Yeah, sorry, I was just fixated on the word "primary". Using primary sources opens the floodgates to articles written from the most biased perspectives. I know it's just semantics, though, because everything else you wrote makes perfect sense.
It suddenly occurred to me that 'primary source' is often used with a different meaning at Wikipedia. In real life, most peer-reviewed scientific papers are considered primary sources (1+3*). I believe many wikipedians consider them to be secondary. I had the real-life definition in mind when I used the word 'primary'.
The guideline at wp-en is actually recommendable, however it is sadly tucked away as a subsection of the original research guideline. Nevertheless, it got translated literally to a number of other projects (I found translations of an older version of the English text in the Spanish, Czech and Dutch guidelines on original research). The equally recommendable German guideline is shorter and very clear. It forbids the use of primary sources of any kind (1+2+3*)and states that Wikipedia content should entirely be based on secondary literature (3*+4).
I personally don't object to the use of 'primary sources' when they are peer-reviewed (1*+3*) and used with care by a user specialized in the field. For factual information (the population of Wyoming, the dimensions of a certain crystal system) I think even non-peer-reviewed primary sources (2+3*)can be used, since they are often the only way to accomplish verifiability (in these two examples: the U.S. census and the mindat.org crystallographic database can respectively be used as refs). However, I think MOST of Wikipedia/Wikiversity/Wikibooks content should be based on a balanced combination of secondary sources (3*+4) to be really reliable. Primary sources (1+2+3*) should only be used when there is no way around it, for example when a certain number or fact is quoted in an infobox template. Using hundreds of references to primary sources (1+2+3*) in (even featured -example) articles is a plague at wp-en. Such a sheer number of unsound sources has two unwanted effects:
- It makes editing more difficult, and
- It wrongly introduces the suggestion that the content can be trusted.
Somehow the guideline at wp-en isn't working well, while the very strict guideline at wp-de (in my eyes: too strict) works generally well (there are exceptions too). I have no explanation for this. Wild guesses: English admins don't dear to implement their own guideline or perhaps the thing is indeed tucked away too far?
Terminology's going to get in the way here. Can you edit the above, so it's clear when you're meaning:
- A primary source in the sense of "a personal or eye witness account, or the original writing" - one which represents the words of its author alone.
- A writing as above, but from a source usually recognized to be reliable despite no formal peer review (government and certain major organizations' official/formal records, etc)
- A source such as a peer reviewed paper that represents the view of its author but has some kind of credible review
- A source where an author discusses and analyzes various of the above dispassionately, to provide coverage of a topic.
I numbered your four types and put red numbers in my last reply. A star means "part of type ...". I feel we should find better terminology/definitions first if we want to continue this discussion. But do we? Our task force could for example end up giving advice on the creation of wizards/guidelines for every source type we can identify ('when to cite a peer-reviewed paper about history'). I'm not sure if this discussion leads to one of the 2-4 most effective recommendations this task force could give for improving quality though.
Often seemingly self-evident primary sources are not reliable. In fact very often the converse is the case. They are explicitly unreliable, erroneous, biased, slanted, POV or downright lies and propaganda. Even basic level history students are taught that in evaluating evidence one has to be critical of its provenance, its intention and the careful handling of all such testimony as corroboration. The current obsession with citing sources has led to rafts and reams of citations which fulfill policy criteria at the marked expense of quality.
That was exactly the point I tried to make in the other thread at this talk page. The use of primary sources should be discouraged when the contributor doesn't have extensive knowledge about a subject. In practice, that means only specialist users should be using primary sources. Other users should be actively discouraged to do so.