See also: Wikipedia:Identifying reliable sources, Wikipedia:No original research, and Wikipedia:Reliable sources (medicine-related articles)

Wikipedia's science articles, while not intended to provide formal instruction, are nonetheless an important and widely-used resource.^[1] Therefore, it is vital that scientific information in articles be based on reliable published sources and accurately reflect current knowledge. These guidelines supplement the general guidelines at Wikipedia:Identifying reliable sources with specific attention to sources appropriate for articles treating topics in science. Ideal sources for these articles include comprehensive reviews in reliable, third-party, published sources, such as reputable scientific journals, widely recognized standard textbooks written by experts in a field, or standard handbooks and reference guides. It can also be useful to reference widely cited seminal papers on a subject to document its history and provide context for the experts' conclusions. While the social context or political impact of science topics may be covered at the same articles using such sources, news reports should generally not be used for reporting actual scientific results or theories.

See Wikipedia:Reliable sources/Noticeboard for queries about the reliability of particular sources for a given purpose, or ask at relevant Wikiprojects such as WikiProject Science, WikiProject Physics, WikiProject Chemistry, WikiProject Biology, or any of the more narrowly targeted science-focused Wikiprojects. While medical topics inherently rely on good science, in case of conflict the guideline Wikipedia:Reliable sources (medicine-related articles) takes precedence for articles primarily treating medical topics.

• A primary source in science is one where the authors directly participated in the research. They filled the test tubes, analyzed the data, or designed the particle accelerator, or at least supervised those who did. Many, but not all, journal articles are primary sources.
• A secondary source in science summarizes one or more primary or secondary sources, usually to provide an overview of the current understanding of a topic. Literature reviews, systematic review articles, topical monographs, specialist textbooks, handbooks, and position statements and literature reviews by major scientific associations are good examples of secondary sources. News reports are also secondary sources, but should be used only with caution. A good secondary source from a reputable publisher will be written by one or more experts in the field, be editorially or peer reviewed, and contain reference to any uncertainty or controversy within a field. University presses and other publishing houses known for publishing reliable science books will document their review process. Do not confuse a scientific review (the thing) with peer review (the activity).
• A tertiary source usually summarizes a range of secondary sources. Encyclopedias, general textbooks, popular science books, and tables of values are tertiary sources.

In general, scientific information in Wikipedia articles should wherever possible be based on published, reliable secondary sources or widely cited tertiary sources. This is to ensure that articles reflect the current state of knowledge about a field, lavishing the most detail on those aspects considered most important by the experts in the field. Reliable primary sources may add to an article, but must be used with care because of the potential for misuse. Edits that rely on primary sources should only describe the conclusions given by the source, and should describe these findings clearly so the edit can be checked by editors with no specialist knowledge. In particular, this description should adhere closely to the interpretation of the data given by the authors, or by other reliable secondary sources. While articles should be kept up to date by citing current literature, care should be taken to avoid recentism focusing too much on new primary sources that have not yet been evaluated by the relevant community. Primary sources should never be cited in support of a conclusion that is not clearly made by the authors or by reliable secondary sources, as defined above (see Wikipedia:No original research). Tertiary sources can provide a valuable overview of a topic, but often oversimplify complex material; it is usually better to cite the secondary sources directly.

Respect secondary sources

Individual primary sources should never be cited or juxtaposed so as to "debunk" or contradict the conclusions of reliable secondary sources, unless the primary source itself directly makes such a claim (see Wikipedia:No original synthesis which advances a position). Controversies or areas of uncertainty should be illustrated with reliable secondary sources describing the varying viewpoints. The use and presentation of primary sources should also respect Wikipedia's policies on undue weight; that is, primary sources favoring a minority opinion should not be aggregated or presented devoid of context in such a way as to undermine proportionate representation of expert opinion in a field. If mainstream secondary sources in a field do not consider a detail or opinion relevant, it is not appropriate to cover it at that article; such details and opinions may be desirable at an article on a sub-topic or at a separate article, with linking governed by WP:SPINOUT and WP:ONEWAY.

