Omission impossible: the ergodics of time

Critical Writing
Record Status: 
Abstract (in English): 

1. Precautions

This paper concentrates on temporal aspects of ergodic narratives. In that respect it runs counter to the still strong spatial emphasis in hypertext theory. Sadly, this emphasis often goes hand in hand with complete ignorance of narratology and with favoring the narrative models and ideals of 19th century mainstream fiction as is the case for instance with Janet H Murray´s recent book Hamlet on the Holodeck, the past of narrative in cyberspace. In order to avoid such unimaginative mistakes certain precautions had to be made.

We´ll have four points of departure: Espen Aarseth´s typology of cybertexts, Stuart Moulthrop´s web fictionHegirascope, Gerard Genette`s narratology reread through his recommendations of inventing practice, and more inherently the concept of digital metamedia that we take to mean, among other things, that it is in principle possible to time every appearance and to transform any kind of bits to any other kind of bits. (There´s nothing new or extraordinary in this if we think works like David Rokeby´s interactive installation Very Nervous System that transforms people´s movements into sounds and music). Consequently digital literature could borrow and embed elements, devices and features previously associated with other forms of art and non-art. Despite the fact that the opposition between print and digital media is for the most part a false one even Aarseth´s analysis confirms that books can´t avoid being both intransient and static (having constant scriptons). Certainly, this is the model that also the classical hypertext narratives, Michael Joyce´s Afternoon and Stuart Moulthrop´s Victory Garden, follow. However, the case is very different with Hegirascope.

2. The scope of Hegirascope

Hegirascope limits the reaction or response time of its users to thirty seconds per node. In that respect it operates like an average adventure game and can be considered to be a hybrid of narratives and games. To put it briefly it is a game of narrative, a bit like Michael Joyce´s Afternoon except that Hegirascope takes the game one step further. Both Moulthrop and Aarseth think that Hegirascope should not be defined as a novel. While this might be a tactically necessary precaution, we should not waste our time in taking author´s word as its face value, especially since Hegirascope only seems to continue a long tradition of experimental writing and artistic planning by adding just another constraint to its readers. While early or previous hypertexts introduced spatial constraints to the random access tradition of print fiction, Hegirascope adds a temporal one. Certainly the use of guarded fields in Afternoon could already be experienced as an effective temporal constraint, but of an unmeasurable kind. Labels aside, this new twist needs all our attention since it has no predecessor in print in contrast to the well-known spatial and navigational problems in Julio Cortazar´s Hopscotch and Vladimir Nabokov´s Pale Fire.

Hegirascope is not alone.William Gibson´s Agrippa can be read and observed only once, Mark Amerika´s introducing section to his Grammatron as many times as you like but those processes cannot be affected by a non-hacker reader. The former functions like a regular self-destructing artefact and the latter like subtitles in film, but they don´t have the thing that makes Hegirascope more interesting to us, that is, a double interface. This means it can be operated both by following links and by letting the client-pull proceed (and change one node to an other in every 30 seconds). When seen through Aarseth´s typology Hegirascope seems then capable of occupying both transient and intransient positions at the same time although this possibility strictly speaking depends on the awareness of its user. If some reader of Hegirascope succeeds in choosing the next link always within 29 seconds, such a reader will never have a clue that there is any client-pull operating at all. If we don´t take that awareness into account then Hegirascope is definetely a transient text of Aarseth´s typology since the mere passing of its user´s time causes the scriptons to appear.

However, Hegirascope has enormous potential for hybridity. If its now simultaneously available operating possibilities were turned either into successive phases or mutually exclusive versions then it would definetely be a hybrid of Aarseth´s genre positions. It is important to keep in mind that we´re dealing here with textonomical and not textological genres as we know that such things as pure genres do not exist, such is the law of genre as Jacques Derrida once wrote it. There are also a few inherent dimensions in Hegirascope that merit a closer look. The first of them is the recurrent nature of its transformations. It is always possible to come back to the same node after a while, access is not denied but only deferred in time. Here Moulthrop´s monster and masterpiece is in marked opposition to ordinary text generators that are linear (nonrepetetive) in their developements. The second hidden dimension is closely related to the first one: the chances to reread and revisit. Despite the fact that those thirty seconds might not be enough for grasping or even merely reading the node, the recurrent cycle ofHegirascope allows one to reread its nodes as many times as one wishes. Thirdly, the double interface could be arranged to function yet another way. The poles of that double could be controlled or determined by each other, let´s say you could follow links thirty times at most and follow the client-pull also the same thirty times but whenever either of these limits is reached, there would be no more Hegirascope for you. The question concerning the autonomy of parts has come to stay.

