2025-3-16
Open-world games use exploration as a core dynamic within expansive environments. How do players understand these spaces and what features do they use to navigate through them? A theoretical framework that explains player conceptualization of open-world spaces would be useful for analyzing player behavior and understanding how elements of game spaces affect the player experience, since it would offer a way to categorize game elements and some expectations about their effects and interactions. Thankfully, Kevin Lynch’s work in “The Image of the City” (Lynch 1964) provides just such a framework, although it was developed based on real-world experiences of cities. Through the analysis of numerous interviews, Lynch developed a theory of five elements which contribute to one’s image of a city: landmarks, paths, edges, nodes, and districts. Our central question is: Does Lynch’s framework apply to open-world games, despite the differences between real-world cities and virtual play spaces?
To explore this question, we had participants play 2 open-world games: A Short Hike (Robinson-Yu 2019) and Sable (Shedworks 2021). Echoing Lynch, we asked them to describe how to navigate between places they had visited, and using thematic analysis, looked for references to Lynch’s 5 categories in their answers. We found that some but not all of Lynch’s categories were clearly present, while a new category emerged: “characters.”
In the The Image of the City (Lynch 1964), Kevin Lynch develops a theory of how people understand cities. This theory was widely influential in urban planning and related fields. Through interviews with people in Los Angeles, Jersey City, and Boston, Lynch concludes that for any city, one’s mental image includes five important aspects: paths, edges, districts, nodes, and landmarks. Decades later in his retrospective, Reconsidering The Image of the City (Lynch 1984), Lynch explains some limitations of his original study, including the limited scope of the data, but notes that it has been replicated in multiple cities across the globe. Lynch’s aspects potentially form a good starting point for understanding virtual spaces.
To analyze our data, we used thematic analysis (Braun and Clarke 2006). The two main approaches to thematic analysis are inductive (let the data guide the themes identified) and deductive (look for themes relevant to an existing theory) (Braun and Clarke 2006, 12). We use deductive thematic analysis based on Lynch’s elements, but also tried to pick out other strong themes we saw.
Others have proposed frameworks for understanding exploration in open-world games. Nicole Vickery and Peta Wyeth investigate exploration in Open-World Games in the The Witcher 3 (Vickery and Wyeth 2023). They used a diary study of 12 participants playing three times a week for two weeks, with participants to completing diaries on their own time, and selecting a primary and secondary activity for each session. Their framework differs from ours in that it is based on Ernest Adams “Fundamentals of Game Design,” which posits 4 challenges associated with exploration: identifying spatial relationships, finding keys, finding hidden passages, and mazes and illogical spaces (Adams 2014).
In contrast to our work, Vickery and Wyeth’s approach centers motives, supported by Bartle’s 4 Player types (Bartle 1996), and focuses more on the exploration process than on mental representations of the space. Their research used diary entries completed individually after gameplay as opposed to interviews and observation during play. Their results complement ours, providing a window into player motivations and experiences during exploration.
Vickery & Wyeth’s framework is useful because it provides a background for motivations behind player exploration. Our research hopes to categorize the basic elements that players use for navigation, which could give analysts like Vickery & Wyeth an established lexicon of game elements to talk about when describing how a game shapes player exploration.
Others, such as Nathan Hughes, have studied open-world games with a broader focus. In “Understanding Specific Gaming Experiences: The Case of Open World Games,” Hughes examines what it means to experience an open-world game and how experiences are unique to specific games (Hughes 2023). He establishes a consensus among players that the open-world game experience involves giving the player freedom to self-pace what they engage with, which allows the player to enter unique context-situated experiences and exercise the agency to choose which experiences they partake in. He identifies two potential frameworks for understanding player action, a motivation framework and a goal-based framework. He concludes that the relationship between action selection and player goals is not 1-to-1, and that open-world game exploration is a flexible negotiation between players’ numerous goals and immediate desires.
Melissa Kagen has also studied exploration in games broadly, with a focus on wandering and its implications for societal ideas of work, gender, colonialism, and death (Kagen 2022). She discusses how walking simulators present a radical potential for spatial storytelling, and how wandering in games is an activity to provoke unproductive, anticapitalist play. Player complaints that these games are “boring” are tied to late-stage capitalism and anxieties around purposelessness. This is interesting in the context of our study where we ask participants about their feelings around exploration with questions such as: “Do you usually find exploration in games more exciting or more frustrating?”
