Theory-building: working the theory-data spectrum

This article illustrates some implications of the view that data, theory and methods should be treated as a matched set. They are inseparable companions, equally essential parts of processes of scholarly research and dissemination. When we focus on one of these dimensions, the others are always implicit. When focus on all of them, our work is more complete, consistent and defensible. This contribution to this special issue highlights the value of focusing more explicitly on the inter-relations between all four levels of scholarly work.


Introduction: Disciplinarity
What makes the study of religion/s a discipline? Why has it come to occupy a separate institutional space in universities around the world (ALLES, 2008). 1 One answer is to suggest that the study of religion/s is just that, the discipline ("study") dedicated to its unique subject ("religion"). This approach fails immediately, because many other disciplines also study religious phenomena, including but not limited to anthropology, sociology, philosophy and psychology.
In the 1960s through 1980s, primarily, scholars of religion/s attempted to get around this counter-claim by focusing on the concept of 'religion': they proposed two inter-related criteria that serve to demarcate the study of religion/s as a separate discipline: sui generis or religionist views of religion; and a distinctive method, broadly "phenomenological," for studying religion seen in this way (SEGAL, 2006, p. xiii-xvii;see ENGLER;STAUSBERG, 2011, p. 129-30). This approach has largely been abandoned due to (i) critiques of religionist views of religion (e.g., WIEBE, 1984;MCCUTCHEON, 1997;FITZGERALD, 2000;NONGBRI, 2013) and (ii) the facts that the alleged method is neither unitary nor unique to the study of religion/s, as well as being problematic for various reasons (HANEGRAAFF, 1995;STAUSBERG;ENGLER, 2011;FUJIWARA, THURFJEL;ENGLER, forthcoming). Nor can scholars of religion claim to have unique theoretical approaches: we are far more likely to appropriate theories from other disciplines than to generate theories specific to our discipline (STAUSBERG; ENGLER, 2016). In sum, it is simply not the case that the study of religion/s is unique in terms of its subject matter, its method(s), or its theoretical approaches. These epistemic characteristics cannot effectively define, delimit or justify the existence of our discipline.
1 Thanks to faculty members and students in the Programa de Estudos Pós-graduação em Ciências da Religião, Pontifícia Universidade Católica de Minas Gerais -PPGCR PUC Minas, for the invitation to present at the 6th Colloquium of the Research Group on Religion and Culture in 2018 and for stimulating discussions during that event. Parts of this article draw on Engler (forthcoming). Meta-theory should also be considered part of this more defensible approach to research. I omit it here for the sake of space. My own meta-theoretical allegiance to semantic holism is set out in 2017); 2018); forthcoming.
Too much ink has been spilled in rehashing these epistemic issues, for example in the interminable "religious studies vs. theology" debates. We need to look elsewhere for the specificities of our discipline: Academic disciplines are not defined by epistemic properties alone. They also have distinct cultural and social ('tribal'), as well as institutional and organizational aspects […]. As cultural entities, disciplines have their specific taken-for-granted preferences, attitudes, instrumentalities, legitimate practices, moral orders (including visions of research ethics), modes of asking questions, standards and genres of writing, styles of thought, and tacit knowledge, which to some extent determine professional language, terms of appraisal, notions of academic credibility and credit, standards of excellence, intellectual substance, fashions and classics, and recognized ways of achieving reputation and status and of assessing the appropriateness of academic performance. As social ('tribal') entities, disciplines have their forms of cooperation and competition, heroes and foundational figures, brokers, entrepreneurs, and gatekeepers, their rules and rituals of admission, their traditional wisdom, their pecking orders and elites, their performances of loyalty and rituals of rebellion, their fairs, festivals, and meeting places. As institutional entities, the sustainability of disciplines depends on their recognition by academic institutions such as academies, universities, and research funding agencies, which allow for the disciplines to be practiced on a daily basis. They are 'institutions that demarcate areas of academic territory, allocate privileges and responsibilities of expertise, and structure claims on resources' […]. As organizational entities, disciplines are represented by