Joris Dormans (2008)
This paper won Top Paper Award at the IADIS Gaming Conference 2008
This article is available as an audio article from IndustryBroadcast.com.
Realism remains a prominent topic in game design and industry research. Yet there is a strong academic case that games are anything but realistic. This paper frames realism in games in semiotic terms as iconic simulation, and argues that games can gain expressiveness when they move beyond the current focus on iconicity. In parallel to natural language, indexical and symbolic simulation and especially the clever configuration of indexical and symbolic simulative rules are proposed as an important framework through which the communicative and expressive power of games can be understood.
Realism still sells games, and therefore a lot of industry research is aimed at making games more realistic. Realism features prominently in the "top ten hurdles facing game designers today" recently published on the website of the magazine Popular Science. All concern the accurate and realistic simulation of real-life phenomena. Getting water and fire effects right made that list, as did realistic movement, rendering human faces and artificial intelligence designed capture realistic behavior (Ward, Canor and Cary 2007).
On the other hand, in certain circles of game critics and scholars, it is in vogue to point out realism is not what games are about. Steven Poole's Trigger Happy deconstructs the supposed realism of games. He argues that most players play games because they allow them to do things they cannot do in reality. A thoroughly realistic game would require a player to undergo equally thorough training before he can even try to play. A game that is totally realistic seizes to be a game (Poole 2000: 77). He concludes: "videogames will become more interesting artistically if they abandon thoughts of recreating something that looks like the 'real' world and try instead to invent utterly novel ones that work in amazing but consistent ways" (Poole 2000: 240).
The sentiment that games are different from realistic and accurate simulations can already be found in the early work of Chris Crawford who states: "accuracy is the sine qua non of simulations; clarity the sine qua non of games. A simulation bears the same relationship to a game that a technical drawing bears to a painting. A game is not merely a small simulation lacking the degree of detail that a simulation possesses; a game deliberately suppresses detail to accentuate the broader message that the designer wishes to present. Where a simulation is detailed a game is stylised" (Crawford, 1982: 9).
More recently Jesper Juul points out that: "games are often stylized simulations; developed not just for fidelity to their source domain, but for aesthetic purposes. These are adaptations of elements of the real world. The simulation is oriented toward the perceived interesting aspects of soccer, tennis or being a criminal in a contemporary city" (Juul 2005: 172).
Despite the considerable influence of Crawford, Poole and Juul in academic circles, the arguments against realism in games are hardly picked up by game designers. Reasons for this are plenty. To name just one, realism is a very convenient selling point. After all, the realism of a game can be easily shown and compared with other games. However, in this paper I wish to explore an important neglect on the part of the academics: to my knowledge, nobody seems to answer in detail the question how non-realistic games look like. In what ways are artistic or abstracted games different to realistic games, and on what grounds could one claim that such games are 'better'? If game scholars expect game designers to move away from realism in games, we should at least provide them with a little more to go on.
As the title of the paper suggest, I understand realism in games as a form of 'iconic simulation'. This view of games combines insights from contemporary ludology with notions taken from linguistics and semiotics. I will discuss iconic simulations as can be found in games as well as examples of non-iconic simulation as also can be found in games. Iconic simulation typically have a high level of detail, although this seems to be a good way to create expressive games with a lot of depth and gameplay, the analogy of with language suggests that non-iconic simulation can reach very high level of expressiveness and articulation by harnessing the power of emergence in much simpler, non-iconic structures that consists of only a handful of building blocks. This is the point I will make in the fourth section.
If the whole narratology versus ludology debate has learned us something, it is that games constitute a new form of representation that is fundamentally different from static representation through non-interactive text, images and sound. According to Rune Klevjer, simulation is a form of procedural representation; simulation represents rules instead of events (2002). Gonzalo Frasca classifies simulation as an alternative to narrative or representation (2003: 223). Ian Bogost picks up on Frasca's work when he defines simulation as follows: "A simulation is a representation of a sources system via a less complex system that informs the user's understanding of the source system" (Bogost 2006: 98). This definition closely resembles the semiotic triparte model of the sign. In the light of the narratology versus ludology debate I find this an intersting observation, an observation worth pursuing.
The tripartite model of the sign was drafted by Charles S. Peirce. In this model a sign is connected to an object (that what the sign represents) and an interpretant, or the mental concept the sign invokes. This model of the sign is best known for the classification of signs into icons, indexes, and symbols. This classification is based on the nature of the relation between the sign and its object: when a sign resembles its object it is an icon, when the sign has an existential connection to its object it is an index, and when the connection is arbitrary it is a symbol (Kim 1996: 19-21). Figure 1 combines the Bogost's definition of simulation with Peirce's model of the sign.
