The semantics of Afro-Brazilian spirits: Applying Davidson on prior and passing theories

Abstract

Based on fieldwork with two related Afro-Brazilian religions, Umbanda and Quimbanda, this article explores the value of Donald Davidson’s semantic theory for making sense of ethnographic fieldwork. Specifically, we look at the role of scriptedness in communication, including religious ritual. We first clarify the role of social externalities in Davidson’s view of communicative interpretation, which is broader than his initial framework of radical interpretation. We then offer an account of what constitutes communicative and interpretational success, by drawing on Davidson’s account of prior and passing theory. Prior theories are interpreters’ initial hypothetical frameworks, ranging from general (e.g., the rational, intentional nature of self and other, and a shared perceivable world) to local (e.g., assumptions about cultural, social, and institutional contexts). Passing theories are tactical, on-the-fly modifications that we hypothesize in order to get mutual understanding back on track. We introduce the concept of ‘semantic reduction’ to operationalize the view that specific, local social externalities provide clues that help keep interpretation on track. In the case of religious ethnography, these include ritual, doctrinal, narrative, symbolic, material, temporal, and spatial frames that constrain the generation of passing theories. Examples from fieldwork illustrate the potential value of our appeal to Davidson’s ideas.

Certain communicative encounters are at least partly scripted: for example, actors on stage; superficially polite exchanges of greetings; a physician taking a patient’s history; a waiter taking an order; a salesperson’s pitch; a job interview; or an airline attendant giving safety instructions. This is also true of much religious language and behaviour, especially ritual. This article explores how a particular theory of meaning, tied to the work of Donald Davidson, can help us make sense of the role of scriptedness in communication, including forms of religious behaviour. We ask how the clarification and application of some of Davidson’s ideas might help make sense of talking to spirits in Afro-Brazilian religions.¹

The religious context

This article places semantic theory in dialogue with the ethnography of two Afro-Brazilian religions, Umbanda and Quimbanda (Borges et al., forthcoming; Engler 2016). These religions present four features that offer challenges for interpretation, and this makes them a useful test-case for our dialogue between semantic theory and religious ethnography. They raise methodological questions about interpreting mediums versus spirits. The majority of those attending rituals are non-members who come for ritual services, especially healing; and this prompts us to place more emphasis on ritual contexts, de-emphasizing doctrine. Beliefs and practices in the two religions vary greatly between individual groups and regionally. And their historical origins and religious classification are not clear, which rules out comparison to supposed root traditions: given the lack of evidence, Umbanda might have originated in Afro-diasporic ritual traditions, in the European esoteric tradition of Kardecist Spiritism, or in some mixture of these and potentially other traditions (Engler 2020; Engler, forthcoming). The same appears to apply to Quimbanda, though little is clear regarding this seldom-studied religion. As a result of these factors, interpretations must emphasize observations and interviews in specific groups, rather than inferring from allegedly comparable cases.²

Umbanda is the largest Afro-Brazilian religion.³ It is a Brazilian Spiritist tradition that emerged in the early twentieth century. (‘Spiritist’ is an umbrella term for esoteric possession traditions in which members communicate with the spirits of the dead, as these incorporate in mediums.) The basic doctrinal frame of Umbanda comes from Kardecist Spiritism (a mid-nineteenth-century French offshoot of Spiritualism). In the public rituals of Umbanda, different types of powerful, highly evolved, caring spirits (previous incarnate in their former lives) incorporate in mediums during rituals. They offer one-on-one consultations, advice, and ritual healing services. The majority of people attending rituals are non-Umbandists – usually Catholic or Kardecist – who attend for ritual services. A few dozen to several hundred clients might attend a given ritual, each being seen and spoken to by a spirit who has incorporated in one of the half-dozen to fifty or more mediums. The needs of the spirits (organizing clients, spreading incense, providing liquor, lighting a cigar, or fetching herbs for healing smoke) are catered to by ritual assistants, called combones, who are often mediums in training. Quimbanda is a related tradition that, in general, shares all these features. It differs in its more exclusive work with a sub-set of morally ambivalent spirits – potentially dangerous but powerful protectors – that offer not only healing but also more pragmatic magical rituals. For example, often for payment in cash or kind, Quimbanda’s female Pombagira and male Exu spirits can assist with romantic and business rivalries.

The task of making sense of the rituals of Umbanda and Quimbanda presents a useful challenge for the study of religion and for philosophy. We suggest here that Davidson’s idea of the dynamic relation between prior and passing theory – supplemented by our concept of semantic reduction – provides a useful theoretical and methodological resource.

Meaning

It is commonplace to think that the meanings of words are tied to what they denote or describe. To be sure, reference is an important part of how we arrive at meanings. In the case of objects or situations that are directly perceptible to our senses, we match what people say to what we can see, hear, and so on. If someone says that it is raining, that a bowl of soup has too much salt, or that a certain rock is surprisingly light, we can look, taste, or lift to assess those statements. Nonetheless, there are problems with reducing meaning to reference, a view often called representationism in the philosophical literature. The operating assumption of representationism is that words mean by pointing to objects or states of affairs. The main hallmarks of this theory of meaning are that the core function of language is to represent, and that the meaning of an expression is given by what it represents (i.e., you know what an expression means when you know what it represents). Methodologically, then, to find out what someone means, you should trace back from their words to the world. While this appears common sense when talk is about empirical things, it leaves us in a difficult position when people speak of things that are less directly accessible to our senses. What is the referent of abstract concepts like justice, atonement, social structuration, mana, or phenomenological reduction? How do we make sense of talk about non-empirical things like gods, spirits, and afterlife states? How do we match words to referents in such cases? There is no doubt that we do interpret such talk, easily or with difficulty, rightly or wrongly, amiably or contentiously, and this should be enough to make us leery of representationism, or the idea that reference is foundational to meaning.

