May 5, 1981
We had assigned in passing three diagrammatic positions… Either the diagram tends toward seizing the whole painting, displaying itself in the whole painting… generally speaking, the so-called “Expressionist” tendency. Or else a second diagrammatic position, the diagram existing there but reduced to a minimum, overlaid… by the domination of a veritable “code”… perhaps, it seemed to us, the tendency known as “abstraction” in painting. And in third diagrammatic position,… the emergence of something that escapes from the diagram. And this “something”… no more than in the other two cases, is not a resemblance or a figuration, not something figurative. But what we can call a “figure”, a non-figurative “figure”, that is, which does not resemble a thing. A figure escapes from the diagram.
Seminar Introduction
Following nearly a decade teaching material included in his and Félix Guattari’s A Thousand Plateaus, and also in his first year at St. Denis, Deleuze turned his attention, first, to Spinoza, and then to eight sessions on painting and its intersections with philosophy. These newly transcribed, annotated and now translated sessions are devoted to diverse questions: What relation does painting establish with catastrophe and with chaos? How does a painter ward off greyness and attain color? What is a line without contour? What is a plane, an optical space, a regime of color? Why paint today? Through works by Cézanne, Van Gogh, Michelangelo, Turner, Klee, Pollock, Mondrian, Bacon, Delacroix, Gauguin and Caravaggio, Deleuze entertains important philosophical concepts, such as the diagram, code, modulation, the digital and the analogical, renovating these concepts in dialogue with his students.
English Translation

After outlining the previous discussion of three “diagrammatic positions”, Deleuze recalls the problem of the eye-hand relationship, asserting the manual aspect of diagrams in painting, and points out his disagreement with American critics who describe expressionism as “pure optical space”. Then, the second diagrammatic position consists of dealing with a code that’s difficult to decipher, delineating a contour determining a tension (e.g., Kandinsky codes immanent to painting). Deleuze then examines the nature of codes (meaningful and discontinuous units, in binary relations or choices) and how they manifest in languages, i.e., as articulation. Then, turning to a third, middle path rendering the eye of the hand-eye relationship into a “third eye”, Deleuze indicates that to define painting, he must consider the difference between diagram and code, but then considers and rejects several approaches for defining analogy (e.g., Peirce’s distinctions of symbol and icon; the uses of code, e.g., for making stories and illustrations; computer binary code in this context; analogy simply as similarity). He finally draws on Bateson’s hypothesis of analogical language as used for relations, opposed to the conventional language of codes for states of affairs, and then he shifts to a third determination for analogical language. With reference to linguistic approaches to language, Deleuze proposes the opposition of articulation and modulation, i.e., the latter referring to the values of a non-articulated voice, thus concluding that painting is to modulate and, as a medium, is modulated based on a signal, the motif or model, modulating light and/or color, resulting in the figure on a canvas. From this, Deleuze suggests that the diagram is the matrix of modulation, the modulator, and to develop modulation into a concept, Deleuze draws first on literary information (Rousseau’s On the Origin of Languages), then through music, moving toward harmony’s demands, in contrast to the voice’s modulation as defining its non-articulated status. Rousseau’s idea is that language comes from passion, in contrast to modulation, and Deleuze closes with the question: what will become of the melodic voice?
Gilles Deleuze
Painting and the Question of Concepts
Session 04, 05 May 1981
Transcriptions: Parts 1 & 2, Paula Moore (duration 1:07:16); Part 3, Guy Nicolas (duration 46:45); Part 4, Sandra Tomassi (duration 40 :32) ; time stamp and additional revisions, Charles J. Stivale
Translation by Billy Dean Goehring
Part 1
… We tried to categorize paintings base on the position of what we were calling a “diagram.” And so, we said this diagram—the diagram—we were looking to flesh out our understanding of diagrams. We said that a diagram could assume several positions. And that perhaps some pictorial categories could be defined as positions of the diagram—to put it in a more complicated way, as diagrammatic positions. And that it was necessary to lay down these pictorial categories wasn’t motivated by a concern over figuration, but was actually— maybe, rather, it might have been—based on the position of the diagram. So, we delineated three diagrammatic positions.
The diagram—these were tendencies for the different positions. They were position-tendencies. The diagram can tend to take up the entire painting, spreading over the whole painting. Broadly speaking, this seemed to be the tendency with so-called “expressionism.”
Or, in the second diagrammatic position, the diagram is there, but it’s kept to a minimum, and it tends to be replaced or “crowded over,” dominated by a genuine code. Notice that this gets complicated, but we’re playing it loose with our wording—because we haven’t said anything yet on what a diagram is, or what a code is. We’re just trying to lay out our terms, our categories. And in this second tendency, keeping the diagram to a minimum—the diagram is and continues to be the real seed of the painting—but keeping the diagram to a minimum and the substitution or application of a code: we figured that might be the tendency of “abstraction” in painting.
And then the third diagrammatic position: the diagram neither takes up the entire painting nor is it minimized. It’s like a rather exterior path; you might call it “a subdued path.” It’s there. It acts like a diagram, but it doesn’t take up the whole painting simply because the diagram fully realizes its effect, namely, it summons something out of the diagram. And this “something” that emerges from the diagram isn’t a resemblance or figuration, isn’t anything figurative—no more so than with the other diagrams. We can call it a “figure,” a non-figurative “figure,” that is, one that doesn’t resemble anything. A figure emerges from the diagram.
But what I examined last time—and we were nearly there, I almost finished—was the first tendency or the first position, the expressionist position. And I said, see—right away, we’re introducing an idea we’ll also have to try and develop—it’s as if the diagram were developed amidst overwhelming interference. Interference. Why this sudden interest in the concept of interference? Because our three diagrammatic positions are: the diagram that stretches out until it becomes genuine interference; second, the diagram that’s crowded over or determined by a code; third, the diagram that works as a diagram.
Well, but if we’re going to get to a logic of diagrams, we still have a lot left to do. So, in order to wrap up the first position: do you remember what it consisted in? What is this diagram that eats up the whole painting? Take your pick: either lines, line strokes [le trait ligne], or blotches of color—the two main pictorial elements that don’t trace contours. Either lines with no contour or blotches with no contour. Expressionism necessarily achieves a level of abstraction far beyond that of so-called “abstract” painting.
Because obviously all painting is abstract. But where things get interesting is when we look at how it’s defined in any given movement, the definition of abstraction corresponding to each movement. It’s obvious that, for an expressionist, abstract painters—again, I’m not about to say that one is better than the other; I’m just trying to figure out our categories. It’s obvious that, for an expressionist, so-called “abstract” painting doesn’t suffer from being too abstract. It suffers from not being abstract enough. How so? Because as abstract as they might be, their lines still trace a contour. Their lines are still outlines; in abstract painting you can easily make out circles, half-circles, triangles, etc. And in the most abstract Kandinsky, you can still make out triangles, i.e., a particular contour. Maybe not always, in Kandinsky’s case. But perhaps he’s not just an abstract painter. And in a Mondrian, you’ll find his famous squares, etc. These are all contour lines.
So, in a way, expressionists could say, “We are the true abstract painters.” How so? Because for them, their problem—and indeed, it’s a problem in painting that… — I believe they were the first to pose in a conscious and deliberate way. That’s my way of protecting myself against the obvious objection that it had already existed in painting. Painting has actually always used and drawn lines with no contour. But we need to grasp what’s significant about—what is a Pollock line? Well, all we can say about it is that it’s a line that constantly changes direction and doesn’t follow a contour. But what’s important about… or Morris Louis’s stain painting, these are all painters specifically known as “abstract expressionists”. Well, they have no contours. These stains[1] or lines have no contours, that is, they delineate neither an interior nor an exterior, are neither concave nor convex. They don’t go from one spot to the other, even virtually, but move between spots, the spots of paint Pollock throws down—it’s a snaking, breaking, convulsing line that constantly changes direction at every turn.
I was saying, think about what’s happening with how this line is laid out. It’s an odd sort of line because, ultimately, it’s a line that’s more than one-dimensional. In other words, it’s a line that’s almost commensurate to a plane. Consequently, this leads the plane itself to stretch and become commensurate to volume. In other words, it’s a line whose dimension could only be mathematically expressed as a fraction, between one and two. Whereas ordinary lines, which outline contours, are one-dimensional. Flat figures are two-dimensional. Volume is three-dimensional. Alright, it’s clear that abstract expressionism addressed the problem of depth in a totally new way. If you end up with fractional measurements, you end up with characteristically intermediate measurements between one and two, that is, between line and surface, and consequently between surface and volume. Ultimately the line takes up the entire painting, hence abstract expressionism is famously known as “all over” painting, i.e., from one end to the other, from one edge of the canvas to the other. Well, in this regard there is something left out. It is sort of a probabilistic approach to painting that refuses to privilege any particular position. Every part of the painting is equally weighted, whereas with classical approaches to painting there was always the center, the edges, and so on.
Well, as I was saying, see, I think it’s obvious that, when it comes to the problem that’s plaguing us, uh… which we’re still discussing: trying to figure out—since there are all kinds of things at stake in the concept of the diagram—trying to figure out the relationships between the eye and the hand in painting. I said, right, we have to assess the relationships between the eye and the hand in painting in keeping with our diagrammatic positions. At the very least, it’s worthwhile because, as I said, writings on the eye and the hand don’t seem to… what critics have written doesn’t appear to have fully accounted for the problem, for the tension there is in painting, at any rate, between the eye and the hand—the fact that painting is a certain resolution of this tension and requires the tension between the eye and the hand. All right. And you’ll recall that I really leaned into the fact that diagrams in painting are fundamentally manual. They’re arrangements of manual strokes [traits][2] and blotches. So maybe… obviously it produces something visual, but that’s not the point. When the diagram starts to take over, when it seizes and charges the entirety of the painting, the prevailing order is clearly a manual one.
And I think obviously so, when it comes to abstract expressionism. This more-than-one-dimensional line—a line that doesn’t outline a contour, with neither inside nor outside, which is neither concave nor convex—this line is a manual line. It’s a line the eye literally[3] has trouble following. It’s a line that the hand can trace only to the extent that it bucks its subordination to the eye. It’s a line expressing the hand’s rebellion against the eye. And how does this kind of conversion from eye to hand figure into abstract expressionism? The triumph of manual lines and manual blotches. This is reflected in the fact that—not to say that this is always the case—in how abstract expressionists have abandoned the easel. There are many ways to abandon the easel. After all, the canvas is never reducible to its place on the easel.
You see why I keep focusing on position or placement. Even when a painter paints with an easel exclusively, it’s obvious that the canvas is a lot better when it’s off the easel. But concretely, for so-called “abstract” expressionism—for Pollock, for Morris Louis, and so on, for [Kenneth] Noland—for all these painters: what is it that’s technically essential? Well, it’s the necessity that drives them… Pollock especially… to abandon the easel in order to paint on the ground [sol], to paint on the ground with an un-stretched canvas. Now, I think that’s really important. So, when American critics baptize this whole movement as “action painting,” what exactly do they mean? Well, they’re referring to what they consider to be a kind of frenetic action, where the painter flings paint, etc.…using sticks, basting syringes, etc.… while walking around the canvas at their feet.
