August 12, 2005

Clarifying Ergodic and Cybertext

by Noah Wardrip-Fruin · , 5:36 pm

Given the enormous influence of Espen Aarseth’s Cybertext: Perspectives on Ergodic Literature on studies of elit, it’s no surprise that people new to the field have asked me, from time to time, to clarify for them what Aarseth’s neologisms “ergodic” and “cybertext” mean. I’ve been happily supplying people with my understandings of what the terms mean, and have only in the last week or so begun to realize that I was probably wrong in my explanation — every time.

I’ve been telling people some variation of this: ergodic literature requires the reader to undertake “non-trivial” effort in order to traverse the text, and cybertext is the kind of text one reads ergodically. Two sides of the same coin. And I’d point them to this paragraph (p. 1-2):

The concept of cybertext focuses on the mechanical organization of the text, by positing the intricacies of the medium as an integral part of the literary exchange. However, it also centers attention on the consumer, or user, of the text, as a more integrated figure than even reader-response theorists would claim. The performance of their reader takes place all in the head, while the user of cybertext also performs in an extranoematic sense. During the cybertextual process, the user will have effectuated a semiotic sequence, and this selective movement is a work of physical construction that various concepts of “reading” do not account for. This phenomenon I call ergodic, using a term appropriated from physics that derives from the Greek words ergon and hodos, meaning “work” and “path.” In ergodic literature, nontrivial effort is required to allow the reader to traverse the text. If ergodic literature is to make sense as a concept, there must also be nonergodic literature, where the effort to traverse the text is trivial, with no extranoematic responsibilities placed on the reader except (for example) eye movement and the periodic or arbitrary turning of pages.

So far so good. Reading a little further in chapter 1, those I referred to the paragraph above might have found a section (pages 9-13) giving some examples of ergodic literature. These include: wall inscriptions of the temples in ancient Egypt that are connected two-dimensionally (on one wall) or three dimensionally (from wall to wall or room to room); the I Ching; Apollinaire’s “calligrammes” in which the words of the poem “are spread out in several directions to form a picture on the page, with no clear sequence in which to be read”; Ayn Rand’s play Night of January 16th, in which members of the audience form a jury that chooses one of two endings; Marc Saporta’s Composition No. 1, Roman, which is a novel with shuffleable pages; Raymond Queneau’s One Hundred Thousand Million Poems; B. S. Johnson’s The Unfortunates; Milorad Pavic’s Landscape Painted with Tea; Joseph Weizenbaum’s Eliza; Will Crowther and Don Woods’s Adventure; James Meehan’s Tale-spin; William Chamberlain and Thomas Etter’s Racter; Michael Joyce’s Afternoon: A Story; Roy Trubshaw and Richard Bartle’s Multi-User Dungeon (aka MUD1); and James Aspnes’s TinyMUD.

Of this list, the only immediately problematic entry is Meehan’s Tale-spin — which is a program for generating simple, linear stories. It’s hard to understand how traversing the output from Tale-spin requires nontrivial effort. All it requires is eye movement, which is explicitly listed (in the paragraph quoted above) as part of what one does for nonergodic literature. Speaking of which, eye movement appears to be the only extranoematic responsibility for the reader of Apollinaire’s “calligrammes” — what makes them ergodic? Rereading this recently, I started to feel confusion rising. But I figured it was only a list, and I thought I understood the basic idea of ergodic literature and cybertext, so I let it pass.

Or, I let it pass until page 75, when I came upon this passage:

I suggest the term cybertext for texts that involve calculation in their production of scriptons. This criterion corresponds nicely to all the texts in the west half of figures 3.2, 3.3, and 3.4, while in the east it applies to Afternoon, which is not really a pure hypertext, since some of its links are conditional. The concept of cybertext is therefore highly relevant to the interpretation of our analysis, since it almost perfectly follows the division established by the main axis.

At this point I was confused more than I could dismiss, and not because of the odd definition of hypertext that Aarseth was using. Let me explain… This is the list of works in the west half of 3.2, plus Afternoon: Robin Waterfield and Wilfred Davies’s The Money Spider, a path-choosing game book; Will Crowther and Don Woods’s Adventure; Mark Smith and Jamie Thomson’s Falcon 5: The Dying Sun, a game book which “adds indeterminacy by having the player roll dice to decide between paths”; Joseph Weizenbaum’s Eliza; Trevor Hall’s adventure game Twin Kingdom Valley; the I Ching; William Chamberlain and Thomas Etter’s Racter; Roy Trubshaw and Richard Bartle’s Multi-User Dungeon; James Aspnes’s TinyMUD; John Cayley’s Book Unbound; James Meehan’s Tale-spin; Michael Joyce’s Afternoon: A Story. This is the list of works in the east half of 3.2, minus Afternoon: Stuart Moulthrop’s Victory Garden; Julio Cortázar’s Hopscotch; Allen S. Firstenberg’s Unending Adventure, a web-based forking text that accepted reader contributions; William Gibson’s encrypted poem Agrippa; Apollinaire’s Calligrammes; Vladimir Nabokov’s Pale Fire; Raymond Queneau’s One Hundred Thousand Million Poems; Herman Melville’s Moby Dick; Randi Strand’s artist’s book Norisbo; Marc Saporta’s Composition No. 1, Roman; and Jenny Holzer’s LED art installation I am Awake in the Place Where Women Die.

The two lists above contain 11 works that are also among the examples listed for ergodic literature in chapter 1: the I Ching; Apollinaire’s Calligrammes; Saporta’s Composition No. 1, Roman; Queneau’s One Hundred Thousand Million Poems; Joseph Weizenbaum’s Eliza; Crowther and Woods’s Adventure; Meehan’s Tale-spin; Chamberlain and Etter’s Racter; Joyce’s Afternoon: A Story; Trubshaw and Bartle’s Multi-User Dungeon; and Aspnes’s TinyMUD. Of this 11, eight appear on the “west half plus Afternoon” list, but three (Calligrammes, One Hundred Thousand Million Poems, and Composition No. 1, Roman) appear on the “east half minus Afternoon” list. This means that I was utterly incorrect when telling people that ergodic literature and cybertext were two sides of the same coin. And as I thought about it I realized something further — while I’d been telling people that “cybertext” was a category that included hypertext, Stuart Moulthrop’s well-known hypertext novel Victory Garden is explicitly excluded by Aarseth (appearing, as it does, on the east half of 3.2). Yet, at the same time, The Money Spider appears in the west (cybertext) list, though the amount of calculation involved in its path-choosing seems less than Victory Garden‘s.

