Game Studies

Genre, Prototype Theory and the Berlin Interpretation of Roguelikes

Abstract

Fans of “roguelike” video games have long been engaged in an intense debate over what this genre label means, and how to understand the task of thinking about it. This article argues that the attempt to define “roguelike” known as the “Berlin Interpretation” is an example of the “prototype theory” of categorization in action, and then moves to a consideration of the strengths and weaknesses of this type of categorization for understanding video game genres. Using examples from roguelike fan discourse, this article shows how the problem with the Interpretation lies in how it privileges game mechanics as the salient features of games. A more productive place to ground discussion might be experience and skills, which can also be extrapolated from prototypical examples.

Keywords: roguelike, genre, prototype theory, philosophy, Wittgenstein

Introduction

Roguelikes are a type of video game that has existed and endured in popularity since the early 1980s, when Rogue was released. Rogue inspired many imitators and much experimentation on its original formula, leading to the use of the term roguelike to describe the games that came after it -- roguelike games are gameslike Rogue [1]. This label emerged from online discussions and debates within the roguelike fan community, and it has become the established genre label for this type of game. However, despite this term becoming accepted and widely used, a remarkable, long-lasting debate has taken place among roguelike fans over what the label means. Part of the reason for this lies in the nature of the label itself: saying something is “like” something else is vague because there are numerous ways things can resemble each other, and resemblance comes in degrees. As such, fans and developers have found it difficult to pin down exactly what makes a game “like Rogue,” and how this task should be conceptualized has been a matter of some debate. The most famous attempt at defining “roguelike” is known as the “Berlin Interpretation,” a list formulated at the 2008 Roguelike Development Conference of fans and developers that specifies several “high value” and “low value” game mechanics used to determine how Rogue-like a game is. The Berlin Interpretation and the history of the discussion of the term “roguelike” represent one of gaming history’s most public and most detailed attempts to concretely specify the meaning of a particular genre label.

This study examines the Berlin Interpretation of roguelikes (and its subsequent criticism) from the perspective of genre and prototype theory in cognitive science; a view that can be traced back to Eleanor Rosch’s work on natural categories. Prototype theories suggest that when learning and formulating concepts, the human mind works by anchoring its understanding of a concept in the best, ideal, “prototypical” examples of that concept being instantiated. Rather than working from rigid Aristotelian categories, criteria, or “necessary and sufficient conditions” for belonging to a category, human categorization is more fluid, with more porous boundaries. Prototype theory argues that when we as human beings categorize things, we begin with prototypical examples and work outwards from them, formulating our concepts based on the most representative examples of the things we want to categorize, and then we assign things membership to particular categories by comparing them to those examples. I will argue that the Berlin Interpretation, by grounding itself in five games it deems “canon” for roguelikes and formulating a definition of “roguelike” based on their salient features, can be understood as an example of prototype theory in action. However, the Berlin Interpretation has received much criticism in the roguelike community, either for being an overly restrictive practise or for misunderstanding what counts as the salient features of a game. One particularly pertinent critique has been that the Interpretation operates solely in terms of mechanics and lacks an understanding or theorization of the experience these mechanics engender or the skills required to play the game, which are arguably more productive starting-points when it comes to defining a type of game. By examining the Berlin Interpretation and its criticism, I will consider how prototype theory could be implemented in an understanding of video game genres and evaluate the benefits and limitations of this approach.

Video Games and the Study of Genre

“Roguelike” (like “strategy,” “puzzle,” or “first-person shooter”) is a video game genre label, and since this article discusses prototype theory as an approach to understanding this label, I will begin by discussing genre. In the public consciousness, genre is most often understood in a way that resonates with the Oxford English Dictionary’s definition: “a particular style or category of works of art […] characterized by a particular form, style, or purpose” (Oxford English Dictionary, 2024). So even in this most basic sense, genre can be understood as a form of categorization. But academic study has demonstrated that genre is an immensely complex and wide-ranging phenomenon. In the handbook Genre: An Introduction to History, Theory, Research, and Pedagogy, Charles Brazerman describes this complexity as such:

Many aspects of communication, social arrangements, and human meaning-making are packaged in genre recognition. Genres are associated with sequences of thought, styles of self-presentation, author-audiences stances and relations, specific contents and organizations, epistemologies and ontologies, emotions and pleasures, speech acts and social accomplishments. Social roles, classes, institutional power are bound together with rights and responsibilities for producing, receiving, and being ruled by genres. Genres shape regularized communicative practices that bind together organizations, institutions, and activity systems. Genres, by identifying contexts and plans for action, also focus our cognitive attention and draw together the dynamics of our mind in pursuit of specific communicative relations, thereby exercising and developing particular ways of thinking. (Brazerman in Bawarshi & Reiff, 2010, pp. xi-xii)

Given all this, it is unsurprising that there are multiple approaches to and understandings of genre. Within the realm of art, the concept of genre has had a broad range of applications as it is widely used in its study, categorization and marketing [2].

