Thursday, January 26, 2023

A Thousand Plateaus, Chapter 4



Chapter 4: November 20, 1923—Postulates of Linguistics


In this chapter, somewhat less imposing than the previous one, but nevertheless very dense, D&G tackle four “postulates of linguistics,” which are more accurately strawman arguments representing postulates of linguistics-as-a-major-science which D&G will argue against. The date refers to an announcement of the German government re-valuing the Deutschmark to end hyperinflation, and thus stands for the power of the order-word. Because of all the detail in this chapter I went with a paragraph-by-paragraph outline instead of a regular summary.


First postulate: “Language is Informational and Communicational.”

1. Language is not communicational; rather it commands. The teacher orders the students, she does not inform them. The fundamental unit of language is the order-word, a pun on mot d’ordre, meaning slogan or password, and here treated also as “word of order.” “Language is made not to be believed but to be obeyed” (76). “Language is not life; it gives orders.” There is an opposition made between language and life, and language, the order word, becomes a sequence of death sentences, or “Judgments.”

2. They then turn to the “status and scope of the order-word,” arguing that it “is only a language-function, a function coextensive with language.” Language, in other words, is not informational; it is not about transmitting information, for instance about something seen or experienced; instead it is a repetition of the already heard, hearsay. The authors here draw heavily on the language theory of Volosinov: all language is indirect discourse. They draw on a distinction made by Benveniste in arguing that bees do not have true language: bees, according to Benveniste, can describe a food source they have seen, but cannot repeat such a description that they have witnessed.

3. They underline this with a turn to Austin’s theory of speech acts, arguing that his concepts of performative utterances, and illocutionary acts, are internal to language and thus distinct from any idea of reference or information external to [that is, excludable from an analysis of] language. The theory of performance and illocution has, they argue, three effects: 1) language can no longer be thought of as a code (like dna is a code: the mere transmission of information) [recall also their discussion in the previous chapter of coding, defined by B&P as “the process of ordering matter as it is drawn into a body;” language, as in that chapter, is something much more]; 2) pragmatics becomes the most important aspect of language, and no other aspect (“semantics, syntactics, or even phonematics”) can be thought independently of pragmatics (77); and 3) the distinction between speech and language, so important to diverse thinkers as Saussure and Chomsky, becomes unworkable (they here reference Labov, but their position owes much to Volosinov’s theory of the utterance).

4. They next clarify another theory of language which they are arguing about under the name “communicational,” namely Benveniste’s appeal to intersubjectivity. Performance is “that which one does by saying it,” and illocution is “that which one does by speaking” (78); there is a temptation to start with performance and derive illocution from it, but this is not correct, as the performative is too easily dismissed, by Benveniste, as a “self-referentiality” in language, which is ultimately founded on communication between pre-existing subjects. Naturally, supposing pre-existing subjects begs the question of where these subjects come from; both Volosinov and Foucault, key interlocutors here, have had much to say about this; in any event “subjectifications are not primary but result from a complex assemblage” (79), and it is this complex assemblage which has to be examined. D&G will approach this by instead deriving performance (corresponding to subjectification) from illocution, which itself is “explained by collective assemblages of enunciation” (78) and by juridical acts or their equivalents (cf. Volosinov on language permeating into, or rather composing, the entire consciousness of the speaker), and they will further explore this through the discussion of indirect discourse in the Bakhtinian/Volosovinovan tradition.

5. To set up order-words as having a function co-extensive with language, they define them as “the relation of every word or statement to implicit presuppositions” and to the speech acts that can be accomplished in these statements; Language, in turn, is defined as “the set of all order-words, implicit presuppositions, or speech acts current in a language at a given moment” (79).

6. Instead of being informational and/or communicational, language is “the transmission of order-words, either from one statement to another or within each statement, insofar as each statement accomplishes an act and the act is accomplished in the statement” (79). Thus performativity, not to mention power in the Foucauldian sense, is what characterizes language. Their meaning can be taken as a somewhat more subtle and felicitous version of what Foucault actually meant when he said there was nothing outside of the text: the subjects who speak or use language, and the world of information and references they speak about, and even the criteria of truth or falsehood by which the relation between the statements of the subjects and what they are talking about is evaluated, cannot be presumed or treated as some pre-existing entities within and among which language takes place, as some kind of add-on. The relationship between statement and act is not one of identity, but rather of redundancy, which has two forms: frequency, corresponding to signifiance/information, and resonance, corresponding to subjectification/communication.