In fact, in looking for the term of Aarseth’s meant to encompass all of hypertext, I was probably looking for “ergodic” rather than “cybertext.” My current guess is that ergodic literature includes literature involving any choice — including the choice of what link to click and the choice of which way to move one’s eye across a page or temple wall. On the other hand, cybertext literature is a rather different category, only including works that involve calculation in their production of scriptons (the texts read by the reader). But how this calculation is defined, precisely, is hard for me to understand — given that The Money Spider has it but Victory Garden and One Hundred Thousand Million Poems lack it. Perhaps, like the difference between Afternoon and Victory Garden, it has to do with conditional connections. This speculation is supported by the following passage I found on Demian’s Gamebook Web Page about the game book series of which The Money Spider is a part:

These mystery gamebooks cast the reader as a detective known as T.S. (short for “troubleshooter”). No dice are used during play, but each book includes a “web” on which numbers are written when certain clues turn up. If the right clues are found, things progress to a successful conclusion; otherwise, the player is taken off the case and the criminal escapes.

But, to me, this feels like a pretty provisional understanding.

Anyway, I write all this for two reasons:

Oddly enough, all that said, I don’t have the same confusion about textons and scriptons that Matt Kirschenbaum discusses. To me they’re pretty straightforward terms, referring to the elements of a work of literature (and not to other applications involved, the operating system, or other technical layers that aren’t part of the work). More relevant to Kirschenbaum’s questions is probably Aarseth’s first paragraph on page 176, which discusses the question of what we view as the “text” when looking at the tower of user/developer hierarchies that underlie any modern software use.

Finally, I’d like to return to my confusion about the presence of Tale-spin among the examples of ergodic literature. The only explanation I’ve been able to find for it is a rather different definition of “ergodic” found on page 94: “The adjective I propose for this function is ergodic, which implies a situation in which a chain of events (a path, a sequence of actions, etc.) has been produced by the nontrivial efforts of one or more individuals or mechanisms.” This is really quite different from saying, as chapter 1 does, “In ergodic literature, nontrivial effort is required to allow the reader to traverse the text.” If we can include nontrivial effort by mechanisms, then Tale-spin definitely fits. I’m tempted to start using this sentence from page 94 as my explanation of “ergodic” — but then I’d probably be running against the grain of almost everything in our field that uses the term, and I haven’t found further support in the text. In fact, I’d be running against the other major passage that defines the term, on page 179:

So what exactly is the difference between the ergodic and the nonergodic work of art? If we are to define this difference as a dichotomy (and such a definition may well end up serving the ideology it is trying to unmask), it would have to be located within the work rather than within the user. The ergodic work of art is one that in a material sense includes the rules for its own use, a work that has certain requirements built in that automatically distinguishes between successful and unsuccessful users. The usefulness of this definition is limited not so much by the concept of ergodics as by the concept of the work of art, which, in the case of ergodic phenomena such as MUDs, becomes notoriously unclear.

Obviously, this isn’t a very useful way to think about reading Tale-spin‘s output. So I’m left — despite having finally clarified the fact that ergodic and cybertext refer to rather different categories — without a clear feeling about the meaning of either term. Can anyone out there help me? Are there, perhaps, other pieces of Aarseth’s that present the terms in a way that could help clear up the confusion that my re-reading of Cybertext has induced?

44 Responses to “Clarifying Ergodic and Cybertext

  1. babylona Says:

    Hm. That’s unusual to me, since “ergodic” already has a specific meaning, and one at some odds with the above, depending somewhat upon interpretation, in game evolution theory (or evolving game theory, or whatever the author in question is calling it.) To quote “Individual Strategy and Social Structure,” by H. Peyton Young, “…ergodic, that is, its long-run average behavior is essentially independent of the path taken; furthermore it is independent of initial conditions.” (p. 10)

  2. Jill Says:

    I wonder whether the potential for productive misunderstandings may be some of the reason why cybertext is a concept that caught on?

  3. noah Says:

    babylona, yes, one interesting thing is how differently “ergodic” is used in elit circles than elsewhere. While my German is basically non-existent, I think that gets some discussion in Frank Furtwängler’s recent Dichtung Digital essay.

    Jill, you may be right, and this is hinted at a bit in Cybertext. For example, in the last chapter (p. 182-83) we read:

    My extensive construction of and use of neologisms, such as cybertext, ergodic, and intriguee is a sure sign of the tentative, rapidly changing phase we are going through at the moment… My ambition is to make them both readable and writeable (and in a way that indicates the problem with these Barthesian terms, or at least with my understanding of them): readable, in the sense that their denotation should be as clear as possible (admittedly, I am, or try to be, one of what Gayatri C. Spivak recently called the ‘clarity-fetishists’); and writeable, in the sense that I want you, the reader, to be a user in a transcending, cocreative, author mode. Please use these terms in any way you find pleasurable, please rewrite them, refute them, or erase them, if you want.

  4. scott Says:

    Thanks Noah, I thought I had a handle on Cybertext. Now I’m thoroughly confused. And ergodic means something else in another part of the universe? Clarifying? I’m covered in mud.

    I suppose its useful to consider that hypertext is not a subset of cybertext. Writings that branch or perform on request are not necessarily writings that require calculation in their production of scriptons. On the other hand, we could parse that sentence [“I suggest the term cybertext for texts that involve calculation in their production of scriptons.”] as non-exclusive. Just because he suggests cybertext for those types of texts doesn’t mean that he doesnt’t suggest it for something else as well. I don’t have the book in front of me so I just shut up.

    But regardless, we can still say that hypertext and cybertext are both forms of ergodic literature, yes? I hope so. I was always clearer on ergodic than cybertext, and find the concept of “non-trivial effort to traverse the text” useful. I particularly enjoy asking my students to work harder on their non-trivial efforts.