Donald Davidson’s response was to reverse the order of priority: meaning is prior to reference. Certainly in non-empirical cases, you cannot know what someone is talking about until you first of all understand their words. In his view, meaning consists solely in what is involved in coming to understand, or interpret, another’s words. Hence, we label his view ‘interpretationism’.⁴ Over the course of his long career he developed that view in remarkably sophisticated ways. In this article we concentrate on one aspect: the role of theorizing in interpreting, and extending it to the case of understanding a particular type of ritual: spirit incorporation in Afro-Brazilian religions.⁵

Radical and communicative interpretation

In his earlier writings, Davidson focused on radical interpretation (RI) (e.g., Davidson 1974/1984). RI involves an interpretative encounter between two people – a speaker and an interpreter – where the interpreter has no prior knowledge of the speaker’s language or thoughts. He argued that if such a radical interpreter can come to understand the speaker’s words, then meaning or meaningfulness itself will be fully grounded in whatever is necessary for interpretative success. (His interests then were largely to naturalize semantics, which he saw as a form of the general trend towards naturalizing epistemology.) In a slogan, meaning is grounded in what is required to radically interpret.

Davidson argued that radical interpreters track speakers’ total behaviour, including their speech behaviour, against an environmental background. Interpretation is triangular: speaker, world, interpreter. But it is not enough for the speaker to make vocalizations correlated with changes in the environment in regular and predictable ways. It must also be the case that the speaker intends to convey meanings. This distinguishes between intentional communication from other sorts of predictable auditory behaviour: for example, squirrels chirping when someone encroaches or the ticking of clocks. As a methodological norm, Davidson proposes the Principle of Charity (PoC): the interpreter starts off assuming the speaker to be an intentional language-using agent who is as rational and sensitive to evidence as themselves. PoC is presumptive: it must always apply at the beginning of an interpretative attempt, but it can be overturned for any number of reasons. For example, it would not take long to decide that the clock’s ticking behaviour was better explained by its internal mechanics than by an intention to mark time, or we might decide that the squirrel’s behaviour, while intentional in some sense, is not best described as linguistic (propositional) in nature.⁶

In later writings, Davidson shifted his focus from what is required for meaningfulness to what is required for us to have thoughts in general, such that we can mean something by our words (e.g., Davidson 1994). The original model of RI, correct in outline, was too simple and needed supplementation. Three of his ideas here are important for this article: active participation, shared worlds, and theorizing.

First, the model of the radical interpreter as neutral observer of the speaker’s behaviour against the environmental background was too passive to ensure even a modicum of success. Davidson began by emphasizing two-way communication, where meaning is co-constituted by the communicants in interaction, though still within a shared environment (i.e., the third point of the interpretative triangle).⁷ We call this interactive process communicative interpretation (CI) to distinguish it from radical interpretation (RI). These are not competing views: CI is possible only because RI is, but RI only gets our foot in the interpretive door. CI is ubiquitous: we rarely, if ever, radically interpret, but we communicatively interpret all the time. We are never not communicatively interpreting whenever we speak, read, or communicate in any form. In the normal fast-moving back and forth of communicating – especially with those we are close to, personally, culturally, and so on – we do not experience RI, but we would need to fall back on it when ease of communication breaks down radically. This article applies Davidson’s explicit theorizing in order to give a fuller account of how challenging interpretive contexts, like speaking to spirits, are normalized as communicative (not radical) interpretation by certain sorts of contingent and locally contextualized conventions.

Second, communicants must have a good deal in common. This was established to some degree by PoC, but it needs to be extended. In particular, the third point of the interpretative triangle, the world, must be understood as more than just shared. It must also be the case that the communicants’ understandings of it must overlap considerably. Davidson’s well-known denial that there can be alternative conceptual schemes about how reality is arranged makes a case for considerable shared background knowledge in CI (Davidson 1974/1984). This is not to say that speakers and interpreters always agree; rather, disagreement is detectable only against a larger backdrop of agreement.⁸

For Davidson, our beliefs about the world, at least a good deal of the time, are caused, at least in part, by our interactions with it. Knowing those causes, then, is pivotal in understanding or interpreting what is believed. Therefore, there are external contributions to meaning; interpretationism embeds a form of semantic externalism. (It is not referential, because it does not hold that the meaning of a word is given by what it refers to.) ‘Rabbit’ may refer to rabbits, but its meaning (its role in CI) does not reduce to that reference relation. Its meaning is given by the role that it plays in interpretation, and that role, even when originating in responding to rabbits, reflects or just is a vast web of holistically interconnected usages. Thus, Davidson’s view also embeds a form of semantic holism. Holism allows that my understanding of words can be somewhat different from yours, but not so different that I can never understand you.

Davidson distinguishes two types of externalism, perceptual and social (Davidson 2001). In perceptual externalism, the world apex of the triangle is shared perceptually. To oversimplify, this is the empirical, natural, physical world, though it is not knowable apart from language. By contrast, if I know nothing about chess but regularly hear you say ‘checkmate’ when moving a chess piece late in a game, followed by your opponent cursing and walking away, then I may be in a good position to decide that your utterance means something to do with winning conditions having been met. Perception plays a role in this particular case, but is it insufficient. Many non-empirical background assumptions are built in: I must believe that you’re playing a game, that the game is rule driven, and the rules include the winning conditions, and so on. (Dummett goes on to argue that I must also presume that you are trying to win (Dummett 1959/1978), and so that there may be a normative element to the shared world.) But I need not assume that all relevant background is part of the natural world: social facts and social worlds are central to CI.

Perceptual and social externalism, natural and social worlds, differ in degree not kind. Think of the semantic difference between talking of rocks and talking of megaliths. We are born into the same natural world but not the same social world: we are all humans but not all, say, Zoroastrians.⁹ Three Davidsonian theses minimize this difference to a considerable extent.

First, the commonality of our beliefs must overshadow their differences, whether they are beliefs about the natural or a social world. This guarantees that, at least in principle, we can interpret the worlds and beliefs of ‘insiders’ even when we are ‘outsiders’ (Gardiner and Engler 2012). We have enough shared resources that I can generally come to understand, to a useful extent, how you conceive of your social world. (This is not necessarily the case. Our argument rests on the weaker claim that ethnography and many other social sciences presume that people do generally arrive at such understandings.)¹⁰ ‘Useful’ here marks the ability to interpret another’s words and thoughts about that world to an extent that meets the contingent goals of communicative encounters. To underline a crucial point, pragmatic success is the only measure of interpretative success on the Davidsonian model. There is no meaning to be grasped ‘out there’, independent of our interpretative encounters. This still allows what counts as pragmatic success to vary from context to context. We return to this point in the following section.

