If a querent’s grip on consensual reality is already fraying, every card you lay down is not just a symbol—it is an instruction about what is real.
Most of us read as though divination is essentially descriptive. The cards mirror, the spread arranges, the querent makes meaning. But at the psychotic edge, the oracle stops being a mirror and starts acting like an operator: it writes directly into the querent’s reality model. A single Ten of Swords can function less as “a difficult ending” and more as, “Proof I am under attack.” A Page in a “What are they thinking?” position becomes, “Evidence my neighbour is surveilling me.”
If you ignore that shift, your beautiful Celtic Cross is just a high‑gain amplifier on a failing reality‑testing system.
This is not a call to turn diviners into amateur psychiatrists. It is an acknowledgement that divination has always been recognised as dangerous to judgement, that modern psychosis theory gives us language for *how* it can be dangerous, and that spread architecture, protocol, and stance are the only levers you actually have at the table.
### Fragile reality‑testing is not a modern invention
The fantasy that “pre‑modern people just called psychosis ‘demons’” dies quickly if you actually read the sources. The argument about what visions *are* is as old as visionary culture.
Johann Weyer, writing in 1563 against witch‑hunters who saw demons everywhere, describes “melancholic illusions” in which “images impressed upon the imagination” are taken as external fact. These people, he notes, “believe their visions in such wise that no argument can move them.” That is a serviceable phenomenological description of fixed delusion centuries before DSM.
Robert Burton’s *Anatomy of Melancholy* catalogues how sermons, symbols and scripture become *fuel* for certain temperaments. The melancholic visionary, especially in “Religious Melancholy”, pores obsessively over signs and prophecies, finds themselves in every passage, and takes each textual “hit” as confirmation. Burton explicitly advises that some people must *avoid* certain stimuli: no dwelling on apocalyptic verses for those already “inclined to fear and fancy apparitions.”
Aquinas sits earlier again, worrying in the *Summa* about *sortilegia* and *divinatio astrologica* because they hand the will over to external agencies. Divination, in his scheme, is not just illicit content; it is a practice that reshapes the soul’s posture to authority. He distinguishes divine revelation, bodily illness, and demonic deception, but in all three the imagination is the channel, and what passes through it can be either tested or swallowed whole.
On the other side, Bodin and Kramer angrily reject “melancholy” as an explanation. To downplay demons, for them, is naivety. That very polemic is evidence that a proto‑clinical frame existed and was being resisted.
The occult literature is more operational. Ficino praises the Saturnine visionary and in the same breath warns that the same configuration can tip into catastrophic imbalance; one does not bandy oracles around the ungrounded. Agrippa devotes a whole book to imagination as a force that can heal, harm, and “make its own oracles.” Lévi warns that unstable minds, exposed to astral forces without discipline, fall into “hallucination” and “astral intoxication”—his vocabulary anticipates, and historically precedes, Jung’s concept of “inflation.” The Golden Dawn’s protocols for testing visions and insisting on mental balance before astral work are likewise earlier than the clinical literature, not derived from it. Crowley is blunt: those who take every inner voice literally end in “Lunatic Asylums.”
Nobody in this tradition thought that imaginal practices were harmless.
What modern psychopathology adds is not the discovery of fragile reality‑testing—that was always visible—but a technical vocabulary: *delusion* as belief impervious to counter‑evidence; *ideas of reference* where unrelated events are experienced as messages “about me”; *thought disorder* and loosening of associations; *boundary weakening* where self/other and thought/world distinctions blur. Bleuler’s “autistic thinking”—a term now retired from clinical use, but the underlying observation remains precise—describes the collapse of shared referential logic into private symbol-systems that are internally consistent but impenetrable from outside; it is practically a description of the oracular mindset taken to pathological extreme. Jaspers hammers home that primary delusion is not simply eccentric belief; it is an unshareable shift in the structure of experience itself. Jung, later, calls out “inflation”: archetypal contents overwhelming ego structure and colonising identity. Grof’s “spiritual emergency” framework adds a functional criterion: the question is not only what the person experiences but whether they can still modulate intensity, distinguish imaginal from literal, and integrate over time.
Taken together, this older demonological/melancholic discourse and modern psychopathology converge on a practical point: certain minds, in certain states, will treat oracular material not as “maybe” but as “is;” not as symbol but as fact. For those minds, your spread is not neutral information. It is a piece of active machinery in their epistemology.
