(For the record, this article is going to use examples from Amnesia: The Dark Descent, Alien: Isolation, Outlast, and Near Midnight, so be wary of potential spoilers if you want to play these games).


After writing my last article on evoking horror in level design, I got a lot of positive feedback from people at my college, to the point where I was invited onto a team for what would become my senior game, The Last Light. I was really excited to apply what I had learned, designing scripted sequences to simultaneously scare the player and teach them about their foe, and squeezing a chase sequence onto a subway train. I felt like it truly was a culmination of my college career.

It wasn’t very successful at being scary, though.

I was brought on after the core narrative and gameplay had been determined: the player was navigating a subway system evading a cloud of flesh-eating microbes, which would act adversely to light sources. I thought this was a smart move for a college team: instead of having one of our artists need to model, rig, and animate a creature, we could just get by with a particle effect. Unfortunately, during our weekly class sessions, the antagonist was a source of consistent criticism. People outside the team said the enemy wasn’t intimidating at all, and pointed towards the art direction more often than not.

Given about a year’s distance, I’m starting to think it was more of a design issue. It certainly made the scares me and the other designers had planned less effective, but the problem felt like a tautology at points: we weren’t able to scare players using the monster because the monster wasn’t scary. Keeping this question in mind, I started looking at successful horror games and how they establish their antagonists, paying attention to things they did that we weren’t able to in The Last Light.

I think I’ve got an idea now of what the issue was. But first, we need to cover some ground:


For a project in college, I did some research into worldspace UI, which made me stumble onto this fantastic thesis written by Erik Fagerholt and Magnus Lorentzson, with assistance from DICE. Surprisingly, on page 67 they pivot from discussions on UI into defining immersion. Many of us have a gut-level reaction to hearing the word “immersion”, either thinking it’s a word co-opted by marketing as a euphemism for “good graphics”, or believing it’s some elusive quality that your game might be lucky to have.

For most people who have experienced immersion, though, they often describe the sensation as “believing you’re in another world”, or experiencing a loss of self in the game. Fagerholt and Lorentzson cite several sources claiming that this is a simplified and somewhat inaccurate description: one definition, which they take from Rules of Play by Katie Salen and Eric Zimmerman, argues that the sensory experience we engage in is inseparable from its framing:

“In a first-person shooter such as Halo, part of the experience is the sensual vertigo of navigating a coherent, imaginary 3D space. But playing the game also involves an awareness of the game interface, the strategic use of the frame-breaking options, the use of text-based chat, fluctuating server speeds, and the sharing of tips with friends in the larger social context of play.” (Salen and Zimmerman, Chap. 27, Pg. 33)

Basically, when we enter an immersive experience, we don’t lose our real-world knowledge, and actually require it to engage with the game. For instance, when a player plays Skyrim, they’re engaging in a fictional world where there is no concept of a computer mouse, a keyboard, or a game controller. However, when players enter an immersive state in Skyrim, they are still able to, and in fact need to, operate these input devices to continue engaging with the game.

Similarly, Salen and Zimmerman argue that immersion is engaging in something called “metacommunication”, communication about the communication itself, which we often associate with role playing. If you ask a player in World of Warcraft who they are, they might be inclined to say “I’m a Death Knight”, or something similar. The player on the other end does not literally believe that they are a Death Knight, but is actually making a more complex metacommunicative statement saying “In the context of this game I am assuming the role of a Death Knight”. This, paired with the use of real-world knowledge, suggests that immersion is the ability to extend oneself into the game, having one foot in the virtual and the other in the real.

Furthermore, Fagerholt and Lorentzson argue that there’s two ways of cultivating immersion: immersion through perception, which is the common “good graphics” assumption, and immersion through reasoning, which is described as being attained through logical consistency in the game. When the rules of gameplay are consistent enough, this allows the player to think in terms of the game’s fiction while performing actions, instead of reminding themselves of the rules, calling attention to the fact that they’re playing a game.

The important thing to note, though, is that Fagerholt and Lorentzson argue that immersion through reasoning does not mean that the logic of the game has to conform with real-world phenomena: again, to use Skyrim, players are able to cast spells, drink potions that have different effects, and interact with fictional species. It only matters that a fire spell consistently lights straw on fire, or that a health potion always restores a certain amount of health.

It’s worth noting that the parts of Rules of Play that I quoted earlier are from a chapter entitled “Games as the Play of Simulation”. At the beginning of the chapter, Salen and Zimmerman propose the definition of simulation as “a procedural representation of aspects of “reality”” (Chap. 27, Pg. 3). In other words, all games are systems trying to recreate real-world phenomena.