We asked our participants to play either Sable (Shedworks 2021) or A Short Hike (Robinson-Yu 2019) for 90 minutes freely. At the 60-minute, 75-minute, and 90-minute marks, we interrupted them to ask “goal probe” questions. At each goal probe, we asked them: “How would you describe what you are doing right now?” followed up with “Why are you doing that?” repeatedly until their answer went beyond the scope of the game. This aimed to get a sense of players’ goals and motivations. After 90 minutes, once the player got to a stopping point, we had them complete a 30-minute follow-up interview, which covered:
Demographics: what age group they belong to, how often they play video games in a week, whether they have played the game before or have previous exposure to it, if they have played similar games, and their self-reported status as underrepresented in gaming communities.
Path questions: we asked questions based on paths within the game, using pre-prepared screenshots and agreed upon place names to help identify places. We pre-specified three landmarks A, B, and C in each game, and then asked about players’ navigation from A to B, from C to one of a set of other landmarks based on where they actually went, and then between two more points that we asked the player to choose. For each path, we asked 3 questions:
Additionally, if time allowed, we asked the participants if they found any of the regions of the world particularly immersive or captivating and their emotional reaction to these aspects.
We employed these questions to get participants to mentally conceptualize aspects of their navigation within an open-world game and see if the features that Lynch mentions in Image of the City were also reflected in their mental models of the space. This paper focuses on the participant’s answers to the path questions, and in particular, which game elements players referenced as they described navigation, mirroring questions that Lynch asked in his study.
After gathering data, multiple members of our team examined recordings of the interviews and coded game elements and other things the players referred to in their responses to path questions. To identify common themes, we used deductive thematic analysis to look for references to Lynch’s 5 elements, or other strong themes. We discussed our codes in a group to establish correspondences and focused on codes where more than one of us had identified the same thing to identify themes. Overall, we followed the six phases of thematic analysis outlined in (Braun and Clarke 2006).
Our methodology has some notable drawbacks: given that we only had players play for 90 minutes, in Sable most players were not able to make it out of the starting area. This was not as much of a limitation with A Short Hike, where most of our participants completed the main objective of climbing to the top of the mountain, or at least got close. Also, by not directly asking about nodes, edges, paths, or landmarks, we may have missed some opportunities to hear from them more directly about their mental models of the world. Our assumption was that if we asked similar questions to those asked by Lynch, we’d hopefully see the same kinds of patterns in the responses, but as we discuss below, this did not turn out as planned. Finally, following Braun’s phases of thematic analysis, we would have ideally completed two rounds of coding, consisting of both individual transcription and team discussion. In contrast, we only completed one and decided instead to discuss together after our first level of coding and agreed on certain terms and their meanings post-hoc. Despite not performing it exactly as its described, we followed Braun’s model on a high-level.
| Theme | A Short Hike | Sable |
|---|---|---|
| edge | X | |
| landmark | X | X |
| cardinal-direction | X | |
| compass | X | |
| in-game-tools | X | X |
| invitation-to-explore | X | |
| NPC | X | X |
| vantage-point | X | |
| visual-cues | X | X |
In addition to the limitations discussed in the previous section, due to bandwidth constraints, we have only been able to fully analyze data from 3 participants. There are still some interesting results that are shaping our thoughts on further study of this topic which we believe are worth sharing immediately (especially since others like Vickery & Wyeth might make use of them). However, it’s difficult to draw firm conclusions from the data we were able to analyze. The findings we’ve uncovered so far suggest that revising our approach is a better path forward than simply gathering more data under the same protocol, which is another reason for us to share our results now before moving on.
We were able to find 9 themes among the game elements that players mentioned in response to our path questions, plus 2 more themes relevant to navigation. Of the 5 elements Lynch describes that we were looking for, only landmarks and edges came out as themes, but non-player characters (NPCs) came out as a strong new theme. Several themes like “compass,” “in-game-tools,” and “vantage-point” relate to how the player explores, rather than what elements of the game world they retain in their mental map of the space. The themes we found were:
A few themes appeared in Sable but not A Short Hike. For example, cardinal directions and vantage points were used more in Sable vs. body-relative directions in A Short Hike. This can be explained by camera differences: A Short Hike has a third-person top-down view where you can’t look around freely; Sable has a first-person view. In A Short Hike, your view rotates as you move, always pointing somewhat towards the center of the mountain, so it’s hard to determine cardinal directions.
