Theorizing is not abstraction but horizontal translation

ABSTRACT

The literature on theorizing usually implicitly assumes that theorizing is writing. A recent focus on theoretical diagrams seeks to correct this idea. But even this focus on diagrams measures them against writing. Instead, taking up ideas developed by the sociology of translation, I propose to understand theorizing as a specific kind of translation, namely horizontal translation. Horizontal translations create new worlds through conceptual invention and re-definition and by creating new geometries of how concepts relate to each other. Because such relationships are geometric forms, rather than sequential texts, diagrams are core media of theoretical work, because they allow the visualization of such conceptual forms.

KEYWORDS:

• Theorizing
• actor-network-theory
• translation
• diagrams
• media
• writing
• visualization

Introduction

How can we theorize theorizing? How are theories produced? Are theories whatever theorists write? Is theorizing whatever is left after empirical work has been completed? If not, how can we develop a positive theory of theorizing?

There has been are new interest in theorizing within sociology. Within the US context, this interest has resulted from trying to justify theorizing that is seen on the retreat, partly because of its lack of internal logic and teachability (Lizardo, 2014). What is seen as an art, cannot be a science. Within the European context, where theorizing is less threatened as a practice, the interest in theorizing is more spurred by an attempt at overcoming the logic of theoretical schools by paying more attention to its own internal logics that transcend such historical divisions (see the contribution by Schlechtriemen and Werron et al. in this special issue).

Both traditions have focused on understanding how theorizing works as a practice, usually without questioning what theorizing is. If anything, authors have tried to overcome protracted definitional issues by thinking about different kinds of theorizing, thereby multiplying the practice (Abend, 2008). One focus of this multiplicatory tendency has been diagrams as an overlooked form (Silver, 2020; Swedberg, 2016).

In this article, I would like to follow the focus on diagrams, not as an overlooked form, but as a probe into understanding what theorizing is and how it works. I use the analysis of the practice of diagrams for an attempt at clarification of theorizing itself. My aim is to develop a theory of theorizing that achieves two goals: First, I develop a concept of theorizing that allows us to understand it as a specific and distinct practice. By doing so, my aim is to follow other contributions that have learned from and built on approaches from practice theory and STS that assume that we need to analyze the practice of theorizing. My aim is to characterize theorizing not as a left-over of empirical work, or in a purely contextual or tautological way (theorizing is what sociologists think is theorizing/theorizing is what theorists do), but as a substantive practice. Secondly, my aim is to develop a concept of theorizing that understands theorizing as operation and therefore does not align with ascriptions to people, whether self or other (she is/I am a theorist) or works (‘Identity and Control’ is a work of sociological theory). Instead, I will show that theorizing can occur everywhere and always, in ways that do not align with canonical ideas of theorizing.

Diagrams are useful for such a probe because their status as minor medium allows us to understand the practice of sociology through its media practices and related moral economies (Daston, 1995). Through understanding the scepticism towards diagrams, and attendant attempts to rehabilitate them, we can see a path towards an idea of theorizing that does not derive from the logics of specific media (writing), but from qualities of theorizing itself.

If we analyze the ideas of theory diagrams, what we come to see is that the prevailing view understands theorizing as writing (section 1), and diagrams as mostly useless duplication that serves specific pedagogic or artistic purposes. But we know from accounts of theorizing and media use, that the relationship between the two is far more complex (section 2). To understand this relationship better, I suggest to abandon the usual idea that theorizing is abstraction and turn to the sociology of translation for help (section 3). The first step is to understand the difference between a structural idea of theory and its specific mediations as text or diagram (section 4). Finally, we can understand theorizing as horizontal translation and see its specific properties, namely conceptual coherence, constructive openness, and potential for transposition (section 5). By way of conclusion, I outline the repercussions of this for critiques of theories and standpoints.

The default assumption: diagrams as the poor relative of writing

I begin by analyzing some shared elements that amount to a default position of sociology. Because most texts about theorizing implicitly assume that theorizing is always writing, I analyze the relatively rare contributions that focus on diagrams as the excluded other. As I will show, even these accounts measure diagrams against texts as default.

The first shared assumption is that texts are the standard medium for social theory, which does not need any justification. This assumption is most visible in the fact that until recently, there were few texts that focused on theory diagrams at all. But even texts that focus on diagrams compare diagrams against the standard of text, such as in the opening statement of Lynch’s classic article, where he says that diagrams ‘rarely show much beyond what a text already says in its writing’ (Lynch, 1991, p. 1), a view repeated for example by Swedberg in his more positive view of diagrams (Swedberg, 2016, p. 258).