Second, Davidson argues that there is always a social dimension to meaning, even in purely perceptual cases.¹¹ Social worlds are constructed, not given, and they can be shared in the relevant sense needed for CI. Davidson is committed to much the same for the natural world. What accomplishes the important interpretative work is not so much the goings-on in the natural world itself, but our conceptualizations or understandings of those goings-on. Siding with Sellars, there is no ‘given’ of perception (Davidson 1997/2005; Sellars 1956/1997). What it is we perceive can only be decided descriptively: it is semantically individuated; our sense of what it is comes about through (and is not prior to or presupposed by) acts of CI. The meaning of any word is not fixed by something we can point to, but only by the myriad holistic uses we put the word to.

Third, the initial description of the radical interpreter having no prior knowledge of the speaker’s language is qualified in the case of CI. The transcendent necessity of PoC as an enabling pre-condition for interpretation is tantamount to an assumption that the speaker is a linguistic agent. Davidson’s early work (on truth-conditional semantics, drawing on Tarski) had argued, in effect, that communicants must bring some semantic and grammatical presuppositions to the table, even if those presuppositions get quickly shaken up. Many linguists and philosophers of language have supposed that these include a presupposition that linguistic signifiers are arbitrary and hence can only acquire a meaning by convention. This would be a form of social externalism, but not one that Davidson endorses (Davidson 1986). First, he points out that communication often succeeds even when such conventions are violated, such as with malapropisms. This suggests that, even where such conventions are present, they do not impose conditions for correct and incorrect usage in any interesting sense. Second, Davidson argues that such conventions presuppose meaning and meaningfulness, and so cannot be their source: there are no literal meanings that are somehow fixed prior to and guide successful communication. This is not to deny that there are social linguistic conventions, only that they cannot serve as the foundation for a viable semantic theory. They are useful shortcuts in many, if not most, of our communicative encounters. As is the case with reference and representation, conventions are useful tools in the process of arriving at meanings, but they are not the basis of meaningfulness.

More dramatically, Davidson infamously writes ‘there is no such thing as a language, not if a language is anything like what many philosophers and linguists have supposed’ (Davidson 1986, 446). By this he means that, semantically, the idiolect has priority over the dialect: languages do not have a being and structure prior to acts of interpretation. Languages are what drop out of successful communication; they are not its enabler. The fact that overlapping usages tend to clump together in a statistical sense – correlated with ethnic, national, regional, and institutional populations – falls out from his semantic theory. Once we reject the idea that conventions have a basic normative role in constraining meaning, we end up with a methodologically productive view that languages and cultures are open-ended, amorphous, overlapping, fuzzy, empirically tractable groupings that are constantly changing, diverging, converging, being born, dying, and so on. In other words, from a Davidsonian perspective, the challenges of ethnography are the same, if more challenging, as the challenges posed by every conversation we have ever had.

Prior and passing theory

Davidson calls the set or range of prior semantic assumptions that we bring into communicative encounters ‘prior theory’. CI works smoothly – in the sense that both parties understand each other at least moderately quite well – when prior theory is shared. It is not that each employs an independently existing set of linguistic conventions, but rather that the two parties’ sets of such assumptions are highly similar. When prior theory is sufficiently shared, we move on, not even noticing how well mutual interpretation is going. When communication is not so smooth, we adjust our prior theories into what Davidson calls a ‘passing theory’. Prior theory is what we bring to the table, and passing theories are tactical, on-the-fly modifications that we hypothesize in order to get mutual understanding back on track.¹² If passing theories converge, we move on. If not, we continue to tinker and modify, leading to a refinement of the passing theory until we achieve success (or give up). We do this in all communicative encounters.

As noted above, communicative success does not involve grasping real meanings: there are none independent of interpretative encounters. For Davidson, there can be no external standards for correct interpretation.¹³ Any such criteria must be found in the interpretative encounter itself. This in no way abandons the semantic externalism mentioned above. That view requires only that, when interpreting, we must pay attention to more than just words.

As the criterion of understanding, Davidson talks about convergence on passing theories (1986, 445). Care must be taken in understanding this. No passing theories are literally shared: yours remains yours and mine remains mine. Rather, they ‘converge’ when they are sufficiently similar, in a way that warrants ascription of mutual understanding, which is just to say that they enable communicative success. Davidson recognizes that this is circular: ‘This characterisation of linguistic ability is so nearly circular that it cannot be wrong: it comes to saying that the ability to communicate by speech consists in the ability to make oneself understood, and to understand’ (1986, 445). This is a pragmatic measure: we mutually understand each other when we can successfully interact. As Davidson says, ‘we have erased the boundary between knowing a language and knowing our way around in the world generally’ (1986, 445–446). More specific instances of understanding may, and will, involve more specific ways of making our way in the world.

Davidson acknowledged ‘degrees of application’ of passing theories (1986, 445). Pragmatic success in a conversation with a server in a restaurant is probably best measured by getting the meal you wanted. In a corporate setting, it might consist in the adoption of a collective course of action. In an academic setting, it might consist in convincing peer reviewers that your ethnography is worth publishing.

From a more general perspective, Davidson argues that reasons are the causes of intentional behaviour, and that understanding the content of those reasons is inseparable from explaining the causes of the behaviour (and vice versa) (1991/2001). Thus, we can take the general aim of the ethnographer to be that of explaining the behaviour of some individual or group in a given context through understanding their beliefs and intentions relevant to that context. The more consistent and coherent the interpretation or explanation is over a broad range of contexts, the more it connects to the material elements of the context, and the more projective it is to new cases, and so on, then the more successful it will be, from an academic perspective.

What we aim to do in all such interpretative encounters is to figure out the semantic content of each other’s actual linguistic intentions. Prior theory is the set of assumptions that we bring to this interpretive table, but, unfortunately, Davidson says little about what prior theories look like. It clearly forms a spectrum from general to local, but he pays little attention to the latter. At the most general level, prior theory consists of the set of beliefs and hypotheses that prepare communicants to treat themselves and each other as rational, intentionally acting beings in a given context of CI. At a more local level, it consists of assumptions about the social, cultural (e.g., religious) and institutional (e.g., ritual) contexts of specific communicative interactions.