Figure 1. A tripartite model of signs and simulation
The link between realism and iconicity in simulation should be obvious. We call a simulation realistic when the simulation (as a system) closely resembles the source system; we call a simulation realistic when it is iconic. From this analogy two other forms of simulation suggest itself: indexical and symbolic simulation. If games are ultimately not realistic, indexical and symbolic simulation might be interesting notions to help us understand games better. As we will see in the next section constructions that we could call indexical or symbolic have been used in games to great effect.
If we push the analogy between signs and simulation one step further, I like to point out an interesting discrepancy between the current focus on iconic games and the highly symbolic nature of language. Natural language is by its nature very abstract. This notion can be traced back a long time. It was already apparent in the works of seventeenth century philosopher John Locke who observed: "Men making abstract Ideas, and settling them in their Minds with names annexed to them, do thereby enable themselves to consider Things, and discourse them, as it were in bundles, for the easier and readier improvement, and communication of their Knowledge, which would advance but slowly were their words and thoughts confined only to Particulars" (Locke 1975: 420).
It is on similar grounds that, roughly a century later, Edmund Burke attaches greater aesthetic and evocative power to poetry, which through words, "obscures" its image, than to the realistic paintings of his age (1990: 55). These days the development of abstract art has changed all this and has increased the expressive power of the image dramatically, as is exemplified by the names used by art history to identify particular genres: Impressionism, Expressionism, Abstract Expressionism, and etcetera.
Ferdinand de Saussure identifies the arbitrary character of the linguistic signs as its principal characteristic. Although he does not rule out the possibility of non-arbitrary signs, he argues that in human languages most signs are arbitrarily linked to their meaning. There are usually no characteristics of what we are referring to that are connected to the words we use (Saussure 1983: 67-69). In other words, Saussure rules out the possibility of many linguistic icons and indexes. For Saussure too, it is the human faculty to construct a "system of distinct signs corresponding to distinct ideas" that makes language possible (ibid. 10). Through the human capability to take abstract meaning but handle them in bundles as they were particular things human understanding is efficiently facilitated and taken beyond the level of the particular into the realm of general knowledge. In other words, there is more expressive power in abstract, non-iconic presentation.
Ian Bogost's definition of simulation quoted above is not complete. Bogost emphasizes the subjectivity inherit to simulation: "A simulation is a representation of a source system via a less complex system that informs the user's understanding of the source system in a subjective way" (Bogost 2006: 98). In a simulation a system is represented through another system and the choices made in the construction of the second system reflect the values of its creator: "no simulation can escape some ideological context" (Bogost 2006: 99). This subjectivity can be partly attributed to Bogost's insistence on the fact that with simulation the simulating system is by necessity less complex that its source system. A simulating system always deviates from its source system and the choices made in that deviation reflect the understanding and/or ideology of the person that set up the simulation. What exactly Bogost means with 'less complex' is not made explicit. I choose to understand it here as consisting of more abstract parts that are fewer in number, and which are therefore easier to apprehend. The parts themselves are usually larger than their counterparts in the source system. For example the parts that make up a simulated weather system are much larger (and more abstracted) units that the actual air-molecules that make up real weather. This makes the simulation more convenient to handle, or to paraphrase Locke: it enables us to consider the multitude of parts of a simulated system in bundles for easier and readier understanding, and for easier and readier communication and improvement of that understanding. 
Thus, there always exists a gap between a simulated system and its simulation, and that gap always renders the simulation subjective to a lesser or greater extend. However, this subjectivity is the price we pay for the convenience and enhanced understanding that simulations allow. The question whether or not this price is too high, is beyond the scope of this paper to explore in detail, although I am inclined to say that it is not. Especially when one realizes simulations are by necessity subjective, and one should approach them critically because of it. In the end, I think the gains in expressive power outweigh the loss in resemblance to particular instances.
When one considers a simulation as essentially subjective, it is worth noting that any claim to realism becomes an ideological maneuver in itself. For example, the high level of verisimilitude in America's Army can be read as the rhetoric claim that its apparent realism and correctness in visual representation can be extended into the ideological domain: the game has its physics right, so its ethical claims must be realistic too (Bogost 2007: 78). On the other hand, in commercial entertainment games realism is often rendered as a special effect. In these games realism and authenticity becomes a spectacle designed to impress and to be appreciated by the audience. Realism, with its high poly-count, plasma effects and particle engines, is fore-grounded and hyper-real. Or to use the words Geoff King used to describe a very similar phenomenon in blockbuster films: it is "the hyperrealistic spectacle-of-authenticity rather than authenticity itself" (2000: 136).