7. This next paragraph largely translates Volosinov into Deleuzian terms: “There is no individual enunciation. There is not even a subject of enunciation.” Rather, the assignation and distribution of individuality takes place according to the requirements of the collective assemblage of enunciation. They provisionally define this collective assemblage as “the redundant complex of the act and the statement that necessarily accomplishes it” (80); but they immediately admit (and correctly so), that this is unsatisfying. “If we wish to move to a real definition of the collective assemblage, we must ask of what consist these acts immanent to language that are in redundancy with statements or constitute order-words.”

8. They now go into more detail clarifying just what will be meant by an act, which is something much less than “action” or “acts” in general (this presumably is less forced with the French acte than with English act; their examples are pretty much all literal speech acts, “media acts,” etc.). Acts are incorporeal transformations which are attributed to bodies; “body” is here meant “in its broadest sense” to include pretty much anything, including souls and other “mental bodies.” Bodies interact with each other through actions and passions; “acts” are distinct from these in being incorporeal. Acts are also simultaneous with their effects (and thus datable): “I declare war,” makes war happen that instant; “I now pronounce you husband and bride,” etc. They refer to this as “the illocutionary.” As all of their examples involve speech or writing, it is not quite clear where they would put an action such as kneeling before a king (and thus submitting); presumably the illocutionary effect of this would be considered an act, in their sense (since kneeling in different contexts has different meanings: we are back to the relevance of the assemblage).

9. They discuss history, or the telling of history, as recounting the actions and passions of bodies, but more especially as being about order-words. The simultaneity/instantaneity of order words is reflected in their datability; they give their example of November 20, 1923, the day the German Reichsmark was replaced with a new currency, to bring about an end to inflation.

10. They explore further the (social, etc.) assemblage or range thereof within which language has meaning, including the circumstances in which “I love you,” “I declare a general mobilization,” etc., are meaningful or nonsensical. They echo Volosinov’s arguments against a linguistics that excludes consideration of all this as “exterior,” in order to focus only on internal “constants” of language; the pragmatics they are for, in contrast, is about both the exterior and what is internal, immanent, to language (this “incorporeal”); the order-word is the “something else” beyond language itself which establishes this relation between the interiority of language and the bodies, passions etc. upon which it acts.

11. They discuss Lenin’s text “On Slogans,” focusing on examples in which order-words create their subjects: “Workers of the world, unite!” “invents” the working class; Lenin’s announcement transferring power from the soviets to the Party predates the existence of the Party to which it is attributed. Against the objection that this is “politics” (and thus merely external to language) they respond that “it must be observed how thoroughly politics works language from within” (83), and give a summary of their pragmatics:

A type of statement can be evaluated only as a function of its pragmatic implications, in other words, in relation to the implicit presuppositions, immanent acts, or incorporeal transformations it expresses and which introduce new configurations of bodies. (83)

12. They propose two new names for the collective assemblage of enunciation: “regime of signs,” and “semiotic machine;” they add the very Bakhtinian observation that “society is plied by several semiotics, ... its regimes are in fact mixed” (83-4). They then discuss the primacy of indirect speech over direct speech and tie this to schizophrenia and the hearing of voices. There is a very interesting footnote on Elias Canetti’s concept of the command as a “sting on the soul, which forms a cyst” (525n17); [in contrast to the “shocks” experienced by the Baudelarian hero] this sting is enabling, the one who follows orders feels like a victim of the orders, and thus feels innocent of their consequences, which further situates them to follow future orders. According to D&G, Canetti tries to limit the consequences of this insight by attributing it to a particular psychological disposition in the minds of those who succumb to it, rather than to the regular functioning of the order-word itself; “rational,” “common-sensical” individuals [aka abstract subjects] are thus immune. D&G observe:

The whole classical rationalist theory—of “common sense,” of universally shared good sense based on information and communication—is a way to cover up or hide, and to justify in advance, a much more disturbing faculty, that of order-words. This singularly irrational faculty is best safeguarded by gracing it with the name of pure reason, by saying that it is nothing but pure reason... (525n17)

[To which could be added that, once again, the “universal” qualities of the rational abstract subject are invoked precisely when some population (in this case those who are prone to fascism) is being excluded.]

