Basics. In single player the connection between the environment and the story is stronger. There are more opportunities to place events, mechanics and other gameplay features that make the player feel immersed in another world. The better the level is, the better that immersion factor becomes.
AI placement and behaviour. The first thing that Hourences comments is that enemies should not be statues. They should have some function within the game's world. They have to have dialogues, walk around, patrol, do something other than just stand still doing nothing. I'd add my comment here that this is heavily a programmer's task. If you look at doom and duke 3D for example, all enemies were placed standing still. They had no behaviours programmed in at all, other than attack the player. Taking me as an example. Without knowledge in scripting or coding I can know what actions I want an enemy to perform, but I don't have the required knowledge to do it. The most basic example that I could think that relates to what Hourences says about making the enemies be part of a living game world is Super Mario or Sonic. All enemies move in a way or another. They are never statues waiting for the player, unless it's a trap. As a side note: I'm not trying to say that you have to be a programmer to make the enemies have "jobs" in your game or level. Games may have forms to do that without requiring previous programming skills, as in the case of Unreal's actors which provide some basic form of dictating what an enemy should do without requiring you to write explicit code.
Next it's enemy's placement. Repeating the same pattern over and over is a bad idea. For example: if every time the player opens up a door there is an enemy behind it, it becomes predictable and breaks the immersion. If the game is about terror, repeating the same moments over and over defeats the purpose of scaring the player. The other extreme is to never repeat any pattern. Randomness also breaks the immersion because our world is not completely random. There is also a risk of not creating an identity by having too much randomness and no recognizable patterns. Hourences discusses that a player should feel empowered by having patterns which he or she is able to recognize and then take advantage of. I'd add to that by resorting to a lesson from Mark Rosewater. The player wants to have control and power and this creates a deeper attachment between himself or herself and the game or level.
Repetition can also happen in the form of enemies which are all the same. A good example is F.E.A.R. and Alan Wake. The same enemies repeat over and over everywhere, which becomes boring over time. Now for the cause of this I can only speculate: time or money constraints being the primary suspects. Good examples of enemies that were designed to fit in the environment are Sonic and Bioshock. The enemies really feel as if they were part of that world. Unfortunately I don't have knowledge in arts and because of that I can't comment on stylistic matters. The only thing that I can say is that in the realm of storytelling we have a lot of freedom and sometimes we are allowed to mix completely different styles if there is a good reason to do so. The opposite problem, too much variation in enemies, is also bad. The balance between repetition and uniqueness in terms of enemy types is not a simple task.
For the matter of how hard or easy enemies should be, I really don't know much to tell. What Hourences discusses is that enemies and bosses should be reasonable and feasible. There should be clear ways to win and the game should clearly show how to win, providing enough clues and tools to accomplish victory. The key point that he tries to make is that when things are too hard this leads to player's frustration. In turn, players quitting the game. Not something that you really want to happen as a designer, developer or even a business man. One example that comes to my mind is that many games employ enemies with health bars. Remove the health bar as many other games do and the player can never know whether they are winning or just wasting their time and energy. If health bars are not an option, then use animations, changes in clothes, changes in the environment or even dialogues to tell the player that they are making progress.
Scripted events and gameplay variation. In simple terms, are events that happen in the game which helps to bring variation to the gameplay. From the point of view of who makes the game or level they are expected to happen. If they don't we have bugs to solve. From the point of view of players, it's an event that breaks the repetition of performing the same actions over and over. That's the key point that Hourences makes here, variability and breaking repetition. The classical example are cutscenes, where the games take control over the character and sets in motion a series of recorded actions that always happen in the same way no matter how many times you replay it. Here is a list of examples:
Doom didn't had a scripting language to allow for total control over what happens and when, but it did have events such as fighting a boss and defeating the boss lowers a wall or opens up a door. Or picking up an item, which in turn, triggers enemies teleporting from elsewhere to fight the player;
In F.E.A.R. there are sudden appearances of Alma, troops dropping off an helicopter, explosions that force the player to go in another direction, hallucinations that make the player see events from the past, etc;
In Bioshock infinite there are many many times when Elizabeth intervenes in the gameplay or the story itself.
Doom 2016 have the glory kills and they could be considered a form of scripted event too, because they add another layer of interaction and a bit of variation to an, otherwise, already saturated game's genre.
In here Hourences talks about the immersion factor once again. There is a huge difference in telling the player that something happened and making the player see the event happening. Even better if they can experience it themselves. The key point is that by making the world "alive" we make the player feel the game world as if it was larger than it really is. Not everything has to be about the player and put the player at the center of attention. It's important to note that. A bad example is Prey 2006. They placed monitors which would play the news and the news reporter would tell what is happening on Earth, outside the alien structure that the player is in. A better example is Max Payne and its graphic novel. The player can at least hear, read and see the events. Anything that happens within the level's space, without a loading screen and that the player can see themselves help in creating a sense of a world in motion and the player takes part in. An even further step is Uncharted. In this game there are some chase sequences which are sort of interactive cinematic experiences where the player does not lose control over the characters movements.
