Whose Problem Is This? Stakes, Askability, and the Difference Between Good and Bad Ethnographic Questions

Abstract

This article develops a stakes-first method for distinguishing good ethnographic questions from bad ones. It begins with two questions from fieldwork on Bengali Hindu worship. The first, “Does Ma Kali actually eat the food offered to her?”, produced divergent and sometimes irritated answers. In retrospect, the question imported an external ontology organized around matter and spirit, literal and symbolic meaning, and transactions between separate domains. None of these distinctions captured what was actually at stake for worshippers, who disagreed about divine consumption but agreed on the ritual sequence through which food was offered, protected, transformed, and redistributed. The second question, “How can I know my iṣṭa-devatā?”, opened immediately onto locally consequential concerns: divine intensity, ritual competence, personal readiness, safety, and the dangers of approaching a deity improperly. Drawing on Living Value Theory, the article argues that good questions disclose stakes already active within a mesocosm, whereas bad questions impose stabilized scholarly problems whose stakes belong elsewhere. Stakes are what anchor felt misalignments in living coordination; they explain why some disturbances demand articulation while others remain inconsequential. The article extends this argument through Axel Michaels’ question of what “holds Hinduism together,” showing how an externally generated problem of religious coherence produces the concept of “identificatory habitus.” It then traces the text/practice problem to Pauline Christianity, where the relation between authoritative scripture and transformed ritual practice genuinely carried salvific and communal stakes. The article concludes by proposing mesocosmography as stakes-first inquiry: before interpreting beliefs, texts, categories, or practices, researchers should ask what can go wrong, what must continue, and whose problem their question actually addresses.

I. What Matters

I have been rereading my fieldwork notes on popular Hindu practices: transcripts of conversations recorded in Kolkata, with Bengali Hindus, on the everyday life of goddess worship, illness, food, and obligation. Instead of focusing on people’s answers, I was curious which questions worked well, and which did not. Among the hundreds of questions I asked across those fieldwork years, two are selected here because of how differently they reveal what a good question is and what a bad question is.

One of the questions I asked, in one form or another, in temple courtyards, in cramped front rooms, in queues outside Kalighat, was about what happens to the food that devotees offer to the deity. A vital part of the worship of Hindu deities is the offering of food. In most cases, this food is strictly vegetarian: rice, chutney, dal, vegetables. In Kalighat and other Kali temples, it includes non-vegetarian items such as mutton (sacrificially killed on the temple premises), fish, and prawns. The idea is that the deity eats the food offered (called bhog). The “leftovers” of the food offering are then given to the devotees and they eat it. This is food as prasad. By eating the prasad, the devotees partake in the divine energy and substance. Some of the prasad is given to the sacrificer who sponsored the food offering, some of it is redistributed by the temple pundits to others, either for free (for beggars) or for a token fee to other devotees. My question, then, was quite straightforward: does Ma Kali actually eat the food that is offered to her?

In hindsight this was not a good question. It imported a Western concern about people’s religious “beliefs” as opposed to “real” processes. If you offered food to a human for a meal and they took the meal, you wouldn’t wonder if they “actually” ate the food. It imported a mind/matter distinction that none of my interlocutors held. It assumed a transactional ontology in which devotee and deity are separate parties exchanging a good, food handed across a boundary from one domain to another. But bhog/prasad cannot be understood this way. There are not two domains with a transfer between them. There is one continuous field of mediation, intensifying, returning, redistributing.

It’s not that people would have refused to answer the question, or question the question. Everyone I asked gave an answer. Some said yes, definitely she eats, with the power of her eyes rather than her mouth, niścaẏ māẏer cokher dṛṣṭi diẏe khāoẏātā khẏe nāẏ, she eats the food with the gaze of her eyes. Others said no, of course not, a mud statue cannot eat, the offering exists only for moner śāntanā, mental satisfaction, and the closed door or drawn curtain at the moment when the deity is eating is a courtesy rather than a necessity. One man told me, with some impatience, that if Ma Kali really ate, she would be a family member, and the very extravagance of believing it literally, thousands of plates, lakhs of devotees, every single day, shows that it cannot be true in that sense. Another, a regular in the queue at Kalighat, insisted that anyone who claimed to have personally witnessed her eating was simply lying: nobody has actually seen it, so how could anyone know?

What struck me, rereading these notes, are not the disagreements. It is what nobody disagreed about. Every person I spoke with, whatever they said about the eating itself, agreed that the door must be closed, that it must stay closed for some time to let the goddess eat without disturbance, that nobody may look inside during that time, that a brahmin must be the first to re-enter, that the food which re-emerges has changed, however little its quantity has changed, and that the sequence, offer, close, wait, open, distribute, may not be skipped or reversed without risking the relationship the whole procedure exists to maintain. The dispute over whether Kali eats turned out to be a dispute about almost nothing that mattered to the people having it. The procedure mattered to all of them, regardless of which position they took on the question if the goddess “actually” eats. What struck me also is that no one challenged my question. They should have said: this question makes no sense from within our world. If Ma Kali eats or not is not at stake.

