3 Assemblages and the Weapons of War

The political, economic, and social regime of the peoples of the steppe are less well known than their innovations in war, in the areas of offensive and defensive weapons, composition or strategy, and technological elements (the saddle, stirrup, horseshoe, harness etc.). History contests each innovation but cannot succeed in effacing the nomad traces. What the nomads invented was the man–animal–weapon, man–horse–bow assemblage. Through this assemblage of speed, the ages of metal are marked by innovation. The socketed bronze battle-ax of the Hyksos and the iron sword of the Hittites have been compared to miniature atomic bombs . . . It is commonly agreed that the nomads lost their role as innovators with the advent of firearms, in particular the cannon . . . But it was not because they did not know how to use them. Not only did armies like the Turkish army, whose nomadic traditions remained strong, develop extensive firepower, a new space, but additionally, and even more characteristically, mobile artillery was thoroughly integrated into mobile formations of wagons, pirate ships etc. If the cannon marks a limit for the nomads, it is on the contrary because it implies an economic investment that only a State apparatus can make (even commercial cities do not suffice).

Deleuze and Guattari, A Thousand Plateaus1

The whole composed of a human being, a fast riding horse, and a missile-throwing weapon like the bow is the best-known example of an assemblage of heterogeneous elements, cutting as it does across entirely different realms of reality: the personal, the biological, and the technological. This emergent whole can itself be composed into larger assemblages, like a nomad army made up of mobile cavalry formations in which warriors could fight alone or coalesce into teams, variably adjusting to the conditions of the battlefield. Among the emergent capacities that these larger assemblages had was the ability to use topographic features of the terrain for ambush and surprise, and the ability to profit from fleeting tactical opportunities: if a portion of a sedentary army lost its nerve and ran, creating a temporary opening in its ranks, a stream of nomad warriors would spontaneously ride into the breach before it closed, fanning out and spreading panic among the outflanked enemy troops. As with all assemblages, these emergent collective abilities were caused by interactions between the parts, interactions in which the parts exercise their own abilities: the flexibility of a nomad army depended on the capacity of its warriors to exercise initiative in the battlefield, a capacity slowly bred into them through intensive daily training. By contrast, the kind of military assemblage that sedentary peoples created, such as the phalanx, was an inflexible block of infantry soldiers that could exercise no personal initiative and which was therefore hard to control once the order to attack had been given. The advantage of these tight formations was that if a commander placed seasoned warriors at the borders of the phalanx and novices inside, then the formation could police itself, the experienced soldiers literally preventing those sandwiched between them from running or going to ground. This gave a phalanx the emergent capacity to hold on to terrain and to stop cavalry assaults.

This distinction between nomadic and sedentary military assemblages, one more flexible or deterritorialised, the other more rigid or territorialised, should not be taken to imply the existence of two general categories of armies. Rather, the degree of territorialisation or deterritorialisation should be treated as a variable parameter that can change historically. The sedentary armies of Europe, for example, underwent a gradual nomadisation in the last four hundred years of the second millennium. First, the phalanx was flattened in the sixteenth century by Maurice of Nassau, from the original eight-men-deep formation to three men deep, drilled daily to make effective use of muskets. These weapons were inaccurate and slow to load, but if these operations were broken down into individual movements, and if the latter were instilled as routines into soldiers, then as the first line of the phalanx was firing, the second could be aiming, and the third loading, creating a continuous stream of projectiles as the cycle was repeated. Second, the phalanx was flattened to one man deep during the Napoleonic Wars, with drill used to instil into groups of soldiers the ability to rapidly switch from a line formation to fire, to a column formation to march, and back to line to continue firing. This made tight formations much more flexible.2 And third, after a century of delay and under the continuous deterritorialising pressure of improved firearms (rifles, machine guns), tight formations were finally abandoned. They were still used in the First World War, but once portable radio transmitters became available the phalanx was broken down into relatively autonomous platoons capable of making tactical decisions on their own.

