Latour - semiotics and science studies

By Roar Høstaker
Associate Professor
Lillehammer University College

Introduction

The main question asked in this paper is how, and to what extent Bruno Latour makes use of theories from the Paris school of semiotics to transform the theoretical framework of science studies. It is commonly known that terminology like 'actant', 'translation', 'shifting in' (embrayage) and 'shifting out' (débrayage) in Latour's texts come from the Greimasian or Paris school of semiotics. Latour has himself characterized the employment of semiotic concepts as just one of several instruments "on the bench of any well-equipped 'science studies' lab" (Latour, 1993a: 131). I believe this is a too modest expression for the function of semiotics in his theory.

Instead this paper will suggest an understanding of Latour's work as fundamentally dependent upon concepts of the Paris school, although he has chosen concepts selectively from this version of structuralist semiotics. His major strategy seems to have been to extend the use of linguistic concepts to include the real and the social. In this way both the real and the social have become immanent to language. This transformation of semiotics represents a substantial contribution to the theoretical discussion in the social sciences, but his theory also inherits some of the problems entailed with this form of semiotics.

Latour's ontology

A host of theoretical influences can be traced through the texts of Bruno Latour. Of the more contemporaneous one could mention Serres, Garfinkel, Deleuze and Stengers and of the more ancient one could mention Nietzsche, Spinoza, Hume, James and Whitehead. However, the suggestion of this paper is that there is a certain homogeneity in Latour's texts coming from two major sources.

The first is the ontological principles summarized in his Irreductions a philosophical precis printed as the second part of The Pasteurization of France (1988a). Irreductions states the differentiatedness of the world and the centrality of trials of strength1: "Whatever resists trials is real" (ibid. p.158). Trials of strength was a central theme from the laboratory studies within Science Studies in the 1970s and 1980s, and the ambition of Irreductions was to generalize this principle to the ontological level. This is the basis of Latour's constructivism, and he also describes these principles as an 'associology' by which entities (also sometimes called actants) is understood to connect with each other, resist each other and gain strength by association (ibid. p.160). This ontology of 'Trials of Strength' has been much criticized, especially in Latour's work from the 1980s (cf. e.g., 1983, 1987, 1988a, 1988b). It was claimed that science was portrayed as expressions of political processes. A sort of generalized Machiavellianism or even worse: For Latour everything is war! (cf. Haraway, 1997:33-34). Although Latour has tried to accommodate his concepts to this critique, he seems faithful to the essence of the ontology. His aim has been, in his own words, to "maintain 'software compatibility'" (Crawford, 1993:266) with Irreductions.

The ontological principle of Trials of Strength is fused with the second major theoretical basis - semiotics2. Latour's dependence upon Greimasian semiotics is not always that easy to trace in his texts, as his references are often oblique or missing. As we will see, the principle of Trials of Strength is also a part of Greimas' narrative theory, since the actants go through different ordeals. The main transformation of this theory, in the hands of Latour, is the extension of the model of language taken from Greimas and Hjelmslev to both the social and the natural world. In this way Latour is able to make a field of immanence where there is no outside from which science can be used to denounce humans, or where humans can denounce science, or where the moderns can denounce the primitive or religious.

The linguistic model thus becomes part of his ontology, but it is also his epistemology. His studies always tend to throw us back into the analysis of texts where Greimas' theories of narrative analysis meet us. But even here he goes further and transforms the analysis of meaning into dramas of existence.

Greimasian semiotics - a rudimentary outline

The semiotics of A.J. Greimas was one of several different attempts in the 1950s and 1960s to develop a scientific approach to the analysis of language and texts. This usually included a rigorous theoretical apparatus in order to unveil the underlying structure of texts or the wider discourse. This could be done through a systematic reduction of the diversity of texts into a small number of functions or elements. A major aim was to make the structure of the texts plain without any recourse to the intuition of the reader or the intentions of the author. The center of attention was the texts themselves and the discourse they were a part of.

Greimas' field of semiotics is usually called 'structural semantics' and a starting point was Louis Hjelmslev's phonological model of language (1993 [1943]) in which the Saussurean model of the sign is transformed. Hjelmslev distinguishes between two parallel planes of language - that of expression (signifier) and that of content (signified). These planes presuppose each other reciprocally. In addition, within these two planes he distinguishes between form and substance (ibid.). Substance in this context is usually understood as meaning or purport3 inasmuch as they are taken on by the semiotic form (Greimas & Courtés 1982:322). The form is thus necessary for the substance to be part of the signification. This division serves Greimas and his colleagues as a way to determine the object of linguistics to include only the form of the expression or of the content. The substance, or the reference to the world outside the text, is better served by other disciplines like philosophy, sociology, literary history, etc. This is part of an undecidable domain for the semiotician. The reason why the semiotician can say something about signification is that the form organizes itself into systems of relations (cf. Hénault, 1979:28-29). The semiotician do not connect form to objects (as do other sciences), but only study relations of forms. Structuration of meaning is the particular object of structural semantics and not "problems concerning the appreciation and evaluation of the substance of meaning" (ibid.p29, my trans.).

Still following Hjelmslev, Greimas and his colleagues held that language should be studied both as system and process. In the formulation of utterances4 (enoncés) elements are combined in strings called syntagms. The elements of syntagms are recognized because they have a "both .... and" relation. Example:
"Eve bought a red dress"
The formulation of such strings represents the process of language along the syntagmatic axis. This axis is seen to be (vertically) opposed to the paradigmatic axis representing the language as a system of interchangeable elements. A paradigm is a class of elements that can occupy the same place in the syntagmatic string, and elements can be recognized by a "either .... or" relation. In the example the dress could have had another color (blue, green, pink, yellow, etc.); Eve could have bought another type of garment (trousers, skirt, sweater, etc.); or some other person than Eve could have done the purchase (Anne, Sheila, Frank, etc.); or the transaction could have been totally different (theft, loan, present, etc.). The paradigmatic axis is also called the axis of substitution and what elements are possible to substitute depends on a given context. This possibility of substitution is a virtual existence, i.e. an existence in absentia, constituted by the given context (Greimas & Courtés, 1983: 224-225, 327, 371).

The utterances (enoncés) is not produced out of nothing, but presumes the enunciation (énunciation). Enunciation is defined in two ways by Greimas and Courtés. It can either be defined as " a) the non-linguistic (referential) structure which underlines linguistic communication, or b) as a linguistic domain which is logically presupposed by the very existence of the utterance (which contains traces or markers of the enunciation)" (1982:103-105). If we follow the first definition we will delve into discussions about the communication situation and its social and psychological context, while if we follow the second definition enunciation becomes the domain of mediation that guarantees the process "by which language virtualities become discourse utterances" (ibid.p.103). This is a domain assumed to govern the passage from linguistic competence to linguistic performance. Not surprisingly, Greimas and Courtés choose to stick with the second definition in order to stay within linguistics.