### Divination as an operator, not a mirror
Under ordinary circumstances, a reading is at least partly buffered by the querent’s intact symbolic function. They know the Devil can stand for addiction rather than a literal goat‑footed entity in the walls. They know “he’s thinking of you” is a story about attachment, not an audio tap into another mind.
At the psychotic edge, several things happen at once.
First, *concrete thinking* pushes out symbolic distance. The metaphor *is* the thing. Jung remarks on this in psychosis: the archetypal image is experienced as literal ontological fact. Psychotic phenomenology research says the same less poetically.
Second, *referentiality* goes into overdrive. The world talks *to* me and *about* me. The DSM calls this “ideas of reference.” The predictive processing crowd calls it aberrant salience: noise tagged as crucial signal. Divination institutionalises that move: by design, a shuffled pattern of paper is “about you.” When you tune a referential mind to an oracle, you connect a hyper‑salience engine to a device whose explicit purpose is to turn randomness into personal message.
Third, there is *externalisation* of internal material. Intrusive thoughts and nascent voices feel “not me.” Tools that speak—cards, runes, pendulums—become convenient transducers. Instead of “I had a thought that…”, the report becomes “The cards/spirits said…” Weyer would have called these “images impressed upon the imagination” and Burton would note that the form in which they appear—a sermon, a verse, a card—matters for how they are believed.
Fourth, procedural authority. Divination is full of rules: if card X appears in position Y, read as Z. That move—this configuration entails that meaning—is precisely the kind of hard inference that delusional systems exploit. “I drew the Tower in the health position, therefore I will have a stroke this month.” For someone with fragile ego boundaries, the reading protocol is not game; it is law.
You can see why Aquinas worries about “handing the will over.”
It is not enough, in these cases, to “read gently” or to give “empowering” interpretations. The architecture you choose and the permissions you grant to the oracle change what counts as evidence, how much noise is canonised as signal, and what sort of behaviours can be justified on the basis of the spread.
### Spread as filter: managing signal, noise, and gain
The analogy that clarifies this best comes from cybernetics. It is only an analogy: spreads do not compute, and querents are not control systems. What the analogy does is clarify what kind of thing spread architecture affects, without claiming to explain the mechanism. In a noisy environment, a filter or controller decides what information the system is allowed to update itself on. Too permissive, and every fluctuation causes over‑correction. Too rigid, and the system ignores crucial shifts.
Reality‑testing is that kind of filter: we check perception against consensual reality, we distinguish thought from fact, we let some anomalies slide as noise. Psychosis is in part a failure of that filter: aberrant precision on noise, excessive meaning assigned to irrelevant cues.
A spread is not a brain. It does not alter neurotransmitters. But structurally, it is a filter: it defines what counts as relevant signal in the reading, how many channels speak at once, where feedback loops are allowed.
For a querent whose inner filter is already compromised, your layout can either tighten or loosen it further.
Highly open, position‑light layouts—“let’s just throw twelve cards and see what the universe says”—maximise degrees of freedom. They are beautiful when the symbolic function is strong and you want creative association. They are lethal when associations are already unbounded. They invite exactly the thought disorder Bleuler describes: loosening of associations, tangentiality, private logics that become self‑sealing.
By contrast, a simple three‑card past/present/next‑step with tightly defined positional meanings and a single, contained question acts as a dampening filter. It constrains proliferation. It does not cure psychosis, but it reduces the number of loci where mis‑assigned meaning can hook in.
Think in those terms and several practical consequences follow:
– You choose **low‑freedom spreads** by default with fragile reality‑testing: few cards, no floating “clarifiers”, fixed positions tied to domains the querent can verify.
– You avoid architectures that encode referentiality you do not want: spreads about “what X is thinking” or “what your enemies are doing” are invitations to third‑person mind‑reading and persecutory ideation.
– You treat iterative pulling and re‑asking the way an engineer treats feedback loops. High‑gain loops oscillate. In oracular terms: the more often the querent re‑asks and you reshuffle, the more you train obsessive checking and install the oracle as arbiter of every micro‑event.
There is no data showing that these moves alter the *clinical course* of psychotic disorders. Nobody has run a trial of “Celtic Cross vs three‑card” on relapse rates. We are not talking about treatment. What they obviously do change is moment‑to‑moment reality‑testing *inside the reading*, and the range of behaviours the querent can plausibly justify by appeal to the cards.
That is the level at which you actually have leverage.
### Symbol boundaries as ego scaffolding
Depth psychology has spent a century worrying about ego boundaries: where self ends and other begins, where inner image stops and outer world starts. Psychosis is in many ways the collapse of those boundaries. “The TV is talking to me.” “Those are not my thoughts in my head.”