    But now I’m curious about the etymology of “ergodic,” and I can’t read German either. Does Espen’s use of the term relate to this other, Babylonaian usage? I hope that Espen responds.

  5. nick Says:

    I only have a moment, but…

    Of this list, the only immediately problematic entry is Meehan’s Tale-spin — which is a program for generating simple, linear stories. It’s hard to understand how traversing the output from Tale-spin requires nontrivial effort.

    Noah, why do you say “traversing the output”? As you quoted, “In ergodic literature, nontrivial effort is required to allow the reader to traverse the text.”

    TALE-SPIN requires input from the user; the “work” of providing this input is what makes the text (or text/machine) TALE-SPIN ergodic, as I understand it.

  6. Espen Aarseth Says:

    Just to clarify, I lifted the *word* ergodic from physics, but not its meaning. It is not a scientific metaphor. Semantically, my use of the term should be understood solely in light of its Greek etymology, a work-path. Any resemblance to real physics concepts and ideas are unintentional.

    As for Tale-spin, it is the textual machine itself that is the object of classification, not the individual fragments it produces. And as Nick points out, the user configures the program prior to story generation.

    The Money Spider is classified in detail in the table on p. 69, where the differences between it, Afternoon etc should be clear. The main difference is that MS is personal (roleplaying) and changes (the ‘web’ page where the player marks the progress). For other differences and classifications as well, see the detailed typology rather than the top-down concepts of ergodic and cybertext.

    The relation between cybertext and ergodic is described by the model on p. 64. Hypertexts are those ergodic texts that are static, while cybertexts can easily emulate/subsume both hypertext and linear text. (An example of the latter would be machinima.) So this means that hypertexts are not a kind of cybertext, just as linear text is not a kind of hypertext, while the reverse claim could also be made: a hypertext is a static subset of cybertexts, and linear text is a mono-linked subset of hypertext.

    I rarely use these two concepts anymore, however, since they are of little use in the study of games. And I hope people will come up with better terms, that is, terms that can explain and describe differences and properties more clearly, accurately, and more nuanced than mine could. At any rate, misunderstandings are unavoidable, since people will read any text in the way which best serves their own purpose, my own reading of my own work included.

  7. noah Says:

    Hey Nick — thanks for your comments. I know it’s a busy moment for you.

    I haven’t read Meehan’s dissertation yet (it’s on my near-term list) so it’s entirely possible that much of my confusion about “ergodic” is because I’ve been misunderstanding Tale-spin. I thought Tale-spin just produced output. Are you saying it runs interactively? What kind of input does it accept from the reader, and at what point in the process?

  8. babylona Says:

    Unfortunately my German is well past non-existent, so unless I can find a translation of that essay, I’m afraid I won’t be able to get much out of it, since I would bet that it says a lot more than “I am a Berliner” and “Speak German?” and “Badger dog.”

    As for ergodic, to take the subject at hand and not only change it but in a brief sentence manage to question an enormous body of work, as far as I am concerned the word in game theory (of the evolutionary sort, not the playing-games sort, jeez we need new names for things) refers to an entirely mythological state.

    Back to the point at hand: first, I would suggest that if one is going to go about fetishizing clarity, which is a position I personally support wholeheartedly, one might consider choosing words that do not already have established meanings in massive, very detail-oriented, highly academic fields, at least if the meaning you are proposing is at odds with the existing meaning.

    Second, to bring back tangent point and try to address Scott’s question: ergodic and non-ergodic are core concepts in game theory. Game theory purports to use a number of mathematical formulae to state several evolutionary and behavioral principles, explain and predict behavior of individuals and groups, and define probability sets for a wide range of activities and long-term trends. If that sounds absurd, it kind of is. The problem from an intellectual standpoint, at least for me, is that while it is patently absurd, and reading the texts is extremely surreal, the formulae are actually often correct, that is, mathematical formulae written to predict group behavior and evolutionary trends often reflect what has already occurred in the real world.

    This absurdity is exacerbated by implementations of the concept of ergodic. I say I believe it refers to a mythological state because I cannot accept that there are any states which are not in some part dependent on both initial conditions and paths chosen. However, from the probabilistic standpoint, I am wrong. There are a number of probability frameworks that will consistently produce the same outcome regardless of initial conditions and paths chosen; probably the most famous and most easily-understood of these is the segregation framework. See Sugarscape and the book I mention above for some examples, but the segregation framework is described in a number of ways in any game theory book you might pick up – it forms a central pillar for evolutionary game theory.

    Third, I certainly can’t speak to Mr. Aarseth’s choice of the word, but I would be very surprised if he did not intend it to bear some relationship to the existing meanings of the word. It seems that the problem here is at least partially rooted indeed in that etymological question, since his statements present a marked departure from the game theory meaning. There is further confusion in every field I have faced – perhaps greater confusion, in fact, than that presented by the mixed meanings of the words brought up here – in addressing what is called here a “user.” Some game theory books say “people,” to the enormous detriment of the frame of reference in those books, since the “people” therein described are extremely simplistic, barely humanOID, let alone human, representations of some aspect of probability or formulae. “User” is also confusing in that it tends to refer to a person. I prefer the Sugarscape (and others’) approach of terming the behavioral unit in question an “agent,” since the only thing that denotes is some capability to perform some level of action, with no connotation of complexity or intelligence.

    I’ll be very interested to see if Aanseth doesrespond – I would love to hear his take on the question(s.)

  9. babylona Says:

    (Great. Aanseth replied after I had reloaded the page. Please imagine that my last coment is a few notches above its actual location :)

  10. noah Says:

    Espen – Sorry, I just found your comment in the moderation queue. Otherwise I wouldn’t have responded only to Nick a moment ago. Also, my question to him doesn’t make sense, now that I’ve seen your response.

    So, forgive my ignorance about Tale-spin. Is the important point here that it is distributed in a way that allows (or requires?) readers to configure it before it runs? I’ve never actually seen a version of it running, and a brief web search has only turned up programs that are kinda based on it. It sounds like my thinking about this would definitely be clarified if one of you wouldn’t mind telling me a little more about what it’s like to run/read Tale-spin.

