The Last of Us: Using Game Elements to Subvert Expectations

In March 2016, whilst working at Guerrilla Cambridge, I wrote an article titled ‘Subverting Player Expectation’ which was a discussion about making a game engaging and impactful by altering established mechanics and gameplay. This focused on examples from the original “The Last of Us” and “Batman: Arkham Knight”.

My original article on subverting player expectation raised these points:

• You should have established a pattern before considering subverting the pattern.
• You should not foreshadow the event, otherwise, it makes it predictable and removes the potential for drama.
• The reveal of the subversion must be clear, concise, and sudden to avoid any confusion and maintain the drama.

I continued with this subject and did a talk at Develop in Brighton, in the UK, later the same year and furthered my thinking with this additional guideline/rule: You should give context before the event so that the player doesn’t feel cheated by something that doesn’t fit into the world.

Shortly after I gave this talk, and now 4 years ago, in March 2017, I was lucky enough to join the incredible team at Naughty Dog and work on my dream game, “The Last of Us Part II” as a Designer.  Furthermore, I was fortunate enough to get an opportunity to create something of my own that subverted expectation, where I could implement my previous thoughts and findings from the original article.

It is particularly exciting that I get to create a follow-up article using some of my own work (the original article was about the inspiring work of others) as I discovered some new things that I wanted to add in order to further a conversation on the topic.

WARNING: This contains spoilers for “The Last of Us Part II”.

If you don’t recognize this street (below) I would recommend not reading this article just yet.

If you do recognize the street, but don’t recognize this optional building, then I wouldn’t read the article just yet if you don’t want spoilers.

If you recognize the optional building but don’t recognise specifically this workbench, then please stop reading until you’ve played this section, unless you’re fine with this event being spoiled!

Level Brief

I worked on 4 levels in “The Last of Us Part II”, one of which was called ‘The Seraphites’ where Ellie left the theatre by herself in search of a character named Nora whom she wanted to exact revenge on. I had a brief for this level that Ellie should feel alone as she had left Dina behind in the theatre and we wanted to emphasize what Ellie would sacrifice in order to pursue revenge.

I started to think about ways I could emphasize the player’s loneliness through game mechanics as I believe the strongest way to tell a story is through engagement with the gameplay mechanics and the gameplay beats themselves; a particularly strong point of Naughty Dog games.

I thought about the mechanics we had that involved a buddy character and how I could use them to make the player directly feel firsthand the drawback of being by themselves. I initially thought about showing a ‘buddy boost’ opportunity to reach a higher ledge that the player couldn’t use on their own, which would require a dumpster to progress, which then led me into thinking about how I could emphasize loneliness in a puzzle. The physics dumpster puzzle was created as a twist on the buddy mechanic of opening a garage door together.

This was a good starting point, but I felt like one moment wasn’t enough, so I looked for other opportunities throughout the level where I could keep referring back to the brief of emphasizing loneliness.

The Inspiration of an Idea

One of the aspects I loved about the original “The Last of Us” was the backpack menu and how it didn’t pause the game. It was perfect for keeping the tension raised during combat as the player could be attacked at any moment; it is such a simple but elegant feature that upholds the game’s core pillars, even during its menus.

I had an idea from thinking about this tense menu experience and how it could also be extended out to the workbench menu as well, as this also didn’t pause the game when the player was using it. I felt there was an opportunity to emphasize loneliness by creating an impact through subversion as if I interrupted the player in this menu it was still consistent with the rules of the Last of Us world and game mechanics.

My idea was to ambush the player during the workbench upgrade menu so that it would support two aspects. The first was for referring back to the brief of emphasizing loneliness; if the player wasn’t alone, Dina would have been able to warn or protect Ellie and she wouldn’t have been ambushed whilst her back was turned. The second was that it emphasized that you were never safe in the Last of Us Part II world; you should always be on your toes and tense about whether you’re ever safe. I’ve read many a tweet from people saying they didn’t feel safe at a workbench after this encounter, and despite it being stressful for players, it meant to us that it was successful – our world was supposed to be tense throughout.