The second question I asked, rarely as a direct question at all, was about how to know who your inner deity is. It surfaced in informal conversations, almost as an aside: how can I, or anyone, know their, their inner god, or personally most cherished form of the divine? One man explained that you cannot see God directly, that you need a devatā, a particular relation, the way Hanuman had Rama, and that finding that relation safely is something that has to be worked out, not assumed. The iṣṭa-devatā question concerns Hindus and non-Hindus alike. Hinduism does not proselytise but some streams allow entry to those not born Hindu, largely based on the principle that All is One, and that different forms of religion are just different in appearance. “All paths lead to the same pond” is the most common line in Kolkata. Unlike the Kali eating question, this one had real stakes for everyone present. Everyone had a view on how you can identify your inner deity and commence a closer relationship with them. In this context, everyone told me that I, as someone who came from a Christian world and would return to a Christian world, should never dare to worship Ma Kali as iṣṭa-devatā. She was too dangerous for anyone for without the habituation, the caution, the ritual know-how to bear what closeness to her demands. Krishna and Shiva were both mentioned as easier, more forgiving devatas, content with a little puja and unlikely to punish a small mistake, while Ma Kali, several people insisted, would not forgive carelessness so lightly. If Ma Kali becomes someone’s iṣṭa-devatā or not is greatly at stake.

This article takes these two questions and asks what the difference between them reveals about method. The argument can be stated in a single sentence. Figure out the stakes of a question, and you will find the ontology of the mesocosm it was asked into. The question about whether Kali eats does not open onto any stake that belongs to the mesocosm of her worship. It opens onto stakes that belong to a different problem, carried in by the person asking it. The question about the iṣṭa-devatā opens directly onto what is actually at stake for the people answering it: safety, appropriateness, intensity, the management of a relationship that, handled wrongly, has real consequences. The rest of this article works out what makes one of these a good question and the other a bad one, and why getting that distinction right matters well beyond the study of Bengali Hindu practice.

II. What's the Problem and Whose Problem Is This?

Most research on religion, and arguably most social and historical research on any topic, begins in the same place: with a concept. The researcher arrives already equipped with a question that has a settled shape, a question that already distinguishes between possible answers in a way some prior discipline, debate, or tradition has found useful. Does the deity really eat, or does the offering merely symbolize something else? What unifies a tradition's diverse beliefs and practices into something nameable as a single religion? Is this text descriptive or prescriptive? These are not neutral questions waiting on the facts to settle them. They are themselves a kind of fact: stabilized classifications, already formed, that the researcher carries into the field and applies to whatever is found there.

Living Value Theory calls this stabilized classification at fourth-level recursivity: a category, doctrine, binary, or theoretical concept that has hardened, out of a long prior history of use, into something that can be picked up, taught, cited, and applied across very different fields without further justification. “Does Kali eat the food? Yes/No?” is an question in exactly this sense. It already contains a finished ontological architecture, a domain of matter and a domain of spirit, a literal pole and a symbolic pole, a transaction between two separate parties. The fieldworker does not invent this architecture in the moment of asking. It arrives ready-made, the residue of a much older history, one this article returns to later through the case of Paul and the New Testament.

The L4 question is then used to interrogate L3 material: the things people actually say, in interviews, in casual remarks overheard during participant observation, in the explanations volunteered when a stranger with a notebook asks why a door is closed. L3, in this framework, names the level at which felt experience becomes articulable, sayable, offerable as an account to another person. It is the level testimony and narration are made of, the level transcripts capture.

The persistent mistake, repeated across a great deal of careful and well-intentioned scholarship, is to treat these L3 statements as exemplifications of the L4 question that prompted them: as so many data points confirming, complicating, or contradicting the imported category. When a Bengali interlocutor says that Kali eats with her eyes and not her mouth, this gets filed as an instance of literal and symbolic ambiguity in popular belief. When another insists that no statue can eat and that the closed door is mere convention, this gets filed as evidence of generational change or rationalization within the same belief system. Both responses are treated as answers to the question the researcher asked.

But this is precisely what gets lost. The statement she eats with her eyes is not an exemplification of a prior metaphysical category about literal versus symbolic transaction. It is an articulation of an L2 felt misalignment: a small disturbance, a thing that needed saying, that arose because something in the local mesocosm required tending. L2 is the pre-conceptual register at which coordination falters, just enough to need attention: not yet a sentence, not yet a position, simply a felt pressure that something matters here and might go wrong. The articulation that follows is shaped by the question asked, certainly, but it answers to a misalignment that has nothing to do with that question's own architecture. It answers to whatever, in this particular informant's experience of puja, has actually felt uncertain, threatened, contested, or worth defending.

