So, my first real blog post, but I'll restrict the self-indulgence to this sentence.  The point:  relatively recently I committed myself to making a game, having never done anything like that before but having an idea for a game that I wanted to explore and also to give myself the challenge of completing an ambitious project by myself, which I have often struggled with in the past.  (It will almost certainly be a Source Mod, for the record.)  I'm very much still in the thinking stage rather than building anything in a level editor anytime soon, but I've run across an issue which is fairly hard to reconcile and yet fairly central to what the game will be like.  I'm genuinely interested in what suggestions people have to get around the issue.
 
The actual point/issue this time (in condensed form):  There has to be fail states in games where right decisions need to be made in order to progress, but being able to retry after having failed strains the sense of verisimilitude the player might feel with the game and is compounded by any subsequent repetition of game elements that the player then experiences.   That sentence was considered carefully but it may still have flaws so feel free to point out any amendments to it if you are able to argue them clearly.   When saying ‘verisimilitude’ I’m meaning a gaming experience that shares enough similarities to reality such that the player instinctively starts to interact with the world as they would in real life and has expectations of the world that they have of the real one (I think the same probably holds true even if the player is consciously role-playing).   That above phrase doesn’t apply to the majority of games since many don’t try to achieve that (fair enough) and many revel in their ‘gameyness’ – personally I’m particularly interested in games that try to create that verisimilitude since there are things that can only be explored and encountered when that is the case.   You may think this is weird because as gamers we’re used to dying and then just trying again so don’t see what the big deal is, but I’m fairly convinced this is because we’re often not playing with a real sense of verisimilitude with the game world and this is precisely because that’s so easy to lose and so hard to maintain.

For the game I’m thinking of creating, this issue is made more complex because at the beginning of the process, when I was deriving the core elements of the game that should be upheld during the creative process, I decided that it was key that the player felt like they were thinking for themselves and applying their own intelligence to the problems they’ll face in the game.   This is important and awkward because the problem created by fail states (dying) could simply be reduced by making the game have no real challenge or by doing a lot of hand-holding and guiding of the players - exactly what I don’t want to do.

So what suggestions do people have for a good solution?   The obvious way would be to try to cleverly reduce the number of places in which a player could die, but maybe someone has ideas that work completely differently.   Here are a few of my ideas so far:   if a player fails a challenge early on, the like of which could lead to failure/death later in the game, they do not lose if they fail but are instead taught (in some way that makes sense within the world) what they should have done differently.   Hopefully this will start to get the player thinking successfully for themselves to the point where less hand-holding needs to be done.   To generally reduce the number of places in the game where the player could fail, but to have negative lasting consequences for other failures that do not lead to a complete game over.   To include some randomness in the events that occur (at least in their order) so that if a player fails because of one event, they are not confronted with the exact same one again immediately which keeps the game fresh and I think gives a greater chance of keeping the verisimilitude of the gaming experience.   If the player dies/ fails try, as much as is possible, to have the last save point at an occasion and location which makes sense within the fiction of the story and generally feels natural to the player.   To have an in-game help system (again, making sense within the world) that offers help if it looks like the player is struggling to complete the necessary tasks, thereby reducing the chance of failure – need to do this carefully to make sure the player feels like they solve any challenges themselves.

Wow, that's some excellent written expression and great first blog post. I particularly like the italicised thesis statement. I think what you might be after is what someone else (whose name escapes me) has theorised - that games need 'permanence', ie. if you fail something you can't retry it and instead have to live with the consequences. There are a few games I can think of that use such design choices, namely Farcry 2 and Heavy Rain.
This would probably be the best way to maintain world consistency. Death, even if you have random resets and stuff (like say a rogue-like) almost always breaks immersion. Or reloading trying to complete some objective that you've failed. Having to live with a decision is something you rarely have to do in games and can increase the moral engagement of the player. What if you had the chance to save Aeris, but failed, and then had to live with the consequences for the rest of FFVII? It could make for an interesting game.
I'm not a designer but those are some ideas I've come across whilst perusing the internet. Good luck with the project.

Wait, wut? No offense, but this reads more like a thesis than a blog post or a design doc. 
 
Also:   

This is important and awkward because the problem created by fail states (dying) could simply be reduced by making the game have no real challenge or by doing a lot of hand-holding and guiding of the players - exactly what I don’t want to do.

 if a player fails a challenge early on, the like of which could lead to failure/death later in the game, they do not lose if they fail but are instead taught (in some way that makes sense within the world) what they should have done differently.    

