Disclaimer: I've read a lot of stuff on the Forge and on game design blogs and such, but I'm still pretty new to game design. It's a little scary to finally put my ideas out in the open, but here goes.

I recently started designing my first narrativist game, Fall of Prometheus. The game is about fallen gods who are trying to get back into heaven. What are the characters willing to do to get back home? The first playtest went better than expected, considering my lack of experience with GMing or true narrativism. The mechanics did not fall apart as I thought they would (I won't bore you with specifics, since the system is very loosely defined at the moment), and the players got really involved with the premise. I set our playtest in Ancient Greek mythology, and Max and Nick decided that their characters would be the two halves of Zeus, Maximus the giant and Minimus the mouse. We worked together to construct the background, in which Hera took Hades as her lover and threw Zeus from Mt. Olympus by splitting him into these two parts, weakening him. I narrated the first scene: Minimus and Maximus travelled to the Oracle at Delphi to ask what they had to do to regain heaven. She told them that must release those they had imprisoned, and Minimus, winning the conflict to see if knew what she was talking about, decided that that meant the Titans. However, Maximus' womanizing got the better of him (i.e. he failed the roll to overcome one of his Provocations), and despite Minimus' efforts to talk him out of it, the situation devolved into combat with the Oracle's guards. Maximus handily--and cinematically, I might add--dispatched them with his spear and lightning bolts, and the pair headed off in search of the Titans. Max and Nick then spent a long time planning how they would get to the Titans, finally grabbing a book of Greek myths to look at all the obstacles they would have to face. Their excitement was very encouraging!

Max (player of Maximus) was very enthusiastic, and dominated both the action and the gameplay. Understandably, he came out loving it, itching to play again. Nick (player of Minimus), on the other hand, did not have as much fun. In our post-game discussion, he said he preferred more structured and complex rules which got into the nitty-gritty of the scene to the relatively freeform, narration-based rules of Fall of Prometheus. He struggled with the more abstract nature of the game, and he tended to do what Minimus would do instead of using Minimus as a tool to do what he (Nick) wanted to happen in the story. Although we came to an understanding that he just had a different Creative Agenda than Max and I, I wonder if the game itself is flawed. Right now the winner of a conflict gets to take over the narration, so although Nick tried several times to get involved in the scene, the only way he could do so was to bring up conflicts between Minimus and Maximus--conflicts he lost time and again. I want players to be heavily involved with narration, but I think what I need is a better system for handling narrative control. I have another player who wants to join in (playing Athena as the expression of wisdom that Zeus tried to suppress), so I hope to do a second playtest and get more feedback soon.

So, my (perhaps ill-posed) question is this: what sorts of practical system techniques can I use to effectively juggle narrative authority? And credibility, for that matter.

I suppose the potential answers to your question would depend a lot on how your system works, and what determiners it uses (dice, cards, etc.).  Some games you could look at are:

--InSpectres, in which your die value determines both your success level as well as whether you or the GM narrates the outcome.
--Dust Devils, which is based around poker hands.  Best hand wins the conflict, but highest card narrates.
--Primetime Adventures; again, card-based with high card narrating.
--Otherkind, which is unpublished but used dice assignment categories, one of which was narration.  It used to be up on the web, but I can't seem to find it anymore.  Vincent Baker of Lumpley Games wrote it.  If you email him (lumpley@earthlink.net) he could tell you how to find it, I'm sure.
--Dogs in the Vineyard, also by Vincent, prettymuch has the players and GM share narration in a turn-taking manner.

Of course, there are plenty other fine examples, but that's all that I've had personal experience with that springs to mind right now.

Nick

Points for every time he looses, so he can horde them and when he REALLY wants narration, he can spend them to buy a bonus to roll? The more points, the bigger the bonus.

I think you might find some fine-tuning of conflict resolution more effective than rules about narration.  A few times in your post, I read things like "conflict to see if she knew what the oracle was talking about."  This strikes me as task-resolution.  Conflict resolution typically has goal-oriented stakes and consequences, rather than "do I succeed at this action/do I know this thing?" 

It's possible that Nick's disapointment arises from disproportion in player abliity to affect the story.  Task resolution might produce player situations where one player (say Maximus) has all the power to affect the game because his character has all the ability ratings that have the widest scope, while another finds their abilities of limited applicability.  One of the keys of a good conflict resolution game, is giving all players equal ablity to affect any situation (or, as in PTA, rotate their effectiveness from episode to episode).  You can see this in Trollbabe where the same situation might be resolved through Fight, Social, or Magic attributes -- Or in Hero Wars, where Golden Tongue 17 is as effective as Spear Fighting 17.  Many also give the players considerable input into what kind of conflict shows up when their character is featured.

Some narrativist-supporting games do use task resolution -- The Riddle of Steel or Burning Wheel, for example -- and they have a kind of compensating system -- like the points Callan suggested.  Interestingly, both TROS and BW keep narration for the GM.  Narration mechanics really don't make a narrativist game.

For good examples of pure conflict resolution, I second the recommendations of Primetime Adventures and Dogs in the Vineyard.  I'll add my fave, Trollbabe.