In moving beyond iconic simulation, indexical and symbolic simulations are obvious points of departure. Examples of both such constructions can be found in games. I have chosen two, one of each, to discuss below.
In the classic game Super Mario Brothers one way of disposing enemies is by jumping on top of them. Although the precise implementation differs from enemy to enemy, and certainly does not work against all enemies, it is a feature that is frequent and remains constant throughout the game and the series it belongs to. I am not the first person to observe that this method is a little odd, to say the least. It is has, however, become a convention within platform games that is instantly recognizable to gamers, and ties in with that genre's defining action of jumping from platform to platform.
I argue that it is an example of symbolic simulation. The connection between jumping on top of something and defeating something in real life is arbitrary and based on convention (which parallels the definition of a symbolic sign in language). Although one could argue that there is a connection between the two as some creatures (all manner of bugs) can be squashed, I still have to encounter the first insect that is lethal when bumped into, but not when stepped upon. What is more, I think that this method of fighting in Super Mario Brothers is motivated by the genre of the game and its most prominent action: jumping, and therefore the link between the simulation and what is simulated is both arbitrary and conventional.
There is, however, a resemblance between the skills needed to defeat enemies in Super Mario Brothers. and in real life. In the game it requires timing and accuracy, which are among the skills involved in real fighting. The point is, the simple representation in the game allows us to do more than to hone and train those skills. The simple metaphor of jumping on top of enemies is easy to grasp by the player, but the game then goes on by inviting the player to experiment and develop strategies. In most platform games each level ends with a 'boss' enemy which is typically set-up to test the validity of player's strategy. It is the ultimate test for the player to demonstrate she understands and has mastered the simulation, and is able to combine different moves. These lessons carry over to situations beyond the game. The mentality of the people that have learned these lessons is excellently described by John Beck and Mitchell Wade: they know that solutions will eventually present themselves, and they have mastered a trial and error approach to many problems in life (2004: 11-14).
The power and pleasure of interacting with a symbolic simulation does not emerge from its resemblance to a system in the real world, it comes from a resemblance of that system to cognitive schemata we build up to represent systems in the real world. The arbitrary nature of symbolic simulation does not hinder understanding of the source system. On the contrary, it invites us to test and verify our understanding. Its internal coherence invites us to actively anticipate and explore its consequences, and if the symbols are chosen well the knowledge we gain is conveniently bundled in abstracted packages, and has application beyond the particularity of the simulated system.
Figure 2. Inventory screen from Diablo
The second example is the 'inventory system' that first occurred in Diablo and has since featured in many other games (see figure 2). It inspired Warren Spector, developer of Deus Ex, into saying that: "Diablo got Inventory right. There's no sense messing with something that works..."  For quite some years now, many computer games have included an 'inventory': the game allows the main character to pick up objects and carry them around. The player can manage these objects in the game's inventory screen. Most games restrict the number of objects the character can carry in some way. There might be a fixed number of objects the character can pick up, or all the game objects might have a weight value attached to it and the character can only carry objects up to a particular load. Sometimes the number depends on the number of containers the character is carrying, which can lead to strange situations when containers can be stored within other containers.
Diablo's inventory system takes size as it main restricting factor. Each item has takes up a number of inventory 'slots', the available slots are limited and organized in a grid. An item may take up 1x1, 2x2 or 1x4 slots for example. Depending on the available room in the inventory an object can be picked up or not. Figure 3 is a screenshot of Diablo's inventory screen. The lower portion of the screen is the grid that allows the player to store objects in. The upper half of the screen is dedicated to the objects the main character is currently carrying or wearing.
I argue that this is an example of iconic representation in games. The main restricting factors for somebody to carry objects in real life (shape, size and weight) are represented by very simple two-dimensional shapes. These shapes and their relative size can be said to be existentially connected to the size and weight of their simulated counterparts. Therefore the simulation qualifies as an indexical construction as it is parallel to indexical signs in which the relation between the sign and its object is also based on a existential connection (rather than resemblance or arbitrary convention).