If an important scientific result is so new that no reliable reviews have been published citing it, it may in some cases be helpful to cite the primary source that reported the result. The more revolutionary the result, the less likely that an exception should be made. Although popular-press news articles and press releases often tout the latest experiments, they often exaggerate or speak of "revolutionary" results where the authors refer to the context of the gradual progress of the field. Including an accessibility link to such a source may aid in reader comprehension, but the language of the actual study should be used; more detailed and less sensational lay sources are preferred. After enough time has passed for a review to be published in the area, the review should be cited in preference to the primary study. If after a reasonable period of time no review is published and few or no researchers outside the originating lab have cited a study, the primary source should be removed as not reporting an important result. When in doubt, omit mention of the primary study, as per WP:RECENTISM. It is not the intention of Wikipedia articles to provide a running blow-by-blow of developing science, but to record how the experts in a field have synthesized the literature.

Summarize scientific consensus

Scientific journals are the best place to find primary source articles about experiments. Any serious scientific journal is peer-reviewed. Be careful of material in a journal that is not peer-reviewed, especially if reporting material in a different field (see Marty Rimm and the Sokal affair).

The fact that a statement is published in a refereed journal does not make it true. Even a well-designed experiment or study can produce flawed results or fall victim to deliberate fraud. (See the Retracted article on neurotoxicity of ecstasy and the Schön affair). There is an informal hierarchy of journals, abetted by the publish or perish culture of academia. Preference should be given to citing articles in top tier journals wherever possible. Similarly, if you find dubious unreferenced or poorly referenced text in an article, your first question should be does including this material add to the full and accurate summary of the topic rather than can I track down a source somewhere that supports this.

The fact that a statement is published in a refereed journal does not make it relevant. Many ideas are proposed and disregarded in the context of scientific discourse. If an idea is cited by only a few research groups ("few" varying by field), it should receive very little or no weight in our articles. Very new papers should be used sparingly if at all until enough time has passed to make this assessment - there is no deadline. Additionally, material that is appropriate for a highly focused article on one specific part of a field may not be appropriate for a higher level article about the field as a whole.

The Neutrality and No original research policies demand that we present the prevailing scientific consensus, which can be found in recent, authoritative review articles or textbooks, high quality journal articles, and some forms of monographs. Although significant-minority views are welcome in Wikipedia, such views must always be described as such and presented in the context of their acceptance by experts in the field. It can be helpful to create a subarticle devoted to the minority position, leaving a brief summary in the main article; both the summary and the subarticle must give due weight to the state of the field as a whole. Positions generally rejected or ignored by the experts in a field should not be mentioned in the main article on a topic except potentially in a historical context; such articles should be described in context of and link to the main article, though the reverse does not generally hold. The views of tiny minorities need not be reported if they are not notable or prominent.

Make readers aware of legitimate uncertainty or controversy within the particular field of study. A well-referenced article will point to specific journal articles or specific theories proposed by specific researchers. Many values, such as the masses of transuranian elements or the isotopic composition of the solar system, have an associated uncertainty, and even up-to-date highly reliable sources may report slightly different values. Where there is no clear reason to report solely one of several values, cite a few of the highest quality sources after discussion on the article's talkpage or the appropriate Wikiproject. For values or classes of values affecting many articles, consistency across articles and Wikiproject-level discussion should be preferred.