Communicants’ prior theories do not need to correspond to any particular degree. All that is necessary is that each communicant have a theory, however rudimentary, that will kickstart interpretation. Nothing guarantees success or failure of a given communicative encounter, but the more general beliefs – for example those embodied in the PoC – are required to get it off the ground: prior theory is ‘prior’ only in that sense.

Here are two cases, to illustrate that CI is more efficient when communicants start with similar prior theories. This efficiency is a sliding scale. The first case is where communicants share similar linguistic, cultural, religious, family, or institutional contexts, and/or where they have a history of communicating successfully with each other on similar topics in similar contexts. This case results in very smooth, less problematic, communication. When prior theories start off as similar, passing theories converge more quickly: each interpreter’s first hypotheses about what the other ‘means’ with a given statement (or action) are effective at furthering mutual understanding. This allows the communicants to move on with minimal confusion, ambiguity, uncertainty, or incomprehension. The second case involves more challenging instances of CI, where less prior theory is shared in the relevant sense. Ethnographic fieldwork is an obvious example. Communicants might have to struggle with every statement that they make to each other. Communication might break down completely. The two cases represent relative differences in the likelihood that passing theories will converge, not qualitatively different phenomena.

Davidson’s account needs more clarification regarding where and how revision of prior theory takes place. The general end of the spectrum of prior theory includes, for example, the assumption that one’s interlocutor is rational (PoC), a systematic collection of presumed conventional meanings and grammatical rules of some recognized language (e.g., English), basic logic, and so on. The local end of the spectrum includes the issues that we highlight in this article: expected behaviours, norms of conduct, standard modes of discourse, typical topics of conversation in particular social, institutional and cultural contexts. These are also presuppositions that are brought to the table and that, when shared by communicants, make CI more effective.

This broadens the notion of prior theory to include any evidence the communicant is predisposed to bring into the interpretative encounter, not just general assumptions of the PoC, lexical, grammatical, and logical sort. Crucially, this includes not only assumptions brought by the speaker to the communicative encounter (i.e., the beliefs about language etc. that the interpreter has on their way to meet someone for a conversation) but hypotheses about empirical aspects of the encounter itself (e.g., inferences from the nature of the setting and the self-presentation of the other person).

This relative distinction between general and local aspects of prior theory is, we argue, where Davidsonian semantics has direct implications for the study of religion. Each successful interpretative encounter delivers convergent passing theories, which, at least in a good many cases, can then serve as prior theory in future encounters. Assuming some degree of semantic inertia – that people tend to talk and act in similar ways in similar contexts – this move is justified as at least a starting point. Expectations might need to be adjusted when one arrives to observe a ritual or conduct an interview: the ritual is in a home, not a religious space; the interviewee is a person with a different religious role than expected; or they choose to speak in English, not the local language. Prior expectations are abandoned, but they are replaced with a new set of expectations, some still prior to the communicative encounter, and some in effect passing theory. This view converges on methodological best practice: ethnographic fieldwork should be extended over time and space, and ethnographers should be able to piggy-back off each other. A published ‘theory’ of a particular ritual, say, can be thought of as an ossified form of a prior theory for students to learn and apply to new cases.

Prior theory involves not only what we bring to a communicative encounter but what we initially infer from the perceptual and social externalities of the immediate context of that encounter. Davidson notes this in passing: ‘I assume that the interpreter’s theory has been adjusted to the evidence so far available to him: knowledge of the character, dress, role, sex, of the speaker, and whatever else has been gained by observing the speaker’s behaviour, linguistic or otherwise’ (Davidson 1986, 100). Knowledge and inferences about dress, role, gender roles, religious doctrines of a group, particular ritual contexts, and so on rest on wider cultural and historical knowledge. Insofar as this can reasonably be taken as interpretative evidence, it serves as prior theory. This sort of knowledge functions as interpretative evidence only within specific contexts. A ritual gesture, for example, stands in need of interpretation. ‘What the gesture means’ is not to be thought of in symbolic or semiotic terms. It is understood not by cracking a code, but by seeing the role that it plays as prior theories morph into passing theories, in the search for interpretational convergence.

There is an ambiguity here, regarding what ‘has been gained by observing the speaker’ (Davidson 1986, 100). The evidence that we accumulate during a given communicative encounter can be prior (insofar as it shapes ongoing interpretation future words and actions) and/or passing (insofar as it constitutes a revision to previous assumptions).

This underlines a crucial point: the distinction between prior and passing theory is context-relative, because it reflects the role of beliefs in dynamic acts of interpretation, and because some of these beliefs are about the immediate contexts of those acts. I once accompanied an informant to what they called ‘a Candomblé’, where that person frequently attends public rituals. The layout of the space was different from Candomblé rituals that I was familiar with. I began to consider the possibility that this was a different religion, Umbanda. When the ritual participants began to gather, their modes of dress strongly suggested that it was a third religion, Quimbanda. This was confirmed by other aspects of the ritual. As far as core participants (as opposed to non-member clients of religious services) label their groups in this way, we might be led to reify these as distinct things called ‘religions’, granted their many overlapping features. This would be unhelpful. It would also be inconsistent with Davidson, whose ideas help make better sense of my shifting interpretations in this case. My prior theories were shaped by my interlocutor calling the group, which I had never visited, ‘Candomblé’. My initial observation, based on previous experience, led me to a passing theory: it was a terreiro (‘grounds’ or house) of Umbanda. Further observation led to a new passing theory: it was a terreiro of Quimbanda. Further observation confirmed the latter as effective (not ‘true’) for guiding interpretation. My attempts to make sense of what I was seeing took shape across a shifting set of hypotheses about religious and ritual context. The divide between prior and passing theory was fluid and dynamic, in large part because aspects of prior theory were about the context itself, about particular social externalities. My reading of the actions and words I was witnessing shifted as those assumptions were challenged. Passing theory revised my understanding of where I was and what I was seeing. But this passing theory (revisions proposed to make sense of what I had just seen) immediately became prior theory (contextualizing what I had yet to see). By the time I sat down after the main ritual to talk to Exu Sete Caveiras (‘seven-skulls Exu’) – the powerful, morally ambivalent spirit incorporated in the body of the leader of the group – my passing theory now constituted (hopefully) shared prior theory for our conversation.

