Chapter 2 // Overinterpreting Texts - Umberto Eco





In "Interpretation and history" I looked at a method of interpreting the world and texts based on the individuation of the relationships of sympathy that link microcosm and macrocosm to one another. Both a metaphysic and a physic of universal sympathy must stand upon a semiotics (explicit or implicit) of similarity. Michel Foucault has already dealt with the paradigm of similarity in Les mots et les choses, but there he was principally concerned with that threshold moment between the Renaissance and the seventeenth century in which the paradigm of similarity dissolves into the paradigm of modern science. My hypothesis is historically more comprehensive and is intended to highlight an interpretive criterion (which I call Hermetic semiosis) the survival of which can be traced through the centuries.

In order to assume that the similar can act upon the similar, the Hermetic semiosis had to decide what similarity was. But its criterion of similarity displayed an over-indulgent generality and flexibility. It included not only those phenomena that today we would list under the heading of morphological resemblance or proportional analogy, but every kind of possible substitution permitted by the rhetoric tradition, that is, contiguity, pars pro toto, action or actor, and so on and so forth.

I have drawn the following
list of criteria for associating images or words not from a treatise on magic but from a sixteenth-century mnemonics or ars memoriae. The quotation is interesting because - quite apart from any Hermetic presumption - the author has identified in the context of his own culture a number of associative automatisms commonly accepted as effective.

1. By similitude, which is in turn subdivided into similitude of substance
(man as a microcosmic image of the macrocosm), quality (the ten figures for the ten commandments), by metonymy and antonomasia (Atlas for astronomers or astronomy, the bear for an irascible man, the lion for pride, Cicero for rhetoric).
2. By homonymy: the animal dog for the constellation Dog.
3.
By irony or contrast: the fool for the sage.
4.
By sign: the spoor for the wolf, or the mirror in which Titus admired himself for Titus.
5.
By a word of different pronunciation: sanum for sane.
6.
By similarity of name: Arista for Aristotle.
7.
By type and species: leopard for animal.
8.
By pagan symbol: eagle for Jupiter.
9.
By peoples: the Parthians for arrows, the Scythians for horses, the Phoenicians for the alphabet.
10.
By signs of the Zodiac: the sign for the constellation.
11.
By the relationship between organ and function.
12.
By a common characteristic: the crow for Ethiopians.
13.
By hieroglyphics: the ant for Providence.
14. And finally, pure idiolectal association, any monster for anything to be remembered.

As can be seen,
sometimes the two things are similar for their behavior sometimes for their shape, sometimes for the fact that in a certain context they appeared together. As long as some kind of relationship can be established, the criterion does not matter. Once the mechanism of analogy has been set in motion there is no guarantee that it will stop. The image, the concept, the truth that is discovered beneath the veil of similarity, will in its turn be seen as a sign of another analogical deferral. Every time one thinks to have discovered a similarity, it will point to another similarity, in an endless progress. In a universe dominated by the logic of similarity (and cosmic sympathy) the interpreter has the right and the duty to suspect that what one believed to be the meaning of a sign is in fact the sign for a further meaning.

This makes clear another
underlying principle of Hermetic semiosis. If two things are similar, the one can become the sign for the other and vice versa. Such a passage from similarity to semiosis is not automatic. This pen is similar to that one, but this does not lead us to conclude that I can use the former in order to designate the latter (except in particular cases of signification by ostension, in which, let's say, I show you this pen in order to ask you to give me the other one or some object performing the same function; but semiosis by ostension requires a previous agreement). The word dog is not similar to a dog. The portrait of Queen Elizabeth on a British stamp is similar (under a certain description) to a given human person who is the queen of the United Kingdom, and through the reference to her it can become the emblem for the UK. The word pig is neither similar to a swine nor to Noriega or Ceauscescu; nevertheless, on the grounds of a culturally established analogy between the physical habits of swine and the moral habits of dictators, I can use the word pig to designate one of the above-mentioned gentlemen. A semiotic analysis of such a complex notion as similarity (see my analysis in A Theory of Semiotics) can help us to isolate the basic flaws of the Hermetic semiosis and through it the basic flaws of many procedures of overinterpretation.