One menu I knew I didn’t want to touch was the pause menu; that should be used when people want to pause the game so they don’t need to interact with it. I particularly wanted to make sure that the ambush happened within the first few seconds of going into the workbench menu so that it accounted for players who assumed it was ‘paused’ in the traditional sense. This was because if they took their attention away from the game momentarily, the ambush would happen within a short enough time that they would still be able to react to the game, even if that action was just to pause the game for real.

Established Pattern

I wanted the player to go into an apartment that was currently being occupied, take all the items from these people, then get ambushed whilst Ellie’s back was turned and she was vulnerable; something that wouldn’t have happened had Dina been with her.  I didn’t want it to just be a random jump scare out of nowhere, but rather subverting player expectation of a learned mechanic (the workbench upgrade menu) in order to create a memorable moment that strengthened the Last of Us Part II world and narrative through gameplay.

One of the rules that I discussed in my first article was: You should have established a pattern before considering subverting the pattern.

My original article talked about 3 being the smallest pattern you can have, so the earliest time you could consider subverting expectation was on the 4th interaction with the pattern.  This workbench was the 8th workbench at that point in the game, so it was well within the established pattern rule.  At this stage, players had also got complacent that the workbenches were ‘safe’, so it was a good place to introduce something that would ensure they were engaged.

Initial Development

I knew that to have the biggest impact when pitching this event it had to be playable. To give someone the desired reaction from subverting their expectation of a learned mechanic they needed to experience it first hand. The safest course of action would have been to pitch it on paper first, but I knew this wouldn’t have anywhere near the impact I was aiming for in the final game.

I did the very risky thing, and usually not recommended, of asking people on the team to invest time and effort into work that hadn’t been approved. I approached my scripter and my animator on this section at the time, Banun Idris and Jonathan Cooper, and pitched to them my idea, what I wanted to achieve, why, and what I needed them to do to help me prototype it. Luckily, they were both on board with the idea and keen to help as they thought it was really interesting and something worth testing, given how it supported the initial ‘loneliness’ brief. We also talked to Derek Mattson, the scripter on the workbenches, and he supported us too by providing us with a unique workbench that would show the initial menus, but let us run a custom ambush setup afterward.

I created a small environment that was a hide-out inside an abandoned office building. It had evidence of people living there, such as hanging clothes and repurposed structures like makeshift toilets and tents. It had lots of scavenging items including parts to use at the workbench. I asked some other designers to play it and act exactly as they would if it were the real game; pick up ingredients, craft items, upgrade weapons at workbenches and upgrade player abilities with any upgrade manuals. I was getting great reactions for the ambush moment itself, so I knew I was on to something impactful, but the initial testing showed some interesting discoveries about how clear it was.

Another one of the rules from my original article was: The reveal of the subversion must be clear, concise, and sudden to avoid any confusion and maintain the drama.

The ambush animation itself was very clear and very sudden; the player was grabbed by an enemy and entered into a button-mash struggle that filled the screen. It couldn’t be missed and it was very understandable as to what was happening. Nobody had any issues understanding what was happening at the moment, but after the attack was completed there were confusion issues.

People were spending time clearing the area completely before using the workbench, and so when they were ambushed they were confused about where the enemy had come from. I had added a hole in the roof near to the workbench from which, once the ambush had been triggered, there was a rope hanging. The story was that the enemy had been hiding up in the roof and had pulled the rope up with them so no one could follow them, and thus the player couldn’t initially get to them. In order to get down to the workbench to ambush the player, they’d let the rope down. Once they were defeated the player could climb the rope to get into their hiding place and receive some bigger rewards, including an upgrade manual, so they didn’t feel so hard done by from an attack, given the valuable item they received as a result.

