Technology is Society made Durable

Bruno Latour, ‘Technology is Society made Durable’, pp.103-130 in:

Law, J. (Ed.), 1991. A sociology of monsters: essays on power, technology, and domination. Routledge, London New York.


In this paper I argue that in order to understand domination we have to turn away from an exclusive concern with social relations and weave them into a fabric that includes non-human actants, actants that offer the possibility of holding society together as a durable whole.

To be sure, the distinction between material infrastructure and symbolic superstructure has been useful to remind social theory of the importance of non-humans, but it is a very inaccurate portrayal of their mobilisation and engagement inside the social links. This paper aims to explore another repertoire for studying this process of mobilisation.


[on weighted hotel keys]

This minor innovation clearly illustrates the fundamental principle underlying all studies of science and technology: the force with which a speaker makes a statement is never enough, in the beginning, to predict the path that the statement will follow.

But the order that is obeyed is no longer the same as the initial order. It has been translated, not transmitted. In following it, we are not following a sentence through the context of its application, nor are we moving from language to the praxis. The program, ‘leave your key at the front desk’, which is now scrupulously executed by the majority of the customers is simply not the one we started with. Its displacement has transformed it. Customers no longer leave their room keys: instead, they get rid of an unwieldy object that deforms their pockets.

The statement is no longer the same, the customers are no longer the same, the key is no longer the same – even the hotel is no longer quite exactly the same.


This little example illustrates the ‘first principle’ of any study of innovation in science and technology: the fate of a statement is in the hands of others.


Any vocabulary we might adopt to follow the engagement of non-humans into the social link should consider both the succession of hands that transport a statement and the succession of transformations undergone by that statement.

Even with such a simple example, we can already understand that when studying science and technology, we are not to follow a given statement through a context. We are to follow the simultaneous production of a ‘text’ and a ‘context’. In other words, any division we make between society on the one hand and scientific or technical content on the other is necessarily arbitrary.

The only non-arbitrary division is the succession of distinctions between ‘naked’ and ‘loaded’ statements. These, and only these, are the distinctions and successions which make up our socio-technical world. These are the ones we must learn to document and to record.

We wish to be able to follow both the chain of speakers and their statements and the transformation of speakers and their statements. We thus define two dimensions: association (akin to the linguist’s syntagm) and substitution (or paradigm for the linguists).

To simplify even further, we can think of these as the AND dimension, which is like latitude, and the OR dimension, which plays the role of longitude. Any engagement of non-humans can be traced both by its position on the AND-OR axes and by the recording of the AND and OR positions which have successively defined it.


He [the hotel manager] is the speaker, or the enunciator – that is, the one who emits the statement. The track that the manager wishes his customers – the listeners – to follow we will call the program of action.


Such a diagram does not retrace the displacement of ‘ a n immutable statement within a context of use or application. Nor does it retrace the displacement of a technical object – in this case a key weighed down by metal – within a context of use or application. Instead, it retraces a movement which is neither linguistic, nor social, nor technical, nor pragmatic. The diagram keeps track of successive changes undergone by customers, keys, hotels, and hotel managers. It does this by recording the ways in which a (syntagmatic) displacement in the associations is ‘paid for’ by a (paradigmatic) displacement in the substitutions.

The degree of attachment of an actant to a program of action varies from version to version. The terms ‘actant’ and ‘degree of attachment’ are symmetrical – that is, they apply indifferently to both humans and non-humans. The key is strongly attached to the weight by a ring, just as the manager is very attached to his keys. It does not matter here that the first link is called ‘physical’ and the second ’emotional’ or ‘financial’.

We notice in the diagram that the social group of the hotel customers finds itself transformed little by little. The accumulation of elements – the will of the manager, the hardness of his words, the multiplicity of his signs, the weight of his keys – ends up trying the patience of some customers, who finally give up and agree to conspire with the manager, faithfully returning their keys.

This gradual transformation, however, does not apply to the ‘hotel customers’ social group alone; it also applies to the keys.


Herein lies the whole point of following innovations. Innovations show us that we never work in a world filled with actors to which fixed contours may be granted. It is not merely that their degree of attachment to a statement varies; their competence, and even their definition, can be transformed.

