From the Tree to the Labyrinth: Historical Studies on the Sign and Interpretation

by Umberto Eco

  For the moment let us consider a telling example. When Aristotle says, in De interpretatione (16a 19–20, 26–29)—this at least was the way he was read in the Middle Ages—that a name is a vox significativa by convention, and that no sound is a name for natural reasons but is such only when it becomes, by convention, a symbol, he adds that inarticulate sounds, like those made by the beasts, manifest (delousi) something, though none of them is a name.

  Aristotle does not say that the sounds made by the beasts signify or designate something, he says they manifest it, as a symptom makes manifest its cause. But the Middle Ages, as we shall see, has no trouble translating the Greek delousi with the Latin significant. Boethius’s translation, by rendering symbol and index with the same term nota, obliterated the distinction and favored their identification.28 But the Middle Ages will have no problem interpreting the sounds made by animals as voces significativae, even though not the same as nomina (and various commentators explain that in such cases Aristotle is not talking about voces but about soni, because not all animals, on account of the structure of their phonatory organs, can utter voces, and many simply produce sounds).29

  The barking of the dog, which means that the dog is angry, appears in Boethius as an example of a vox significativa, though not ad placitum, but naturaliter: “canum latratus iram significat canum”—and, by the same token, voces naturaliter significativae are also the moans of the sick.30

  And so, under the genus voces significativae we find a species that according to Aristotle should have belonged among the semeia. In this category, Boethius and those who follow him lump together, along with the barking of the dog, the gemitus infirmorum, the whinny of the horse and the sounds made by those animals that have no vox but have “tantum sonitu quodam concrepant.”31

  Boethius assuredly understands that these voces signify naturaliter, because they evidently reveal their cause according to the (symptomatic) model of inference, but, having obfuscated the distinction between the doctrine of indices and the doctrine of names, he neglects an important fact: that natural sounds do not have an emitter, unless, as sometimes occurs in certain processes of divination, they are interpreted as if they had been emitted intentionally by a supernatural agent. The moaning of the sick and the barking of the dog, however, have an emitter, though we are not in a position to affirm that the emission was deliberate. But Boethius also singles out the whinny of the horse: “hinnitus quoque equorum sepe alterius equi consuetudinem quaerit,”32 when the horse whinnies to call another horse, and hence whinnies with a definite intention. In fact, in the same passage, Boethius says that “ferarum quoque mutorum animalium voces interdum aliqua significatione praeditas esse perspicimus.” We are dealing, then, with voces endowed with some meaning. But endowed by whom (before the advent of Abelard’s idea of an “active will”)? By the beast emitting them or by the human hearing them?

  Figure 4.3

  We can immediately see that the latratus canis (and all the other sounds animals make) may enjoy a double status: on the one hand the dog speaks to other dogs and on the other the dog speaks to humans. But in the second case, the alternative is still twofold: either humans understand the dog’s bark because they have acquired a habit that makes them apt at interpreting symptoms (like the sailor who has learned how to interpret the signs in the sky), or else humans have acquired a habit that makes them apt at interpreting the language the dog uses to talk to them. These are two distinct zoosemiotic problems (while yet another problem remains on the back burner: if and in what way the dog understands the language the human uses to address him).

  We must conclude, then, that with Boethius a classification of voces is inaugurated which has two characteristics: it melds together the Stoic classification of signs (as voces significativae naturaliter) and the Aristotelian classification of voces (as nomina ad placitum), and it leaves in abeyance the problem of the intentionality of the utterance of the vox.

  Consequently, this classification—consolidating a series of basically analogous positions taken by a number of authors, from Boethius to Peter of Spain, Lambert of Auxerre, Garland the Compotist, and others—would appear as in Figure 4.3.33

  4.2.5. The Thomist Reading of De interpretatione 16a

  Thomas Aquinas will not depart from this classification, though his taxonomy is more complex. Taking into account the remarks appearing passim in his commentary on the De interpretatione, whereas, at IV, 38 et seq., he seems to concern himself with a classification of the voces, reserving the appellation signum only for the voces significativae, in IV, 46, on the other hand, where he attempts to explain why Aristotle, in speaking of animal sounds, used the term soni and not voces (it was necessary, Thomas explains, to take into consideration the sounds of animals which, not being furnished with lungs, are not capable of uttering vocal sounds), he suggests the possibility of a more detailed classification, which considers the sonus as a genus.

  The Aristotelian translation available to Thomas was not yet that of William of Moerbeke (who was shrewd enough to translate symbolon and semeion correctly as two distinct notions), but basically that of Boethius, whose Latin term for both semiotic phenomena is nota (which the commentary then proceeds to read as the equivalent of signum).

  Bearing in mind the probable sources he was relying on,34 his classification could be summed up as in Figure 4.4.

