Diminishing ability to contribute
Vasiliy Faronov has hit it right on the nail: the idea of helping to build something new and potentially very important was exciting and inspiring back then; there was no lack of glaring holes in the coverage of notable topics, and new editors had much more the feeling than now that they were part of a vanguard community, going against the common idea that a project like this was destined to end up soon on the garbage heap of history. An editor then was much more likely to recognize an other editor's name, and have a notion of their style and interests on Wikipedia.
So I think that today's lower retention can at least partly be ascribed to a lesser sense among new editors of belonging to a community, and also increasingly less commitment to a "cause" as Wikipedia is finding more mainstream acceptance. Wikipedia being less of a cause, and the attendant reduced commitment, are unavoidable consequences of (by itself desirable) wide acceptance. But I feel more could be done about the sense of belonging to a community.
As a start, I think it will definitely help if newbie editors get timely feedback on their edits. It doesn't have to be WikiLove-style feedback, but should of course be positive for good edits, and friendly but not condescending for good-faith not-so-good edits. It could be questions like "do you have a source for that?" or "did you really mean that?", anything that makes clear the newly recruited editor's work does not simply vanish to insignificance in a huge stream of unseen and unacknowledged database updates. --Lambiam 11:56, 11 March 2011 (UTC)
P.S. I just found out about User:EpochFail's NICE script, that makes it easy to see when you're about to revert a newbie's edit and to interact with them. I've not tried it out yet, but would nevertheless encourage people to try it out and help to develop this fledgling tool.
I agree with Vasiliy Faronov and Lambiam that the Gold Rush is over, and we are no longer in the phase where any sort of content was wanted and applauded. The 'pedia is now in a more considered and sophisticated phase where quality matters more than quantity, and this requires a different sort of editor - one who has a willingness to learn and follow our quality standards. Such an editor is harder to find than the slap and dash editor, so numbers will fall. A more pertinent study would be: has the quality growth of the 'pedia occured despite the fall in numbers of editors or as a result of it? SilkTork 12:46, 11 March 2011 (UTC)