Since different media are studied and thought about differently, it can be problematic to apply an understanding of genre in one media to the study of another. Mark Wolf has argued that this is the case with film genres and video game genres, even though video games often overlap with films in terms of narrative themes, structure, tone and iconography. Film genre labels usually indicate something about a film’s features to audiences. The genre labels “comedy,” “romance,” “horror,” “drama,” “war,” “science fiction,” or “Western” give audiences a glimpse of what kind of experience they are in for: they indicate a film’s themes, narrative structure and perhaps its setting or tone. Wolf acknowledges that these types of features are important and productive for film genre analysis but argues that this type of approach would have limited productivity for the categorization of video games. Wolf’s argument comes through most explicitly in his discussion of iconography:

Just as different forms of dance (foxtrot, waltz, ballet, jazz) are defined by how the dancers move rather than how they look, an examination of the variety and range of video games reveals the inadequacy of classification by iconography of even narrative-based games. While some video games can be classified in a manner similar to that of films (we might say that Outlaw is a Western, Space Invaders science fiction, and Combat a war game), classification by iconography ignores the fundamental differences and similarities which are to be found in the player’s experience of the game. (Wolf, 2002, p. 115)

The gameplay experience tends to be what modern video game genre labels convey to players, with labels like “third-person shooter,” “puzzle,” “turn-based strategy,” “beat-em-up” (and so on) immediately conveying a sense of what the game will be like to play, as well as its base mechanics. “Third-person shooter” tells a potential player that the game will take place from a third-person (rather than first-person) perspective, and that there will be ranged combat mechanics, most likely involving gunplay. “Colony management simulator” tells a player that they will be managing the lives of a group of people, which will highly likely involve construction and management mechanics. But while gameplay experience and game mechanics seem like a more productive approach for video game genre labels, they nonetheless pose their own problems for genre classification: gameplay experience is difficult to specify and gameplay mechanics often overlap between genres. All of which suggests an important question: how exactly can it be known that something belongs to a particular genre? And regarding video games in particular: is listing mechanics the best way to convey what a gameplay experience involves, or the best way to approach genre categorization? If genre is at least partially about how we categorization and communication, these questions clearly bear importance for the study of genre.

One way to think about genre might be to think as philosophers have often thought about knowledge: that is, in terms of necessary and sufficient conditions. Ever since Plato discussed such conditions in his Theaetetus, philosophers have debated whether knowledge can be understood as “justified true belief” (Plato/Campbell, 1883). This account entails that to know something I must believe it, it must be true and I must be justified in believing it. For something to belong to the category of knowledge, it must meet these necessary and sufficient conditions. So perhaps there could be necessary and sufficient conditions for a film to be a comedy, or a game to be a first-person shooter. But given how diverse games (or films) can be, it seems unlikely that everything we call a “first-person shooter” or a “comedy” could all share exactly the same attributes, or partake in the same necessary and sufficient conditions. Of course, there will always be borderline cases that defy rigid genre labels.

Indeed, this sort of rigid classificatory thinking about necessary and sufficient conditions is what Wittgenstein famously cautioned against in his remarks on the category of “games”:

Consider, for example, the activities that we call “games”. I mean board-games, card-games, ball-games, athletic games, and so on. What is common to them all? - Don’t say: “They must have something in common, or they would not be called ‘games’” - but look and see whether there is anything common to all. - For if you look at them, you won’t see something that is common to all, but similarities, affinities, and a whole series of them at that. To repeat: don’t think, but look! - Look, for example, at board-games, with their various affinities. Now pass to card-games; here you find many correspondences with the first group, but many common features drop out, and others appear. When we pass next to ball-games, much that is common is retained, but much is lost. - Are they all ‘entertaining’? Compare chess with noughts and crosses. Or is there always winning and losing, or competition between players? Think of patience. In ball-games, there is winning and losing; but when a child throws his ball at the wall and catches it again, this feature has disappeared. Look at the parts played by skill and luck, and at the difference between skill in chess and skill in tennis. Think now of singing and dancing games; here we have the element of entertainment, but how many other characteristic features have disappeared! And we can go through the many, many other groups of games in the same way, can see how similarities crop up and disappear.

And the upshot of these considerations is: we see a complicated network of similarities overlapping and criss-crossing: similarities in the large and in the small. (Wittgenstein, 2009, p. 36)

Wittgenstein goes on to call these overlapping similarities “family resemblances,” and suggests that language, the use of language and everyday understandings language’s meaning do not operate according to necessary and sufficient conditions. The meanings of words shift and change with time; new uses of words crop up and old ones fade away. Definitions of words are not strict, solid boundaries but rather porous, fluid and capable of change and adaption. So, which is true of genres? Are genres strict and rigid categories to which you either belong to or not, dependant on whether you share particular attributes? Or should we think of them as more fluid and open as elements in a complicated network of overlapping similarities or “family resemblances” à la Wittgenstein?

“Roguelike” video games present an interesting case study for thinking about these issues. “Roguelike” is a genre label, but a curious one insofar as it (1) does not directly convey any particular mechanics and (2) admits a degree of flexibility since it indicates that a game belonging to this category is like another particular game (Rogue). This is also particularly interesting because a very public attempt to define this term has played out in the roguelike fan community for years. Perhaps the most famous moment in this process involved a controversial list of particular game mechanics called the “Berlin Interpretation,” explicitly conceived as an attempt to define “roguelike.” “The purpose of the definition is for the roguelike community to better understand what the community is studying.” (Rogue Basin, 2008) In one sense, the Berlin Interpretation reads like an attempt to specify the necessary and sufficient conditions that a game must meet in order to be a “roguelike.” In another sense, however, the list is conceptualized in a manner more suggestive of Wittgenstein’s philosophy of family resemblances, since it takes five examples of games which exemplify the category and then extrapolates a definition of the category from the family resemblances between its members. In what follows, I argue that the Berlin Interpretation is actually representative of prototype theory in cognitive science, a theory in part inspired by Wittgenstein’s remarks about family resemblances. Examining the Interpretation and its subsequent criticism allows an evaluation of prototype theory as a means of approaching not just the roguelike genre but also video game genres more generally. But first, I need to clarify what prototype theory is.

