier is a source for some of these attributions. ⁷²

A column on misquotation itself carried a misattribution. ⁷³ It gave the line that if men could get pregnant, abortion would be a sacrament as Gloria Steinem's 1971 report of a salty observation by John Kennedy, which Steinem decades later said had come from an elderly Irish woman driving Kennedy around Boston in the early 1970s. ^{71, 73} A reader questioned the dates: Kennedy was assassinated in November 1963, so he could not have been riding in a cab in the early 1970s. ⁷³ The columnist could not find his original source, but his drafts and notes matched what had been published. ⁷³ A search turned up no reference to President Kennedy, but the same story attached to Florynce Kennedy, an attorney and civil rights activist of the time noted for her sharp comments; The Quote Verifier, one of the sources the column had used, records that Steinem published a compendium of Florynce Kennedy's observations in 1973 and, a decade after attributing the remark to her, admitted that its real author was the elderly Irish woman driving the two of them around Boston. ⁷³ Somewhere, by the columnist's own judgement or someone else's, "Kennedy" had been taken to mean President John F. Kennedy; the column's quotations had all come from different, supposedly reliable sources. ⁷³ He concluded that the mistake helped make the column's point: be very careful with all quotations. ⁷³

How common reference errors are

Reference errors are common enough to have been measured. ^{74, 3, 75} A study of references in 30 articles from 10 journals, published in 2014, found an overall error rate of 18%, with a fall in errors over the period it covered, 2005 to 2011, which its author attributed to more careful editorial checks and increasing use of reference management software. ⁷⁴ A letter to the editor called for the apparent decline to be confirmed, and explained how errors in references can undermine metrics and citation-based assessments of academics, and how they often reflect authors' failure to read the papers they cite. ⁷⁴

Another study examined the references of all 200 original articles published in 2008 in two Pakistani medical journals, counting journal citations only and excluding books, internet articles, websites and newspapers. ³ Only 19 of the 200 articles, 9.5%, were completely free of any error in the references, with no significant difference between the two journals. ³ Of 3,783 references assessed, 26.8% were incorrect, 31% in one journal and 23.4% in the other, a difference that was not statistically significant. ³ Author errors were commonest, in 13% of references, followed by page numbers at 7.9%, article title at 5.9%, journal title at 5%, volume at 0.7% and year at 0.6%; one journal had more article-title errors and the other more journal-title and page-number errors. ³ The authors proposed that the higher than expected level of reference error could be rectified by more careful formatting of the initial manuscript and by giving the final manuscript to the author to proofread. ³ An article containing many reference errors may lead the reader to doubt the accuracy of its other information, and the reference list is also used to help calculate a journal's impact factor, so reference citations should match the source exactly. ³

A spot check of references in a book of 55 papers by different authors, recalled by a manuscript editor in 2008, checked the first ten references in each paper against the original sources in libraries and found that about half had errors that would make a difference. ⁷⁵ The authors were all important enough to have been asked to present new work on thrombosis at a three-day conference, and from then on the editor's division knew it could not trust its authors and that things had to be checked. ⁷⁵ In that editor's view, authors are notoriously bad at copying things down accurately. ⁷⁵ Science is supposedly built on science; if a citation says Jones and Brown 2003 and names a journal, and it turns out to be Brown and Jones, or 2004, or a different journal, it is much harder or impossible to track the reference down and see whether it supports what the manuscript says it supports. ⁷⁵ To that editor references are the most annoying and time-consuming part of a manuscript, especially a long one: a 300-page manuscript can have several hundred references, and he had worked on books with over 2,000. ⁷⁵

The reference section is one of the most neglected parts of scholarly articles and can contain careless spelling, typographical and formatting errors. ⁵⁷ One author blames some of the more easily missed referencing errors in draft manuscripts on excessive haste. ⁷⁶