The second assumption follows from the assumed primacy of the text, as diagrams are seen in a temporal relationship to it. Both Lynch and Swedberg write about what the ‘text already says’, and Turner relates to diagrams as emerging ‘from the text which gave rise to it’ (Turner, 2010, p. 115):¹ They assume that the process of theorizing moves from first writing to then producing a diagram (I will return to this crucial relationship in the next section in more detail). However, they have no proof that the relevant theories were produced in this order. If it is the reader (i.e. Lynch), who produces the temporal order, it tells us more about the reading logics of specific readers, than about the relationships between texts and diagrams themselves. As an ethnomethodologist, Lynch may well support a theory of the primacy of reading over the primacy of production. But he would then need to refrain from attributing the temporal logic to the media themselves. Also notice that the primacy of texts in the reading practices of sociologists – at least as professed in their writings – is so ingrained, that examples of readers claiming a primacy of diagrams as tool to understand a theory are exceedingly rare (for one such exemption see Baldamus, 1992).²

The third assumption, which follows from this assumed temporal relationship, is that diagrams run the danger of duplication, of being supplementary: ‘Is the diagram an adjunct to the words on the page … ?’ (Turner, 2010, p. 113).

Fourth, because of this danger of duplication, diagrams need justification. They do not just exist as an equally valid mode of theorizing among others, but need to be questioned, and need specific legitimacy to simply be there, a burden that texts do not carry.

Fifth, Lynch makes an unstated reason for this need for justification explicit: theory diagrams suffer from their comparison with other types of diagrams, particularly in the natural sciences, such as Galileo’s diagram of a rolling ball on a plane. These diagrams allow, following Latour’s theory of inscriptions (Latour, 1990), to measure distances on paper, and extrapolate from these paper measures back to the world.³ The (theoretical) diagram is an extension of the empirical onto paper. In social theory diagrams, Lynch observes, such measurements are not possible, because the objects of the diagrams do not have a defined relationship to empirical objects.

For Lynch, therefore, as per his title, theory diagrams are ‘pictures of nothing’. Note that Lynch fails to ask the attendant question of what theory texts are texts of. According to his logic, we should probably call them ‘texts of nothing’. But they are not, since Lynch understands that whatever theory texts do, they do something sui generis that does not need further justification. Other authors, faced with this nothingness try to save this lack of reference of theorizing as a specific, ‘disjunctive’ mode of theorizing ‘independent of data’ (Silver, 2020, p. 873).

Sixth, because diagrams are always seen as deficient, they are seen as not contributing to sociological theorizing, but as a minor practice: Because of the potential emptiness of the diagram, there must be another reason for its production – for why else would it exist? – which cannot be truly sociological: is a specific diagram ‘a mere pedagogic device, or does it have an independent cognitive potential?’ (Turner, 2010, p. 113). As a pedagogic device, the unstated assumption is that it serves the purpose of helping a reader, confused by unfamiliar words, to make sense of theory (see also Toth, 1980). The diagram is a medium for the uninitiated, A-level students (Lawson, Jones, & Moores, 2000), ‘the undergraduate seeking a shortcut through the diagrams’ (Turner, 2010, p. 114), but not professional sociologists. Again, we could ask the reverse question: If a diagram allows to understand a theory better and faster, then why should it not be considered the actual theory, rather than the more cumbersome, complex, and confusing text? (I will return to this in the next section in more detail).

Taken together, these elements amount to a standard account of theorizing, which is based what I have called an ‘unequal media determinism’ (Guggenheim, 2015). Unequal media determinism describes the idea that sociologists take one medium, writing, as a medium for granted and being able to do all kinds of things (produce novels, poems, theory, empirical descriptions), whereas another medium, diagrams, is identified with specific effects produced by its inherent problematic features. Unequal media determinism results in writing being free of any need for justification. In contrast, diagrams are in need of justification and are constantly seen as endangering the true practice of sociology, unless they are improved, specified, and made relevant. What this view of diagrams fails to explain is why diagrams are ubiquitous, successful, used by many authors, and in many cases, the better-known elements of specific theories than their attendant texts. It neither helps to understand what theorizing is because it leaves the relationship of theories and media undertheorized. Worse, it conceives of theorizing implicitly in the logic of writing as the preferred medium.

Theorizing and Theories: Media differences between process and result

A first step to undo this unequal media-determinism is to look closer at the temporal relation between writing and diagrams. As I have shown above, the temporal relation is a mere assumption by the analyst, rather than an empirical observation. Let us now look at a rare case where we have empirical insight into the temporal relation of different media namely Alfred Gells’ article on ‘Strathernograms or the semiotics of mixed metaphors’ (Gell, 1999). In the article, he tries to make sense of Marilyn Strathern’s book ‘the Gender of the Gift’ (GG). He begins by noting how intractable the book is and how hard it is to understand Strathern’s theory of Melanesian society. To make sense of the book he resorts to turning the book into a series of diagrams and he invents his own diagrammatical language:

The virtual absence of a diagrammatic channel of communication in GG is particularly puzzling in that, as one reads this text, images of forms, relations and transformations continually materialize before the mind's eye. (Gell, 1999, p. 31)