Prototype Theory

Inspired by Eleanor Rosch’s work Natural Categories (1973), prototype theory is a theory of categorization in cognitive science that seeks to explain how the mind understands, makes and orders categories. In a separate article, Rosch cites Wittgenstein’s Investigations-era philosophy as having “radical implications for categories” because his reflections about family resemblance lead to certain conclusions about what categorization is and how it works (Rosch, 1987, p. 153). It may be tempting to think (and the earlier Wittgenstein of the Tractatus may have thought) that categories function as clear, solid boundaries picked out by language. For the early Wittgenstein, “language is necessarily a description of the world and words and propositions get their meaning through the objects to which they refer” (Ibid.). Language functions by describing the logical form of the world, a logical form with which language necessarily correlates. The words in language get their meaning by logically connecting to and picking out a particular thing in the world. The word “table” necessarily connects to the object “table” and picks out every instance of a table in exactly the same way, to the same degree, via the same mechanism. By combining words together, which each necessarily pick out their object in the same way, language gives us a clear, logical “picture” of how things stand in the world. On this view, the words in language explicitly connect to the particular object to which they refer. The generic category of “first-person shooter,” for example, clearly picks out every first-person shooter in the same way, and each example that it refers to is an equally legitimate member of that category, belonging to it just as legitimately as the others. Categories are rigidly defined groups with clear boundaries, groups to which a thing either does or does not belong, depending on its attributes or whether necessary and sufficient conditions are met.

Rosch argues that the later Wittgenstein totally reverses this view. Categories, and words in general, do not obtain their meaning from some mysterious referential relationship between language and the logical form of the world, but instead get their meaning from people, their actions, the contexts in which words are used and what they do in these contexts. As Rosch explains:

Rather than being referential, rather than giving us the “facts”, language is part of our actions, part of the most basic practices which make up our physical and social “forms of life.” This view has radical implications for categories. They are no longer objects of words or knowledge but are part of our delicately shifting forms of life. (1987, p. 153)

Examining how our language actually works in practise shows that our categories, and the meanings of words in general, are shifting and changing all the time. Meaning is not stable or fixed once-and-for-all, but rather constantly evolves in response to what we do with language. So our understanding of language must reflect the fact that our words do not actually have stable, fixed meanings, but are dependent on context and constantly evolve. This is as true for categories as it is with any other word. The consequences of this view for categories, as Rosch explains, are that “1. [categories] need not be precise; boundaries can be ill-defined. […] 2. They need not have anything in common, any common defining attributes […] [and] 3. All members need not be equally good members” (1987, p.153). Prototype theory can be thought of as an attempt to understand these insights from a cognitive-scientific perspective, to integrate them into our understanding of the human mind, and to conceive of human categorization explicitly along cognitive lines, as “the result of psychological principles” (Rosch, 1978, p. 27). But if our categories are fluid, flexible and evolving in the way that Rosch describes, then how do we understand categories, and how does categorization work?

Rosch argues that categories, and our understanding of categories, develop around salient “natural prototypes” of members of those categories. Natural prototypes are particularly clear, simple, easy to grasp, or perceptually salient members of a category which most clearly represent it. The quiz show Family Fortunes (US: Family Feud), for example, presents its contestants with a particular category and challenges them to guess what most people said when asked to name members of that category. If asked to name items of clothing, for instance, most people will say “shirt,” “dress,” “trousers,” “hat” (and so on) more readily than “cravat,” “dressing gown” or “tuxedo” because the former are clearer, easier examples of members of theategoryy “items of clothing.” In Rosch’s language, these items would be the “natural prototypes” of their categories, and our understanding of categories revolves around and is anchored in them. We ground our learning of and engagement with categories in natural prototypes, working outwards from them, and thinking about them based upon the examples that best represent what we are thinking about. If we are thinking about furniture, we are thinking about the stuff best exemplified by chairs, couches and tables.

Categories arise from psychological principles and social practises, including the way people communicate about and teach meanings of words to others. If we teach people the meaning of the word furniture, we might well point to tables, chairs and sofas and say, “that sort of thing is what we call furniture.” For Rosch, this is how we should think about categories in general, and the way the mind works with them. We take salient examples which easily suggest the idea of membership in that category, and we ground our engagement with categories in these examples. The Berlin Interpretation of roguelike video games, I claim, is an example of prototype theory in action, and will be used as a case study for evaluating the efficacy of prototype theory for understanding game genres.

The Origin of the Roguelike Genre

While a comprehensive history of the roguelike genre is out of scope for this essay, a few remarks about its history are necessary [3]. As Craddock writes:

Rogue is popularly credited as the progenitor of the roguelike, a subgenre of computer role-playing games known for procedurally generated levels, turn-based gameplay, items bearing randomly assigned properties, and irreversible death. While the genre bears its name, Rogue was not the first of its kind. Don Worth got there first. (Craddock, 2022, p. 1-2)

Don Worth was the creator of Beneath Apple Manor (or BAM), a game designed for the Apple II computer conceived as “a computer simulation of Dungeons and Dragons” (Craddock, 2022, p. 2). BAM was released in 1978, predating Rogue by two years. The objective of BAM was to navigate the dungeons below Apple Manor in search of the fabled golden apple, and it was one of the first games to make use of procedural generation, which randomly generated the layout and contents of the environment every playthrough. Though BAM shares some of the same characteristics as Rogue, the creators of both games claim not to have had any knowledge of the other, and roguelikes are not called BAM-likes. In fact, “few gamers remember BAM, [but] its influence should not go overlooked” (Craddock, 2022, p. 6). The reason for its relative obscurity compared to Rogue is that Rogue was designed for and released on computer systems used by colleges across America, while BAM was “released on a home computer that most users never took online” (Craddock, 2022, p. 6).