When communication becomes difficult, when we fail to understand some statement or action during CI, where do we start revising? The first site of revision is our reading of the perceptual world, for example, checking to see if we understand correctly just what the person we are talking to is pointing at or referring to. We then move to revise prior theory, starting with the local end of the spectrum, that is, social externalities: if you are unable to understand what another person says at some point in a conversation, you will first question if you took something out of context, or if you yourself are unclear on the nature of that context, before you contemplate the possibilities that your interlocutor is speaking a different dialect than you thought or, ever at the last resort, that they are not even a rational being.

In cases of confusion or failed interpretation in standard, everyday contexts, we often question our assumptions regarding perceptual before those of social externalities. We generate hypotheses like the following: perhaps they are referring to bullet point five in the other document, not this one; perhaps they were talking to someone standing behind me and not to me when they said that; perhaps I was confused by their accent and they meant this different word; perhaps they are using this ambiguous word in its other sense. In ethnographic contexts, social externalities come to the fore. I have found myself asking more contextualized questions when words or actions leave me unexpectedly confused: perhaps the group’s leader was trained in or is borrowing from Kardecist Spiritism; perhaps these unexpected esoteric elements reflect affiliation with that other lineage in São Paulo; perhaps the prominent Catholic elements reflect the religious landscape of this particular city; perhaps this is a hybrid ritual variant that I have not seen before? Without more contextualized knowledge, such questions are unanswerable. In this light, the role of prior and passing theory in communicative interpretation leads to a nuanced characterization of ethnographic fieldwork: methods like participant observation and interviews lead to a more detailed sense of social externalities, which in turn result in more effective prior theory and more convergent passing theories, in cases where interpretation appears to break down.

At this point it might be useful to pause and reflect back on the difference between radical and communicative interpretation (RI and CI). In the former, the radical interpreter’s prior theory includes no historical or cultural knowledge: the third apex of the triangle consists only of the natural world. In CI, the social world becomes the primary dimension of externality. In the former case, the world is largely taken for granted; in the latter, it is itself the product of interpretation. The fact that a person is dressed in such-and-such a way serves as evidence about how they intend to be understood by their linguistic behaviour, in light of other sorts of knowledge, especially regarding the social or ritual context. I know many Brazilian umbandist mediums socially. If they are wearing street clothes and I run into them at a restaurant, this gives me an immediate cue that our conversation will be constrained only broadly, by normal social conventions and shared social context. If I run into them on the street near the Umbanda centro where they are a member, and they are wearing white ritual clothes, this tells me that I am talking to a medium, and the conversation will centre more around the topic of Umbanda. If, while attending a ritual in the centro, I am led to them while they are in a trance, then this context tells me that I am talking not to the medium but to a spirit: the conversation will be highly constrained as a result. My prior and passing theories – and my sense of theirs – will reflect my sense of relevant context.

In discussing the relation between prior and passing theory, the key issue is not the conditions under which interpretation succeeds: it is what communicants do when it fails. In pragmatic terms, prior theory is the set of beliefs that are potentially revisable in efforts to make sense of hiccups, false starts, or failures in CI. Questioning basic presuppositions – which Davidson discussed most – is extremely rare. Most work with passing theory takes place at the local level of social externalities, as we try to get a sense of each other’s social worlds.

In this sense, ethnographic fieldwork is an exercise in teaching and learning, as is all communication: you help me come to understand how you understand your social facts and norms. Social worlds are constructed and shared. One’s idiolect can be forged in part from elements of another’s. The teacher does not pour meanings into the learner’s head. The learner has the basic semantic norm of wanting to be understood, and their ongoing hypotheses of what the teacher takes their own words to mean contribute towards the realization of that intention. The learner’s norm is not ‘I should speak this way because I am a member of this linguistic community’. Nor is it ‘I should speak this way if I want to be a member of this linguistic community’. It is ‘I should speak this way if I want to be understood’. As the learner communicates with more and more people, they will have more and more teachers (and they themselves will act as teacher to other learners). In this Davidsonian light, a linguistic community is best understood not as a determinate group with a shared language, but as a vaguely overlapping group set of communicants who have already fine-tuned their prior theories to harmonize well together, with inevitable variation. In terms of fieldwork with religious groups, the outsider scholar, whose prior theory has not been pre-tuned, must look at as many people’s behaviour as possible over as wide a range of contexts as possible. This includes looking at people who seem to opt in to and out of particular practices: our understanding of what a typical member looks like emerges alongside our understanding of what they are members of. Our individual intentions are independent (hence, the primacy of the idiolect), but our meta-intention to be understood in context will tend to unite the individual intentions of communicants in particular ways that are describable as shared. This yields something much like what the social realist wants, without having to reify concepts like culture, tradition, religion, and so on.

The move from prior to passing theory – in the dynamic sense described above – offers a philosophically robust theorization of the work of learning that is central to ethnographic fieldwork and, by extension, the study of religions. We understand religions by interpreting the intentional behaviour – spoken, written, ritualized, and so on – of those who we take to be members of a religious community. We fine-tune our understanding by revising our initial assumptions (prior theory) into more convergent hypotheses (passing theories) about the meaning of what we observe and experience. Most of this work takes place not at the most immediate level of reference to the physical world nor at the most general level of basic presuppositions of rationality, intentionality and constraints on meaning itself (Davidson’s basic framework). It takes place at the local level of making sense of others’ social worlds. The following section further clarifies this work using the concept of ‘semantic reduction’, which we propose as a tool for conceptualizing this work.

Semantic reduction

In RI, communicants look to perceptual externalities as they revise prior to passing theory; in CI they look primarily to social externalities. As each person attempts to make sense of the intentional behaviour of the other (their words and actions), they are in effect trying to learn about a different socially constructed context. We propose the term ‘semantic reduction’ to highlight the relevant methodological factors.¹⁴ This supplements Davidson’s partially developed ideas about social contexts of interpretation.