We are going backwards here and start with a specific reader, and for this reader, the text becomes a diagram that ‘materialize[s] before the minds eye’. Note how Gell uses the term ‘materialize’ to describe a cognitive process. Diagrams, at least for this reader, are not useless duplications, but necessary for understanding. Gell continues:‘ … one has the strongest impression that while Strathern was writing, she likewise spent a lot of the time seeing forms, … . in her mind’s eye’ (Gell, 1999, p. 31). For Gell, as a hypothesis, the process of theorizing itself must have included such visual forms, otherwise such a theory would not exist. His hypothesis, it turns out, is correct: ‘Indeed, [Strathern] has since told me that she does use a series of diagrams as schemata in the process of writing, but that she eliminates them subsequently since the text has to stand by itself (pers. comm.)’ (Gell, 1999, p. 31). This theorizer admittedly needs diagrams to theorize, but then decides that the media needed for composing for themselves, are not needed for understanding by others. This is surprising, given that at least one specific reader, Gell, for his own understanding, needs to recompose those diagrams, but without the aid of the author’s original ones. Asked about this process of deletion, Strathern tells Gell ‘that diagrams can give a spurious logic to texts which are, in fact, discursively incoherent’(Gell, 1999, p. 31).

We have arrived at a surprising conclusion, given that the standard view reconstructed above amounts to critique diagrams for being implicitly ornamental and superfluous. For Strathern, texts are inherently incoherent and the problem with diagrams is that they create a ‘spurious logic’. Note that her critique is not that the empirical world is incoherent, and theories (in whichever medium) create a spurious logic – a well-rehearsed critique of theorizing itself–, but that theorizing as text is incoherent, which she understands as a good thing, otherwise she would not destroy the diagrams.

As SibylleKrämer has shown, this fear of false coherence through diagrams has a long history in philosophy, where a number of famous philosophers (Plato, Descartes, Wittgenstein) are known to have been philosophizing through diagramming, but where followers and publishers have subsequently omitted these diagrams from publications(Krämer, 2014, pp. 12–14). Also, in Sociology we have the case of C. Wright Mills, who developed his practice against Talcott Parsons, whose theory was built on cross-tabulation diagrams. As Stefan Bargheer reconstructs in his study of Mills’own cross-tabulations, ‘Mills told the reader that none of his famous works had been produced without some cross-classification taking place in the process of writing the first draft, although he did not usually display such diagrams in the published version’ (Bargheer, 2021). But, unlike Strathern, ‘it was precisely this practice of filling out empty cells in a 2 × 2 table that Mills provided as an example for the sociological imagination’ (Bargheer, 2021, p. 266).

Just like Strathern, Mills routinely produced diagrams and then destroyed them for publication, in his case ostensibly for distancing his own practice from those of his opponent Parsons. What matters however is that in all cases writing occurs not before, but along or even after the production of diagrams, and the deletion of diagrams occurs because of prejudices about the media, despite the obvious usefulness in the production process to the authors.

Research as translation: why theorizing is not abstraction

If we understand theorizing as a practice that can occur in different media, and if various authors use different media at different points in the theorizing process, and if the deletion of diagrams is due to some shared assumption of how theories should presented, how better to conceive of theorizing as a practice? I suggest to understand diagrams as being repressed for historical and sociological reasons and to use this urge to diagram during theorizing as an important indication of its actual logic. That is, diagrams allow us to understand theorizing not as opposed to empirical work and abstraction.

Authors who return to this distinction usually do so, because they cannot give a positive definition of theorizing. They resort to a negative one, where theorizing is everything that is left over after empirical work has been done. Theorizing as opposed to empirical work turns the former, as Lizardo has pointed out, into ‘a mystery’ because in the analytical tradition, theoretical discovery is an ‘embarrassment’ that can only be dealt with in the ‘context of justification’, while in the continental tradition, while celebrated, it must be ‘protected from explicit dissection’ (Lizardo, 2014, p. 19).

Usually, the embarrassment stems from the idea that theorizing is ‘abstraction’ and ‘generalization’, constantly in danger of ‘loosing contact with the underlying empirical phenomenon’, which is deemed ‘concrete’ (Swedberg, 2017, pp. 192–94). Sociology has inherited this idea from the philosophy of science, where theory, whether as verification or falsification, is defined with reference to data. The production of theory itself, inasmuch as it is not merely an extension of data, is ultimately metaphysical, beyond of what can be considered scientific work (for an overview see Abend, 2008, p. 190; Carleheden, 2019; Hallberg, 2013). Even weaker versions, trying to resolve the issues by pointing to the overlap between theory and empirical work and to the theory-ladenness of empirical research, are ultimately based on this underlying distinction. Swedberg for example, resolves the difference between theory diagrams and empirical diagrams by resorting to the idea of ‘colligation’, whereby when ‘facts’ are rearranged ‘in some way’ indicates ‘methods’, whereas if they are connected by an ‘explicit idea’ it means ‘you have introduced some theory’ through colligation (Swedberg, 2016, p. 254). But the abstract-concrete dualism produces misleading ideas about both practices.⁴

To understand why this distinction is problematic, it helps to begin with the underlying idea of empirical work. To speak of concrete, underlying phenomena assumes that these exist in the world independently of the work of researchers.