Yet another function of scripted events is to showcase things or to preview the next part or to foreshadow something. For example: a character showcases how a weapon works and it's even better if the player is able to test it himself or herself. In Max Payne there is a scene where the camera cuts to the next area, previewing laser trip mines. Many many games do that. One of the most used is to have the camera focus for one second on a boss's weak spot or some interactive piece of the level that must be activated to progress. Another example is when the camera focus on a window and there is nothing there, but when the player reaches a certain position an enemy breaks in from that window.
I should add that this area of scripting and level design is an hybrid between design, arts and programming. This is certainly not an easy accomplishment to be skilled in all three at the same time.
Floorplan. For single player there is more freedom because there is the storyline, scripted events and there is a stronger atmosphere with more room for development. Hourences highlights some basic techniques:
Previewing. In short, it means to preview something from a certain distance before the player can interact with it or reach that place. Many many games do that. I have shown multiple examples of this in this site.
Evolution. His idea is that the same area can be reused. When the player revisits a previous area they have something new to see or experience. A really bad example of this is Duke Nukem Forever. There is a level that the player traverses with a car while in shrunk state. After the player is enlarged back to normal size they go back through the same areas but there is nothing new, the only difference is that the player at normal size traverses it much faster than before in shrunk state. Max Payne did that in the hotel part. First Max Payne gets out of the hotel. Later on he goes back and some new areas and enemies are revealed. From a developer's or designer's standpoint I can see where the "evolution" comes from. It's about evolving states of the same level or area. But from a player's standpoint I'd call this "shifting", "living" or "reactive".
Recycle. This is a simplified form of the previous. It's literally reusing the same area or level. Jedi Outcast did that in the first level of the game. There is an area which is both a landmark and a reused area that the player comes back multiple times once he or she is done with one of the side areas. The player grabs a key or code and returns to a previous area. This approach is meant to save time and resources, while at the same time adding a bit of depth by not discarding areas once the first mission, event, has taken place or an objective is completed. I don't have much knowledge to comment on this, except that it very much depends on how you approach your storyline. Among other factors such as available resources.
Landmarks. I have already explained it in my site. It pretty much matches what Hourences says in the book.
Interactivity. This is about making the world really alive. The difference between Doom and Duke 3D is that the latter added interactivity. There are stripers, vending machines, fire hydrants, breaking glass. Little details that make the world in Duke 3D much more alive than in Doom. Max Payne did the same thing and the technological advancements allowed for different footstep sounds and bullet markings depending on the surface. Unreal did that with native Nalis that would interact with the player and reveal secrets. Bioshock did that with water puddles that the player could take advantage of by freezing or electrocuting enemies. I'd argue that this requires more from the designers side than from the programmers, because hardware limitations are more of an excuse than a a limitation that cannot be overcame.
Traps. The difference in SP over MP is that traps are meant to slow down the player if they are in their way or provide an optional challenge if it's the case. What Hourences advices is that traps should not be placed in a way that kills the player without warnings. The player should be able to escape or avoid it. Traps that are unavoidable and/or inescapable are bad.
Items. It's the same concept from multiplayer: risk vs reward. I'd take what Hourences said and stretch it a bit. What should be avoided at all costs is to make the game feel as if it was a ponzi scheme or create unnecessary tension by punishing the player right after they get an item with low value. That would feel as if the player is being tortured or conditioned to feel bad. Now I can't stretched it further than this because delving into whatever reasons the developers or designers had to do this is outside my reach.
In SP the items are good breadcrumbs for the player to follow.
Difficulty. I can't say much about this. From a personal experience I could comment about Bunny Track. It's a gametype for Unreal Tournament where all maps are racing tracks with many obstacles to be traversed. I'd say that most maps are incredible hard and they are more frustrating than anything else and this is specially caused by high latency. Bioshock for instance is incredible hard because there isn't enough money to buy a lot of ammo and weapon upgrades or power upgrades. At least in the beginning of the game, later on it becomes easier. Bioshock infinite on the other hand is way easier than the previous games. This is a very tricky point of games with multiple challenges to tackle. Difficulty depends on how good the AI is, on the level providing strategic choices, on item placement, on core gameplay. Each part contributes to make the game easier or harder.
The key point that Hourences tries to make is that, ultimately, the game is about fun and not frustration. You want to please the player, not anger him or her. Which resonates quite well with what Mark Rosewater says about Magic the Gathering.
Pace. I don't have experience with this, admittedly. It's about progression, climax and anti-climax. The difficulty should make sense by being easier in the beginning and harder near the end. This has a lot to do with the storyline or plot itself, on how the story progresses and is developed over time. I'd say that we have two extremes here: one is to be impatient and rush things, not giving the player enough time to experience it; the other end is similar to procrastination in psychological terms. You delay, delay and delay. In which case the player himself or herself is going to become impatient. I'd guess that professional storytellers or script writers have a better understanding on such matter. It's their job to keep the audience interested over shorter or longer periods of time. However, let's also bear in mind that script writing for TV and cinema is not the same as for games because of one crucial difference: passively watching something vs actively interacting with it.
Hourence's conclusion is pretty much that each of the previous topics should be carefully intertwined to produce the best outcomes. I agree with him. To split the job into multiple departments and people is not bad in itself, but as teams grow larger it becomes incredible difficult to manage all the pieces and keep everything cohesive.