Ritual magic, long before any of that psychology, invents circles, openings and closings, banishings and licences to depart. These are ways of saying: “Inside here is one kind of space; outside is another. What comes in has a limited remit. When we close, it goes.”
Divination, particularly as it is casually practised now, tends to omit explicit containment. The table is opened and closed, if at all, only symbolically. Yet spread structure and reading protocol *are* a kind of boundary‑work. Positions like “inner state” vs “outer event”, or “self” vs “other”, are symbolic demarcations. How much you lean on them—how explicitly you name and enforce them—matters when ego boundaries are thin.
Working at that edge, it is not enough to keep positional meanings in your own head. You have to make them collaborative scaffolding:
– You say aloud: “This column is strictly about your internal experience—feelings, thoughts, images. It is *not* telling us what your neighbour is actually doing.”
– You counterpart it: “This row is about things that, in principle, could be checked in the ordinary world: appointments, who said what, whether a bill arrives. We will interpret these cards only in those terms.”
– You bracket ontology up front: “For the purposes of this reading, we are treating the cards as representations, as images that can help us think about possibilities. They are not permitted to give us absolute verdicts on what other people are, think, or feel.”
You are not sneaking in a covert “it’s all in your head” undermining of the querent’s metaphysics. You are building an as‑if frame in which symbol and referent can be distinguished at all. Jung’s “active imagination” technique hinges on this: inviting images in whilst maintaining that they are images.
Opening and closing become more than theatre:
– At opening, you bound time (“half an hour”), scope (“this situation, this week”), and permissions (“cards do not adjudicate on demons, diagnoses, or surveillance”).
– At closing, you explicitly de‑commission the spread: cards gathered, images “go back to being just pictures”, any “messages” moved into the category of hypotheses to be tested in lived life.
Will that procedure restore ruptured ego boundaries? No. Again, we are not mistaking metaphor for neurology. What it does is support whatever boundary‑function remains by coherently marking domains: inside the circle is imaginal work; outside is ordinary life. For some querents, that is the only labelling of domains they get.
### Access control: what you simply do not let the oracle decide
A more unusual but practically sharp way of modelling this is via information security. In a computer system, access control decides which processes can read or write which parts of memory. Code that runs “sandboxed” can explore, but not change core configurations. Code running as root can rewrite the entire system.
Referential delusion is like giving root access to every sign. A car backfires and the world‑model updates: “They are sending me signals.” A pigeon lands on a sill and the script rewrites: “Angels are confirming the prophecy.” There is no effective distinction between “this is an interesting thought” and “this is a binding fact about reality.”
Unstructured divination, in that setting, is privileged code. The deck is granted write access to core beliefs: if the cards say it, it is simply true. That is how you get “I stopped my antipsychotics because the cards told me my psychiatrist was an agent”, or “I cut off my entire family because the oracle confirmed they are demons.”
You cannot change the querent’s internal access model. You can absolutely change what permissions you grant the oracle in your practice.
That means formulating, for yourself, something like a permissions schema:
– **No‑access domains** – topics on which you simply refuse to read, because any “answer” is either unfalsifiable or functionally hazardous:
– “Am I being surveilled by the government, spirits, demons, or neighbours?”
– “Is X literally a demon/alien/AI agent?”
– “Am I the sole chosen messiah, twin flame of a celebrity, or architect of a cosmic shift?”
Aquinas would have said such questions belong to God alone, or to delusion. Modern psychiatry says they are canonical content of psychosis. Either way, it is not an oracle’s job to adjudicate them.
– **Read‑only, hypothesis‑only domains** – areas where you are willing to use cards as exploratory tools but refuse to let them decide anything:
– “How am *I* experiencing my relationship with my neighbour/therapist/partner?”
– “What feelings are activated when I hear footsteps upstairs?”
– “What might help me cope with the fear that my thoughts are not private?”
You state overtly: “The cards can help us explore how this feels and what options you have. They cannot tell us what is objectively, finally true about other people or ultimate reality.”
– **Normal‑access domains** – bounded, mundane questions that can be checked and do not plug directly into persecutory or grandiose themes.
– “What might support me this week?”
– “How can I structure my day to lower stress?”
– “What do I need to be aware of in managing this work situation?”
You enforce these permissions as hard rules, not guidelines you negotiate under pressure. A querent at the psychotic edge will often push for root: “But the cards *must* be allowed to tell me whether the voices are angels.” That is precisely where you say no.