STS has taught us however that empirical phenomena are the outcome of scientific practice through what Callon, Latour, and Akrich would call translations. The sociology of translation, or Actor-Network Theory, was developed originally through the analysis of the work of natural scientists (Callon, 1986). Translation is not a linguistic process, but a process of transforming objects into something else. In empirical work, objects are translated into representations. Such translation always entails a process of selection and abstraction, in the sense that moving from an object to its representation makes the representation unlike the represented object. A falling body is turned into a measurement. Soil becomes a colour code. In the natural sciences, as Latour has stressed, many of these translations turn objects into visual representations (Latour, 1990). Among the reasons for this preference for visual inscriptions are that they can be scaled up or down, they can be measured, they can be overlayed, they can be written on, and many more.

The same concept of translation can be used to describe empirical work in the social sciences. For example, in telephone surveys, spoken answers are translated into written words or crosses in tick boxes, which are then in turn translated into graphs or pie charts. In conversation analysis interactions are audio- or video-recorded, then translated into transcripts via transcription rules, then coded and annotated. In social research, translations often occur outside of the control of researchers, for example when doctors translate physical bodies into written causes of death or teachers translate pupil performance into grades.

In empirical research, translation as inscription moves away from the original object, transforms it and makes it ever more unlike the original object. The more translation steps, the more abstract a representation becomes by losing properties of the original object (such as the sound of the voice, nuance of the sentence, materiality of the body), undergoing a shift of medium (i.e. sound to transcription), becoming enriched with other data and media, such as annotations and codes, all brought to the object by the researcher and their own logics. This process of moving away from an object to a scientific representation is a process of abstraction. It goes without saying, that in each translation step, theoretical considerations play a part in the choice of translations. Deciding whether and how to transcribe and code representations is impossible without theoretical input, if by theoretical input, we mean anything that is not a feature of the empirical object itself (this is what in the philosophy of science is usually called the theory-ladenness of observation). But it is not theorizing.

Theorizing is a mediation of a structural theory

If all empirical research is a process of abstraction, then how else can we characterize theorizing as a practice, and how can we make sense of the role of diagrams in this process? How can we differentiate between the unresolved ‘thorny issue’ between the ‘visualization of data and the visualization of theory’ (Swedberg, 2016, p. 253). One way out, as Bargheer suggests, is to understand theorizing as a purely contextual – and fundamentally arbitrary – designation given to some works, that is unrelated to their mode of operation (Bargheer, 2017). This would also have the advantage of allowing to move away from theorizing as a heroic, male, standalone practice and understand it as woven into all research, as Stöckelová posits and as Werron in this issue elaborates (Stöckelová, 2013)? While I agree with the goal, I also aim to preserve it as an identifiable, specific, and distinct practice. To do so then urges us to understand how it can be a distinct practice within sociological practices, rather than as a designation of works or people.

I suggest focusing on the very process of mediatization invoked in Strathern’s example above. Remember that a justification for Gell’s diagrams was that Strathern was ‘seeing forms, … . in her mind’s eye’. Theorizing for Gell, includes a moment before theory becomes mediatized, whether as diagram or text. Also note the phrase ‘visualization of theory’ by Swedberg in the previous paragraph, which suggests that the visualization is a mediatization of theory (rather than a transformation of a different mediation of theory), it must be something else. I would suggest to call this a structural idea of theorizing, but not in the sense mentioned above, where all theorizing is considered metaphysical tout court. Rather, a structural idea of theory assumes that a non-mediated version of a theory pre-exists its mediations, namely theory as an internal conversation, a shared horizon, or an unmediated idea, or indeed, in Deleuze and Guattari’s vocabulary, an ‘abstract machine’ or a ‘diagram’:⁵

Defined diagrammatically in this way, an abstract machine is neither an infrastructure that is determining in the last instance nor a transcendental Idea that is determining in the supreme instance. Rather, it plays a piloting role. The diagrammatic or abstract machine does not function to represent, even something real, but rather constructs a real that is yet to come, a new type of reality. (Deleuze & Guattari, 1987, p. 142)

Luhmann for example invokes such a structural idea of theorizing when he distinguishes between a theory and its representation: ‘The representation of a theory does not convey an adequate idea of its structural design plan’ (Luhmann, 2017, p. 264, translation by author). This non-mediated design plan contains ‘functional elements’ of a theory that serve its ‘internal flexibility’, ‘darkrooms’ where theorizers experience their ‘internal successes’ and where they can ‘feel safe in their own construction’ (Luhmann, 2017, p. 264). Turner expresses a similar idea when he addresses ‘inventories of purposes’ of diagrams and lists as for example ‘making structures clearer’. Randall Collins discusses philosophizing, which I count as a cousin of theorizing, as an ‘internalized conversation’ (Collins, 2009, p. 46 ff.).

Such structural notions of theory are rendered visible in the following ubiquitous instances:

At its most basic, a structural idea of theory is invoked whenever a ‘theory’ is referenced by a name, title, or an identification with an author such as in ‘field theory’, or ‘Weber’s theory of bureaucracy’, rather than by specific references and quotes. This routinely occurs in introductory texts to specific theories or theorists, in overviews, and in theory chapter sections of empirical articles. For these to be possible and recognized as versions of theory x is only possible if producers and readers share an understanding that a theory is not the same as a given mediatization. Structural ideas of theories are also invoked when authors allude to qualities of theories that cannot be found in specific mediations, say when someone accuses field theory of not being able to account for social change. Finally, structural versions of theories are invoked when a specific contribution is discussed as belonging or not belonging to a given theory (‘This book pretends to be a contribution to Marxist theory, but is not’).