It is
indisputable that human beings think (also) in terms of identity and similarity. In everyday life, however, it is a fact that we generally know how to distinguish between relevant, significant similarities on the one hand and fortuitous, illusory similarities on the other. We may see someone in the distance whose features remind us of person A, whom we know, mistake him for A, and then realize that in fact it is B, a stranger: after which, usually, we abandon our hypothesis as to the person's identity and give no further credence to the similarity, which we record as fortuitous. We do this because each of us has introjected into him or her an indisputable fact, namely, that from a certain point of view everything bears relationships of analogy, contiguity and similarity to everything else. One may push this to its limits and state that there is a relationship between the adverb 'while' and the noun 'crocodile' because - at least - they both appeared in the sentence that I have just uttered. But the difference between the sane interpretation and paranoiac interpretation lies in recognizing that this relationship is minimal, and not, on the contrary, deducing from this minimal relationship the maximum possible. The paranoiac is not the person who notices that 'while' and 'crocodile' curiously appear in the same context: the paranoiac is the person who begins to wonder about the mysterious motives that induced me to bring these two particular words together. The paranoiac sees beneath my example a secret, to which I allude.

In order to read both the world and all texts suspiciously one must have elaborated some kind of obsessive method. Suspicion, in itself, is not pathological: both the detective and the scientist suspect on principle that some elements, evident but not apparently important, may be evidence of something else that is not evident - and on this basis they elaborate a new hypothesis to be tested. But the evidence is considered as a sign of something else only on three conditions: that it cannot be explained more economically; that it points to a single cause (or a limited class of possible causes) and not to an indeterminate number of dissimilar causes; and that it fits in with the other evidence. If at the scene of a crime I find a copy of the most widely circulated paper, I must first of all ask (the criterion of economy) whether it might not have belonged to the victim; if it did not, the clue would point to a million potential suspects. If, on the other hand, at the scene of the crime I find a jewel of rare form, deemed the unique example of its kind, generally known to belong to a certain individual, the clue becomes interesting; and if I then find that this individual is unable to show me his own jewel, then the two clues fit in with each other. Note, however, that at this point my conjecture is not yet proved. It merely seems reasonable, and it is reasonable because it allows me to establish some of the conditions in which it could be falsified: if, for example, the suspect were able to provide incontrovertible proof that he had given the jewel to the victim a long time before, then the presence of the jewel on the scene of the crime would no longer be an important clue.

The
overestimation of the importance of clues is often born of a propensity to consider the most immediately apparent elements as significant, whereas the very fact that they are apparent should allow us to recognize that they are explicable in much more economical terms. One example of the ascription of pertinent to the wrong element provided by the theorists of scientific induction is the following: if a doctor notices that all his patients suffering from cirrhosis of the liver regularly drink either whisky and soda, cognac and soda, or gin and soda, and concludes from this that soda causes cirrhosis of the liver, he is wrong. He is wrong because he does not notice that there is another element common to the three cases, namely alcohol, and he is wrong because he ignores all the cases of teetotal patients who drink only soda and do not have cirrhosis of the liver. Now, the example seems ridiculous precisely because the doctor fixes upon what could be explained in other ways and not upon what he should have wondered about; and he does so because it is easier to notice the presence of water, which is evident, than the presence of alcohol.

Hermetic semiosis goes too far precisely in the practices of suspicious interpretation, according to principles of facility which appear in all the texts of this tradition. First of all, an excess of wonder leads to
overestimating the importance of coincidences which are explainable in other ways. The Hermeticism of the Renaissance was looking for 'signatures', that is, visible clues revealing occult relationships. The tradition had discovered, for example, that the plant called orchis had two spheroidal bulbs, and they had seen in this a remarkable morphological analogy with the testicles. On the basis of this resemblance they proceeded to the homologation of different relationships: from the morphological analogy they passed to the functional analogy. The orchis could not but have magical properties with regard to the reproductive apparatus (hence it was also known as satyrion).