That bit about “aspects of “reality”” is important, as a simulation cannot be a 1:1 recreation of what it’s simulating. We (currently) are unable to provide sensory feedback for taste and smell, and are only somewhat able to do this with touch. However, this selection of certain traits is prioritization as well as inability: while Arma 3 might be a very in-depth military simulation, it doesn’t have systems that keep track of climate change, flower pollination, or thermodynamics, simply because they’re not required or important to the fantasy of armed combat the game is trying to pursue.

So, to recap and inelegantly tie these two concepts together:

  • A simulation is a recreation of parts of reality, and immersion is a metric that can be used to measure how effective the simulation is.
  • Immersiveness is fostered through removing obstacles preventing the player from role-playing as an agent in the game.
  • A good method to preventing obstacles to role-playing is to have logical consistency within the world, even if the phenomena in the game are not based in reality.

So, if we want to make a horror game with an effective monster the player can roleplay against, it should be logically consistent with the other aspects of our simulation. However, there’s still a bit of a snag: most antagonists in horror games are based off of things that we have no reference point of in reality (ghosts, aliens, demons, etc.). So, if we want to make a simulation of a monster, but we have no real-world object to simulate it from, how do make the player better able to roleplay with it?

Game Feel and Communicating Presence:

I have this feeling that Game Feel by Steve Swink is the most-cited game design book, or at least will be very shortly. In order to figure out what makes controls that “feel good”, Swink covered a lot of ground, researching perception, cognitive psychology, and proprioception in order to create some workable guidelines to cultivating good game feel. Because the book draws from so many areas, though, its implications are similarly far-reaching.

One basis of Swink’s argument is that game feel is based on engaging a player’s proprioception, a sense that not only allows a person to sense where all the parts of their body are in relation to another, but to extend that sense to a tool the person is using. One example Swink uses is a car: after a while, the driver is able to sense properties of the car, such as how tightly it turns and at what speed it handles most comfortably. The driver starts thinking of the car as an extension of themselves, and good-feeling controls in a game are similarly cultivated from this phenomenon.

But getting that sense with a car requires actually feeling the weight of it, which we can’t do with virtual objects. Fortunately, it seems, our senses try to intuit this information anyways. As Swink puts it:

“As soon as we encounter a virtual space, we piece together whatever clues we have about the physical laws that govern it into a mental model… based on what can be gleaned from the limited stimulus available: visuals, sounds, tactile feedback and motion. When all these harmonize, the fake physics are seamless; every tiny clue serves to support the same impression of physicality…” (Swink, Pg. 30).

Observing virtual objects, like real ones, allows us to construct a mental model that reflects this physicality. Fortunately, this model has a definition in psychology: a schema. Basically put, a schema is a mental concept that we have of an object, made up of its properties and uses. A schema of “knife”, for example, includes properties like “metal”, “blade”, and “handle”, as well as uses like “slice”. Physical properties are also part of this: we already have a sense of how heavy a butcher’s knife is. Similarly, our schema of “cinder block” also includes descriptors like “heavy” and “dense”.

I’m not sure if animators discuss cognitive psychology in their day-to-day work, but they’ve known about conveying the physical properties of something for quite some time now. Richard Williams devotes a chapter of his book The Animator’s Survival Kit to conveying weight, mostly through the posture of characters interacting with objects of certain weights, and the reaction of those objects to the forces they’re subjected to. This was taken to new lengths in Who Framed Roger Rabbit, which had animated characters interacting with real actors and real-world objects frequently. Fitting, since Williams was the animation director for the film:

Figure 1: A video by kaptainkristian going into detail about how Roger Rabbit’s animation needed to accommodate real-world movement and lighting

Truthfully, we’ve known how to do this in games for quite some time now. When level designers introduce new gameplay elements, typically we see them in action with an NPC or another object before the player is allowed to interact with it. Valve games are masterful at this, and Half-Life 2’s repeated use of this is one of the reasons it’s considered a high watermark for game design.

Figure 2: A video by Mark Brown goes into detail on how Half-Life 2 conveys rules of the game.

So note, specifically, how the characters in Roger Rabbit are unrealistic, and yet we’re not distracted by their presence in the real world. Despite these fantastical elements, Roger Rabbit has achieved immersion through reasoning by simulating real-world phenomena, including physical interactions and lighting. Similarly, to get the player to believe in a monster in a horror game, we need to show their reaction from and impact on the real world in order to more thoroughly construct a schema for them. Given that most of this construction happens when we first observe an object, like the way Half-Life 2 introduces gameplay elements, the first impression is the most important step to achieving its fearsomeness.

Let’s look at a few examples from horror games when they introduce their monsters, paying attention to how they attempt to convince the player they’re part of the world and how effective these attempts are.

Case Studies:

The Last Light:

Let’s start with my game, as I’m most able to contrast the intent and effectiveness of this example. Taking cues from Outlast (which we’ll get to shortly), I knew that the most effective scares could also teach players rules about the game.