And the L2 misalignment is itself not free floating. It registers as a misalignment, rather than as simple noise, because of something at stake in the L1 world beneath it: the absorbed, unreflective, ordinarily seamless coordination of daily life, in which puja simply gets done, food simply gets offered, the door simply gets closed, without anyone needing to settle a metaphysical question about literal eating at all. L1 is where life happens before it needs to be defended, explained, or argued over. What gives an L2 disturbance its weight, the reason it demands articulation rather than passing unnoticed, is what is at stake in the L1 coordination it threatens. This anchoring of L2 in L1, the fact that a felt misalignment is felt only because something that matters is under pressure, is what this article calls stakes. Stakes are not a fifth level alongside L1 through L4. They are the name for the relation between L1 and L2: what makes a disturbance register as a disturbance at all.

This gives the method its central diagnostic question, which can be put more simply than any of the foregoing: what is the problem, and whose problem is it? A question asked in fieldwork, an interview prompt, a note made to oneself during participant observation, is itself an L3 articulation, produced by the researcher, designed to elicit either a further L3 articulation or, far more often, an L4 articulation, a classification, from the person being asked. A good question is one that is aware, in its very construction, that what actually matters is L2 and, through L2, L1: it is built to surface the respondent's own felt stakes rather than to harvest confirmation or complication of a category the researcher already holds. A bad question is an L4, an already stabilized classification, imported wholesale by the researcher, that then overrides the respondent's own L3, L2, and L1 alike, forcing an answer to be produced within a frame that has nothing to do with what the respondent actually has at stake.

Compare this to the ethnoindological works by Axel Michaels. In terms of detailed, mesocosmically faithful L3 descriptions of Hindu concepts and rituals, Michaels is second to none. His Hinduism: Past and Present (first edition 2003) is an indispensable reference. But in terms of his L4 conceptual work, it is far less compelling. One of Michael’s key questions is: “What holds Hinduism together despite the absence of a canonical text?” His answer to this question is: Hinduism is an “identificatory habitus.” This is an L4 stabilization that imports a whole range of problematic assumptions, such as that a religion normally needs a stable identity, or that religious unity is normally based on a canonical text and Hinduism “lacks” such a text. But nobody inside the mesocosm that I encountered in Kolkata has ever been troubled by this question. It’s never been at stake for anyone I talked to. I’m sure there were and are Hindus who are bothered by this question, but no one I ever met was. In their world of Kali worship, Durga puja, household devotion, or relations to an iṣṭa-devatā, the question whether their tradition coheres as a nameable, classifiable system comparable to Christianity or Islam just never came up. That is a problem for comparative religion, for a discipline that needs traditions to be sortable into a shared taxonomy. It does not register anywhere in the L1 coordination of a Bengali household doing its evening puja. That is precisely why I argue that is so important to ask, always, not just “what’s the problem?” but also “whose problem is this?”, before asking asking any question in the field. If “Hinduism” is held together by an “identificatory habitus” turns out to be Michael’s question, not theirs. And if it’s not theirs, it’s not a good question.

III. The Bad Question: Does Ma Kali Eat the Food?

Returning to the transcripts with the distinction between L4 import and L2 stake in mind changes what the disagreement over Kali's eating looks like. Read as answers to my question, the transcripts are a mess: contradictory claims about an empirical matter, whether the food's quantity decreases, competing theories of perception, she eats with her eyes and not her mouth, generational splits, older relatives believe it literally and the young generation does not, open scepticism, a mud statue obviously cannot eat, the offering exists only for mental satisfaction, and outright accusations of dishonesty against anyone claiming direct evidence either way. Read as evidence about the question I actually asked, none of this resolves into a coherent position, and a researcher committed to extracting one is forced either into condescension, these are contradictions popular religion has not yet worked through, or into a story about belief decline across generations that the material only partly supports.

Read instead as L3 articulations of L2 felt misalignments, the same material becomes orderly almost immediately, because what generates each statement is not a settled metaphysics but a specific, locally felt pressure that my question happened to touch. The man who told me that the arrangement would be too expensive if a deity literally ate, fish, rice, sweets, for thousands of devotees every single day, was not offering a theory of divine ontology. He was protecting a felt sense of economic and practical proportion: an arrangement this costly could not survive if it required literal consumption, so whatever is happening must be something other than consumption, full stop. The man at Kalighat who called other people's claims to have witnessed Kali eating lies, I haven't seen it with my own eyes, how can I know, was not contesting cosmology. He was defending a standard of personal testimony, the same standard he would apply to any claim about an event he had not personally witnessed.