In actual fact, as Bacon later explained ('Parasceve ad historiam naturalem et experimentalem', in the Appendix to Novum Organum, 1620), the orchis has two bulbs because a new bulb is formed every year and grows beside the old one; and while the former grows, ,the latter withers. Thus the bulbs may demonstrate a formal analogy with the testicles, but they have a different function with respect to the fertilization process. And, as the magic relationship must be of a functional type, the analogy does not hold. The morphological phenomenon cannot be evidence of a relationship of cause and effect because it does not fit in with other data concerning causal relationships. Hermetic thought made use of a
principle of false transitivity, by which it is assumed that if A bears a relationship x to B, and B bears a relationship y to C, then A must bear a relationship y to C. If the bulbs bear a relationship of morphological resemblance to the testicles and the testicles bear a causal relationship to the production of semen, it does not follow that the bulbs are causally connected to sexual activity.

But the belief in the magic power of the orchis was sustained by another Hermetic principle, namely the short circuit of the
post hoc, ergo ante hoc: a consequence is assumed and interpreted as the cause of its own cause. That the orchis must bear a relationship to the testicles was proved by the fact that the former bore the name of the latter ('orchis' = 'testicle'). Of course, the etymology was the result of a false clue. Nevertheless, Hermetic thought saw in the etymology the evidence that proved the occult sympathy.

The Renaissance Hermetists believed that the Corpus Hermeticum had been written by a mythical Trismegistos who lived in Egypt before Moses. Isaac Casaubon proved at the beginning of the seventeenth century not only that a text which bears traces of Christian thought had to be written after Christ but also that the text of the Corpus did not bear any trace of Egyptian idioms. The whole of the occult tradition after Casaubon disregarded the second remark and used the first one in terms of
post hoc, ergo ante hoc: if the Corpus contains ideas that were afterwards supported by the Christian thought, this meant that it was written before Christ and influenced Christianity.

I shall show in a while that we can find similar procedures in contemporary practices of textual interpretation. Our problem is, however, the following: we know that the analogy between satyric and testicles was a wrong one because empirical tests have demonstrated that that plant cannot act upon our body. We can reasonably believe that the Corpus Hermeticum was not so archaic because we do not have any philological proof of the existence of its manuscripts before the end of the first millennium AD. But by what criterion do we decide that a given textual interpretation is an instance of overinterpretation? One can object that in order to define a bad interpretation one needs the criteria for defining a good interpretation.

I think, on the contrary, that we can accept a sort of Popperian principle according to which
if there are no rules that help to ascertain which interpretations are the 'best' ones, there is at least a rule for ascertaining which ones are 'bad'. We cannot say if the Keplerian hypothesis are definitely the best ones but we can say that the Ptolemaic explanation of the solar system was wrong because the notions of epicycle and deferent violated certain criteria of economy or simplicity, and could not coexist with other hypotheses that proved to be reliable in order to explain phenomena that Ptolemy did not explain. Let me for the moment assume my criterion of textual economy without a previous definition of it.

Let me examine a
blatant case of overinterpretation a propos secular sacred texts. Forgive me the oxymoron. As soon as a text becomes 'sacred' for a certain culture, it becomes subject ot the process of suspicious reading and therefore to what is undoubtedly an excess of interpretation. It had happened, with classical allegory, in the case of the Homeric texts, and it could not but have happened in the patristic and scholastic periods with the Scriptures, as in Jewish cutter with the interpretation of the Torah. But in the case of texts which are sacred, properly speaking, one cannot allow oneself too much licence, as there is usually a religious authority and tradition that claims to hold the key to its interpretation. Medieval culture, for example, did everything it could to encourage an interpretation that was infinite in terms of time but nevertheless limited in its options. If anything characterized the theory of the fourfold sense of Scripture it was that the senses of Scripture (and, for Dante, of secular poetry as well) were four in number; but senses had to be determined according to precise rules, and these sense, though hidden beneath the literal surface of the words, were not secret at all but, on the contrary - for those who know how to read the text correctly - had to be clear. And if they were not clear at first sight, it was the task of the exegetic tradition (in the case of the Bible) or the poet (for his work) to provide the key. This is what Dante does in the Convivio and in other wrings such as the Epistula XIII.