What’s the significance of having “the canvas, un-stretched, on the ground” instead of “the canvas on an easel”? It amounts to a fundamental conversion. It means converting the horizon into the ground [sol]. It means passing from an optical horizon to the ground… to the ground under one’s feet. Well, feet… In this case, hands and feet are the same. The manual line actually does a good job of expressing this kind of… or it’s expressed by this kind of conversion of the horizon into the ground [sol]. The horizon is fundamentally optical. The ground is fundamentally tactile. Anyway, I said—this is where we left off—sure, but nevertheless, there’s one hiccup: American critics—especially those writing on Pollock and his followers… American critics are excellent, and I brought up two in particular: [Clement] Greenberg and [Michael] Fried. And they’ve written really, very beautiful pieces on this movement, on so-called “abstract expressionism.” But how do they go about defining it? They say, “It’s wonderful, and it’s modern.” And what makes it modern? It’s modern because it involves developing a pure optical space.
I mean, what bothers me… it bothers me that, if I’m being honest, I have exactly the opposite impression. I mean, I agree that Pollock is great; he’s really something else. Because for the first time, a purely manual line is freed from any visual subordination. For the first time, the hand is completely liberated from any visual directive. And here these critics are saying the exact opposite. So, it isn’t possible. So that presents us with one last problem.
Georges Comtesse: Perhaps that’s why American critics talk about a pure optical space when it comes to Pollock. Perhaps if you can’t see that there’s no contradiction between the manual line and pure space, it might be because of your concept of the pictorial diagram. Since you define the pictorial diagram as a hand detached from the eye, one the eye can’t keep up with, a rebellious hand. Okay, but in painting, in painting’s process of experimentation, the hand’s diagrammatic detachment from the eye, freer than the hand of the painter… there might be something else… you aren’t saying: the hand’s detachment from the eye… it’s specifically an optical machine of detachment that has nothing to do with the eye, the optical machine of the gaze, the painter’s gaze, which is neither the eye of perception, the sensitive eye, nor is it any possible eye whatsoever. There’s a gaze machine that… in the detachment, the painter… the painter’s hand is certainly still framed by this machine that’s irreducible to the eye, and which would certainly shift your concept of the pictorial diagram. The painter, meanwhile, … I don’t mean that the painter turns into this gaze machine while painting. But there’s like a sort of constant shifting when it comes to this gaze machine, which itself is primarily geared toward the task at hand.
Deleuze: All right! That’s one response. That’s our first possible response.
Anne Querrien: [Some inaudible words] … Everything that you are saying is perfectly correct regarding the painter’s point of view, the Action Painting, the action of painting, but there is a pure optical space which is created from the point of view of the passive affect of looking at painting which you don’t mention at all. You talk about the act of painting. And throughout the entire Kantian era, the Romantic era, the Impressionist era, etc., and I would say, from everything we’re taught in school tells us to look at painting by putting ourselves in the painter’s shoes. Thus, we had to have a tactile approach in how we view painting, to see how the layers were put down, etc., and furthermore, we were taught to paint in order to appreciate aesthetically others’ paintings, following Kant’s model of universal humanity [indistinct words] … but now there is a gap separating the positions exactly like in mathematical spaces [indistinct words] …, where the painter and the viewer are no longer in the same position when it comes to… and what’s more, canvases painted on the ground [sol], right, that we walk around, they aren’t displayed on the ground. They’re displayed in an optical space. They’re viewed vertically… [indistinct words] … [Interruption of the recording] [20:29]
Deleuze: Excellent! Okay, then, that’s perfect. That’s a second one… That’s good because I have a third response. But they don’t cancel each other out—to the contrary, we have to account for… there are that many fewer problems. What I’m wondering is: why do Greenberg, Fried, etc. … call space—in Pollock, in Morris Louis—why do they call it a “purely optical space”? We have to follow them carefully. They did so for a very specific reason. The reason is the following: such space is opposed to so-called “classical” pictorial space. So-called “classical” pictorial space is classically defined as a tactile-optical space. In other words, space in classical paintings is—as we’ll see later, we’ll come back to this point—is a tactile-optical space. Which means what? That it’s a tactile space with tactile referents on the canvas. What are these tactile referents? One example of a tactile referent: the contour. Why? Things have contours, but they have a tactile contour just as they have a visual contour. Yes and no. There is a tactile referent whenever the contour remains self-identical, no matter the degree of luminosity. You have so many remarkable paintings that develop a tactile space, and you can tell when there’s a tactile referent when, for example, you see a contour that’s still, say, intact… intact—the reference is tactile—intact, for example… under a bright light or in shadow. Whether perspective doesn’t also involve tactile referents—I think it’s obvious that perspective does involve tactile referents. So, it’s clear what we mean by visual space with tactile referents, and such a space would be “tactile-optical.”
That being said, it’s clear that the line without contour breaks with any tactile reference. There is no more form; there is no more tactile form. The tactile-optical form is thus decomposed into a line without contour. So, I think that when American critics define abstract expressionist space as an optical space, they mean that it’s a space which has cast aside all of its tactile referents. Now, stay with me; let’s take this literally. A space whose tactile referents have been cast aside. Okay. Does that settle the matter? Is this what a purely optical space is, then?
Despite both these analyses just now, you can see why I need—why I’m emphasizing this additional, uh… or I’m adding it…It’s funny, because I almost feel like it’s the other way around. There is a pictorial direction or movement that achieves a purely optical space, but it isn’t expressionism. It’s abstract painting. In abstract painting, you get something that could actually be called a pure optical space. But not at all in expressionism. Why not? It’s true that all tactile referents are eliminated. But why is that? It’s not because space has become optical but because, once again, because the hand has managed to become independent from the eye. Because now it’s the hand that imposes itself on the eye. Right. It’s the hand that imposes itself on the eye like a foreign power that, again, the eye struggles to follow.
Consequently, tactile referents, which express the hand’s dependence on the eye, are effectively suppressed. Not because it’s a pure optical space, but because the hand is no longer subordinate to the eye and breaks free. So, it’s because it’s a pure manual space that tactile referents—which express the hand’s subordination to the eye—are driven out, cast out from the canvas. But again, it was enough for me that there weren’t any contradictions.
Anne Querrien: It’s possible there are both at the same time?
Deleuze: Of course, of course… Yes, but then, uh…
Anne Querrien: A pure and manual optique…
Deleuze: Okay… you say it’s… suddenly it becomes purely optical, but that’s a different question. It becomes purely optical from the viewer’s perspective. Sure, but at that point… Personally, at any rate, that’s an issue I can’t get into yet, because… we’d have to figure out what optics, what sort of optics come from the hand, are produced by a purely manual gesture.
Anne Querrien: No… it’s the idea that there’s no longer any communication anywhere, really, that there are no commands to place oneself in the painter’s position to look at painting, well, that seems really important! There’s a liberation of the eye of the viewer from the position of painting as well… whereas in all the education we’ve received at school, we’re told that you can only appreciate painting if you yourself are some kind of amateur painter (peintre du dimanche).
Deleuze: For the viewer themselves, that doesn’t change the fact that this optical conquest is still a conquest. Because the violence done to the eye remains. Thus, there is kind of a need for the eye to learn to accept this violence done against it.
Anne Querrien: It’s not necessarily a violence. I find that we are emerging from a kind of Hegelian dialectic in which there is only the active which is positive or negative, and we have an active affect of painting which is to paint, and a passive affect which is that of viewing… [Inaudible comments]
Deleuze [laughing, reacting to the inaudible remarks]: This line is not a calm line.
Another student: I think we have to approach it in terms of contemporary physics. Because there’s a transformation in optics that results perhaps in it becoming manual or something. I don’t know.
Deleuze: She’s saying, for those who can’t hear—she’s saying we’d have to account for a type of physics with which, in fact, some so-called “informal” expressionists align themselves, and which accounts for—for example, the whole physics of signals—which accounts, oddly enough, for some novel relationships between the optical and the manual. Yeah, okay, then.
Comtesse: Perhaps we shouldn’t forget that, in the major periods—Pollock’s first major periods—when compared, for example, comparing Jackson Pollock with Robert Motherwell, that all their problems, it’s not just the paintings or the new painting techniques. What was important was the pictorial line, pictorial creation, and this is exactly how they pose the question: as they related to the unconscious. It’s a crucial problem for Pollock and in, for instance, in Motherwell’s writings on Pollock. And the whole issue with Pollock over his botched interpretation of [Carl] Jung, and so on… painting was a way of revealing the unconscious. So that poses the problem of the relationship between the pictorial diagram and the unconscious, since they themselves keep posing it in their own artistic processes, the problem of this relationship, using just their lines.
Deleuze: Yes, Comtesse… that’s not entirely uniquely… certainly, that’s an apt description of diagrams, but it isn’t unique to expressionism because some will say, “the diagram, or its equivalent, is an instance of randomness.” Others will say, “it’s an instance of the involuntary.” Others will say, “it’s an instance of the unconscious.” Ultimately, all can agree that the diagram, the kind we first loosely defined as a chaos-seed, is sort of the unconscious of painting—yes, of the painter. See, that has so many ramifications, it’s perfect.
Moving onto the second diagrammatic position. This time it isn’t the extended diagram, what [Paul] Klee calls “the grey point that takes up the whole painting.” That’s not it. Instead, the diagram is only… it’s totally constricted, as if—it’s so complicated—the painter wanted to somehow suppress everything obscure about the diagram. Anything that’s, say, unconscious, involuntary, etc., etc. What about this tendency to reduce the diagram? My hypothesis—this will certainly get us tangled up and get us off track. These painters, for us, the viewers… you get a strange feeling, the feeling that once again we’ve reached painting’s boundary limit—but all painting is at the limit of painting—we’ve reached the boundary limit of painting because now we feel like we’re dealing with a sort of code we don’t know how to decipher. And what makes this form of painting verge on code? Once again, this is off the cuff. My initial thought: these painters are painters. They wouldn’t be painters if they applied a code or painted based on a code. That’s not what I mean. But it might be a fine line.
When a painting comes down to applying a code, what do you say? “Well, any computer could do that, obviously.” Any computer can turn out paintings using code; that’s easy. Anyway, that sort of nonsense is not what I have in mind with abstract painters. What I mean is that it’s as though we’re shown what was to serve as code in painting, as a uniquely pictorial code. So, we have to—I’m trying to get across how… To quote a 19th century painter who, incidentally, isn’t an abstract painter, strictly speaking, but I don’t think he’s far off. I’ll read the quote: “Synthesis consists of making all perceived shapes conform to the small number of shapes that we are capable of imagining. To the small number of shapes, we are capable of imagining: straight lines, a few angles, arcs of circles and ellipses.”[4] Isn’t that a sort of code—ultimately, a geometrical code? A geometrical code for which geometry is all that’s required. Geometry has a code. So, again, it’s not about applying geometric shapes.
Kandinsky clearly distinguishes between so-called abstract shapes and geometric shapes. He says—alright, this is what Kandinsky calls an abstract shape: “It’s a shape representing nothing other than itself.” All right. Then abstract shapes and geometric shapes appear to be the same thing. I mean, the triangles Kandinsky paints and triangles delineated geometrically both seem to qualify as “shapes that only represent themselves,” as opposed to concrete shapes [figures].[5] And he goes on: but [the abstract shape] “is a shape that’s internalized its own tension.” Tension is the movement that characterizes it. It has internalized its own tension. That’s what geometric shapes do not do. You see why an abstract painter—taking things one step at a time—why an abstract painter can say, “It’s abstract even though the line forms a contour, even though it has a contour.”
The problem is very different from that of expressionism: there is a contour, and yet the contour no longer determines a concrete figure; the contour only determines a tension. The contour no longer determines an object; the contour only determines a tension. For Kandinsky, that’s the pictorial definition of abstraction. And the idea of tension will be crucial throughout everything Kandinsky says about painting. Okay. What do we take away from this? What do we make of this tension? In Kandinsky’s writings, you constantly run into passages that allude to the invention of a code. What do I mean? In Kandinsky’s best-known writings, for example, he says—after a long inquiry, after extensive commentary—this is just the conclusion, so it’ll seem a little arbitrary—he says: “Vertical, white, active. Horizontal, black, passive or inert. Acute angle, yellow, building tension. Obtuse angle, blue, weakness.”
That’s an excerpt. There are long lists in Kandinsky’s work. You get the sense that it’s not just a table of categories. These are elements we’re talking about. Synthesis consists of making all perceived shapes conform to a small number of set forms.[6] What small number of forms? I’ll try to clarify this idea of a pictorial code.
You’ll notice that in Kandinsky’s case, it must be said that there isn’t one code. Nearly every abstract painter invents a code. Just like in language [langage], where there are all sorts of possible spoken versions [toutes sortes de langues], there are all sorts of codes in a virtual pictorial code, to the point that perhaps every abstract painter is the inventor of a code. So, how would we define this code? What would a code “immanent” to painting be? One that doesn’t exist in advance, waiting to be invented by one painter or another? I’ll use Kandinsky’s terms. See, he always has three criteria: vertical lines, horizontal lines, obtuse angles, right angles, etc., etc. And then he’ll use that to form squares, rectangles, circles, half-circles. So, there is line or form: the first category.
The second category: an active, passive dynamic. We could add to that. We could imagine a code with more than two values. There wouldn’t just be active and passive. There could be active, passive, baseline. I know of some painters who describe three rhythms, three fundamental rhythms: Active rhythm, which tends to grow. Passive rhythm, which tends to diminish. Baseline, constant rhythm. And it works on canvas: you have elements on a baseline, you have elements on a falling level, you have elements on a rising level.
So, all I’ll say is: the first category refers to lines or figures. A second category of dynamics, referring to the dynamic, referring to activity/passivity. And you have a third category with Kandinsky, which he never loses sight of, referring to a kind of affective disposition, sort of a category of affect. And then a category referring to color. For example: vertical, white, activity, joy.
What does that mean? Just what is a code? It seems like one of the criteria for a code is, for one thing, whether you can identify meaningful units that are discontinuous— discrete, as it were. A finite number of “discrete meaningful units,” which can be very large—it could be large or small, but it’s always a finite group of discrete meaningful units. I’m putting it in abstract terms for now. And the second condition is that these meaningful units ought to bear out—each one ought to bear out a number of binary relations. Actually, it isn’t just for the sake of convenience that codes are binary. There’s something crucial about binarity and code that binds them together. What do I mean by that?
I’ll take a familiar example, that of language. How might there be a code in language [langage]? Or how might “language” involve a code? Linguists have been telling us for a long time—first of all, that language breaks down into so-called “meaningful” units known as, for instance, “monemes.” [Pause] But these meaningful units can be broken down into smaller elements. These monemes, these meaningful units, are broken down into smaller elements called “phonemes.” And phonemes do not exist outside of binary relationships. Let’s say the meaningful unit is “vent.”[7] You mishear it. These are well-known, you know, these ubiquitous examples in phonology. Then I clarify: “I said vent, not ‘dent.’” A relation between V/D. It’s a binary relationship, a phonemic relationship. Not “bent”—a relation between V/B. Not “meant”—V/M. Etc., etc. These binary relationships are features [traits], what are called “distinctive features [traits distinctifs]” in linguistics. Such that phonemes strictly depend on the set of their binary relationships to other phonemes. Anyway! I said there’s a linguistic code because there are meaningful units which have the possibility of being broken down into elements caught up in binary relationships.
What have I just described? And this might help us later on, so I’d like to keep our thumb on it as we move along. It’s a detour I can’t avoid. In a way, what I’ve just described is the concept of articulation. Let’s back up. There is—as [André] Martinet tells us, there’s even a double articulation. [Pause] Language is articulated. Meaning what? Language is articulated—that doesn’t just mean that there are glottal movements that articulate language. It’s not solely a question of articulatory physical movements. Language is actualized via articulatory physical movements because it is in itself articulated. And what does it mean to be articulated? It means being composed of discrete units [Pause] which themselves refer to elements tied up in binary relationships. That seems to me like the best way to define code. But how does that help me? What does that do for me?
I could elaborate later on, but for now I’ll just give you my conclusion. That is, we tend to associate code with articulation. Are they interchangeable? Beats me—that’s not what interests me. In any case, I can say that there is some overlap between the two concepts, code and articulation. There is no inarticulate code. One last example. The idea of code has two basic components: (1) discrete units—a finite number of discrete units—and (2) these discrete units are selected according to a series of binary choices.
Why bring up the idea of “choice”? In order to account for the relationship between units and elements, between meaningful units and elements caught up in binary relations. To take a common example from computer science: how do you select “six” out of eight given numbers?[8] You select six based on three binary choices. Three successive binary choices. You take your set: one, two, three, four, five, six, seven, eight—ah, this is rough, it’s annoying because today’s when I have to get pretty abstract. So, there’s: one, two, three, four, five, six, seven, eight. The first binary choice: you divide your set in half. You select the right half, greater than four. You take that set, or subset—four, five, six, seven, eight—and divide it in half. You select the part that includes six. That gives you a subset with two terms.
The third binary choice: you select six. You’ll select the half that includes six. You’ll get a sub-set with two terms. The third binary choice: you select six. So, it’s always possible to reduce a code-based decision to a sequence of binary choices. I don’t need any more, I won’t go any farther. Code = articulation. Articulation = units determined by a series of binary choices. A unit determinable by a series of binary choices. So, what do I think is crucial about that? I’ll bring it back to painting—more specifically, to my hypothesis that abstract painting is the elaboration of a code to which the diagram itself is subjected. In the case of abstract painting, that actually works like this: you have a certain number of discrete units. That doesn’t mean that it’s easy to paint or anything. But it’s painting by code. It’s the invention of a properly pictorial code that only exists in painting, and that only exists insofar as it is invented.
And then painting would mean inventing a code, right, inventing a uniquely optical code. The idea of an optical code seems to lie deep at the heart of abstract painting. And according to abstract painters, that would be the modern understanding of painting. You’re presented with an internal, optical code. I briefly mentioned Kandinsky’s writings—how does it show up there? Meaningful units are definitely there. “White, active, vertical,” for example, is a meaningful unit. These meaningful units cannot be broken down into smaller units, but they can easily be broken down into elements subject to binary choices. What sort of binary choice? They’re choices about figure, choices about color, choices about affective disposition—active/passive.
You have a binary choice. You might say, “That doesn’t work for color.” Yes, with color there is a series of binary choices. Exactly like my computing example. And the entire color wheel itself is a kind of binarization of color relationships. The complementary relationships between colors, etc., intermediate colors, etc.—everything about the color wheel and its opposing relationships gives you a system of binary choices between colors. And that’s why I say that literally, for Kandinsky, there are two clear pictorial levels of articulation. On the one hand, meaningful units bundling a whole series of binary choices. Yet what do we call these binary choices that allow us to establish meaningful units? You know what they go by? They’re called “digits.”
And what’s going on with this word, “digit”? Binary-digit. What a great word. What is the “digit,” or finger? What role does the finger play here? It’s reduced to… a finger pressing a keyboard. The finger as a runaway simplification of the hand. It’s funny, the finger is what remains once humanity loses its hands. A finger pressing keys. What’s this about? It’s a handless humanity. The digit is the manual state of handless humanity.
What do I mean by that? I have in mind a passage from [André] Leroi-Gourhan. Leroi-Gourhan talks about future humans, and he says it’s a form of humanity that lies prone. It doesn’t need to move.[9] All right. It’s kind of science fiction. Humanity is more and more infantilized, and then it loses its hands. But humans still have one finger left to type with. In our future evolution, we won’t have hands anymore, right? We’d only have one hand, and we’d press stuff like this. [Laughter]
Which leads me to make one adjustment. With the “problems surrounding the eye/hand relationship” I brought up before, things get complicated because the hand alone can take so many forms… I could list out categories of hand. To begin—it’s pure conjecture at this point, but we’ll see it bear out later on — for starters, I’d make a distinction between the manual, the tactile, and the digital.
I’d say that tactile—allow me to offer a few suitable definitions; obviously, they’re standard, they’re standard definitions. That way, there won’t be any complaints. “Tactile” is what I’d call the hand subordinated to the eye. When the hand follows the eye’s commands, then the hand becomes tactile.
When the hand shakes off its subordination to the eye. When it imposes itself on the eye, when it does violence to the eye, when it strikes back against the eye—that’s what I’d call “properly manual.” And the digital, on the other hand, is the hand’s absolute subordination to the eye. It’s not even that the hand’s tactile qualities are enlisted in the eye’s service. The hand has dissolved; only a finger remains, for picking between visual binaries. The hand is reduced to a finger pressing on a keyboard. In other words, it’s the computerized hand. It’s the handless finger. In a way, isn’t that the “Ideal”? But in a very qualified sort of way: the ideal of abstract painting as a pure optical space. A pure optical space such that the hand is undetectable.
What does that mean, the hand is undetectable? Well, the hand is undetectable. It’s funny because you run into this expression everywhere in painting. Painters say to each other: “What a beautiful painting—you can’t detect their hand.” In other words, not being able to detect the hand seems to be a flaw. Is it possible for the hand to be undetectable? Isn’t abstract painting the painting of a handless humanity? What would that mean? Clearly not—it’s not that. What makes us certain that it isn’t?