For all these cases the following holds: First, a theory cannot be reduced to a defined set of mediations. Second, it could contain other elements and it could be rendered differently (sentences could be written differently, diagrams could be replaced by texts and vice versa) and would still be the same theory. Third, it could be shortened, lengthened, turned into more complex or simpler mediations, it could be written in a different style and for different audiences (‘Judith Butler for lawyers’, see my point above about Turner’s idea that diagrams may be mere pedagogic devices) and would still be the same theory. Fifth, it could contain additional elements; it could cover different fields and topics (I will return to this below as a transposition) and would still be the same theory.

Whatever else it is, someone who theorizes is engaged in (re-)producing structural theory and finding specific mediations for it. The structural form does not exist as text or diagram, but as something else (a thought, an internal dialogue, an idea, a structure, a blueprint, a framework), to which we have no empirical access, at least not if we only analyze final mediations. The only empirical access we have are self-reports such as the one by Luhmann cited above.

A counterproposal: Theorizing as horizontal translation

But what happens during the process of turning structural into mediated form? We can take as a starting point the idea of Niklas Luhmann that theorizing is a form of ‘selection’ and that the practice of the theorist is to have ‘better selection strategies to understand higher complexity’ (Luhmann, 2017, p. 252). With Esposito, who follows Luhmann, we could say that such selection always includes ‘comparison’, with other concepts, ideas, and cases (Esposito, 2013, p. 141). Both selection and comparison have the advantage of understanding theorizing as writing in that they conceive of theorizing as a specific kind of operation of elements that is independent of media. These individual elements do not invoke a specific medium or a specific form of relation, and they do not call, as Czarniawska suggests, for ‘plots’, ‘narratives’, or any other form of textual and therefore linear relationship (Czarniawska, 2013, p. 107 ff.).

But selection and comparison can only be seen as the first step, otherwise it is unclear why for example Gell is compelled to create diagrams to understand Strathern. While we could say that the reduced number of concepts Gell uses in his diagrams aim to clarify the selection steps that Strathern undertook and that are – intentionally, as Strathern makes clear – obscured in her text by too many other words, it does not explain what the diagrams do. Gell could have left us with a list of concepts. But neither Gell nor Mills produce a list of concepts, but a geometrical form relating concepts to each other. The concepts are not additive, linear or sequential, but relational: they gain meaning and specificity not by measuring them against data, but by relating them spatially against other concepts.

Theorizing is not Fiction

If we return to translations, we could look to Latour’s later attempts to produce a comparative ontology of what he calls ‘modes of existence’. Translation as a theorization of empirical research is but one such mode called ‘Reference’ or ‘REF’ by him. Remember that translation as ‘REF’ moves from an object to a representation. Of the other modes he offers, ‘Fiction’ or FIC may look like a candidate to describe theorizing(Latour, 2013a, chapter 9). Fiction is different because unlike REF that ‘come back home’ from the world to the inscription, FIC reverses the direction of travel and ‘carry us away’ from the inscription to the world(Latour, 2013a, p. 251). Whereas for REF the task is to identify elements of an object (say the words in a sound recording) and translate them into a representation (a string of words on paper), in FIC representations (such as words on paper) are used to create worlds that spring from the mind of a creator. These representations in FIC may even be indistinguishable on the surface from the representations in REF: For example, an author of a historical novel may invent sources, such as private letters, but these were invented by the author. They are FIC rather than REF not due to a feature of their mediated form (they look indistinguishable from actual letters), but due to their production logic: one springs from the mind of the author, the other of a historical person. The production logic of FIC is invariably driven by the producer and targeted to the audience (even in case of incoherence, say in aleatoric or absurd works): the author will invent the letters in ways that drive the story. The production logic of REF however is driven by the original materials and the logic of specific methods.

But whatever theorizing is, it is neither FIC nor REF: While it does not return from the world as in the case of REF, it neither carries us from the theorizer into the world as in FIC. Social theory is not the imagination of alternate (or even aleatoric or absurd) realities (even though good social theories allow us to conceptualize alternate realities).

Theorizing is not abstraction

Theorizing, I suggest is a different type of translation, a type of translation that I call horizontal translation. As a specific type of translation, it has the following elements:

First, horizontality is defined through its production logic of elements on the same level of abstraction or concreteness. If empirical work is a process of abstraction, away from objects towards representations,⁶ theorizing is a process of moving laterally, from one theoretical concept to another. It neither adds nor reduces abstraction. Theorizing can happen as close or as distanced (measured in REF-translation steps) from specific sociological objects and these objects can have any size or complexity. Theorizing is agnostic with regard to how many translation steps it involves or the density of translation steps of the representations with which it operates. Both ‘the raw and the cooked’ (without further translation steps) as well as a specific practice, translated through interviews, photographs, video-recordings or fieldnotes of eating raw fish can become the operational step of theorizing. While sociologists debate which translations should count as good or bad theorizing (i.e. debates about social theory versus sociological theory, or sterile grand theory versus empirically rich theory, or macro-, meso- or micro-theory) from the viewpoint of horizontal translations, these debates do not make sense. Figure 1.