Assess evidence quality

Several systems exist for assessing the quality of available evidence on scientific subjects, and these should be kept in mind while assessing whether a particular idea or viewpoint is well-accepted by the relevant academic community. Such evidence should include reviews of the literature including the work of several different research groups. Individual papers often disagree with each other, but there are several indicators that may be assessed even without specialist knowledge to differentiate high quality papers from low, including:

• methods are clearly explained and are appropriate to the experiment
• model fitting and statistical analysis are meaningful and appropriate
• uncertainty and the paper's place in the wider scientific discourse are acknowledged
• funding sources and any potential conflicts of interest are disclosed
• the authors and the paper itself are widely cited by other researchers in the field; in most scientific fields, the order of the author list usually indicates importance of each researcher's contribution to the article, except that the final author is commonly the senior researcher in charge of the laboratory or research group where the work was done. Both of these conventions may vary by field, journal, and paper.
• recognized experts in the field have commented or offered informal opinion

Cutting edge science is built on the foundation of previous research, and paradigms almost always change only slowly. Preliminary results, whether reported in the popular press, a conference abstract, or a peer-reviewed journal, are a form of anecdote and generally fall below the minimum requirements of reliable science sources. Exceptional or surprising claims should not be presented as authoritative, nor should the description of a broad consensus view be presented as less well-founded until such exceptional claims are replicated or widely cited.

Speculative proposals and early-stage research should not be cited in ways that suggest wide acceptance. For example, results that have been reported only in conference proceedings or a researcher's website are unlikely to be appropriate for inclusion except when reported as such in the author's biography. A secondary source reporting on preliminary results might be appropriate as part of a well-documented section on research directions in field. To prevent misunderstandings, the text should clearly identify the level of research cited. If a result does not indicate its place in the scientific discourse, it is unlikely to be reliable.

For example, every year, people propose modifications to general relativity or publish results that call some aspect of the theory into question. Usually these ideas are proposed by serious researchers who pose the question how can these results be understood in terms of the theory they seem to contradict? as part of an endeavor to understand the results more deeply. Such nuances are often missed in popular press reports, but should be included in articles if the proposed modification is cited. Sometimes "revolutionary" ideas are proposed by cranks or are otherwise ignored by researchers; such ideas should be presented only in the context of the broader field and only in articles devoted to the proponent(s) or specific to the idea. Until a significant fraction of the astrophysics community indicates doubt as to the general validity of the theory, the articles treating general relativity should not imply any such doubt.

Use up-to-date evidence

Here are some rules of thumb for keeping an article up-to-date while maintaining the more-important goal of reliably reflecting the current state of a field of research. These guidelines are appropriate for actively-researched areas with many primary sources and several reviews, and they may need to be relaxed for mature fields or in areas where little progress is being made and few reviews are being published.

• Look for reviews published in the last five years or so, preferably in the last two or three years. The range of reviews examined should be wide enough to catch at least one full review cycle, containing newer reviews written and published in the light of older ones and of more-recent primary studies.
• Within this range, things can be tricky. Although the most-recent reviews include later research results, do not automatically give more weight to the review that happens to have been published most recently, as this is recentism. The prominence of the publishing journal, the quality and comprehensiveness of the review, and the respectability of the authors should also be taken into account.
• Prefer recent reviews to older primary sources on the same topic. If recent reviews do not mention an older primary source or result, the older source is dubious. For example, the articles superconductivity and List of superconductors might mention the hot-off-the-presses latest material or model found to undergo the transition, but such observations should be treated as tentative until confirmed by another research group or affirmed by a broad review of the field. More detail should be devoted to discussion supported by recent reviews.

These are just rules of thumb. There are exceptions:

• History sections often cite older work, for obvious reasons.
• An older primary source that is seminal, replicated, and often-cited in reviews is notable in its own right and can be mentioned in the main text in a context established by reviews.
• Consider scope and focus: articles on broader topics and more mature fields should contain less primary research than articles on narrow, actively researched topics.
• Editors should be especially leery of citing papers making exceptional claims until the relevant community has evaluated the evidence. If a result is cited only by the research group originating the claim and ignored by the rest of the field, it should probably not be included even if present in a review authored by the group. Blogs by relevant subject matter experts may be useful in talk page evaluation of the relevance of very new results, though they should rarely be cited themselves (see below).
• Sometimes scientific results have or are taken to have political or social relevance. Wikipedia articles should avoid sensationalism, and should strictly follow the relevant research community in according weight to such results. Reporting on political and social impacts is often done in separate articles, and is beyond the scope of this guideline.

