by Umberto Eco
If The Three Musketeers were a sci-fi novel (or an example of self-voiding fiction), there wouldn’t be a problem. I could easily write the story of a space navigator who leaves Hesperus on January 1, 2001, and reaches Phosphorus on January 1, 1999. My story could posit the existence of parallel worlds in which there is a time gap of two years. One of these planets is called Hesperus; it has a million inhabitants and a king named Stan Laurel. The other is Phosphorus; its inhabitants number a million minus one (Stan Laurel doesn’t exist on Phosphorus, which is a republic) and are exactly the same as those of Hesperus (same names, same characteristics, same individual histories, same family relationships). Or else I could imagine that the space navigator travels backward in time and reaches a past Hesperus, when it was still called Phosphorus, just half an hour before its inhabitants decide to change its name.
But one of the basic fictional agreements of every historical novel is that however many imaginary characters are introduced in the story, everything else has to more or less correspond to what happened in that era in the real world.
One good solution of our conundrum could be the following: Since according to some maps it appears that, at least about 1636, the rue des Fossoyeurs, after a certain point southward, was named the rue du Pied de Biche, then d’Artagnan lives on the rue des Fossoyeurs and Aramis on the rue du Pied de Biche. D’Artagnan, who thinks the two streets are different because they have different names, knows he lives on a street which is the continuation of Aramis’ street, and by mere error believes that Aramis’ street is called the rue Servandoni instead of the rue du Pied de Biche. Why not? Perhaps he has met a Florentine called Servandoni, great-grandfather of the architect of Saint-Sulpice, and his memory has pulled a fast one on him.
But the text doesn’t tell us that d’Artagnan arrives at what “he believes” is the rue Servandoni. The text tells us that he arrives at what the reader should believe is the rue Servandoni. How can we resolve this most perplexing situation? By accepting the idea that up to now I have been caricaturing discussions about the ontology of fictional characters. What actually interests us is not the ontology of possible worlds and their inhabitants (a respectable problem in discussions of modal logic) but the position of the reader.
That Holmes isn’t married we know from the Holmes saga—that is, from a fictional corpus. In contrast, that the rue Servandoni couldn’t have existed in 1625 we can learn only from the Encyclopedia; and the Encyclopedia’s information is, from the point of view of the textual world, irrelevant gossip. If you think about it for a moment, it’s the same sort of problem that was posed by the wolf in “Little Red Riding Hood.” We know very well as empirical readers that wolves don’t speak, but as model readers we have to agree to live in a world where wolves do speak. So if we accept that there are speaking wolves in the wood, why can’t we accept that there was a rue Servandoni in Paris in 1625? And in reality that’s what we do and what you will continue to do if you reread The Three Musketeers, even after my revelations.
In my books The Limits of Interpretation and Interpretation and Overinterpretation, I insisted on the difference between interpreting a text and using a text, but I said that it is not forbidden to use a text for daydreaming. In this lecture I have “used” The Three Musketeers to allow myself an exciting adventure in the world of history and erudition. I must admit that I very much enjoyed walking the streets of Paris to find the ones named by Dumas and examining seventeenth-century plans of the city (all very inaccurate, by the way). You can do what you like with a fictional text. I enjoyed playing the role of paranoid reader and of checking to see whether seventeenth-century Paris corresponded to Dumas’ descriptions.
But in doing this I did not behave like a model reader, or even like a normal empirical reader. To know who Servandoni was you have to know a lot about art, and to know that the rue des Fossoyeurs was the rue Servandoni you have to have a great deal of specialized knowledge. Dumas’ text, which presents itself via stylistic signals as a popular historical novel, can’t claim to have such a sophisticated sort of reader. So Dumas’ model readers are not supposed to know this irrelevant detail—that in 1625 the rue Servandoni was called the rue des Fossoyeurs—and can carry on happily with their reading.
Does this solve everything? By no means. Let us imagine that Dumas had made d’Artagnan leave the Treville residence in the rue du Vieux Colombier and had made him turn through the rue Bonaparte (which already existed at that time: it was perpendicular to the rue du Vieux Colombier and parallel to the rue Férou, and in those days was called the rue du Pot de Fer). Well, no, this would be too much. Either we would throw away the book indignantly or else we would try rereading it, convinced we had made a mistake in setting ourselves up as model readers of a historical novel. We would be dealing not with a historical novel, apparently, but with one of those stories that are called uchronian—that take place in a historical time all upside down, where Julius Caesar fights a duel with Napoleon, and Euclid finally manages to demonstrate Fermat’s theorem.
Why can’t we accept that d’Artagnan walks up the rue Bonaparte, whereas we can accept that he walks up the rue Servan-doni? It’s obvious: because almost everyone knows that it was impossible for the rue Bonaparte to have existed in the seventeenth century, while almost no one knows that the rue Servandoni couldn’t have existed; not even Dumas knew that.