One woman, asked directly whether the quantity of food shrinks after the door reopens, answered with an anecdote rather than a proposition. Suppose, she said, I offer five sandesh to Ma Kali. I close the door. I open it again, and there are still five sandesh there. She cannot have come and eaten them from the plate; we just think she has eaten. This is a remarkably careful piece of reasoning, offered without hesitation by someone who would never have described herself as doing theology. It distinguishes cleanly between what is materially observable, five sandesh remain five sandesh, and what the practice nonetheless accomplishes, we think, we believe, that she has eaten, without treating the gap between the two as troubling. Nothing here registers as an L2 misalignment requiring resolution. The misalignment, if there is one, belongs entirely to the question, not to the practice it interrupted.

Other interlocutors resolved the same gap differently, by declining to claim the standing to know at all. We are sinners, papi, one man told me, we cannot see what Ma Kali is doing; she has great power and can do as she wishes, and our part is simply to believe. This is not evasiveness. It is a coherent way of holding open a question that the practice itself does not require anyone to close, by locating the limit on the side of the worshipper's own capacity rather than on the side of the goddess's reality.

The clearest sign that the question does not belong to this mesocosm is the resentment it sometimes produced. One older man, having answered patiently for some time, eventually turned the interview back on me. People come, he said, with their research, their interviews, and in the end they insult God, make a fool of God. If you understand even half of what I understand, ask your questions, and I will enjoy answering them as much as you enjoy asking them. If you understand nothing of what you are asking about, what is the point? This is not the reaction of someone whose ontology has been productively probed. It is the reaction of someone who has correctly identified that the question on offer answers to a problem that is not his, a problem belonging to the symbolic apparatus of the person asking it, dressed up as an inquiry into his own world.

What every one of these responses shares, regardless of where each speaker landed on the literal eating question, is unwavering agreement about procedure: the door closes, nobody watches, the interval lasts roughly half an hour, a brahmin enters first, what comes back out is treated differently from what went in whether or not its mass has changed. This is where the real stake sits, in the maintenance of an unbroken sequence whose interruption, as a later section of this article will show through a single vivid case, produces consequences that are immediate, embodied, and entirely real to everyone involved. The metaphysical question of whether eating occurs sits, for the people I spoke with, almost entirely outside that sequence's actual stakes. It is a question my own training equipped me to ask and theirs gave them no particular reason to have settled, because nothing in their L1 world depended on settling it.

IV. The Good Question: How Can I Know My Iṣṭa-Devatā?

The iṣṭa-devatā question rarely surfaced as a direct exchange in my interview transcripts in the way the Kali question did, because nobody needed a researcher to ask it. It was already alive in ordinary conversation, in the stories people told about their own devotional history, in the cautions older relatives gave to younger ones, in small disputes about which deity suited which kind of person. Where the Kali question produced hedging, contradiction, and occasional irritation, the iṣṭa-devatā question produced fluent, confident, practically reasoned answers, because it was a question the mesocosm was already asking itself.

One man explained the underlying structure plainly: you cannot reach God directly, only through a devatā, the way Hanuman approached the divine only through unwavering devotion to Rama. Everyone, he said, needs one devatā, not because the other gods and goddesses are false or unworthy, but because a relationship with the divine, like a relationship with a parent, has to be particular before it can be anything at all. You have one mother and one father, even though everyone else also has a mother and a father; what makes yours yours is the relationship, not the category. The same logic, he insisted, applies to choosing whom among the gods to approach.

That choice, once raised, immediately became a question of safety and appropriateness rather than doctrine. Several of the people I spoke with were explicit that Kali requires a level of ritual competence and personal readiness that not everyone has, and that approaching her carelessly carries real risk. She is dangerous, one man told me; if anything goes wrong in her puja, if a mantra is mispronounced, if a single procedural step is skipped, the consequence will not be minor. Other deities are more forgiving. Krishna and Shiva were both named, by more than one person, as gods who are happy with a little puja and will not destroy a household over a mistake the way Kali might. One description of Krishna, offered with visible affection, had him relaxed and easygoing, playing his flute, untroubled by his devotees' errors, simply forgiving them the way a kind employer forgives an apprentice's clumsiness. Kali, by contrast, was repeatedly described as a goddess approached only once a person is ready to bear the intensity of what closeness to her demands.

This is exactly the kind of practical reasoning that the Kali eating question never produced. Nobody hedged about whether Krishna's forgiveness was real or merely symbolic. Nobody worried that describing Kali as dangerous imported an illegitimate category. The reasoning moved fluently between divine temperament, ritual competence, personal history, and consequence, because every term in it answered to something genuinely at stake for the person doing the reasoning: which relationship to take on, how much risk to accept, what kind of preparation a given intensity of divine closeness demands.