An army, sedentary or nomadic, is an assemblage of assemblages, that is, an entity produced by the recursive application of the part-to-whole relation. Thus, a nomad army is composed of many interacting cavalry teams, each composed of human–horse–bow assemblages. Similarly, and simplifying a bit, a modern army is composed of many platoons, composed of many human–rifle–radio assemblages, the human and technical components of which are themselves assemblages. Let’s illustrate this point by analysing one of these sub-assemblages: the rifle. These weapons had a highly disruptive effect on nineteenth-century warfare. During the Napoleonic Wars artillery caused 90 per cent of the casualties, but by the end of the nineteenth century this proportion of war casualties had dropped to 10 per cent because of the increasing use of rifles: the American Civil War was still fought with an even mix of muskets and rifles, but the 1870–1 Franco-Prussian war used only rifles. The increased level of lethality of the rifle can be explained if we disassemble it into its components. In particular, a musket must be loaded from the front because for gunpowder to work the breach of a musket must be perfectly sealed to contain the expanding gases the powder produces as it burns. But loading a spherical projectile through the muzzle is slow and it prevents any further aerodynamic improvements. When the projectile was given a metal jacket so that the required gunpowder confinement was created in the bullet itself, the latter could now be loaded through the breach, in effect deterritorialising the muzzle which could now be given a groove to impress a rotational motion on the bullet, thus increasing its aerodynamic stability and accuracy. Thus, to move beyond the musket the conoidal bullet itself had to become the assemblage projectile–explosive–detonator and liberate the muzzle. By modelling armies as assemblages of assemblages and allowing each nested level to have its own parameters, we can capture the complex interactions between levels. Changes in the parameters at one level (weapons) can be shown to have a cascading effect on the parameters of the larger assemblages of which they are parts (increased accuracy and range for individual soldier–rifle assemblages) that, in turn, affects the parameters of even larger assemblages: the increased lethality of entire armies that forced the abandonment of tight formations.

Let’s discuss the relations between the levels of a nested set of assemblages in more detail. At any given level causality operates in two directions at once: the bottom-up effect of the parts on the whole, and the top-down effect of the whole on its parts. On the one hand, the properties and capacities of a whole emerge from the causal interactions between its parts: many human–horse–bow assemblages, trained intensively to work together, form a whole with the emergent capacity to exploit spatial and temporal features of the battlefield. Because of this bottom-up causality the emergent properties and capacities of a whole are immanent, that is, they are irreducible to its parts but do not transcend them, in the sense that if the parts stop interacting the whole itself ceases to exist, or becomes a mere aggregation of elements. But on the other hand, once a whole emerges it can exercise its own capacities not only to interact with other wholes, as when two enemy armies face each other in battle, but to affect its own components, both constraining them and enabling them. Thus, belonging to a team of warriors makes its members subject to mutual policing, a constraint by the whole, since any loss of nerve or display of weakness by one member will be noticed by the rest of the team and affect his or her reputation. But the team also creates resources for its members, as they compensate for each other’s weaknesses and amplify each other’s strengths.

The existence of top-down causality implies that the evolution of any given assemblage will be partly autonomous and partly influenced by the environment created by the larger assemblage of which it is a part. For instance, whether a particular technical object should be considered a weapon or a tool is in some cases determined this way. Let’s take the example of the communities of hunter-gatherers of which human beings were component parts for hundreds of thousands of years. In these communal assemblages there was not only a division of labour but also collective monitoring of behaviour: the community as a whole could enforce a more or less egalitarian distribution of the spoils of the hunt. They also possessed a material culture of stone artifacts that was passed from one generation to the next. Ordinarily, the stone artifacts were merely tools, used to hunt as well as to butcher and skin a carcass, but when one community faced another in violent conflict, the same object became a weapon: the object’s properties remained the same but it was used in a very different capacity, not directed with controlled movements towards a carcass, but projected towards, or even thrown at, an enemy. In other words, the communal assemblage selected some capacities of the stone artifacts when in ‘work mode’, and other capacities when in ‘war mode’. Thus, Deleuze and Guattari argue that

the principle behind all technology is to demonstrate that a technical element remains abstract, entirely undetermined, as long as one does not relate it to an assemblage it presupposes . . . [It is the social] assemblage that determines what is a technical element at a given moment, what is its usage, its extension, its comprehension, etc.3