Designed by Michael Toy and Glenn Wichman, and later Ken Arnold, Rogue was released in 1980. Maria Garda correctly calls it a “timeless classic of computer games” whose place in the history of gaming is assured because of the sheer breadth of its influence on video game design (Garda, 2013, p. 3). Though BAM did some of the things Rogue does first, it is with Rogue that the genre’s formula first crystalizes; conventions that continue to be imitated and experimented with today. Rogue was initially made available as an executable file on University of California campuses, which became popular among their students. In 1983, it was released as part of the Berkeley Software Distribution (BSD) operating system. This directly contributed to Rogue’s growth in popularity. Like BAM, Rogue was also influenced by Dungeons and Dragons and made use of procedural generation, randomly generating the game environment and its contents anew on each playthrough. Rogue challenges the player to descend the Dungeons of Doom in search of the Amulet of Yendor, retrieve the amulet and escape. But the game implements “permadeath,” meaning if you die, you die. The player cannot save their progress and restart from their previous save; they must begin again from the beginning, every time.

One key innovation of Rogue is its use of ASCII graphics, in which the elements of the game world are represented by letters and punctuation marks: the player character is an @ icon, doors are + signs and so on. This enables it to be played on non-graphical computers. Remarkably, as Mark Johnson points out, “this style has continuedlargely unabated within the roguelike community, despite massive advances in computer graphics” (Johnson, 2015, p. 2). The goal of making Rogue, as Toy put it, was not only to create a highly replayable game, but “to be able to be surprised by our own game” (Craddock, 2022, p. 16). The fact that the game environment is different in every playthrough goes a long way toward assuring this element of surprise, coupled with the fact that the identity of consumable items is always initially unknown to the player. That “blue potion” you just picked up could heal you or just as easily set you on fire, so you have to either roll the dice and drink it, or find some other way to identify it.

It was not long before the college students who made Rogue popular began to crave other games like it. But Rogue’s creators did not release its source code, and it would not become public until 1986, when it was added to the BSD operating system. Therefore, Rogue inspired imitators who attempted to replicate the basic formula, with added improvements and twists. This led to the creation of Hack (1982), which was basically a remake of Rogue that added shops and expanded the number of in-game monsters, items and spells. Not long after came Moria (1983), inspired by the writings of J.R.R. Tolkein, in which players braved the mines of Moria to kill the Balrog dwelling in the mine’s lowest levels. Though not much played anymore, Hack and Moria inspired people to create new games based on their software, most notably NetHack (1987) and Angband (1990). These two early iterations on the Rogue format are two of the most popular and most historically significant roguelikes, and two members of what the Berlin Interpretation deem as “canon” for the roguelike genre. They also happen to be two of the oldest video games in existence which are still being developed and added to, over three decades later. Both of these games have numerous variants and other forks that grew out of their source codes -- as did Rogue once its source code was made public. It is from out of this creative context -- of imitation, borrowing and experimentation with Rogue’s formula -- that the roguelike genre was born.

The Origin of the “Roguelike” Term

For a while, however, there was no established name for these types of games. Santiago Zapata, in a well-researched talk On the Historical Origin of the ‘Roguelike’ Term (2017), traces the origin of the term to Usenet discussion groups in July 1993, where an attempt was made to organize games that resembled Rogue under a single umbrella term. At the time, this primarily meant Rogue, Hack, NetHack, Moria and Angband:

the efforts to define this hierarchical relationship, intended mainly to facilitate discussion, ended up serving the purpose of creating a community of niche developers and players who through the constant usage of the term ended up giving it an evolving, collective meaning that was relatively stable until the 2010’s decade, with the growth of the indie games scene and diffusion of the more action focused “Roguelike” labelled games. (Zapata, 2017)

This collective effort was meant to create a concept that describes the types of games these groups of players were interested in. The community emerging around these games facilitated the creation of new online spaces on Usenet where players could discuss them. Initially, the term “rogue-type” was proposed, as were several others, including “roguish,” “dungeon-type,” “hacklike,” “single-use computer role playing game” and “ASCII-Dungeon.” These were all eventually superseded by the term “roguelike” (Zapata, 2017).

Zapata shows what “roguelike” meant, outlines the debates around its origins and discusses how the community judged its suitability. Numerous participants in the Usenet discussions disliked the term, arguing that it was too vague and required too much knowledge of Rogue. Moreover, these participants questioned the idea that the genre should be named after only one game. Some even claimed that the games in question had already “diverged too far from each other to be unified into one hierarchy” (Zapata, 2017). As another Usenet user said, “try playing NetHack like it was Moria and see how far you get: they are two separate, even if related, games. The only real connection is their Rogue ancestry, a connection that has no relevance from a player’s point of view” (Zapata, 2017). Additional complications arose from the fact that, initially, the roguelike’s main characteristics were proposed as being “character-based” (meaning having ASCII graphics) and “highly portable” -- characteristics which neither explains what kinds of games they are, nor what they are like to play. The potential name for roguelikes continued to be fiercely debated, up to the point where Andrew Solovay, the person who had started and quasi-moderated these discussions, claimed that the debate “had seemed to reach its conclusions; I don’t think many people will change their minds at this point, and people are mainly repeating the same points” (Zapata, 2017). Usenet users who were fans of what would come to be known as roguelike games participated in a series of votes on creating new Usenet spaces specifically dedicated to roguelikes, which included the word “roguelike” in their titles. These votes were accompanied by persistent resistance from some users calling the whole exercise “pointless,” “insane,” “overly limiting,” “generally abhorrent,” and “silly” (Zapata, 2017). Some votes passed, and other votes held six months later about a similar issue passed even more decisively. Ever since these early discussions, the debate over the suitability and meaning of the term “roguelike” never dissipated, despite relative stability and consensus. Zapata draws two conclusions about these early attempts at specification: 1) “They make no mention at all of the game design features that we currently consider being inherent to the games of the genre” and 2) “They make a great emphasis on features which nowadays may seem distant and secondary except in the most traditional roguelike circles” (Zapata, 2017). As the genre has evolved and its games have grown in complexity, the terms of the roguelike debate and what is viewed as significant in it have changed dramatically since these early discussions.

