Hypnosis
from First
Principles

The Myth and
the Mechanism

The popular understanding of hypnosis is almost perfectly wrong. The stage hypnotist has produced, over a century of entertainment, a specific set of misconceptions so durable that they contaminate even clinical thinking: that hypnosis involves sleep, or something like it; that the hypnotised subject is under the control of the hypnotist; that hypnosis grants access to hidden memories with special reliability; that only gullible or weak-willed people can be hypnotised; and that the phenomena are theatrical compliance rather than genuine altered experience.

Each of these is false. Hypnosis is not sleep (EEG studies consistently show that hypnotised subjects are awake, and in many cases more focused and alert than baseline. The hypnotist does not control the subject) subjects who are given instructions that violate their values under hypnosis simply refuse them, and no competent researcher has produced a controlled demonstration of hypnotically-induced behaviour that the subject would otherwise find genuinely unacceptable. The idea that hypnosis provides reliable access to repressed memories is not only unsubstantiated but actively dangerous, the hypnotic state appears to increase confabulation, producing vivid false memories with high phenomenal confidence. And susceptibility to hypnosis correlates positively with intelligence, imaginative capacity, and the ability to become absorbed in experience, not with gullibility.

What hypnosis actually is has been the subject of serious theoretical dispute for over a century, and that dispute is not fully resolved. But three things are agreed: first, that hypnotic phenomena are real, they are not simply the product of social compliance or voluntary pretence. Second, that they involve genuine alterations in neural processing, measurable by neuroimaging. Third, that they reveal, more clearly than any other phenomenon in normal human experience, the architecture of the relationship between conscious awareness and the systems that generate experience.

Susceptibility: Who and How Much

Hypnotic susceptibility (or hypnotisability) is a normally distributed trait in the human population. Standardised scales, principally the Stanford Hypnotic Susceptibility Scale (SHSS) and the Harvard Group Scale, measure responsiveness to a battery of standardised suggestions. Approximately 10–15% of the population scores at the high end, these are the people who can experience vivid positive and negative hallucinations, complete amnesia, and deep analgesia under hypnosis. Approximately 10–15% score at the very low end and show essentially no responsiveness to suggestion. The majority fall in the middle and can experience lighter phenomena, arm levitation, increased relaxation, partial amnesia.

Hypnotisability is remarkably stable across the lifespan. Piccione, Hilgard, and Zimbardo (1989) retested subjects on the Stanford scale 25 years after initial assessment and found a test-retest correlation of 0.71. This is higher than the 25-year stability of most personality traits. Hypnotisability appears to be a genuine individual-difference variable (probably heritable) that remains stable through adult life. It is not a skill that can be significantly increased through practice. The implication is that the capacity for hypnotic phenomena reflects something structural about the brain's organisation rather than a learnable technique.

Highly hypnotisable subjects share certain characteristics. They tend to score high on absorption, the tendency to become deeply immersed in imaginative, aesthetic, or perceptual experiences. They tend to have rich, vivid, detailed internal imagery. They have thin boundaries between self and world, between reality and fantasy, in the sense that imaginative engagement carries high subjective reality for them. These traits suggest that hypnotic susceptibility reflects the baseline precision weighting of the generative model: people who assign high precision to internally-generated predictions are more readily able to enter states in which the model runs with reduced reality-testing.

Milton Erickson:
The Architect
of the Method

Milton Hyland Erickson (1901–1980) is the single most important figure in the history of clinical hypnosis, not because he invented it but because he fundamentally reconceived what it was. His work transformed hypnosis from a formal, authoritarian procedure requiring a compliant subject and a commanding operator into a fluid, naturalistic, profoundly attentive clinical art that worked with the subject's own mental processes rather than attempting to override them.

Erickson's biography is itself relevant to his method. He contracted polio at seventeen and was told he would not survive. He did survive, but with paralysis of his entire body. Over the following year, lying immobile in his childhood home, he began to reconstruct his own muscle function through what he would later describe as intense, systematic imaginative rehearsal, spending hours visualising the movements his body could not make, observing and learning from his infant niece's motor development, and gradually re-establishing voluntary motor function. He later contracted a second bout of polio in his fifties, which left him in permanent pain and partial paralysis. He managed his own chronic pain through self-hypnosis for the rest of his life.

This personal history is not incidental. It gave Erickson a depth of experiential knowledge about the relationship between conscious intention, unconscious motor function, and the malleability of experience that no purely academic student of the subject could acquire. He knew, from the inside, that the conscious mind's explicit understanding of what it is doing bears little relationship to the unconscious processes actually producing the result. His entire therapeutic philosophy followed from this.