    I’m glad to know that the hunch about The Money Spider that I formed from my web search (about the book’s fill-in “web” page) was correct. So, let me see if I understand. It sounds like what separates The Money Spider and Afternoon from Victory Garden is that: (a) they maintain state in some manner and (b) different behavior can result while the work is in different states. Is that roughly accurate? Would you say, more generally, that is the way (or one of the ways) that you were separating between cybertexts and non-cybertexts at the time you wrote that chapter?

    As for the model on page 64, I think I understand it, especially now that I have your additional gloss here. (Of course, I have to note that I disagree that hypertexts are static by definition, though we should of course use your definition when trying to understand your book.) But I guess the important point that I should come away with, on this subject, is that static hypertexts are not part of the category of cybertext (and neither is Queneau’s Poems or anything else that doesn’t maintain state and behave differently based on state) though a static hypertext’s behavior could be emulated in a cybertextual environment?

  11. michael Says:

    I haven’t read Meehan’s dissertation yet (it’s on my near-term list) so it’s entirely possible that much of my confusion about “ergodic” is because I’ve been misunderstanding Tale-spin. I thought Tale-spin just produced output. Are you saying it runs interactively? What kind of input does it accept from the reader, and at what point in the process?

    Tale-spin lets the reader set up initial situations (characters, locations, initial character goals). The story is produced as the characters pursue their goals. During the generation process, there is no interaction, though one could imagine adding interaction by passing to the reader choices that occur during the planning process; this is in fact the approach some of my students have taken to “interactivizing” the Universe story generation system. Incidentally, you can download and play with Warren Sack’s reimplementation of micro Tale-spin. I usually demonstrate and talk about micro Tale-spin in my interactive narrative class. One year one of my students modified micro Tale-spin to generate anime stories in the high-school romance genre. Another amusing use of Tale-spin is Michael Cox’s a drug-addicted Elvis version of Tale-spin. He wrote it for use in his work on the meta-Aqua story understanding system; the modified Tale-spin produced stories for meta-Aqua to understand).

    Ah, I see the comments have shifted out from under me while composing this; well, I’ll let it stand, even if some of it is no longer as relevant.

  12. nick Says:

    While it isn’t Meehan’s original system, the Common Lisp “MICRO-TALESPIN” that Warren Sack wrote is pretty relevant here. You can look through the file to find *story1* through *story6*, which show some of the things a user would need to specify to generate a story:

    It’s also pretty easy to run it. OS X/Fink users can type “fink install clisp” to get Common Lisp set up, then just run clisp from the directory where you’ve put the file and (load “micro-talespin.txt”).

  13. michael Says:

    It sounds like my thinking about this would definitely be clarified if one of you wouldn’t mind telling me a little more about what it’s like to run/read Tale-spin.

    Micro Tale-spin is pure Common Lisp, so it’s pretty easy to play with. Just download a free Lisp (like GNUs), and have at it. You of course do need a basic proficiency in interacting with the Lisp top level. Micro Tale-spin doesn’t have an interactive loop; you interact by binding initial situations to a global variable and then evaluating an entry-point function. But it would be trivial to add a loop that asks you questions about the initial situation and builds it dynamically. When I first joined the Oz Project in 1996, I remember coming across Peter Wehyrauch’s Tale-spin implementation in the Oz text world while spelunking old Oz code. In that version, you set the initial situation by selecting stuff from various menus.

    Update: Looks like Nick wrote about micro Tale-spin between my two references to it. OK, enough about micro Tale-spin :).

  14. Ian Bogost Says:

    Espen, a couple questions on this subject as long as the line is open (of course, anyone else can toss in their thoughts of well):

    (1) This is an extremely pedantic question, and one I normally wouldn’t have interest even in posing, but why not in this context. In the book you’re very clear about borrowing the word “ergodic” from physics, but as I understand it physics in turn borrows the term from mathematics (a statistically representable recurring state) for use in periodic systems theory. I’m not a physicist and I could be wrong about this. I’m just curious if there was any reason you refer to physics in particular and not mathematics (or statistics).

    (2) Given that your interest in the term ergodic is driven by etymology rather than usage, and given that it’s so rarely that I get to talk about Greek, I’d like to take advantage of the opportunity. Hodos does indeed mean “path,” but it can also mean “way” or “method” or “manner”, as we might talk about a “way of doing things” (or even how the japanese talk about a condition or a predilection … “gyoza no hoo ga suki desu” = “I prefer potstickers” but literally “I like the way of the potstickers”). Hodos comes close to suggesting the function of a system; even in the more usual sense of hodos as “path” it can mean something like “course”, like the course of a river. That kind of path implies mechanism more than does “road,” which suggests the notion of choice to my Robert Frost-weary ears. Here’s an example of the usage I’m thinking of, from Plato’s Phaedrus (256b):

    Greek: Sôcratês: oukoun ton mellonta tekhnên rhêtorikên metienai prôton men dei tauta hodôi diêirêsthai, kai eilêphenai tina kharaktêra hekaterou tou eidous…

    English: Socrates: Anyone who intends to make use of the craft of rhetoric must first determine the various methods, and get a clear impression of each of the forms…

    So essentially I’m wondering if the types of ergodicity at the heart of this question can be teased out of the Greek itself… the ergodic texts that you exclude from cybertexts, such as “ordinary hypertext,” if I’m getting it right, are ergodic in the sense that they require effort (ergon) to choose a path (hodos); the ergodic texts that you include as cybertexts, even such as Afternoon, a hypertext, are furthermore ergodic in the sense that they deploy an internal method or a logic, presumably any logic of any kind so long as it is formalized in and effectuated by the work itself. (?)

    (3) As a potential short-circuit for (2) above, you started a rather lovely sentence in your last comment but then didn’t follow through with the parallel statement I was expecting. You say Hypertexts are those ergodic texts that are static, while cybertexts … and then you talk about emulating hypertext in cybertext. I’m wondering how you would complete this statement: Hypertexts are those ergodic texts that are static, while cybertexts are those ergodic texts that are…

  15. Espen Aarseth Says:

    If I as a game researcher have offended economists or mathematicians by appropriating ‘ergodic,’ I guess that would be simply poetic justice, given their massive misuse of the rather common word ‘game’. Terms by themselves are of course meaningless (or Babylonian). I made clear the meaning of ‘ergodic’ in my book, and most readers seem to have had little trouble understanding that it was not dependent on its use in other scientific fields.