“Roguelike” is the term that emerged out of these discussions, and it is the name that has stuck for the genre. However, in the early stages of searching for a name, it appeared difficult for the community around Rogue to identify what unites these games. “Roguelike” might have stuck, but the debate continued. As Luke Winkie attests, the debate has been intense, often fractious, and long-lasting: certain roguelike communities on Reddit, for example, have now “essentially banned discussion on ‘the nature of roguelikes’”, and the online roguelike community can often appear ‘defensive’, ‘confrontational’ and ‘obsessive’” (Winkie, 2021). Jose Ge, a roguelike developer and moderator of two roguelike Reddit pages, claims that,

The propensity to discuss this particular topic in so much detail […] likely stems from the type of players who enjoy traditional roguelikes in the first place, a pretty analytical, detail-oriented bunch, a huge portion of which are themselves programmers or at least work in IT fields and are big on breaking things down and categorizing them as part of a problem-solving process. (Winkie, 2021)

The intensity of these debates likely stems from several factors, including the complexity of the games themselves, the kinds of people that play them and the inherent vagueness of the term itself. A testament to the persistence of this discussion can be found in the fact that arguably the most famous event in its history did not occur until many years later, in 2008.

The Berlin Interpretation

Participants of the 2008 International Roguelike Development Conference in Berlin, which included both developers and players, took it upon themselves to address the intense debate which had become such a significant part of the genre’s history. This took the form of an attempt to define the term “roguelike,” one which has become notorious in roguelike fandom. Dubbed the “Berlin Interpretation,” this scheme lists 9 “high-value” and 6 “low-value” factors which can be used to determine how roguelike a game is (Rogue Basin) [4]. The Interpretation begins with some qualifications about its intended purpose and its choice of five games that best represent the concept it tries to define:

“Roguelike” refers to a genre, not merely “like-Rogue”. The genre is represented by its canon. The canon for Roguelikes is ADOM, Angband, Crawl, NetHack, and Rogue. This list can be used to determine how roguelike a game is. Missing some points does not mean the game is not a roguelike. Likewise, possessing some points does not mean the game is a roguelike. The purpose of the definition is for the roguelike community to better understand what the community is studying. It is not to place constraints on developers or games. (Rogue Basin, 2008)

As Garda notes, “the authors were not trying to be dogmatic and the notion of genre blurriness was included in their methodological approach” (Garda, 2013, p. 3). In a sense, the Berlin Interpretation reads as an attempt to specify a set of conditions for a game to be “roguelike,” but it does not understand the conditions it specifies as necessary and sufficient conditions, and it admits some flexibility in terms of the shared attributes -- a game it does not need to have all the attributes on the list to be a roguelike. However, labelling some factors “high-value,” [5] might reasonably lead to the impression that they are necessary, if not sufficient, conditions for a game to be roguelike. The Interpretation is an explicit attempt to specify what “roguelike” means and develop a list of factors which can be used to determine how roguelike a game is. As can already be seen, the Interpretation derives its definition from a list of five games that are unquestioningly accepted as roguelikes by the fan community. Each game functions as a natural prototype that strongly exemplifies the category “roguelike”’ and the Interpretation formulates a definition grounded on an identification of the salient resemblances between them.

The “high-value” factors are listed as follows, alongside the Interpretation’s original explanations for each item:

1. Random Environment Generation. The game world is randomly generated in a way that increases replayability. Appearance and placement of items is random. Appearance of monsters is fixed, their placement is random.

2. Permadeath. You are not expected to win the game with your first character. You start over from the first level when you die. (It is possible to save games but the save-file is deleted upon loading.)

3. Turn-Based. Each command corresponds to a single action/movement. The game is not sensitive to time, you can take your time to choose your action.

4. Grid-Based. The world is represented by a uniform grid of tiles. Monsters (and the player) take up one tile, regardless of size.

5. Non-Modal. Movement, battle and other actions take place in the same mode. Every action should be available at any point of the game. Violations to this are ADOM's overworld or Angband's and Crawl's shops.

6. Complexity. The game has enough complexity to allow several solutions to common goals. This is obtained by providing enough item/monster and item/item interactions and is strongly connected to having just one mode.

7. Resource Management. You have to manage your limited resources (e.g., food, healing potions) and find uses for the resources you receive.

8. Hack'n'Slash. Even though there can be much more to the game, killing lots of monsters is a very important part of a roguelike. The game is player-vs-world: there are no monster/monster relations (like enmities, or diplomacy).

9. Exploration and Discovery. The game requires careful exploration of the dungeon levels and discovery of the usage of unidentified items. This has to be done anew every time the player starts a new game.

(Rogue Basin, 2008)

This, I claim, is an example of prototype theory in action. The Berlin Interpretation explicitly states that it is not looking for a list of necessary and sufficient conditions, and instead begins from prototypical examples of roguelikes. In the collective imagination of roguelike fandom, these games are prototypical roguelikes, and they share mechanics and gameplay experience that are prototypical for the genre. The Berlin Interpretation is grounded in the shared understanding of the prototypicality of the games in question.