We can already extract some categories by breaking down and combining these possibilities. Agrippa can be neither regulated nor repeated, the first part of Grammatron can be repeated but not regulated, and Hegirascopecan be both repeated and (partly) regulated. How about a text then that could be regulated but not repeated, a demolition process at reader´s pace or a text that can be navigated but not restarted? The latter possibility was presented by Gonzalo Frasca in a poster at the 9th ACM Hypertext conference last June. So there are at least the availability and access dimensions in all relationships between the reader and the text, measurable or not. Access to the whole text or parts of it or both can be temporally contolled by limiting either the duration or the number of visits or both. Likewise the existence of the whole text or the parts of it or both can be permanent or temporally limited.

In addition to these categories we can map out at least three kinds of hybrids. First, there are the spatial ones where different parts have different genre positions. Then there are the temporally successive ones that have various phases of different genre positions following each other as in our fantasy case above. Thirdly, there are those like Hegirascope that can be simultaneously operated by at least two different ways. More possibilities could easily be created either by combining those above or by differentiating between levels, most basically between parts (nodes or screenfuls) and wholes, or more to the point or pixel, between visible and invisible structures and behaviours. Still, perhaps the most crucial move Hegirascope made was the crossing of the boundary between the measurable and calculable true time and the pseudo-time of classical narratology (Chatman, Genette, Prince). And you really have to be a Liestol not to know or notice the difference.


3. Narratology and Hypertext: linearities as trivialities

There still doesn´t exist any reasonable study or survey about how much the order in which the nodes are actually read affects the concepts and comprehensions of the (hyper)textual whole (since for some curious reason that still seems to be the goal) put forward by actual readers. I guess the more one is familiar with heterarchic and achronic print narratives the less one shares the obsession of order and the resulting hypebabble of non- or multilinearities in average hypertext theory. In other words there are texts and parts of texts that enjoy temporal autonomy since they cannot be arranged into a chronology without a reasonable doubt. Artists like Alain Robbe-Grillet and Robert Coover have long ago shown us that narratives can proceed in an aporetic order creating and destroying possible causalities along the way. Maybe their audiences have also learned to expect more than just simpleminded linearities and hierarchies.

Genette discusses time in terms of order, speed and frequency. Hypertext renders the category of order almost useless with the almost tautological exception of achronic texts. The case may seem to be the same with frequency, which could at least be modified both by quantitative and qualitative analysis of link structures. On the other hand, if we think that the necessity of navigation only increases the probability of rereading and there is nothing special in revisiting a node, then the category of frequency looks as valid as before. If there´s only one car accident there´s only one car accident. And we can count how many times our middle-aged fraud of a narrator comes back to recount it and quite possibly also how many times his reader bears with him.

The category of duration or speed is also in some need of modification. There is an ellipsis, a pause, a summary and a scene in Genette`s and quite reasonably in addition to them a stretch in Chatman´s model, measuring the pseudo-time or the relation of the time of the story (in years, days etc.) to the length of narrative (in words, lines and pages). Let me take an example. If there´s one node of 15 lines for 15 years in Gunnar´s life we`ll say it is a summary, and it is still the same summary until either the content or the number of those lines can be changed. That could be accomplished in cybertexts like John Cayley´s Book Unbound but not in classic hypertexts that are static (constant scriptons, remember) and intransient and exist only to be explored and interpreted. So it really doesn´t matter how many times Gunnar reads that node or if he doesn´t read it at all. The act of not reading does not constitute an ellipsis and various rereadings do not constitute a stretch contary to what Liestøl seems to believe in his article because the discourse time is not the same thing as the time of the reading at least under the usual hypertextual conditions described above.