When it comes to distinguishing a fake Mondrian from a real Mondrian, what do they do? There’s a famous passage from a critic on this.[10] What do they distinguish, what do you look for to tell if it’s fake or not? Critics say: It’s not very hard, ultimately, with a little practice, if you get the square up close: you’re told that this is a Mondrian, you look at where the square’s two sides overlap, and you see what’s going on with the painting’s layers, with the overlap. You’ll likely notice that it’s a little… there’s a good chance you’ll notice, especially if the square is colored, if the layers of color overlap. Then you might be able to tell whether or not it’s a forgery. Which means you can detect the hand. But it’s worse in Mondrian. But with Mondrian it’s worse—worse or, uh, better, depending how you look at it—than Kandinsky. Because Mondrian has a sort of fantasy—Mondrian has some theoretical writings—of boiling everything down to two binary units. Now, it’s a kind of code that exaggerates the horizontal and vertical. On multiple occasions Mondrian claims, with the utmost austerity, in all his spiritual asceticism, to reach the point where everything is depicted via horizontal or vertical lines. Nothing else is required.
What does he mean when he says nothing else is required? See, from the perspective of code, that’s code’s ideal. Because as I was saying, a code—normally, a code is a finite number of meaningful units, i.e., more than two, that are determined following a series of binary choices. The supreme ideal for code is for it to only have two meaningful units and therefore only one binary choice. Then you’d have a code of code. The code of code is when instead of a set number of meaningful units determinable by a series of binary choices, you only have one binary choice between two meaningful units.
Yet that’s never what Kandinsky tried to do. But Mondrian goes a long way with the horizontal / vertical, with this high level of pictorial asceticism: “With a horizontal line and a vertical line, I can give you the world,” “the world in its abstraction.” It’s all there. Just to give you a sense—you see it already—a sense that, ultimately, aesthetic categories are well-founded, but everything’s mixing together. Because among the finest writings on Mondrian are those by Michel Butor.[11] There’s something Butor wonderfully demonstrated: that Mondrian’s squares, for example, very often do not have the same thickness in length or breadth. That’s obvious, actually. He didn’t need much to demonstrate that.
However, this difference in thickness has a very peculiar optical effect: it’s that the now crucial and even increasingly crucial intersection—between the thinnest length and the thickest breadth, for example—now this intersection will determine a virtual line. There’s something very odd—just as one talks of a virtual line in music, here we have a virtual line in painting. The fact that both sides of the square don’t have the same thickness makes the eye, the eye of the viewer, follow a diagonal line that Mondrian doesn’t need to draw himself.
This virtual line, this virtual diagonal—what can I say about it? I might call it an abstract version of a line with no contour since it isn’t traced by the painter. Likewise, what makes Kandinsky so complicated is that he has some wonderful paintings where you get these elements, these meaningful units. But they are strangely traversed by lines that, in Kandinsky, have a very simple source, a really gothic source, by lines—nomadic lines, really—lines with no contour. By lines that pass between figures, that pass between points, that have neither beginning nor end, and that are characteristically “expressionist” lines, and that nevertheless do not manage to disrupt the painting’s harmony or rhythm. That tells you that these categories, I think—it’s not because things overlap that the categories aren’t well formed.
All that I’m trying to say is that, when it comes to abstract painting, the one who went the furthest, I think, is the painter I told you about, who I think is a tremendously great abstract painter: [Auguste] Herbin. And Herbin goes a long way. He just invents a code. He doesn’t borrow a code from nearby. What do I mean when I say, “he invents a code”? He does something he calls… he calls his painting a “plastic alphabet.” A plastic alphabet—and it takes four forms, four basic forms. For him, there are four meaningful units: the triangle, the sphere, the hemisphere, the quadrangle (the rectangle and the square). He has four forms. He has four forms for units, and he puts them through binary relationships, a bit like Kandinsky: this time with regard to color, with regard to affective disposition, and then he adds—is this just embellishment, or is something more profound happening here?—He adds letters of the alphabet, such that he’ll often base his paintings on them, or vice versa, he’ll determine a painting’s title according to the letters determined by his pictorial units. For example, he titles a painting Nu. And you have to break down Nu into N – U. Then see what plastic form corresponds to N, what plastic form corresponds to U, what color, etc., etc. You know, a bit like how Bach played with the word “Bach” in music. So, he takes the idea of a “plastic alphabet” pretty far.
What’s the upshot of all this? In fact, it turns out masterpieces. He’s a great colorist. Is that just tacked on? No. I have a hard time imagining how one could “look” at abstract painting without “looking” at it not as applying a preexisting code but as the invention of an optical code. Again, this optical code being based on double articulation. First, meaningful units articulated pictorially. Second, the articulated elementary binary choices that determine these units. So, we get a definition of a sort of “pure coded optical space,” where the hand ultimately tends—but it’s only a tension—where the hand tends to give way to the finger, to a digital space.
Anyway, as I was saying, there’s a third option. How does it go? See, my two diagrammatic positions are opposed to each other, point-for-point… [Interruption of the recording] [1:07:13]
Part 2
… [-dage] and code,[12] hence the response of these painters, who appear to take the third route, i.e., one that’s sort of moderate, middle-of-the-road—but only ostensibly moderate and middle-of-the-road… only nominally.
Just who are these painters? To borrow a term… terms, for example, I’m borrowing terminology from [Jean-François] Lyotard that I think are very fitting and very… when Lyotard contrasts the “figurative” with what he calls the “figural,” it’s not figurative painting because in fact there is no figurative painting—again, it’s a “figural” painting.[13] That is, I’m taking diagram in the full sense, with the exception that…that is, in particular, I don’t intend for it to be a code, and at the same time, I’m preventing it from overwhelming the painting, from muddling the painting. In other words, I’m using the diagram in order to produce the pure “figural” or figure.
Anyway, see, I’m going to get completely new hand/eye relationships. It will no longer be the hand opposed to the eye or imposing itself on the eye as with expressionism, broadly speaking. It will no longer be the eye reducing the hand to the point that only a finger remains. What will it be? A hand/eye tension such that the manual diagram makes… what emerges? There’s only one answer since it’s not a figurative figure.
It’s that it gives the eye a new function, that the hand forces the eye to take on a new function, i.e., a real third eye. The hand causes a third eye to emerge, right—which tells you that this isn’t a moderate path. It’s only moderate compared to the other two, meaning that it doesn’t stretch the diagram over the entire painting, nor does it submit the diagram to a specifically pictorial code. Otherwise, it comes with many risks, including the double risk of verging into the abstract or into expressionism instead of forging its own path. But that would be a third path, the pure diagram’s position, the purely diagrammatic position.
So, at the point we’ve reached, well, today we have to… even though it would take up a lot of our time today, a long detour, I’d like to get to the point where we can propose—everyone is entitled to one—a definition of painting. There are so many possible definitions, so everyone can give their own… we could play that game, sure, at this point, we have to kind of forget painting. I told you that my goal was twofold: my goal was to talk about painting, but I also wanted to sketch out a theory of the diagram. Well then, at the point we’ve encountered, it’s fine, now we have to try to manage with: well, what is a diagram? And what is the difference between a diagram and a code? What’s that about? The matter of diagram and code? And for me, that’s what I’d like to manage to derive from this a sort of definition of painting.
If it’s true that painting is diagram, what is the relationship between a diagram and a code? By no means am I trying to say… it’s certainly a complicated relationship, since painting absolutely includes the endeavor to invent optical codes that seemed characteristic of abstract painting, right, as a diagram… so we’re back to square one, very well, let’s start over. As a diagram, we’re back in a purely logical element now, it’s this element of pure logic that will get us back into painting.[14]
I’m saying that the couple as a diagram, there is another couple after all, we have to use everything. We looked at the digital—digital, a code is digital in the sense I worked out before: what we call “digital” is the binary choice behind the unit. A code is digital; you’ll grant me that much. And typically, in every theory of information and even linguistics, what is the opposite of digital? Analog. Analog and digital—synthesizers today, for example, are either analog synthesizers or digital synthesizers. The processes of retransmitting signals are either analog processes or digital processes. Well, it’s still technological, but it isn’t anything complicated. It’s a distinction today regarding code/diagram—I’ve got my two pairs: code/diagram, digital/analog. What does this have to do with painting? Why do I sound like I’m talking about something else if I’m not talking about something else?
Is painting a language [langage], or is it not a language? Personally, I find this question interesting. What I’d also ask is: what is an analogical language? Not so easy to define an analogical language—is there an analogical language? Is painting an analogical language? Is painting the analogical language par excellence? What else could it be? Cinema? Is cinema an analogical language? We ought to consider the ideal setting: when film was silent. Or when film had sound but no talking—is that an analogical language? After all, everyone in silent film thought they had invented what they themselves constantly referred to as a “universal language.”
Silent film as a “universal language”—hence their frustration when… When talkies came onto the scene, all of cinema’s pretensions as a universal language were called into question. Okay, and is there a connection between painting and silent film? Maybe… maybe there is, but never where you think. We all know that film is at its worst when a director thinks they can make a scene or a shot as beautiful as a painting—it’s a disaster—everyone cracks up, everyone falls asleep, or else it only works when it’s funny, when it’s [Luis] Buñuel. But with Italian cinema, for example, you get scenes that make you think, “Oh jeez, what a disaster!” right out of—you think that, but it’s a straight up Raphael. It doesn’t get any cheesier in cinema than that. So, if film and painting have anything to do with each other, it’s not in that regard.
All right, analogical language—what is it? All I can say is that we already know a bit about what makes a code “digital.” And what the expression, “digital code,” means is that code is the basis for a digital language. Like Americans always say: language is digital. What does it mean that language is digital—it means, see, it doesn’t mean that it’s done with fingers like a sign language for the deaf; it means something very specific: “language is constituted by meaningful units determinable through a sequence of binary choices.” That—that’s what it means to say that language is digital. Well, are there analogical languages? And if so, how should we understand analog in that case? … What time is it? What’s that?
Claire Parnet: It’s 12.
Deleuze: Do the two ever mix? If there are analogical languages, how do we define them? Would painting be an example? Would painting be the analogical language par excellence? Why not pantomime? Why not all of the visual arts? How would painting stand out from the rest? On the other hand, will it suffice to draw a crude distinction between digital code and analog language—what would that be?
Well, there’s no avoiding it, let’s see our hypothesis through to the end. It’s the diagram that would be analogical: the analog diagram and the digital code. But would this be a simple opposition, or would there be some code grafted onto analogical language, onto analog diagrams? Here we have a whole series of confusing problems surrounding the diagram’s logic. So, I’ll try to be brief because I want to get back to painting, but I’m starting with a initial approximation. Digital code would imply “convention”; analog diagrams or analogical language would be a language of “similarity.” So, my two concepts would be distinguished in the following way: similarity for analogy, or for the diagram; convention for digital code.
See, that doesn’t take us very far—why I say that is because the notion of the diagram and its extension and its eruption into logic, into philosophy, resulted from the general approximation laid out by an incredible author (whom I’ve already discussed with you in past years): [Charles Sanders] Peirce, P-E-I-R-C-E. An English-speaking logician who invented a discipline that went on to enjoy great success, semiology, which was based—I’ll stick to what’s relevant for our concerns here and now—which was based on a very simple distinction between what he called icons and symbols. He was saying that icons have to do with similarity, generally speaking. An icon – an icon? [Deleuze asks about “un” or “une” for the noun gender; students respond with both] – an icon is determined by its similarity to something. A symbol, on the other hand, he said, is inseparable from a conventional rule.[15]
You might say we don’t need to refer to Peirce because this doesn’t add all that much, but this is Peirce’s starting point, and he takes it even further. And he takes it further only in order to put this validity into question. Then what can I say? I start from this simple problem, do I define analogical language by similarity and coded or digital language by convention? Recall the Saussurean point [that] language is conventional, etc. The linguistic symbol is conventional—immediately it’s obvious that it’s not. But what’s interesting, what ought to interest us, are the reasons for which this first dualism is so, so inadequate. And I’ll be brief, but here you need a sense for their order since we’re tentatively dipping into an area of logic.
There are two reasons for saying that this similarity/convention duality isn’t satisfactory. It’s because on the one hand, there are instances of similarity in code, and on the other hand, similarity isn’t sufficient for defining analogy. Those are my two points, the two points which I’d like to unpack.
The first point is that we can’t simply oppose convention and similarity because a code necessarily includes—I’d almost say that it necessarily produces instances of similarity. I mean, it’s clear in Peirce’s work; here’s what Peirce says on the matter. I’ll paraphrase; it’s a very complex thought—gorgeous, really, but I’ll only paraphrase. Basically, Peirce says that there are two sorts of icon based on similarity: there’s a similarity in quality, of similar quality—for example, you paint with blue because the sky is blue; that’s a qualitative similarity, where you look for the blue that’s closest with the blue of the sky. And then there’s a similarity which is that of relation, so there are particular icons that are icons of relation.
Now what he calls diagrams are icons of relations; thus, you can see why it interests us but, but, but he maintains a definition of the diagram that depends on similarity. That’s why, for our purposes, we cannot follow him, and all of the Americans afterward who developed theories of the diagram have hung onto Peirce’s “iconic” principle, that is, the diagram as primarily defined by a similarity in relation. Which is what leads Peirce to think that the exemplary diagram or diagrammatic process is algebra. Algebra—he says that algebra isn’t actually a language because it is an icon, so it’s a matter of similarity in relation. The algebraic diagram extracts similarities in relation. Okay, and at the same time he adds that, on the other hand, algebra as such is not separable from certain conventional symbols that belong to the other pole. Which implies a code, that is, at which point Peirce is aware of mixtures of code/analogy or code/similarity. All right, I said why, then, we weren’t going to follow Peirce too closely.
Coming back to my question, the first one: there’s one main reason that the diagram cannot be defined by similarity, to wit, I cannot imagine a code that doesn’t involve or produce instances of similarity inseparable from it.
In fact, what can you do with a code? As I see it, there are two things done with code: you can tell stories; you can make illustrations. With a code, you can do three further things: you can make sub-systems; you can make codes; you can make sub-codes—but that doesn’t tell us very much. What could you do with code? So, we can make stories and we can make illustrations.
Right, keeping things simple, how do you make an illustration with a binary code? A distinctly digital exercise—so again, not a pictorial code—a computer can give you a portrait. All you have to do is encode the model’s data according to a purely binary code consisting of 0/+ or 1/0—the binary system. Your computer can be programmed to render the portrait.
Thus, code as such, the simplest binary code, can give you—for example, current computers—a huge range of illustrations. All you need is to encode the details, encode the data. But what does encoding data imply? Binarization: fundamentally it implies binarization; if you binarize a figure you can very easily render it by computer.
Now in this case, I’d say that there is a resemblance produced by way of a code and encoding. Code more commonly—especially when it comes to language—results in stories rather than illustrations. What do I mean? In my first example, the computer generating a portrait once it was programmed to, you have a direct connection between the encoded program and the end result.
In language, what differentiates language from a computational function? It’s that with language you necessarily have a third term, as linguists say: you have the signifier, you have the state of affairs, but in an illustration, you have a signifier that produces a state of affairs—the encoded signifier.
It doesn’t work that way with language. What characterizes language is precisely a third entity, the signified—the signified. The signified is not the same thing as the designated state of affairs. Yet what to make of the well-known principle: “linguistic symbols are conventional”? It’s been said by all sorts of linguists; what does it mean, exactly? It means that there is no similarity between—what? Between the signifying word, the signifier, and the designated state of affairs. There is no similarity between the word, “cow,” cows, and the cow’s state—their relationship is purely one of convention, that is, it’s by convention that this morpheme designates the thing with horns, etc.
On the other hand, the word “signifying unit” has a signified—what is the signified? It’s the way in which the state of affairs appears in correspondence with the word. Consider languages where there are two words for cattle depending on whether they’re dead [beef] or alive [cow]. Each of these words correspond to a different signified, dead-cattle/living-cattle—see what I mean? When it’s said that language is a conventional system, that means that the relationship between a word and the state of affairs it designates, the external state of affairs that it designates, is arbitrary. By contrast, if it’s true that the relationship between words and what they designate is always arbitrary in language, by contrast, the relationship between the signifying word and the signified is not arbitrary.
Why isn’t it arbitrary? Because they’re two sides of the same reality, the same phonological or sonorous reality. The signified and signifier are two sides of the same sonorous reality. In other words, there is necessarily a similarity between the signified and signifier. Simple as that. Inevitably this is what linguists call “isomorphism”, whereby linguists are led to amend Saussure’s principle, that linguistic symbols are conventional, and they amend it by adding that, yes, [it is conventional] insofar as you determine them, insofar as you hold them up to designated states of affairs.
However, there is a perfect isomorphism, i.e., a similarity of relation between the signified and the signifier. That is, the signified and signifier necessarily have similarly formed relations. Every linguist—isomorphism is the principle stressed by every linguist. So, I’ll stop there because that’s all deadly boring — all I’m after is something very straightforward… the two ways a digital code implies similarity: it implies illustrative similarity; it implies narrative similarity. In other words, it implies a similarity in quality, and it implies a similarity in relation.
Counterexample: can analogy be defined by similarity? Of course not. Why? It can’t be, for one very simple reason: for starters, it wouldn’t sufficiently distinguish it from code. Once again, if code necessarily implies and involves instances of similarity, there’s no way to both directly oppose them and tie similarity back to analogy. But I also need a reason inherent to analogy—just like I asked earlier when it came to code, what can you do with it? Well, with analogy, with analogical language—we don’t quite know what it is yet, since we’re looking for its definition—with analogical language, as obscure as it is for the time being. What can you do with it? You can do two things, I think: you can reproduce, and you can produce. What do I mean by that? I’d claim that it’s reproduction when what’s conveyed is a resemblance or a similarity in relation. When you convey a similarity in relation, you produce a resemblance; analogy is thus the formative principle behind resemblance.
I’d call this type “figuration”. That’s the first form of analogy. I’d call this first form common analogy, analogia communis, because we have to squeeze in a little science. “Analogia communis” is the conveyance of resemblance. The conveyance of the relations of resemblance, because if anything’s conveyed it’s clearly the relations, the relations that are conveyable. When you’ve conveyed a rate of similarity, you’re dealing with common analogy. In other words, you produce a resembling image; you make it “resembling.” Whereas—coming back to painting—painting is never this way.
However, I wonder whether, notwithstanding its pretensions and ambitions, photography isn’t necessarily and always this way. Because what is photography, ultimately? What makes it different from painting? Well, photography very generally—I’m putting this is in really rudimentary terms—is about capturing and conveying contrasts in lighting [rapports de lumière]. I understand that that opens up all kinds of possibilities. Namely, you can set out enough room in your mode of conveyance to obtain the deepest of variations, the most extreme degrees in the resemblance, vast variations of similarity. I’d say, you could obtain a fainter and fainter resemblance. That doesn’t change that fact that there is no photograph if it doesn’t convey a contrast in light. So much so that I can’t see how photography could overcome what we might call the figurative aspect. By figurative, I don’t mean the extent to which it resembles something, but the extent to which the image is produced by conveying a similar relation, by a similarity in relation, however faint the similarity might be.
Anyway, it’s like the photo “lives” and possesses its condition of possibility in common analogy, the conveyance of similarity, but analogy isn’t bound to that. We can do something else with analogy, this time producing rather than reproducing—we can produce resemblance. What does it mean for a resemblance to be produced rather than reproduced? Notice that code could also produce resemblance—it could make us a portrait, but that resemblance was produced by way of the detour of a code and a binarization of its input.
Whereas I have something else in mind: an analogy that’d be capable of producing a resemblance without conveying anything, any similarity—now things are coming to a head, because if we manage to define this sort of analogy, one that produces resemblance without conveying any sort of similarity, we’ll have a possible definition for painting. Indeed, painting does produce resemblance, or figures.
Here I am reintroducing the word, “resemblance”. You’re going to see why I’m bringing it back; there’s nothing stopping me from saying, “Painting produces resemblance through non-resembling means.” It produces resemblance through means that are completely different from conveying a similarity, conveying similar relations. You approach a painting—a Van Gogh, a Gauguin—you see a figure: you don’t need to see the model to be convinced that… that you’re looking at an icon. Only this icon is produced via non-resembling means. You reproduce resemblances through non-resembling means. That’s what analogy means. What are these non-resembling means?
But you see, I’m already getting ahead of myself. Why? Because, in a qualified way, I’ve characterized code through articulation, or through the “common sphere”—I said there was a “common sphere” between digital code and articulation. Articulate, and you get a code. Which led us, to a lesser extent—as a result, I’m committed to define analogy and as a result, the diagram as the analogical principle. I have to… I can’t avoid it, see, it’s great when your hands are tied when it comes to concepts. My hands are tied; I have to either give up — it would be perfect, everything is perfect — whether we give up or whether we manage to define analogy and the diagram in analogy depends… on something as straightforward as articulation, and this “something” would be to the diagram what articulation is to code. I already know that it won’t involve any resemblance, won’t convey any similarity, and it won’t involve any code.
So, what does the diagram do that’s opposite of articulation, distinct from articulation, which can be defined neither by its conveying similarity, nor by code, nor by encoding? At least the conditions of our problem are well-defined. Then we have to press on, we have to press on, and so we saw—for now I’ll just say that, as code doesn’t rule out similarity but rather implies similarity, on the flipside, analogy cannot be defined by similarity. Only vulgar, common analogy is defined by similarity. Aesthetic analogy isn’t defined by similarity since it only produces resemblance through wholly different means.
Well, then, if similarity isn’t able to define analogy—as things stand—what is able to define it? Let’s take a look. We’ll move onto the second step. See, our first test was whether analogy can be defined by similarity. Our second hypothesis is that analogy, or analogical language, can be defined as a language of relations. That’s [Gregory] Bateson’s hypothesis, B-A-T-E-S-O-N, who is really such an interesting writer.
Analogical language would be one of relation—as opposed to what? As opposed to conventional language, that of codes—and what would that be? Bateson says, keeping things simple in order to demonstrate something very peculiar—well, it would be a language of states-of-affairs. Our coded language, our digital language, would be one suited for designating, determining, or translating states-of-affairs. While analogical language would express and would be used for relations.