academic associations, typically both on a national and an international level. … Practitioners of disciplines typically attend regular meetings (conferences) and publish in journals or with publishers that have an acknowledged disciplinary reputation (which brings the discussion back to the above-mentioned cultural and tribal aspects). Journal editors, series editors, and conference committees negotiate the intellectual territory of disciplines by safeguarding its professional standards and by excluding illegitimate immigrants from invading the tribe. While there are no hard criteria to establish scores for legitimate or illegitimate claims of disciplinary status with regard to these aspects, it seems to us that there is reason to believe that the social aspects of academic study of religion/s\s are sufficiently developed for its disciplinary status to be acknowledged. Besides these epistemological and social (cultural, tribal, institutional, and organizational) elements, there is a third aspect to disciplines that often tends to be forgotten: their pedagogical dimension. A discipline is a subject that is taught at universities and that is transmitted and inscribed pedagogically […]. STAUSBERG, 2011, p. 130-131;emphasis added). 2 The history of the study of religion/s in countries around the world has been one of slow consolidation of these cultural, social, institutional, organizational and pedagogical factors. Epistemic debates are secondary and largely unproductive. The take-home point for present purposes is that there is nothing specific to or unique about the study of religion/s in terms of data, methods, theories or meta-theories. These dimensions of research are interrelated, and this applies to all disciplines. At the same time, however, exploring their inter-relations offers a valuable window onto certain debates in our discipline.

Data
In light of post-positivist views of science, it would be hard to find a scholar of religion today who does not recognize that data are theory-laden. 3 This means that data and theory always travel hand-in-hand, whether we admit the fact or not. But what does it mean more specifically?
A useful point to start is J. Z. Smith's often-cited claim that ... while there is a staggering amount of data, phenomena, of human experiences and expressions that might be characterized in one culture or another, by one criterion or another, as religion -there is no data for religion. Religion is solely the creation of the scholar's study. It is created for the scholar's analytic purposes by his imaginative acts of comparison and generalization. Religion has no independent existence apart from the academy. (SMITH, 1982, p. XI, original emphasis). 4 Smith is neither a realist nor a radical relativist here. If we read the passages that precede and follow this "there is not data" passage -almost always quoted out of that context -we see that Smith is not concerned about the existence of some unitary thing to which the word 'religion' refers or not. His approach is pragmatic: he holds that "the scholar is not interested in religion per se, i.e., as it is in-itself, nor in knowing what it denotes or the reality it represents. Rather, she is interested in understanding it, in imagining it in a way that is fruitful or has some sort of pragmatic payoff" (GARDINER; ENGLER, forthcoming).
Smith should be read as restating the post-positivist view that data always arrive theory-laden on our table. 5 The early twentieth-century attempt to limit meaning to verifiable facts drew a distinction between interpreted data and 'brute data': "Verification must be grounded ultimately in the acquisition of brute data […] data whose validity cannot be questioned by offering another interpretation or reading"; this distinction should be rejected because of "the perpetual threat of skepticism and solipsism inseparable from a conception of the basic data of knowledge as brute data, beyond investigation". (TAYLOR, 1971, p. 8).
It would be a mistake to read Smith as saying that the case of religion is special insofar as there are no brute data with respect to that object of study.
This would reinforce the specialness of religion, a move that Michael Stausberg (2010, p. 356-357) calls "reverse-sui-generis-rhetoric". By reminding us that there are no brute data, Smith repeats post-positivism's well-known lesson.
However, by speaking of 'data for religion', he risks implying that there is something special about religious phenomena in this sense. Using "facts" to describe less formally what Taylor calls "brute data," Smith (1995, p. 413) makes this point crystal clear elsewhere: it "is only theories and concepts that convert facts into data, that render them significant as examples of larger intellectual issues which comprise the agendum, debated though it may be, of a field".