So, here’s where the player first encounters the antagonist:

Figure 3: MrKravin’s playthrough of The Last Light, where he first encounters the monster

During this scripted sequence, we learn the following about the monster:

  • It’s a cloud-like entity that inhabits dark areas
  • It has the potential to damage the player
  • It reacts adversely to having light shone on it, and emits a scream in response
  • It retreats into air vents when hurt

While I thought this was an effective way to communicate the rules of the game, it failed to make the antagonist a believable presence. The fiction of the monster is that it’s a swarm of microbes that consumes organic material, flesh included. So, why it’s floating around a couple meters above the ground isn’t explained in the fiction, nor is it believable to the player. Seeing a swarm of microbes scuttling along the floor to the player may have proven more effective. Additionally, throughout the game, there aren’t any real indications of its effects in the environment: no half-eaten remains of people or clothing, no scratches on the walls or floor hinting at its corrosive properties.

A lot of these decisions were spurred by our lack of time and resources, as we were a dozen college students trying to finish a game while balancing three or four other courses. The programmer on our team who was dedicated to the AI, Hughes, did an excellent job with the time he had, but I doubt having the particles of the monster spread out and slither would’ve been in scope. Our artists, Nick and Angelo, knew the limits of their time, which we tried our best to accommodate by minimizing the number of assets we asked for. They also had a very hard rule of “no humans on screen”, which was very understandable.

As I mentioned at the beginning of this article, though, this is just something that didn’t occur to us that we needed to do.

Near Midnight:

I’m only familiar with this game via Jim Sterling’s video on it, and boy is it not good. Watching gameplay of it, though, made me start thinking about the topic of this article. The game is about a ghost leading the player through the house they used to inhabit when they were alive. Here’s Jim’s reaction to first seeing it:

Figure 4: The introduction of the ghost in Near Midnight.

One of the biggest reasons this isn’t very effective is that we don’t learn much about it. From its first appearance, we can gather:

  • The ghost is of a woman
  • The ghost is talking to the player to get them to accomplish something

Much like the rest of the game, the properties of this ghost are sparse and missing some very obvious attributes. First, it just pops in and out of existence at the designer’s whim, with very little fanfare. No fade-in, or phasing through walls like we’d expect a ghost to do, the mesh component is simply flipped on and off. Additionally, the only sounds associated with this ghost are non-diegetic music stings.

There are many things that don’t work about this game, but not being able to believe the antagonist is part of the game’s world is chief among them.


Yay, a good game! I used this example in my previous article as well, as the introduction of Chris Walker serves as both an effective scare and a great scripted sequence that manages to teach the player rules about the game:

Figure 5: The sequence in Outlast where the enemy Chris Walker is introduced

From this sequence, we learn the following things about this enemy:

  • Chris is a man with a very large stature
  • Chris is hostile towards the player, though we don’t understand why
  • Chris is able to speak, though his simple vocabulary implies some sort of disorder
  • Chris is able to reach into narrow gaps in furniture where the player is sliding through and grab them. This is consistent with the rest of the game, and Chris is the only enemy capable of doing this.
  • Chris has immense strength, being able to pick up the player and throw them through a glass pane

There’s a lot more information communicated here than in the previous two examples. The fact that Chris Walker is a person helps immensely, though, as we project our understanding of the physical properties of human beings onto him. In addition to learning properties about Chris and some gameplay rules, his introduction provides an effective scare and advances the plot by getting the player stuck on the first floor of the asylum. Now that’s effective design.

Using this lens, though, makes me understand more about why I thought the ending of Outlast wasn’t as effective, mainly because of it being centered around paranormal entities called Walriders. When first introduced, they’re used sparingly and are usually at an arm’s length. Here’s the first time they’re shown in the game:

Figure 6: The first time a walrider is seen in Outlast, taken from Markiplier’s playthrough

From this, we can gather:

  • Walriders are entities capable of flight
  • Walriders appear to be emitting some energy that looks like flames
  • Walriders take the form of a human head (though later on there are full-body walriders as well)

Unfortunately, the rules of the walrider, both physically and gameplay-wise, aren’t clear at all. Most of the time they’re limited to scripted instances of charging at you. When they are able to pursue you, though, their rules aren’t communicated well, being able to pass through certain walls but not the walls of decontamination chambers. The walrider also appears throughout the last areas of the game out of thin air, instead of appearing from offscreen or phasing through walls like it previously had, undercutting the illusion of it as a persistent entity.

Figure 7: The walrider simply fading in, from Markiplier’s playthrough

Lastly, the walriders seem to have a disproportionate physical impact on the environment, being able to lift the player and Chris Walker despite their ethereal appearance. I imagine animating Chris Walker being thrown around like a ragdoll is a difficult undertaking, but it doesn’t look quite right in-game.