13. They note that “order-words, collective assemblages, and regimes of signs cannot be equated with language. But they effectuate its condition of possibility” (85, emphasis added), and this is the key point they are making. Without these, “language would remain a pure virtuality,” which is why all the attempts to define it as informational or communicational miss the point. They also tie in the concept of “superlinearity” from the previous chapter.


Second Postulate: “There is an abstract machine of language that does not appeal to any extrinsic factor.”

Their repudiation of this is naturally tied to their previous discussion of pragmatics and performativity. This is one of the points where their method of arguing against straw-man propositions is at its most ridiculous, because obviously anyone actually making the above claim would not use the distinctly Deleuzo-Guattarian term, “abstract machine.”

1. They discuss the relationship between content and expression, which, in the previous chapter, had correlated with the first and second articulation, respectively; here they refer to these as formalizations, emphasizing the independence of these two strata (content is not dependent on expression, nor vice versa). They trace their argument back to the ancient Stoic distinction between the actions and passions of bodies, body here defined as “any formed content” (86).

2. Following the usage of the Stoics, anything that can interact through actions and passions (i.e., in the [material]/corporeal world) is a body. What counts as a body for D&G thus immediately begins to proliferate in a manner reminiscent of what happened with “strata” in the previous chapter:

The purpose [of language, and of this chapter] is not to describe or represent bodies: bodies already have their proper qualities, actions and passions, souls, in short forms [emphasis added], which are themselves bodies. Representations are bodies too! (86)

The speech act being, as was previously stated, an incorporeal transformation, this is here distinguished again from representations, which are bodies with actions and passions. This is all related back to the independence of content and expression (as each having their own forms) from the previous chapter. Language-as-speech-act thus intervenes, rather than representing, and it does so in a non-corporeal way. The image of a loom or fabric is used to clarify these two forms: “The warp of the instantaneous transformations is always inserted into the woof of the continuous modifications.” But the relation between these two registers is at the same level: “An assemblage of enunciation does not speak ‘of’ things; it speaks on the same level as states of things and states of content” (87). Here, they appear to move beyond their examples of words to explicitly discuss signs in general (thus answering my earlier questioning about kneeling): “...the same x, the same particle, may function either as a body that acts and undergoes actions or as a sign constituting an act or order-word, depending on which form [content or expression] it is taken up by....” [They make a reference to the “causality of contents” which makes me wonder if their distinction between the incorporeal and the corporeal, while insisting also on their flatness and interdependence, is related to the ancient Stoic attempt at resolving the causality/free will relationship.]

3. They qualify their assertion that incorporeal speech acts “intervene” in the continuous modification of the corporeal world, by noting that this might imply a new idealism; instead, both modification and intervention are characterized by deterritorializations which make possible subsequent reterritorializations. The external circumstances, and the internal factors, of bodies, languages, etc. are variables of content and variables of expression. Again, this is all about denying a traditional understanding of language as representation: instead, “...forms of expression and forms of content communicate [with each other, not the communication between subjects which was dismissed earlier] through a conjunction of their quanta of relative deterritorialization, each intervening, operating in the other” (88).

4. From this, they draw “some general conclusions on the nature of Assemblages” (88). First, a horizontal axis including two segments, one of content (a machinic assemblage of bodies, actions, and passions) and one of expression (“a collective assemblage of enunciation”); as we have already seen, these two “segments” exist together on this axis interwoven, as warp and woof. The vertical axis gives the assemblage two sides, one resting on a process of re/territorialization, the second a “carrying away,” “cutting edge” of deterritorialization. Kafka’s writings are discussed as providing insight.

5. This is then re-expressed as the four qualities (or “tetravalence”) of assemblages: 1) “interminglings of bodies,” 2) incorporeal transformations, 3) “territorialities and reterritorializations,” and 4) deterritorialization (89). This is illustrated with the example of the feudal order.

6. They then revisit their argument against a base/superstructure model in which content (as economic base) is seen as determining expression (as ideology). Their key argument against this is again that there is a form of content and a form of expression, and these two forms are thus independent rather than one being dependent on the other. Furthermore, they argue that ideological models of language ignore these two forms, or separate them out as somehow abstract and eternal, rather than seeing them also as affected by historical struggle. They suggest that “expressions and statements intervene directly in productivity, in the form of a production of meaning or sign-value.” Production, then, as a way of understanding language, avoids the problems already pointed out, with seeing language as representation, information, or communication; however, it has its own problems, in that it “appeals to an ongoing dialectical miracle of the transformation of matter into meaning, content into expression, the social process into a signifying system” (90).