So our problem doesn’t concern the ontology of the characters who live in fictional worlds, so much as the format of the model reader’s Encyclopedia. The model readers foreseen by The Three Musketeers are quite keen on historical reconstruction (so long as it’s not too scholarly) and know who Bonaparte was; they have only a vague idea of the difference between the reigns of Louis XIII and Louis XIV, so that the author supplies them with a lot of information both at the outset and during the story, and they do not intend to forage in the French national archives to see if there really was a Count de Rochefort in those days. Should they also know that at that time America had already been discovered? The text neither says so nor implies it, but it’s reasonable to suppose that if d’Artagnan were to meet Christopher Columbus in the rue Servandoni, the reader should be amazed. “Should be,” because I’m only supposing. There are readers who are certainly ready to believe that Columbus was a contemporary of d’Artagnan, because there are readers for whom everything that is not present is “past,” and for them the past may be very vague indeed. So once we’ve said that the text presupposes a reader’s Encyclopedia of a certain format, it’s quite hard to ascertain what that format should be.
The first example that comes to mind is Finnegans Wake, which foresees, demands, and requires a model reader endowed with an infinite encyclopedic competence, superior to that of the empirical author James Joyce—a reader able to discover allusions and semantic connections even where they escaped the notice of the empirical author. In fact the text presupposes (as Joyce said) an “ideal reader affected by an ideal insomnia.” Dumas didn’t expect—would, on the contrary, have been irritated by—a reader like me, who checks to see where the rue des Fossoyeurs was. Joyce, on the other hand (even though the wood in Finnegans Wake is potentially infinite, so that once you’re in you can’t get out), wanted a reader who was able at any time to leave the wood and think of other woods, of the infinite forest of universal culture and intertextuality.
Can we say that every fictional text designs such a model reader, so similar to Borges’ “Funes the Memorious”? Certainly not. The readers of “Little Red Riding Hood” are not supposed to know about Giordano Bruno, as the readers of Finnegans Wake certainly are. So, what is the format of the Encyclopedia that a “normal” narrative work requires of us?
Roger Schank and Peter Childers, in their book The Cognitive Computer, allow us to approach the problem from another point of view: What is the format of the Encyclopedia one should give to a machine so that it may write (and understand) Aesop-like fables?5
In their program Tale-Spin, they first started with
a small-scale Encyclopedia: the computer was told how—given a set of problematic situations—a bear might plan to get some honey.
At the beginning of the computer trials, Joe Bear asked Irving Bird where he could find some honey, and Irving replied that “there was a beehive in the oak tree.” But in one of the early stories generated by the computer, Joe Bear became miffed because he thought Irving hadn’t answered him. In fact, his encyclopedic competence lacked the information that at times you can indicate the location of food by using metonymy—that is, by naming the source instead of the food itself. Proust praised Flaubert for writing that Madame Bovary drew near the fireplace and for considering it unnecessary to tell his readers that she was cold. Moreover, Flaubert took it for granted that his readers would know a fireplace produces warmth. In contrast, Schank and Childers realized that they had to be more explicit with a computer, and they supplied it with information on the relationship of food to its source. But when Irving Bird repeated that there was a beehive in the oak tree, Joe Bear walked over to the oak tree and ate the whole beehive. His Encyclopedia was still incomplete: the difference between source as a container and source as an object still had to be explained to him, because “finding a refrigerator will do when you are hungry [only] if you know you have to look inside it, and not eat it. None of this is obvious to a machine.”6
Another unforeseen incident occurred when the machine was told how to use certain means to obtain certain ends (for example, “if a character wants some object, then one option he has is to try bargaining with the object’s owner”). And so this happened:
Joe Bear was hungry. He asked Irving Bird where some honey was. Irving refused to tell him, so Joe offered to bring him a worm if he’d tell him where some honey was. Irving agreed. But Joe didn’t know where any worms were, so he asked Irving, who refused to say. So Joe offered to bring him a worm if he’d tell him where a worm was. Irving agreed. But Joe didn’t know where any worms were, so he asked Irving, who refused to say. So he offered to bring him a worm if he’d tell him where a worm was . . .7
To avoid this loop, the computer had to be told “not to give a character a goal if he already has it [that is, if he’s already attempted it without success] but to try something else.” But even these instructions caused problems, because they interacted badly with later information—for example, “If a character is hungry and sees some food, he will want to eat it. If a character is trying to get some food and fails, he will get sick from the lack of food.”
Here is a story that the computer came up with. Bill Fox saw Henry Crow sitting on a branch and holding a piece of cheese in his mouth. Bill was hungry and obviously wanted to eat the cheese, so he persuaded Henry Crow to sing. Henry opened his mouth and the cheese fell to the ground. Once the cheese was on the ground, Bill saw it again and should, under normal circumstances, have wanted to eat it. But the computer had been instructed not to give a character the same goal twice, so Bill could not satisfy his hunger and got sick. Too bad for Bill. But what happened to Henry Crow?
