One way to implicate the players is to offer them horrific, impossible choices. (Let me go back a bit to explain this.)
Conventionally, in horror games, we present players with horror. We show them something horrific, usually a monster, and expect them to respond.
Instead, try giving choices. For example, do you kill the person who's trapped inside the monster, or leave them to be slowly digested? Do you rescue your comrade, risking your life, or escape yourself? Do you give your life to save the town? Or, to take an example from Jason Morningstar's The Black Drop: one of you must be sacrificed. Who will it be?
That way, the players are implicated. Damn right they're implicated: they chose the horror.
You need to make sure it's horrific either way. It's not hard. Usually, it involves a choice between human horror ("If I don't kill/torture/blind someone...") and supernatural horror ("...then this monster grows stronger/wreaks havoc/does horrible stuff to people I care about"). And you need to make sure it's a real choice, too, and you're not railroading them into one option.
I like to put choices like this at the end of scenarios. Often, players/characters will fight over them, and that's a great thing. One character wants to perform a sacrifice, so another turns a gun on them. Often, you'll end up with a bloodbath, and all you have to do, as a GM, is sit back and watch.
There are other techniques too. Most obviously you can make horror scenarios actually creepy. It sounds obvious, but so many horror scenarios aren't: they recycle tentacles and zombies, which we've seen many times before. Put a zombie child in there. Put a river of running bile. Put something that you find unpleasant and, often, the players will find it unpleasant too.
Another technique comes from improv: be obvious, without self-censoring. This is particularly effective in horror. For example, I recently wrote a scenario about Daoloth, a god who helps you "see the world clearly". Daoloth sent one of my NPCs mad. I decided that, to see the world clearly, my NPC had carefully removed his eyes with a dinner knife (they were getting in the way). This seems obvious, to me, but it's clearly horrific.
Honestly, though, offering choices is the best technique I know. And it's the one that actually implicates players in the horror.
Building Tension to Build Horror
I’ve played RPG’s since I was a kid, but I’ve also written narrative as a hobby for years. RPG’s combine the intellectual stimulation of board games with the deep engagement of storytelling (books and films). When the players calmly intellectualize their attempts to WIN, they’re in board game mode. When players FEEL something, they’re in engagement mode. As you’ve already guessed, horror can only be experienced by engaged players. Players experience horror as a form of tension. Tension engages them laying the ground work for horror.
The concern your question addresses is The Law of Diminishing Returns. For example, players may find zombies scary at first (a form of tension), but if they keep running into them, the zombies soon become familiar, then boring, and eventually comedic (a total lack of tension). Consider the zombie meme. It began with the fearful “28 Days Later” but was quickly played out by copy-cat films and computer games eventually producing “Shaun of the Dead” - a 180 degree shift from horror to comedy. Further, almost everything is subject to the law of diminishing returns. As the other answers suggest, a campaign seeking to remain tense by existing within the horror space for extended time, must tap the needs of tension to counteract the law of diminishing returns. Specifically, it must push high-stakes dilemmas on the players to build and sustain tension.
A dilemma is a difficult choice for the players. For clarity, a dilemma is more than a decision, which players make all the time. “Do I fire an arrow or cast a spell?” In contrast, dilemmas are the tough ones, coming in only two forms; lesser of evils OR irreconcilable goods. An irreconcilable good is when the player/s desperately WANT all of the choices, but they can only have one. This works great for treasure rewards in gaming by the way. Lesser of evils is when the player/s passionately HATE all of the choices, but MUST choose one. Needless to say, a horror campaign benefits greatly from lesser of evil dilemmas. With this focus on dilemmas, consider the following.
Pacing
Pacing is not the scale of an event, it’s inherent drama, or the speed of travel. Rather it is the frequency of dilemmas over time. Many gaming sessions feature a battle-royale taking hours. Counter-intuitively, this can present as a slow pace. Why? Because of the lack of dilemmas while the battle transpires. Effectively the players make one choice, to fight, then play out that choice for hours. In contrast, a fast pace might look like this:
GM: A hoard of zombies surround you, closing in, but you notice a stair descending into blackness. Vapors seep from the opening. What do you do?
PLAYERS: We don’t like the look of those stairs, but there’s too many zombies. We’ll have to chance the stairs.
GM: You rush down the stairs, barely able to see. The sound of pattering feet signals the zombies pursue close behind. The stairs stop before a pit, water rushes by at its bottom, barely visible in the blackness. If you drop your backpacks you could leap across. What do you do?
PLAYERS: No way we’re dropping our stuff and we still can’t fight the zombies, but who knows what’s in that water. Damn, we’ll have to swim for it.
GM: You plunge into icy blackness, the water sweeping you into a large cavern. You barely make out zombies crowding either shore. As the current races you along, something slithers past your paddling feet. What do you do?
This quick succession of lesser of evils dilemmas constitutes a fast pace. Pacing, fast or slow, is not horror itself, but a fast pace creates tension priming the players to feel horror conveyed in other ways.
Mood
Many of the other answers detailed suggestions for establishing mood so I will not elaborate save to observe that mood can target the players (for instance, dimming the lights in the room) or the characters (descriptions of gore in the game setting). For maximum effectiveness, use both.
Rhythm
Alluded to in your question, rhythm is variation in the pacing and mood. Since both pacing and mood are subject to the law of diminishing returns, a GM must vary both to avoid draining the effectiveness of either. I’m sure you’ve heard that movie description ‘roller-coaster ride of excitement.” This references an effective variation of rhythm. Just remember, when varying pacing and mood, avoid repeating patterns. If the players detect the pattern, they’ll predict the surprises because the pattern itself is subject to the law of diminishing returns.
Remove Safe Ground
Like mood, this tension-building technique can target both the players or their characters. Don’t provide the CHARACTERS convenient ‘home bases’ (such as a room with a door they can lock) to sleep or heal in, where they can control their safety to turn off threats. Even more important, deny the PLAYERS safe ground. Avoid quantifying opponents and environment when feasible as this allows players to intellectualize their odds of success, taking emotion out of the equation. Also, it is common for GM’s to allow players the ability to stop game time while they tackle dilemmas. This kills the tension horror demands. I often employ the prop of a cooking timer to drive home the point. “Do you fight the zombies or run down the dark stair? You have 10 seconds to decide. tick, tick, tick, tick...
Deny Closure
Easily the most common way game masters defeat their own efforts to raise tension is by making monsters fight to the death; for once monsters are destroyed, relief replaces tension allowing players to relax. These easy victories also sabotage the players’ sense of triumph. Consider a pack of undead wolves pursuing the players, employing hit or run tactics. Met with strong resistance, the wolves melt into the forest, but the threat remains. “What happens if the wolves attack while we’re fighting the zombies?” If the tension starts to wear off, the wolves attack again, or maybe it’s enough to hear rustling in the bushes.
My point in all of this, is that horror requires the players to FEEL tension as opposed to dryly KNOWING that a monster is there. When the players know what’s going on, they feel in control, and mentally switch into board game mode. Pacing, the removal of save ground, and denying closure build tension. Variations in rhythm prevent the tension from wearing off allowing mood to transform the tension into horror.
Best Answer
This is a pretty large topic. I'll offer up my essay on Horror in Roleplaying as a partial answer. A quick summary of the approach: