In its analysis of communication, information theory considers messages as organized systems governed by fixed laws of probability, and likely to be disturbed either from without or from within (from the attenuation of the text itself, for instance) by a certain amount of disorder, of communication consumption—that is to say, by a certain increase in entropy commonly known as «noise.» If the meaning of the message depends on its organization according to certain laws of probability (that is, laws pertaining to the linguistic system), then «disorder» is a constant threat to the message itself, and entropy is its measure. In other words, the information carried by a message is the negative of its entropy.8
To protect the message against consumption so that no matter how much noise interferes with its reception the gist of its meaning (of its order) will not be altered, it is necessary to «wrap» it in a number of conventional reiterations that will increase the probability of its survival. This surplus of reiterations is what we commonly call «redundancy.» Let’s say I want to transmit the message «Mets won» to another fan who lives on the other side of the Hudson.
Either I shout it at him with the help of a loudspeaker, or I have it wired to him by a possibly inexperienced telex operator, or I phone it to him over a staticfilled line, or I put a note in the classic bottle and abandon it to the whims of the current. One way or another, my message will have to overcome a certain number of obstacles before it reaches its destination; in information theory, all these obstacles fall under the rubric «noise.» To make sure that neither the hapless telex operator nor a water leak is going to turn my victorious cry into the rather baffling «Met swan,» or the more allusive «Met Swann,» I can add «Red Sox lost,» at which point, whether the message reaches my friend or not, its meaning will probably not be lost.
According to a more rigorous definition, «redundancy,» within a linguistic system, results from a set of syntactic, orthographic, and grammatical laws. As a system of preestablished probabilities, language is a code of communication. Pronouns, particles, inflections—all these linguistic elements tend to enrich the organization of a message and make its communication more probable. It might be said that even vowels can contribute to the redundancy of a message, because they facilitate (and make more probable) one’s ability to distinguish and to comprehend the consonants in a word. The sequence of consonants bldg suggests the word «building» more clearly than the vowels uii; on the other hand, the insertion of these three vowels between the consonants makes the word easier to utter and to understand, thus increasing its comprehensibility.
When information theorists say that so percent of the English language consists of redundancy, what they mean is that only so percent of what is said concerns the 5o percent is determined by the statistical structure of the language and functions as a supplementary means of clarification. When we speak of a «telegraphic style,» we generally refer to a message that has been stripped of most of its redundancy (pronouns, articles. adverbs)—that is. of all that is not strictly necessary to its communication. On the other hand, in a telegram the lost redundancy of the message is replaced by another set of conventions also aiming at facilitating its communication by constituting a new form of probability and order. Indeed, linguistic redundancy is so dependent on a particular system of probability that a statistical study of the morphological structure of words from any language would yield an x number of frequently recurring letters which, when arranged in random sequences, would reveal some traits of the language from which they have been taken.°
Yet this also means that the very order which allows a message to be understood is also what makes it absolutely predictable—that is, extremely banal. The more ordered and comprehensible a message, the more predictable it is. The messages written on Christmas cards or birthday cards, determined by a very limited system of probability. are generally quite clear but seldom tell us anything we don’t already know.
The Difference between Meaning and Information
All of the above seems to invalidate the assumption, supported by Wiener’s book, that the meaning of a message and the information it carries are synonymous, strictly related to the notions of order and probability and opposed to those of entropy and disorder.
But, as I have pointed out, the quantity of information conveyed by a message also depends on its source. A Christmas card sent by a Soviet official would, by virtue of its improbability, have a much higher information value than the same card sent by a favorite aunt. Which again confirms the fact that information, being essentially additive, depends for its value on both originality and improbability How can this be reconciled with the fact that, on the contrary, the more meaningful a message, the more probable and the more predictable its structure? A sentence such as «Flowers bloom in the spring» has a very clear, direct meaning and a maximal power of communication, but it doesn’t add anything to what we already know. In other words, it does not carry much information. Isn’t this proof enough that meaning and information are not one and the samething?
Not so, according to Wiener, who maintains that information means order and that entropy is its opposite. Wiener, however, is using information theory to explore the power of communication of an electronic brain, in order to determine what makes a message comprehensible. He is not concerned with the differences between information and meaning.
And yet, at a particular point in his work, he makes an interesting declaration: «A piece of information, in order to contribute to the general information of a community, must say something substantially different from the community’s previous common stock of information.» To illustrate this point, he cites the example of great artists, whose chief merit is that they introduce new ways of saying or doing into their community. He explains the public consumption of their work as the consequence of the work’s inclusion within a collective background—the inevitable process of popularization and banalization that occurs to any novelty, any original work, the moment people get used to it.’°
On reflection, one sees that this is precisely the case with everyday speech. whose very power of communication and information seems to be directly proportional to the grammatical and syntactic rules it constantly eludes—the very same rules deemed necessary to the transmission of meaning. It often happens that in a language (here taken to mean a system of probability), certain elements of disorder may in fact increase the level of information conveyed by a message.
Meaning and Information in the Poetic Message
This phenomenon, the direct relationship between disorder and information, is of course the norm in art. It is commonly believed that the poetic word is characterized by its capacity to create unusual meanings and emotions by establishing new relationships between sounds and sense, words and sounds, one phrase and the next—to the point that an emotion can often emerge even in the absence of any clear meaning. Let’s imagine a lover who wants to express his feelings according to all the rules of probability imposed on him by his language. This is how he might speak: «When I try to remember events that occurred a long time ago,
I sometimes think I see a stream, a stream of smoothly flowing, cool, clear water. The memory of this stream affects me in a particular way, since the woman I then loved, and still love, used to sit on its banks. In fact, I am still so much in love with this woman that I have a tendency, common among lovers, to consider her the only female individual existing in the world.
I should add, if I may, that the memory of this stream, being so closely connected to the memory of the woman I love (I should probably mention that this woman is very beautiful), has the power to till my soul with sweetness. As a result, following a procedure that is also fairly common among lovers. I like to transfer this feeling of sweetness to the stream that indirectly causes me to feel it, and attribute the emotion to it as if the sweetness were really a quality of the stream. This is what I wanted to tell you.
I hope I have explained myself clearly.» This is how the lover’s sentence would sound if, afraid of not being able to communicate exactly what he wants to say, he were to rely on all the rules of redundancy. Although we would certainly understand what he says, we would probably forget it shortly thereafter.
But if the lover were Petrarch, he would do away with all the conventional rules of construction, shun all logical transitions, disdain all but the most daring metaphors, and, refusing to tell us that he is describing a memory but using the past tense to suggest it, he would say: «Chiare, fresche e dolci acque—dove le belle membrapose colei the Bola a me par donna» («Clear, fresh and sweet waters where she who alone to me seems woman rested her lovely limbs»).» In fewer than twenty words, he would also succeed in telling us that he still loves the woman he remembers, and would manage to convey the intensity of his love through a rhythm whose liveliness imbues