This attitude toward sacred texts (in the literal sense of the term) has also been transmitted, in secularized form, to texts which have become metaphorically sacred in the course of their reception. It happened in the medieval world to Virgil; it happened in France to Rabelais; it happened to Shakespeare (under the banner of the 'Bacon-Shakespeare controversy' a legion of secret-hunters have sacked the texts of the Bard word by word, letter by letter, to find anagrams, acrostics, and other secret messages through which Francis Bacon might have made it clear that he was the true author of the 1623 Folio); and it is happening, maybe too much, to Joyce. Such being the case, Dante could hardly have been left out.

Thus we see that - starting from the second half of the nineteenth century up to now - from the early works of the Anglo-Italian author Gabriele Rossetti (father of the better-known pre-Raphaelite painter Dante Gabriel), of the French Eugene Aroux, or of the great Italian poet Giovanni Pascoli, up to Rene Guenon, many critics have
obsessively read and re-read Dante's immense opus in order to find in it a hidden message.

Notice that Dante was the first to say that his poetry conveyed a non-literal sense, to be detected 'sotto il velame delli versi strani', beyond and beneath the literal sense. But not only did Dante explicitly assert this; he also furnished the keys for finding out non-literal senses. Nevertheless, these interpreters, whom we shall call Followers of the Veil (Adepti del Velame), identify in Dante a secret language or jargon on the basis of which every reference to erotic matters and to real people is to be interpreted as a coded invective against the Church.Here one might reasonably ask why Dante should have gone to such trouble to conceal his Ghibelline passions, given that he did nothing but issue explicit invective against the papal seat. The Followers of the Veil evoke someone who, upon being told 'Sir, you are a thief, believe me!' replies with: 'What do you mean by "believe me"? Do you perhaps with to insinuate that I am distrustful?'

The bibliography of the Followers of the Veil is incredibly rich. And it is incredible to what extent the mainstream of Dantesque criticism ignored or disregarded it. Recently I encouraged selected young researchers to read – maybe for the first time - all those books. The aim of the research was not so much to decide whether the Followers of the Veil were wrong or not (it happens that in many instances, by a felicitous case of serendipity, they were probably right), but rather to evaluate the economic value of their hypotheses.

Let us examine a concrete example in which Rossetti deals with one of the paramount obsessions of the Followers of the Veil. (3) According to them, Dante in his text depicts a number of symbols and liturgical practices typical of the Masonic and Rosicrucian traditions. This is an interesting question that runs into a historical-philological problem: while documents exist which attest to the rise of Rosicrucian ideas at the beginning of the seventeenth century and the appearance of the first lodges of symbolic Freemasonry at the beginning of the eighteenth century, there are none – none at least that are accepted by serious scholars – attesting to the earlier existence of these ideas and/or organizations. On the contrary, reliable documents exist which attest to how in the eighteenth and nineteenth centuries various lodges and societies of different tendencies chose rites and symbols which would demonstrate their Rosicrucian and Templar lineage. Indeed, any organization that claims its own descent from an earlier tradition chooses for its emblems those of the tradition to which it refers back (see, for example, the Italian Fascist party's choice of the lictor's fasces as a sign that they wished to consider themselves the heirs of ancient Rome). Such choices provide clear proof of the intentions of the group, but do not provide proof of any direct descent.

Rossetti sets out with the conviction that Dante was a Freemason, Templar, and member of the Fraternity of the Rosy Cross, and he therefore assumes that a Masonic-Rosicrucian symbol would be as follows: a rose with the cross inside it, under which appears a pelican that, in accordance with traditional legend, feeds its young with the flesh it tears from its own breast. Now, Rossetti's task is to prove that this symbol also appears in Dante. It is true that he runs the risk of demonstrating merely the only reasonable hypothesis, namely that Masonic symbology was inspired by Dante, but at this point another hypothesis could be advanced: that of a third archetypal text. In this way Rossetti would kill two birds with one stone: he would be able to prove that only that the Masonic tradition is an ancient one, but also that Dante himself was inspired by this ancient tradition.