Use independent sources

Many scientific claims lack independent replication or confirmation of the legitimacy of statements made by proponents. In such cases, reliable sources may be much more difficult to find and unreliable sources can often be more readily available. Especially when writing about ideas not supported or contradicted by mainstream research, it is vital that third-party, independent sources be used. Sources written and reviewed by the advocates of such marginal ideas can be used to describe notable personal opinions, but extreme care should be taken when using such sources lest the more controversial aspects of their opinions be taken at face value or, worse, asserted as fact. If the only independent sources discussing a subject are of low quality, then it is likely that the subject itself is not notable enough for inclusion. For example, coverage of individual perpetual motion machines should focus on their importance to the creator's biography (if notable) or actual impact (did a large company invest in the inventor? did an eminent scientist comment on the device?) rather than a detailed recapitulation of the supposed principles involved.

A Wikipedia article should cite the best and most-reliable sources regardless of whether they require a fee or a subscription. When all else is equal, it is preferable to cite a source whose full text is freely readable so that readers can follow the link to the source and editors can double check the content. Journals more likely to be available at a reader's local university library should also be preferred. While most high-quality journals require a payment or subscription for access, some, such as Proceedings of the National Academy of Sciences publish only freely-readable articles; others, such as Physical Review, publish a few freely-readable articles even though most are not free; still others use delayed open access.

When searching for sources, it is wise to skim-read everything available, including abstracts of papers you cannot fully access, and use that to get a feel for expert opinion on the most important aspects of a topic. However, when it comes to actually writing a Wikipedia article, it is generally not a good idea to cite a source having read only its abstract, as the abstract necessarily presents a stripped-down version of the conclusions and omits the background that can be crucial for understanding exactly what the source says. You may need to visit a university library to access the full text, or ask somebody at the WikiProject Resource Exchange or at a relevant WikiProject either to provide you with a copy or to read the source for you and summarize what it says. If neither is possible, you may need to settle for using a lower-impact source or even just an abstract, with an eye to updating or replacing the text when a better or more complete source becomes available.

Sometimes, a paper or series of papers will be summarized by an expert in the field, usually in a research journal with a target audience of other researchers in the field. Such summaries are likely to be more detailed and reliable than similar commentary in the popular press while providing context for the impact of a result or relative importance of a line of research. If you have access to both the original source(s) and the summary and you find the summary helpful, it is good practice to cite both sources together (see Formatting citations for details).

Research journals

Articles published in respected peer-reviewed research journals are ?preferred for up-to-date reliable information. Scientific literature contains two major types of sources: primary publications describe novel research for the first time, and review articles summarize and integrate a topic of research into an overall view. Research journals generally publish a mix of primary and secondary sources, though some may concentrate on particular types. In science, primary sources include descriptions of an individual experiment or a series of experiments by the same research group; secondary sources include independent review articles summarizing a line of research or rectifying apparently discordant results. It is usually best to use reviews where possible, as these give a more balanced and general perspective of a topic, and can be easier to understand.

In many fields, research journals serve their community by also publishing less technical material such as biographies and obituaries. Although almost all such material will count as a reliable source, not all the material is equally useful. Journal articles come in many types, including: original research, reviews, expert summaries, news, editorials, advocacy pieces, speculation, book reviews, correspondence, biographies, and eulogies. Original research papers are primary sources; although they normally contain a previous-work section that functions as a secondary source, these sections are typically less reliable and comprehensive than reviews. A general narrative review of a subject by an expert in the field makes a good secondary source that can be used to cover various aspects of a subject within a Wikipedia article. Such reviews typically contain no original research but can make interpretations and draw conclusions from primary sources that no Wikipedia editor would be allowed to do. A systematic review uses a reproducible methodology to select primary studies meeting explicit criteria in order to answer a specific question. Such reviews should be more reliable, accurate and less prone to bias than a narrative review.^[2] However, a systematic review's focus on answering one question limits its usage as a source on Wikipedia.