    When terms become more important than the the models and concepts they support, we are in the field of ideology, not wissenschaft. I remember a psychologist once reprimanding me for using the word “multi-modal” in a way he did not recognize from his own discipline. Luckily for me this was at an explicitly interdisciplinary symposium, so others reprimanded him in turn.

    Which brings me to hypertext, a concept I never found very useful as an analytical tool, since nobody could agree on what it was and was not. We’ve had this discussion before, Noah, so I will not repeat myself here, merely say that with my more detailed model of tetual behavior (or someone else’s better future model), we don’t need the term hypertext to distinguish between things like John Cayley’s Clock, Queneau’s Poemes and Crowther and Woods’ Adventure, all of which both are and are not hypertext, depending on who you talk to.

    Queneau’s Poemes *is* a cybertext (according to the figure on p. 64), because the user configures the poem. The important thing to note here is that it allows the user some creativity in the *shaping* of the textual output, unlike, say Victory Garden. Given the table of classifications on pp. 68-9 the differences between the texts should be quite clear (or as clear as I could make them), and while there are probably better ways of doing this (e.g. using concepts like state machines, petri nets, matematical combinatorics and whatnot), I will leave that task to others.

    (1) no reason, at least none that I still remember. Probably carelessness mellowed by ignorance.
    (2) Not really, no. A cybertext can be effectuated solely by the choices of a user following a recipe, as in the case of Queneau. I never meant to make a sharp distinction between way and mechanism. Instead, such distinctions are taken care of in the multidimentional model.
    (3) I probably said too much, which is a problem when people quote from what I write in places like this, rather than, like Noah, actually bothering to read the book (the first chapter is still online somewhere). So let me instead correct myself: There are also ergodic, static texts that are not hypertext (linked fragments) or hypertext-like. Instead of simply constructing cybertext as a category, my idea in the book was to use it as a perspective. As an analytical category it is too broad to be of much use, as a perspective it draws attention to aspects and genres that other perspectives may have overlooked.

  16. Jonathan Says:


    Boltzmann was responsible for the development of statistical mechanics, and I think the word was coined in that context.

    I wonder if “conatus” or “nisus” might be more appropriate terms.

  17. noah Says:

    Michael and Nick – thanks for the pointers to Micro Tale-Spin! I have to admit, however, that it doesn’t entirely clear up my confusion about “ergodic.” It turns out that Tale-spin runs just the way I thought it did when I wrote my initial post: you load it, you run it, and it tells you a story. It certainly didn’t require any non-trivial effort for me to traverse the text I got by following the load/run instructions: it just told me about Joe getting a drink of water and not being thirsty any more. It’s the same effort as running Agrippa (if I understand Agrippa, it outputs a linear poem after you run it, following the instructions). The difference between Tale-spin and Agrippa, it seems to me, is that Tale-spin has a mechanism that would allow me, if I chose, to create a different story kernel and tell it to use that to generate a story. But this doesn’t feel captured by the phrase, “nontrivial effort is required to allow the reader to traverse the text.” Rather, it seems like the point is that the reader has the *option*, supported by the system, of expending non-trivial effort that would generate a text. Or have I simply compounded my misunderstanding?

    Espen, thanks for setting me straight about Queneau’s Poems. It sounds like my confusion about the term “cybertext” may actually have, at its root, a small error on page 75. On that page the book says (as I also quote above):

    I suggest the term cybertext for texts that involve calculation in their production of scriptons. This criterion corresponds nicely to all the texts in the west half of figures 3.2, 3.3, and 3.4, while in the east it only applies to Afternoon….

    But Queneau’s Poems appears in the east half of figure 3.2, leading me to believe (when I wrote my initial post) that it wasn’t a cybertext. If you tell me that it is a cybertext, then the whole thing makes more sense to me. Basically, it sounds like a cybertext is any text that has some sort of built-in mechanism that can produce differences at the level of the scripton (rather than just, say, the connections between scriptons). So The Money Spider is only a cybertext because there are places for the reader to write things in, thereby changing one of the scriptons. Similarly, if Queneau’s piece were implemented as a choose-your-own-adventure poem (“if you want to now read line two from the second original sonnet, turn to page 15”) rather than as a way of flipping individual lines to make up a single page, then it wouldn’t be a cybertext any more (because each scripton, each page, would be unchanging). Or am I on the wrong scent again?

    In any case, I appreciate everyone (and especially Espen, close to a decade after he wrote all this) taking the time to work through these two terms with me. BTW, we’re still having issues with people’s comments getting caught in the moderation queue. I believe the way around this is to provide a consistent email address and web address when you comment. (These can be made up, and, if you provide both, only the web address will be visible.) Once we’ve approved one of your comments with a certain set of addresses, your future comments with that set will show up immediately.

  18. nick Says:

    It certainly didn’t require any non-trivial effort for me to traverse the text I got

    To repeat: I don’t believe any of Espen’s distinctions are about traversing “the text you got” or about “traversing the output.”

    A central point of Cybertext, as I understand it, involves enlarging the traditional concept of text to something more like text/machine. TALE-SPIN is a text/machine. Espen wrote about the system and about how his perspective could be used to learn about the entire system, not just about the “text” (in the more narrow sense) that it outputs.

    As to whether or not TALE-SPIN is a cybertext, if you refuse to do anything different with a cybertextual system (don’t turn the strips of paper in Queneau’s Poèms, don’t click on Afternoon, type only “(micro-talespin-demo *story1*)” but never anything else), then, if it’s a deterministic system, of course you get the same output every time. But that’s just because you’re not looking at the whole system and what it can do, you’re just trying to guess something about it from reading the output.

    It turns out that Tale-spin runs just the way I thought it did when I wrote my initial post: you load it, you run it, and it tells you a story.

    You should really study the program further. You have made it generate more than one story, yes? In a discussion of cybertext and user agency, your reply makes me think of a user who believes “ls” or “dir” do the same thing every time because he always runs these commands from his home directory without parameters, and he never creates or destroys files in his home directory. MICRO-TALESPIN is not just a program that you load and run, resulting in a story. At the very least, some story kernel, chosen by the user, must be supplied as an argument to micro-talespin-demo, but you can also define your own story kernel.