At base, a semantic reduction is an imposed limitation on the form that communication can take in a particular context. For example, when a doctor asks a patient how they are feeling, the doctor’s interest in the answer is limited to the condition the patient seeks treatment for; the patient implicitly knows this, and guides their response appropriately. This is not a matter of the doctor/patient interaction being governed by a certain ritual form: rather, scripted conversations, including medical consultations and ritual discourse, involve semantic reduction. This specifies the most semantically relevant dimension of the scriptedness of much communication.¹⁵

The metaphor of the holistic web of meaning clarifies what we mean by semantic reduction. The meaning of a given term, expression, performance, and so on, is always tied to the meaning of others. The interconnections form a web. In principle, all parts of the web are traceable to every other, but in practice certain connections are more salient than others. In normal cases of interpretation, we understand one node by tracing it, either explicitly or implicitly, to others that are nearby. ‘Nearness’ is a relative and contingent notion, involving such factors as two nodes sharing connections to other nodes and co-occurrence in a speaker’s past discourse. A semantic reduction involves a particular focus on a particular region: limiting semantic salience (to a relative extent) to a delimited set of local semantic nodes, and decreasing or ignoring the salience of more distant semantic nodes. We argue that, at least in the case of religious phenomena, those reductions are normatively imposed, and themselves play a constitutive role in understanding their meanings. One can step outside of the script to some extent when talking to one’s doctor without clearly stepping out of the patient-doctor relationship, but doing so in a ritual takes one outside of it altogether. In other words, we argue that a theory of a ritual is tantamount to a proposal about a specific form of semantic reduction.

If the full meaningfulness of an expression is given by its placement in a complete web, then semantic reductions curtail its meaning. Expressions in less constrained contexts are more meaningful than those in semantically reduced contexts. The distinction is relative, because, we suggest, every actual interpretative encounter imposes, either explicitly or implicitly, some degree of reduction. For example, the range of inferences that can be drawn from expressions in semantically reduced contexts will be less. Part of what it is to understand a particular ritual, we argue, is to understand the particular limitations to the meaningfulness of words and actions within it. Nonetheless, there are limits to how far the reductions can go: they cannot go so far as to reduce them to mere noises or bodily movements.¹⁶ Where that line lies can only be pragmatically determined by reference to the contextual goals of particular encounters.

‘Semantic reduction’ points to ways in which social externalities (social, cultural, religious, and ritual contexts) require more frequent recourse to passing theories that is the case with perceptual externalities (the natural world). In RI, the interpreter assumes that they perceive and refer to the same natural world as their interlocutor, and part of the move from prior to passing theory involves recognizing that aspects of this perceptual context are relevant – and other aspects not relevant – for making sense of the interlocutors’ intentional behaviour. The ‘world’ as seen through the pragmatic lens of the RI encounter is a selectively reduced model of the hypothesized world-in-toto: each communicant must learn what aspects of perceptual externalities are relevant for interpretation. Each speaker/interpreter must learn, in a relative and relational sense, what world is relevant for making sense of the other: to interpret is, in part to learn about others’ worlds; and passing theory is convergence between worlds.¹⁷

A greater degree of convergence in prior theory can be expected with respect to perceptual than with social externalities. That is because, in part, semantic reductions are part of the material from which social externalities are constructed. The more obvious socially constructed nature of social, as opposed to natural, world results in a generally greater need to shift from prior to passing theory. The natural world is socially constructed as well, in the sense that we have no access to this world outside of language: the shared sense of that world (central to RI but less so in CI) is also subject to the interpretive process. However, social externalities are more likely to vary between interpretive context, especially while doing ethnography.

Interpretation involves learning about other people’s social contexts, about frameworks of significance in which they have been trained, educated, enculturated and habituated over a long period of time. Words like ‘culture’ and ‘religion’ point to these social externalities – granted that we reject their reification. The ethnographic encounter is, in this sense, a model for all CI. It underlines the fact that interlocutors, to greater and lesser extents, swim like a fish in the relatively distinct waters of their own languages, cultures, traditions, religions, and so on – where these concepts refer to contingent groupings of semantic networks with relatively greater overlap. The concept of semantic reduction points to a methodological aspect of the CI encounter: a specific challenge for the generation of passing theories. In the case of religious ethnography, specific, local social externalities provide clues that help keep interpretation on track: ritual, doctrinal, narrative, symbolic, material, temporal, and spatial frames constrain the generation of passing theory as we try to interpret those we observe and talk to.

In the ritualized spirit incorporation that is central to most Afro-Brazilian religions, the scholar faces the interpretive challenge of distinguishing between mediums and spirits (or other embodied entities) as interlocutors. ‘Semantic reduction’ points to the use of contextual cues that allow members, as well as ethnographers, to distinguish the more semantically constrained contexts of ritualized interactions from the more open contexts of everyday conversation. For example, the interpretative possibilities of making sense of what a caboclo (Indigenous) spirit says and does in an Umbanda ritual are constrained by umbandist doctrine (nature, types, roles of spirits), by ritual scripts (rituals begin with chants to spirits, following by incorporation in mediums, followed by structured one-on-one consultations between ‘clients’ and spirits), by the physical layout of the space (with spirits manifest close to the altar, spatially opposed to the seating space of the clients), by indexical ritual cues (distinct types of spirits are recognizable by characteristic accents, discourse, noises, gestures, postures, artefacts etc.),¹⁸ and by the formalized nature of the ritualized ‘consultation’ itself. Conversations with spirits are limited in a way that reflects their ritualized context. This is strictly comparable to the constrained conversations that we have with doctors, as they solicit our medical history and perform examinations: ‘any heart problems in your family?’; ‘open and say “ah”’. It is easier to interpret when we know the script; children learn this skill as they grow older. Introductory superficial chit-chat aside – a social ritual that has its own scripted dimension – doctors’ words and actions are reduced by their functional role, as are patients’ responses. Similarly, instructors are semantically reduced in the classroom, to a lesser extent than doctors, but for comparable professional reasons.