Normally one accepts the idea that
if document B was produced before document C, which is analogous to the first in terms of content and style, it is correct to assume that the first influenced the production of the second but not vice versa. One could at most formulate the hypothesis of an archetypal document, A, produced before the other two, from which the two later ones both drew independently. The hypothesis of an archetypal text may be useful in order to explain analogies between two known documents that would otherwise be unaccountable: but it is necessarily only if the analogies (the clues) cannot otherwise, and more economically, be explained. If we find two texts of different periods both of which mention the murder of Julius Caesar, there is no need to suppose either that the first influenced the second or that they were both influenced by an archetypal text, because here we are dealing with an event that was, and still is, reported in countless other texts.

Worse can happen, however: in order to show the excellence of C, one needs an archetypal text A on which B and C depend. Since, however, A is not to be found, then it is fideistically postulated as being in all respects identical to C. The optical effect is that C influenced B, and thus we have the
post hoc, ergo ante hoc effect. Rossetti's tragedy is that he does not find in Dante any remarkable analogy with Masonic symbology, and having no analogies to lead him to an archetype, he does not even know what archetype to look for.

If we are to decide whether the phrase 'the rose is blue' appears in the text of an author, it is necessary to find in the text the complete phrase 'the rose is blue'. If we find on page 1 the article 'the', on page 50 the sequence 'ros' in the body of the lexeme 'rosary' and so on, we have proved nothing, because it is obvious that, given the limited number of letters in the alphabet that a text combines, with such a method we could find any statement we wish in any text whatsoever.

Rossetti is surprised that in
Dante we find references to the cross, the rose, and the pelican. The reasons for the appearance of these words are self-evident, In a poem that speaks of the mysterious of the Christian religion it is not surprising that sooner or later the symbol for the Passion should appear. On the basis of an ancient symbolic tradition, the pelican became the symbol of Christ very early on in the Christian tradition (and medieval bestiaries and religious poetry are full of references to this symbol). As regards the rose, because of its complex symmetry, its softness, the variety of its colours, and the fact that it flowers in the spring, it appears in nearly all mystical traditions as a symbol, metaphor, allegory, or simile for freshness, youth, feminine grace, and beauty in general. For all these reasons, what Rossetti himself calls the 'fresh, sweet-smelling rose' appears as a symbol of feminine beauty in another poet of the thirteenth century, Ciullo d'Alcamo, and as an erotic symbol both in Apuleius and in a text which Dante knew well, the Roman de la Rose (which in its turn intentionally makes use of pagan symbology). Thus, when Dante has to represent the supernatural glory of the Church triumphant in terms of splendor, love, and beauty, he resorts to the figure of the spotless rose (Paradiso, XXXI). Incidentally, since the Church triumphant is the bride of Christ as a direct result of the Passion, Dante cannot avoid observing that 'Christ made (the Church) his bride by his blood'; and this allusion to blood is the only case among the texts presented by Rossetti in which, by inference, the rose can be seen in a reference (conceptual, but not iconographic) to the cross. 'Rosa' appears in the Divine Comedy eight times in the singular and three in the plural. 'Croce' appears seventeen times. But they never appear together.

Rossetti, however, wants the pelican as well. He finds it, on its own, in Paradiso XXXI (its only appearance in the poem), clearly in connection with the cross, for the pelican is the symbol of sacrifice. Unfortunately, the rose is not there. So Rossetti goes in search of other pelicans. He finds a pelican in Cecco d'Ascoli (another author over whom the Followers of the Veil have racked their brains for the very reason that the text of L'Acerba is intentionally obscure), and Cecco's pelican appears in the usual context of the Passion. Moreover, a pelican in Cecco is not a pelican in Dante, even though Rossetti tries to blur such a minor difference by confusing the footnotes. Rossetti believes he has found another pelican in that incipit of Paradiso XXIII, where we read of the fowl that, waiting impatiently for the dawn, sits alert among the beloved fronds on a leafy branch watching for the sunrise so as to go and find food for its young. Now, this bird, graceful indeed, searches for food precisely because it is not a pelican, otherwise it would not need to go hunting, as it could easily feed its young with flesh torn from its own breast. Second, it appears as a simile for Beatrice, and it would have been poetic suicide had Dante represented his beloved by the awkward features of a billed pelican. Rossetti, in his desperate and rather pathetic fowling, could find in the divine poem seven fowls and eleven birds and ascribe them all to the pelican family: but he would find them all far from the rose.