  19. noah Says:

    Nick, I’m not trying to get into some fight about details. I’m genuinely trying to understand what Espen meant by “ergodic” when he wrote Cybertext. It sounds like, if Tale-spin is a canonical example of ergodic literature, then ergodic literature isn’t that literature for which “nontrivial effort is required to allow the reader to traverse the text.” Really, it sounds like it’s the literature for which “it is possible, and in some cases required, for the reader to expend non-trivial effort and thereby produce different sequences or bodies of text.” Maybe that’s not the best way of putting it, but do you see what I’m getting at? Telling Tale-spin which kernel to use isn’t nontrivial effort, and writing your own kernel isn’t required.

    As for my confusion about what “cybertext” meant, when I wrote this initial post, I’m guessing it came down to (a) not knowing enough about The Money Spider, and (b) an error on page 75 of Cybertext, where it should have said “This criterion corresponds nicely to all the texts in the west half of figures 3.2, 3.3, and 3.4, while in the east it only applies to Afternoon and Cent mille milliards de poemes…

  20. Espen Aarseth Says:

    Noah, I was classifying Tale-spin based on Mehan’s description in his 1976 PhD dissertation ( e.g, chapter 11.) and not on Micro-Tale-spin. Even so, when programs operate in several modes, it can be difficult to pidgeon-hole them: Is Doom a single-player or multi-player game, or both? The fact that Tale-Spin has a secondary, automatic mode does not diminish its ergodic qualities.

    You are of course right in pointing out the omission of Cent Mille… from the sentence on page 75, thanks! I can only conclude that the use of a bottom-up empirical model once again works to correct top-down, error-prone categorizations. Having readers like you also helps, of course.

  21. Ian Bogost Says:

    Espen —

    On (1), I was just curious. I’m not suggesting that your usage of the term is incorrect nor that one bears responsibility to every possible usage of a term. As I said, it was a pedantic question.

    On (2), my point, however experimental, is that it is possible to tease out a few different notions of actual utility from the Greek terms ergon and hodos, including one between path and mechanism. One could, it would seem, adopt the term ergodic to both ends. This was not your goal nor is it necessarily mine, but it might be a possible extension/adaptation/perversion of the concept that could suggest resonances between non-trivial effort and generativity.

  22. nick Says:

    Noah, I’m not fighting, I’m trying to help you understand. Sorry that I’m not able to.

  23. Harald Says:

    Espen Aarseth said:

    >If I as a game researcher have offended economists or mathematicians by >appropriating ‘ergodic,’ I guess that would be simply poetic justice, given their >massive misuse of the rather common word ‘game’

    Surely it is more permissible – or at least traditional – to appropriate very common words for technical use than to appropriate technical terms that already have another meaning? I have not yet met anybody who minds the (different) uses of the words “stack” in mathematics and computer science, say; yet, if you call your newborn concept after something completely different in an unrelated field, people will suspect you of choosing sexy names to get grants from confused committees. A curious example: this is the kind of suspicion that many mathematicians feel towards the chaps who chose to call their subfield of maths “quantum ergodicity”; here the filched sexy name is not “ergodic”, but “quantum”!

    Of course, chance has it that “game” is by now a central technical term for people who have at least as much of a right to it as mathematicians and economists do
    (but weren’t we there first?). “Ergodic” has at least two different meanings in maths, by the way — neither of which seems to have much to do with what I am reading here.

  24. Harald Says:

    babylona says:
    >Hm. That’s unusual to me, since “ergodic” already has a specific meaning, and one at >some odds with the above, depending somewhat upon interpretation, in game evolution >theory (or evolving game theory, or whatever the author in question is calling it.)>To quote “Individual Strategy and Social Structure,” by H. Peyton Young, “…ergodic, >that is, its long-run average behavior is essentially independent of the path taken; >furthermore it is independent of initial conditions.” (p. 10)

    This is much, much closer to the meaning of “ergodic” in maths (ergodic theory, to be more precise; there is another, possibly obsolescent meaning with which I am not very familiar).

  25. noah Says:

    Nick, apologies for my tone. I meant to assure you that I wasn’t deliberately looking for a fight. Sorry it sounded like I was implying that you were.

    In terms of “ergodic” and Tale-spin, I definitely plan to read Meehan’s dissertation and spend more time with Warren Sack’s reimplementation. In the meantime, I think Espen’s comment about Tale-spin having an ergodic mode of operation, much the way Doom has a multi-player mode of operation, helps me understand what Nick was trying to tell me earlier. Basically, if I understand correctly, even if Tale-spin can produce batch-mode stories, it can also run interactively — as demonstrated by story kernel 6 in Sack’s version. So while the whole of Tale-spin doesn’t “require” non-trivial effort (it can be run with non-interactive settings) it is a system that’s built to accommodate experiences that do require non-trivial effort.

    As for “cybertext,” let me see if I can summarize what I feel I’ve learned:
    – The terms “cybertext” and “ergodic” are not two sides of the same coin.
    – The category of “cybertext” explicitly excludes static hypertexts, so hypertexts are not a subset of cybertext whether we view the term “hypertext” as Espen does or as I do.
    – When we say cybertexts “involve calculation in their production of scriptons” that can mean a number of things. First, it can mean that the scriptons are generated differently out of textons. In this case, Queneau’s Poems is a cybertext when it’s presented as re-arrangeable strips (textons) that are combined into different sonnets (scriptons). If Queneau’s Poems was, instead, presented in a choose-your-own-adventure format (with a page for each first line, and instructions of where to turn from each of these to a page with each second line, and so on) it wouldn’t be cybertext because the scriptons would always remain the same. However, you can still have a cybertext where the scriptons always stay the same, such as Michael Joyce’s Afternoon. In this case, the calculation involved in the production of scriptons is that the connections between the scriptons (and therefore the potential orders in which they can be traversed) change when the work is in different states. Also, if I understand correctly, the possibility of textons being altered or added would make something a cybertext.