These transformations undergone by actors are of crucial importance to us when we follow innovations, because they reveal that the unified actor – in this case, the hotel-customer-who-forgets-the-key- is itself an association made up of elements which can be redistributed. It is opening and closing these black boxes that, until now, have made understanding the entry points of innovations such a delicate process.


Whereas the asymmetry between the feasible and the unfeasible, the real and the imagined, or the realistic and the idealistic dominates most studies of innovation, our account only recognizes variations of realization and de-realization.

The front line traced by the exploration of what holds and what does not hold together records the compatibilities and the incompatibilities of humans and non-humans – that is, the socio-logics of the worlds in which we live.

These two possible scenarios in our example show how difficult it is to avoid the twin pitfalls of sociologism and technologism. We are never faced with objects or social relations, we are faced with chains which are associations of human (H) and non-humans (NH).

Of course, an H-H-H assembly looks like social relations while a NH-NH-NH portion looks like a mechanism or a machine, but the point is that they are always integrated into longer chains. It is the chain – the syntagm – we study or its transformation – the paradigm – but it is never some of its aggregates or lumps. So instead of asking ‘is this social’, ‘is this technical or scientific’, or asking ‘are these techniques influenced by society’ or is this ‘social relation influenced by techniques’ we simply ask: has a human replaced a non-human? has a non-human replaced a human? has the competence of this actor been modified? has this actor – human or non-human – been replaced by another one? has this chain of association been extended or modified? Power is not a property of any one of those elements but of a chain.


The main difficulty of integrating technology into social theory is the lack of a narrative resource. We know how to describe human relations, we know how to describe mechanisms, we often try to alternate between context and content to talk about the influence of technology on society or vice-versa, but we are not yet expert at weaving together the two resources into an integrated whole.

This is unfortunate because whenever we discover a stable social relation, it is the introduction of some non-humans that accounts for this relative durability.

The most productive way to create new narratives has been to follow the development of an innovation (Bijker et al. 1986; Bijker and Law 1992; Hughes 1983). Those recent histories allow one to go from powerless engineers to domination that is so complete that it has become invisible. It is now the landscape in which human action and will flow effortlessly.

Consider Jenkins’s story of the simultaneous invention of the Kodak camera and of the mass market for amateur photography (Jenkins 1975, 1979). Let us abridge this story by identifying each program and anti-program and by successively recording all the new actors, be they human or non-human, single or collective.


This table summarizes a success story, that of the simultaneous building of a new object (the Kodak camera) and of a new market (the mass-market). What is remarkable in the story is that you are never faced with two repertoires – infrastructure and super- structure, techniques and economics, function and style – but with shifting assemblies of associations and substitutions.

The film is substituted to the plates, the dry collodion is substituted to the wet collodion, capitalists replace other capitalists, and above all, average consumers replace professional-amateurs. Is the final consumer forced to buy a Kodak camera? In a sense, yes, since the whole landscape is now built in such a way that there is no course of action left but to rush to the Eastman company store. However, this domination is visible only at the end of the story.

At many other steps in the story the innovation was highly flexible, negotiable, at the mercy of a contingent event. It is this variation that makes technology such an enigma for social theory.


The first of these enigmas is the notion of trajectory.

For example, the curator of a museum of technology trying to put together an exhibit on the history of photography might be tempted to link succeeding versions of early cameras in a display case.

From the perspective of the trajectory of a glass-and-wood object moving through society, these two innovations should no more be linked in a museum display case than a sewing machine and an operating table. By cutting across the translations, the notion of trajectory invents surrealist ‘cadavres exquk’.

And yet, from the perspective of the flow of associations and substitutions, there does indeed exist some link, established by Warnerke and Eastman themselves. But this link is not supported by wood, reels, or glass. The two inventions do not have a single non-human in common: they only appear to do so in retrospect.

Either we give this work a place in our analyses, in which case the link is not fortuitous, or we don’t, in which case the link between the two is nothing but an artefact of the technical history of technology.


Rather than confusing the secondary mechanism of attribution with the primary mechanism of mobilization, we should stick to the latter. An innovation is a syntagmatic line (AND) containing just as many humans and non-humans as were recruited to counter the anti-programs. If even a single segment differs from one version to the next, the innovation is simply no longer the same.