Examples of this kind abound in Rossetti's work. I will cite only one other, which appears in Canto II, which is generally considered one of the most philosophic and doctrinal of the whole Paradiso. This canto exploits fully a device which is a basic element in the whole of the third book: the divine mysteries, otherwise inexpressible, are represented in terms of light – in full accord with theological and mystic tradition. Consequently, even the most difficult philosophical concepts must be expressed with optical examples. It should be noted here that Dante was led to this choice by all the literature of the theology and physics of his time: Arabic treatises dealing with optics had reached the Western world only a few decades earlier; Robert Grosseteste had explained cosmogonic phenomena in terms of light energy; in the theological field Bonaventura had debated the difference between 'lux', 'lumen', and 'color'; the Roman de la Rose had celebrated the magic of mirrors and had described phenomena of the reflection, refraction, and magnification of images; Roger Bacon had claimed for optics the dignity of a major and fundamental science, reproaching the Parisians for not considering it enough, while the English were investigating its principles. It is obvious that, having used the similes of a diamond struck by the sun, of a gem, and of a mass of water penetrated by a ray of light to describe a number of astronomical phenomena, Dante, faced with having to explain the different brightnesses of the fixed stars, should have recourse to an optical explanation and propose the example of three mirrors which, placed at different distances, reflect the rays of a single source of light.

For Rossetti, however, in this canto
Dante would be 'whimsical' if we did not take into account that three lights arranged in a triangle – three sources of light, note, which is not the same as three mirrors reflecting the light of another source – appear in Masonic ritual. Even if we accept the principle of post hoc, ergo ante hoc, however, this hypothesis would at most explain why Dante (knowing Masonic rituals of a later date!) chose the image of three sources of light, but it does not explain the rest of the canto.

Thomas Kuhn observes that to be
accepted as a paradigm, a theory must seem better than the other theories in the lists but need not necessarily explain all the facts with which it is concerned. Let me add, however, that neither must it explain less than previous theories. If we accept that here Dante is speaking in terms of medieval optics, we may also understand why in verses 89-90 he speaks of the colour that 'turns through glass – which hides lead behind it'. If, on the other hand, Dante is speaking of Masonic lights, the other lights of the canto remain obscure.

Let me now consider a case where the rightness of the interpretation is undecidable, but where it is assuredly difficult to assert that it is wrong. It can happen that certain more or less esoteric interpretive practices recall those of certain deconstructionist critics, but in the shrewdest representatives of this school the hermeneutic game does not exclude interpretive rules.

Here is how one of the leaders of the Yale deconstructionists, Geoffrey Hartman, examines some lines from Wordsworth's 'Lucy' poems, in which the poet speaks explicitly of the death of a girl:

I had no human fears:
She seemed a thing that could not feel
The touch of earthly years.
No motion has she now, no force;
She neither hears nor sees,
Rolled round in earth's diurnal course
With rocks and stones and trees.


Hartman sees here a series of funeral motifs under the surface of the text.

Others even show Wordsworth's language penetrated by an inappropriate subliminal punning. So 'diurnal' (line 6) divides into 'die' and 'urn', and 'course' may recall the older pronunciation of 'corpse'. Yet these condensations are troublesome rather than expressive; the power of the second stanza resides predominantly in the euphemistic displacement of the word 'grave' by an image of gravitation ('Rolled round in the earth's diurnal course'). And though there is no agreement on the tone of this stanza, it is clear that a subvocal word is uttered without being written out. It is a word that rhymes with 'fears' and 'years' and 'hears', but which is closed off by the very last syllable of the poem: 'trees'. Read 'tears' and the animating, cosmic metaphor comes alive, the poet's lament echoes through nature as in pastoral elegy. 'Tears', however, must give way to what is written, to a dull yet definitive sound, the anagram 'trees'.