The structural semantics became possible because the expression plane was considered ".. to be made up of differential gaps and considering these gaps of the signifier to correspond to gaps of the signified" (Greimas & Courtés, 1983:272). This approach made it possible to analyze lexical units by decomposing them into smaller underlying units. For Greimas the goal of this line of work was to map the minimal units of signification. These were called semes. On an elementary level semes could be arranged into sememes, which is a composition of semes for a given term in a given context. In the book Structural Semantics (Greimas,1983: 50-55) Greimas discusses the semantic field of the lexeme (term) "head". This is a term that may enter into many different contexts, for instance, expressions like "the head of the column" or "the needle-head" or "he could not leave it out of his head". Nonetheless, the sememe /head/ will always contain the abstract semes of /spheroidity/ + /extremity/ + /superativity/. These are nuclear semes that will always be a part of the sememe /head/, but the context of the lexeme "head" will also enter the meaning of /head/. These are the contextual semes or classemes. Classemes represent general classifications that are present in a given natural language and is arranged in binary oppositions. The search for minimal units of signification thus entailed a reduction of the variety of possible significations into a few abstract notions, and texts can be analyzed on several different levels in order to reveal meaning.

Relevant classemes for /head/ in the phrase "he could not leave it out of his head" is (among others) /human/ and /male/ and /singular/. In Western culture /human/ is usually opposed to /animal/. The classemes of /human/ and /animal/ taken together represent /animate/ as opposed to /inanimate/. In this way classemes build hierarchical relations:

Classemes are important because as general categorical schemes they contribute to the continuity of signification. A coherence of discourse is made possible at the elementary level by the repetition of elements that are similar or compatible. In the phrase "the cat purrs" /purr/ implicates the classemes /subject/, /singular/ and /animate/. As we find from Figure 2 /animate/ can be either /animal/ or /human/. The subject "cat" in the phrase implies /animal/ to be the right choice among the two classemes. The presence of the element "cat" in the phrase thus reduces the possible meanings of the phrase and it can be condensed into a 'common denominator' /animality/ (Hénault, 1979:80-82). This is called isotopy and can be followed along a syntagmatic chain of classemes that secure the homogeneity of utterances and discourse. Similarly Greimas and colleagues hold that other types of isotopies can be found. Texts also often follow actors, and actorial continuity is usually secured by anaphorization. In the phrase "Eve bought a red dress, but it didn't fit her well" actorial continuity is secured because the terms "Eve" and "her" denominate the same actor as do "dress" and "it". Greimas and colleagues call this form of continuity actorial isotopy. Furthermore, semantic isotopy makes possible a uniform reading of longer texts where the reader tries to resolve ambiguities and find dominant isotopies underlying the texts (Greimas & Courtés, 1982:163-165).

The subject of semantic or dominant isotopies in texts can be analyzed further in the relationship between a category and its opposite. Linguists early discovered that binary oppositions somehow governed texts. An example could be the common opposition between 'beautiful' and 'ugly'. However, this opposition differs from the mere contradiction of 'beautiful' vs. 'not beautiful'. If a person is said to be not beautiful, he or she does not have to be ugly, but you cannot say that someone is not beautiful and beautiful at the same time. The assessment 'not beautiful' presupposes beautiful. It is also implied that someone deemed ugly is also not beautiful. Similarly, to be not ugly is not the same as being beautiful although by implication a person who is deemed beautiful is also not ugly. Greimas collected these relations of contradictions and oppositions5 in the semiotic square:

Following this figure the meaning of 'beautiful' is presupposed in its oppsition to 'ugly', its contradiction to 'not beautiful' and by being a complement to 'not ugly'. In analyses of texts this give rise to transformations of meaning between the poles of the square. In the book Maupassant Greimas analyses the short story Deux Amies (Two Friends) according to the 'tendency' of the text to move between the oppositions of life and death (1976). Already the first phrase of the text sets the tone of the story:

"Paris était bloqué, affamé et râlant" (Paris was besieged, starved and rattling)6

Between life and death we can form the following semiotic square:

The opposition between life and death might be seen to be absolute - either there is death or there is life - no exception allowed. Greimas, on his side, argues that the lexemes "affamé" (starved) and "râlant" (rattling) introduce dimensions between life and death in the text. There is an uncertainty about the fate of the subject of the phrase ("Paris") - will it survive or will it die. While "affamé" describes a situation of living that is precarious, and can be transformed in the direction of life or non-life, "râlant" describes the process of dying - a transformation from the condition of non-death to death (1976:23-25). The expression "râlant" signify, of course, a way of being alive, but it still contains a previous negation of life through the absence of death. On its side, lexemes of non-life can be said to express be the absence of life, but still contain a previous negation of death. The tension between the lexemes of "affamé" and "râlant" portrays the slow exhaustion of the city.

In the phrase above "Paris" can also be described as a humanized actor in a drama (cf. ibid.p.23-24), in other words, as part of a narrative. Greimas based much of his work on narrative structures on Vladimir Propp's study of 100 Russian folktales. Propp formulated 31 different functions to describe situations in these folk-tales. Any folktale could be described as a combination of these functions (1968). Greimas condensed and generalized this schema into three functions and six actantial categories7. The categories were defined by their role in relation to the development of the story:

1. The subject lacks a certain object (understood in wide terms). In the folktale the object might be person or thing missing after a misdeed by a villain.
2. The sender passes a contract with the receiver in order to liquidate the lack. In the folktale the sender is often a person from the social hierarchy (queen, king, mother, father) giving the receiver an obligation.
3. The subject, with or without a helper, accomplishes to obtain the object. This act may take the form of a combat with an opponent (or anti-subject). This is the test.

Terms in Italics are actantial categories while terms in bold are functions of the narrative8.

The actantial categories are only roles in the narrative and often different actants form the same actor by syncretization. The receiver and the subject are usually the same actor, but not always. In a text like The Quest for the Holy Grail the sender is God, the receiver is Mankind, the subject the Hero and the object is the Holy Grail (cf. Greimas, 1983:204). The sender is somehow thought to be transcendent in relation to the action that takes place. The sender delegates a task to the receiver-subject and (often) sanctions the results of action. The sender possesses some knowledge that he transmits to the receiver and also frames the action. What makes a text a narrative text is, according to this theory, a state of dispossession or possession of some valued object leading to some action to produce the opposite state of dispossession or possession (Hénault, 1979:145).