Without any kind of idea what type of game you're making, or the context, it is impossible to offer any meaningful suggestions.
 
Suspension of disbelief is hard, very hard - particularly upon player death.  It is of course significantly easier if you are dealing with fantastical elements, or with a game that doesn't rely on the tension created by the risk of death to the player.  For example, the Prince of Persia doesn't fall to his death, Elika just catches him and throws him right back to where he was.  I want to say that Prey had a mechanic for this too, where you entered a death state and had to fight your way back to life or something (I didn't play it myself).  GTA dumps you back at a hospital or cop shop etc.  
 
These examples are fudges at best, because whilst they may be trying to gloss over the fact that the player has died, they're no more believable than just dying and reloading, so it doesn't really achieve what you're looking for.  The popular Lucas Arts adventure games are famous for not letting you die - though of course, there is still plenty of challenge there.  Maybe this is closer to what you're looking for.  A game where failure does not equal death, but it does stop you progressing... which you don't seem to want either.
 
There was a game in development once, called Sex n' Drugs n' Rock n' Roll, where you could supposedly sleep with a hooker in the first half of the game, and if you didn't use a rubber, you would subsequently die of AIDs in the second half of the game.  Amazingly, this game never came out.
 
You could always go for the 'endless trapdoor' method, where failing a 'challenge' doesn't result in death, but merely drops the player through the floor to be presented with a different challenge - or even back to the start of the current one.  Maybe that failure will prove to have consequences later on, and maybe it won't.

" Wow, that's some excellent written expression and great first blog post. I particularly like the italicised thesis statement. I think what you might be after is what someone else (whose name escapes me) has theorised - that games need 'permanence', ie. if you fail something you can't retry it and instead have to live with the consequences. There are a few games I can think of that use such design choices, namely Farcry 2 and Heavy Rain. This would probably be the best way to maintain world consistency. Death, even if you have random resets and stuff (like say a rogue-like) almost always breaks immersion. Or reloading trying to complete some objective that you've failed. Having to live with a decision is something you rarely have to do in games and can increase the moral engagement of the player. What if you had the chance to save Aeris, but failed, and then had to live with the consequences for the rest of FFVII? It could make for an interesting game.I'm not a designer but those are some ideas I've come across whilst perusing the internet. Good luck with the project. "

If someone could pull off what you describe in the last point, I'll die happy as a gamer. The problem is just simply how much work goes in to creating games, and if you truly had a fork in the game's progression where someone major like a character death happens, you are essentially doubling the amount of work needed to finish the rest of the game. Especially if the character plays a meaningful role in the game, you'll have to create multiple sets of dialogue, since characters would obviously respond differently if the major character dies, the plot would have to be completely reworked to be coherent with and without the dead character.

This is the kind of stuff that drove me crazy in ME2. The whole "moral choices from ME1 effecting the world" thing felt like such a gimmick, but at the same time it's completely understandable.

Also, completely agreed on the save-reloading thing. One of the few things I think Fable 2 got right was how it one had 1 save per character and immediately saved after almost any decision you made. I love that concept, but I don't have the self-control to impose it on myself when playing something like ME2, which let's you keep tons of saves.

" @Gaff:  You're right, fail-state isn't the right term for it. Nevertheless permanency might have some relevance to what he's doing, if he wants to get rid of failstates in some situations or altogether. "

It sounds like you're going for a sense of reality, so a good solution would be to do away with death altogether. There can be other ways to fail/make failure apparent, and, as you've already said, failure doesn't have to end the game.

Using big words will make it more good.
 
/s

" It sounds like you're going for a sense of reality, so a good solution would be to do away with death altogether. There can be other ways to fail/make failure apparent, and, as you've already said, failure doesn't have to end the game. "

Kinda funny to see that this blog by dankempster has directly influenced your post, and you've got more responses.
 
I think Video Game King had a good point in that you could write death out of the game for good. Hell, do away with game over screens. Just have serious ramifications for failure that last the entire game. Your only problem there is that those consequences probably can't affect the storyline much. 
 
You appear to be attempting to revolutionise a certain aspect of video game design, and for that I wish you all the best. We need innovation in this industry.