Core basic science journals include such publications as Science, Nature, and the subject-specific journals published by the several national associations.

Books

Science textbooks published by academic publishers are often excellent secondary sources. If a book has as its declared target audience students, it may not be as complete as a monograph or chapter in a book intended for professionals or postgraduates. Ensure the book is up-to-date, unless a historical perspective is required. Major academic publishers (Elsevier, Spinger Verlag, Informa, and the several university presses to name a few) publish specialized book series with good editorial oversight; volumes in these series summarize the latest research in narrow areas usually in a more extensive format than journal reviews. Specialized encyclopedias published by such established publishers are often of good quality, but may be too terse for detailed articles.

Popular science books can be useful tertiary sources, though information may be oversimplified or lacking in nuance or the full range of opinion in a field may not be adequately represented. Even in such cases, it may be useful to seek them out as an example of the material being presented in a fashion accessible to non-scientists.

Most self-published books or books published by vanity presses undergo no independent fact-checking or peer review and consequently are not reliable sources.

Scientific organizations

Statements and information from reputable major scientific bodies may be valuable encyclopedic sources. These bodies include the various national societies, such as the Royal Society, the American Physical Society, or the Royal Australian Chemical Institute. The reliability of these sources range from formal scientific reports, which can be the equal of the best reviews published in journals, through public guides and service announcements, which have the advantage of being freely readable but are generally less authoritative than the underlying literature. Such organizations often contain working groups and subcommittees, which cannot be presumed to speak for the society as a whole.

Advocacy organizations formed for a specific purpose or to advance a cause may be composed of scientists and mimic the structure and naming conventions of the general purpose societies. Statements and reports from such organizations are not reliable except to cite the organization's opinion or position. If such statements are necessary to the coverage of a topic, they must be attributed and the role of the organization made clear.

Popular press

The popular press is generally not of itself a reliable source for science information in articles, though they are often readily accessible and can contain valuable supplemental information of a social, biographical, current-affairs, or historical nature. Most science news articles fail to discuss important issues such as the certainty of a conclusion, how a result has been received by experts in the field, the context of related results and theories, and barriers to wide-spread adoption or realization of an idea. Articles in newspapers and popular magazines generally lack the context to judge experimental results; be particularly wary of any result reported as surprising or revolutionary, which may be an indication of exaggeration or worse. Popular press articles tend to overemphasize the certainty of any result, for instance presenting a new theory as substantiated or a new technology as just around the corner. Newspapers and magazines may also publish articles about scientific results before those results have been published in a peer-reviewed journal or reproduced by other experimenters. Such articles may be based uncritically on a press release, which can be a biased source even when issued by the public relations department of a university or national laboratory. News articles also tend neither to report adequately on the scientific methodology and the experimental error, nor to express risk or uncertainty in meaningful terms.

A news article should therefore not be used as a sole source for a scientific fact or figure, nor should they be considered when describing what aspects of a field the relevant experts consider interesting, surprising, or controversial. Editors are encouraged to seek out the scholarly research behind the news story; good quality science news articles will name their sources. One possibility is to cite a higher-quality source along with a more-accessible popular source, for example with the |laysummary= parameter of {{Cite journal}}.

On the other hand, the high-quality popular press can be a good resource for presenting science to a non-technical audience, and often as a source in its own right to supplement (but not supplant) the peer reviewed literature. For example, while popular science magazines such as Scientific American, Discover, and Popular Science are not peer reviewed, they sometimes feature articles written by experts that explain scientific subjects in plain English. As the quality of press coverage of science ranges from excellent to irresponsible, use common sense, and see how well the source fits the verifiability policy and the general reliable sources guideline.