  This classification betrays a number of influences. In the first instance, along the lines of Saint Augustine, Thomas calls every meaningful vox a signum. But, in II, 19, he speaks of a signum also apropos of the military trumpet (tuba), which is evidently not a case of sonus vocalis. It would appear, then, that a signum for him was any case of meaningful utterance, vocal or nonvocal. But, having translated the two Aristotelian terms with signum = nota, he takes no account of the inferential nature of the Stoic semeia, ignores every type of index except sonus, and places evident indices like the moaning of the sick and the sounds made by animals among the meaningful voces.

  4.2.6. Transcribability and Articulation

  In the second place there appears in Thomas a distinction between voces litteratae and articulatae and voces illitteratae and non articulatae. The opposition litteratae/illitteratae appears to go back directly to the text of Aristotle, where he speaks of utterances that are agrammatoi (like those of the animals), that is, which cannot be transcribed with letters of the alphabet. Which would explain why blitris (a typical example used by the Stoics and subsequently in the Middle Ages, along with buba and bufbaf, of vocal utterances that, though transcribable, signify nothing) appears among the voces litteratae, though it is not meaningful. The problem is rather that of defining what is meant by articulata (with its opposite non articulata).

  Figure 4.4

  It is unclear whether articulation concerns the sounds only or their graphemic transcription as well, and Aristotle’s texts are not explicit on this issue.35 In his commentary on the De interpretatione, Boethius (col. 395 D) appears determined to unite the two types of articulation. Some ideas become clearer if we go on to read Priscian:36 we recognize, in fact, a line of thought that is found in the grammatical tradition, and also in authors like Vincent of Beauvais.37 A vox articulata for Priscian is one that is “copulata cum aliquo sensu mentis eius qui loquitur,” and utterances are no longer classified according to a binary taxonomy, but following the matrix represented in Figure 4.5.

  Figure 4.5

  This matrix presents two distinct problems. The lesser problem concerns its internal coherence, given that the croaking of the frog appears to be transcribable in letters of the alphabet—see Aristophanes’s brekekex koax koax—while the lowing of the ox is not. But the classification probably represents current linguistic usage (simply put, that it was more customary to spell out the croaking of the frog than the lowing of the ox), and Priscian was likely referring to a panoply of examples handed down to him by the Greek tradition. The second problem concerns Thomas’s solution. If Thomas is following Priscian, it is hard to un
derstand why the difference between articulation and nonarticulation appears to distinguish nonmeaningful utterances (non significativa), while it is absent from the branch devoted to meaningful utterances (significativa)—in which the names are articulate and lettered, but not the animal sounds.

  The fact is that behind this complex of questions there lurk a number of semiotic problems that are by no means negligible. Thomas’s classification is anticipated by Ammonius.38

  After he has made it clear that for him being litterata (in other words, transcribable in letters) is the same thing, in the case of a vox, as being articulata, Ammonius seems to place the difference articulata / non articulata twice under two different genera, so that his classification can only be transcribed in the form of a matrix, as was the case for Priscian’s (Figure 4.6).

  Thomas (In 1. De Int. Exp. IV, 38) appears to take up only the first part of Ammonius’s suggestion, and he writes: “Additur autem prima differentia, scilicet significativa, ad differentiam quarumcumque vocum non significantium, sive sit vox litterata et articulata, sicut “blitris,” sive non litterata et non articulata, sicut sibilus per nihilo factus.” Why does Thomas seem to be embarrassed by a classification that would suggest a matrix rather than a tree?

  Figure 4.6

  The problem seems to lie with the very nature of a tree (of Porphyrian inspiration), which proceeds by genera, species, and specific differences. In other words, the problem arises when, starting from a series of definitions given in discursive form, one attempts to regiment them in the form of a Porphyrian tree (something Thomas did not do, though it was precisely because he did not do so that the problem facing him did not become evident). As we demonstrated in Chapter 1 apropos of the Porphyrian tree, in order to give an account of any organization of the universe (even, as in the present case, a classification of signs and voces), the same pair of differences ought to be reproduced over and over again under different genera. If Thomas had followed this procedure, the difference litterata/illitterata, like that between articulata and inarticulata, should have appeared under two distinct genera (Figure 4.7).

  Figure 4.7

  But, once he had accepted the principle that the same difference may be placed under two different genera, he ought to have reproduced it also under the signs that are meaningful ex institutione, where tuba or trumpet appears, seeing that—according to Aristotle, De anima 420b, 5–8—we have articulability among musical sounds too (to say nothing of their transcribability in musical notation).

  In that case, the tree-like structure would have given way to a system of interconnected nodes, which, as Chapter 1 suggested, Thomas may have caught a glimpse of but could not admit. Independently, however, of these considerations, what is clear is that the contradictions of Thomas’s solution stem from the fact that he is playing a double (and irreconcilable) game. One is grammatical, and it required that the voces (lettered or not) be distinguished according to their articulatory possibilities—and this is what Priscian, as a good grammarian, did. The other was semantic, and it required that the difference be posited between meaningful and nonmeaningful voces. The two taxonomies could not coexist. Thomas appears convincing when he speaks of one issue rather than the other, but he reveals all his uncertainty when he attempts (albeit motivated simply by a desire for taxonomical clarity) to put the two discourses together into a single system.