During the 1970s Greimas transformed the narrative schema described above by concentrating on the relationship between the subject and the object as the principal actants of transformations. Utterances about the relation between a subject and an object were seen to take two basic forms - either an utterance of state (être) or an utterance of doing (faire). The justification for this theoretical articulation was that it allowed for a unified conceptualization of descriptions and transformations (Hénault, 1983, ch.2). In syntagmatic chains utterances will take other utterances as their object and modalize them. When an utterance (of state or doing) governs another utterance (of state or doing), the first utterance is said to be a modal utterance while the second one a descriptive utterance (ibid.).

When an utterance of doing modalizes an utterance of state we have narrative performance. In the phrase "Eve bought a red dress" the action (buying) changes the state of Eve from a state of non-possession to a state of possession (of the dress). Performance amounts to the realized action in narratives. However, this realization presupposes the existence of some virtuality or potentiality for action. In other words, some state must lead to this or that form of doing. The subject in the phrase (Eve) must have a certain competence to do what she does. Greimas and colleagues distinguish between four modal values of competence. The subject can be seen to have knowing (savoir) about what to do, to be wanting (vouloir) to do something, to have to (devoir) do someting, and to be able to (pouvoir) do something (ibid.). Typically a hero acquire different forms of competence during the course of a story9.

At a superficial level Greimas has come to distinguish between two types of narrative dimensions: the pragmatic dimension and the cognitive dimension. The pragmatic dimension includes the programmed forms of doing in the narratives - the action - while the cognitive dimension includes a knowledge and the reflections over the action (Greimas & Courtés, 1982:32-34, 240-241). This is similar to what has been said above about the relationship between sender and receiver. In narratives of this type utterances of doing are often modalized by utterances of doing (called manipulation10) or utterances of state is modalized by utterances of state (called veridiction). In traditional stories manipulation takes place in the initial phase of the story where the sender have some knowledge that he reveals to the receiver in order to make the contract. This situation forms the basis for later action11. Veridiction concerns those parts of a narrative that install a distance between what seems and what is. Trickery and discovery is the source of endless plots in traditional stories, and when the mission of the hero is accomplished, there is usually a scene where the sender sanctions the actions of the hero (and rewards him)12. Many stories thus follow this schema:
Manipulation - Qualification/Action - Sanction
Through manipulation and sanction the sender frame the action from a transcendent level, and the pragmatic dimension serves as an internal referent for the cognitive dimension.

Many texts are not narrative texts of the traditional kind and in all texts where some object of knowledge is sought may be called cognitive texts. The identification of the cognitive dimension opens up for the analyses of discourses where the knowing subject is important, as in history, literature, politics, advertising and science. This knowing (or cognitive subject) (sender or receiver) is formed by the enunciator in different ways. Greimas distinguishes between different types of cognitive doing. For instance, in informative doing the subject reports or receives (factual) information, while in persuasive doing the enunciator (e.g. the assumed author of the text) tries to have the enunciatee (the assumed reader) to accept a contract. In interpretative doing the focus is on the cognitive subject as enunciatee assessing and summoning under what conditions it is willing to accept a given interpretation (ibid. pp. 33, 157, 160-161, 230-231).

Latour's anthropology of science

How do Greimas' theories enter into Latour's work? It is possible to recognize two different ways in which it is used. The first one is the merger of some notions like actant and translation with the ontology of 'Trials of Strength' as found in Irreductions (Latour, 1988a). Here actants are minimal entities associating with other actants and forming joint vectors of agency through translation, etc. This understanding of actant fits well into the definition as "that which accomplishes or undergoes an act, independently of all other determinations" (Greimas & Courtés, 1982:5). However, the Greimasian project is concerned with meaning, and not with the construction of an all-encompassing ontology. Language is only one of several domains. The second way Latour uses literary theory seems to be more in line with this view. Latour in many ways extends the linguistic model to the world drawing on some of the insights from the Hjelmslevian distinction between form and substance13: the substance (or the matter) of the world cannot have any signification without linguistic form. Substance has to take on form (of expression and of content) to establish signification. Since linguistics is the study of relations between forms (cf. above) it might be a handy tool for the study of how humans establish meaning. Enter Greimasian semiotics.

The advantage of literary theory is seen by Latour to be its openness in relation to the constructions of the texts: "Semiotics is the ethnomethodology of texts. Like ethnomethodology, it helps replace the analyst's prejudiced and limited vocabulary by the actor's activity of world. To be sure, one cannot stop at the study of one text in isolation - but when adding other documents, other sources, other methods, the lessons learned from semiotics must be retained. There are mediators all the way down, and adding sources will only add more mediations, none of them being reducible to mere 'document' or 'information'" (Latour, 1993a: 131). This passage contains some of the essence of Latour's methodology: follow how the agents themselves construct their world and do not at any point make any additional assumptions about how real their associations are (cf. Latour, 1987:205).

I will in the following try to follow some of the application by Latour of arguments and models derived from Greimasian semiotics. In a way, we might say that I try to read his texts in the light of Greimasian theory.

The mobilization of the world

In comparison with much previous sociology of science, constructivist science studies have followed in detail the 'mobilization of the world' made by scientists. Latour has combined this form of research with discussions of how to understand the construction of scientific objects and realist theory of science has usually been a target for critique (Latour, 1987, 1999a, Latour & Woolgar, 1986). This theory of science, so much part of our common sense, presupposes a fundamental discontinuity between word and world. Our words are mere descriptions or depictions of a not really knowable real world. Here Latour's extension of Hjelmslevian linguistics to the world (cf. above) forms the basis of his argument.

We can follow an example from Latour's participation in a pedological and botanical fieldwork in Boa Vista in Brazil (1999a, Ch.2). The group of scientists should study the border between an area of the tropical rainforest and the savanna. Before they go to work there is only undifferentiated forest and savanna. They could, of course, describe the forest or the savanna after having walked around in it, and that has been done many times in literature. The scientific enterprise, however, presupposes the establishment of a systematic reference to the object under study. This is obtained by the group by first dividing the area into numbered squares and then take samples of leaves and soil within each square. All samples are numbered and thus linked to their particular square. The soil samples are collected into a frame - the pedocomparator - mimicking the square structure of the area. The pedologist thereby obtains a synoptic representation of the soil in the area by this organization of the lumps of earth. This device makes it possible for him to assess the qualities of the soil at a glance and it makes it possible to produce a graphical representation of the soil in the area. Later these samples are moved to Paris for further analysis and the results entered the scientific literature in the form of reports and papers (ibid.).