The shared factors of the games can be thought of as the “family resemblances” between roguelikes, but I should add the caveat that the Interpretation represents a very traditional type of roguelike game, and the genre has evolved a lot since then. For example, many games influenced by roguelikes retain turn-based combat but, as the popular subgenre of “roguelike deck-builders” shows, turns can take different forms, such as moves in a card game. The games in this subgenre hold many of the characteristics of the Berlin Interpretation, so should they still be called roguelikes? If the Interpretation is not meant to constrain developers, and missing some high-value factors does not disqualify a game from being a roguelike, how much flexibility should be allowed in calling a game a roguelike or not? What if turn-based combat is removed entirely, and replaced with real-time action combat, as many so-called “roguelite” games have done? What if permadeath were removed, or made more forgiving? Or introduced persistent progression (meta-progression) between runs? What if only very little procedural generation is used? These sorts of questions have animated roguelike fans and developers ever since the publication of the Berlin Interpretation, and even since the genre’s inception. Some factors -- like procedural generation, turn-based combat, being grid-based and permadeath -- seem less negotiable than others. In fact, out of all of the list’s points on the genre, these factors come closest to being “necessary” conditions for a game to be roguelike because they are the ones that occur in every traditional roguelike without exception. Many recent games influenced by roguelikes have effectively taken the Berlin Interpretation as a starting point, creatively remixing the listed game mechanics to produce different types of experience that nonetheless bear the stamp of their roguelike origins. It remains a point of contention in the roguelike community that many such games are often called “roguelikes” and listed as such on platforms like Steam, even though terms like “roguelite” are more commonly used for describing these types of game.

Screw the Berlin Interpretation?

Some of the more interesting debates in the roguelike community have concerned the validity and usefulness of the Berlin Interpretation itself. Roguelike fans and developers have debated whether this practise was misconceived, had negative effects, or that established the wrong type of conditions. Perhaps through examining some of these criticisms, prototype theory can be examined as a means of understanding video game genres.

One of the most well-known critiques of the Berlin Interpretation is probably roguelike developer Darren Grey’s blog post Screw the Berlin Interpretation! (2013), which highlights the negative effects the Interpretation could have on roguelikes’ design and within the roguelike community. It boldly claims that “definitions are about excluding things. They ultimately draw a line in the sand and say “if you stray beyond, you are forgotten” (Grey, 2013). This is arguably not a plausible claim: definitions are as much about including things as excluding them. The Berlin Interpretation explicitly includes all the games in its specified five-game “canon,” and presumably many others like them, but excludes others (perhaps “roguelites” like FTL, Spelunky, The Binding of Isaac). Grey takes issue with this exclusionary aspect, arguing that it places unnecessary constraints on developers, could lead to creative stagnation, and could cause great, more experimental games to be forgotten by the roguelike community [6]. While these are harmful effects, it is not clear that they are caused by the Berlin Interpretation. The roguelike community continues to be experimental, prolific, creative and even quite keen on games that the Berlin Interpretation seems to exclude. Furthermore, given Rosch’s prototype theory and Wittgenstein’s family resemblance view, it seems that the Interpretation is more on the Wittgensteinian side of categorization rather than a rigidly defined approach to categories; it is more open and flexible than Grey seems to allow. Perhaps the listing of “high” and “low” value factors gives a misleading impression that a set of conditions must fulfilled for a game to be a roguelike. But this list was drawn up via an examination of prototypical examples of roguelikes games and their most common shared characteristics. The very inclusion of a list of “low value” factors can be thought of as a concession that games do not need to include every factor in order to be a roguelike, and moreover, nowhere is it stated that a game needs to possess all the high value factors to be one either. The counterargument to this would be that procedural generation and permadeath are so foundational to the roguelike genre that not including them might immediately disqualify a game from being a roguelike. Therefore, they are the two closest factors in the Berin Interpretation’s list of factors to being necessary and sufficient conditions.

Another point Grey makes is that the Interpretation unjustly reduces incredibly complex games to an “ingredients list” of their core gameplay mechanics, which is akin to saying that a cake can be defined as “flour, butter, eggs, and sugar” (Grey, 2013). These ingredients are involved in making a cake, but to claim that the concept of cake is equivalent to them is patently false. This formulation does not get to the heart of what cake is, which is the more-than-the-sum-of-its-parts product produced when these ingredients are combined according to certain methods. This is what the Berlin Interpretation fails to capture about roguelikes. A roguelike game (perhaps any game) is what happens when certain mechanics are combined and experimented with in different ways by developers according to design choices, which in turn produces something more than the sum of its parts that cannot be adequately reduced to a list of its components.

This is arguably a stronger critical point about the Berlin Interpretation, and by extension of the application of prototype theory to video game genres. Even if the prototypes started with are unquestionably accepted as members of a category, that still leaves the problem of deciding which characteristics are salient or not. This is arguably a problem the Berlin Interpretation runs into: all the items it lists as a definition of “roguelike” are game mechanics. Like the cake example, it misses out the extra dimension produced by the combination and interaction of the ingredients.

The Berlin Interpretation misses out on with roguelikes: the surplus, the extra dimension produced by the interactions of their components as arranged by the designers. While the Interpretation manages to include games as disparate as Rogue and NetHack, it does so at the cost of saying very little, if anything, about the deeper factors that unite them beyond the mechanics they share. Ultimately, pointing out that Rogue shares mechanics with NetHack says little about what it is like to play either of them, or what really makes these two starkly different games worthy of being united under the same label. If there are deeper factors, or extra dimensions, that really unite these games, what are they? If the Berlin Interpretation misses the point, what is the point? If mechanics aren’t what we are looking for, what should we be looking for?