We still have the diffusionist’s (Latour 1987b) bad habit of considering that one particular segment of a program of action is the essence of an innovation, and that the others are merely context, packaging, history, or development. But the only essence of a project or of a knowledge’s claims is its total existence.

This existentialism (extended to things!) provides a precise content to the distinction between questions of rhetoric (or packaging) and substantive questions.

Is the invention of the word ‘Kodak’ important or not? Is merely deciding to build a market enough? Or is such a decision superfluous? Is the whole thing simply a marketing problem? All these questions should acquire a precise meaning: does the actor ‘the name Kodak’ lead to a modification in the durability of the syntagm, and if so how much of a modification?


Symmetrical to the illusion of a trajectory crossing a context is that of a context crossed by innovations. We need to dismiss this other sociological ghost as well if we wish to understand how the weaving of humans and non-humans is done.

Can one say that the amateur professionals of the first days of photography closed their minds to technological progress as of 1886, and that the larger public opened its mind to progress as of 1892? Can one explain the diffusion of photography by examining the nature of the social groups interested in it? In other words has the notion of interest to be stabilised in order to account for the path of the knowledge claims? No, because the social groups themselves were deeply transformed by the innovations.

They actively sorted the proposed innovations, but they also were altered, modifying their laboratories and delegating the task of plate, then paper, preparation to individual companies. What we observe is a group of variable geometry entering into a relationship with an object of variable geometry. Both get transformed. We observe a process of translation – not one of reception, rejection, resistance, or acceptance.


The same applies to the amateurs. The amateur […] who only has to click the Kodak camera, thereby imitating millions of other amateurs, and who does not need any laboratory since he can send the camera with the films to be developed at Eastman’s factories, is no longer the same as the one […] who bought intimidating cameras whose film got stuck and produced fuzzy pictures.


The amateur market was explored, extracted, and constructed from heterogeneous social groups which did not exist as such before Eastman. The new amateurs and Eastman’s camera co-produced each other. We see neither resistance to, nor opening of, nor acceptance of, nor refusal of technical progress. Instead we see millions of people, held by an innovation that they themselves hold.

And what about Eastman? Is he a fixed actor? Not at all. The contours of what Eastman can and wants to do, as well as the size and the design of his company also vary in this story. Contrary to the claims of those who want to hold either the state of technology or that of society constant, it is possible to consider a path of an innovation in which all the actors co-evolve. The unity of an innovation is not given by something which would remain constant over time, but by the moving translation of what we call, with Serres, a quasi-object (Serres 1987).

By dissolving the difference between that which mutates and the surroundings in which an innovation mutates, we should remove yet another problem: that of the asymmetry between the realizable and the unrealizable.

The real is no different from the possible, the unrealistic, the realizable, the desirable, the utopian, the absurd, the reasonable, or the costly. All these adjectives are merely ways of describing successive points along the narrative.


A major result of this manner of recording socio-logics is that ‘reality’ is not a final, definitive state demanding no further effort. A chain of associations is more real than another one if it is longer – from the perspective of the enunciator designated as a starting point in the story.

Nothing becomes real to the point of not needing a network in which to upkeep its existence. No gene pool is well adapted enough to the point that it needs not reproduce.

The only possible thing to do is to diminish the margin of negotiation or to transform the most faithful allies in black boxes. The only absolutely impossible thing is to diminish the number of associated actors while pretending at the same time that the existence of the innovation continues to be just as ‘real’. Domination is never a capital that can be stored in a bank. It has to be deployed, black-box, repaired, maintained.

Network analysis and field work have been criticized for giving interesting demonstrations of local contingencies without being able to take into account the ‘social structures’ which influence the course of local history.

Yet, as Hughes has shown in a remarkable study of electrical networks (Hughes 1979, 1983) the macro-structure of society is made of the same stuff as the micro-structure – especially in the case of innovations which originate in a garage and end up in a world that includes all garages – or, conversely, in the case of technological systems which begin as a whole world and end up on a dump.


If a version does indeed represent a progressive change of scale from micro to macro with the inclusion of greater and greater numbers of black boxes (each of which counts ‘as one’), then we can also document, using the same tool, the progressive re-opening, dispersion, and disbanding of actors passing from the macro level to the micro level.