    Finally, in terms of the text of Cybertext, I think we’ve come to the conclusion that on page 75 Queneau’s Poems should be added to Afternoon as an example of things on the east half of figures 3.2, 3.3, and 3.4 that are cybertexts. I don’t know if JHUP would possibly add a note of this sort in future printings, but it might avoid additional confusions like my own. Personally, if they were willing to consider any changes, I’d also like to advocate for a note giving slightly more information about how The Money Spider functions. Reading through Cybertext as carefully as I could, I thought this was just a “typical gamebook,” and I think people would benefit from an explanation of how it has conditional linking. In addition, while I know this is probably much too much to ask, it would be great if JHUP were willing to add a note to chapter 6 that indicates in some way that Tale-spin can be run interactively, and this is when it is ergodic. Perhaps even if it’s not possible to add these to future printed versions, they could be put on the online errata page for the book?

  26. Espen Aarseth Says:

    Harald (you are not my old physics professor from Bergen, by any chance?),

    if my use of “ergodic” turns out to be problematic (if the other concepts could be fruitfully applied to the field of comparative media theory in a way that causes confusion because of the name conflict) then I will be the first to regret it. Until that happens, though, I don’t really see a problem. I explicitly appropriated a term, not a concept. Instead, as I point out in the book, my concept is very close to the original, very common Greek meaning. If, as you say, there are several techincal meanings of the term in your field, then I am happy this is not the case in mine.


    thanks for the wishlist. I’ll see what I can do, but I doubt there is room for thicker descriptions of the sample texts without a major revision, and that is unlikely. (And as for explaining ergodic using the term interactive, well…) There are a number of issues and problems with the book that could be fixed, but I’d rather let it float on by itself while it still can, and wait for someone else to come up with a more “systematic and rigorous theory” of these matters.

    The problem you still seem to have with cybertext might stem from the fact that you are trying to peg it as a formal category, which it really isn’t intended to be, as I pointed out in the book. My critique of the concept of hypertext isn’t solved by the concept of cybertext, but by the multimodal typology, by which it should be clear that these broad concepts cover very different individual works and genres, which are always dangerous to lump together under one term. Instead, the formal categories of the typology could be used to find out precisely which characteristics of texts that are the important differences that makes a difference. Or a better typology, which can be built from scratch or by modifying the present one.

    As an example of how the typology can be used to clarify the conceptual problems, let us look at your Queneau example. The problem here is that the hypothetical text you describe would be formally different from Queneau’s original. This cannot be explained in terms of ergodic or cybertext, but by applying the formal categories and concepts of the typology:

    What you propose, it seems, is a book of 140 pages where all ten alternative page numbers for each following sonnet line are presented as choices on the first ten pages, followed by the next ten alternatives for the next ten pages, and so on, 13 times. So far so good. The scriptons in this case, however, would not be identical to the scriptons of the original, since they would only contain one line each, *not* the assembled 14. And the access to the scriptons would be controlled, not random, as in the original, where you can change individual lines in any order, unlike in your version. So you are not discussing Queneau’s work, but a hypertextualisation(?) which reduces the user’s possibilities and the format of the result. The original is configurative and random access with no links, yours is explorable and controlled access with explicit links.

    Ian, I am completely open to use ergodic to mean auto-generated paths as well as user-generated paths, so by all means do (just don’t tell anyone I said so…). The main rationale for the concept was to distinguish between texts that are unicursal *and* static, and texts that are somehow different from this ideal. As you suggest, it would certainly be within the semantic frame of Greek. And we all know how non-trivial successful text generation is…

  27. noah Says:

    Espen, thanks for your further thoughts. I was just trying, in my most recent comment, to get a clear idea of what it means to say that cybertexts “involve calculation in their production of scriptons.” But I do understand that it’s the traversal functions that have proven most useful to people using the concepts from your book for analysis.

  28. Espen Aarseth Says:

    Noah, I certainly agree that the notion of calculation is not very clear. Indeed, in the original paper presented at ACH/ALLC in Santa Barbara in 1995, I had written “some sort of calculation.” The definition of cybertext that I am most happy with is the one on page 3: “A cybertext is a machine for the production of variety of expression.”

  29. Patrik's Sprawl Says:

    interesting discussions

    Currently I am following two online discussions carried out in different platforms: Coglist email list discussion on conceptual metaphors Grantextauto blog discussion on ergodic and cybertext It is interesting to think about how (if?) the materiality o…

  30. Mark Bernstein Says:

    ASIDE: I believe the history of Aarseth’s “ergodic” is instructive to those who wish to coin neologisms.

    It’s quite common for science and mathematics to borrow common words for new concepts: game theory, strange particles, strange attractor, garbage collector, electrical resistance. But in almost all these cases, the common meaning is immediately understood and is clearly unrelated to the new technical meaning. Nobody expects a garbage truck to appear weekly to clean out your program’s garbage collector, or offers to introduce a strange particle to the other guests so it will feel at home.

    It’s also common for science and mathematics to invent words, often from roots in ancient languages. This practice descends from the time when all scientists spoke Latin and Greek, but most laypersons did not; embedding a Latin term in vernacular signalled the technical meaning.

    The catch with “ergodic” is that it already possesses a technical meaning that’s widely used in thermodynamics. In the US, for example, you’re almost certain to run into the term if your studies lead you to a 3rd-year physics or chemistry course. *But* it’s still a specialized and esoteric term, one that students quickly forget. (“Intensive” and “extensive”, similarly, are terms from physical chemistry that Steve DeRose borrowed literally for his taxonomy of links; DeRose borrowed both the words and the meaning, but I think students have a hard time decoding them)

    And, as Espen says, the scientific meaning of ergodic doesn’t really illuminate the literary meaning. (There’s also an applied-math meaning that’s related to, but distinct from, the physical term. The technical term for this situation is, “Oy veh”)

    I used to wish for another word — perhaps the Latin opus/operis or labor/laboris or factus would have been a useful root. Or you could reach for “toil” and its Latin source, “tudicula”, an olive press. Tudicular literature might have been less trouble for everyone.

    But it didn’t happen that way.

  31. Espen Aarseth Says:

    Mark, you seem to imply that my appropriation of ‘ergodic’ has caused actual trouble (as opposed to mere quibbling, as seen here for the first time in a couple of comments). Perhaps you can give a concrete example of this?