An important argument for Latour here is that the world can nowhere be seen apart from words. If we limit ourselves to the non-scientific description where we walk around in the forest, we can of course sense the forest, but we cannot produce signification about the forest (even in our thoughts) without adapting linguistic form to substance. In the research enterprise this is even clearer. When the researchers divide the area into numbered squares it is by the means of the forms made by the science of geometry and arithmetic. When the researchers collect samples of the foliage and the soil, the leaves and the lumps of soil are not just leaves and earth. They have taken on the form given to them by the researchers. They have taken on significance beyond being just leaves or pieces of earth - they have become representatives of some part of the area. They re-present the forest and the lumps of earth and the leaves can again be re-presented in the graphic table and the graphic table can be re-presented by descriptions in texts, etc. (ibid.).

Interpreted in semiotic terms this process of collection and re-presentation of the forest and the savanna shows the process how a science produces its internal referent. Each step of the re-presentation involves the cognitive activities of the enunciator (the researchers) through successive steps of shifting out (débrayage)14 by which they frame and stage the object they want to say something about (the relation between forest and savanna). First, the shifting out from the enunciator to the 'frame' of the geometrical pattern, then from these numbered squares to the samples of earth, from the samples to the pedocomparator and from this device to the graphical drawing. The enunciator can at each step, without problems, shift in and say something about the condition of the forest and the savanna in Boa Vista15. For each step in the chain of shiftings out something is delegated with the task of re-presenting the object the enunciator really wants to say something about.

This chain of translations from the forest in Bahia to the transportable samples to the even more transportable graphical drawing, constitute a fundamental part of what we might call the enunciative regime of science. A translation is defined by Greimas & Courtés as "the passage from a given utterance to another utterance considered as equivalent" (1982:351). Latour's use of the term is more material as he lets objects stand in as equivalents of each other. This is another example of how he extends the linguistic model to the world. However, this process of translations depends at each stage on the relevance and the accuracy of the 'framing' made by the enunciator. Only with a particular 'framing' is the chain of translations allowed to be seen as information about the object in question. It is this 'framing' that allows the sciences to claim that they speak about an external referent and not only an internal one. However, for Latour every notion of an external referent is meaningless16. Reference can only mean the chain of translations of internal referents. To what degree scientists speak truthfully about the nature depends upon the quality of this chain (1999a: 310).

A consequence of this view is that there is no known or knowable reality that is a non-linguistic reality. That does not mean that everything is language, but science has become immanent to language. There is no place where reality can hide outside language, a view that amounts to a rebuttal of realist epistemology. The rupture between word and world stated by this theory involve presuppositions of an undiffereciated background existing independent of human knowledge. When scientists propose some state of fact about the world, a common argument is that this factual entity always has been present, even before the fact. The facts are potentially knowable for us in a slumbering state and when they are discovered they just manifest their potency. The real hero is the entity itself and not the humans who describe them and who make it possible for this knowledge to be formulated. This model is not only essentialist by presuming a hidden truth waiting to be able to unleash its potentiality, but also religious as it intervenes from a world beyond. In no other situation is this clearer than in scientific controversies. From a realist standpoint the production of an artefact is usually described in social terms while the acceptance of a fact is described as being the thing itself. Everything happens as if the reality of the beyond somehow confirms or denies the efforts of man in his study of nature.

Enunciation and context

We have seen so far that Latour tends to extend the linguistic model to encompass also what Greimas and colleagues tend to see as the non-linguistic domain. If the external referent has disappeared and become part of the internal referent of language, a number of the distinctions of semiotics might have become obsolete. We saw above how Greimas and colleagues distinguished between a non-linguistic and a linguistic definition of enunciation. The non-linguistic definition presupposed some "communication situation" or "psycho-sociological context"17(Greimas & Courtés, 1982:103) to explain the production of utterances. This is similar to sociology where a common view claims the existence of a social context in which the agents have to be situated in order to explain why they do what they do. This form of contextual explanations has been one of the main targets of Latour's critique of the social sciences.

Within the social sciences the acts of individual agents is often tied to social interests, outer determining causes or similar entities. Realized acts are aligned with this pre-established context and suddenly the researcher has an explanation for why agents actually did what they did. Classical sociology ".. knows more than the 'actors'; it sees right through them to the social structure or the destiny of which they are the patients. It can judge their behavior because it has fixed reference frames with respect to which the patients behave in a pathological fashion" (1996a: 199). Latour does not deny the existence of a social context, but action cannot be explained by context. Action does not reside as some potential either outside agents or inside agents, because we cannot know why agents really do what they do. That is a form of information we do not have access to (1988a, 1996a). All presuppositions of access to why agents really do what they do amounts to another version of the intervention from something beyond our knowledge. In other to maintain the immanent character of the research enterprise, Latour chooses an understanding of social context that is compatible with the linguistic model.

Greimas & Courtés defines context as "the entire text which precedes and/or accompanies the syntagmatic unit under consideration and upon which the signification depends" (1982:58). They further distinguish between explicit (or linguistic) context and implicit (or extra-linguistic or situational) context. The implicit context is important for the understanding of texts, but can only be called upon for semantic interpretation when it has been made explicit (ibid.p.58). Latour solves the problem of making the situational context explicit by analyzing how agents involved in the same field or controversy link together and combine actors: "whom they endow with qualities, to whom they give a past, to whom they attribute motivations, visions, goals, targets, and desires, and whose margin of maneuver they define. (..) For a given actor, this is the way the strategy of the other actors is interdefined" (1996a: 163). Here the term 'actors' should be understood linguistically as involving both individual, collective, figurative (anthropomorphic or zoomorphic) or non-figurative ones (cf. Greimas & Courtés, 1982:7).

In this way social agents could be seen as engaged in contextualization: they analyze and interpret their social context and direct their actions in relation to some part of this context. Social context becomes a resource for the agent in his/her production of agency. The different elements of the context might thus be made explicit in the same way as meaning is produced through the realization of language: by inter-definitions (1988a: 9-10). In the case of the ARAMIS project, Latour shows how the agents negotiate between different contexts and try to build isotopies connecting them in order to define the meaning and the development of the project (ibid. pp.171-176). All these social actions are connected to the (inter-) definitions of meaning, and may thereby become the object of semiotic-based analyses.

These different interdefinitions of meaning nonetheless throw us back at the particular utterance - and especially texts - as the starting point for the analysis of a controversy or a field. Latour always seems to move from particular texts (and what they do) to the 'wider picture'.

Scientific texts

The study of scientific texts as part of a scientific controversy has become one of the central elements of the Latourian repertory. It is in and through these texts he has found much of the drama of science that he is so concerned with. Many of his concrete analyses of texts are highly popularized, and I will in the following try to give the analysis of scientific texts a slightly more technical interpretation both to enhance the heritage from Greimas and colleagues and to point at Latour's originality. I will do this by a re-reading of Latour's analysis of Louis Pasteur's famous article on lactic acid fermentation18 (Latour, 1993a).