Figure 1. Theorizing as horizontal translations.

Theorizing is also not the invocation of pre-existing theoretical concepts. To avoid theorizing from being a mere re-reading of past authors, or the re-deployment of already existing theoretical concepts, it needs to either re-define and re-conceptualize old concepts or produce new ones. As a process and practice, it needs to do something with and to an existing theoretical structure. Theorizing is creating, developing, and reorganizing a structure that allows the re-definition or production of new concepts.

Theorizing depends on conceptual coherence and constructive openness

The structure never consists of a defined set of concepts, but of generative rules of how to make sense of the world.⁷Horizontal translations are further defined by their conceptual coherence and constructive openness. This conceptual coherence operates on the level of its structural backbone. Concepts relate to other concepts (and not to the words and sentences in texts next to them). By doing so, they allow the creation of new concepts or new forms of relationships: ‘Capital’ allows for different kinds of capital, because the notion of ‘capital’ is conceptually coherent since the idea that capital can be accumulated and used to compare human beings can be transposed from money to social relations. Different kinds of capital open new avenues for construction, by inviting the question whether the relationships between different kinds of capital are hierarchical or not, or whether different kinds of capital relate to specific conditions or are particular to specific fields, and so on. The constructive openness is not a property of the notion of capital as such, but of a more general geometry of theorizing as horizontal practice. A critique of binary concepts can be carried from one theory to another, and binary concepts within theories can be replaced by triangular or even more complex logics.

In FIC there is no structure: there is only a given mediation, which is constructively closed. A written work is defined by its narrative or formal logic: different literary or artistic works are defined by their specific mediation and are appreciated for how they render the same underlying logic. There are innumerable versions of Leda and the Swan, both in written and painted format, and they each count as individual works. Producing a new version of Leda and the Swan is legitimate, welcome, and seen as a new work, rather than a form of plagiarism because what matters is the success of the mediatization of the work, rather than the structure of the narrative. Conversely, art works that are defined by their structure (say improvisations over the chords of ‘Ornithology’) are neither defined by conceptual coherence nor do they invite constructional openness: Changing and adding to the chords of ‘Ornithology’ is pointless, as it turns ‘Ornithology’ into a different composition.

In theorizing, by contrast, the specific mediation is secondary and is always measured against its metaphysical or structural version (hence the discussions around the supplementary character of diagrams, or whether an article could have been written more succinctly). What counts as theorizing is a modification of the structure. A new mediated form of the same structure (such as Gell’s article) counts as interpretation or application. Criticism in theory aims at dismantling the narrative logics and metaphors with the aim of demonstrating that the structural logics is either not new or wrong.

Most importantly, as a horizontal translation, theorizing turns a structural element of theory into a mediated form. It neither moves from world to form, nor from form to world, but from structure to form.

Like translations for REF, horizontal translations can come in different media, diagrams, and texts. As we have seen, there is no logical reason why one medium should be temporally or logically prioritized. But such a hierarchy has developed historically: The moral economy (Daston, 1995) of the social sciences has come to prioritize, texts over diagrams.

Why Theorizing mostly occurs in texts and diagrams, but not in other media

To understand horizontal translations, instead of asking about the value of diagrams, it helps to ask why theorizing occurs in diagrams and text, and not in other media, specifically in mechanical recording technologies such as video, photography, or audio-recordings (Guggenheim, 2015). REF translations can begin with any recording technology: writing (of field notes or interview answers), audio recording (of field sounds or voices in interviews), photographs (of human subjects, movements, buildings), drawings of bodily movements, or video-recordings (of social situations). REF translation steps can move from any of the above to other media, but a typical movement is towards drawing, numbers, and texts (audio-recordings to textual transcriptions, texts or photographs to statistics, textual annotations of drawings, drawn highlights on photographs, etc). When empirical translations turn objects into texts or diagrams, they move away from the details of mechanical recording to the abstraction and reduction of writing and drawing. Because theories are not translations of empirical objects, it is hard to theorize with mechanical recording technologies.

Unlike mechanical recording technologies, which are media of representation (a photo of a human being), drawing and writing lend themselves as media of projection. They can produce an account of something that does not physically exist. If we follow the logic of theorizing as translation from structure to mediation, we can also understand why diagrams are such useful media: Translation is a slow movement from one stage in research to the next, hence: the more translations and manipulations the better (Latour, 2013b). Diagrams as a translation step are closer to the metaphysical structure than texts. We can see this nicely in Gell’s pig diagram, where a drawing of a pig and a sow are used to indicate how two capacities inscribed with gender render a vehicle or appearance ‘pig’ (in the upper rhomboid). Figure 2.

Figure 2. Alfred Gell’s Pig (from Gell, 1999, p. 42).