Henry Crow saw the cheese on the ground, and he became hungry, but he knew that he owned the cheese. He felt pretty honest with himself, so he decided not to trick himself into giving up the cheese. He wasn’t trying to deceive himself, either, nor did he feel competitive with himself, but he remembered that he was also in a position of dominance over himself, so he refused to give himself the cheese. He couldn’t think of a good reason why he should give himself the cheese [if he did that, he’d lose the cheese], so he offered to bring himself a worm if he’d give himself the cheese. That sounded okay, but he didn’t know where any worms were. So he said to himself, “Henry, do you know where any worms are?” But of course, he didn’t, so he . . . [and so on].8
You really have to know a lot to read a fable. But however much Schank and Childers had to teach their computer, they didn’t have to tell it where the rue Servandoni was. The world of Joe Bear was always a small world.
In order to read a work of fiction, one must have some notion of the economic criteria that rule the fictional world. The criteria aren’t there—or rather, as in every hermeneutic circle, they have to be presupposed even as you are trying to infer them from the evidence of the text. For this reason, reading is like a bet. You bet that you will be faithful to the suggestions of a voice that is not saying explicitly what it is suggesting.
Let’s return to Dumas and try reading him as if we were readers brought up on Finnegans Wake—that is, as if we were authorized to find evidence and clues all over the place for allusions and semantic short-circuits. Let’s try to overinterpret The Three Musketeers.
One could suppose that naming the rue Servandoni was not a mistake but a trace, an allusion—that Dumas dropped this name in the margins of the text in order to alert his readers. He wanted them to realize that every fictional text contains a basic contradiction just because it’s trying so hard to make the fictional world coincide with the real one. Dumas wanted to demonstrate that every fiction is a self-voiding fiction. The title of the chapter “The Plot Grows Tangled” refers not only to the love affairs of d’Artagnan or of the queen but to the nature of narrativity itself.
Here, however, the economic criteria come into play. We said that Nerval wanted us to reconstruct his story, and we could say this because the text of Sylvie contains an abundance of temporal signals. It is difficult to believe that those signals are fortuitous; it could scarcely be an accident that the only precise date in the novel comes right at the end, as if we were being invited to reread the novella to rediscover the sequence of the story, which the narrator has lost and we have not yet found. But the temporal signals scattered throughout Nerval’s text all come at crucial moments in the plot, just ‘when the reader feels lost. Those signals function like dim yet perceptible traffic lights at a foggy crossroads. On the other hand, anyone hunting for anachronisms in Dumas would perhaps find a good many, but none of them in very strategic places. In Chapter 11 the narrating voice focuses on the jealousy d’Artagnan is feeling, a drama that wouldn’t have been altered no matter what route he followed as he walked. True, one might observe that the whole chapter revolves around a confusion of identities: first we see a shadow, then this is identified as Madame Bonacieux, then she speaks to someone d’Artagnan believes is Aramis, then we discover that that someone was a woman, at the end of the chapter Madame Bonacieux will be accompanied by someone d’Artagnan again believes is her lover, but then we find out that it is Lord Buckingham, the queen’s lover . . . Why not think that the mix-up over the streets is intentional—that it functions as a sign and allegory of the mix-up over people and that there are subtle parallels between the two kinds of misunderstanding?
The answer is that, throughout the novel, cases of mistaken identity are followed by sudden recognitions, as is usual in nineteenth-century popular novels. D’Artagnan continually recognizes in a passing stranger the infamous man of Meung; many times he believes that Madame Bonacieux is unfaithful and then discovers that she is as pure as an angel. Athos will recognize Milady as Anne de Breuil, whom he married years before discovering she was a criminal. Milady will recognize in the executioner of Lille the brother of the man she drove to ruin. And so forth. The anachronism concerning the rue Servandoni, however, is not followed by any revelation, and Aramis keeps living in that nowhere until the end of the novel and probably afterward. According to the rules of nineteenth-century cloak-and-dagger novels, if we follow the Servandoni trail we find ourselves in a blind alley.
Up to now we’ve been conducting some amusing mental experiments, asking ourselves what would have happened if Nerval had told us that the carriage hadn’t been pulled by a horse, if Rex Stout had situated Alexanderplatz in New York, if Dumas had made d’Artagnan turn onto the rue Bonaparte. All right, so we’ve enjoyed ourselves, as philosophers do sometimes; but we must not forget that Nerval never said the carriage lacked a horse, that Stout never did put Alexanderplatz in New York, and that d’Artagnan never turned
onto the rue Bonaparte.
The encyclopedic competence demanded of the reader (the limits placed on the potentially infinite size of the maximal Encyclopedia, which none of us will ever possess) is limited by the fictional text. Probably a model reader of Dumas should know that Bonaparte couldn’t have had a road named after him in 1625, and in fact Dumas doesn’t make that mistake. Probably that same reader is not supposed to know who Servandoni was, and Dumas can permit himself to mention him in the wrong place. A fictional text suggests some capacities that the reader should have, and sets up others. Regarding the rest, the text remains vague, but of course it doesn’t oblige us to explore the entire maximal Encyclopedia.
The precise format of the Encyclopedia that a text requires a reader to have remains a matter of conjecture. To discover it means discovering the strategy of the model author—that is, not the figure in the carpet but the rule by which many figures can be traced in the fictional carpet.