The same innovation can lead us from a laboratory to a world and from a world to a laboratory. Respecting such changes of scale, induced by the actors themselves, is just as important as respecting the displacement of translations. Given the tools of network analysis that we have at our disposal, trying to endow actors with a fixed dimension as well as a fixed form is not only dangerous, but simply unnecessary.

It is worth noting one last consequence of substituting socio-logics to asymmetric notions of the real and the possible. The passage of time becomes the consequence of alliances and no longer the fixed, regular framework within which the observer must tell a tale. The observer has no more need for a regulated time frame than for actors with fixed contours or predetermined scales.

Just as we let actors create their respective relationships, transformations, and sizes, we also let them mark their measure of time; we even let them decide what comes before what.

Referring back to the Eastman example, thirty years elapse between versions (1) and (15), but only a few months go by between versions (25) and (30). Should we then conclude that the innovation ‘drags its feet for thirty years’ and ‘accelerates brusquely’ in 1887 as historians so often say? We could indeed reach this conclusion, but words such as ‘fast’ or ‘slow’, ‘mature’ or ‘premature7,’feasible’, ‘utopian’, ‘real’, merely float on the surface of translation movements without explaining anything.


The number and speed of events depend entirely on movements of alliance or rupture performed by the actors.


If you can reconstitute these movements, you obtain the dimension of temporality as well; if you cannot reconstitute these movements, the regular passage of time won’t tell you anything.

Admitting that we are now capable of displaying the fine variations of a socio-technical exploration, how does this ability help us explain the contingent shape adopted by a particular trajectory? The three Graces of Truth, Efficiency, and Profitability, so handy for providing causes in science, technology, and economics, are obviously unusable, as they are the result and not the cause of these displays.

But there is no question of substituting sociological interests for the three Graces as the motor of history. Stable Interests, like good Efficiency or sure Profitability, need stable networks and instruments to be able to make predictions. But the amateurs do not know that they need photography before version (36). Stockholders wait twenty years to decide whether their interests are better served by plates, films, or Kodak cameras.

Since an explanation of an innovation’s path cannot be retrospective, it can only spring from the socio-logics of programs and anti-programs. Can anti-program actors be either recruited, ignored, or rebuffed? Can program actors maintain their association if such and such an actor is recruited, ignored, or rebuffed?


To understand the path taken by an innovation, we must evaluate the resistance put up by the successive actors that it mobilizes or rejects. Explanation does not follow from description; it is description taken that much further.

We do not look for a stabilized and simplified description before we begin to propose an explanation. On the contrary, we use what they do to an innovation or a statement to define the actors, and it is from them and them alone that we extract any ’cause’ we might need. Paradoxically our explanation are ‘internalist’ in the sense that they all come from the inherent topography of specific networks.

We define an actor or an actant only by its actions in conformity with the etymology. If an innovation is defined by a diagram in which its essence is co-extensive to its existence – that is, the ever- provisional aggregate of its versions and their transformations – then these versions and transformations are in turn completely defined by the actants that constitute them.

We define an actor or an actant only by its actions in conformity with the etymology. If an innovation is defined by a diagram in which its essence is co-extensive to its existence – that is, the ever- provisional aggregate of its versions and their transformations – then these versions and transformations are in turn completely defined by the actants that constitute them.


And as for the metal weight, it does not merely intervene as a modest attachment to a hotel key. It undergoes many other tests, which define it much more completely: it melts at 1800° in a furnace, it is made up of iron or carbon, it contains up to 4% silicon, it turns white or grey when it breaks, etc.

The longer the list, the more active the actor is. The more variations that exist among the actors to which it is linked, the more polymorphous our actor is. The more it appears as being composed of different elements from version to version, the less stable its essence. Conversely, the shorter the list the less important the actor.


The more diversity it encounters among the different actors it meets, or the more difficult it is to open its black-box, the more coherent and firm it is.


The list of tests undergone by a given actor defines its historicity, just as a socio-technical graph defines the historicity of an innovation or knowledge claim.

Just as an innovation can become increasingly predictable by black-boxing longer and longer chains of associations, an actor can become so coherent as to be almost predictable.