I had moved the camera to face the rope during the ambush and had enemies coming from that direction, but I found players were either getting turned around during combat, pushing further into the environment to give themselves some breathing room, or naturally moving around the environment during combat, so that by the time the fight was over they had left the rope behind and missed the reward area altogether. This led me to make the reward area a room that was further away from the workbench and subsequently the last thing to pass on the way out of the space. This positioning remained as it is in the final game with the open door to further encourage players to find the room and make the discovery.

I had also locked the door before the ambush in the final game so that when players tried to open it and found it locked, they’d understand that the entire space hadn’t been cleared. This helped people to realize after the ambush that as they hadn’t cleared the whole apartment they’d left people alive in the locked room, even if they couldn’t have done anything about it – apart from laying trap mines down outside the door for super cautious players!

When we made new hires during the project, my scripter on this section, Banun, moved to a different level and I was then teamed up with Alex Stewart for the last 18 months of the level’s development. He scripted in the behavior that if you lay trap mines either outside the locked door, along the corridor leading to the workbench, or just around the workbench itself, it will trigger a different scenario of blowing up the ambushers. It’s super satisfying and I love that we supported it for either cautious players or second play-throughs.

Focus Testing

After a lot of iteration, I had a layout that was pretty similar to the final version that shipped in the game.  The rest of the iterative process was spent fine-tuning how much context we gave in the environment without foreshadowing the event. We got the workbench ambush into the game and I watched the focus test videos of the public playing this level with keen interest.

The additional guideline/rule I added to this topic during my Develop talk, was: You should give context before the event so that the player doesn’t feel cheated by something that doesn’t fit into the world.

To give an example of what I mean by context; if you were in a new, clean, office building and you went up the stairs, if the stairs collapsed so you couldn’t get to the exit you’d feel it was out of place and the event was a bit of a cheat. If, however, you were in an office building that had been long abandoned and it was falling apart, and you went up the creaky stairs only for it to collapse, that would be fitting with the world. The context was there to say the stairs were not structurally sound and so a breakage doesn’t seem out of place.

For the workbench ambush apartment, I added living plants as one of the main signifiers that this space was currently occupied. I did this even in the blockmesh phase before any art was introduced as it was super important to give the sense that space was lived in. Usually, I wouldn’t dictate how a room should look down to the props, as I like to give the artists on the area the freedom to inject their take on the visuals here, but I had great discussions early on with one of my incredible artists on this section, Ben Springer, about what we needed to include in order to sell the space without being too obvious; I didn’t want to run the risk of foreshadowing an ambush and sapping the drama from the event. The living plants by the window also had one of my favorite arrangements of contextual pickups; a trowel (blade ingredient) and some fertilizer (explosive ingredient) to not only draw the player over to see the plants but to also show that someone was currently tending to these plants as a food source.

The inspiration for a food source came from the film “The Survivalist”, which Neil Druckmann had recommended for me to watch when we were talking about the inspiration for The Last of Us Part II, where seeds were a bargaining chip for survival in that post-apocalyptic world. It reminded me of what the most important things in this world were and what people would immediately start doing when they set up a new place of refuge; get a food source.

The end result of the context in the final game to show that the apartment was being lived in worked really well, thanks to the careful detail in the artwork. Many focus testers would comment on the plants, in particular, looking healthy and generally out of place in the desolate Last of Us II world, to then recall them being suspicious after the ambush.

This level was actually used as the ‘Press Preview’ demo a month before the release of the game, and it was awesome to hear some of the previews mention not just this ambush moment, but the context prior to it as well. Aofie Wilson of Eurogamer had the exact experience I intended, of it being noticed but not being foreshadowed, and said in her preview:

Early Positive Feedback

After one of the early focus tests, Anthony Newman, one of the Co-Game Directors, walked up to my desk and said he’d just had Evan Wells, the president of the company, call him about the workbench ambush. My heart sank – oh god, what could this mean? Did he hate it? Have I wasted my time? Am I fired? Maybe I’d become accustomed, from previous companies to a company meeting or a phone call from the studio head often meaning that people were being made redundant, as I had a mild panic.