Other sources

Press releases, blogs, newsletters, advocacy and self-help publications, and other sources offer a broad spectrum of scientific information ranging from factual to fraudulent, with a high percentage being of low quality. As much as possible Wikipedia articles should cite the literature directly, and editors should bear in mind that a particular source may introduce a spin not present in the original paper or present a result not supported by the research. Conference abstracts are often incomplete and preliminary, and may be contradicted if and when the data are published; they should be avoided. Personal or group blogs from prominent scientists writing in their field of expertise may be usable when properly attributed. Nature Blogs, ScienceBlogs, and Discover blogs host many such experts, as do more specific portals such as the public outreach and service blogs at the Large Hadron Collider blogs or the more STEM policy oriented blog hosted by the American Physical Society.

Search engines and academic databases are commonly used to find sources. Each system has quirks, advantages, and disadvantages, and may not return the results that you need unless used carefully. It typically takes experience and practice to recognize when a search has not been effective; even if you find useful sources, you may have missed other sources that would have been more useful, or you may generate pages and pages of less-than-useful material. A good strategy for avoiding sole reliance on search engines is to find a few recent high-quality sources and follow the citations backwards and forwards to see what your search engine may have missed. Limiting a general search using the key words (usually listed under a paper's abstract), or using a semantic search engine may help focus results to the relevant topic. Some resources, such as Google Scholar and Physical Review, also list the papers citing a particular paper; these results may not be comprehensive, especially tending to miss citations that are not well-formed, but the results can be useful both in finding additional sources and as a rough metric of the impact of a particular paper on the field in general. It can also be helpful to perform a plain web search rather than one of scholarly articles only.

Other useful search engines include

• Web of Science
• InfoTrac
• Scopus
• Google Scholar
• Google Books often offers readers peek at a few sentences even when full access is not granted, and can help editors find reliable sources quickly, either by looking at the book's references or by citing the book itself. Check that a particular book is published by a reliable academic publishing house.
• arXiv is a preprint server; near-final versions of many physics and astronomy papers may be read freely, but these papers have not yet undergone peer review, and any citation should be checked against the final version.
• Astrophysics Data System covers astronomy and physics papers.
• University librarians are often aware of specialized resources, and can be exceedingly helpful when approached in a friendly and open fashion.
• Journals occasionally devote all or most of an issue to a particular topic or sub-field. Such issues can provide a valuable snapshot of the current state and research directions of a field.

Approaching the problem from the other end, many large research organizations and funding agencies publish research highlights. These summaries can be helpful in recognizing the most important result of a piece of research or in ascertaining current research directions.

A citation should document precisely how to access a source. Normally, citations should contain a digital object identifier (DOI) if available. A common practice is to supply a uniform resource locator (URL) to a source if and only if full text is freely readable. Check that the URL given does not depend on a cookie on your machine or IP-based subscription access. Some journals offer free access for only a limited period after publication, so check for linkrot when updating references. WP:CHECKLINKS semi-automates this process. If the {{Cite journal}} template is used, all this information can be supplied with the |doi=, and |url= parameters, respectively. If you are citing a source along with an expert summary, it is helpful to list them together, with the main source first to indicate that it is more authoritative. For example:

Griffin SO, Regnier E, Griffin PM, Huntley V (2007). "Effectiveness of fluoride in preventing caries in adults". J Dent Res. 86 (5): 410–5. doi:10.1177/154405910708600504. PMID 17452559.{{cite journal}}: CS1 maint: multiple names: authors list (link) Summary: Yeung CA (2007). "Fluoride prevents caries among adults of all ages". Evid Based Dent. 8 (3): 72–3. doi:10.1038/sj.ebd.6400506. PMID 17891121.

If a source is available in both HTML and some other form, normally the HTML form should be linked, as it is more likely to work on a wider variety of browsers. If the full text of a source is found in a location other than at the publisher's website, check that the copy does not violate copyright before linking it and be aware that the text may have been altered from the original version.

• Dispatches: Sources in biology and medicine. The Wikipedia Signpost (2008-06-30)
• Reliable source examples in physical sciences, mathematics and medicine