I "finished" Super Mario 3D Land a few days ago, making my way through all 8 worlds and defeating Bowser at the end. Each world had 5-6 levels in it, ending with a boss fight. But in world 8, after defeating Bowser, I was taken to a special "extra" world, where I was able to play two more levels and then defeat him again.
Secret levels! Awesome. The game had shown me from the very start that there were 8 worlds, so getting a few extra levels in a 9th world felt like a bonus.
So then, after defeating Bowes a second time, I won the game properly. Credits rolled.
And then... 8 MORE worlds popped up! I'm only half way through the game's content! But this time, instead of feeling excited, I felt mostly cheated. I thought I was going to be done with the game - able to pass it off to my kids. I thought that the game was a bit easy, and short, but I was willing to accept that and just feel as though those impressions were a result of my extraordinary skill.
One sort of secret levels got me excited, and another caused me disappointment. What was the difference?
The most significant difference, obviously, was that minor adjustments to my expectations help to keep experiences fresh & exciting, while huge revisions to my expectations make me feel as if I have been lied to.
It can be really hard to anticipate what a player's expectations will be - is it "expected" that Dr. Wily will have extra levels within his castle? Is it "expected" that Mario will have the ability to warp between levels? This is an important role that maps serve - they create expectations of how much time / distance separates us from our goals. Adding things on the map can be exciting, but changing the map itself is probably a no-no.
I'm struggling with this same issue in my professional life - how much content should I give away for free, and how much should I give to my paying customers? In this case, I'm actually trying to create the opposite feeling - I want free players to feel as if they have experienced most of the game - but then surprise them with the revelation that there is a lot more for them to learn and experience.
Discussions of hidden content often lead to a discussion of books - because it's hard to have "extra content" in a book. You can see how many pages a book has when you pick it up, and you can see the typeface & how many pictures it has at a glance. One book has successfully pulled off this trick however - the fantastic "Godel, Escher, Bach" abruptly ends about 60% of the way through, and then starts a new book completely. You "finish" the main discussion, with the characters and goals you are introduced to, long before you run out of pages. The book even hints at this experience during the discussions of incompleteness, twining, and complexity. But that's probably just the exception to prove the rule.