Because this is a definite case where TL, DNR is a reasonable response, I encourage people who are less interested in the full analysis to skip to the few paragraphs following the line "If the sources cited do not explicitly reach the same conclusion, or if the sources cited are not directly related to the subject of the article, then the editor is engaged in original research."
As a counterpart to my line-by-line reading of WP:V, I've had a look at NOR. It is, for the most part, and mercifully, not as bad as WP:V. That said, it still has some serious problems, including what is probably the worst sentence ever put into a Wikipedia policy.
But for the most part, unlike WP:V, it doesn't need a top-down reconsideration, except in one area - it fundamentally misunderstands the relationship between sources and research. But even this can probably be solved through addition rather than deletion.
In any case, the specifics (again, cross-posted to WT:NOR):
* ""Original research" is material for which no reliable source can be found. The only way you can show that your edit is not original research is to produce a reliable published source that contains that material."
Certainly reliable sources are our best practice, and are the simplest way to show that something is not original research. But I know of few areas that have reliable, encyclopedia-like published sources that cover their details to the degree we want to. Certainly my field, literary studies, has a mass of conventional wisdom that is passed around via oral tradition, not written tradition. There are topics that we unquestionably should have articles about that one cannot write a general overview of without relying on that oral tradition. And we should not consider that original research - generally speaking it's the opposite - the stuff that is part of the oral tradition is often the most obvious and basic stuff that is just so obvious that it's not worth anybody's while to publish a book saying it, because everyone would go "Yes, of course, we knew that."
* "Article statements generally should not rely on unclear or inconsistent passages, nor on passing comments. Passages open to interpretation should be precisely cited or avoided"
Oh dear. Now, admittedly, part of my objection here is that, as a good and mainstream literary scholar, I have no idea what a passage that is not open to interpretation would be. Assuming a passage is written in language, it is open for interpretation, and for all but the driest technical literature this opening is significant.
* "Drawing conclusions not evident in the reference is original research regardless of the type of source."
And here we get the sentence that most of our actual practice falls into - not evident to who? We have articles on very hard, very technical topics in multiple fields. How are references in these fields meant to be used? The problem here is *not* the lack of a clear standard, but rather the word "evident." We would do better even with "If the cited source does not clearly support the claim being made, it is original research."
* "For that reason, anyone—without specialist knowledge—who reads the primary source should be able to verify that the Wikipedia passage agrees with the primary source."
This is very possibly the single worst sentence ever enshrined as Wikipedia policy. This sentence is where the first and third problems I identified come to a terrible head. Because secondary source publication is generally a commercial act, what gets published in a secondary source is heavily determined by what is financially viable. That is a very, very different concern from what is useful or important. As a result, primary sources are vital to research - not just scholarly research, but all research. And this becomes even more true as you get to more and more specialist topics - this sentence effectively guts our coverage of science and mathematics.
Let me be clear here, and using a credible expert (my wife, a PhD student in chemistry). It is simply not possible to write an article on [[Single molecule magnets]] from overview-providing secondary sources. Any such article would be badly out of date. Specialist- requiring primary sources are *necessary* to write these articles.
* "All interpretive claims, analyses, or synthetic claims about primary sources must be referenced to a secondary source, rather than original analysis of the primary-source material by Wikipedia editors."
Secondarily, this sentence sets up a distinction that just makes no sense. taking a topic from my field this time, yes, secondary sources exist and are plentiful on [[Jacques Lacan]]. But the good ones are hard too - often just as hard as Lacan's primary texts. Nothing is gained by limiting "interpetive" claims (which, again, I think is a deeply meaningless term) to primary sources only - it's an arbitrary distinction that requires arbitrary source usage that, more to the point, is not in line with respected practices or with how people actually use sources.
This points also to an issue about credentialism. The current policy is all but anti-expert. That is, it basically requires that articles be written either by people who do not know the material or by people who are going to act like they do not know the material. This is *not* what our anti-credentialist position started as, and it's a terrible evolution of it. The original concept of our anti-credentialism was that you could replace a process of credential verification with a high-speed (i.e. wiki-based) open peer review process. That is, a mass of people whose credentials you don't verify can, if given the proper tools, provide as good a peer review as credentialed experts. But this does not assume non-knowledgeable participants - it merely says that we don't check the credentials. The assumption can safely be made that if somebody is editing an article in good faith, they know stuff about the topic. That doesn't eliminate the need for sources (any more than it does in academic research), but it does mean that this "articles must be able to be written by a non-specialist" policy is, frankly, an idiotic poison.
* "If the sources cited do not explicitly reach the same conclusion, or if the sources cited are not directly related to the subject of the article, then the editor is engaged in original research."
This bit, and really the whole section it's in is where the policy most falls apart. What it's trying to do is clear - and the plagiarism example a bit further down is (mostly) spot-on. (It's not unreasonable to mention the Chicago Manual for context there - it's the explicit conclusion-drawing that is a problem)
But as it stands, this sentence describes a research practice that is impossible. The idea that it is possible to simply and unambiguously transport conclusions from a source into a piece of research would be rejected by any credible school of teaching research skills that I am aware of. It is, frankly, the rhetoric and composition equivalent of spontaneous generation.
Indeed, the opposite is increasingly widely accepted. One of the major composition texts these days is called _Everything's an Argument_, and makes the case, essentially, that one cannot organize information without advancing an argument. Research is always an interpretive and synthetic process, and any presentation of research advances a position. The position our articles try to advance is a NPOV position, but it is still a position. NPOV is not "No point of view."
Given all of that, this phrasing of the policy is untenable and inaccurate. The section should be heavily cut down, and coupled with a section that needs to be written. That section must explictly accept that the basic act of organizing information into a NPOV presentation is an act of synthetic research. Connections, interpretations, and organizations are going to have to be introduced, not all of which can be drawn straightforwardly from reliable sources. It should openly acknowledge that determining what the best NPOV presentation and what the most significant viewpoints are is hard and requires a process of open and good faith communication among editors. How to write an encyclopedia article is not something that can be determined mechanically or obviously by an absolute standard or by outsiders brought in to mediate or intervene.
* "This is welcomed because images generally do not propose unpublished ideas or arguments, the core reason behind the NOR policy. "
This is not true. Or, at least, it's no more true of images than it is of words. This section is a frankly bizarre hedge, and a weird moment of liberalness in the policy where none is warranted - especially because the use of images to subtly advance points of view is one of the most insidious and subtle problems we have in this area.
The policy ought to be something along the lines of "images are used to illustrate aspects of the article. Images that are modified or are structured so as to clearly imply or argue for a position are a subtle form of original research that must be watched for."
As I said, the policy is, largely, better than WP:V - it has only two egregious problems, both of which are closely related - its bewildering sense of "self-evident" sources, and its deeply flawed belief about the transparency of assembling information into a tertiary source. This can largely be fixed by new language and by careful rephrasing, but it is a fix that is desperately needed, as right now this page provides bad advice that is not and cannot be usefully applied towards writing an encyclopedia, and that should frankly be largely, if not totally, ignored by responsible editors.
-Phil