Alternatives to the Berlin Interpretation

I will now examine what I think are two of the most interesting types of response to these questions, which focus not on game mechanics, but on “experience” and “skills.” Grey seems to favour the first of these, claiming that what makes a roguelike a roguelike is not its mechanics, but the type of gameplay experience it engenders, which he expresses as follows:

In my view [the roguelike] is inherently replayable, capable of surprising the player on many playthroughs. It rewards cleverness and tactical thinking. It cannot simply be learnt by rote, but it can be mastered with experience. It emphasizes gameplay before aesthetics, concentrating on making the replayable experience fresh and engaging on each play. It’s unforgiving, but all the more rewarding when you perform it well, offering an honest sense of achievement and satisfaction. Much of this satisfaction comes from the internal knowledge of having done well at the game itself, rather than artificially constructed rewards. (Grey, 2013)

Here, Grey defines the roguelike not in terms of mechanics, but in terms of its gameplay experience. A roguelike offers a difficult, punishing, but rewarding experience that relies on tactical thinking which cannot be learnt by memory, it prioritizes deep gameplay at the expense of aesthetics, and it is inherently replayable. While this may not be a complete description of the gameplay experience of a roguelike, it nonetheless shows how “experience” is perhaps a more promising starting-point for defining a type of game than a list of mechanics. This can also be done by working from prototypical examples. For instance, a first-person shooter could have a procedurally generated environment, permadeath and several other factors canonized in the Berlin Interpretation. However, a description of the experience playing a first-person shooter compared to a roguelike would immediately differentiate the two.

A second alternative to the Berlin Interpretation’s focus on mechanics could be to focus on the skills a player is required to master to be able to play the game well, eloquently expressed by Reddit user aaron_ds in a comments thread about “roguelike definitions.” Their basic idea is that what differentiates a traditional roguelike from other types of game is the primary skill required to master the game, which is:

accurate threat assessment in tactical scenarios generated by systems. […] The player needs to learn this skill to win the game. This means that the primary means of progressing is by developing this skill in the player. This asks more of players than simply putting in time. The player is required to learn this skill to progress. (aaron_ds, 2020)

Roguelikes require players adapt to an ever-changing environment and accurately assess and respond to threats in tactical scenarios in this environment. A player cannot succeed in a traditional roguelike just by spending a lot of time on it. In my personal experience, I’ve found that when playing frantic twin-stick shooters like Binding of Isaac or Enter the Gungeon, my chance of success almost naturally increases over time because my muscle memory, response times (etc.) improve through repetitive practise. But making progress in classical roguelikes did not come so naturally at first because success in them does not rely on things like muscle memory or response time. In classical roguelikes, the player must learn an intellectual skill to succeed: how to accurately assess threat. This is partially what differentiates roguelikes from other types of game. The weakness of this view is that many types of non-roguelike games will require similar skills from the player -- strategy games are a clear example of this. While many roguelikes could accurately be referred to as strategy games, not all strategy games are roguelikes, even though strategy games usually require the player to be able to accurately assess threat.

Having identified a required skill which distinguishes roguelikes from other types of game, aaron_ds develops their critique of the Berlin Interpretation as follows:

I see the Berlin Interpretation as a list of second-order effects. Second-order effects can be easier to identify (ex. symptoms), but they are a consequence not a cause of something. Roguelikes having these properties doesn’t mean that having these properties makes something a roguelike (affirming the consequent). Being a game that primarily values threat assessment makes something closer to a roguelike than checking off all the Berlin Interpretation boxes. (aaron_ds, 2020)

aaron_ds seems to admit that even if a game values threat assessment, this can only make the game closer to being a roguelike rather than definitively a roguelike. This indicates a level of weakness in his theory that aligns with the criticism made earlier: it does not definitively show that a game that possesses this characteristic is a roguelike. In fact, he correctly identifies a case of affirming the consequent that relates to this: roguelikes having certain properties does not mean that having these properties makes a game a roguelike. However, threat assessment is a key aspect of the experience of playing a roguelike, and so rather than being a separate category on its own, perhaps this important skill could be included under a description of the experience of playing a roguelike. Many types of game can require the same skill while involving very different experiences, so a description of exactly how threat assessment features in and informs the playing of a roguelike could be a good place to start in formulating a definition of roguelike which is grounded in the experience of playing them.

A second important point in the above passage is that aaron_ds seems to indicate that adhering to or seeing value in the Berlin Interpretation makes what the philosopher Gilbert Ryle calls a “category mistake” by mistaking consequences for causes. According to aaron_ds, roguelike mechanics are a consequence of the desire of these games’ creators to deliver gameplay experiences which require accurate threat assessment. The desire to create a particular type of experience causes the deployment of mechanics, not the other way round. To posit that mechanics are the fundamental constituents of a roguelike is therefore to make a category mistake.

What these alternative approaches to the Berlin Interpretation show is that, while working from natural prototypical examples of the roguelike genre might make sense as a starting point for defining the genre, this approach still does not determine what the salient characteristics of roguelikes are. The Berlin Interpretation opts for shared game mechanics, but mechanics alone do not convey the extra dimension which occurs when these mechanics are combined according to certain methods. Experience is perhaps a better place to start, rather than thinking about what mechanics the prototypical cases all share. Perhaps we should be thinking about what the different cases share in terms of gameplay experience should be further considered from a player’s perspective. From there, a more general description of the gameplay experience of the genre could be extrapolated from the prototypes, which might well include a description of the skills necessary for playing the game.

Conclusion

The history of the roguelike genre is also the history of an ongoing, fan-led analysis of this genre which has questioned not just what “roguelike” means but also how the concept should be approached when figuring out what it means. I have shown, following Zapata, how the quest for a name for the roguelike genre originated in debates on internet message boards, which eventually led to its attempted codification in the Berlin Interpretation, and then subsequent debates that challenged the validity of the Interpretation. I analysed the Berlin Interpretation of the term “roguelike” in terms of prototype theory, which argues that human categorization works by anchoring its understanding of categories in the clearest, most salient, “prototypical” examples of members of that category. But as seen, the Berlin Interpretation is problematic, not because it begins with prototypes but because it arguably seeks after the wrong type of shared characteristic and makes a mistake about what type of factor (i.e., game mechanics) should be thought of as the salient feature of the roguelike genre. Fans have proposed and debated alternative avenues, two of the most important of which focus on experience and skills. Rather than listing shared mechanics, a better definition of roguelike might be one that focuses on the structural aspects of the gameplay experience, including references to the necessary skills required for gameplay success.