Checking each reference against its source

Editing references means correcting the citation format of the references in the text, checking the accuracy of all references, and formatting the references to the journal's guidelines. ⁵⁷ The references should be carefully checked for consistency and correctness, and the reader should be able to find the references quoted. ¹⁹ The final review of a manuscript should always include checking the correctness and completeness of the references. ⁷⁷ All reference citations should be verified against an online bibliographic source such as PubMed. ⁵ By 2008, almost any paper could be found online in one form or another, at best as a facsimile on screen, so exactly who a paper's authors are and exactly what its title is could be checked without a telephone or a visit to a library. ⁷⁵

In an editing exercise in a 2019 manual on writing medical publications, refining a raw reference list with author queries removes a duplicate entry and renumbers the rest, completes a page range that gave only the first page, corrects a page range given in reverse order, adds a supplement number, expands a bare report title into a full entry with the issuing organisation, address and date of access, uses en dashes rather than hyphens for page ranges, and drops a personal communication and an internal document label from the list. ⁷⁸

The bibliographic database can itself supply an error. ^{79, 28} In one case in 2008, the name of the senior author of a sleep-medicine paper was spelled with a terminal e in the Medline record and occasionally elsewhere online, while the paper itself and the author's own hospital page spelled the name without it. ⁷⁹ A usage columnist's answer was that mistakes in lists of references should not be perpetuated, that the name should be spelled correctly, and that sic should not be used; Medline has its own rules and style, and the reprint is the horse's mouth. ²⁸

A translator took the von in a German byline, which there meant simply "by", to be part of the author's name, and the publication was cited for years under that name in English-language medical literature until a more experienced translator corrected it. ^{28, 66} A misspelt name in a quotation should not be changed without indicating the correct form, because the work may well be cited in PubMed or elsewhere under the incorrect name, and then the citation cannot be found. ⁶⁶

One medical editor, having read decades earlier, from what she called an impeccable source, that 75% of reference lists contain errors of one kind or another, insisted from then on on getting actual reprints of the cited articles, which presumably are error-free since they come from the authors themselves, and reports that the practice paid off across more than a thousand edited articles. ²⁸ Of a list in which one reference differed from its reprint, she expected that examining the reprints of the other references would turn up other errors. ²⁸

Whether a citation supports its claim

A reference citation used to support a statement should convey the actual meaning of the cited author; one experienced author has seen too many references used to support statements when the cited paper says something entirely different, and PubMed and other sites make it easy to match an author's assertions with the actual words of the cited authors. ⁴ Someone else's list of citations should not be borrowed: a published list can be the basis for research, but each original publication must then be sought out and read carefully to be sure that it says what the citing writer says it does, or someone knowledgeable in the field will spot the discrepancy and correct the writer in a letter to the editor. ⁵ When a paper is finished, the writer should keep the copies of the reference files reviewed and the permission letters received. ⁵

How often a citation misrepresents what it cites has been estimated more than once. ^{80, 81, 82} A 2016 meta-analysis of 15 studies evaluating the accuracy of cited claims in medicine selected studies from keyword searches of PubMed/MEDLINE and Google Scholar that evaluated medical specialties and had a random selection of quotations evaluated by experts. ⁸⁰ Across medical fields, the average rate of quotation errors normalised to the total number of claims was 21.6% (840 of 3,880), and reported error rates by specialty ranged from 6.8% in psychiatry to 38.2% in orthopaedics. ⁸⁰ Studies of the same specialty reported rates that varied considerably: a 2010 study of the orthopaedic literature reported 38.2% for a 2007–2008 sample, while a subsequent study reported 17.9% for a 2009 sample. ⁸⁰ The meta-analysis attributed the high variability between and within specialties partly to differences in sampling and in how experts evaluated claims, and concluded that the average from multiple studies, given in its conclusion as 21.1%, might be a better overall estimate of misquotation in medicine than any individual study. ⁸⁰

