Carl Beckhorn wrote:
On Sat, Jan 03, 2009 at 07:07:37PM -0500, WJhonson@aol.com wrote:
Our policy was fashioned in a deliberate way to prevent the use of primary sources where there is no secondary source mention. That was deliberate.
We have always permitted the use of academic research articles published in peer-reviewed journals. These are crucial both for the results they contain and for their link to the historical record. The difficulty is that these sources have to be considered "secondary sources" in order to mesh our best practices with the literal wording of NOR. But many people like to consider them "primary sources".
The idea that these sources should be avoided entirely would simply be silly. The idea that it's better to avoid primary sources entirely is more applicable when "primary source" means "blog post".
I think it's perfectly applicable to journal articles as well. I personally, at least, think it's usually inappropriate to directly cite a new-research result to the journal article, since evaluating journal articles, and placing them in proper historical and disciplinary context is itself a quite difficult bit of original research. That's what survey articles, textbooks, summary mentions in other papers, works like Mathematical Reviews, and so on are for---much better to cite those.
To take an even more direct example, in the medical field, summarizing the results of all the studies that have been done on a particular subject is a "meta-analysis", and a publishable, first-class research project in itself. If no prominent meta-analysis in an area exists, it would be original research for Wikipedia to attempt to directly crawl through the primary literature and write our own, beyond something simple and non-committal like "studies have found both positive [1,2] and negative [3,4] results".
An exception might be important but entirely uncontroversial results, which are not likely to ever get a whole lot of critical analysis. So if some mathematical theorem is proven, I don't have a problem with citing the paper that proves it. But if, say, an antidepressant was "shown to be no better than placebo"---now we're in a controversial, murky area, where anyone can cherry-pick primary sources to make an argument for all possible conclusions.
-Mark