    At any rate, I don’t quite see the logic behind the idea that it is ok to appropriate a term from a large linguistic domain (i.e. everyday language) but not from a specialized one, if the goal is to reduce confusion.

  32. Ian Bogost Says:

    I feel responsible for starting this quibble, so I’ll stick my head back in. I agree that the term hasn’t caused actual trouble in practice. My interest in its origins in the comments here was purely one of empty curiousity, and then additional interest in the etymological implications of the term, no matter its use in any domain of practice.

    Neologisms that confuse terms common within a specific knowledge domain might indeed be problematic, I suppose, but I’m not sure we could (or really need to) devise a ruleset for their coinage.

  33. Espen Aarseth Says:

    Ian, you give yourself too much debit, especially given the existence of comment #1, which surely was not written by you? In fact, I was not referring to you at all, and had half a mind to add “not you, Ian” to the parenthesis in my previous message, but unwisely decided against it.

    As for the term causing actual trouble or not, I withhold conclusion while I await Mr. Bernstein’s elaboration, which should settle the matter.

  34. Ian Bogost Says:

    Indeed, comment 1 was not mine, so you’re right, and kind to make light of it. But nevertheless, I agree strongly enough that one field shouldn’t feel indebted to the happenstance of usage in another (unless it desires to be) that I don’t mind putting a word in.

  35. Francois Lachance Says:

    A very stimulating exchange that has enabled me to revisit the term
    erogdic and to temporalize it. The nontrivial traversal of a text depends
    upon both the construction of the the semiotic matter that gives access
    the text and its world as well as the competencies of the reader or
    readers seeking that access. Of course, in such a formulation “text” is a
    technical term (demonstratig the influce of possible world semantics and
    discourse analysis). From a reader’s perspective, the traversal of a text
    whose semiotic material is assembled in a language not within the regular
    competencies of the reader, the text is ergodic. Whether the reading agent
    is a human being or an electronic machine, whether the language is natural
    or formal, traversal is non-trivial if the deciphering is laborious. Once
    competencies develop, the traversal of the text can become non-trivial and
    hence the text classified as non-ergodic. I think the confusion
    arises/arose from the claim that the readerly activitiy diminshes to the
    point that there is “no extranoematic responsibilites … except …”.
    Extranoematic resposibility is key here. Outside the object created in
    consciousness. Ergodic literature would be akin to the Barthesian category
    of the “writerly”. The object held in consciousness triggers the agent to
    consider the act of consciousness that holds the object present. The way
    the story has been told in certain strains of modernism is that the
    noematic forces a reflection upon the noetic. The object makes one think
    about how one thinks. I think that that discussion of Espen Aarseth’s work
    highlights the other direction: how we may think affects the object held
    in thought. A cybertext is made. Ergodic literature is played. And a text
    signed Espen Aarseth is not yet trivially traversed. Place, for example,
    the question of making-playing alongside the suggestions about
    amplification and reduction (see index under Don Ihde) from C.A. Bower’s
    on computing and education: “The use of technology, in effect, amplifies
    certain aspects of human experience and reduces others”. Worth asking what
    is amplified _and_ what reduced in the traversal of textual machines and
    in their traversals.

  36. Jill Says:

    I somehow came across the Wikipedia entry for ergodic literature today – and it’s really terrible. I suppose the cool thing is that people who obviously haven’t actually read Cybertext have cared enough to write such an article, but if someone thought this article was accurate, they’d be – well, come to think of it, maybe it explains some of the student papers I’ve read recently. I’ve marked it as needing attention from experts and if any of you would help that’d be great. So I’m trying to improve it just a little but very fast (funny how sitting down to do Real Writing always mamkes me go edit the Wikipedia instead; i really should force myself back to what I’m supposed to be doing but…) Anyway, Noah, I hope it’s OK that I filched a few bits from here, like the list of works Espen calls ergodic?

  37. nick Says:

    I just flagged that article for possible copyright violations.

  38. nick Says:


  39. Jill Says:

    Ha. Yeah, I know. Sigh.

  40. noah Says:

    Hey Jill — sorry to just be seeing your comments now. Totally fine to use whatever you like of my writing from this post in performing the Wikipedia edit. And glad to see the link pointing back here!

  41. Jill Says:

    Thanks, Noah. Actually the wikipedia edit is rather exhausting. Everyone keeps adding stupid examples that they’re sure are ergodic literature. And hell, maybe they are, who knows?

    House of Leaves: ergodic? The wikipedia entry for it calls it an example of ergodic literature, and at first I just thought that was ridiculous – but then rereading this I suppose it’s just as ergodic as Apollinaire’s Calligrammes. The problem is, all the wikipedians want to add that sort of footnote-ish text as an example, which entirely skews the concept.

    The Wikipedia’s great in some ways, but the entries about electronic literature, ergodic literature and so on and so forth are really not very good. They mostly read as though they were written by students after one course on new media where they’ve sort of half-read the textbooks, if that. Most of the discussions are clearly by people who haven’t read the literature at all but just form an approximate idea of something and fling in some links and assertions and examples based on that.

    And yet the wikipedia’s entry on ergodic literature, say, is likely to be thought of as authoritative by students. Mine sure seem to assume the wikipedia’s authoritative, despite any of my I think instead of just telling my students that quoting the Wikipedia ain’t good enough and having them disbelieve me (I do tell them it’s often a very good place to start one’s research) I’m going to have to devise some kind of exercise where they actually research specific articles and consider the history of them and who edited them and compare sources and so on. Should have done that long ago, really.

  42. jim Says:

    I’m just a casual visitor who was also puzzling over the use of ‘ergodicity’ applied to literature. I learned the statistical meaning first. Gooling ‘ergodic etymology’ brought me here.

    The antonym of ‘entropy’ is ‘ectropy.’ Would ‘ectropic’ be a reasonable alternative to ‘ergodic.’

  43. Grand Text Auto » Ergodic Histories in the Cybertext Database Says:

    […] g of the potentially-puzzling terms “cybertext” and “ergodic,” see our early discussion (ignoring, if you will, my ill-informed characterizati […]

  44. Siswo Harsono Says:

    the materials about hypertext and cybertext are worth for teaching electronic creative writing.

Powered by WordPress