A Christian convert I spoke with answered the iṣṭa-devatā question without any of the hesitation that attended the Kali question. He did not need to investigate or choose. His relationship was already established, with Jesus, already safe, already settled by his conversion, so the question did not require fresh deliberation. This too is informative. It shows that the question is not a doctrinal puzzle awaiting a single correct resolution, applicable identically to everyone. It is a live, practically oriented question about an actual, ongoing relationship, one that different people answer differently depending on where they already stand, the way a question about which doctor one trusts, or which parent one turns to with a particular problem, has a real and variable answer rather than a single correct one.

None of this required me to ask the question the way I asked the Kali question, directly, with a notebook, expecting a proposition in reply. It surfaced because it was already part of how people narrated their own caution, their own sense of what they were and were not yet ready for. That is the clearest signal available that a question has found its mesocosm: it does not need to be imported, because it is already there, doing work, before the researcher ever arrives.

V. Whose Problem Is "What Holds Hinduism Together?"

Axel Michaels poses a serious scholarly question: what holds a tradition as internally diverse as Hinduism together, given that it has no single founder, no single creed, no centralized authority enforcing doctrinal uniformity? His answer, identificatory habitus, a learned, embodied capacity to recognize oneself as Hindu across enormous regional, sectarian, and practical variation, is a careful and illuminating response to that question.

But the question is not asked by anyone inside the mesocosm it is asked about. Nobody offering food to Ma Kali at Kalighat, nobody choosing between Krishna and Kali as an iṣṭa-devatā, nobody objecting to the brahmins' insistence on goat sacrifice, is troubled by whether their practice coheres into a single classifiable system alongside Christianity and Islam. That is a problem generated by comparative religion's own classificatory needs: the discipline requires traditions to be sortable, nameable, comparable, and Hinduism, lacking the familiar markers of doctrinal centralization, becomes puzzling only within that prior requirement. Identificatory habitus answers a genuine scholarly puzzle. It does not answer a question the mesocosm itself is asking.

This does not make Michaels' question illegitimate as scholarship. It makes it, in the terms developed above, an L4 question whose stakes belong to comparative religion rather than to the field it is applied to. The diagnostic test from the previous section applies cleanly here: what is the problem, and whose problem is it? The coherence of Hinduism as a classifiable tradition is the discipline's problem. The problems that are actually live within the mesocosm, which devatā to approach, how to manage a dangerous relationship safely, what a closed door protects, what a goat sacrifice might really be for, are different problems, with different and far more locally legible stakes.

VI. Paul's Problem: A Good Question Native to Its Own Mesocosm

The contrast between a good question and a bad one is not a contrast between question types. The very same formal question, how do authoritative texts relate to practices that have changed, can be a good question in one mesocosm and an imported, alien classification in another. The clearest demonstration of this is the early Christian movement itself, the historical source of the text and practice problem that later attached itself, illegitimately, to the study of Hinduism.

Paul's situation in the middle of the first century was a crisis of coordination, not a crisis of meaning. Can Gentiles join the movement without becoming Jews first? Must Gentile men be circumcised? Can Jews and Gentiles eat at the same table without fracturing the very communities the movement depended on for its survival? Are the dietary laws binding on those who enter through faith rather than through covenantal birth and observance? These were not abstract puzzles. They were questions on which the continuation of a fragile new community, and in the participants' own understanding their salvation, genuinely depended. Paul's answer retained the authority of the inherited scriptures while suspending many of the practices those scriptures had prescribed. Circumcision, dietary law, and purity regulation stopped functioning as the boundary markers of belonging; grace, faith, and baptism took their place.

This generated, for the first time, a real and pressing version of the text and practice problem: the texts remained authoritative, the practices they prescribed no longer were, in the same way and for the same reasons, and the question of how to reconcile the two carried the full weight of communal survival and personal salvation. For early Christians, asking how scripture related to practice was not an imported classification. It was the direct articulation of an L2 felt misalignment grounded in the most basic L1 stakes the movement had, whether the community would hold together at all, and whether its members were right with God.

The question did not stay where it started. Patristic allegory, medieval exegesis, the Reformation, and Protestant scripturalism all extended the same configuration, an authoritative text, a changed practice, and an unavoidable question about their relationship, carrying ever fainter traces of the original salvific urgency but retaining its structural shape. By the time European scholarship turned its comparative attention to other traditions, the text and practice question had been generalized into something that looked like a neutral methodological requirement: any serious study of a religious tradition, the assumption ran, must account for both its texts and its practices and explain how they relate. The assumption felt obvious because its origin had become invisible.