Landmarks and edges were the only themes from Lynch that were prominent in our results (edges only in Sable). Although some of us coded paths, nodes, and/or districts, we didn’t reach consensus for any participant. While our study design and small dataset likely contributed to this, there are differences between real cities and game spaces that might explain it. For example, in both Sable and A Short Hike, paths exist but are not explicitly named like in cities. Thus, in order to refer to a path, a player would have to describe it using other features such as landmarks. In that case, it’s easier just to describe navigating by the landmarks directly, instead of trying to refer to the path. In contrast, in cities there are many paths that lack distinctive landmarks but they each have a name that can be easily observed, so referring to them is efficient. Nodes and districts may simply have been less prevalent in A Short Hike (and in the parts of Sable our participants reached). Another factor is familiarity: Our players were new to the game spaces they explored, whereas Lynch interviewed long-time residents. An unfamiliar space might be easier to understand in terms of obvious visual features like landmarks, while over time players might refer more to nodes like the Ibex Camp just by their names instead of by their prominent landmarks (like the watchtower). Districts similarly might be used once a player is more familiar, and in Sable, our players never made it out of the starting area into the broader world where there are more distinctive districts.
We did identify “NPCs” (“characters”) as an important element for navigation. This theme is prevalent across all participants and both games. NPCs serve as reference points for navigation, but can also dispense information or directions. For example, Participant 3 describes the NPC Sizo as “pointing to the place you’re supposed to go” and granting them a navigator to find it. Of course, the presence of and the possible interactions with NPCs in a game are usually dependable, in contrast to humans in the real world, which explains why “people” aren’t a category of Lynch’s theory. NPCs don’t fit into any existing category: they are most similar to nodes, since they usually support a specific set of affordances (unlike landmarks, they are usually not visible from a distance). However, unlike nodes and landmarks, they often move around within the game space, appearing in dependable places at dependable times such that they are useful in navigation, but potentially appearing in many different places, and/or moving on dependable routes. Navigation instructions like “follow [NPC]” or “go to the clearing where [NPC] will appear in the mornings” wouldn’t make sense for landmarks or nodes.
Other themes beyond Lynch’s, such as in-game-tools, compass, visual-cues, and vantage-point describe game elements that players interact with during exploration. While these aren’t part of one’s mental image of the game world, they are important for navigating it, which is why one might refer to them when giving directions. Similarly, they are more uniform and predictable in a game than in the real world.
Wyeth and Vickery’s categories for gameplay in The Witcher 3 include exploring, conflict, and directing narrative. They focus more on player motivations than game elements, but another notable difference was their ‘combat’ category. We used games without violence as a central mechanic, but it’s a prevalent mechanic in many open-world games, and how that might affect one’s conceptions of game spaces is an interesting question. Wyeth and Vickery found that combat furthers worldbuilding and guides players’ exploration without the need for explicit barriers. This makes ‘challenge level’ another possible aspect specific to games, although it was not emphasized by our participants.
Our findings are still limited by our small sample size, but our preliminary results point to paths, nodes, and districts possibly being less relevant to players’ navigation in open-world games than in real-world cities, and they definitely point to NPCs as an important element for open-world exploration. More work is necessary to investigate whether or not paths, nodes, and districts are actually less important in games than in cities, and if so why and whether that changes as players become more familiar with a game.
Given our mixed results, we plan to revise our study design before gathering more data. In particular, asking players more explicitly about their mental models of the world may be useful, as well as using a journal response format to allow players to become familiar with the game worlds over multiple play sessions. Trying to study how players’ conceptions of the space change over time might also be interesting. We could also focus on known systematic differences between real-world cities & game worlds (like the relative lack of path labels in game worlds), which this data suggests might lead to differences in conceptualization. Finally, including games that feature combat as a central mechanic might lead to extra themes in the results.