In this article, the term SYNTH refers to Wikipedia's policy of forbidding original research by synthesis, and to its forms and nature. SYNTH cautions against original research by synthesis, where an editor combines reliably sourced statements in a way that makes or suggests a new statement not supported by any one of the sources. This essay is intended to help explain the spirit of that policy. In particular, this essay is intended to oppose taking an excessively strict interpretation of the policy in many cases. After all, Wikipedia does not have firm rules.

"SYNTH" as a policy is not a re-statement of Wikipedia's policy against original research. SYNTH provides a guideline for helping to determine the difference between summarizing the information from sources and extrapolating new information from sources.

Here's a hypothetical example of SYNTH:

The editor who added the citation-needed tag is correct.

The sourced numbers would seem to imply, by simple arithmetic, that 500 men went missing. But if the editors working on the article had read the full text of the hypothetical order instead of only the excerpt quoted in source #1, they would have found that men are counted as "missing" for the purposes of the desertion order only if they still haven't been reunited with their unit after five days, whereas source #2 was reporting on the number of men who did not make roll call the evening after the battle, and that only 148 men deserted during the battle of Salamander Creek. To avoid this sort of error, we need a reliable source who has made the same inference, rather than having editors bring together disparate pieces of information themselves. This helps provide a clear solution to many content disputes.

Note that reliable sources used by Wikipedia editors often contain a great deal of synthesis. Wikipedia editors are allowed to use all of this synthesis, since they did not create it but are instead reporting what reliable sources have said. "SYNTH" refers both to a policy forbidding original research by Wikipedia editors by synthesis, and to such synthesis itself. The policy is not unnecessary. You should not be able to use Wikipedia to publish a crackpot theory, even if you can cite sources for all the premises of your arguments. You should not be able to count something as verified when it really isn't, even if you can cite sources for all the premises of your arguments.

Here's an example of using SYNTH to push a fringe theory:

The first and last sentences in that paragraph makes claims which are not supported by any source. Even assuming that each quote is supported by the in-line citations, those two claims are not only SYNTH, but factually incorrect. None of the quotes state explicit support for the hypothesis, nor do any of the quotes indicate that the quoted party is actively pursuing a geological model based on it. Even if reading all of those sources provides one with the distinct impression that the paragraph and all of its claims are true, it is not necessarily so, and it would be wrong to present this view as a fact.

The purpose of the policy is to ensure the accuracy of Wikipedia. The above quote pushes the view that the hypothesis of a Hollow Earth is a serious subject of scientific inquiry, when the truth is that it is not. Absent this policy, articles about controversial subjects or fringe theories could easily devolve into pure advocacy or POV pushing.

Unclear writing can cause editors to commit SYNTH unintentionally. If your writing is unclear to the extent that it seems to contain claims you didn't intend to make, those claims are still being made. However, if a passage is unclear, the first response shouldn't be to challenge its perceived claims as synthetic. You can ask for clarification on the article's talk page, ask the original author directly on their own talk page, or you can boldly clarify the passage yourself.

If someone doesn't like what was said, and they therefore cry SYNTH, others almost certainly will be right to cry foul. Virtually anything can be shoehorned into a broad reading of SYNTH, but the vast majority of it shouldn't be.

A talk page is the right place to claim that something in an article is SYNTH. The policy does not forbid inferences on talk pages that would be SYNTH if made in an article.

Drawing non-trivial inferences is the heart of argument, and on talk pages, you're supposed to present arguments. As the policy consensus says, "The quality of an argument is more important than whether it represents a minority or a majority view." Likewise the guideline PNSD says, "Wikipedia works by building consensus. When conflicts arise, they are resolved through discussion, debate and collaboration." Of course, these are arguments about what the sources and policy say, or what will or won't improve an article, not arguments about the substantive issues themselves. As the talk page guidelines say, "Talk pages are not a forum for editors to argue their personal point of view about a controversial issue. They are a forum to discuss how the points of view of reliable sources should be included in the article, so that the end result is neutral."

Wikipedia doesn't have them, supposedly. But if a policy gets enforced zealously, it can be hard to tell the difference. The solution is to not enforce policies zealously. Never use a policy in such a way that the net effect will be to stop people from improving an article.

SYNTH is when two or more reliably-sourced statements are combined to produce a new thesis that isn't verifiable from the sources. Given just about any two juxtaposed statements, one can imagine that something might be insinuated by the juxtaposition. Don't. If the juxtaposition really does constitute SYNTH, the insinuation will be obvious to everyone. Gray-area cases aren't SYNTH, just unclear writing.

Nothing is insinuated by the mere fact that these sentences are in the same paragraph. The reader would get the same meaning from these sentences if they were in separate paragraphs, or in different parts of the article.

SYNTH is when two or more reliably-sourced statements are combined to produce a new thesis that isn't verifiable from the sources.