But this panic was short-lived, as Anthony continued with the next sentence – “Evan loved it and it’s his favourite part of the game so far, so nice one!” Oh my word, I couldn’t have been happier! What a result that was! The president of the company had played it and enjoyed it so much he’d actually called one of the directors to complement it. That was not the reaction I had expected, but boy was I glad it had turned out that way (and not just because I could keep my job!). This gave me more energy to pour into this section, and I was straight off to the desks of the people who were working on it with me to tell them the good news and to share the praise.

In fact, it was getting such positive feedback that Derek Mattson even put the effort in to extend the standard workbench script so that any workbench in the game supported being attacked during systemic gameplay as well. If at any time in the game there are enemies present in the same space as the player whilst the player is using the workbench, they can potentially be attacked and kicked out of the menu. An area this was most likely to occur in was the Capitol Hill area of Seattle, just after the Eastbrook Elementary school fight (another level of mine and Alex’s!), where you can interact with a workbench inside a petrol station and then some enemies enter the area on patrol. Spend too long upgrading your weapons and you run the risk of being attacked.

Overstepping Into Foreshadowing

The early focus testing with the public was going well, but it wasn’t super obvious to every tester as to why the ambush had happened or where the enemy had been. There were a few confused experiences or comments, and occasionally the odd negative reaction. Although it was still early in development, I wanted to make sure that every tester was having the desired experience; one where they were caught out by the enemy, but understood what had happened; the enemy had been hiding in the locked room and had previously been occupying the apartment. I started to look more at the context prior to this event to see what could be improved.

I had been discussing with one of the audio designers working on this section with me, Jesse Garcia, about wanting some audio cues to give more prior context as to what happened. One of the things I really pushed for was the muffled sound of a door closing when you approached the apartment; something Jesse had delivered exactly as I’d intended on his first implementation! The player had either smashed a window on the second floor to gain access to the building or triggered a sound trap of hanging bottles downstairs if they approached through the laundry room on the ground floor.  These two noisy entrances were being used as the reason for the enemy hearing the player approach and then hiding in the locked room to protect themselves. This muffled closing door was a great audio cue that people could think back to and realize that they’d heard the evidence of people occupying the space.

One thing Jesse had discussed with me that I kept thinking about was the idea that Ellie could get her gun out and go into an uneasy tension mode/animation when she heard the muffled door closing. I’d thought about it for some time and it was really interesting, but I also wasn’t sure if it was too much; would it foreshadow the event and thus remove the element of surprise if the player character was expecting it? I had thought it was a cool idea though, and I’d been inspired by it, so thought it was worth trying out as I wasn’t sure what the outcome would be. One of the fantastic things we do at Naughty Dog is constant focus testing with new players, so I asked my extremely talented scripter and teammate Alex if he thought it was worth trying to see what the result was.  He said it was super easy to try out and we implemented it on my machine with a couple of lines of the script he knew. It looked great having Ellie get her gun out in reaction and go into a new tension mode, as well as an additional line of dialogue to accompany her tension change, and I was giddy with excitement that it had been that easy to try out. We watched the following focus test eagerly.

I was totally wrong, however, and it worked against us for what I had been trying to achieve. We had oversteered too heavily into foreshadowing the enemy, so much so that some testers were throwing bricks into the apartment and entering listen mode, just to see where the enemies were hiding. It made the inevitable workbench ambush feel like a cheat as these testers had combed the whole apartment making sure it was 100% safe before upgrading their weapons. Thank goodness it was only a couple of lines of script and within seconds of us seeing the results, Alex had already stripped the lines out.

