The more we talk about this, the more I feel the term "narrative space" is misleading, and the more I lean into the "three pillars" model - Combat, Exploration, Social. What we call narrative abilities are abilities that fall under the last two pillars.
No, Starfox, no!
I like your "narrative space" formulation much better than a "3 pillars" formulation - even though, as I posted upthread, I want to break narrative space into a "plot determined by action resolution mechanics" component and a "scene framing determined by scene framing guidelines/mechanics" component.
The problem with focusing it on the "three pillars" is that you don't have the tools to properly analyse the following two conrasting options: the ranger has mechanics for tracking, navigating, etc - ie
engaging an exploration challenge - whereas the wizard has mechanics for scrying in on the missing whatsits and then teleporting to where they are - ie
bypassing an exploration challenge and reframing it as a "pick up the whatsits" scene.
Or in the social sphere: there is a big difference between a Burning Wheel Duel of Wits, or a 4e skill challenge run per the DMG guidlines, which require the players to
engage the situation by declaring action, making rolls and having the GM adjudicate the unflding situation; compared to a Charm Spell or the most literal reading of a 3E Diplomacy roll, which doesn't engage the challenge but simply reframes it as one containing a friend rather than a potentially hostile independent personality.
Suppose that the player of a fighter had a "scene reframing" power of some sort such that, in any non-combat situation, s/he could change it into a combat one (eg it might be a taunt power, that turns non-hostile NPCs hostile; or a "challenge the spirits" power, so that in an exploration situation the nature spirits can be forced to duel with the PC, and if the PC wins they will lead the party where the PCs want to go). In this case, it wouldn't matter that the fighter's action resolution abilities were confined to combat, because the player would never have to engage any other sort of situation.
A further element that I didn't mention upthread but that [MENTION=27570]sheadunne[/MENTION] and [MENTION=6696971]Manbearcat[/MENTION] have both flagged is "content introduction": ie who has the authority to introduce new content/backstory into the game. Giving players of martial PCs this sort of power (eg the player of a rogue can simply "declare" old friends and contacts in a city which is, from the point of view of actual play events, a new city - d20 Conan permits this sort of thing with Fate Points) can be one way of trying to balance spellcasters scene reframing power.
If you turn the discussion into a 3 pillars discussion you lose what I think are these key analytical distinctions for understanding how the game unfolds.
I still think that the Background Traits need to be where we look to for inspiration for how 5e might assemble director stance and author stance narrative authority.
I agree with this. They're the most interesting part to date of D&Dnext design, in my view.