This is the precise mechanism by which a good question becomes a bad one without changing its form at all. For Paul's community, the question was diagnostic: it tracked real stakes, and answering it mattered enormously to the people who asked it. Applied to the Kali worship mesocosm, the same question tracks nothing. Nobody offering food to Ma Kali is suspended between an authoritative scripture and a practice that scripture no longer licenses. The question of how text relates to practice does not arise from any felt misalignment in that mesocosm, because the mesocosm was never organized around the specific historical crisis that made the question necessary in the first place. Importing it produces, at best, the appearance of rigor, a comparative observation that both traditions have texts and both have practices, floating free of the entirely different stakes that made the relationship between the two an urgent problem in one case and a non-issue in the other.

The lesson generalizes well beyond the comparison of Christianity and Hinduism. Whether a question is good or bad cannot be read off its grammatical form, its theoretical sophistication, or its track record of producing answers elsewhere. It can only be read off whether it tracks a stake that is actually live in the mesocosm being asked about. The same applies, this article would argue, to social and historical research on any topic whatsoever, well beyond the study of religion: a question that worked because it tracked real stakes in one historical setting carries no automatic license to be asked, in the same form, of a different setting whose stakes have not yet been independently established.

VII. Stakes: The Anchor Beneath the Felt Misalignment

If a felt misalignment, an L2 disturbance, registers as a disturbance only because something at stake in the underlying L1 coordination is under pressure, then the clearest way to see a stake in operation is to watch what happens when that coordination actually breaks.

One story from my fieldwork notes makes this unusually vivid. A respected pujari, known for performing a midnight offering to Ma Kali every single night without exception, was engaged to officiate a wedding at a hotel in the city. He expected to be home in time for the midnight offering, but the ceremony ran late, so he telephoned his wife and asked her to make the offering in his place at the usual hour. She agreed, then sat down to watch television, lost track of time, and fell asleep before midnight without having made the offering. Back at the wedding, as the hour passed, the pujari, still reciting at full volume with several assistants beside him, began, without apparently noticing, to recite the mantras used for death rites instead of the mantras appropriate to a marriage. His assistants, alarmed, did not dare correct him directly; eventually one of them shook him and asked what he was doing. He came back to himself, immediately telephoned his wife, and demanded to know whether she had made the offering, telling her that Ma Kali was disturbing him because she had not. She got up at once, bathed, and made the offering. Within minutes, by every account, he became calm again, and the ceremony continued.

Nothing in this story requires us to settle the question of whether Kali ate anything that night. What the story shows with total clarity is what was actually at stake: an unbroken nightly relationship, maintained without exception for years, whose interruption produced an immediate, embodied, public, and consequential disturbance, registering first in the pujari's own coordination, his mind wandering uncontrollably into the wrong ritual register, before it registered anywhere as a sentence anyone could say aloud. The misalignment here was not a felt discomfort about metaphysics. It was a felt disturbance in an L1 world organized around an unbroken sequence of nightly care, and it surfaced, the moment that sequence broke, with a force that no question about literal eating ever produced in my interviews.

A second kind of stake becomes visible in a different register, not ritual breakdown, but moral contestation from within. Several of the people I spoke with were sharply critical of the requirement, maintained by local brahmins, that certain pujas to Ma Kali include the sacrifice of a goat. Their objection was not scepticism about the goddess's reality or power. It was a tightly reasoned argument: Ma Kali is the mother of everything, including the goat; how, then, could she possibly want her own child killed and offered to herself? The suspicion several of them voiced was that the brahmins wanted the meat for themselves and used the goddess as cover for a desire they could not admit to directly, eating and drinking in private what they claimed in public never to touch. This is not the disorganized scepticism of people whose ontology has begun to fray under modern pressure. It is an argument internal to the mesocosm's own logic, a mother does not demand her child's death, aimed at exposing a specific, locally legible form of exploitation. The stake here is fairness, the honesty of those who claim ritual authority, and the moral coherence of the goddess's own relationship to her devotees, all of it entirely independent of whether anyone present believed Kali literally ate anything at all.

Both cases illustrate the same general point. A stake is not a belief, not a value statement, not a position one could be talked out of by a better argument. It is what makes a disturbance feel like a disturbance in the first place: the reason a missed midnight offering produces a public breakdown rather than a shrug, the reason a goat sacrifice produces a moral objection rather than indifference. Strip away the stake, and the same events would register as nothing at all. The question of whether the deity eats has, by contrast, no comparable purchase on anyone's L1 world. Nobody's coordination collapses, nobody objects on grounds of fairness, nobody telephones anyone in a panic, over whether the literal versus symbolic status of divine consumption has been correctly determined. That asymmetry, between what produces real disturbance and what produces only hedged or indifferent answers to an outsider's question, is the most reliable evidence available for where the stakes of a mesocosm actually lie.