Ultimately, the categorization of video game genres is an open task, and video game genres do not always have sharp boundaries. Many games blend genres, experimenting with established conventions to push the boundaries of what is possible within the medium. Naming a genre “roguelike,” rather than “post-Rogue” or “neo-Rogue” and thereby indicating a level of similarity with a prototypical example, seems to acknowledge this complexity and the often-blurry nature of the boundaries between different types of video game (Garda, 2013). Prototype theory is a productive approach to analysing the categorization of games, but caution must be paid to exactly what type of things are thought of as the salient features of prototypes. The roguelike is a remarkable genre label because of the level of debate and scrutiny it has received in fan analysis. It is, however, far from the only example of prototype-based thinking in video game discourse. Up until the late 90s, first-person shooters were often referred to as “Doom clones” for their resemblance to Doom. Now, there are “Survivors-likes” named for their similarity to Vampire Survivors, and “Diablo-likes” for games resembling Diablo. Thinking via prototypes seems to be quite prevalent in video game categorization, and saying a game is “like” another game is a particularly easy way to communicate certain aspects of the gameplay experience. But as this article has shown, mechanics might not be the salient feature for establishing prototypes. The experience these games create in players through innovative manipulation, combination and mechanical interaction may provide a more solid foundation for understanding the roguelike genre.

Endnotes

[1] Despite being very niche, the roguelike genre has blossomed into a creative and diverse type of video game, one that has been very influential on game design. One way this can be observed is in the recent wave of games now commonly referred to as “roguelites,” a term which indicates the clear influence of roguelike games while also acknowledging their significant departures from the initial roguelike formula. Another way the roguelike’s influence can be observed is in the prolific use of procedural generation throughout video gaming today.

[2] I will not wade into the debate over whether video games can be considered art, but will take it as assumed that they are.

[3] In the currently available literature, David Craddock’s Dungeon Hacks is unmatched as a history of roguelikes, and John Harris’ Exploring Roguelike Games is also a very valuable resource.

[4] The creators of the Berlin Interpretation used a list of 13 “roguelikeness factors” proposed by Slash and The Temple of the Roguelike as a starting point for the discussion. Slash lists “high-value,” “middle-value” and “low-value” factors. Many of his points are included in the Berlin Interpretation but phrased differently, while others were removed and new ones added. (Temple of the Roguelike). 

[5] It must be noted that the roguelike genre has evolved massively since the Berlin Interpretation, and the prominent games of the “roguelite” movement experiment with the original formula so much that several of even the “high-value” factors of the Interpretation are often absent in them. Covering this point in detail is beyond the scope of this article, and I will largely focus on the earlier, traditional roguelikes.

[6] The comments on Grey’s Reddit post take issue with several of his points, but especially this one, pointing out that many games that do not fit the Berlin Interpretation are actually very popular among roguelike players (Grey).

References

Brazerman, C. (2010). Series editors’ preface. In A. Bawarshi & M. Reiff (Eds.), Genre: An introduction to history, theory, research, and pedagogy. Parlor Press.

Craddock, D. (2022). Dungeon hacks: How NetHack, Angband, and other roguelikes changed the course of video games. CRC Press.

Garda, M. (2013). Neo-Rogue and the essence of roguelikeness. Homo Ludens, 1(5).

Grey, D. (2013). Screw the Berlin Interpretation! Games of Grey [Blog]. http://www.gamesofgrey.com/blog/?p=403

Johnson, M. (2015). The use of ASCII graphics in roguelikes: Aesthetic nostalgia and semiotic difference. Games and Culture, 1-21.

Oxford English Dictionary. (2024). “Genre.” OED.com. https://www.oed.com/search/dictionary/?scope=Entries&q=genre

Reddit user aaron_ds. (2020). Comment on roguelike definitions. Reddit. https://old.reddit.com/r/TraditionalRoguelikes/comments/eq1rmz/roguelike_definitions/ff1sly1/

Rogue Basin. (2022). “The Berlin Interpretation.”http://www.roguebasin.com/index.php?title=Berlin_Interpretation

Rosch, E. (1973). Natural categories. Cognitive Psychology, 4, 328-350.

Rosch, E. (1987). Wittgenstein and categorization research in cognitive science. In M. Chapman, et al. (Eds.), Meaning and the growth of understanding (p. 153). Springer-Verlag.

Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.), Cognition and categorization. Erlbaum.

Temple of the Roguelike. (2008). Roguelikeness factors. Rogue Temple. https://blog.roguetemple.com/roguelike-definition/roguelikeness-factors/

Toy, M., Wichmann, G., Arnold, K., & Lane, J. (1980). Rogue. A.I. Design/Epyx.

Winkie, J. (2021, June 7). People that argue about the definition of roguelikes are annoying, but what if they’re right? PC Gamer. https://www.pcgamer.com/people-who-argue-about-the-definition-of-roguelikes-are-annoying-but-what-if-theyre-right/

Wittgenstein, L. (2009). Philosophical investigations (G. E. M. Anscombe, P. M. S. Hacker, & J. Schulte, Trans.). Wiley-Blackwell.

Wolf, M. (Ed.). (2002). The medium of the video game. University of Texas Press.

Worth, D. (1978). Beneath Apple Manor. The Software Factory/Quality Software.

Zapata, S. (2017). On the historical origin of the ‘Roguelike’ term. Roguelike Celebration. https://blog.slashie.net/on-the-historical-origin-of-the-roguelike-term/