Summary is necessary to reduce the information in lengthy sources to an encyclopedic length -- even when the information being summarized comes from multiple sources. It's not necessary to find a source that summarizes the information. As long as what's in the article is an accurate, neutral summary, and each of the statements is verified by an appropriate source, then the summary is also verified by the same sources. Summary is not forbidden by any Wikipedia policy.

For example, a source that reads:

can be safely summarized as

The fact that the source explains why the sky is blue in a rather technical manner does not affect the validity of the summary.

On the contrary, "coming up with summary statements for difficult, involved problems" has been described as "the essence of the NPOV process". [1]

SYNTH is when two or more reliably-sourced statements are combined to produce a new thesis that isn't verifiable from the sources. If you're just explaining the same material in a different way, there's no new thesis.

When the word "thus", "therefore", or any equivalent appears in an article, that doesn't mean SYNTH has been committed. The word helps explain how facts or ideas are related. It doesn't mean that the previously cited facts are being used as sources for the statement that follows it. More often, if the "thus" statement is not followed by an inline citation, the editor believed it unlikely to be challenged and therefore not one of the four things required to have an inline citation.

If what you consider to be SYNTH includes 90% of what's on Wikipedia, your understanding of SYNTH is wrong. If you consider all instances of reading a table to be SYNTH because reading a table requires "synthesizing" the entry in the table with the label of what the table is, your understanding of SYNTH is wrong. Objective, straightforward, and basic descriptions of an illustration are not SYNTH. If your understanding of SYNTH includes any statement whose source continues from one page to the next of an online article, because that involves "synthesizing" stuff from two different URLs, your understanding of SYNTH is wrong.

If you think the letter of the policy supports such an understanding, you may be right. Even policies are subject to occasional revision, and even with consensus there can be blunders in wording. But there was never consensus to obliterate 90% of Wikipedia. The correct response is to go to the policy talk page and get consensus to clarify it, or just ignore it.

If you want to revert something on the grounds that it's SYNTH, you should be able to explain what new thesis is being introduced and why it's not verified by the sources. You don't have to put the whole explanation in the edit summary, but if someone asks on the talk page, you should have something better ready than "Of course it's SYNTH. You prove it isn't." The burden of proof is light: just explaining what new assertion is made will do, and then it's up to the other editor to show that your reading is unreasonable. But in any disagreement, the initial burden of proof is on the person making the claim, and the claim that something is SYNTH is no exception.

If there's something bugging you about an edit, but you're not sure what, why not use SYNTH? After all, everything under the sun can be shoehorned into a broad-enough reading of SYNTH. Well, because it isn't SYNTH. It's shoehorning. To claim SYNTH, you should be able to explain what new claim was made, and what sort of additional research a source would have to do in order to support the claim.

Please note that this section does not apply to potentially controversial statements about living people, about which Wikipedia policy is stricter than about other statements.

What matters is that all material in Wikipedia is verifiable, not that it's actually verified. By this we mean that it is important that a suitable reliable source that supports this material has been published in the real world, not that someone has gotten around to typing up a specific bibliographic citation in the article.

Citations are not an end in themselves. If there's a statement for which no source is cited, that's normally ok, as with the example on Wikipedia:No original research: "Paris is the capital of France" needs no source because no one is likely to object to it, but we know that sources for it exist. Likewise with very many unsourced statements, regardless of whether they could be deduced from sourced statements in the same article, we know the sources exist.

It's part of a policy: no original research. If a putative SYNTH doesn't constitute original research, then it doesn't constitute SYNTH. The section points out that synthesis can and often does constitute original research. It does not follow that all synthesis constitutes original research.

It's not always obvious whether something is SYNTH. To be able to say that something is SYNTH, you have to be able to understand what it says, what the sources say, and whether the sources suffice to verify the assertion. If you don't understand something, don't say it's SYNTH. Say it's too advanced for the article. Say it's unclear writing. Boldly try to clarify it. Allege on the noticeboard that it's SYNTH. But don't revert it indiscriminately for being SYNTH.

If something is obvious to anyone who reads and understands the sources that are supposed to support it, then it's not SYNTH. An example of a perfectly valid citation is given in the guideline on citations, at WP:Bundling: "The sun is pretty big, but the moon is not so big.^[1]" The bundled citation uses one source for the size of the sun, and another for the size of the moon. Neither says that the sun is bigger than the moon, but the article is making that comparison. Given the two sources, the conclusion is obvious. So a typical reader can use the sources to check the accuracy of the comparison.

Beware however of using vague concepts defined in one source to interpret statements in another source. Many concepts in social sciences have vague definitions. The same words may denote slightly different things in different sources. In such cases it's better to attribute the varying concepts rather than to make definite statements about the most general concept you can synthesize. See the Sorites paradox for an example of slippery slope when using vague concepts.