When Strathern complains about the ‘spurious logic’ of diagrams, she perfectly summarizes this point. Empirical work always reduces the complexity of the empirical object to become more abstract, filtered by the selective interest of the research question. It sheds unnecessary detail, by removing whole channels of the sensory world (sounds are turned into text) and by ignoring many details that are irrelevant from the viewpoint of the research question. By contrast, theorizing increases the complexity of a metaphysical structure, but the increase in complexity of the structure is not the same as an increase in words. The complexity of a structure needs to be inferred through a process of identifying relevant elements and their relations to each other. In the form of diagrams, this increase in complexity is much more direct and apparent than in writing. In the Gell diagram above, the increase in complexity results from the diagrammatical logic (for example, the circle indicates exchange), rather than the precision of the pig drawing.

Diagrams usually contain fewer elements than texts. This is because they reduce theories to their core concepts and because they can spatialize relations. Compare any theory diagram to its attendant text, and you will immediately understand this point. For this reason, diagrams are not only faster to produce during the production of theorizing than texts, but they are also faster to read.

To be sure, the reason for this difference is not primarily founded in the media itself, but in the historically evolved media practices related to these forms. The frontispiece to Hobbes’ Leviathan is an incredibly slow and detailed diagram, both in terms of production and consumption. It has an incredibly complex relationship with the text that follows (Bredekamp, 2003). The frontispiece cannot be reduced to a small number of basic visual elements but unfolds on multiple levels. These include, among others, the overall structure and partition of the image, complex visualization of various kinds of powers, interrelation to textual elements, minute details of drawings, references to a complex history of visualizing state and church power, and so on. But diagrams such as the frontispiece of Leviathan no longer exist, while triangles, networks, and so on continue to proliferate. Historically, theory diagrams have undergone a radical process of reduction and simplification (no doubt inspired by natural science diagrams), during which any step towards a more detailed diagram has become impossible. While it is perfectly possible to imagine complex fictional quasi-empirical scenarios as a basis for theorizing in writing (Farzin & Laux, 2016), these have become all but impossible in drawing. Diagrams have become paragons of Occam’s razor, while written theories are rarely saved from word diarrhoea. Any contemporary visual correlate of Leviathan’s frontispice is classified as art rather than as sociology.⁸ It would be a whole research project to figure out how and why this process occurred. For the present article, what matters is that the outcome of this process places diagrams closer to structural versions of theories and makes their production faster.

Because of this historically evolved visual poverty of diagrams, every single element attracts outsize attention: Whether an element is placed here or there, whether a word is placed in a box or not becomes crucial, because every single one of these elements counts (for an example, see Turner, 2010, pp. 120–35). Diagrammatical critique in existing sociology is always critique of fundamentals, never a critique of ornament and style. If there is one place where the Anglophone horror of flourish in sociological writing has universally succeeded, it is in the shape of diagrams.

Diagrams facilitate transposition

A given form of a diagram can then be, to use a term from Levi-Strauss, be transposed to any other situation and re-used, to create new social forms:

It would seem that the methodological principle which inspires such distinctions [namely the vowel/consonant triangle known from linguistics] is transposable to other domains [italics by author], notably that of cooking which, … , is with language a truly universal form of human activity. (Levi-Strauss, 2008, p. 40)

Levi-Strauss and too many structuralists got consumed with ‘universal forms’, and the idea that then every domain of society should be represented as a triangle and that ‘these ideal positions must be occupied’ in any society (Levi-Strauss, 2008, p. 41). However, the idea of transposability does not necessitate universality. Just because we can transpose, it does not follow that the whole world should be represented through one singular diagram. Levi-Strauss already highlights that the piloting role of the triangle creates ‘empty forms’, that ‘teach us nothing … about any specific society’ (Levi-Strauss, 2008, p. 41). These empty forms are not proof of the vacuity or useless abstraction, but precondition of theorizing. Only because we can create empty forms and then use these to transpose from one object to another are we able to create theory at all. Theorizing in this sense is not about universalizing existing forms, but about using existing forms to research other aspects, fields, and topics. As piloting devices these empty forms do not call for being filled, but for being made productive in understanding seemingly unrelated aspects. Theorizing therefore is neither universal nor general, but transpository.

Conclusion: Theorizing, standpoints, and critique

In this article, I have begun by analyzing sociologists’ ideas of the media of theorizing. I have identified a common discourse that identifies theorizing with writing and then, even when trying to give a positive account of diagrams, measures diagrams against this standard. In this default view, a media determinism is at play, which assumes that texts come first, diagrams are illustrations of these texts, and that diagrams have certain inherent properties which renders them problematic. I have shown that in the process of theorizing at least sometimes diagrams come first. I have then suggested that to better understand theorizing, it helps to conceive of a structural or metaphysical form of theories. Theorizing then means giving such metaphysical versions mediated form, with no inherent temporal or moral order between writing and diagrams. We can then understand theorizing as a specific form of translation, namely what I call horizontal translation. In horizontal translations, structural forms of theories are given form and because such forms are geometric, they lean towards diagrams. Further, the geometric elements allow transposition.