Context and roles perform a semantic reduction in many of our daily interactions. There is no sharp divide between more and less reduced contexts. We can slide between along the scale in a single conversation. For example, in learning about First Nations traditions in Western Canada, I have had several conversations with Cree and Blackfoot Elders. Sometimes, in the middle of a normal, respectful, informative conversation, I have set a pouch of tobacco on the table between us. This indicates that I am about to make a formal request, to an Elder in their capacity as Elder. The conversation shifts into a more ritualized mode: the range of topics narrows; pauses before the Elder’s responses grow longer, their content more ponderous; the hierarchical relation between knowledgeable Elder and knowledge-seeking petitioner is now more clearly marked by tone and posture. The context of communicative interpretation is reduced. Semantic reduction is part of normal CI, but it is more marked in the context of ritual; we propose that this is an important aspect of ritualization.

In the ethnography of religion, semantic reduction manifests as a sliding scale indexed by doctrinal, ritual, and other frames. There can be a reduction within or beyond an initial reduction. Umbandist spirits are semantically reduced echoes of people. Conversations with spirits in rituals are more semantically reduced than semi-structured interviews with mediums; and the latter are more reduced than casual conversations with mediums in social contexts, over a coffee or beer. Neo-Pentecostal versions of umbandist spirits are further reduced. In Neo-Pentecostal churches allegedly the same Afro-Brazilian spirits, called by the same names, are categorized as demons (Isaia and Silva 2019; Macedo 2001/1997). As mid-century healing Pentecostalism began to shift into combative Neo-Pentecostalism in the 1960s and 1970s, these spirits were occasionally seen as demonic monsters (Engler 2011). Canadian Pentecostal minister, Robert McAlister (founder of the New Life Christian Church in 1968) wrote, ‘I began to face this monstrosity considered only folk belief and I began to use the Name of Jesus in prayer to free the oppressed from these “caboclos” and “orixás” [incorporating Afro-Brazilian entities] who are nothing more than evil and demonic spirits’ (Mcalister 1968, 12). The process of interpreting these ‘same’ beings in a different religious context leads to a more reduced network of context-specific associations.

‘Semantic reduction’ refers to this doctrinal and ritual interpretive shift, to the narrowing of semantic connections that emerge as we try to interpret certain types of speech and action in certain contexts. In Umbanda, spiritually evolved, charitable spirits of a wide range of types incorporate in mediums as individualized spirits, each with its own history and personality, and the spirits conduct healing and consultation rituals of various sorts. In Neo-Pentecostalism, a generic malevolent demon, sometimes unnamed, is exorcised; and, according to Neo-Pentecostal theology, these are the same entities. There is less meaning to be found in the caboclos of neo-Pentecostal ritual than in the caboclos of umbandist ritual. This is in part because only pastors speak to the spirits in Neo-Pentecostal rituals, as they exorcise them in a highly scripted ritual encounter; in many Afro-Brazilian religions, we can talk to the spirits ourselves, and we talk about our problems, our worries, our goals, our lives, all within a constraining ritual frame.

The role of semantic reduction can be illustrated by shifting from a semantic to an epistemological register. Truth-conditions, essential to a Davidsonian semantics, can be read as relative to doctrinal, narrative, and ritual frames. This is clearest in rituals, where the denotation or reference of a given word depends on whether we look at it from inside or outside the ritual-frame:

frames shift denotations in a manner that swings free from issues of reference.. That is, we get a more nuanced view of the frame when we recognize that the issue of what is true is relative (i.e., it depends on whether one is talking about inside or outside the frame). ‘This is the body of Christ’, said by the Catholic priest holding up the wafer, is true as long as the denotations of the terms ‘this’ and ‘the body of Christ’ coincide. Outside of the ritual-frame – in the ‘normal’ context – ‘this’ denotes a thin piece of bread and ‘the body of Christ’ denotes a body which would bleed if cut, and those things are simply not the same; i.e., this claim is false (viewed from outside the ritual-frame, from with the ‘normal’ context). ‘Real’ or ‘really’ are terms whose denotations are similarly affected depending on whether they are in or out of a given frame: viewed within the ritual-frame, this (i.e., the object held by the priest) reallyis the Body of Christ (Engler and Gardiner 2012, 12).¹⁹

The role of semantic reduction is clearest when we look not at when CI succeeds but when it fails. In one terreiro of Quimbanda (a ‘left-hand’ tradition often seen as closely related to Umbanda), I was called up to dance with the powerful, transgressive, and sensuous female Pombagira spirit (as described above) that was incorporated in the Mãe de Santo, the human leader of the house. She insisted that I insert money into her bra, which I recognized as a normal and scripted portion of rituals with this spirit in this house (prior theory). Then the Pomba Gira disincorporated and a powerful morally ambivalent, dangerous/protecting male Exu spirit incorporated in the medium: this is less usual, but happens on occasion (prior theory). The Exu is the most powerful spirit and spiritual leader of the house. A cambone (ritual assistant) rushed up to give the Exu his hat. The Exu then spoke to honour me and placed the hat on my head as a gift. At that point, semantic reduction broke down. In terms of the triangular shape of interpretation, my attempt to interpret the actions of my interlocutor (the Exu spirit incorporated in the Mãe de Santo) ran into incongruity, relative to the expected ritual form (social externalities). In my experience and to my knowledge, this special attention given to a visiting non-member was unprecedented. It did not fit with prior theory, and I was unable to come up with a passing theory without including factors beyond the ritual frame, which would normally constrain the generation of passing theory. While the hat was put on my head and the Exu spoke, I thought of two hypotheses. Perhaps the group was honoured by the attention of a foreign researcher. In informal conversations, the Mãe de Santo had made this point. She underlined that this religious community is triply marginalized: stigmatized in religious terms as non-Christian/African, in social terms as consisting largely of gay men and trans-women, and in moral terms as being the spiritual home for many sex-workers. My legitimizing presence may have invoked an unusual form of ritualized reciprocity. Alternatively, perhaps the human leader of the group (or the Exu?) was acknowledging a more material dimension of reciprocity. Shortly before this ritual, I had inferred that one of my co-researchers was passing on research funds directly to this community. This was later confirmed when the Mãe de Santo thanked me, outside the ritual context, for my generosity. The hat now sits in a place of honour on a special shelf in my office back in Canada. But my sense of what the hat means in religious terms remains limited by the fact that my passing theories extend beyond the semantically reduced ritual and doctrinal contexts: I remain unsure about whether it has ritual significance within the context of Quimbanda, social significance as acknowledging and strengthening bonds between researcher and community, or both, in some sense that remains unclear to me.