We can thus begin to deduce the performance of actors from their competence. We are then, but only then, allowed to be normative again, but these norms are not forced onto the data, they are extracted from the actor’s own efforts at rendering each other’s behaviour more predictable. Power and domination are the words given to those stabilizations and not an account of their coming into being.

The lists constructed from the joint story of innovations and actors highlight the continual variation in an actor’s isotopy, i.e., in its stability over time. Its behaviour becomes either more and more or less and less predictable. The list allows us to go from extremely shaky certainty to necessity, or from necessity to uncertainty. The force of habit, or of habitus, will either exert itself or not; it will act or not as a function of the historical records of the actor.


In a universe of innovations solely defined by the associations and substitutions of actants, and of actants solely defined by the multiplicity of inventions in which they conspire, the translation operation becomes the essential principle of composition, of linkage, of recruitment, or of enrolment. But since there no longer exists any external point of view to which we could ascribe the degree of reality or of success of an innovation, we can only obtain an evaluation by triangulating the many points of view of the actors. It is thus crucial to be able to shift easily from one observer to another.


Until now, the starting points of all the narratives have remained stable. We told the story of the hotel keys from the manager’s perspective, and we told the Kodak story from the perspective of Eastman and Jenkins. Yet a program’s capability to counter an anti-program obviously depends on how well an actor’s conception of others corresponds to their conceptions of themselves or of the said actor.

If this convergence is weak, the actor will populate his world with other beings; but these beings will behave in an unpredictable fashion, attaching or detaching themselves to the program from version to version. If, on the other hand, this convergence is strong, the actor can begin to make predictions – or, in any case, to guarantee the consistent behaviour of the beings constituting his world.

We thus have to do more than follow the sequence of events surrounding an innovation: we should compare the different versions given by successive informants of the ‘same’ syntagm. We do not have an outside referee to test the credibility of a claim. The degree of alignment or dispersion of the accounts will be enough to evaluate the reality of a claim.


If we abandon the divide between material infrastructure on the one hand and social superstructure on the other, a much larger dose of relativism is possible. Unlike scholars who treat power and domination with special tools, we do not have to start from stable actors, from stable statements, from a stable repertoire of beliefs and interests, nor even from a stable observer.

And still, we regain the durability of social assemblage, but it is shared with the non- humans thus mobilised. When actors and points of view are aligned, then we enter a stable definition of society that looks like domination. When actors are unstable and the observers’ points of view shift endlessly we are entering a highly unstable and negotiated situation in which domination is not yet exerted.

It is as if we might call technology the moment when social assemblages gain stability by aligning actors and observers. Society and technology are not two ontologically distinct entities but more like phases of the same essential action.

By replacing those two arbitrary divisions with syntagm and paradigm, we may draw a few more methodological conclusions.

The description of socio-technical networks is often opposed to their explanation, which is supposed to come afterwards. Critics of the sociology of science and technology often suggest that even the most meticulous description of a case-study would not suffice to give an explanation of its development.

Yet nothing proves that this kind of distinction is necessary. If we display a socio-technical network – defining trajectories by actants’ association and substitution, defining actants by all the trajectories in which they enter, by following translations and, finally, by varying the observer’s point of view – we have no need to look for any additional causes. The explanation emerges once the description is saturated.


We can certainly continue to follow actants, innovations, and translation operations through other networks, but we will never find ourselves forced to abandon the task of description to take up that of explanation.


Our second conclusion relates to relativism and the heterogeneity of networks. Criticisms of studies of controversy insist on the local, soft, and inconsistent nature of the results.

Yet networks analysis tends to lead us in exactly the opposite direction. To eliminate the great divides between science/society, technology/science, macro/micro, economics/research, humans/non-humans, and rational/irrational is not to immerse ourselves in relativism and indifferentiation. Networks are not amorphous. They are highly differentiated, but their differences are fine, circumstantial, and small; thus requiring new tools and concepts.

Refusing to explain the closure of a controversy by its consequences does not mean that we are indifferent to the possibility of judgement, but only that we refuse to accept judgements that transcend the situation. For network analysis does not prevent judgement any more than it prevents differentiation.

Efficiency, truth, profitability, and interest are simply properties of networks, not of statements. Domination is an effect not a cause. In order to make a diagnosis or a decision about the absurdity, the danger, the amorality, or the unrealism of an innovation, one must first describe the network.