7. Thus, they move beyond the concept of production to a general intermingling of bodies, “including all the attractions and repulsions, sympathies and antipathies, alterations, amalgamations, penetrations, and expansions that affect bodies of all kinds in their relations to one another.” This is governed by alimentary and sexual regimes. They reiterate their understanding of tools as necessarily part of assemblages, a la Leroi-Gourhan, and show how their understanding of language is parallel to this, again showing the flatness of the machinic assemblage of bodies (which has a “primacy” over individual tools and goods) with the collective assemblage of enunciation (which has a primacy over language and words). Having moved beyond the production process as the engine of meaning, they supercede dialectics as well, stating that “the social field is defined less by its conflicts and contradictions than by the lines of flight running through it.”

8. Finally, they directly address the postulate they are refuting, namely that “there is an abstract machine of language that does not appeal to any extrinsic factor;” they explicitly link this to Chomsky’s theory of language. The problem is that this is really not abstract enough, because, excluding “extrinsic factors,” language is only a part of the assemblage. “For a true abstract machine pertains to an assemblage in its entirety; it is defined by the diagram of that assemblage” (91). [Thus, the related question of just where an assemblage ends, in all this intermingling and so on, is related to the concept of the diagrammatic, which will be discussed at more length in the next chapter.] They reiterate one of the key points they have derived from Volosinov: “the interpenetration of language and the social field and political problems lies at the deepest level of the abstract machine, not at the surface.” They end with a reference to two “states of the diagram:” 1) relative deterritorialization, in which variables of content and expression are discernable and heterogeneous; and 2) “an absolute threshold of deterritorialization” in which they can no longer be distinguished.


Third Postulate: “There are constants or universals of language which allow us to define it as a homogeneous system.”

1. This is obviously related to the previous postulate regarding an “abstract machine,” though here they are intent on dismissing the idea of “constants or universals of language” which are part of the way that linguistics is posited as a science, a la Chomsky, etc.

2. They argue against a way of incorporating pragmatics into the scientific view of language, which nevertheless tries to recreate universal or constant aspects of these pragmatics. The Langue/parole distinction, or the competence/performance distinction in Chomsky, is criticized in this way, and also as being too arborescent.

3. They go deeper into this by contrasting the approaches of Chomsky and Labov. Chomsky argues, according to D&G, that a science of language must “carve out” a “homogeneous or standard system” from the “essentially heterogeneous reality” of language (93). Variations, from this perspective, just become deviations from this established norm. However, such an approach is arbitrary. Labov, in contrast, tries to grasp how such variation works at the core of language, rather than reducing it to some secondary phenomenon.

4. They expand on this critique, using the phenomenon of [codeswitching] to argue that language is best seen in terms of “continuous variation” (94).

5. They further explore the concept of continuous variation in relation to their theory of music, chromaticism, etc.

6. They dismiss the objection that music and language are separate phenomena, invoking Rousseau to indicate how the study of language and music together could have taken a different course; Labov’s variable language rules are discussed, along with Dahomeyan chants, and “chromatic” and secret languages.

7. They call for a “generalized chromaticism,” pointing to how music has advanced from the simplistic major/minor organization into more complex and varying chromaticism. They argue that linguistics is stuck in “a kind of major mode,” and has yet to make the chromaticist leap, to understand the “immanent continuous variation” that actually characterizes language (97).

8. In a move reminiscent of Bakhtin, they discuss style as a way that such chromaticism takes shape in the works of particular authors. They summarize the method they are advocating for:

when one submits linguistic elements to a treatment producing continuous variation, when one introduces an internal pragmatics into language, one is necessarily led to treat non-linguistic elements such as gestures and instruments in the same fashion, as if the two aspects of pragmatics joined on the same line of variation, in the same continuum. (98)

Drawing on the French pun, est and et, they contrast the logic of “is” and “and:”

the first acts in language as a constant and forms the diatonic scale of language, while the second places everything in variation constituting the lines of a general chromaticism.