The disciplinary context of Smith's points is crucial. There are two separate issues here. First, there are no brute data, no facts, in any academic discipline, whether the study of religion/s, politics or history. Second, there are no essentially religious facts, the religiosity of which is independent of our scholarly operations. Of course, the latter is just the same point stated a different way. Once again, 'religion' is not special in this sense: in the same light, there are no essentially political or historical facts. What Smith is saying is that there are no brute data for scholarly work in any discipline: all data are interpreted, not just interpretable; they change from uninterpreted facts to interpreted data as soon as they are taken by the methodological procedures of academic research. In other word, there obviously are data for religion, i.e., phenomena that have come to be classified as 'religious' through the conceptual/ theoretical work of scholars of religion\s, among others. Thus, Smith says the opposite of what he is often taken as saying: there is nothing but data for religion, because once we let go of the illusion that brute data are available, everything is potential grist for the scholar's mill. In sum, there are no pure, brute facts independent of our conceptualization; all we have are data that arrive already theory-laden.
Smith develops this point about data in a way that illustrates the close connection between data, methods and theory. 6 He writes, What we study when we study religion is one mode of constructing worlds of meaning, worlds within which men find themselves and in which they choose to dwell.
[…] What we study. is the variety of attempts to map, construct and inhabit […] positions of power through the use of myths, rituals and experiences of transformation. (SMITH, 1978, p. 290-291).
In substantive terms -in terms of the subject matter or data of our discipline -he argues against religionist and sui generis views, against views that that all religious phenomena are manifestations of a unitary core, like the sacred. The objects of study are often purportedly universal religious claims, but, from the proper scholarly perspective, "this claim to universality is relativized by the situation" (SMITH, 1978, p. 300). Smith emphasizes not static content (e.g., a core list of religious phenomena), but the relation between elements identified by the scholar: "there is nothing that is inherently or essentially clean or unclean, sacred or profane. There are situational or relational categories, mobile boundaries which shift according to the map being employed" (SMITH, 1978, p. 291).
In theoretical terms, this highlights the relationship between scholarly claims and their putative objects. For Smith (1988, p. 235;1998, p. 281-282), "religion" is, like "language" in linguistics, a second-order concept: it orders theoretical constructs not objective (or brute) facts. This is his point in the famous quote discussed above -"there is no data for religion." The moment that some fact, figure or observation becomes "data," it is already theorized.
In methodological terms, Smith (1982, p. 15) discusses how scholars of religion construct their maps. This is the motive for his extensive and repeated discussions of comparison in the study of religion/s. His account of this 'mapping enterprise' involves several discrete steps, and his methodological tugof-war between similarity and difference raises important questions like the following: What are the criteria of a good scholarly map of religious phenomena? And are these criteria preserved as we move across different levels of relations between maps and territories, e.g., from mapping religious beliefs, to mapping different scholarly accounts of those beliefs, to mapping theoretical commitments made by different scholars?
2 What Is Theory?
The word 'theory' can mean many things. A first step toward narrowing its scope is to begin by describing scholarly uses of the term. Gabriel Abend (2018) distinguishes seven senses of 'theory' as used in sociology: 1. "a general proposition, or logically-connected system of general propositions, which establishes a relationship between two or more variables"; 2. "an explanation of a particular social phenomenon"; 3. "the development of lexica and schemata with which to talk about the social world"; "an original 'interpretation,' 'reading,' or 'way of making sense' of a certain slice of the empirical world"; 4. "the study of and the students of the writings of [classic theorists, i.e.,] authors such as Marx, Weber, Durkheim, Simmel, Parsons, Habermas, or Bourdieu"; 5. "a Weltanschauung, that is, an overall perspective from which one sees and interprets the world […] providing one with an a priori framework (scheme, grid, map, net, plan), a framework that is independent from experience, logically prior to any contact with the social world"; 6. "accounts that have a fundamental normative component. […] For example, the contemporary projects of 'critical theory,' 'feminist theory,' and 'postcolonial theory' are explicitly normative ones, which usually reject the fact/value dichotomy, and hence the supposedly value-neutral sociological theory"; 7. "the study of certain special problems […] [that] may be described as 'philosophical' problems, insofar as they call for reflection upon the nature of knowledge, language, and reality, and some sort of conceptual analysis." (ABEND, 2008, p. 175;p. 177-181).
Four of the seven senses of 'theory' on Abend's list above are especially relevant in the study of religion/s. 7 Sense 1 (formalizing relations between variables) is associated with a realist ontology and scientific method, e.g., testing hypotheses. Sense 2, explaining social phenomena, and sense 3 (interpreting the social world) are associated with a moderate constructionist ontology and interpretive methods, e.g., investigating the meanings that certain people attribute to certain phenomena. Sense 6 (normative critique) is associated with a radical (i.e., relativist) constructionist ontology and ideological critique, e.g., revealing the role of specific discourses (including those of 'theory' itself) in maintaining systems of domination, and advocating discursive changes as a means of attempting to change these relations.
The variety of senses of "theory" shows the uselessness of superficial debates "over what a good theory is, what constitutes a theoretical contribution, where theory should go, whether [a given discipline] … has made theoretical progress, which theoretical paradigm should be favored, what the functions of theory are, what it is for a paper to be an atheoretical one, and so on" (ABEND, 2008, p. 175-176). There is no hope of resolving these or related issues until participants agree on what sense(s) of 'theory' they share, and many such debates simply disappear after this initial clarification.
We could quibble over details of Abend's list, but three key points are clear: scholars use different senses of 'theory'; different senses will tend to be prioritized differently within a given discipline; and different sense relate in different ways to data and to the methods used to collect and organize those data.
The variety of appeals to 'theory' in the study of religion/s is more diffuse as a result of being less rigorously addressed. 'Theory' in the study of religion/s is a particularly vague category. As Donald Wiebe (1983) noted in a review of "the great 'theory/anti-theory debate'" the concept of theory is particularly amorphous in the study of religion/s: 7 I leave aside here the assessment and recapitulation of 'classic theories' (sense 4). Abend's distinction between sense 2 and 3 is basically that between explanation and interpretation. There are important arguments to the effect that this distinction between explanation and interpretation is not sharp: "the two 'are different cognitive tasks … [that] supplement and support one another" (LAWSON; MCCAULEY, 1990, p. 30;see ENGLER;GARDINER, 2010, p. 279). Hence, in terms of the methodological point that I am trying to make, sense 2 can be approached along the lines of either sense 1 or sense 3, i.e., more scientifically or more hermeneutically.
Even a cursory reading produces a bewildering variety of meanings for 'theory': 'theory' is used synonymously, or nearly so, with method, conjecture, approach, perspective, hypothesis, model, paradigm, explanation, view, way of understanding, conceptual scheme/framework, interpretation, etc. 'Theory' is often used in so loose a manner that it means little more than 'a solution to a problem or a generalization that "goes beyond the facts".' On the other hand, it is also used in so broad a sense as to be undistinguishable from metaphysics and speculative philosophy. (WIEBE, 1983, p. 295-296).
Abend's seven senses and Wiebe's more diffuse list both reflect the state of theory in the study of religion/s. There is no such thing as theory; there are competing perspectives that force us to be clear on how we think of theory in a given research or teaching context.

The Relativity of Theory
'Theory' and 'data' are relative terms -relational, mutually entangled, mutually constituting. Wiebe's list of uses of 'theory' in the study of religion/s suggests that the term can refer to anything that falls between two extremes on a spectrum between 'the facts' and metaphysics. Data are often opposed to theory in stark terms. But data are theory-laden, as critiques of positivist views of science have made clear (ZAMMITO, 2004). At the other extreme, in metatheory -theorizing about theories -theories themselves become data; and meta-theoretical approaches can serve as data for even more abstract and general analyses. In sum, the word 'data' does not pick out a specific type of thing that is radically distinct from theory. Data is more empirical but always still theory-laden to some extent, and theory' is more abstract, but can still itself serve as data. The data/theory distinction is relative. The two can and should be characterized distinctly, but this distinction is strategic and relational, not absolute. We should not think of them as distinct types of things but as contrasting vectors: they point in opposite directions, empirical and abstract, with a force that depends on the case at hand.