It must be noted that, while 'die', 'urn', 'corpse', and 'tears' can be in some way suggested by other terms that appear in the text (namely, 'diurnal', 'course', 'fears', 'years', and 'hears'), 'grave' is, on the contrary, suggested by a gravitation which does not appear in the text but is produced by a paraphrastic decision of the reader. Furthermore, 'tears' is not the anagram of 'trees'. If we want to prove that a visible text A is the anagram of a hidden text B, we must show that all the letters of A, duly reorganized, produce B. If we start to discard some letters, the game is no longer valid.
Top is an anagram of pot, but not of port. There is, thus, a constant oscillation (I do not know how acceptable) between the phonic similarity of terms in praesentia and the phonic similariy of terms in absentia. In spite of this, Hartman's reading sounds, if not fully convincing, at least charming.

Hartman is certainly not suggesting here that Wordsworth actually wished to produce these associations – such searching after the author's intentions would not fit Hartman's critical principles. He simply wishes to say that it is legitimate for a sensitive reader to find what he finds in the text, because these associations are, at least potentially, evoked by the text, and because the poet might (perhaps unconsciously) have created some 'harmonics' to the main theme. If it is not the author, let us say it is the language which has created this echo effect. As far as Wordsworth is concerned, though on the one hand
nothing proves that the text suggests neither tomb nor tears, on the other hand nothing excludes it. The tomb and the tears evoked belong to the same semantic field as the lexemes in praesentia. Hartman's reading does not contradict other explicit aspects of the text. One may judge his interpretation too generous, but not economically absurd. The evidence may be weak, but it does fit in.

In theory, one can always invent a system that renders otherwise unconnected clues plausible. But in the case of texts there is at least a proof depending on the isolation of the relevant semantic isotopy. Greimas defines
isotopy' as 'a complex of manifold semantic categories making possible the uniform reading of a story'. The most flashing and maybe the most sophomoric example of contradictory readings due to the possible isolation of different textual isotopies is the following: two fellows talk during a party and the first praises the food, the service, the generosity of the hosts, the beauty of the female guests, and finally, the excellence of the 'toilettes'; the second replies that he has not yet been there. This is a joke, and we laugh about the second fellow, because he interprets the French term 'toilette', which is polysemic, in the sense of sanitary facilities and not of garments and fashion. He is wrong because the whole of the discourse of the first fellow was concerning a social event and not a question of plumbing. The first movement toward the recognition of a semantic isotopy is a conjecture about the topic of a given discourse: once this conjecture has been attempted, the recognition of a possible constant semantic isotopy is the textual proof of the 'aboutness' of the discourse in question. If the second fellow had attempted to infer that the first one was speaking of the various aspects of a social event, he would have been able to decide that the lexeme 'toilettes' had to be interpreted accordingly.

Deciding what is being talked about is, of course, a kind of interpretive bet. But the contexts allow us to make this bet less uncertain than a bet on the red or the black of a roulette wheel. The funereal interpretation of Hartman has the advantage of betting on a constant isotopy. Bets on the isotopy are certainly a good interpretive criterion, but only as long as the isotopies are not too generic. This is a principle which is valid also for metaphors. A metaphor exists when we substitute a vehicle for the tenor on the basis of one or more semantic traits common to both the linguistic terms: but if Achilles is a lion because both are courageous and fierce, we would be inclined to reject the metaphor “Achilles is a duck' if it were justified on the basis of the principle that both are bipeds. Few others are as courageous as Achilles and the duck. A similarity or an analogy, whatever its epistemological status, is important if it is exceptional, at least under a certain description. An analogy between Achilles and a clock based on the fact that both are physical objects is of no interest whatsoever.