VIII. Temple and Household: Stakes Unevenly Distributed

Stakes are not distributed evenly across a mesocosm. The same goddess who is approached with extraordinary caution in one setting is approached far more casually in another, and the difference tracks neither belief nor doctrine but the local distribution of risk and the resources available for managing it.

Kalighat, Dakshineswar, and Tarakeshwar were named to me repeatedly as places where Kali, Shiva, and other major deities are present with unusual intensity, and where the consequences of a ritual mistake are correspondingly severe. People who do her puja, one man told me, have to be extremely careful; a single error and she will destroy everything, sorbonāś hobe, total ruin. Other gods, he added, Krishna, Radhakrishna, Viśvakarmā, Shiva among them, are satisfied with comparatively little ritual exactness. Ma Kali will not forgive a mistake the same way. The temple, in this account, is not safer because it is more sacred in some abstract sense, set apart from an ordinary profane world. It is safer because, over a long time, it has accumulated the resources that make handling an intense presence survivable: an architecture built for the purpose, generations of correctly repeated performance, a designated brahmin whose competence is itself a kind of infrastructure, a distributed body of people whose task is precisely to manage what a single household could not manage alone.

A further distribution appeared in how people spoke of illness. Ma Sitala, several told me, is the goddess associated with fevers and skin eruptions, chickenpox above all; the illness itself is described as something she gives and something only she, in her own time, takes away, with a period of days during which the family waits, observes restrictions, and afterward bathes in coconut water before any puja of thanks is offered. The danger here is not a generalized divine intensity of the kind associated with Kali. It is a specific, bounded domain, illness and its course, in which a particular goddess is the one whose relationship has to be tended. The stakes shift again: not safety in approaching an overwhelming presence, but patience and correct timing in living through an affliction that has already arrived.

Household worship is more exposed than temple worship for the opposite reason from the temple's safety: fewer resources are available to absorb a mistake, responsibility rests with fewer people, and the relationship is correspondingly more intimate and more dangerous. This is also why, several of my interlocutors made clear, one does not let an inexperienced person approach Kali casually, in the way a child might say a quick prayer to Krishna or pour cooling water and milk over a Shiva liṅgam during Shivaratri. The distribution of danger across deities and across sites is not a symbolic classification superimposed on the practice from outside. It is the practice's own account of where its real risks lie, who can be trusted to manage them, and what has to be in place before a given intensity of divine presence can be safely approached at all.

IX. A Typology of Bad Questions

The distinction between good and bad questions, developed across the cases above, generates a short typology of the ways inquiry goes wrong when it begins with an imported L4 rather than with a mesocosm's own stakes.

The first failure is stake blindness: beginning with a category, a survey instrument, or a theoretical framework before any attempt has been made to discover what actually matters to the people being studied. A researcher who arrives already certain that the interesting question about Kali worship concerns literal versus symbolic divine consumption has decided, before a single interview has taken place, what counts as data and what does not. This is rarely malicious. It is the almost automatic consequence of bringing a discipline's existing resources to a field rather than letting the field generate its own questions first.

The second is imported askability, asking a question that belongs to the researcher's own mesocosm and mistaking it for one that belongs to the field. Does Kali actually eat? is imported askability at the level of a single interview question. What holds Hinduism together? is the same failure at the level of comparative theory. Both questions can produce coherent, even illuminating, answers; the danger is precisely that they do, because an interesting answer to an imported question is easily mistaken for an insight into the field, when it is in fact an insight into the question's own history.

The third is premature universalization, assuming in advance that one's own framework, ethical, legal, medical, or developmental, identifies the stakes that matter everywhere, for everyone, in every mesocosm, without first doing the work of discovering what is locally at stake. A researcher who arrives at the goat sacrifice already equipped with a settled position on animal welfare, and who treats the practice primarily as an instance of a universal ethical question, will miss the entirely different and entirely indigenous moral argument several of my own interlocutors were already making, that a mother goddess cannot coherently want her own child killed, and that the brahmins administering the sacrifice may be using her as cover for an appetite of their own. This argument has its own stakes, fully legible within the mesocosm, and does not need a universal framework imported from outside to be recognized as serious.

The fourth is representational reduction, assuming that an offering, a closed door, or a ritual gesture exists primarily to represent or symbolize a meaning, rather than to participate directly in a relationship that continues over time. The closed door at Kalighat is not best understood as a symbol of anything. It is a practical, repeated, carefully timed procedure whose interruption, as the wedding story above shows, produces consequences with no symbolic distance at all. Treating it as a representation to be decoded misses what the procedure actually does.

Each of these failures shares the same root. They all substitute a question the researcher already has an answer for, in shape if not in content, for the slower and harder work of discovering what question, if any, the mesocosm is already asking itself, and what is genuinely at stake in the answer.