This is harm reduction, not hidden diagnosis. You are not deciding what their experiences “really are.” You are deciding what role your tool will play in their epistemology.
### When you should not be reading at all
There is an obvious objection here: what gives you the right to decide someone is “on the edge of psychosis” in the first place? You are (probably) not a psychiatrist. You are not licensed to diagnose.
The hard answer is: you are not diagnosing. *You are triaging risk inside your own practice.*
You can, and should, stay entirely out of labelling and nonetheless notice certain patterns as red flags that move the situation out of “adjust the spread” and into “defer and refer”:
– The querent consistently reports fixed persecution or surveillance that has no plausible mundane correlate and does not shift at all when gently questioned: government implants, AI mind‑readers, demons in the electricity, neighbours reading thoughts through the ceiling.
– They explicitly experience their thoughts as inserted or stolen (“those are not my thoughts”, “they can hear everything in my head”), and want you to use cards to locate perpetrators.
– Their speech is sufficiently disorganised—derailment, loose associations, neologisms—that you cannot maintain a coherent question.
– They frame you, or the deck, as literally controlled by hostile forces in the moment: “The demons are making you lie”, “The cards are being hijacked.”
– They tell you they are making or considering high‑risk behavioural choices *because of oracle instructions*: stopping medication, cutting all contact with support, harming self or others.
When you encounter that combination of content and conviction, the responsible move is not, “Let me craft a very clever low‑noise spread.” It is to stop, name your limits, and direct them towards people whose scope actually includes managing imminent risk: crisis lines, community mental health teams, trusted clinicians, in some jurisdictions emergency services.
You do *not* say “You’re psychotic” or “You’re delusional.” You stay in your lane: “The kinds of experiences you’re describing are outside what cards can safely address and outside my competence. I don’t want to feed anything that might harm you. Let’s talk about other kinds of support.”
False positives are inevitable: you will sometimes be cautious with someone who, in clinical terms, is not psychotic but is in an intense non‑ordinary state. Erring on the side of not using divination to adjudicate ontological high‑stakes questions is tolerable collateral damage. False negatives—missing a psychotic process and letting the oracle steer dangerous action—are less tolerable.
### What “worsening” looks like, in practice
Another sceptical worry: how do you know when the oracle is actually *contributing* to deterioration rather than just coinciding with it?
You probably cannot know with certainty. You are not running a controlled trial. What you can observe are functional and behavioural signals—not diagnostic criteria, but practitioner-observable thresholds that tell you when the reading relationship has crossed from possibly useful into probably harmful.
The clearest signal is **compulsive high-frequency use with functional deterioration**: the querent who consults cards multiple times daily, whose decision-making capacity is narrowing rather than expanding, who cannot act on anything without first “checking” the oracle. This is not enthusiasm or devotion. It is a function of the oracle displacing rather than supporting the querent’s own reality-testing. Their ability to navigate ordinary life is shrinking in direct proportion to their reliance on the cards.
A second signal is **high-risk actions attributed to oracular directives**: stopping psychiatric medication because “the cards said to,” cutting off all support contacts because the spread “confirmed” they are malicious, making large financial commitments because a particular card “appeared three times.” The shift to note is not the decision itself but the causal structure: the oracle has been granted root access to behaviour. What matters to you, as the practitioner, is not whether the action is objectively risky but whether the justification structure is one that bypasses the querent’s own judgment entirely.
A third signal is **thematic rigidity regardless of outcome**: the querent for whom every spread, whatever cards fall, arrives at the same conclusion—persecution, special mission, betrayal by named individuals. When the reading process has lost the capacity to produce unexpected results, when every configuration confirms the same fixed narrative, the oracle is no longer functioning as an information source. It has become a ritual ratification machine for a closed belief system.
These are not clinical diagnoses. You are not qualified to offer one, and that is not what this framework requires. The distinction that matters is between **symbolic harm**—interpretations that generate unhelpful anxiety, self-limiting beliefs, or magical thinking—and **functional harm**: measurable deterioration in the querent’s ability to navigate ordinary life, maintain relationships, or keep themselves safe. The ethical anchor is the latter. Symbolic discomfort is part of serious oracular work. Functional harm is where your obligations as a practitioner begin to outweigh any benefit the reading might offer.
When you see functional harm—not suspect it, not theorise it, but observe it in the specific behaviours above—the reading stops. Not with diagnosis, not with clinical language, but simply: “I’m not going to continue this right now. The kinds of experiences you’re describing need more support than I’m able to offer here.”