The labels 'theory' and 'data' maintain their relative position (one more abstract and one more empirical) along a scale of varying degrees of generalization and systematization. For example, concepts are parts of theory that categorize or relate empirical materials, yet theory can also be about concepts, taking them as data. Jeffrey C. Alexander's chart of "the scientific spectrum and its components" presents a range of positions, ranging along a spectrum from empirical to metaphysical (Figure 1). From any given point along the spectrum, elements closer to the empirical end serve as 'data' and elements closer to the abstract end as 'theory' (ALEXANDER, 1982, p. 3;see GLAZIER;GROVER, 2002, p. 318;ENGLER, 2011, p. 262-266).

Figure 1. The Data-theory Spectrum
Everything is relative in this conception: nothing is pure theory and nothing pure data. Everything toward the theory end (the top) of Table 1-e.g., definitions, concepts, and general presuppositions-can also serve data. The entire spectrum, for example, is data for the discussion of theory that you are currently reading. Similarly, everything toward the data end (the bottom) of Table 1 is theory-laden: e.g., there are no 'pure,' theory-free observations: each is separated out from its background and context, selected from a certain perspective for a certain purpose and re-described using a certain vocabulary.

Theorizing and Theory Building
If the distinction between theory and data is relative -as illustrated by the theory-data spectrum (Figure 1) -then working with theory is not just a matter of applying a highly abstract conceptual frame to a set of data. Rather 'theory' is in some sense related to all the conceptual work from initial redescription of data, through selecting or generating concepts and categories, to working with more abstract explanatory or interpretive frameworks. The research process is a dynamic activity of leveraging one's position up the datatheory spectrum. No particular level of start-or end-point is required for that movement to qualify as an act of theorizing. Different disciplines and subdisciplines have customs-and individual scholars have habits-regarding the definition of 'theory.' Theory takes place on a variety of levels in the study of religion/s, with definitions typically functioning as entry points.
This leads to a distinction between theory as an independent structure or schema -a sort of form that can be applied to the content of data -and theorizing as a process. Theorizing is not limited to a frame or schema to be applied late in the game, after the data have been neatly arranged: "theorizing is of a practical nature, it draws on a number of basic theorizing tools (such as abduction, abstraction, and analogy), and the area covered by theorizing is considerably larger than that of conventional theory" (SWEDBORG, 2014, p. 189).
This broader view of theory as process problematizes frequent appeals, in the human and social sciences, to a narrow, normative view of scientific method.
It shifts our attention to the context of research, foregrounding pragmatic issues of why a given research project is being undertaken and what impacts it might have: "Theorists often write trivial theories because their process of theory construction is hemmed in by methodological strictures that favor validation rather than usefulness" (WEICK, 1989, p. 516).
Pragmatic concerns -the challenges of moving from empirical materials to high-level theory -along with post-positivist critiques of science and knowledge lead to 'middle-range theories' being seen as paradigmatic: i.e., "theories that lie between the minor but necessary working hypotheses that evolve in abundance during day-to-day research and the all-inclusive systematic efforts to develop a unified theory that will explain all the observed uniformities" of those phenomena considered central to a given discipline (MERTON, 1967(MERTON, [1949, p. 39). The scope of items in Figure 1 represents this middle-range: the spectrum of theoretical levels most commonly met with on our experiences with theory.
The relational nature of 'theory' and 'data' undermines assumptions that 'data' are brute features of 'reality.' Concepts and categories, even theories themselves, can be data in the same sense (i.e., stand in the same relative relation to a more general level of theory). Recognizing this is especially useful for working with the constructionist ontologies and interpretive methods that are prominent in the humanities and, to a lesser extent, in the social sciences (making sense of the different logical levels of the 'raw materials' of construction). This recognition also highlights an important distinction between level of theory and type of theory. The former refers to issues of where a particular instance of theorizing is situated along the data/theory spectrum, as described in the next section, and the latter falls under the heading of metatheory, theory of theories.