The story and discourse times are always accompanied with two other temporal organisations. I refer quite simply to the times of reading and narration (or text). They are uselessly unverifiable in print narratives for they cannot be measured, controlled or calculated there. In Agrippa they both are at last measurable as exactly the time the text takes to scroll its course and vanish. As we have seen Hegirascope too makes an exception that situates it on the threshold of cybertext narratives. And to better under- stand what that might be we must finally take a look into Espen Aarseth´s typology of cybertexts and its explicit and implicit temporalities.

4. Cybertext narratives: from pseudo-time to true and real time

In the chapter devoted to adventure games in Cybertext Aarseth draws a distinction between narratives, hypertexts and cybertexts. That´s why it may seem that the concept and possibility of cybertext narratives is excluded in advance. However, there´s also a perhaps more central division into ordinary texts, hypertexts and cybertexts that compresses Aarseth´s typology. So I take the former to be there for protecting games from narrativizing literary invasions and choose the latter since my aim is more modest: to borrow features and devices from games, muds and generators, and to use them for narrative purposes. These should not be too hard to imagine: for example muds for real time metafiction and deconstructive invagination, generators for oulipian objects and operations, and games for replacing browser fiction by causally more flexible parser fiction. And because I´m quite happy with Gerald Prince´s minimal definition of narrative requiring only a temporal change and a narrative situation I see no reason to suppose one could not plan and write narratives consisting of events, progression and negotiation situated in separate levels. This kind of approach has its risks but they are not any graver than in the opposite one: if narrative is defined too narrowly one may end up like Paul Ricouer who had to humiliate himself by insisting that Virginia Woolf´s The Wawes is not a novel but "a kind of oratory".

There are scriptons revealed or generated from textons by the traversal function in Aarseth´s clear and heuristic model of (cyber)textuality. He lists seven traversal modes and by combining the possibilities these variables contain you´ll have 576 unique (textonomical) genre positions, but there´s no mention of temporally or otherwise hybrid texts that could occupy more than one of these positions either at once or successively. While two traversal modes, those of transiency and access, deal directly with time, the main difference between hyper- and cybertexts cuts across two other: the dynamics and the user functions. Keeping this in mind we can proceed beyond Hegirascope.

In order to better show some new possibilities of verbal narratives outside the pseudo- time of print narratives and narratology I´ll give three more examples. Let´s take an ordinary web fiction and add three things one by one. First, e-mail to furnish it with the textonic and configurative user functions, then soundfiles so it would contain two kinds of verbal information, and finally IRC for real time authorial intrusions and multiuser communication. In the first case, the relations of story and discourse times can be altered by adding, removing or otherwise changing the text. Here we have alterable pseudo-time in contrast to print and ordinary hypertext narratives where that relation is unalterable. Since web texts are easily terminable and the access to them can be limited in time, the time of the text and the time of the reading can have different values of true time from each other. That´s not at all typical for usually those texts function a lot like books: they seem to be there forever and are accessible as many times and for as long as you like. By combining the measurable and unmeasurable true times to the alterable and unalterable pseudo-times we´ll have four temporal categories at our disposal, but there is more to come. It might be possible to alter also the true time parameters. We can say they are either given, chosen or caused.

Soundfiles present different type of verbal information than writing. The main temporal difference here is that oral narratives have clearly measurable discourse time in contrast to written ones. On the other hand the latter are probably more easily alterable in the textonic mode than the former. For that reason only it would be nice to confront a digital novel some day where all dia- and polylogue is situated in sound files that could be listened only once.

Adding IRC to a web narrative opens yet another system of time: the real time or the time of the communication. In that any author can step back into his or her fictional world, but more importantly, this is the surest way to increase the temporal autonomy of the chosen parts of the text, and consequently to embed and support online communities with narrative performatives previously unaccountable in theory and unconfrontable in practise – one needs only to imagine a trivial polymorphic love story with an online chatting and dating service. Coming back to our minimal definition: there´s a change, a temporality and a narrative situation in narrative. A change in the type of verbal information, in genre position, in textons or in scriptons. A temporality of pseudo-, true and real times. And six kinds of mutually nonexclusive situations: the user communicates with either unalterable or alterable objects (narrators and characters), or with either unalterable or alterable subjects (bots, Elizas and other programmable creatures), or with other users or the author during the intrusions of the latter figure (pronounced dead before and for all the right reasons). It goes without saying that the author can play all the above mentioned roles while present on-line. To sum up: traditional narratology knows only the first situation. Under such circumstances not only the narrators and characters but the authors and texts too can turn out to be unreliable, that is, standard devices for sophisticated narration. It´s almost self-evident that the digital dominant of cybertext fiction creates its own set of epistemological and ontological problems in addition to those described and explained (away) by Brian McHale. Since this is a short paper we won´t go any further but instead jump elliptically ahead into our preliminary typology of ergodic time.