What does Bateson mean? He elaborates on how we ought to understand relation. It’s a fascinating development—I’m pulling this out and I’ll need it because it’ll lead us back to painting via some odd twists and turns. Bateson famously wrote on the language of dolphins.[16]
Actually, Bateson has led an extremely eventful life, has done all sorts of things—and he’s still alive! He was Margaret Mead’s husband—now, Margaret Mead is an ethnologist. So, he started in ethnology, but it turned out he was even better than Margaret Mead; his ethnological studies were so intriguing, so profound, so important. And then his career took off in American style, incredibly—he, hmmm, he said no, no. As if Bateson were a perfect example of an American hero, he doesn’t stop, uh, moving on, moving on. Sort of a hippy, a philosophy hippy—so he divorced Margaret Mead, and then he divorced the tribes he studied. Then he stumbled upon schizophrenia, and he couldn’t… he developed a whole theory of schizophrenia—one of the finest there is, well, a theory now known in France as the theory of the double-bind. Using logic—he’s quite familiar with [Bertrand] Russell’s logic—applying the theory of logical types onto schizophrenia. And even so, he lost interest, he… anyway.
So, he threw himself into dolphin language, which was even better—schizophrenia seemed too human for him, too monotone; dolphins are great, he thought, so he works with dolphins. Obviously, he gets a lot of funding from the US military, who are very interested in dolphins, but Bateson’s results are hilarious, because they’re totally useless for the Navy. So, it’s wonderful, it’s excellent work, and you’ll definitely see why I’m going through his career.
He starts by going over, well, very basic things, because that’s the American style, starting from—they aren’t used to our western, European process—they start with extremely simple terms from which emerges… – whereas we make deductions — they take simple bits and pieces and build a sort of hornet’s nest and draw out a paradox. And they always come up with such great paradoxes, and then they use logic to unravel them, which is all completely different from our way of thinking.
I’m talking about when Americans do it well. So, that’s why they invent so many concepts; they invent a lot more concepts than we do because for us the invention of concepts is a very deductive process. They make theirs by tying stuff together, drawing on a wide range of things; Bateson takes a schizophrenic, a savage [un sauvage], and a dolphin—then, see, he’ll draw something out of it. I think it’s some of the greatest philosophy, and it involves as much rigor as ours, since it’ll all come down to the logic of the paradox, which is what makes them logicians in the end; they open onto everything at once… it’s outdoor logic, whereas ours is deduction in confinement—with us it’s a bit like what I was just saying, we do philosophy on an easel. Ultimately, our easel is the history of philosophy. See, so it’s not like that, uh, Americans don’t do that, but then again, it’s rare for them to be at Bateson’s level.
Well anyway, then Bateson says conventional language is the left side of the brain, which controls the right side of the body—remember, analogical language is the [right] side.[17] What do we usually put under analogical language, by comparison [with digital language]? Well, first off, back to one of our benchmarks: conventional or digital language is fundamentally articulated. It’s articulated. Analogical language is thus the right hemisphere rather than the left hemisphere of the brain. It’s not articulated, so what is it? It isn’t articulated—see, we’re getting around to the heart of it—if we found out what it is, since it isn’t articulated, if we found out what it is, then we might get our definition of painting. Well, it isn’t articulated; it’s non-articulated—what is it, then? It’s made of non-linguistic, even non-sonorous things; it’s made of kinetic movement, so to speak; it’s made of emotional expression; it’s made of inarticulate sonorous input—murmurs, cries.
Obviously, if we were talking about music, we’d find a similar problem because what is singing? It’s articulated or inarticulate, analogical or digital? We don’t know, so we won’t throw music on the pile. But then analogical language, see, is in a way an animal language, but we, we—and Bateson’s just poking around—it’s made out of very heterogenous sources, for example: hairs standing on end, a grimace, a yelp. All of that is analogical language. See, we’re already making progress: a scream doesn’t resemble anything; similarity isn’t what defines analogical language—what does it resemble when your hairs stand on end? It’s not a language of similarity; a scream doesn’t resemble the horror that causes it—not in the least, so it’s not that simple.
So, he says, what is it that defines analogical language? He says that it’s a language of relations. What does he mean by “relations”? He doesn’t mean just any relations, because if he said just any relation—there are some writings where he seems to mean just any old relation—then we wind up with similarity again, i.e., analogical language would be one whose function is to convey relations. For example, in a diagram, you have to represent one quantity that’s big and one quantity that’s relatively small, and you make two levels, one level smaller than the other… that’s similarity, that is a language of relations, in fact. But that’s not what it means, because we’ve ruled out the similarity hypothesis.
He means, it’s a language that’s supposed to express the relations between the transmitter and the receiver, between what emits it and its intended destination. In other words, he explains that this language, analogical language, is a language of relations, understood as the relations between transmitter and addressee; in other words, it primarily expresses dependency relations—in all their possible forms. So, right, analogical language would express relations—see, that’s very different from similarity; it expresses dependency relations between a transmitter and a receiver. — One second, I’ll lose my train of thought if you stop me now… — Okay, that’s what he says about analogical language.
Right, and Bateson feels the need, whenever he makes any headway, he feels the need to joke around—but they’re always good jokes. He calls it the mu-function. Why he calls it the mu-function is because mu is the Greek letter equivalent to our M, and whenever he needs an example for something it’s always cats. Cats meow in the morning; the mu-function is the “meow” function—there’s a side… the English and the Americans have never moved beyond Lewis Carroll. [Laughter] What is the mu-function, or Meow-function? Well, Bateson says that when cats meow in the morning, as you’re getting up, they’re saying—through meowing, which is analogical language—they aren’t saying milk, milk; they’re saying Dependency! Dependency!, I depend on you, with all kinds of variations: there are angry meows, where it’s that I depend on you and I’m sick of it—it’s a very rich language. But it always expresses the relation between transmitter and addressee, with all sorts of reversals. That’s the mu function.[18]
And Bateson says it’s a language requiring a lot of deduction, since if you look at the language’s structure, it directly expresses mu functions, that is, functions of dependency, dependency relations, from which one deduces the state of affairs. In other words, I should ought to deduce: “Hey, my kitty wants some milk,” and deduction is no less involved if it’s one animal talking to another using analogical language. For example, he refers to the famous ritual among wolves or dogs, where an individual shows its inferiority by exposing its neck, demonstrating its dependency vis-à-vis the leader or the stronger animal. You have a dependency relation from which one deduces a state of affairs—in our language it might be “I won’t do it again.” States of affairs are fundamentally deduced from relations, from dependency relations. That’s how Bateson defines analogical language.
See, that’s actually going to be very interesting because, on the other hand, what is our coded language, our digital language? Bateson says it’s a language that primarily concerns states-of-affairs; it’s a language essentially intended to designate states of affairs, but that doesn’t mean there aren’t all sorts of analogy behind the scenes, and that takes us a great distance, and I’d like for you to hang onto it, to hang onto it for later. Anyway, I think codes are practically steeped in analogy, analogically glued together… [Interruption of the recording] [1:53:59]
Part 3
… On the other hand, you don’t get on a plane for no reason—there are all these analog motivations. What dependency relations are inscribed therein? What subverted dependency relations? But in the case of our language, our coded, conventional language, I would claim—or rather, Bateson would claim—that language designates states of affairs by convention, from which one induces analogical functions. Whereas in analogical language, it’s almost reversed: language directly expresses analogical dependency relations, from which one deduces states of affairs.
And yet—and this is all I’m trying to say and why I brought up dolphins. Dolphins have a language, and no one can understand a lick of it. Bateson says that no one understands it because it’s likely that there’s not much in their language to understand. Consider the following convoluted process—one wonders who’d be capable of it. I have my two languages: an analogical language for relations, a coded language for states of affairs.
Let’s say I get a bit of a wild idea: to encode analogical relations as such. To encode mu-functions. A language that remains analogical, but which is fed through a code. It’s a very odd sort of language: a code grafted onto analogical flows. At first glance it’s impossible; it’s contradictory. However, it’s sort of what we saw earlier with the computer. Computers have a binary code; they encode something to be reproduced, a design [dessin] to be reproduced, and it makes the design for you. Now suppose that dependency relations, mu-functions, etc., are likewise encoded. Encoded—we considered what “encoded” might mean. It can mean being caught up in a system of binary choices. Then again, why would an analogical language get encoded? Why would you ever need to encode an analogical language? That is, to graft code onto analogy? On only one occasion, Bateson says: in the case of large mammals who’ve abandoned the life on land and have fled to sea. Why? Because large mammals have a robust analogical language. None on Earth have taken analogical language further than mammals. They’re screwed once they take to the water. How so? Because, unlike fish, they aren’t capable of their own analogical language, one particular to a marine environment, and they no longer have the means to use terrestrial analogical language.
In fact, terrestrial analogical language implies a clear distinction between the head and the body, implies hair, implies expressive movements—all things that water’s parameters not only limit, but what’s more, even if they had these, their message wouldn’t be received, since the conditions of visibility underwater are such that terrestrial analogical language doesn’t work. What’s more, the whole body is submerged, preventing analogical expressions. So, he says: we think that dolphins have a mysterious language. Not at all. We think that dolphins have a conventional language, and he warns the American military that they’re setting themselves up for serious disappointment. Is he right? I have no idea. He says no, it’s not like that.
The paradox with dolphins is simply that the maritime conditions to which they’ve had to adapt make it so that they’ve had to encode the analogical as such. They haven’t developed a digital language; they haven’t developed a language of codes; they’ve had to encode analogical language. So, it’s very bizarre. He says he’s personally convinced that if we manage to decipher a bit of dolphin language, we won’t find a linguistic language. What we’ll find in this language is strictly analogical content that simply expresses dependency relations and that expresses nothing about states of affairs. That’s his claim! But why am I going over Bateson’s thesis on dolphin language? It’s because, in light of this, see, there’s something that really intrigues me: the possibility of grafting binary code onto purely analogical language.
Thus, it allows us to somewhat overcome the duality we started with. What I mean by that—you’ve already guessed what I’m getting at. What I’m getting at is: just think about the formula I would end up with; it no longer appears so simple. I sort of get the feeling that an abstract painter is no different from a dolphin—they’re dolphins; they’re dolphin painters. It’s what makes them abstract; their real method is inventing a code for a specifically analogical matter and content. Then they graft code onto the pictorial material, and this code is entirely pictorial, whereby they manage to achieve something awesome. In other words, they aren’t abstract painters; they’re actually marine mammals. It’s equivalent to—it’s exactly the same problem as with the dolphins, it seems to me. But that’s not important.
Let’s just keep going: now we’re kind of stuck because—okay, when it comes to everything he says, even relations, well… Analogical language is no longer defined by similarity but is defined by dependency relations. Does that work for us? Does that change anything? Maybe, but not on its own. I think it still needs tweaking. What needs to be changed? How are dependency relations expressed? What is it that expresses dependency relations? At this point I need a third determination for analogical language. One more. Because strictly speaking dependency relations are the content of such language. But if analogical language has its own specific form, what will this form be? Suppose it’s painting. What then? It needs a form. How are dependency relations expressed? Then we’d be able to define analogical language. We’ve got it, it’s in hand. Suspense… What were you going to say, Anne?