Methods: Coding and Redescription
A common problem with theorizing is the failure to distinguish between levels of theoretical work: e.g., cherry-picking a single concept from a 'classic' theory, thus severing its relation to other concepts in that theory. 'Habitus,' for example, often appears in papers and articles that claim to be using 'Bourdieu's concept' or 'Bourdieu's theory,' yet that make no reference to the concepts of 'field' and 'capital.' Extracting the one concept from its theoretical context is a sign of superficial appropriation: this is no longer Bourdieu's concept.
Theorizing with 'habitus' requires either a fuller appropriation of Bourdieu's work-closer to the top of Figure 1-or the elaboration of another theoretical frame at a comparably general level, linking 'habitus' to other concepts in a different way. In general, where 'habitus' is appropriated superficially, references to 'Bourdieu' function as a rhetorical appeal to authority.
Redescribing empirical materials with a single out-of-context concept is very different than working with a complex theory, and hand-waving references to classic theorists does not make up for the gap between these.
The basic point -that data are theory-laden -repeats itself at various stages of the research process. The theory-data relation already appears in initial research design and data selection. For a scholar of religion to choose certain phenomena as relevant to work in the discipline, that scholar must already have at least an implicit distinction in mind between religious and nonreligious phenomena. The choice of certain facts or observations as being relevant to the study of religion/s already contains an act of classification that is highly theoretical. In other words, the theory-data relation is already present in data collection. This illustrates the inter-relation between data, theory and methods. Theoretically informed concepts and categories are already present in choices regarding what sites or situations to observe, which texts to read or persons to interview, which interview or survey questions to ask etc.
The next point at which this issue appears has especially pressing methodological importance. The scholar must choose a vocabulary with which to describe the empirical materials that have been chosen and collected. Coding In this naïve -and all too common -approach, the scholar reads a text, transcript, image, video, artifact or space in order to intuitively select points, issues or themes that are considered "relevant" or "important" in some vague sense. This sense of what is important comes from academic training: familiar models of reading are implicitly appropriated and applied. This happens at a general level: e.g., scholars of religion are trained to use the category of "ritual" in order to distinguish between "religious" and other action. It also happens at a more specific level: e.g., where interactions between individuals are categorized according to concepts like "authority" and "power." Two critical points. First, these concepts bring theoretical presuppositions with them: e.g., focusing on issues of power predisposes the scholar to adopt certain theoretical perspectives (e.g., poststructuralism) over others (e.g., rational-choice theory). It is not that that such theoretical choices are wrong. Rather, they underline the second point: more formalized coding processes are essential for more effective conceptual and theoretical work (more responsive, less mechanical and less derivative), regardless of one's metatheoretical preferences.
In an informal approach to coding, the process of choosing a sub-set of the empirical materials (e.g., certain passages in a text or interview transcript) proceeds "intuitively." This approach obviously risks the unconscious replication of dominant conceptual frames. Minimally, if coding is performed by rote, with little reflection, solely on the basis of disciplinary training, then new research will tend to replicate old research: novelty will depend on new groups of scholars who continue to code superficially on the basis of divergent training, e.g., replicating newer, trendy theoretical frames.
At the other extreme, a more formalized approach would develop standardized coding procedures. It would also assess their reliability and correct for coding errors, as well as screening data for consistency and establishing procedures for cleaning or regularizing them. This can be overly rigid and also impose preconceived conceptual frames. Approaches like Grounded Theorywhich generates concepts and categories through sensitive, reflexive coding techniques and abduction -offer a useful middle ground (ENGLER, 2011).