Figure 8: The scripted sequence in Outlast where Chris Walker is killed by a walrider, from Markiplier’s playthrough

Obviously, this doesn’t detract from the excellent first 80 percent of the game. However, this does show us that simulating paranormal enemies might prove to be more difficult.

Alien: Isolation

One of the most effective horror games of the past few years, Alien: Isolation took on a gargantuan task of creating a convincing simulation of the xenomorph from the Alien film series. Instead of making them mere grunts, though, Creative Assembly looked to the original film for inspiration, and tried to create an AI that would recreate the fearsome predatory behaviors of the species.

They pulled this off with aplomb, taking the approach of slowly adding behaviors to the xenomorph’s repertoire over the course of the game. Because of this, though, when they introduce the alien to the player, they have fewer aspects that they can initially show. Nevertheless, the xenomorph’s entrance manages to pack in a great deal of information:

Figure 9: The first encounter with the xenomorph in Alien: Isolation

From this sequence, we learn:

  • That the alien can hide in ceiling vents and exit them
  • That the alien makes specific sounds when traversing through and exiting vents
  • What the alien looks like
  • How the alien’s body is affected by gravity. We get an approximate sense of its weight, especially from the animation of its tail draping over the desk
  • What the alien’s footsteps sound like
  • That the alien can trigger motion-activated doors, further reinforcing its physical presence
  • The sound the alien makes when re-entering a ceiling vent (though the action is not visible on screen, so this association may not be the most effective)

Again, much like Outlast’s introduction of Chris Walker, this sequence teaches us a lot about the rules of the game and the properties of the alien itself. The whole game is built around staying mindful of the alien’s location in order to circumvent it, so the creature is quite frequently heard and not seen. Distinct sound effects are played depending on what action the alien takes, with big bassy footsteps for it walking around, specific clanks for it entering and exiting vents, and the awkward bumping shuffle of it moving around in the ducts. Even when not in sight, the game is working to cement the xenomorph’s physical impact on the world.

It also helps that the gameplay’s rules also seem to strive towards immersion through reasoning, with certain aspects of the game being reminiscent of immersive sims like Thief and Deus Ex. For example, the alien reacts consistently to fire by being afraid of it, whether it’s from a flamethrower, a molotov, or a source in the environment. Similarly, the alien reacts consistently to loud sounds by seeking them out, whether that’s the player sprinting, hitting something with their wrench, using a noisemaker, or engaging in a firefight.

Amnesia: The Dark Descent

A devious designer might be tempted to subvert the arguments of this article and ask “Can we imply the physicality of a monster and scare the player without actually showing it?”. Amnesia’s Kaernk has already answered that question with an emphatic “Yes!”. Though it only lasts about 10 to 15 minutes, the “Water Level” where the Kaernk is introduced became a notorious level in the nascent Let’s Play scene on YouTube.

Though I have qualms about needing the player to run towards the monster to get to a safe area initially, the way this enemy is introduced is simple but effective:

Fig. 10: The first encounter with the Kaernk in Amnesia, from theRadBrad’s playthrough

By simply starting the Kaernk a ways away from the player, they’re given an opportunity to see it move towards them while still having some time to get out of the water. By observing the Kaernk, we learn the following:

  • The creature creates a line of splashes as it moves, allowing us to approximate its location
  • The creature is able to follow the player as they move around on top of boxes, but is only able to attack them if they’re in the water
  • The player can throw objects elsewhere in the water to lure it away, suggesting it’s using sound
  • The creature growls while eating body parts towards it, and occasionally otherwise, suggesting it has a mouth and teeth
  • The creature is able to hit and break physics objects, such as boxes and doors

Despite not being able to see the Kaernk, we are given an abundance of sensory clues as to its presence. It inhabiting the water is not just part of its fiction, but a smart move to communicate where the creature is at all times via splashes. An invisible monster with these rules simply would not work on dry land, unless there were some other way of tracking it (such as footprints in snow or sand).


Designing an antagonist requires a lot of know-how and cooperation from different disciplines, including systems design, AI, character art, animation, and audio. All of them are vital to creating a simulation that facilitates the player to more easily maintain immersion. When next creating an enemy, for a horror game or otherwise, keep the following questions in mind:

  • What aspects of this enemy, if any, break immersion and remind the player that they’re playing a simulation?
  • How does this enemy interact with the world of the game and objects within it? What does that look and sound like?
  • Are there any inconsistencies in the enemy’s presence and behavior? How can you change your design to avoid these?
  • Are you showing the player enough aspects of the enemy when they’re introduced to it? What aspects, if any, are alright to show later?