In other words, the “and... and... and...” logic (which is elsewhere used to describe the workings of assemblages) is like chromaticism, adding in additional notes to a chord.

9. Per Hjelmslev, the unexploited possibilities of language (aka the virtuality of language) are part of the language; change and variation are part of the machine, not something that happens to it, or that it somehow produces externally. They use the examples of “atypical expression” from e. e. cummings’ poetry, e.g., “he danced his did” (99). Such an a typical expression is a tensor, which “constitutes a cutting edge or deterritorialization of language...”

10. They summarize their argument against the third postulate: the “abstract machine of language” is “not actual, but virtual-real” (100). Through the concept of virtuality applied to language they are arguing against, e.g., the langue/parole and competence/performance distinctions; the reality of the virtual means that all of this is on the same plane, part of the same process of de/re/te in which language is formed, operates, and is eventually dissolved or transformed. Instead of being organized around constants and invariable rules, language has “optional rules that ceaselessly vary with the variation itself, as in a game in which each move changes the rules.” [This description evokes calvinball; but if so, then this is calvinball as the very essence of game itself, such that baseball, or football, and so on, could be seen as particular instances of calvinball.] To an extent, they replace the old langue/parole (etc.) distinction with abstract machine of language and collective assemblage of enunciation, but the relationship between these two is quite different. There is no primacy or hierarchy of one over the other, they are part of the same process or phenomenon of de/re/te.


Fourth Postulate: “Language Can Be Scientifically Studied Only under the Conditions of a Standard or Major Language.”

1. The four postulates, particularly the last three, are in a way restatements of each other; this last links most directly to the politics of language, and provides D&G an opportunity to drive this aspect of their argument home. They point out the problematic link between the assumptions of the linguistics they are criticizing, and nationalist homogenization and centralization of languages, e.g., French as opposed to its regional dialects. (They add the Boasian insight that these processes of homogenization are distinct and need to be understood historically before they could be studied comparatively). Chomsky’s linguistics is singled out as giving the scientific cover for what is essentially an operation of the state (somewhat ironic, given Chomsky’s politics). “Forming grammatically correct sentences is for the normal individual the prerequisite for any submission to social laws” (101). Here, they re-emphasize the deep imbrication of politics and language, calling to mind not only Althusser’s interpellation, but Augustine on the evil character of babies, due in part to their inability to speak. They end with a quote from Michèle Lalonde’s poem, “Speak White,” which is well worth reading in full, not least because D&G’s selection is far from the most powerful part of the poem.

2. Drawing on Lalonde’s poem, they discuss the concepts of major and minor languages. They note that the concept of “dialect” is complicated and problematic, and note the complicated situations of Quebeçois, which is subordinated both to “proper” French and Canadian English, and Bantu dialects in South Africa, which stand in relation to Afrikaans and English as competing dominant major languages.

3. Having established a distinction between dominant major languages, and subordinated minor ones, they then introduce (of course) two reasons for calling this distinction into question. First, Even minor languages will tend to have the same process of centralization and local dominance that characterize major languages. Second, the very position of being a major language in relation to minor languages means that those minor languages start to deterritorialize and transform the major language; thus, “Chomsky’s and Labov’s positions are constantly passing and converting into each other” (103). In a way these are not even contrary processes, but can be described as “a regulated, continuous, immanent process of variation” (emphasis original).

4. Thus, there are not really two kinds of languages, but two “possible treatments of the same language.” You can either extract constants from the variables of language, or place those variables into continuous variation. The further implication is that “constants” are not opposed to variables (they apologize if they have given this impression “only for convenience of presentation”). Instead, “constants” are just one way of treating variables, the other being continuous variation.

5. “Major” and “minor,” thus, refers not to two different languages or kinds of languages, but “two usages or functions of language” (104). Two “conjoined tendencies” or aspects of minor languages are discussed: impoverishment (or rather, restriction or ellipsis), and proliferation or overload.

6. These aspects of “poverty” and “overload” should be better seen as ways of becoming, or reterritorialization. They discuss “minor authors” and their role in transforming major languages.