Seen superficially Latour invites us to a reading of Pasteur's article according to Greimas' narrative schema where Pasteur is the subject of a fairy tale (Prince Charming), the yeast is the object (Cinderella) and the theories of Liebig and Berzelius the anti-subject (ibid. p.132-135). I will instead follow a model suggested by Latour's (much-regretted) colleague Françoise Bastide. She argues that objects of science could be analyzed as cognitive objects. Cognitive objects are more complex than pragmatic objects (like Cinderella) and go through more profound changes (Bastide, 1981). In the case of the lactic ferment, it changes from a non-existent entity to a plant-like entity in the course of a few pages. The original dispossession of the object is a lack of knowledge and the possession of the object is to obtain this knowledge.

Following Bastide (and Latour) a scientific article is first and foremost an act of persuasion. It is supposed to persuade a reader outside the text to make him or her to believe something. This act of persuasion is built into the text by acts on the cognitive level of making known (faire savoir) or making believe (faire croire). To make someone believe something (e.g. about ferments) is part of a polemical situation, while to make something known is marked by the absence of polemics (Bastide, 1981). To make something known is to present something without alternatives and with little or no modalization, while those parts of the text that make the presumed reader (enunciatee) believe something is rife with modalizations . The descriptive parts of the text refer to acknowledged facts within the field, but also when the article refers to experiments or reports empirical findings it is usually in the form of making known. Bastide analyses this part of a text as an informative doing in which the enunciator places himself/herself as a witness of the performance of humans and non-humans in the experiment. He/she reports (emissive doing) what is done and listens or sees what happens (receptive doing) (ibid.). Often the scientific texts take the overall cognitive structure of:

persuasive doing - informative doing (experiment) - persuasive doing

The parts of the text written as informative doing is part of the overall making believe of the whole article (ibid.). The description of the experiments forms the internal referent of the text and is written as performances and at any time the enunciator can shift in to comment on it and change the course of the narrative.

In the beginning of his article Pasteur poses the problem concerning lactic fermentation: "Everybody today knows that by adding chalk to sweetened water, and in addition a nitrogeneous matter like casein, gluten, animal membranes, fibrin, albumin, etc., the sugar transforms itself into lactic acid. But the explanation of the phenomenon is highly obscure. One ignores completely the mode of action of the plastic nitrogeneous matter20" (Pasteur, 1922:5-6, §5, my trans.). This segment of the text states a dysphoric situation - the disposession of the explanation of lactic acid fermentation. It is this initial lack the enunciator is to do something about. His quest could have ended quite early in the paper since Pasteur already in the next paragraph states that: "The facts seems, however, to be very favorable to the ideas of M. Liebieg or to those of Berzelius21" (ibid. p.6, §6, my trans.). These theories denied the existence of organized beings as the cause of fermentation. Pasteur, on his side, claimed the existence of a particular organized being in the same manner as the alcoholic ferment established by him in a previous article. Following Bastide, this is part of the construction of the initial situation of making to believe (faire croire): there are two different views and the author creates a highly polemical situation by favoring one of the views (Bastide, 1981:35). The dissymmetry created between the views paves the way for the presentation of Pasteur's experiments.

I will only discuss some parts of this presentation of experiments in order to highlight the relationship between performance and informative doing. As we have seen above the modalization of an utterance of state by an utterance of doing forms narrative performance22, and when Pasteur describes what his experiments he follow this particular mode of writing. In § 8 in his article he describes how it is sometimes possible, when carefully examined, to find a gray matter above the deposit of the chalk of an ordinary lactic ferment. It is possible to examine it in a microscope without much result, and often it is not possible to discern the gray matter at all: "Nevertheless, it is it [the gray matter] that will play the major part23" (Pasteur, 1922:7, §8, my trans.). Here we suddenly leave the messy circumstances of the pragmatic dimension and the enunciator shifts in to focus our attention and to change the direction of the text.

One of the consequences from this change of direction is the attempt in the next paragraph to show what the gray matter can do. He describes an experiment with a limpid nitrogeneous solution in which he dissolves sugar; he adds chalk and a trace of the gray matter. He keeps the solution at a good temperature and chases out the air with carbon dioxide. The next day there is a regular and lively fermentation in the vessel, and he describes different observations in connection with the development of this fermentation and similar ones. Finally he concludes: "In one word, we have under our eyes a clearly characterized lactic fermentation, with all the accidents and common complications involved with this phenomenon, well known by chemists by their external manifestations24" (ibid. p.8, §9, my trans.). The enunciator has once more shifted in to inform the assumed reader (the enunciatee) about what is going on in the vessel.

Both of these shiftings in of the enunciator are examples of informative doing. While the descriptions of performances leave us with a bewildering number of actions on changing matter, these shiftings in of the enunciator ensures that the assumed reader gets the point: 1. The gray matter is the important element and 2. The gray matter produces a lactic fermentation under certain circumstances. In the first case we only have his word for the part played by the gray matter, while in the second case the common competence of chemists is presumed by our 'expert witness'. By these acts of informative doing the properties of the gray matter have changed, and one of Latour's major points is that in experiments a substance starts as a 'name of action' (1999a: 308), or a list of performances, and only later will gain its essence (1993a: 136). While the subject of the folktale usually acquires competence before performance, the opposite is the case with the substance of an experiment. It is the nature of this competence that is the searched for cognitive object.

The originality of Latour, compared to the linguistic theories that he uses, lies in his transformation of the interpretation of the Pasteur's article into an ontological drama. In a similar way as described earlier Latour extends the linguistic model both to the social and the natural world. Through the text the lactic ferment comes into existence and is stabilized as an organism, while Pasteur establishes himself as a spokesperson for the ferment. The article is a vehicle for both Pasteur and the ferment whose existence is about to be tested by the tribunals of science: First by the Society of Sciences, Agriculture and the Arts of Lille and then by the Academy of Sciences in Paris. Both the statuses of the ferment and of Pasteur change as a consequence. The ferment enters into the networks of scientific laboratories and agricultural industry, while Pasteur is promoted to a professorship at the École Normale Superieure in Paris. If Pasteur had not succeeded in his persuasion of his colleagues he might have become a less important scientific figure, and the ferment might not have moved beyond his laboratory in Lille (cf. Latour, 1999a, Chapter 4).