The way in which the authors express this idea, however, seems to deny any autonomy to the components of an assemblage. As the above quotation continues, for example, Deleuze and Guattari assert that ‘technical objects have no distinctive intrinsic characteristics’.4 If by that one means that there are no necessary and sufficient properties characterising weapons and tools as general categories, then we may agree with that assertion. Entities like ‘the weapon’ and ‘the tool’ are only reified generalities and as such have no place in assemblage theory. But to say that a technical object remains entirely undetermined runs the risk of transforming the concept of assemblage into something like a Hegelian totality in which the very identity of the parts is constituted by their relations in the whole. But as we have seen, the very definition of assemblage involves relations of exteriority, that is, relations that respect the relative autonomy of the parts. To avoid this problem it is important to distinguish two different ways in which technical objects may be characterised: by their properties and by their capacities. Let’s use a knife as an example. Its properties include its length, weight, and sharpness. These properties characterise the more or less enduring states of the knife and are therefore always actual – at any one point in time a knife is either sharp or blunt – and they emerge from the interactions between a knife’s own components: the sharpness of its blade, for example, is a geometric property of the cross-section of the blade (its triangular or pointy form), a property that emerges from a particular arrangement of its component crystals.

A sharp knife, on the other hand, also has capacities, like its capacity to cut. Unlike sharpness, the capacity to cut need not be actual, if the knife is not presently cutting something, and may never become actual if the knife is never used. And when a capacity does become actual it is never as an enduring state but as an event. Moreover, this event is always double, to cut–to be cut, because a capacity to affect must always be coupled with a capacity to be affected: a knife may be able to cut through bread, cheese, paper, or even wood, but not through a block of titanium. Another difference is that while properties are finite and may be put into a closed list, capacities to affect may not be fully enumerated because they depend on a potentially infinite number of capacities to be affected. Thus, a knife may not only have a capacity to cut but also a capacity to kill, if it happens to interact with a large enough organism with differentiated organs, that is, with an entity having the capacity to be killed. The assertion that a particular technical object is a tool or a weapon depending on the larger assemblage of which it is a part can then be interpreted as meaning that a knife as used in a ‘kitchen assemblage’ is a tool, the assemblage selecting from all its capacities only the ability to cut, while the same knife in an ‘army assemblage’ becomes a weapon, the assemblage selecting its ability to kill. But accepting this does not commit us to agreeing that the knife’s properties are determined by the larger assemblage.

To return to the example of hunter-gatherer communities, whereas their stone artifacts may indeed be viewed as relatively undetermined, the same object having different uses within a community (tool) and between conflicting communities (weapon), once tools and weapons became differentiated in form and function they acquired a certain autonomy. No doubt the progressive differentiation of technical objects was guided by the social assemblages (farms, armies, temples, workshops) of which they were parts. But they nevertheless retained their own properties, a fact that explains, for example, how they could be detached from one assemblage and plugged into another, as when firearms were transferred from sedentary armies to nomad ones. And technical objects have their own history, in the sense that the pace of technological development may accelerate relative to that of institutional development, forcing the latter to catch up. That seems to be the case with rifles and machine guns, the lethal capacities of which demanded a break with tight formations, a demand that went unsatisfied for more than a century as armies lagged behind technology, unable to give up the mutual policing function of the phalanx, and incapable of meeting the challenges presented by the new conditions on the battlefield. Thus, we must preserve both bottom-up and top-down forms of causality within assemblage theory, and never conceive of the latter as determining the very properties of an assemblage’s components.