X. Mesocosmography as Stakes-First Inquiry

The positive method that follows from all of this can be stated as a short list of questions, asked before any category, doctrine, text, or practice is allowed to organize the inquiry. What can go wrong here, and for whom? What must continue without interruption, and what happens when it does not? What forms of care, competence, or caution are required, and who is trusted to provide them? What kinds of failure are feared, and what does that fear reveal about what is actually being protected? What forms of remediation are already under way, before any researcher arrives to name them?

Applied to the material examined here, these questions reorganize what counts as evidence. A closed door is no longer a puzzle about belief, requiring a decision about whether the people closing it are being literal or figurative. It is evidence about timing, trust, and the proper sequence of an ongoing relationship. A goat sacrifice is no longer a puzzle about ritual violence requiring an external ethical verdict. It is evidence of an internal moral argument about motherhood, fairness, and the honesty of those who claim ritual authority. A difference in how cautiously people approach Kali compared with Krishna is no longer evidence of inconsistent belief. It is evidence of a genuine, locally calibrated distribution of danger and competence.

This reorientation also clarifies how comparison should proceed, when comparison is wanted at all. Before claiming that two traditions share a concept of purity, or a text and practice problem, or any other apparently common term, the stakes animating that term in each setting need to be independently recovered. The Pauline case discussed earlier shows why this matters. The text and practice problem was real, urgent, and salvific for early Christians, and a comparison that simply notes that both traditions have texts and both have practices, while ignoring this entirely different configuration of stakes, produces the appearance of insight without any of its substance.

Although the material examined here is restricted to Bengali Hindu practice and to the early Christian case used for comparison, the underlying method is not specific to either. Any research question, in any historical period, on any topic, carries an implicit answer to the question this article has placed at its centre: what is the problem, and whose problem is it? A historian asking why a particular legal reform was adopted, an economist asking why a market behaved as it did, a sociologist asking why a community organized itself in a particular way, all face the same choice between importing an already stabilized category and doing the slower work of recovering what was actually at stake for the people whose actions are being explained. The argument developed here through Bengali Hindu material is, in that sense, a method for social and historical inquiry generally, illustrated through one field rather than confined to it.

XI. Conclusion: Figure Out Their Stakes to Figure Out Their Ontology

The argument of this article can be stated in the same single sentence with which it began. Figure out the stakes, and you will figure out the ontology. The question of whether Ma Kali eats the food offered to her produces hedging, contradiction, and occasional resentment, not because the people answering it have an unstable or incoherent ontology, but because the question was never theirs to begin with. It arrived already finished, an L4 import carrying a metaphysical architecture, mind against matter, literal against symbolic, that nobody in the mesocosm needed and nobody's coordination depended on settling. The question of how to find one's iṣṭa-devatā, by contrast, produced fluent, confident, practically reasoned answers, because it was already alive in the mesocosm, tracking real concerns about safety, appropriateness, and the management of an intense and consequential relationship.

The same asymmetry runs through every case examined here. A missed midnight offering produces a public, embodied disturbance because an unbroken nightly relationship was genuinely at stake. A goat sacrifice produces a sharp internal moral argument because the coherence of the goddess's own relationship to her devotees is genuinely at stake. A difference in caution between Kali and Krishna tracks a real, locally calibrated distribution of danger because safety is genuinely at stake. None of this required settling, or even raising, the question an outside observer might have assumed was the natural place to start.

The method that follows is correspondingly simple to state and harder to practise. Before asking what a tradition believes, what its texts mean, what its categories classify, or what unifies its practices into a nameable whole, ask what is actually at stake for the people living inside it: what they protect, what they fear, what they argue about among themselves, what would count, in their own terms, as getting it wrong. The categories, the doctrines, the unifying theories can come later, if they are needed at all. What cannot come first, without distorting everything that follows, is an already finished question, however sophisticated, however well it served some other field, some other century, some other problem entirely. Stakes come first. The ontology follows.

References

Dunn, James D. G. "The New Perspective on Paul." Bulletin of the John Rylands Library 65 (1983): 95-122.

Ecks, Stefan. Living Worth: Value and Values in Global Pharmaceutical Markets. Durham: Duke University Press, 2022.

Fuller, C. J. The Camphor Flame: Popular Hinduism and Society in India. Princeton: Princeton University Press, 1992.

Michaels, Axel. Hinduism: Past and Present. Translated by Barbara Harshav. Princeton: Princeton University Press, 2004.

Michaels, Axel. Homo Ritualis: Hindu Ritual and Its Significance for Ritual Theory. Oxford: Oxford University Press, 2015.

Sanders, E. P. Paul and Palestinian Judaism: A Comparison of Patterns of Religion. Philadelphia: Fortress Press, 1977.