The classical debate aimed at finding in a text either what its author intended to say, or what the text said independently of the intentions of its author. Only after accepting the second horn of the dilemma can one ask if what is found is what the text says by virtue of its textual coherence and of an original underlying signification system, or what the addressees found in it by virtue of their own systems and expectations.

It is clear that I am trying to keep a dialectical link between
intentio operis and intentio lectoris. The problem is that, if one perhaps knows what is meant by 'intention of the reader', it seems more difficult to define abstractly what is meant by 'intention of the text'. The text's intention is not displayed by the textual surface. Or, if it is displayed, it is so in the sense of the purloined letter. One has to decide to 'see' it. Thus it is possible to speak of the text's intention only as the result of a conjecture on the part of the reader. The initiative of the reader basically consists in making a conjecture about the text's intention.

A text is a device conceived in order to produce its model reader. I repeat that this reader is not the one who makes the 'only right' conjecture. A text can foresee a model reader entitled to try infinite conjectures. The empirical reader is only an actor who makes conjectures about the kind of model reader postulated by the text.
Since the intention of the text is basically to produce a model reader able to make conjectures about it, the initiative of the model reader consists in figuring out a model author that is not the empirical one and that, in the end, coincides with the intention of the text. Thus, more than a parameter to use in order to validate the interpretation, the text is an object that the interpretation builds up in the course of the circular effort of validating itself on the basis of what it makes up as its result. I am not ashamed to admit that I am so defining the old and still valid 'hermeneutic circle'.

To recognize the intention operis is to recognize a semiotic strategy. Sometimes the semiotic strategy is detectable on the grounds of established stylistic conventions.
If a story starts with 'Once upon a time', there is a good probability that it is a fairy tale and that the evoked and postulated model reader is a child (or an adult eager to react in a childish mood). Naturally, I can witness a case of irony, and as a matter of fact the following text should be read in a more sophisticated way. But even though I can discover by the further course of the text that this is the case, it has been indispensable to recognize that the text pretended to start as a fairy tale.

How to prove a conjecture about the intention operis? The only way is to check it upon the text as a coherent whole. This idea, too, is an old one and comes from Augustine (De doctrina christiana):
any interpretation given of a certain portion of a text can be accepted if it is confirmed by, and must be rejected if it is challenged by, another portion of the same text. In this sense the internal textual coherence controls the otherwise uncontrollable drives of the reader. Borges (a propos his character Pierre Menard) suggested that it would be exciting to read the Imitation of Christ as if it were written by Celine. The game is amusing and could be intellectually fruitful. I tried; I discovered sentences that could have been written by Celine ('Grace loves low things and is not disgusted by thorny ones, and likes filthy clothes'). But this kind of reading offers a suitable 'grid' for very few sentences of the imitation. All the rest, most of the book, resists this reading. If on the contrary I read the book according to the Christian medieval encyclopedia, it appears textually coherent in each of its parts.

I realize that,
in this dialectics between the intention of the reader and the intention of the text, the intention of the empirical author has been totally disregarded. Are we entitled to ask what was the 'real' intention of Wordsworth when writing his 'Lucy' poems? My idea of textual interpretation as the discovery of a strategy intended to produce a model reader, conceived as the ideal counterpart of a model author (which appears only as a textual strategy), makes the notion of an empirical author's intention radically useless. We have to respect the text, not the author as person so-and-so. Nevertheless, it can look rather crude to eliminate the poor author as something irrelevant for the story of an interpretation. There are, in the process of communication, cases in which an inference about the intention of the speaker is absolutely important, as this always happens in everyday communication. An anonymous letter reading 'I am happy' can refer to an infinite range of possible subjects of the utterance, that is, to the entire class of persons who believe themselves not to be sad; but if I, in this precise moment, utter the sentence 'I am happy' it is absolutely certain that my intention was to say that that happy one is me and not someone else, and you are invited to make such an assumption, for the sake of the felicity of our interaction. Can we (likewise) take into account cases of interpretation of written texts to which the empirical author, still alive, reacts by saying 'No, I did not mean that'? This will be the topic of my next lecture.