7. They begin this lengthy and dense paragraph by providing new definitions for minority and majority which are not necessarily, nor even primarily, about number. “Majority implies a constant, of expression or content, serving as a standard measure by which to evaluate it” (105) [“it” being the opposition or relationship between majority and minority]. Their argument runs along the same lines as markedness theory, in which “the average adult-white-heterosexual-European-male-speaking a standard language” serves as the constant or standard of “majority” against which any other identity, no matter how numerous, is judged:

It is obvious that “man” holds the majority, even if he is less numerous than mosquitoes, children, women, blacks, peasants, homosexuals, etc. This is because he appears twice, once in the constant and again in the variable from which the constant is extracted. Majority assumes a state of domination and power, not the other way around.

This is a powerful and clear iteration of the unmarked/marked distinction, and they bring in also some of the classic feminist criticisms of traditional marxism’s blindness to gender. [They also tie into a common reading of Ulysses as a representation of colonialism (because he wanders the world, tells the cyclops his name is “Nobody” (and thus stands here for the abstract, unmarked subject), which I have always felt is a bit unfair (particularly to such an incarnation as Poldy Bloom, which they reference); surely Odysseus, the iconic embodiment of metis, could as easily be anti-colonial? Isn’t it Penelope’s suitors who better resemble, for example, the European powers gathering to carve up Africa, or a meeting this year in Davos? But I digress.] In any event, majority is founded on an abstract standard that limits, categorizes, evaluates, and hierarchizes, and minority is a subsystem that places this in variation, such that becoming-minoritarian is about becoming and potentiality.

8. Having defined the major and minor modes, we must now return to discussion of the order-word, which is “the variable of enunciation that effectuates the condition of possibility of language and defines the usage of its elements according to one of the two [modes]” (106). The question thus becomes just what it is about the order word that makes possible its treatment through these two modes. D&G turn again to Canetti, from whom they had earlier derived the concept of the order-word as a “sting” that creates a “cyst;” they now refer to this aspect as a “little death sentence” (which it seems must be an even more charged term in French). But the order word is not only a death sentence or command, it is also “a warning cry or a message to flee” (107). It thus has “two tones,” or two potentials. [There are those who live out their lives in Omelas; others walk away.]

9. They explore the first aspect of the order-word, “death as the expressed of a statement,” as an incorporeal transformation that affects or is attributed to bodies. They link to Canetti’s concept of enantiomorphosis (the editors note that this is translated into English as “prohibitions of transformation” (528n44), which seems to be a very poor translation as the original evokes a doubling of crystalline structures which ties to a philosophical debate about identity going back to Kant]. A sentence later they state that “death is the Figure,” stepping into another deep-rooted concept and debate in art history, which Deleuze has also discussed elsewhere, e.g. in “Plato and the Simulacrum.” [I also find it particularly interesting as a link to how the figure in Latour identifies/circumscribes an agentive subject in an assemblage, the form of a person in a painting, etc.] Thus, the order-word, as death and as the Figure, links again to the abstraction of constants to form a measure by which variation is evaluated and hierarchized, subordinated; there is an uncanny aspect in death/order-word/figure confronting you as your double, limiting and ending you, sending you through an incorporeal transformation. [The uncanny aspect gives an interesting hauntological link between enantiomorphosis and the doppelgänger, Diderot’s phantom, etc.]

10. They reiterate the importance of seeing this as involving both content and expression: “the incorporeal transformation is the expressed of order-words [in the plane of expression], but also the attribute of bodies” [in the plane of content] (108). They emphasize again that these two planes are always presupposing each other, but also independent, to the extent that there is “no analytic resemblance, correspondence, or conformity between the two planes.”

11. Finally, they turn to the other aspect of the order-word, “flight rather than death.” This is not a simple opposition to death/order-word/constants and so on, but a placing into continuous variation: “that is the only way, not to eliminate death, but to reduce it or make a variation itself.” They bring up the distinction between major and minor sciences, which will be returned to in a future chapter. In light of the second aspect of the order-word as flight, the content/expression opposition is dissolved [as binaries always are when used by D&G]:

the synthesizer has replaced judgment, and matter has replaced the figure or formed substance. It is no longer even appropriate to group biological, physiochemical, and energetic intensities on the one hand [i.e., the regime of bodies], and mathematical, aesthetic, linguistic, informational, semiotic intensities, etc., on the other [i.e., the regime of signs]. (109)

The second aspect draws out the “revolutionary potentiality of the order-word,” (110) which they call a pass-word, which allows us to “answer the answer of death, not by fleeing, but by making flight act and create,” and thus “transform the compositions of order into compositions of passage.”