What does it take to run a good learning seminar? I'm thinking about running a graduate student seminar in the summer. Having both organized and participated in such seminars in the past, I have witnessed first-hand that, contrary to what one might expect, they can be rather successful. However I haven't been able to quite put my finger on what makes a good seminar good.
Certainly there are obvious necessary conditions for success, such as having sufficiently many (dedicated) participants and at least some semblance of a goal. But in my experience these conditions aren’t at all sufficient.
And on the other hand there are clear pitfalls that should be avoided, such as going too fast or not going fast enough, or scheduling the seminar at 8 in the morning. But there are also more subtle pitfalls that aren't as easily avoided: for example, having consecutive speakers of a certain style that might put off or discourage other participants. (Of course a plausible solution to this specific problem is to not have such people speak one after another, but often this is infeasible.)
So I turn to the collective wisdom of MO: In your experience, what has made a specific learning seminar feel successful to you? Feel free to interpret the word "successful" any way you want. Anecdotes and horror stories welcome.
(Aside: The seminar I'm planning is a "classics in geometry and topology" type of deal. By this I mean, each participant will select a classic paper at the beginning of the summer and then briefly discuss its contents sometime during the course of the seminar. I would consider this seminar successful if, at the end, each participant walks away with a set of their own notes on each paper, explaining why it's important, and containing a sketch of its main ideas and how it fits in the grand scheme of things; the hope is that such a set of notes might prove useful if one were to take a closer look at the paper down the road. If anyone has any experience about running a seminar of this type, then I'd be especially interested in hearing their comments!)
 A: As a partial answer:  Allow for a 10 minute break in the middle of a long
presentation.  It is rare for someone to follow for two hours straight,
and many topics can not be done thoroughly in less time.  Also, the break 
time can be just as productive as the presentation.
Towards the parenthetical remarks about success:  It is good to have a clearly
defined goal.  If everyone is on board with it, then set policy that helps everyone
acheive the goal.  In your example, I would limit/prohibit any side discussion that is
not helping to educate in the topic being presented.  This implies that questions about applications to other fields, or alternative definitions or problems that take one outside the scope of the presentation ( even if it is covered in some other presentation ), and their responses, should be brief or be cut short.  (I would not apply this prohibition if the goal were to stimulate new research or creative problem solving.)
One opportunity I missed out on was when I was an undergraduate sitting in a small graduate PDE class looking at a paper by Leray on Navier-Stokes.  The students took turns
translating and presenting sections of the paper.  If I had been smarter, I would have
met outside the seminar with the graduate students to go over sections of the paper and
help improve my and their presentations.  If I had done that, I would likely remember
the mathematical impact of that paper, rather than forget most of the technical details
needed to push through the proofs.  Perhaps something similar would be effective for your "Classics" seminar: team presentations.
Gerhard "Ask Me About System Design" Paseman, 2010.02.20
A: One big key is for the organizers/leaders of the seminar to be willing to set the tone by asking frequent and sometimes stupid questions.  If the audience gets into a low-question equilibrium rather than a high-question one then you're in trouble.  You want people in the audience to feel like they can and should understand everything that happens, and should hassle the speaker until they do understand everything.  So make sure that there are enough people in the audience who are willing to do this.
A: I'm currently participating in a learning seminar (a "Classics in Geometry/Topology kind of deal", on Teichmuller theory) which seems to work quite well. It is quite tightly managed, with the professor running it both picking the papers we read and giving an introductory lecture before each paper is presented. I appreciate that, since otherwise I'd have trouble knowing how to pick something interesting and understandable from the huge amount of stuff out there.
On the other hand, I can't see any overarching "goal" in the seminar we are striving towards; it's all just various interesting classical topics in the subject. IMHO, this doesn't hurt anything. 
A: I think it is better to have the seminar focused at too low a level than too high.  When the material is at a low level, people may be bored, but at least everyone learns something.  When it is at too high a level, people end up trying to fake knowledge, which is just a waste of time.  I think it is hard for people to let their egos out of the way and really let this happen.  You need a friendly and understanding atmosphere in order to let a seminar be at a level where everybody learns.  One of the best ways to achieve this sort of camaraderie  is to adjourn to a coffee shop or pub afterward where people can continue discussing what they learned that day.  
A: This semester I organized a seminar on "Modern Geometry"(loosely). Another and myself are presenting out of Maclane and Moerdijk's "Sheaves in Geometry and Logic". He is presenting on the more topos theory oriented chapters, while I am presenting Grothendeick topologies Sheaves etc. and later I will move into Dimca's Sheaves in Topology and cover Triangulated Categories and the Derived Picture. The other three speakers are speaking about Monoidal Categories from Maclane's CFTWM, Algebraic Geometry from Ueno, and Representations of Quantum Groups. (This is why I said loosely before). In addition to the five presenters we have about 3 attendees who don't present, and one professor who just heckles.
So far our seminar has been very successful and I think some of the reasons are these(some days it has been not as successful and it was usually when these elements were lacking);


*

*We decided * together* what we wanted to speak about, and all the presenters are really interested in what they are speaking on.

*We go at exactly the pace necessary, so that everybody understands everything. We don't give homework, we don't make people read on their own, we don't just gloss over a point that everyone gets except for one person. This way, nobody feels hopelessly lost.

*We try prepare for our lectures a lot(this is the hardest).

*We meet 5 times a week and we have it scheduled at specific times.

*We are very engaged during the talks, we ask tons of questions and try to find contradictions to the definitions. Together we come up with examples and test them. Sometimes our seminar moves very slow, but usually we leave understanding very well.

*We are friends outside seminar, and we go out to eat after and hang out besides the seminar, this makes us feel comfortable looking stupid in front of one another.


This is all I can think of right now, if I think of anything else, I will edit this post. Anyways, seminar is about to start so I got to go!
A: I have participated in and organized many such seminars.  The advice above, especially Noah's, is very good.  I will add one suggestion.  In my experience, informal seminars in which the presenter has sole responsibility for studying and explaining the week's reading do not work.  It's better if everyone in the group reads paper X or chapter Y, and prepares for themself a list of questions/comments/worked out examples of general statements/confusions, which can be raised during the seminar meeting itself.
My experience here is really based on graduate seminars aimed at reading a single textbook or paper; I'm not sure how well it applies to a seminar that, like yours, aims at treating a different paper each week.  So I may be answering the question in the post title without actually answering your question.  