There is another difference between the original version of the theory and the one being presented here. We have been distinguishing among the components of an assemblage those playing a material role from those playing an expressive one. The problem with the distinction between materiality and expressivity is that it can be easily confused with that between substance and form: the unformed materials out of which an assemblage is made and the way its overall form expresses its identity. But Deleuze and Guattari are very emphatic in their rejection of this: both material and expressive components have substance and form. That is, the parts of a whole are always the product of two distinct formalisations.5 For this reason Deleuze and Guattari refer to the two kinds of components as ‘segments of content’ and ‘segments of expression’. The only problem with this terminology is that it suggests something related to language, although the authors explicitly denounce this possible misinterpretation.6 A different solution is to retain the terms ‘material’ and ‘expressive’ for the components but to always treat these components as assemblages in their own right, operating at a smaller scale but also composed of formed materialities and substantial expressions. In other words, the solution is to always think in terms of assemblages of assemblages.

Having clarified these conceptual and terminological issues, we may now return to our discussion of military assemblages. Of all the different properties that these assemblages may have, the property of speed can be singled out as one of the most significant. But the concept of speed must be subjected to a philosophical treatment before we can incorporate it into assemblage theory. In particular, if we are to distinguish weapons from tools by their speed, this cannot refer to the mere slowness or rapidity of movement, since a knife may be used very rapidly to chop a vegetable or used in a slow, piercing movement to kill an enemy combatant. What matters is not the absolute speed, but the manner in which the knife is moved (projected at, rather than adapted to, a target) to solve the problem of chopping or killing. One of the components of the man–horse–weapon assemblage can be used to illustrate this. A horse, as a quadrupedal animal, has a variety of qualitatively different ways of moving, called gaits. The different gaits are needed to solve the problem of how to move faster once a critical speed has been reached. Thus, while at low speeds a horse can simply walk, there is a limit beyond which the horse cannot move any faster unless it changes to a trot; and to reach even higher speeds the horse is forced to break into a gallop.7 It is these qualitatively different ways of moving, associated with different speeds, that are important. As Deleuze and Guattari write:

It is thus necessary to make a distinction between speed and movement. A movement may be very fast, but that does not give it speed; a speed may be very slow, or even immobile, yet it is still speed. Movement is extensive; speed is intensive. Movement designates the relative character of a body considered as ‘one’, and which goes from point to point; speed, on the contrary, constitutes the absolute character of a body whose irreducible parts . . . occupy or fill a smooth space in the manner of a vortex, with the possibility of springing up at any point.8

There are several terms in this quotation that need to be elucidated. First, a distinction is made between properties that are ‘extensive’ and those that are ‘intensive’. We will explore the scientific history and usage of these terms in detail in Chapter 5, but at this point we can give a simple definition: the term ‘extensive’ is used for properties like length, area, or volume that can be added to each other yielding only a quantitative change; the term ‘intensive’, on the other hand, is applied to properties in which addition may result in a qualitative change. Examples of these properties are speed, temperature, pressure, concentration, voltage. Thus, given a body of water at 99 degrees centigrade of temperature, adding one more degree does not result in a mere change in quantity (100 degrees) but in a transformative event: the water ceases to be liquid and becomes steam. This change in quality does not occur with all additions – a body of water at 45 degrees centigrade does not transform if a few extra degrees are added – but only at critical points of intensity, those marking the boiling and freezing points of water. And similarly for speed. A given liquid can flow more or less rapidly or slowly, but at critical points a quantitative change can cause a qualitative change. At relatively slow speeds a fluid tends to move in a uniform manner called a ‘laminar flow’, but when the speed reaches a critical intensity this regime of flow becomes unstable, the fluid being incapable of moving faster while retaining uniformity, and a qualitatively different regime appears, a coherent circular motion called ‘convective flow’. At even faster speeds, this rhythmic manner of flowing cannot be kept up and the fluid is forced to change to a regime called ‘turbulent flow’.9