  At this point, we could leave Thomas to his fate, gratified by the fact that once again our pursuit of the barking dog has succeeded in revealing the weaknesses or contradictions of a system. At most, we could point out that Thomas does not use articulata in the same sense as Priscian (that is, “endowed with meaning”), but in the same sense as Ammonius, and that consequently blitris is an example of a vox non significativa that nonetheless has a phonetic articulation and at the same time can be transcribed alphabetically. But it is precisely this observation that leads us to wonder why Priscian (and the grammarians who follow him) attributed to articulata a connection with meaning. Nota bene, it is not that they try to meld a taxonomy of articulations with one of signification, but that they take practically for granted that there is a connection, which they do not however explore, between articulation and meaning. In the first instance we could argue that they assumed that one articulates only to express something—and this was the hypothesis that was made in Latratus canis 1989.

  In fact, when we go back to Ammonius’s commentary, we see that he makes explicit and implicit reference to Plato’s Cratylus, suggesting that there is a close link between articulatio and significatio. In Plato’s dialogue, Socrates expounds the notion according to which whoever invented the first names created them in imitation of things, endeavoring to reproduce, through the coordination of letters and syllables, their nature. In other words, there would be a relationship of an iconic type between the phonological structure of the signans and the ontological structure of the signatum. A theory very close to this is found among the Stoics.39 So it becomes comprehensible why Priscian, heir to a grammatical tradition with its roots in Stoicism, goes so far as to identify the articulatio of the vox with its significatio, followed in this by all medieval grammarians,40 while, in the logical-philosophical tradition (untouched by the grammatical tradition), the articulatio has nothing to do with the meaning, but concerns the litteratio, and hence the possibility of the written translation of the sound.41

  However that may be, it is obvious that among the grammarians the barking of the dog was on track for an unhappy ending. All the grammarian is interested in are the sounds articulated by humans, observant, precisely, of a grammar, in order to express meanings. The sounds made by animals are of no interest. Accordingly, in the texts of the grammarians the barking of the dog is destined to occupy an increasingly marginal position. For, if the first hypothesis (the influence of the Cratylus) were to be valid, then, given that the meaningfulness of the name is the consequence of an original relationship of iconicity, hence the articulatio, the voces of the animals, by common consent neither articulate or articulable, would not represent a subject of great interest. Animals are not aware that nomina sunt consequentia rerum, and they are not capable of imitating the nature of things.42

  4.2.7. Back to Thomas

  Clearly, at this point, we may skip the tradition of the grammarians. What interests us instead is the tradition of the philosophers, who continue to grant the dog and his bark a position of honor in the classification of signs. This is also because the philosophers, in addition to the classifications they elaborate, following the lead of the De interpretatione, are constantly induced to make supplementary observations. Take Thomas, who, in Sententia libri Politicorum (I, I/b), comes back once more to the difference between human and animal voces. Since, he affirms, nature never does anything gratuitously but always has a definite purpose, it is obvious that, although various animals possess a “voice,” only humans possess a locutio and, though there may be animals capable of repeating human words, we cannot say that they talk, because they do not understand what they are saying, but utter the words they have learned out of mere habit. Animal “voices” serve to express sadness or delight and other passions (and once again the barking of the dog is cited and the roar of the lion: “et haec sibi invicem significant per aliquas naturales voces, sicut leo per rugitum et canis per latratum”), while humans, instead of these voces, use interjections. But only human locutio is able to signify things useful and harmful, just and unjust, good and evil.43

  Here Thomas takes a step forward. He recognizes that, just as humans have ways of signifying to each other, alternately and intentionally, sadness and delight, the same is true of animals, and he thus touches on a problem that will be treated at greater length by Roger Bacon, who will distinguish between the moan that the sick man utters inadvertently and the interjection that he utters intentionally, following a certain linguistic convention, to signify the same pain, in however conceptually imperfect a manner.44

  But, in this way, within the same Thom
ist system, the latratus canis changes position, as if, halfway between the voces significativae naturaliter (among which we find the gemitus) and the voces ad placitum (where we find spoken language), we were to locate an intermediate zone, in which humans produce (paralinguistically, we would say today) interjections, while dogs bark. In fact, in this revised classification, the real difference between human and canine language lies not in the opposition intentional/unintentional (vaguely touched upon, but basically eluded), and not only in that between natural and ad placitum, but in the opposition between the interjection and the ability of human language to express abstractions by means of which humans set up domum et civitatem (“ergo homo est naturale animal domesticum et civile”)—an affirmation that Thomas takes up from Aristotle’s Politics 1253 at 9–30, where Aristotle opposes human language, capable of producing concepts and abstractions, to the inarticulate sounds of animals, expressive merely of pleasure or pain.

  4.2.8. Roger Bacon