Here is another example. One night, during the lead-up to the 2018 Brazilian Presidential election, a cambone in an esoteric centre of Umbanda, very unusually, pulled me gently out of the line of clients who were waiting to speak to the caboclos. She explained that she was waiting to take me to a spirit that I would more easily understand. (Caboclo spirits have strong accents, Portuguese not being their mother tongue.) I was led to the powerful caboclo who incorporates in the senior assistant to the leader of the house. The spirit asked if I was having any problems. I said I was ‘worried about the election’. He replied, ‘not just the American one’. Both cambone and spirit demonstrated knowledge of my being non-Brazilian. The spirit also appeared to be up to date on domestic and international politics. My attempt to make sense of this unexpected shift from healing discourse to current events led me to wonder if I was speaking to the medium or the spirit. Faced with a statement that did not easily make sense within the ritual frame, I moved automatically to extend the network of associations beyond the semantically reduced context of the ritual, treating the conversation more like a ‘normal’ one. Then the script reasserted itself. We talked about clearing negative energies, and the gestures of the passe re-established the customary corporeal rhythm. The spirit and I were back on track: our conversation had returned to the semantically reduced form that is normal within the ritual frame.

This points to a normative dimension of semantic reduction. Attempts to interpret spirits soon encounter a pre-set frames of doctrine, ritual forms, stories, traditions, embodied dispositions, material culture, and so on, and these are methodologically productive resources for communicative interpretation in that context. They constitute a form of prior theory, and they orient passing theory. Semantic reduction distinguishes this interpretive context from the wider world, by defining what is normal ‘here’. We are discouraged from seeking meaning outside the ritual frame. They lead us to pre-established, semantically reduced networks of meaning. They force and delimit the expanding exploration of semantic networks. They constrain the very process of interpretation that produces meaning.

Conclusion and final observations

Ritual behaviour is behaviour in a ritual context. All behaviour is meaningful, and meaningful in the same interpretationist way. Context always contributes to interpretation, but ‘ritual context’ is not a special, privileged, or sui generis type of context, demanding its own special method of interpretation. Rather, participants and scholars will identify a context as ritualistic if they find that interpretatively useful, and we have suggested how prior/passing theory, pragmatically guided by learnable semantic reductions, should play a role in those identifications. In particular, contexts identified as ritualistic tend to be correlated with those in which prior theories – those that include particular sorts of semantic reductions more efficiently and effectively – lead to convergent passing theories. A general theory of ritual, which we neither provide nor endorse, could be thought of as prior theory, that is, as a proposal of the contours of semantic reductions most operative in those contexts. A particular theory of a particular ritual can likewise be a proposal of the particular semantic reduction operative in a more limited context. Such proposals are empirically evidenced, and can ground predictions, in precisely the way established in CI, but they are never the last word nor can they be blindly applied ethnographically.

We end with three observations that we do not have space to develop further. First, ‘insiders’ are people who have internalized semantic reductions in particular contexts. Those reductions are grounded by the contents of their actual propositional attitudes, largely shared. Their prior, and where required passing, theories, are already close. ‘Outsiders’ are those in the position of trying to figure what those reductions are. This is inseparable from trying to understand the propositional attitudes of their interlocutors (members), and this is inseparable from trying to make sense of their intentional behaviour. The difference is a relative one tied to actual occasions of interpretation. As the above examples of CI failure in ritual contexts indicate, when an insider starts acting in unexpected ways, diverging from their co-religionists, outsiders will find themselves generating more radical passing theory, to the extent that they have already internalizedat least some of the relevant semantic reductions.

Second, this suggests a methodological strategy for the ethnographer. In addition to attempting to act as much as possible like insiders during participant observation, a strategy of diverging from this emerging model in a controlled way can contribute to getting a sense of relevant semantic reduction(s) by testing their limits. The first strategy seeks the norm and the second the bounds. Davidson underlined the role of prompting assent/dissent in RI (1983/2001) 147). This helps us find the contours of semantic reductions, and it is supplemented by the related issue of prompting response/non-response. One example is my mentioning that I was ‘worried about the election’, in the case described above. Another example is that, during rituals, I sometimes ask a spirit for their name and biographical details. This is outside the doctrinal and ritual frame, but spirits sometimes answer the question. One cabocla gave her name but added that she was relatively new and that ‘the head of the house’ did not want neophyte members sharing personal details. This gave me insight into the formalization of semantic reduction in this particular group. Most often, I am ignored when I stray from a mutual discourse of wellness, energies, and healing. The spirits perform their customary healing gestures as if my discursive divergence went unnoticed. To the extent that semantic reduction has a normative dimension, the method here is to investigate what constitutes semantic transgression. In terms of research ethics, this is defensible because people new to umbandist rituals frequently show their ignorance through precisely these sorts of transgressions: it is part of learning how to act correctly in the religion. In Davidsonian terms, we speculate that dissent/dissent and response/non-response are part of the process of teaching and learning that is central to all CI.

Third, to extend this methodological point to a theoretical level, religious actors are, in a sense, semantically obligated to keep their usages, beliefs, movements, and so on, stable over time. Semantic reduction – reflecting established doctrine, accepted tradition, and ritual scripts – is stable in ways that facilitate interpretation by outsiders, and it is also acts as a sort of semantic inertia. This is crucial to the role of education and training, as found in weekly study sessions and the long apprenticeship of cambones. This stabilizing aspect of semantic reduction is common to most communicative contexts – it is a feature of social externalities as characterized by Davidson – and religion sit at the more static end of the spectrum. In this light, Davidsonian semantic theory, as we begin to apply it here, offers promise for theorizing religious and cultural dynamics: continuity and change, stasis and fluidity, tradition and innovation, and so on. This final observation underlines our guiding meta-methodological, metatheoretical, and disciplinary claims. Davidsonian semantic theory – by shifting emphasis from representation and reference to the conditions of successful interpretation – has enormous potential value for the study.