Anne Querrien: That reminds me of a passage from Thousand Plateaus, in which you speak about order words and the transformation of bodies by incorporeals.
Deleuze: Fair enough—me too, but that’s even more complicated. Since things are complicated enough as is, that won’t make it any easier.
Anne Querrien: And then it reminded me of something else, the vocabulary of masons for cathedrals, etc., particularly, for example, there was a study done by a guy named [André] Scobeltzine, where he analyzes Roman and Gothic sculpture and demonstrates that they all have a code for capitals that directly express dependency relations by how they’re positioned, in the way… [Deleuze interrupts her][19]
Deleuze: Oh, that’s interesting!
Anne Querrien: It’s called Feudal Art and Its Social Significance, by Scobeltzine, an architect, in Gallimard’s library of human sciences collection.[20] And so, he shows how sculpture is an expression of dependency relations and that you can interpret all gothic art in this way.
Deleuze: That’s wonderful! Did you all hear what she said? Would you mind standing up and saying that again? Because they might be interested.
Anne Querrien: There’s an architect named Scobeltzine who wrote a book titled Feudal Art and Its Social Significance, and he says that all cathedral sculpture and architecture is the expression of social dependency relations, using both an architectural code for the vaults—or I mean the way they are constructed — and sculptural code on capitals essentially. He goes into detail about this whole gothic line, outlining the different forms of capitals.
Deleuze: I have to read this book. I need to read it.
Anne Querrien: Yeah, it’s worth a look!
Deleuze: Write down a note with his name for me because I didn’t jot it down earlier.
Anne Querrien: I ought to have my notes at home.
Deleuze: Send them my way—that way I don’t have to read it myself. [Laughter] This is very important. See, there are tons of things I haven’t considered. You could… Okay then, what is it?
I’d say even when it comes to the voice. Consider the example of audio: analogical language exceeds the voice, okay, is more than voice, but there is also analogical language in the voice. Now, in a way, linguists themselves—cutting to the chase—there’s something curious that linguists don’t even bother to conceal. They don’t hide the fact that language, what they call conventional language, is made up of so-called distinctive features [traits distinctifs]—moreover, what these internal distinctive features are specifically amounts to the binary relationships between phonemes. A phonological relationship.[21]
See, a binary relationship between phonemes is an internal distinctive feature of language. But they’ve always claimed that there were other linguistic features. What are these linguistic features? They’re tones, intonations, accents—or to put it more precisely: pitch, loudness, and length. Pitch, loudness, and length form three kinds of stress [trois espèces d’accents]. What about them? Can I now say what it is that’s distinct from articulation… Just what are linguists doing? That’s what bothers me. And actually, these non-internal features, you might even say non-distinctive features. They acknowledge them. [Roman] Jakobson, for example, takes this approach when defining what he calls poetics in relation to linguistics.
But what I think is strange is that, despite everything, they acknowledge the specificity of its domain, but still try to exhaustively encode it. They try to encode it in its entirety, i.e., they apply their binary rules to it. It’s very apparent in Jakobson. Personally, however, I’m working—as we continue to ask “what is analogical language and what would its concept be”—I’m working instead under the assumption that we must not apply rules of code or binarize the realm of so-called “prosodic” or “poetic” features. Then what is it? The non-articulated voice—non-articulated voice has pitch, loudness, length, and it has stress. Stress and articulation present an even bigger problem when it comes to music. Fortunately, we’re not concerning ourselves with music. Also, what role does code play in music? What is the role of the non-articulated? What is the opposite of musical code in music itself? That would be a problem. Everyone knows the opposite of code in music, ultimately. But anyway, we’ll come back to all that later.
Okay, so what’s on the table when it comes to the analogical? Well, there’s a convenient term, but it won’t fix anything because it’ll be hard to develop a concept for it: modulation. In what way is it modulation? I’m not claiming that there’s a simple opposition, although in some regards there is a simple opposition between articulation and modulation. I mean that modulation refers to the values of a non-articulated voice. I can start there. That being said, there are all—and this will help—modulation and articulation can combine in all sorts of ways. Modulation and articulation. But now that we have a hypothesis, you can see why it matters—I’m belaboring the point so you can see what’s at stake in all this.
I’m saying, analogical language would be defined by modulation; where there’s modulation, there’s analogical language—and thus, there’s a diagram. In other words, the diagram is a modulator. See, that does a good job of meeting my requirements: the diagram and analogical language are defined independently of any reference to similarity. Obviously, you have to check; there’s no need to bring any similarity into modulation.
Analogical language is about modulation. Digital, or coded, language is about articulation. All sorts of combinations are possible, so you can articulate a modulation’s flow. You can articulate the modulatory. Then you’re grafting a code, which might be important: maybe code has to be involved to allow for analogy’s full development. Things are getting complicated—how can this hypothesis help us with painting?
Let’s just apply it, since painting is an analogical language, maybe the highest form of analogical language to date. How so? Because to paint is to modulate. To paint is to modulate—but what does it mean to modulate? Mind you, with modulating, you modulate something on the basis of something else. Let’s specify what’s going on with this concept of modulating. At its most basic, you modulate something using what’s called a carrier or medium—a carrier wave or medium. On what basis do you modulate a medium? According to a signal. And now you’re as much of an expert as I am—it’s TV, or whatever, it’s all around us. We’re immersed in the work of modulation.
A carrier, or a medium, is modulated based on a signal. A signal to be conveyed. Modulation does not convey similarity. We still don’t know what it is. We don’t know yet. When it comes to painting, can I apply this very broad definition in a way that isn’t merely an application, such that what we get is undeniably a definition of painting? What is the signal? The signal—let’s stick to worn-out categories. The more worn-out, the better.
The signal is the model. In more complex terms, the signal would be what we called the motif, based on our discussion of Cézanne. Which isn’t the same thing as the model. But anyway. It’s either the model or, in a more particular way, the motif. Or else it’s—these aren’t mutually exclusive—it’s the surface of the canvas; that’s also the signal. The model is the signal, but so too is the surface of the canvas. It too is the signal. No doubt it all depends on my perspective; I can come at it from all sorts of angles.
What do I modulate on the canvas? To paint is to modulate. I can only see two alternatives for what would address modulation. Either I modulate the light, I modulate the color, or I modulate both. Indeed, light and color really are the carrier waves of painting. So that, again—I highlighted this last time—I’m not so sure we can define painting as just line and color. I’d claim that to paint is to modulate light or color, light and color, depending on the flat surface and—these aren’t exclusive—depending on the motif or the model, which plays the role of signal.
But see, what result do we get from modulation? The figure on my canvas. Whether it’s Pollock’s line—with no figure, in fact—or Kandinsky’s abstract figure, or Cézanne or Van Gogh’s figural figure: that’s what I get from modulation. What I might call Resemblance with a capital R—only I produced it via non-resembling means, hence the painter’s motto: “I’d be capable of a resemblance deeper than that of a camera.” “I’d be capable of a deeper resemblance than any other resemblance.” That’s because I produced it through wholly different means, and these different means are the modulation of light and of color. Thus, with all of the countless definitions for painting: “It’s an arrangement of colors assembled on a flat surface, or carved out of the surface, or it’s this or that,” and so on—we can pat ourselves on the back for adding one more, which obviously isn’t a big deal, but we run into at least one issue in particular.
How are light and color the objects of modulation? What exactly is the modulation on a flat surface of light and color? What’s going on here? If I can manage to explain what modulating light is, what modulating color is. Here we go.
So now we can’t get around defining a concept of modulation, one both clearly distinct from the concept of articulation, and at the same time one that makes no reference to similarity and the relation of similarity.
And I could say that the diagram is the matrix of modulation. The diagram is the modulator, just as code is the matrix of articulation. And strange as it is, it’s not at all impossible that we’ll end up—if it gets us closer to the diagram, if it gets us closer to analogical language, if it gets us closer to modulation—involving a period of code. It may very well be that modulation has a lot to gain from a code phase. In other words, it may very well be that abstract painting constitutes a step forward for painting—for all painting—fundamentally, a step forward. From the twofold perspective of the modulation of color, i.e., from the paradoxical perspective—not with regard to the invention of code but with regard to the development of an analogical language. Regarding the modulation of color and regarding the modulation of light. What would it mean to modulate color, to modulate light? … What time is it?
A student: Twenty to, quarter to.
Deleuze: You all must be tired!
Students: No.
Deleuze: Then, to wrap up, we’ll go back to square one. Modulate, modulate, modulate! Modulate, not articulate; modulate, not articulate. But how are we going to make this into a concept—I’ll try to draw from wherever I can. I’d refer us to two sorts of information: literary information and technological information.
When it comes to literary information: easy, there’s a great text. An excellent text which has already been covered from every angle, but I’d like to discuss it in light of our analysis. It’s Rousseau’s On the Origin of Languages. Rousseau’s On the Origin of Languages speaks to the issue at hand. How so? Because, in this extraordinary text, Rousseau’s core premise is that language cannot have originated in articulation. Articulation can at most be the second stage of language. That’s a bit of an overstatement—for Rousseau, all language is articulated.
But articulation can only be the second stage of the voice’s development. Voice existed prior to articulated language—what sort of voice? The melodic voice, Rousseau says.[22] And how does Rousseau define the melodic voice? Rousseau defines it in a very precise way. It’s the intonated, accented voice. Not just intonated or accented, because you might think that all language has intonation [accents], but in fact, for Rousseau, languages have lost all intonation.[23] They have more or less have some, but the key to understanding intonation lies in extinct languages.
The Greeks still had an accented language [langue à accent]. English might somewhat—it’s weird, but it’s not Rousseau who says it. But anyway. Why is he saying that our languages are no longer accented? They have accents, but they’re no longer “accented,” he says. When there are accents, intonation [l’accent] is gone.[24] He means that differences in accent, as he sees it, ought to correspond to tonal differences. However, in our case accents don’t correspond to tonal differences. Differences in accent do not correspond to differences in tonality. So, well, our accents have deteriorated so much—look at what he’s getting at. Why has intonation [les accents] declined so much in our languages? Why has our language ceased to be melodic? Whereas real language is melodic. Okay, well, they ceased to be melodic as soon as they became articulated languages.
And why did they turn into articulated languages? Rousseau’s idea is wonderful but strange. He says languages became articulated once they left their birthplace, because in his view, language was born in the South. That’s where you’ll find the conditions for a language to emerge. Languages are originally southern. But that doesn’t stop them from prevailing in the North. Tough northerners are the articulate ones. Why do tough northerners articulate? Because they are industrial. What’s more, he explicitly says—I hope you check it out, On the Origin of Languages is a very short essay—several times he goes so far as to explicitly state: “Articulation is conventional.” Articulation is a matter of convention. Granted, granted—but what that means in modern terms is that articulations are determined by choice. This puts it unquestionably in the realm of binary choices.