The number of games that have copied this system in one form or another is testimony to the quality of this construction. The internal rules and constraints are immediately apparent (not in the least because they are tailored towards visual representation on a screen). The management problems the system gives rise to are very much like those in real life. The system even allows players to make an inefficient mess of their inventory, teaching them something about the need to organize themselves . Again, the simplicity of the system makes it an effective interactive representation on the one hand and gives it expressive power that stretches beyond the particularities of its simulation on the other.
The expressive power of natural language does not reside in the abstract nature of most of its signs only. On similar grounds, the design of expressive, non-iconic simulation involves more than coming up with many abstracted game systems. A lot of expressive power of language resides in the combination of words, not in the number of quality words. Noam Chomsky observed that language allows speakers to make infinite use of finite means: the number of words we have may be limited (and is vastly outnumbered by particular things in reality), the number of combinations we can make with them is infinite (Chomsky 1972: 17). This characteristic of language is often called discrete infinity. In language discrete infinity is very powerful. Poetry is probably the best example of how a few simple words can be used to great effect.
Discrete infinity arises from a particular construction of the rules that govern the way elements can be combined. Chomsky shows that an infinite number of combinations can be created once the rules of the combination are recursive. Nested and recurrent constructions form the basis of his work on transformative generative grammar that has inspired much work within and outside the field of computer linguistics. For Chomsky and his successors the complexity of language does not really emerge from a large number of words, rather from a relative simple set of rules that govern their combination. Finding and understanding the rules of language (which are not necessarily the same as the grammatical rules we learn in school) is a valid practice in linguistics. Chomsky's framework for transformative generative grammar has been applied in different fields of cultural analysis, story-analysis in particular (see for example Baddeley 1997: 241-246). The premise of this application has always been the construction of sets of rules that govern the construction of stories based on a finite set of elements. Discrete infinity applied to games dictates that we should never be able to design the set of all possible outcomes and combinations. Instead we should design the rules that generate the infinite set (cf. Salen & Zimmerman 2003: 158-159). Games that focus on realism tend to include too much detail. These games neglect the expressive beauty and strength of few simple gameplay mechanics. After all, many are still captivated by the ancient game of Go, even though the number of rules and mechanics of that game are very limited. If we are to understand games as an expressive form we have to look beyond rules and mechanics in isolation and look how they can be combined into larger structures, and how meaningful play can emerge from these structures.
Jumping on top of your enemies in order to defeat them, as discussed above, is a great example of such emerging gameplay. On the one hand, the mechanic is relatively simple. On the other hand, it is easily combined with particular level layouts and characteristics of enemy characters to form puzzles, obstacles and gameplay strategies that are far more articulate than inspection of these simple mechanics alone would suggest. Flying enemies can be used to cross gaps between platforms that would otherwise be impossible to traverse. The turtles from Super Mario Brothers that are launched to one side and bounce around when Mario jumps on them, can be both be a threat to the player as a way of disposing other enemies, often both at the same time.
These days, emergence of complex behavior from relatively simple elements is an important aspect of many fields of research in the domains of mathematics, physics and social sciences. In the research of games, too, emergence is becoming an increasingly important notion. From the computational side, emergence is an important technique used in anything from development of AI to the realistic rendering of water and fire. For Penelope Sweetser the disadvantageous loss of creative control in a system that is set up for emergence is outweighed by the more consistent and intuitive player interactions such systems allow (Sweetser 2006: 14). Likewise, game designer Harvey Smith argues that attempting to design a totally controlled game environment that allows rich interaction is no longer economically viable, as the sheer amount of detail cannot be efficiently produced manually (Smith 2001).
From the techniques currently available to study and express emergence, cellular automata stand out as particular promising. On the one hand, their grid-based structure and computational complexity is well suited to games on a technical level (Sweetser 2006: 158), on the other hand, the general principles that can be distilled from the studies of cellular automata seem to be well suited to game design on a more abstract level: for example, the successful application of object oriented or agent based design have carried over from old MUDs and MOOs to current MMORPGs, and is directly responsible for the complexity and scalability these game allow. Likewise, communication and feedback within the game system are great tools to understand and tweak game balance (cf. Adams & Rollings 2007: 384-389). These aspects can be directly derived from an intermediate level of activity, simple, local rules and (long-range) communication between cells that propel the most interesting behavior in cellular automata (see Wolfram 2002: 76, 106 & 252).