When dealing with entire armies in battle we must take into account different qualitative manners of moving, not just compare the rapid charge of nomad cavalry to the slow march of a sedentary phalanx. What must be contrasted are the ways in which armies flow and occupy space: flexible cavalry formations, in which teams can continuously form and disappear, like vortices in turbulent flow, and in which a temporary breach in enemy defences can be instantly responded to by a piercing cavalry charge, must be contrasted with tight formations having a limited capacity to vary their movement, advancing along predetermined paths to gain control over terrain and hold on to it. We suggested above that the difference between these two military assemblages can be captured by setting the territorialisation parameter to different values. But for this to work, territorialisation must be viewed as an intensive property, subject to threshold effects and able to generate qualitatively different states for the assemblage. In addition, we cannot think of the parameter as quantifying a simple physical property, not even one as ubiquitous as speed. Rather, it must be conceived as a complex function of several properties, some of which may not be physical. For example, the parameter must also quantify the degree to which the making of decisions in a given army is done in a centralised or decentralised way, since the manner of movement of a nomad army depended on the ability of its warriors to exercise initiative on the battlefield. Similarly, in the Napoleonic armies an increase in mobility was coupled to a more flexible chain of command, the conjunction of speed and individual initiative being what made these armies more nomad. The semi-autonomous platoon that emerged late in the First World War (the German stormtroopers), but that came into its own in the Second World War, also needed this conjunction: not only was their mobility increased but decision-making thresholds were lowered so that troops on the ground could make tactical decisions, while their commanders set only the overall strategic goal.10

A battlefield is a special social environment because of the lethality of the objects that populate it – projectiles and shrapnel; shock waves and fire – but also because it presents combatants with the problem of a situation that is in continuous variation. Armies that can take advantage of continuously varying conditions may be considered ‘nomadic’ even if they belong to sedentary organisations, and similarly for other social environments, such as the workshops or arsenals in which the weapons of war are produced.11 In this case, the producers of weapons must deal with critical points of intensity in the materials they work with, like the tendency of iron to melt at 1,535 degrees centigrade, as well as with the variable properties and capacities of the objects that result from these transformations. Thus, if the critical point at which iron crystallises is crossed rapidly (by quenching or submerging the red hot weapon into water), the resulting object acquires rigidity, the capacity to hold on to a shape. This is needed, for example, to craft the edge of a sword, which must hold on to a triangular cross-section. If the critical point is crossed slowly (by annealing or air cooling), the object acquires ductility: the capacity to yield without breaking. This is required for the body of the sword, the part that must bear the loads imposed on the weapon during hand-to-hand combat. Deleuze and Guattari use the term ‘singularity’ to refer to the critical points, and the term ‘affect’ to refer to capacities. Because both properties and capacities express the identity of an assemblage, they are sometimes grouped under the term ‘traits of expression’. As the authors write:

It would be useless to say that metallurgy is a science because it discovers constant laws, for example, the melting point of a metal at all times and in all places. For metallurgy is inseparable from several lines of variation: variation between meteorites and indigenous metals; variation between ores and proportions of metal; variation between alloys, natural and artificial; variation between the qualities that make a given operation possible or that result from a given operation . . . Let us return to the example of the saber, or rather, of crucible steel. It implies the actualisation of a first singularity, namely, the melting of the iron at high temperature; then a second singularity, the successive decarbonations; corresponding to these singularities are traits of expression – not only the hardness, sharpness, and finish, but also the undulations or designs traced by the crystallisation and resulting from the internal organisation of the cast steel. The iron sword is associated with entirely different singularities, because it is forged and not cast or molded, quenched and not air cooled, produced by the piece and not in number; its traits of expression are necessarily very different because it pierces rather than hews, attacks from the front rather than from the side . . .12

Earlier in this chapter we criticised the authors’ assertion that a technical object is entirely undetermined outside of the social assemblages in which it is used, and argued that while a social assemblage (a community or an organisation) does select what capacities of a technical object are exercised, it does not fully determine its properties. Grouping capacities and properties under the term ‘traits’ obscures this important distinction. We also suggested that the very term ‘technical object’ should be dropped as a separate concept and that we should always deal with assemblages of assemblages, technical objects themselves being nothing but assemblages. That Deleuze and Guattari thought along similar lines is clear from the following definition of the term ‘assemblage’:

We will call an assemblage every constellation of singularities and traits deducted from the flow – selected, organized, stratified – in such a way as to converge . . . artificially or naturally. An assemblage is, in this sense, a veritable invention.13

This definition does not require that assemblages be ‘collective assemblages of enunciation’, only that they have expressive elements, whatever these may be: the crystalline sound of metals, their shine or gleam, but also the way they express what they can do as they exercise electrical, mechanical, and chemical capacities. Swords and sabres are as much an assemblage as the workshop of a blacksmith in which the convergence of singularities and affects that gives rise to these weapons is effected. In the case of a workshop in which metallurgical skills are being passed from master to apprentice, we are indeed dealing with a collective assemblage of enunciation, because binding commitments can be created by speech acts, like the commands issued by the teacher or the vows expressed by the disciple. What gives these larger assemblages the power to deal with continuous variation is the fact that human beings possess capacities that can be extended and differentiated in an unlimited number of ways, capacities we normally refer to as skills. Skills, and more specifically manual skills, have been neglected in philosophical analysis of knowledge by a dismissive attitude that we inherited from the Ancient Greeks.14 This contempt does not extend beyond philosophers, since clearly within organisations like medieval guilds, members were aware of the value of practical knowledge and the importance of ritual and secrecy in its transmission. But outside of these organisations the disdainful attitude prevailed until the 1940s, when the British philosopher Gilbert Ryle broke with tradition and carefully articulated the differences between what he called ‘knowing that___’ and ‘knowing how___’. The first type of knowledge is representational, and in most cases, linguistic. What fills the blank after ‘knowing that’ is a declarative sentence like ‘Napoleon lost the battle of Waterloo.’ As such this knowledge is transmitted by lectures or books in which these sentences are uttered (or written) with the illocutionary force of a statement or a claim to truth. By contrast what fills the blank after ‘knowing how’ is a verb in the infinitive, like ‘to swim’ or ‘to ride a bicycle’, a verb expressing the exercise of an ability or bodily skill. But the most important difference is the mode of transmission: know how is taught by example and learned by doing, that is, acquired only after lengthy daily practice.15 Hence, its transmission need not involve language, although words may occur as aids to draw attention to what is being displayed by a master’s actions. Knowing how is embodied knowledge, but it is also flexible knowledge, because skills learned in one context can be adapted to many other contexts. It is this flexibility that gives the workshop, as an emergent whole, its capacity to deal with variation in materials, procedures, and products.

For most of their history, military arsenals and armouries were staffed by this kind of skilled artisan. An arsenal is already different from a workshop in that its authority structure has more hierarchical levels, and commands play a more central role. We can capture this distinction by saying that the parameters of the arsenal assemblage have higher values for both territorialisation and coding, but this difference used to be merely quantitative until the middle of the eighteenth century, when a threshold was reached and the difference became qualitative. The threshold in territorialisation was reached when the raw materials were submitted to an intensive homogenisation, while the threshold in coding resulted from the replacement of flexible skills by rigid optimised routines. The industrial regime that emerged after this transformation is very familiar to us from the factories run by Henry Ford early in the twentieth century, which is why Marxists erroneously refer to it as ‘Fordism’. But the factors that gave rise to it are much older, related to requirements closely associated with the military. Specifically, given the harsh conditions of the battlefield, weapons of war tend to break down and are often in need of repair. Repairing weapons at the rate needed in a continuously changing environment in turn requires a constant supply of spare parts. But if the weapons are created by skilled hands, with the inevitable variation this implies, then the components of one partially disabled weapon cannot be used to fix another. Hence, spare parts had to be manufactured in an industrial regime in which all sources of variation had been systematically removed.