A systematic review by the same researcher, presented as a poster in 2016, evaluated the data and methods of studies of the accuracy of cited claims, or quotations, in medicine. ^{83, 81} Across medical fields the mean rate of quotation errors was 14.7% (95% confidence interval 10.7% to 18.6%), and 67.4% of these content errors (58.7% to 76.1%) were major: cited claims that the referenced source fails to substantiate, is unrelated to, or contradicts. ⁸¹ In the studies reporting quotation source errors, the mean ratio of improper secondary, or indirect, citations to references selected was 17.5 (6.6 to 28.4). ⁸¹ The review describes the 14.7% rate, adjusted for methodological differences between studies, as approximately 5% to 10% lower than previous estimates and as a more accurate estimate of the quotation error rate in original medical research articles. ⁸¹

A 2015 systematic review and meta-analysis of quotation accuracy screened 559 studies and included 28 in its main analysis, estimating a major quotation error rate of 11.9% (95% CI 8.4 to 16.6), a minor error rate of 11.5% (95% CI 8.3 to 15.7), and a total quotation error rate of 25.4% (95% CI 19.5 to 32.4). ⁸² Heterogeneity was substantial, and even the lowest estimate of total quotation errors was considerable, at 6.7%. ⁸² Indirect references accounted for about one sixth of all quotation errors. ⁸² Citations are an essential element of manuscripts, but on that estimate 25% do not serve their purpose. ⁸² Further research should elucidate whether and to what degree quotation errors are detrimental to scientific progress. ⁸² The strategies the review suggested for reducing quotation errors were spot checks by editors and reviewers, correct placement of citations in the text, and declarations by authors that they have checked the cited material. ⁸²

Responsibility for the reference list

In one medical editor's view, the accuracy of the reference list is the responsibility of the article's authors, not the charge of the journal's editors. ⁵⁵ In one case a cited article's journal title had been mis-abbreviated and the journal's editors had not caught it, and in her experience checking every reference against the reprint of each cited article takes many days, not hours. ⁵⁵

A freelance writer should understand whether they are responsible for checking the accuracy of references and of information drawn from them. ¹⁴ One freelancer's advice is to ask whether the client requires the writer to highlight hard copies of references to substantiate the content or location of what is cited, which can be onerous but potentially important if a quality control review is planned, and to ask about the quality control process. ¹³ The writer should also ask whether they are expected to check the accuracy of any citation list, because client-supplied documents and reference lists often contain mistakes, mostly typographical, and checking each citation can be time-consuming; that freelancer includes a clause in her proposals saying that she is not responsible for the accuracy of documents the client supplies, but fixes mistakes she happens to notice if time permits. ¹⁴ The writer should ask whether they are responsible for checking the accuracy of information drawn from literature supplied to them: if an investigator's brochure being overhauled turns out to have wrong reference numbers or citations or, worse, inaccurately cited information such as percentages that differ from the primary source, the writer should alert the client as soon as possible and perhaps negotiate extra time and cost, because even with a disclaimer the writer sometimes needs to clean up past mistakes to prevent their perpetuation. ¹⁴ And the writer should ask which reference format or style the client uses, ideally with an official company style guide or an approved document with a reference list as an example. ¹⁴

The reference list as part of a conversation

In one journal editor's view, a paper's reference list is much more than a pro forma inventory of sources kept to satisfy the staid demands of academic publishing. ⁸⁴ It is an invitation to dive into a subject, a cheat sheet for finding the best thinking on a topic, and a guide for pinpointing the paper's place in an ongoing conversation. ⁸⁴ It is also a reminder that the paper was written in some context: the article is not the first word on its subject and probably not the last, and a reader is eavesdropping on a conversation already in progress. ⁸⁴ For example, from the reference list of an article in which William Van Nostran took issue with what Tom Lang had written on the educational background of medical communicators, a reader could track down Lang's previous contributions and also Patrick Moore's 2006 article on legitimising technical communication in English departments, which gives background on institutional tensions between departments of literature and of technical communication. ⁸⁴