Citations or links aren't required here per se, any more than they are on any other StackExchange site. On site such as ours, however, citations and links are of far greater value and importance than on StackOverflow or other technology stacks in helping the end user evaluate the usefulness of a question.
- There is no Fiddle for English
Answers on StackExchange are supposed to be based on their helpfulness (i.e. usefulness). When the question is about development, system administration, software usage, or other technology tasks, the overriding factor in every case will be the answer to the question
Does it work?
In other words, the original posters can look at the answers provided, plug the solutions into their… solutions… and get nearly immediate results. The package will compile, or install, or execute when it wasn't before. The error message or condition will go away. The animation will play in Netscape Navigator 3.0 Gold. Sometimes, a one-line answer that fixes a typo is all you need.
There is no JSFiddle for English usage— nor for many of the other subject areas StackExchange now covers. There's no tool where you can copy and paste in the solution to Could I fake death to stop being Force-choked? to see if it "works." There's no mathematical model that demonstrates How to avoid procrastination during the research phase of my PhD? There's no test case that confirms Is it permissible to erase the name of G-d from an E-reader?.
Since the end user is the one seeking the knowledge from experts, how then can he or she know what the "right" answer will be? If the answerer's personal authority is not personally established— and aside from the odd Barrie England or John Lawler, it rarely is— the answerer must in turn make some appeal to authority ("here's a paper in a linguistics journal") or to popularity ("here's BNC frequency chart and some Amazon previews from books showing that this phrasing is used").
[We all understand that appeals to authority or popularity are logical fallacies, but almost no questions on EL&U, or anywhere on SE, are formulated as requests for logical proofs, so that objection is not a final one. Prove that Caligula succeeded Tiberius. "Because Tacitus said so" is usually good enough.]
The "softer" the stack, the more that outside citations will substantially improve an answer; the "harder" the stack, the more that the solution is self-evident. A stack like History.SE requires a great deal of outside references to make an answer useful. A stack like SuperUser requires very little, and references to secondary sources in particular will clutter more than elucidate. EL&U lies somewhere in the middle. Note that by using the terms "hard" and "soft" I am not suggesting that the "soft" stacks are less rigorous. Skeptics.SE is extremely strict about requiring outside citations and avoiding original research and personal opinions, and no one would question their rigor.
- Misunderstanding abounds
Many questions about usage can be answered by any reasonably competent speaker. Many of my answers come down to me "knowing" the answer as a native speaker, then hunting for a justification. But establishing one's competence in a written forum is fraught.
Furthermore, unlike the finer points of, say, Sharepoint or Solaris, many people hold themselves to be excellent at English, whether as a mother tongue or as a secondary language, so there is a much larger pool of random answerers lurking the site. And we have also absorbed innumerable "rules" which are either oversimplifications (e.g. "I is nominative and me is objective") or whole cloth inventions (e.g. "starting a sentence with hopefully is ungrammatical"), but which have been parroted textbook to textbook, teacher to teacher, rendering those of us with "excellent" educations overconfident in our knowledge.
Thus, it's far better if someone can identify the origin of their knowledge so that the questioner may evaluate it and its applicability or currency to their situation.
- "Correctness" is not the same as "helpfulness"
Third, even if the answer is correct and a cited rule is sound, the answer may still not be helpful. "How do I get to Carnegie Hall?" "Practice" is correct, but not helpful. "The R Train" is correct, and somewhat helpful, but not complete. "The easiest way is the N/Q/R Train to 57th Street, but there are even more options on their website at www.example.net" is correct, and helpful, and complete.
Not every external link is created equal, of course. I may prefer to link to the official MTA page rather than a commercial site about visiting Carnegie Hall. The point is that the addition of a good reference almost certainly makes an answer more helpful, because it comes from another source that can help answer other questions, or goes into thorough detail that is difficult to achieve in the space of a StackExchange answer.
- Our standards for references and examples are extremely loose
No one is going to mistake any part of SE for a scholarly enterprise. We're hardly demanding that quotations come from peer-reviewed academic journals and papers authored by prize-winning tenured professors at tier one universities. An online dictionary or a language blog may be sufficient. I take many examples from freely available newspapers and magazines. We even widely tolerate Wikipedia and Wiktionary links, even though I would never tolerate them in my own organization.
The failure to find a justification for an answer is thus not usually a problem of paywalls, firewalls Great and otherwise, or other technical, financial, or legal hurdles. In my view, it reflects instead an unwillingness of the answerer to invest any effort into the answer. It isn't against the rules; it shouldn't get you banned. But it shouldn't exactly be encouraged or rewarded, either.