" Kinda funny to see that this blog by dankempster has directly influenced your post, and you've got more responses. I think Video Game King had a good point in that you could write death out of the game for good. Hell, do away with game over screens. Just have serious ramifications for failure that last the entire game. Your only problem there is that those consequences probably can't affect the storyline much.  You appear to be attempting to revolutionise a certain aspect of video game design, and for that I wish you all the best. We need innovation in this industry. "

" Using big words will make it more good.  /s "

Well you could employ an omniscient presence and environment within the game (providing it would make sense to do so within the context of the existing game environment and plot), so that upon death, the player is transported to this 'out of game' environment, and is reprimanded by an omniscient being there for failing/dying. Said NPC could then 'revive' the player to reattempt the challenge. Interestingly, you could in fact keep their corpse, and for failed puzzles you could keep their existing state at the player's death (although this would take some designing so that puzzles remain doable even mid attempt), you could in fact incorporate the player's own corpse into the solution of the game (I'm speaking from a source perspective as source ragdolls have weight).

A lot of what I know about game design I picked up through tabletop games.  What I suggest (and what I practice in my own case) is a real dedication to pass/fail branching.  The core concept of gameplay is that input creates feedback.  Every action necessitates a reaction.  The player's action needs to result in an outcome.  To make this rewarding in a game, you need to influence the player to desire certain outcomes, rather than presenting him with correct/incorrect choices.  Creating pass/fail scenarios that result in branching is key.  The branching has to be relevant to the game world... not being able to save a person needs to have an affect later when you need them, while still allowing the story and gameplay to continue.
 
Most often this is a story consideration.  Catharsis is the easiest way to influence a player, and if you can make him or her care about the game world, the characters, or their own avatar, they will play the game differently.  We've seen attempts at this in karma meters and story decisions... the problem there is that they are 'decisions' and not a result of gameplay.  Success or failure doesn't affect Shepard, because he inevitably succeeds and then chooses whatever he feels.
 
I don't feel as if you can demand the player play the game in a certain way, so the idea of a player reloading and doing it until it's done the way they want must be accepted.  But the idea of a branching narrative is that the player is experiencing a singular experience.  If a fail state presents the player with a different outcome, make sure the outcome is rewarding and defines the experience as much as the success. 
 
"Nothing happens" should be the one thing you avoid.  Unless you're just saying that to the player, and changing things unseen.
 
edit: Reading some replies... I do feel as players require some guiding.  The most important thing a game can do is to present its rules and exceptions clearly.  You don't need big prompts necessarily, but you do need to allow them to discover how to interact with the world.  If there's no control screen or tutorial, the first thing a player does is stand in a safe spot and try all the buttons, looking for feedback.  And when things get convoluted, as a person who has ran tabletop games, it does help to simply present the players with a couple of options as a start, and see if they can figure out the hidden third option.  Or, and this can really only happen in a game where the designer is present, they present you with an option you didn't see.

It depends on the game. For example, if the goal of the game is simply to reach the end, it's practically impossible.
 
If the goal is to get as good an end result as possible, however, it's quite simple. You just create a number of endings, and assign each ending to a number of successes. For example, if there are ten challenges, succeeding at all ten would give you the best ending, while failing all of them would give you the worst ending. If you then want the game to have more depth, you could have the scenario branch depending on whether the player succeeds or fails at certain challenges, and/or have the ending vary not only depending on the number of successes and failures, but what specific challenges the player succeeded and failed at.

Interesting blog. Here are my thoughts:
 
1. Remove or reduce combat (Or make it easier). I don't think you wished to do this, but you could still have a combat system and somehow remove the possibility of death. If the player fails, then you could permanantly injure them; this disability could affect their efficiancy at performing certain actions in the game world, or even remove some other possibilites. But I think the concept of disability works.
 
2. Have the player turn into some sort of ghost. Demon's Souls was a game that did this well. When players die they turn into a spirit and must work their way up into another physical body. Gameplay still remains mostly similar throughout, though in soul form you have less HP.
 
3. Have a system of permanance, where if you die that's it; the story ends and you can't start over.
 
4. Embrace the fact that games are artificial and fictional, so it doesn't matter if you die because everything is essentially infinite. Or you could be some kind of god that could die, but because you're a god you can self-replicate or transform into other beings (Reincarnation perhaps?).
 
Hope that helps.

Verisimilitude = The quality of realism in a movie, book, game etc. 
(For those who had to google it like me)  
I'm surprised I've never heard the term before (I read a lot.)

Also, in Prince of Persia, the prince had a power to turn back time if he died. I thought this was clever.