Finally, let me add a few observations of what follows from understanding theorizing as horizontal translations: First, it decentres the theorizer as a person and instead focuses on specific research operations. If we understand theorizing as a horizontal translation, we begin to understand that theorizing may happen at any stage during any research project. It occurs whenever a researcher shifts attention from translating an object in the world to translating from one concept to another. In the same way, as there is no specific sequencing of the media used in theorizing, there is also no specific sequencing of when specific kinds of translations occur. For the same reason, many practices by people who consider themselves theorists have nothing to do with theorizing and vice versa, a lot of practices of empirical researchers are indeed theorizing. Similarly, a lot of what happens in so-called theory sections of books and articles is not theorizing, as it is usually a recounting and critiquing of a given theory. More importantly, the person of the theorizer or in a more abstract version, what Julian Go in this special issue calls their standpoint, is not particularly important. Standpoint assumes that social location of the person of theorizer matters for theorizing. But as a corollary this deemphasizes the constructive logic of theorizing itself. The issue here is not whether theorizers can transcend their standpoint understood as social position within a field, which they surely can. Standpoint theories focus on the starting point of theorizing, at the expense of the internal operations, the horizontal translations of theorizing itself. Standpoint theory tends to reduce theorizing to a process in which the input determines the output, with little mediation going on between the two. A feminist starting point will result in a feminist outcome. It may be that empirically indeed such an input/output logic can be observed (and the designation of sociology as a relatively non-autonomous science supports this). But if we wanted to improve theorizing as a distinct practice, we would do well to focus less on standpoints and focus more on those horizontal translations as practices that transport a theory into some unexpected place away from the starting point. Such a focus would judge specific acts of theorizing (not theories) according to their powers to transcend their starting points by creating surprises derived from their own geometries. In other words, good theorizing would be work that surprises, disorients, and dislocates the theorizer and its audience and transports them away from where they started. With the language of Imre Lakatos, we could also say that if the endpoint of theorizing equals its standpoint, it is a sign of a failing research programme. While a successful research programme will extend a theory into a space that cannot be mapped to the initial standpoint (Lakatos, 1978).

A second issue follows from this focus away from the person to the process. Notice that an often rehearsed critique of theories has focused on their basic geometries. We can see this in the ubiquitous critique of ‘binaries’, or previously in the critique of Parsons four-field schemes and Levi-Strauss triangles. Such critiques are fundamentally a critique of the basic structure of horizontal translations, rather than their specific mediations. There is nothing wrong in critiquing either the universalizing moves of such very basic forms or the restrictions of these geometries. Such geometric critiques also demonstrate an understanding of theories in their structural logics. But where these critiques usually fail is in spelling out more complex structural alternatives to these specific geometries. If binaries are a bad and restrictive geometry, then no geometry is not the alternative, not least because its transpositional powers will be minimal.

Acknowledgments

The author would like to thank the editors, three anonymous readers, the co-editors, and authors of this special issue for their incredibly helpful and constructive comments. The author would also like to thank Turo-Kimmo Lehtonen for his generous invitation to present this work in Helsinki and all the participants in the Helsinki workshop for their constructive criticism.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

Michael Guggenheim

Michael Guggenheim is a sociologist who works at the Department of Sociology, Goldsmiths, University of London. He is the co-founder and convenor of the MA Visual Sociology at Goldsmiths. He has published widely on expertise, lay people, disasters, change of use of buildings, environmental research and food and social theory. He has developed numerous performative experiments, most recently together with Jan-Peter Voss the exhibition ‘Taste! Experiments for the Senses’ at the Museum of Natural History Berlin.

Notes

1 There are also a number of exceptions to the assumptions of temporal primacy, for example Swedberg discusses what he calls “theory sketches” (Swedberg, 2016) or in Silver (2020).

2 My thanks to an anonymous reviewer for pointing this out.

3 Though note, within the philosophy of natural science the same idea of eliminating diagrams persists (Griesemer, 1991). Empirical analysis of the prominence across disciplines however contradicts this view: the more diagrams, the harder a discipline (Smith, Best, Stubbs, Johnston, & Archibald, 2000).

4 See Esposito for a different attempt “to overcome the idea that abstraction is the opposite of empiricism” (Esposito, 2013, p. 141). For Esposito, the key here is „comparison“ that allows theory to compare radically different objects (elephants and mice via DNA, churches and the military via organisations)(Esposito, 2013, p. 141; also see Stöckelová, 2013).

5 Their use of the word diagram is confusing here, as it does not allow to distinguish material diagrams from diagrams as abstract machines.

6 Note that this process of abstraction is what is usually thought to be theorizing, see for example Alexander (1987, p. 3).

7 This is not an argument for structuralist theories, quite the opposite. It is true even for the post-structuralist forms of theory: They also need to relate a set of defined rules and concepts, and these are just as amenable to diagrams, see for example the diagrams of Deleuze himself.

8 Though see Becker (2007), who argues that such contemporary visual examples should count as sociology, and see Oliver (1979) for a contemporary take making diagrams richer.