So, articulation is conventional. The Northern man, with his industrial needs, is forced to articulate because he no longer knows how to say “love me”; he no longer knows how to say “help me”—for Rousseau, the proof of one’s love is helping. Helping with work. So, the language of work, the language of industry, is one that’s thoroughly articulated. Right, it’s a thoroughly articulated language. There are still inarticulate sounds among Northern men. They articulate—they articulate; it’s very much an articulatory language. But they retain some inarticulate sounds, only they’ve turned into fearsome cries. What does he have in mind? See, they happen in tandem: when articulation begins to rule language, inarticulate sounds turn into paroxysm, a sort of paroxysm. They become fearsome noises. What does he have in mind? The sad setting of Rameau’s opera. And in music, what’s the equivalent of musical code, of articulation. Articulation is akin to what Rameau described as “the matrix of all music.” For Rameau, the matrix of all music was “harmony.”
It was harmony. With its vertical sections of melodic lines and how it establishes chords. Rousseau touches on this—his essay is so artfully written—touches on this very subject when he says that harmony is to music precisely what articulation is to language. It’s the conventional aspect. Only the melody is natural. Harmony is convention; you make music purely by convention. And then this conventional music broke so far away from melody that everything inarticulate, non-harmonic turns into—what? Awful shouting.
And at that point, for Rousseau, our whole relationship with the voice and music is fundamentally distorted. The voice reverts to crying while melody winds up depending on purely conventional harmonics. As a result, in his struggle against Rameau, with what does Rousseau counter? He counters Rameau with a purely melodic music, with very little harmony, in which the voice is non-articulated but doesn’t resort to shouting, [that is,] the pleasant voice of pure melody. So, the triple voice, point and counter point, etc., but without conceding to harmony’s demands. What will melody as opposed to harmony define? It will define the modulation of the voice that will provide a positive definition for the non-articulated voice. Whereas, from the point of view of harmony, the non-articulated voice can only be defined negatively, in the form of awful shouting.
So, what would a present-day Rousseau say, for example, about Italian opera, Wagnerian opera? Obviously, that’s unfair because it is obvious, but anyway, I’ll try to reproduce his schema. You see what he’s thinking: language has two basic stages. First, he says, language couldn’t have been born from interest—an interesting idea—language absolutely couldn’t have been born out of interest or need. This runs counter to the entire 18th century. Give me a second… sorry… Yes, what is it?
Anne Querrien: There’s another extraordinary on the origin of languages text by [Jean-Pierre] Brisset, La Grammaire logique et la Science de Dieu [Logical Grammar and God’s Science] which has been re-edited by Foucault, in which he says humans descend from frogs, emerge from the water and discover their sex, i.e., the fundamental binarism and, based on this sexual matter, little by little they start talking. The code is created based on the frog’s croaking.
Deleuze: Yeah, but that’s no good, as opposed to the example from earlier, it’s purely sonorous similarity. It’s just a matter of similarity.
Anne Querrien: Afterwards, if you will, he gradually builds everything back up … [Inaudible words]
Deleuze: Yeah, but unlike your earlier text, we have to force things in order to bring in Brisset. I feel like Brisset is a whole other problem. We’ll come back to interest because he says that with interest and need and even industry gesture would do. A pure gestural language would suffice. Really interesting. Why? Because gestural language is a language of similarity. It just mimes things. If I [Deleuze pantomimes], can you all tell I mean “pulling on a rope”? [Laughter] Or gestural language in the military: he draws his sword and then points it in a direction—even the most boneheaded cavalry knows to head that way. [Laughter] Yeah, a language of gestures is enough…
A student: There’s a text by Marcel Jousse on this, Anthropologie du geste [Anthropology of the gesture] where he actually traces language back to gestures precisely, claiming that speech was created because humans were lazy and didn’t want to use their whole bodies to express themselves…
Deleuze: A text by who?
The student: By Marcel Jousse. He’s an anthropologist.
Deleuze: An anthropologist… We don’t need to refer to Marcel Jousse because it’s a very 18th century idea. A completely standard idea, that language originated in labor and in the gestures of labor. Rousseau says as much.
The student: Still, he goes a little further than that…
Deleuze: What?
The student: He goes further than that.
Deleuze: I hope so.
The student: Because he analyzes the function of mimèmes, that is, the human capacity to reproduce outward interactions.
Deleuze: Yeah, of course, it might be interesting, to see if there’s an analogous dimension to what he’s calling interactions. But anyway, that’s another topic.
The student: There’s also…
Deleuze: Yeah.
The student: There’s also an analysis of language and, in particular, rhythm-melody functions.
Deleuze: I’m sure there is, [Laughter] there’s all kinds of things. [Laughter] So you understand why, in fact, Rousseau’s idea is very straightforward. It’s that language can have but one source: passion. It is passion. So, in fact, this sort of puts us in an almost aesthetic element. Because it’s precisely what some art critics call the pathic moment, pathos as opposed to logos. You could say that logos is code, but there’s an element which is the pathic element of passion. So clearly, it’s due to passion that language has a southern origin. The young boys and young girls gather around the fountain, says Rousseau. Then they begin to dance, etc. That’s the origin of modulation. You have an idea of Rousseau’s schema: you exclude gestural language because that’s a matter of common analogy; it works via similarity.
The second step is the modulation of the voice. Yes, that’s the second analogy, the aesthetic analogy. It’s no longer defined by similarity but by modulation. The melodic voice. Third, language heads north: the people of the North get their hands on it and, in response to industrial development, introduce laws of articulation and apply them to all melodic language, just as music will be subject to the laws of harmony. Not a bad idea. My takeaway is—and I’m just about done— modulation: Rousseau himself will call it “melody.” All right, so what will become of this melodic voice? … Whew! Anyway, until next time.
A student: Yes. [End of the session] [2:34:28]
Notes
[1] The translation for tache is a “patch” of color. See Daniel Smith’s note on the translation in Gilles Deleuze, Francis Bacon: The Logic of Sensation (New York: Continuum, 2003), p. 184 note 1. In this case, however, it is rendered as “stain” in keeping with Morris Louis’s “stain painting.”
[2] In order to distinguish between ligne and trait, I will typically translate the former as “line” and the latter as “stroke.”
[3] The French reads l’œil a peine à suivre. Understood literally, the line is such that following it hurts one’s eyes.
[4] A line attributed to Sérusier by Maurice Denis. See Denis, “Excerpt from the Journal (1906),” in Conversations with Cézanne, ed. Michael Doran, trans. Julie Lawrence Cochran (Berkeley: UCLA, 2001), 178. I have used Cochran’s translation above.
[5] The French figure is translated as “figure” consistently throughout. At this point in the discussion regarding Kandinsky, however, there is a brief switch to referring to figures as “shapes” in keeping with Cochran’s translation of Kandinsky. The reader can assume that all mentions of “shape” and “figure” refer to figure in French, and that there is no intentional distinction between “concrete shapes” and “concrete figures,” for example.
[6] Translation modified to conform to Kandinsky’s definition of synthesis, cited above. On the diagram and the code, see Francis Bacon. Logic of Sensation, chapter 12, notably pp. 104-108.
[7] The French has vent (wind), dent (tooth), and ment (lie). To preserve Deleuze’s point about phonemes, similarly spelled English words are used for these three French words: “vent,” “dent,” and “meant.” For the French fend, the choice is “bent.”
[8] Deleuze is describing what is better known in English as a “binary search algorithm” or “bisection search.” The translation preserves the language of “choice” or “selection”.
[9] While this could refer to a number of texts by Leroi-Gourhan, it most likely refers to Le geste et la parole, technique et langage (Paris: Albin Michel, 1964), Gesture and Speech, trans. Anna Bostock Berger (Cambridge, MA: MIT Press, 1993). See also references in A Thousand Plateaus, notably pp. 496-498 and p. 574 note 33.
[10] On this famous critic, Deleuze refers to Georg Schmidt, Mondrian (Paris: Réunion des musées nationaux, n.d.), in Francis Bacon. The Logic of Sensation, p. 185 note 10.
[11] See Tout l’oeuvre peint de Piet Mondrian (Paris: Flammarion, 1976).
[12] As the audio returns in mid-syllable, the word preceding “and code” is unclear.
[13] This is no doubt a reference to Lyotard’s Discours, figure (Paris: Klincksieck, 1971), Discourse, Figure, trans. Antony Hudek and Mary Lydon (Minneapolis: University of Minnesota Press, 2011).
[14] The shift toward the “third path” and the analogical roughly corresponds to the shift to chapter 14, “Analogy”, in Francis Bacon. The Logic of Sensation.
[15] The reference to Peirce as well as the entire discussion on analogy corresponds to chapter 13 of Francis Bacon. The Logic of Sensation, notably pp. 116-120 and p. 188 note 5.
[16] Gregory Bateson, “Problems in Cetacean and Other Mammalian Communication,” Steps to an Ecology of Mind (Northvale: Jason Aronson, 1972), 260-9. On Bateson, see Francis Bacon. The Logic of Sensation, p. 188 note 7.
[17] Deleuze mistakenly say “left” here.
[18] Deleuze is loosely paraphrasing passages from Bateson, 261-2. His paraphrase includes broad references to Bateson’s overall body of work.
[19] On this reference, in A Thousand Plateaus, pp. 554-555 note 26, Deleuze and Guattari already have referred to Querrien’s work, Devenir fonctionnaire ou le travail de l’État (Paris: CERFI, no date), a reference seemingly without any bibliographical trace.
[20] André Scobeltzine, L’Art feudal et son enjeu social (Paris: Gallimard, 1973).
[21] In this context, the translation follows linguistic parlance with trait as “feature.”
[22] “[I]l n’y eut point d’abord d’autre Musique que la mélodie, ni d’autre mélodie ; que le son varie de la parole… (408). See Jean-Jacques Rousseau, “Essai sur l’origine des langues,” in Collection complète des œuvres 1780-1789 vol. 8, no. 4. Online edition (27 June 2022): http://rousseauonline.ch/Text/essai-sur-l-origine-des-langues.php
[23] “Si l’on croit suppléer à l’accent par les accens on se trompe: on n’invente les accens que quand l’accent est déjà perdu. […] Il y a plus; nous croyons avoir des accens dans notre langue, & nous n’en avons point.” 378-9. While the translation elsewhere renders accent as “stress” (when Deleuze is talking about contemporary linguistics), I have deferred to Rousseau’s translators whenever possible. See Jean-Jacques Rousseau, On the Origin of Language, trans. John H. Moran and Alexander Code (Chicago: University of Chicago Press, 1966).
[24] “On n’invente les accents que quand l’accent est déjà perdu”, 378. Translation distinguishes “accents” and “accent” by referring the former to accent marks and to the latter as “intonation.”
For archival purposes, the transcript was completed by Paris 8. The translation was completed in June 2020, and the time stamp as well as updates to the transcription and translation were completed in June 2022, with additions and a revised description completed in September 2023. Both the transcript (and the translation) have benefited greatly from the corrections brought to the transcription in preparation for publishing Sur la peinture (Paris: Minuit, 2023), edited by David Lapoujade, with additional revisions completed in February and April 2024.