Most Wikipedia articles, those on subjects of general interest, should be comprehensible to a typical secondary-school student. It does not follow that a secondary student should be able to evaluate whether the cited sources suffice to verify a particular assertion. Inevitably, many sources are more advanced than the article. Normally, however, an ordinary educated layperson can understand the sources adequately. If such a person can confirm that the sources suffice to verify the claim, then the claim is not SYNTH – even if a typical secondary-school student would have trouble evaluating the question.

The policy gives one example of something that's not SYNTH and says, "The first paragraph is fine, because each of the sentences is carefully sourced." SYNTH is when two or more reliably-sourced statements are combined to produce a new thesis that isn't verifiable from the sources: it's about what the writing says, not the grammatical structure of how it says it.

If you want to take one source that says "allergic reactions can lead to death" and a second source that says "some people are allergic to apples", and you want to add these up into a new conclusion, "so nobody should eat apples unless they're trying to commit suicide", then there is no possible way to say this without violating SYNTH. It does not matter what grammar or structure you use: introducing your own new idea is prohibited.

On the other hand, if every single idea (considered separately as well as the overall effect) is taken from reliable sources (rather than your own new ideas), then there is no grammar structure or way of expressing these ideas that will make the material violate SYNTH. There are many ways of expressing these concepts that make the article badly written, but bad writing that introduces no novel ideas is bad writing, not a violation of SYNTH.

SYNTH is original research by synthesis, not synthesis per se. In 2004, Jimbo Wales actually contrasted synthesis with original research: "In many cases, the distinction between original research and synthesis of published work will require thoughtful editorial judgment." [2] It seems clear that "synthesis of published work" was assumed to be part of the legitimate role of Wikipedia.

Some old versions of NOR even said "Wikipedia is a secondary source (one that analyzes, assimilates, evaluates, interprets, and/or synthesizes primary sources) or tertiary source ..."[3] (emphasis added). Which is really helpful for those editors with time travel capabilities who can go back and edit Wikipedia before community consensus changed the policy to specifically remove that connotation.

Very old versions of NOR focused on crackpot theories being published whole, not on single statements within the exposition of a body of well-established fact. The under-the-microscope level of scrutiny often practiced now, which demands removal of a single clause until a source can be found that presents the material the same way, is more a result of creep than of well-thought-out policy.

Crackpot theories should be removed quickly; their exclusion is a good reason to have NOR be a central tenet of Wikipedia policy. By contrast, unsourced but uncontroversial individual statements can be left in place indefinitely, with at most a citation-needed template. If such statements were what the policy was primarily about, it would be a minor guideline.

The policy forbids "synthesis of published material that reaches or implies a conclusion not stated by the sources" (emphasis added). A NPOV article gives appropriate weight to all positions, if there are multiple positions on a subject, by including multiple statements. A single assertion, by contrast, can be NPOV only if it doesn't advance any particular position to begin with. So if a single statement, taken in isolation, is NPOV, then it's not SYNTH.

When you look at a case of putative SYNTH, apply the following test. Suppose you took this claim to a journal that does publish original research. Would they (A) vet your article for correctness, documentation, and style, and publish it if it met their standards in those areas? Or would they (B) laugh in your face because your "original research" is utterly devoid of both originality and research, having been common knowledge in the field since ten years before you were born? If you chose (B), it's not original research -- even if it violates the letter of WP:SYNTH.

Treatment of numeric data is an encyclopedic issue: summarization by sum, average, etc. are necessary expedients, and should not be confused with original research. As an example, if a source shows (without any total following it) "1+1+1+1", a Wikipedia article can express the same data with summarization "1+1+1+1=4". (Whether to express only the result 4, if it is not explicitly given by any source, could be a point for discussion, but in any case it is not SYNTH.)

Summarizations based on statistical methods, however, are original research by synthesis, as they involve the reinterpretation of data, and decisions about which statistical methods and significance levels are appropriate. It is common to confuse the arithmetic mean (summarization) with the expected value (mean and standard deviation calculated by a reliable source). If in any doubt about summarization vs. statistical reinterpretation, discuss it first.

Providing a brief, neutral, in-text definition of a new term on first mention in an article is not SYNTH. The Wikipedia Manual of Style at WP:LINKSTYLE asks us to provide a brief, neutral definition of a new term on first mention in an article whenever possible. We are asked not to assume readers can follow links. For example, "...according to Climatic Change,^[1] a peer-reviewed scientific journal." That the defined term has its own Wikipedia article and is wikilinked does not satisfy our obligation to write articles that can be read and understood, standalone, by a wide variety of English language readers, and the brief, neutral, in-text definition is not SYNTH.

• Wikipedia:Core content policies
• Wikipedia:These are not original research (essay)
• Wikipedia:Don't build the Frankenstein (essay)
• Wikipedia:Inaccuracy (essay)