The goal of Pasteur in our example is to show that the ferment is an entity of nature independent of man, but at the same time he has to construct the experiment. In other words, does science say something about nature or is it constructed by humans? Scientists often tend to switch between social explanations or natural explanations according to how successful an experiment has been. If it is a success the substance of the experiment is part of nature independent of man, while if it is not, the setup of the experiment was wrong, the researchers were biased or incompetent, the substance does not exist at all, etc. (Latour & Woolgar, 1986). The view of Latour is not to judge whether scientists are wrong when they seem to contradict themselves by switching between social and natural explanations. Instead he argues that scientific discourse forms a frame of reference where humans are active and another where the substances are active (1993a: 141-142). The function of the informative doing in Pasteur's text is to make us shift out from the frame of human action and focus our attention and see the process of independent nature. But even in the descriptions of the performances in §9 show the change of agency from humans to non-humans. The first half of the paragraph describes the setup of the experiment by humans, then it describes the performances of the liquid, the chalk and the gas, then again the humans intervene to disseminate the results of the fermentation, and finally the enunciator shifts in and tells us what has happened. In other words, when the experiment has been set up the experimenter steps back and lets the natural substances take their course. In the experiment "Pasteur acts so that the yeast acts alone" (ibid. p.141, emphasis in orig.).

What makes Pasteur's paper into an event is the fact that a new object is about to enter the world. The ferment is at its most unstable. It is at its institutional zero-point where it has not yet entered into the wider networks of scientific laboratories, dairies, textbooks, etc. It has not yet become an object with a taken for granted essence, but is still fragile. The shifting in the text between the world of human agency and the world of objective agency shows how the ferment is constructed through human action in order to be able to act alone. The ferment is both constructed and real, a hybrid or a quasi-object25 (cf. Latour, 1993b). This initial weakness of the object makes it possible for Latour to criticize the properties of some of our cultural categories. Since scientific objects clearly are forms of a nature-culture we cannot still make utterances about nature as if it is completely independent of humans. Nature cannot be seen as independent of our efforts to study it, and society cannot be seen as independent of nature or material objects. Our ways of producing discourse about culture or society implies that it is warm, changeable, talkative and rife with conflicts, while science is cold, solid, silent and consensual. Science can therefore be used to make the chattering multitude shut up. Latour criticizes this view as a part of a modernist constitution in which we divide ourselves from the past, from 'primitive societies' and from God by the reference to our 'solid' scientific objects. Simultaneously we divide our own contemporary world in two - nature and culture - where the objects from the former can be used as a resource to denounce the latter. Simultaneously the inhumanity of science can be denounced by the very much human culture (1993b, 1996b, 1997, 1999a).

Latour's movement from the construction of a scientific object to a major critique of our cultural categories can be interpreted as a movement from the text and towards the classematic categories, or contextual semes, that form our enunciations about science and society. As we have seen above, classemes form a tree-like structure with sets of clear oppositions like nature/culture, nature/society, scientific/social, object/subject, fact/fiction, fact/fetish, etc. These are the oppositions that Latour wants to transform. Since the objects were so clearly constructed and held together socially in the experiment, it makes them less hard and society (or culture) less weak. They have become part of the same immanent field held together by language. There is no longer an outside from which we can denounce society. Neither can we make society into an outside from which we can denounce science for being inhuman.

Critical remarks

We have so far seen that Latour tends to move from the analysis of particular texts and 'outwards' to the context made by the social agents themselves in the texts. The agent contextualizes through the texts and implicit context is made explicit by the study of different texts engaged in the same discourse. Through their texts the agents give each other a history, visions, goals, interests, targets and desires. The agents are defined by their mutual 'interdefinitions'. In this way Latour avoids the predefined context of traditional sociology where the researcher is able to 'see through' the motives of agents and to 'explain' their behavior by reference to the social context. Context-oriented sociology seems to presuppose a superior or underlying level where the meaning of the multiplying actions of agents seems to converge. Context understood as interdefinitions, on the other hand, lead us to another conclusion. The more texts we include the more the context will diverge in different directions. While there might be local centers where a certain degree of globality is produced, even these are local and particular (Latour, 1988a: 253). There is no superior or underlying level in Latour's theory. Everything is on the same plane and will pull in different directions.

There is a question whether Latour here enters into difficulties that he shares with much actor-oriented sociology. Since he depends upon the agents' own inter-definitions he is limited to the agents' own taken-for-granted world. By the analysis of inter-definitions he can make explicit many of the implicit notions in texts, but the analysis stops at what the agents have in common. This shared implicit world becomes a setting that he can describe and analyze, but whose own principles cannot be analyzed (cf. Alexander, 1982). In his book about Pasteur and how microbiology changed the understanding of contagion and illness, Latour mentions that an important motivation for the scientific effort in France in this period was to avenge the war of 1870 - 1871 and to regain the lost provinces of Alsace and Lorraine. Everything that made the French people stronger would make this goal more attainable. The regeneration of man was the slogan of the day. Latour made this into a part of the setting in order to study how Pasteur transformed the field of hygiene (1988a), but an important question is how does this setting influence the agents and the events taking place. Latour rather stops short when he reaches the implied social context and the result is a form of 'endogenization' where the setting closes itself around the events taking place (cf. Chateauraynaud, 1991:478).

This 'endogenization' may be seen as a result of Latour's extension of the linguistic model by making the social immanent to language. This immanentism makes all references to the outside of a setting into something that intervenes into the setting. The outside suddenly explains the action of the agents, or the researcher can take a divine perspective and understand the agents better than they understand themselves, etc. I will in the following try to show that the outside of a given setting (or the social transcendence) does not necessarily have to be understood in this way. The crucial question is how the setting itself is constituted. In his book about Pasteur Latour presupposed the national state as the major frame for scientific activities in France at the end of the 19th century. The national desire for revenge somehow produced a convergence of very diverse scientific efforts (1988a). This does not imply that the only meaning of all these activities was to strengthen the nation. The national desire for revenge cannot be used to 'look through' the motives of the agents or to reduce their actions to this project. Nonetheless, this setting must have imposed strong limitations concerning possible actions. In the same way as Pasteur had to align his interests with the hygiene movement, before he could change it, he had to be a French patriot or nothing. Seen isolated this is all intelligible through Latour's perspective of the social as immanence, but the constitution of the setting cannot itself be studied. The setting seems just to be present, and in this way the force of the limitations of the constituted setting cannot really be understood.

Latour claims to build upon the works of Gilles Deleuze and Félix Guattari26, but these authors have suggested a fundamentally different solution to the relation between social immanence and transcendence with their concepts of the molecular and the molar (1984, 1987, e.g., Chapter 9). By these concepts the interactions of agents within a given setting is seen as being of a different order than the constitution of the setting itself. The molecular concerns the active creation of multiple connections between liberated flows. These connections can be between units that are themselves multiple and all sorts of connections are in principle possible. On the other side, all civilizations or societies imply a certain regulation and steering of these flows. This external regulation of 'mass action' or gregariousness is what Deleuze and Guattari call the molar27. The regulation of the flows usually takes place by the 'political machine', which in archaic societies might be the family or the clan and in modern societies the state. By the establishment of national states there is an extensive standardization of language, culture and society compared to earlier civilizations.

