But, what sense?
Taking my cue from Walton's book Mimesis as Make-believe again, let's say that a roleplaying game is a game of make-believe. Depending on what is happening around the table, we are to imagine certain things. But is everybody to imagine the same things? It would seem so, at least on the surface. If I, playing my trollbabe Ingirid, state that I run towards the troll shaman and punch him on the nose, then surely everybody is to imagine that Ingirid runs towards the troll shaman and punch him on the nose.
But there is a difference, and is related to the role of the appreciator. Everybody is to imagine Ingirid punching the shaman, but in addition I am to imagine being Ingirid and punching the shaman on the nose. It is not just fictional that Ingirid punches the shaman, it is fictional that I punch the shaman.
Maybe putting this in terms of having to imagine things, and things being fictional, seems wrong. It would be more intuitive to say that some systems enable me to imagine myself making those decisions and performing those actions, and other systems do not enable me to do so - they merely enable me to imagine that some character makes those decisions and performs those actions. But I suspect that the parts of the system that enable me to imagine myself as fictionally identical to the character are the same parts of the system that make it fictional that I am identical to the character. (Still following me?) For instance, "each player is allowed to make the decisions of his/her character alone and unhindered by other players" is one such part. (Having or not having this in place is one big difference between Trollbabe and Universalis.) It enables me to identify myself with the character (in the strong sense of imagining being the character), but it arguably also makes it fictional that I am the character.
Anyway, let's not dwell on that too much. What is more important is to identify techniques that will aid / hinder immersion. I'll make several observations, the first of which is probably rather uncontroversial.
Authorship over a character
Fictional identification with a character (immersion) is helped by mechanics that give total authorship over the character's beliefs, desires, decisions and so forth to one player. In Universalis, you do not have authorship over the decisions of any character, which hinders identification with any character. In Sorcerer or Trollbabe you do, which aids identification. In My Life with Master, you only have authorship within certain bounds (some decision are made by the dice and, more importantly, the kind of scene you're in severly limits the possible desires and choices of your character), which puts it in the middle.
Please note that I'm talking about authorship, not about authority! You can have authority about your character without having authorship - that is, although you are allowed to have the final say about what is the case, you can leave it up to others to actually think up and narrate what happens. Authority as such does not aid immersion.
Actor / author /director stance
The division between actor / author /director stance (and whatever other stances you want to identify) has absolutely nothing to do with aiding or hindring immersion. It is not the case that actor stance is in the interest of immersion, or that director stance is detrimental to it. If my character is in a bar and a fight breaks out, I may yell "I grab a chair and hit the sucker on the head!", when no chair has been mentioned before. This is director stance. What goes through my head as a player is this: I realise that chairs make good weapons, I realise that I probably should imagine that I see a chair, and I start imagining that I hit the sucker on the head with the chair. The player and the character are in close correspondence. Yelling what I did makes it fictional that my character looks around, sees a chair, realises that chairs are good tools for hitting people, grabs the chair and hits the sucker on the head. What I, as a player do, and what my character does are closely related. This helps immersion.
That bears repetition and emphasis: immersion is easier when the actions of the player and those of the character are closely related.
Using this principle, one can make a case that playing this scene using actor stance actually mildly hurts immersion. "Do I see a chair?", the player asks the GM. This corresponds to the character looking around, so there is still a close relation between what the player does and what the character does. However, the split second look of the character is more like the split second realisation of the player using director stance, than it is like the several second exchange between player and GM in the case of the player using actor stance.
In general, however, I don't think there is any clear relation between stances and immersion.
(I don't think that actively pursuing a narrativist CA hurts immersion either, but that's something else.)
An analysis of Polaris
Polaris is not, in my experience, a very immersive game. How is that possible? It is, after all, a game in which one has not only authority, but even authorship over the thoughts, feelings and decisions of one's own character. That should help immersion, shouldn't it?
Yes, and I think you can immerse best when you are in the stage of 'free play'. However, once conflicts get going, the immersion is mostly lost. Why? The principle I identified in the previous section will help us out. In a conflict, I am negotiating possible events with my Mistaken. As a player, I am considering alternatives, weighing conditions against each other, trying to get the Mistaken to accept what I really want to happen, and so on. I, as a player, am doing all these things; but it is not fictional that my character is doing anything like this at all. My character is making all kinds of important decisions, having strong feelings, and so forth, while I am doing something else entirely. It would be hard for anyone to imagine that I am fighting a demon lord when what I'm actually doing is negotiating possible outcomes of the fight that do not even cross my character's mind. The very mechanics themselves seperate me from my character (in the fictional world), casting me as a benevolent hand of Fate instead of as my character.
If this is right, then immersion should be bigger in Polaris when I'm not using the conflict system. And actually, it is: I feel most immersed when I'm playing my character as a Moon. Because when I'm a Moon, most actions I, the player, take that have to do with my character correspond neatly with things my character actually thinks and does.
Obviously, the above is not a criticism of Polaris, lest anyone think it is. It is a wicked cool system, and immersion is not the 'way to play'. But I hope it has made for an interesting illustration. O, and while we're at it...
Immersion might be helped by not having too much rules stuff in play, but I actually don't think this is very important. It is easy to drift in and out of immersion. You can immerse one moment, do some complicated rules stuff the next, talk about stakes the next, and immerse yourself back in. Please take care to notice that in my Polaris analysis, one fails to immerse not because one has to do things that don't correspond with character actions; no, one fails to immerse because the character makes his/her important decisions and has his/her strongest feelings while the player is involved in something else entirely. If one's important decisions all took place in the free play stage, immersion would not be hurt at all. (Of course, one should also stop playing Polaris in that case; it wouldn't be helping you at all.)
I hope this has been interesting. I thought I was going to write a very short entry, but I was wrong.