The story of this transformation begins in 1765 when a general of the French army, Jean Baptiste de Gribeauval, initiated a series of reforms in the production of military equipment. The goal was the creation of weapons with interchangeable parts. The idea was brought to the United States by Thomas Jefferson, who had witnessed a demonstration of the new industrial regime while an ambassador to Versailles. Jefferson’s efforts to import the new organisational regime impacted both individual inventors (Eli Whitney) as well as government organisations, like the Ordnance Department and the Springfield and Harpers Ferry armouries, but the latter played a much more significant role. A network of military officers and factory superintendents transformed what was basically a concept into what came to be known as the ‘American system of manufactures’, involving the replacement of flexible skills by rigid routines, and the homogenisation of metals to make their behaviour predictable. As one historian puts it, ‘the engineering of people assumed an importance equal to the engineering of materials. As conformity supplanted individuality in the workplace, craft skills and other preindustrial traditions became a detriment to production.’16 Achieving full interchangeability of component parts also demanded detailed regulation of every aspect of manufacture and accounting. Thus, while before 1823 the inspection of the finished product was performed by eye, with the exception of the use of a caliper to check the dimensions of a gun’s barrel, after 1832 new gauges were introduced to routinise the inspection process itself. Acting as commands frozen in metal, steel gauges replaced the skill of the inspectors in the assessment of the quality of every single part of a weapon during and after the manufacturing process, so that an inspector equipped with a master set of gauges could enforce full uniformity.17 Any civilian factory that wanted to become a subcontractor to the army had to adopt the new regime, so the latter slowly propagated through the population of large industrial organisations.

These two industrial production regimes are a good illustration of top-down causality. The relatively flat hierarchy of the workshop, its reliance on flexible skills, and its ability to work with heterogeneous raw materials indicate an assemblage with low values for the territorialisation and coding parameters. This, in turn, affects the territorialisation parameter of the assemblages (weapons) that are produced: their parts are not uniform and interchangeable. The authority structure of an arsenal, on the other hand, with its deeper and more rigid chain of command, its practices of material homogenisation, and its replacement of skills by coded routines, has high values for both parameters, defining a qualitatively different regime, and this affects the assemblages produced: mass-produced identical replicas. Nevertheless, it is not the case that the social assemblage (workshop, arsenal) fully determines the identity of the ‘technical objects’ produced: an arsenal producing muskets and another producing rifles under identical conditions can still be distinguished from each other by the differences between muskets and rifles. And as we argued above, the replacement of muskets by rifles had bottom-up effects on larger assemblages, causing the further deterritorialisation of the phalanx.

To conclude, let’s compare the lessons learned here to those from the previous chapter. In both cases we were interested in understanding how the different levels of a nested set of assemblages interact. In the case of language, and also of genes, we are dealing with entities that by their very nature can affect many levels at once. Thus, a gene is an assemblage of nucleotides and, as part of a chromosome, it is itself a component of a cellular assemblage. But within the body of an animal, genes do not just affect what goes on at the level of cells, guiding the process of differentiation that produces nerve, muscle, blood, and bone cells. They also affect, through the proteins and enzymes they code for, the activity of the organs formed by populations of individual cells, and in some cases, the actions of the organism formed by many individual organs. Similarly, as we saw, a language can be viewed as an assemblage, as a component of a communal or organisational assemblage, and as a parameter of those assemblages. It is this ability to operate at many levels of scale at once that makes genes and words so special, allowing for the micro and the macro, the molecular and the molar, to be related ‘independently of order of magnitude’.18 This chapter, on the other hand, dealt with a more general form of the relation between the micro and the macro, the ascending and descending causal relations between the two: the properties of a whole are produced by the ongoing interactions between its parts, while the whole, once it is stabilised, reacts back on its parts. One of the forms that downward causality takes is selective, the whole promoting the exercise of some of its parts’ capacities, while inhibiting others. But we can also add another form that is easier to conceive thanks to the parametrised version of the concept of assemblage: if both parts and wholes are assemblages, then they both have their own parameters, and this implies that changes in the settings of the whole’s parameters can affect the settings of those of its parts, and vice versa. When a nested set of assemblages has many levels, we need to be able to keep track at what level of the nested set a given deterritorialisation or decoding is taking place, then follow its cascading effects. We will explore this problem in the following chapter.

Notes