Following this point of view there are historically constituted units that try to fixate, organize and totalize the flows. The example cited above shows how the national state was able to channel the desire for revenge after the 1870-1871 defeat, and this desire was able to capture science into its service. There is a dynamic between the molar and the molecular where all sorts of new inventions and variations are made through the lateral molecular relations, while the hierarchical molar units try to capture the inventiveness and lead it in certain directions. Latour, on his side, builds his whole theory on the molecular connections, while the molar dimension remains a blind spot. By always taking the molecular view he cannot show how molar units provide the basis for what molecular connections are possible and against which they insurrect. Taken linguistically we can say that Latour has a preference for the syntagmatic axis as opposed to the paradigmatic axis. The syngtagmatic axis represents the realization of the language where new connections and events take place. What interests him is how scientists build new objects and form new networks in ever expanding syntagmatic chains. The paradigmatic axis, however, represents the limitations of what connections that can be made in a linguistic situation by its "either .... or" structure. What expressions can be substituted and thus have a virtual existence within a given setting? Latour's analyses along this axis are either made in the most abstract ways as critiques of our basic classematic categories (cf. above) or by the analyses of regimes of enunciation (cf. 2001a, 2002).

'Regime of enunciation' is a concept formed to counter the argument for social transcendence by searching for types of coordination present in certain immanent contexts (Latour, 2001a: 1). He thus identifies political representation with a certain form of utterances where multiplicities of views are re-presented as one common view. Whenever we want to represent the views of others we tend to start speaking politically with all the necessary distortions involved (ibid.). Regime of enunciation is thus an attempt to identify transcendent forms within language. It is a sort of position of enunciation we have to occupy in order to speak about a certain subject matter. This is an interesting view taken by itself, but is also an important clue to the understanding of the Greimasian heritage at work in Latour's theories: His studies always concerns forms of utterances. We have seen above how Greimas and colleagues defined the forms of expression and of content as the object of the field of linguistics, while the substance of expression and content were left to other disciplines. Since the substance (or the matter of the world) cannot have any signification without linguistic form, the study of forms is a handy point of entry to study what different disciplines do. This is exactly what Latour commits himself to and he does not want to go beyond this study of forms. In his case studies he can leave judgement of substance to the agents themselves. They have to be: "... left to their own devices. It's a laissez-faire sociology" (1996a: 170). In this way he does not have to touch the objects himself, but can leave them to the scientists and engineers.

This is the basis for Latour's affirmative view of science. Social scientists and humanists tend to criticize science from external positions wholly alien to how science and technical development is practiced. However, the merciful view that one should not pass judgements on those one studies nevertheless reaches a limit when Latour wants to do politics. Although the practices of science and technology have consequences for the world, he will not pass judgements on their applications. His critique of political institutions concerns their forms and not their substance. If he should have engaged himself with the latter, he would have been forced to take into consideration the already vast assemblages of political enunciation. Instead he limits himself to sketch forms of representation that will include science and technology in democracy (cf. 1999b), but this politics of nature has no direction. It is without substance and hence without any politics at all! (cf. e.g. Caillé, 1991).

Conclusion

The originality of Latour's conception of science studies lies in his transformation of nature, culture, science and society, object and man, into entities on the same plane of immanence. In order to do this he needs the Hjelmslevian model of the sign in its Greimasian interpretation. Several 'disappearing acts' follow this transformation. The external referent disappears and becomes an internal referent. References to the world 'out there' turn into steps of translations between different frames of reference. The same disappearance happens to the transcendent social context, which now and then tended to intervene in order to influence what agents do or think. Only a context that might become explicit through the agents' own interdefinitions is relevant. The major gain with a plane of immanence of this type is the absence of an external point of view from which accusations and denunciations can be formed. A major problem with Latour's conception is, however, the absence of relevant ways of handling the virtual dimension. How do historically constituted contexts - taken for granted by the agents - limit and form the possible choices made by agents.

A point where Latour distinguishes himself from the semiotics of the Paris school, is the ontological reading of texts. While the semioticians limit themselves to the study of structuration of meaning, Latour studies the structuration of entities. He turns the study of meaning into the study of existence. However, in the same manner as the semioticians he excludes any problems concerning the appreciation or evaluation of the substances of entites. This is probably why he cannot say anything about what science does to the world, only how scientists form and maintain scientific objects and how they are forced to utter themselves in certain ways to do so. This study of forms of agency and of enunciation reaches its limit when Latour also wants to do politics without making any choices about political direction.

Footnotes

1)For those easily hurt by words like 'strength' or 'force', Latour claims that trials can also be of weakness. (ibid. p.155).

2)In the interview with Crawford he sais (jokingly) about the fusion of ontology and semiotics: "But then I add to this implausible semiotics a still more implausible ontology, and strangely enough, the whole thing, instead of being a complete mess, ends up being reasonable, even common sense!" (Crawford, 1993:264).

3) matiére in French.

4)Defined as "any entity endowed with meaning, belonging to either spoken strings or written texts prior to any logic or linguistic analysis" (Greimas & Courtés, 1982:362)

5)Also called contrariety

6)My translation. Unfortunately the official English translation does not give the same wealth of lexemes as in the French original.

7)This is often called the narrative schema.

8)This summary is based partly upon Greimas (1983:202-207) and Hénault (1979:146) and gives a early version of the model.

9)Hénault uses Star Wars as an example: The young Luke finds a magnetoscopic message showing a princess in distress. This leads to the search for Obi Wan Kenobi (first aquisition of knowing to do as well as wanting). The latter reveals to Luke his origins and what he should do (first acquisition of having to do), but also informs him about the galactic struggle to come (second acquisition of knowing to do). Later the old man becomes the Mentor of Luke and teaches him in the practices of combat (acquisition of being able to do) (cf. Hénault, 1983:57)

10)Manipulation must not be understood in a pejorative sense.

11)In Star Wars the magnetoscopic message and the revelations of Obi Wan Kenobi forms the cognitive manipulation of the young Luke Skywalker.

12)In Star Wars the hero and his helpers are awarded medals under a big ceremony.

13)This is my interpretation of some of Latour's arguments. I have not found any explicit refrence to Hjelmslev's work in Latour's texts.

14)The translators of Greimas & Courtés (1982) prefer the term disengagement, but I will keep the Latourian usage. Similarly, for embrayage I will use shifting in and not engagement, as preferred by Greimas & Courtés' translator.

15)For an extensive treatment of shifters by Latour, cf. 1988b.

16)Greimas, on his side, presupposes the division between external and internal reference. The external reference is some extra-linguistic reality. However, he is aware of that all sciences (through their discourse) must build an internal referent (cf. Greimas, 1990, ch.2, and Greimas & Courtés, 1982:259-261).