5. Preliminary typology of ergodic time

Taking into account the inherent dimensions of temporal hypertexts like Hegirascope, the possible hybridization of Aarseth´s model, various rereadings of and plug-ins to classical narratology and some well-established practices from the other forms of art, we can draw a preliminary typology of cybertextual temporality, an ergodics of narrative time. There are at least ten categories.

1. Permanence. If the existence of the ergodic work is temporally limited, then it is temporary. Otherwise it is permanent. There´s also a third possibility: the ergodic text can be partly temporary and partly permanent.

2. Hybridity. If the ergodic text occupies only one genre position in cybertextual typology it is non-hybrid, otherwise it is hybrid. There are two elementary types of hybrids: successive (consecutive phases of different genre positions) and simultaneous (having at least two genre positions available all the time as in Hegirascope). The time is served either concurrently or consecutevely.

3. Response time to the whole text. It is either unlimited, limited and more than what is needed, or limited and less than what is needed. The third possibility means that the text can be read or used only partly.

4. Response time to the part(s) of the text (individual screens or nodes). It too can be either unlimited, or limited two ways as described above. 5. The possibilities to revisit the whole. Either none, limited or unlimited. It is all about rereading.

6. The possibilities to revisit the parts. Either none, limited or unlimited like above.

7. The number of time zones in the whole text. Either one zone (homogenous text), several synchronous or co-dependent zones (multiple) or several autonomous zones (heterogenous).

8. The number of time frames within the visible part (screen or node) of the ergodic text. Like the text above, the screen is either homogenous, multiple or heterogenous. The difference between frames and zones corresponds roughly to the difference between the surface and the archive or storage of the text. Or to put it differently: the former can be perceived at once.

9. The type of change (if there is any). It can be either cyclic (recurrent) or linear (nonrecurrent).

10. In all previous categories there is an inherent division into given, chosen and caused parameters.



1. Aarseth, Espen. 1997. Cybertext. Baltimore: Johns Hopkins University Press.

2. Amerika, Mark. 1997.Grammatron. URL:

3. Cayley, John. 1995. Book Unbound. London: Wellsweep.

4. Chatman, Seymour. 1978. Story and Discourse. Ithaca: Cornell University Press.

5. Derrida, Jacques. 1980. "The Law of Genre." In Glyph 7, pp.176-232.

6. Frasca, Gonzalo. 1998. "Blueprint for Laura." Poster at the 9th ACM Hypertext Conference.

7. Genette, Gerard. 1983. Narrative Discourse. Ithaca: Cornell University Press.

8. Genette, Gerard. 1988. Narrative Discourse Revisited. Ithaca: Cornell University Press.

9. Gibson, William. 1992. Agrippa: A Book of Dead. New York: Kevin Begos.

10. Liestol, Gunnar. 1994. "Wittgenstein, Genette, and the Reader´s Narrative in Hypertext." InHyper/Text/Theory, edited by George P. Landow, pp. 87-120. Baltimore: Johns Hopkins UP.

11. McHale, Brian. 1987. Postmodernist Fiction. New York: Methuen.

12. Moulthrop, Stuart. 1995.Hegirascope. URL: Hegirascope.html

13. Murray, Janet H. 1997. Hamlet on the Holodeck. New York: The Free Press.

14. Prince, Gerald. 1982. Narratology. Berlin: Walter de Gruyter & Co.

15. Ricouer, Paul. 1985. Time and Narrative, part 2, p.97. Chicago: University of Chicago Press.

16. Rokeby, David. 1987. Very Nervous System. Interactive installation.

(Source: DAC 1998, Author's abstract)

Works referenced:

Critical writing referenced:

Research Collection that references this Critical Writing:

The permanent URL of this page: 
Record posted by: 
Scott Rettberg