However, in some ways, computer games seem to be moving against the trend of emergence. Jesper Juul differentiates 'games of progression' from 'games of emergence' as historical newer category associated with computer games. While in games of emergence dynamic play emerges from a large possible combinations and strategies, games of progression control the player's progression through the game. In a game of progression the player effectively has to perform the moves the designer has laid out beforehand. The rise of computer games, and adventure games in particular, has made games of progression possible, as without a computer the amount of data and the number of special case rules would have become unwieldy (Juul 2005: 5). Games of emergence are typically classic board games and strategy games with a lot of units and a high level of connectedness (ibid. 81-82).
I regard this trend as a form of regression; although an understandable form of regression as game designers are struggling to cope with the loss of authorial control over their creations. We come from a culture where authorship is deemed very important. The hand of the master is one, if not the most important, aspect of what sets great pieces of art apart from other works. It is difficult and perhaps impossible to accept that works (partly) generated by the computer could possible possess the same quality (cf. Aarseth 1997: 129-131). From the perspective of communication it is not straightforward just how the "more approximate control" emergent systems allow (Sweetser 2006: 159) translate to effective communication. Hence, it is not strange that with the development of an essentially new medium its authors hesitate to plunge themselves in uncharted territory, and try to stick to time-tested practices instead. Striving for realism and controlling the interaction as much as possible seem to be the easiest first steps in finding out what games can do. After all, it took painting many centuries to develop an aesthetic that moved beyond increasingly sophisticated techniques to capture reality towards more abstract and expressive forms that invite many different interpretations from its audience. 
I would say that many games would do well to strive for non-iconic, discrete infinity rather that detailed realism. Not only is this economically more feasible, it is also more interesting artistically and allows for more effective communication. The Legend of Zelda series is a great example of gameplay design in which only a handful of game objects and associated mechanics are combined in many interesting challenges. The value of each of these mechanics does not stem from its power to represent some sort of realistic aspect of adventuring through a dungeon, but from potential combination with other mechanics. The exploration challenges, which the series is famous for, are almost always the result of combinations of simple, reusable gameplay mechanics. Players appreciate this sort of structures as these have the advantage of being inheritably coherent. And as has been pointed out before, coherence is a strong contributing factor to gameplay (Poole 200: 64-66). One can even argue that the appreciation of such structures is in its essence an aesthetic appreciation. It is the appreciation of the craftsmanship of the game designer, it forces to pay attention to the way the game was constructed and how it is structured (cf. Ryan 2001: 176). The meaning that emerges from these games is not necessarily less detailed or less valuable than games that aim for detailed and realistic simulation. On the contrary, as the challenges of exploration in a Zelda game are more abstract, the skills and knowledge the game addresses are more generic; the message of a game that is less iconic is much better applicable outside the particular settings of the game. This is especially useful when one wants to express something through a game that has value beyond the game and its immediate premise.
The power of stylized simulation, such as games, lies not in its power to accurately model a source system, rather in its expressive potential by using indexical and symbolic constructions. These constructions, especially when combined in clever ways, facilitate a more generalized and abstracted understanding of the system that is being simulated. Knowledge becomes more useful as its application becomes more general. It stands to reason that abstract knowledge is best communicated with general and abstract means. Natural language has been a powerful tool to represent knowledge, especially because in its very nature, natural language is abstract. If games want to be vehicles for culture or knowledge, they need to start reflect the nature of culture and knowledge better. They need to develop more articulate, indexical and symbolic structures of expression.
Although it is hard to estimate whether the potential of iconic representation has been fully explored, it is clear that much more progress can be made by developing indexical and symbolic building blocks for simulation, and, more importantly, investigate the effectiveness of particular configurations of such building blocks. After all, just as it is the art of the skilled orator to find appropriate and simple words to construct articulate arguments, it is the craft of the game designer to create complex systems from appropriate and simple elements. There is no art in creating complex simulation with equally complex (or worse, more complex) means. Yet this seems to be what a lot of developers aim for.
 For Bogost a simulation is always less complex than its source system, although I agree that for games such simulations that are more complex are the (less interesting) exceptions, I do not wish to rule out the possibilty altogether.
 Quoted on Stephan T. Lavavej's Deus Ex webpage. Available at
 Although it is not always the best gameplay decision to burden the player with the upkeep of his inventory.
 The history of art and the development of an abstract esthetic are more complex than this statement suggests. The point here is not to make a caricature of art history, rather to point out that development of many media seem to progress from a focus on realism to a larger diversity of forms.
I would like to thank Eleonore ten Thij, Swen Stoop and my promotor Remko Scha for reading and commenting on earlier drafts of this paper. I am grateful to the Institute of Information Technology and Computer Science of Hogeschool van Amsterdam for supporting this research.