17)These expressions are placed in brackets by Greimas & Courtés to indicate that they are not linguistic concepts.

18)Original title: La fermentation apellée lactique (Pasteur, 1922), published for the first time in 1857.

19)This point is developed furthered by Latour in Science in Action (1987, Chapter 1) where he tracks the positive and negative modalizations of utterances.

20)In the original: "Tout le monde sait aujourd'hui qu'en ajoutant à de l'eau sucrée de la craie, plus une matiére azotée telle que le caséum, le gluten, les membranes animales, la fibrine, l'albumine, etc., le sucre se transforme en acide lactique. Mais l'explication des phénomènes est très obscure. On ignore tour à fait le mode d'action de la matière plastique azotèe".

21)In the original: "Les faits paraissent donc très favorables aux idées de M. Liebig ou à celles de Berzelius".

22)Performance is also connected to what Greimas and colleagues call narrative programs by which the subject follows certain trajectories (and subtrajectories) of performances in order to reach a given object (cf. Hénault, 1983, Chapter 3)

23)In the original: "C'est elle nèanmoins qui joue le principal ròle".

24)In the original: "En un mot, on a sous les yeux une fermentation lactique des mieux caractérisées, avec tous les accidents et toute la complication habituelle de ce phénomène, bien connu des chimistes dans ses manifestations extérieures".

25)Latour changes his terminology now and then to describe this situation.

26)Cf. for instance, his interview with Crawford (1993: 263) where he at least refers to Deleuze.

27)The distinction between the molar and the molecular might be seen as a reinterpretation of Sartre's distinction between group and series (Donzelot, 1972). Groups are marked by mutual relations between humans, while the series lacks this quality. A series is defined by an outer common point of reference, which gives those who are part of the series a common identity (Sartre, 1985 [1960]). Important differences between group/series and molecular/molar are that the latter couplet is not limited to humans and that they are co-present as two opposite principles.

Literature

Alexander, Jeffrey C. (1982), Theoretical logic in sociology, vol. 1: Positivism, presuppositions, and current controversies, London: Routledge & Kegan Paul

Bastide, Françoise (1981); La demonstration. Analyse de la structure actantielle du faire-croire, in Documents de recherche, (Groupe de Recherche sémio-linguistiques, EHESS) vol.III, no.28, pp.5-38

Caillé, Alain (2001), Une politique de la nature sans politique, i Revue du M.A.U.S.S, no.17, pp.94-116

Chateauraynaud, Francis (1991), Forces et faiblesses de la nouvelle anthropologie des sciences, i Critique, nr.529-530,

Crawford, T. Hugh (1993); An interview with Bruno Latour in Configurations, vol. 1 no.2, pp.247-269

Deleuze, Gilles & Félix Guattari (1984); Anti-Oedipus: Capitalism and Schizophrenia, London: Athlone Press

Deleuze, Gilles & Félix Guattari (1987); A Thousand Plateaus; Capitalism and Schizophrenia, vol.2, London: Athlone Press

Donzelot, Jacques (1972); Une anti-sociologie, in Esprit, nr.12, (December), pp.835-855

Greimas, A.J. (1976), Maupassant. La semiotique du texte: exercices pratiques, Paris: Éditions du Seuil

Greimas, A.J. (1983), Structural Semantics: An Attempt at a Method, Lincoln & London: University of Nebraska Press

Greimas, A.J. & J. Courtés (1982), Semiotics and Language: An Analytical Dictionary, Bloomington: Indiana University Press

Greimas, A.J. (1990), Narrative Semiotics and Cognitive Discourses, London: Pinter Publishers

Haraway, Donna J. (1997), Modest_Witness@Second_Millenium.FemaleManÓ_Meets_OncomouseÔ, Feminism and Technoscience, New York and London: Routledge

Hénault, Anne (1979), Les enjeux de la sémiotique, Paris: Presses Universitaires de France

Hénault, Anne (1983), Narratologie, sémiotique générale, Les enjeux de la sémiotique, tome II, Paris: Presses Universitaires de France

Hjelmslev, Louis (1993 [1943]), Omkring Sprogteoriens Grundlæggelse, (Prolegomena to a Theory of Language) Copenhagen: The Linguistic Circle of Copenhagen

Latour, Bruno (1983), Give Me a Laboratory and I will Raise the World, in M. Mulkay K. Knorr-Cetina (eds.), Science observed, Perspectives on the Study of Science, London: Sage, pp.141-170

Latour, Bruno (1987), Science in Action: How to Follow Scientists and Engineers through Society, Cambridge, Mass.: Harvard University Press

Latour, Bruno (1988a), The Pasteurization of France, Cambridge, Mass. and London: Harvard University Press

Latour, Bruno (1988b), A Relativistic Account of Einsein's Relativity in Social Science Information, vol.18, pp. 3-44

Latour, Bruno (1993a); Pasteur on Lactic Acid Yeast: A Partial Semiotic Analysis, in Configurations, John Hopkins University Press, vol.1, no.1, pp. 129-145

Latour, Bruno (1993b); We have never been modern, New York: Harvester Wheatsheaf, 1993.

Latour, Bruno (1996a), ARAMIS or the Love of Technology, Cambridge, Mass. & London: Harvard University Press

Latour, Bruno (1996b); Petite réflexion sur le culte moderne des dieux faitiches, Paris: Les Empêcheurs de penser en rond

Latour, Bruno (1997), Socrates' and Callicles' Settlement - or The Invention of the Impossible Body Politic, in Configurations, John Hopkins University Press, vol.5, no.2, pp.189-240

Latour, Bruno (1999a); Pandora's Hope: Essays on the Reality of Science Studies, Cambridge, Mass. & London: Harvard University Press,

Latour, Bruno (1999b); Politiques de la nature: Comment faire entrer les sciences en démocratie, Paris: La Découverte

Latour, Bruno (2001a); Si l'on parlait un peu politique? mimeo, accepted by Politix, (http://www.ensmp.fr/~latour/Articles/83-POLITIQUE.html) (27 September 2002)

Latour, Bruno (2002); Jubiler - ou les tourments de la parole religieuse, Paris: Les Empêcheurs de penser en rond, Seuil

Latour, Bruno and Steve Woolgar (1986), Laboratory Life: The Construction of Scientific Facts, Princeton, NJ: Princeton University Press

Pasteur, Louis (1922), Oevres de Pasteur, tome 2: Fermentations et générations dites spontanées, Pasteur Vallery-Radot (ed.), Paris: Masson et Cie, Editeurs

Propp, Vladimir (1968), The Morphology of the Folktale, Austin, Texas: American Folklore Society biographical and special series, no.9

Sartre, Jean-Paul (1985); Critique of Dialectical Reason, vol.1; London & New York: Verso