There is a sliding scale, not a sharp distinction, between informal and formal coding processes. This can be illustrated by the process of redescription. Figure 2 illustrates three potential paths or approaches in redescription: two extremes and a mid-point on a spectrum. All three paths begin with the same set of empirical materials. Path 1 immediately uses highly abstract concepts to redescribe these materials, and in doing so, it jumps too far up the data-theory spectrum. It redescribes the data using concepts that immediately import a given theoretical perspective, foreclosing on others. Path 2 uses less abstract concepts in the initial redescription, but caution is still required to avoid importing or presupposes an over limited conceptual line. Path 3 is closer to Grounded Theory in that it takes smaller steps of redescription, abstracting little-by-little, in closer dialogue with the data.

Figure 2. Redescription
To exemplify, if interviews reveal that a Protestant church is planning to enter into a partnership with a non-religious charity, it would be a mistake to begin by labelling or coding this as 'secularization.' This would presume what needs to be demonstrated: that it is an issue of a decline in religiosity in some sense. In the same way, labelling it a case of "the extravasation of the sacred" would also presuppose too much: it would impose an interpretation in the initial act of redescription (HAMMOND, 2000). Coding and redescription should arrive through a series of cautious steps at more abstract concepts and categories, not apply them at the outset. Highly abstract concepts that come already rooted in certain theoretical perspectives should be avoided in early stages of coding or redescription: regular activities should not be labelled 'habitus'; moral formation should not be labelled 'governmentality'; appeals to scientific metaphors in religious texts should not be labelled 'disenchantment'; etc. To give a more concrete example, classifying the plays of Nelson Rodrigues as "gnostic" would be to jump from a recognition that themes of good and evil are prominent to a high-level abstraction that brings with it a whole series of ontological, ethical and soteriological themes that go far beyond what is found in the plays. Figure 2 illustrates three paths of coding. Path 1 is a case where a highly abstract concept from a specific theory was imposed right at the beginning of coding. this imposes an interpretation based solely on that theory: the conclusion comes pre-packaged in the theoretical assumptions that are smuggled in at the outset. Path 2 moves more cautiously, but still imports assumptions by coding with overly abstract concepts too early in the process.
Path 3 moves more cautious, advancing in small steps, using codes (themes, concepts, types, categories etc.) that increase the level of abstraction only slightly in each round of coding. This allows for an interpretation to emerge to a greater extent from the data; it avoids to a great extent imposing an interpretation.
This over-extension of coding and redescription is often found in the work of graduate students. Readers of this article can no doubt think of examples of the following process: 1. first look for a high-level theory to be 'applied' to the chosen case; 2. choose such the theory based on an initial characterization of the case, emphasizing a small number of features that present prima facie salience; 3. redescribe the case using the technical terminology of this theory;' 4. label the resulting theoretically ad hoc interpretation as a 'conclusion.' This sort of process forces the case into the shape called for by the theory, as opposed to letting the salient features of the case emerge in more gradual steps of abstraction, through coding and redescription. This can be a useful, if limited, exercise in concept work, but it should not be mistaken for serious research. Dozens of different theories could be used to interpret any given case that we choose, yielding widely divergent interpretations: e.g., some sub-set of Marxist, Freudian, Durkheimian, Weberian, Bourdieuian, Foucauldian, rational choice, Bergerian social constructionist, Luhmannian system-theoretical, post-structuralist, decolonial, cognitive, evolutionary theories, etc. Choosing among theoretical options at the beginning of a study, based on an initial, and hence still superficial, acquaintance with the details of the case, imposes one reading at the expense of others. A more responsible approach is to work the case in more detail, moving more cautiously toward greater levels of abstraction. If work converges on a given pre-existing theory, that fit will be richer and more defensible.
In sum, just as middle range theory is more useful and common in the study of religion/s, redescription and coding should work toward smaller steps of abstraction, avoiding highly abstract concepts, except potentially in late stages of analysis.

Conclusion
We cannot think or talk about any of data, methods or theory without also taking account of the other two. We can separate them analytically, but they are not independent things. They are mutually constituting, overlapping, a sort of meta-theoretical Trinity. We can and should distinguish them for contingent purposes in specific contexts of research, yet without losing sight of their fundamentally relational nature.