Giving Context After the Event

I was dismayed at the test results, so Alex and I went and talked to Anthony about what had happened, to see if he had any thoughts; I’d been working on it for so long I needed a fresh pair of eyes, and ears.  I explained how I thought it was important to give context prior to an event happening so it didn’t feel unfair and too out of the blue. He listened to my theory and guidelines that I had previously come up with and gave some great feedback.  I think I had mentally ‘ticked off’ working more on the context after the ambush during my work early on in the blockmesh phase. This was where I had moved the reward room to be the last section the player passes on the way out of the space, and I had felt that I didn’t need to revisit it again. Not only was he feeling that I’d focused too much on the prior context and my newest rule of the importance of prior context, but he had a great bit of insight. He suggested it was more important to show the reason that something had happened after the event had occurred, as players would struggle to think back to before an interesting event. This was especially important if the event involved combat, as that’s great for wiping players’ memories as it takes all the attention away from the previous situation.

We discussed it at great length, throwing ideas back and forth, and with this post-event focus in mind, we leaned heavily into emphasizing the post context and included some more elements to make it even clearer. I went back to my artist, Ben Springer, and begged him for some alterations, but thankfully he not only understood the reasoning, but he also offered great suggestions for stronger visual sells.

The elements we included to help players realize what had happened – that the enemy had heard the player enter, went and hid in the locked room, and then ambushed the player when their back was turned – were:

• Opening the locked door outwards, rather than inwards, so it acted like a funnel for players into the newly opened room when they were exiting the apartment down the main corridor. We also added a new element on the back of the door; a hanging green towel that grabbed the player’s eye and further showed the door was in a different state, as the towel couldn’t have been seen from the other side when it was locked.

• Lured players to the open door with shiny pickups, something that hadn’t been there previously.  Pickups are also great breadcrumbing to ensure players take desired routes.

• We had environmental storytelling that the enemies had all been sleeping in one room, and thus why there were 4 of them.

• We added a note inside the room which further explained that they were deserters from the WLF and had been hiding in that apartment to avoid being found by soldiers who had been sent by Isaac (the head of the WLF) to find them. Additionally, a note that you could pick up from the last infected in the bar encounter, a few gameplay beats earlier on the same street, linked to finding these deserters. The infected had been the WLF soldiers sent to find these deserters and they had ambushed you assuming you were after them. This was also mentioned by Aofie Wilson in her preview for Eurogamer, saying: “What was even more interesting is that what we found in that building had been previously hinted at by a letter dropped by an infected earlier in our playthrough.”

All these elements strengthened the context given after the ambush to explain what had happened, which is important for my original rule of making the subversion clear and concise to avoid confusion and maintain the drama. They were really successful in creating the desired player experience and it was a pleasure to then see every subsequent tester enter the room and proclaim “Ah!  So THIS is where they were hiding!” before claiming their reward of an upgrade manual, sat nicely in front of them at the foot of the bed.

Conclusion

My original article on subverting player expectation had raised these points:

• You should have established a pattern before considering subverting the pattern
• You should not foreshadow the event, otherwise it makes it predictable and removes the potential for drama
• The reveal of the subversion must be clear, concise and sudden to avoid any confusion and maintain the drama

I continued with this subject and during my talk at Develop I furthered my thinking with this additional guideline/rule:

• You should give context before the event so that the player doesn’t feel cheated by something that doesn’t fit into the world

I tried to push this new guideline of prior context as far as possible in the development of “The Last of Us Part II” and I found that I went too far and crossed the line into foreshadowing, completely ruining the intended experience. From testing the results and discussing with Anthony and Alex I wanted to add this new guideline/rule:

• The most important context to give for an event is after it has happened, rather than before it happens. This is because players will find it harder to think back to before an engaging event that grabs their attention than to experience any context afterward. This is especially important if there is combat involved as it is good at wiping people’s memories as to what had previously happened.

I hope this has been insightful and useful for anyone interested in subverting a mechanic in their work. I look forward to playing any future games that pleasantly take me by surprise and create an impactful drama that strengthens the game’s world and narrative through gameplay.

Pete Ellis, Game Designer at Naughty Dog