The Interspersal Technique and Embedded Suggestion

One of Erickson's most analysed techniques is what he called the interspersal technique, later elaborated by linguists and cognitive scientists into what is now understood as the structure of embedded command. In its basic form, Erickson would embed a direct suggestion within an apparently neutral narrative or observation, marking the embedded suggestion with subtle shifts in vocal tonality, rhythm, or emphasis that were below the threshold of the listener's conscious monitoring but registered by the unconscious.

A simplified example: "I'm not sure whether you'll begin to feel comfortable now, or whether that might come in a few moments, but most people find, sitting just like that, that it becomes surprisingly easy to just let go." The italicised phrases function as direct commands embedded within hedged, speculative framing. The conscious mind processes the hedge ("I'm not sure whether..."). The marked phrases register as direct instructions at a pre-conscious level.

This is not sleight of hand. The mechanism is a genuine property of how the auditory processing system handles parallel streams of linguistic information. Research on dichotic listening (presenting different messages to each ear simultaneously) established decades ago that unattended speech is processed for meaning even when the listener cannot consciously report its content. Embedded suggestion exploits the same architecture: the message is processed and actioned by systems that operate below the threshold of deliberate monitoring.

The Confusion Technique

Among Erickson's more unusual induction methods, the confusion technique is perhaps the most intellectually interesting. Erickson would deliver a stream of statements that were syntactically complex, logically contradictory, or rapidly alternating in their apparent meaning, designed to tax the subject's conscious processing to the point of overload. When the conscious analytical system was sufficiently overwhelmed by the attempt to track and evaluate the incoming stream, Erickson would insert a simple, clear instruction, and because the evaluation system was occupied elsewhere, the instruction slipped through without conscious scrutiny.

The confusion technique is a direct exploitation of the architecture of attention. Conscious evaluation requires cognitive resources. Those resources are finite. Overloading them with one task leaves them unavailable for another. The technique works not because it is mysterious but because it is a precise application of the psychology of attentional capacity. It is, in this sense, a laboratory demonstration of the narrow bandwidth of conscious processing, the same 40 bits per second identified in Artifact I.

The Neuroscience
of Trance

The neuroimaging of hypnosis has, over the past two decades, moved from curiosity to rigorous inquiry. The central finding can be stated simply: hypnosis produces measurable, replicable, specific changes in brain activity, changes that correspond precisely to the phenomenological reports of hypnotised subjects, and that are not produced by equivalent instructions to simply imagine or pretend the suggested experience.

The Colour Experiment: Staining the Brain

The most elegant early neuroimaging experiment on hypnosis was conducted by Pierre Rainville and colleagues in 1997, and a complementary study by Kosslyn, Thompson, Costantini-Ferrando, Alpert, and Spiegel in 2000. The Kosslyn study gave highly hypnotisable subjects coloured or grey-scale images and asked them to perceive them as either coloured or drained of colour. The instruction was given both as a direct (non-hypnotic) request and under hypnotic suggestion.

The result was unambiguous. When subjects perceived colour under hypnotic suggestion (even when looking at a grey-scale image) the colour-processing areas of the right hemisphere (specifically the fusiform colour area) showed increased activation. When subjects drained colour from a genuinely coloured image under hypnotic suggestion, those same areas showed decreased activation. Under non-hypnotic instruction to imagine the same changes, the modulation was absent. The subjects were not pretending. Their visual cortex was processing colour differently (or not processing it at all) in accordance with hypnotic suggestion, not the physical properties of the stimulus.

This is perhaps the single most important finding in the neuroscience of hypnosis, because it demonstrates that hypnotic suggestion can modulate early sensory processing, not merely the conscious interpretation of sensory data, but the neural processing of perceptual content at a level that precedes conscious awareness. The suggestion was acting on the generative model of the visual system, not on the narrator's interpretation of it.

Hypnotic Analgesia: What the Brain Does With Pain

The analgesic effects of hypnosis are the most clinically relevant and the most neuroimaged. Pain is a two-component experience: the sensory-discriminative component (where is it, how intense is it?) and the affective-motivational component (how much does it matter? how unpleasant is it?). These components are neurally dissociable, they involve partially different systems.

Rainville et al. (1997) used hypnotic suggestion to modify the unpleasantness of a painful stimulus (hand in hot water) without changing the reported intensity. Neuroimaging showed that the affective component of pain (processed in the anterior cingulate cortex) was reduced in proportion to the hypnotic suggestion for reduced unpleasantness. The sensory intensity, processed in the somatosensory cortex, was unchanged. Hypnosis had dissociated the two components of pain, modifying one without touching the other. A subject could feel intense pain but find it not particularly unpleasant, a state that is difficult to even describe in ordinary terms, but that is neuroimaging-confirmed.

The Default Mode Network and the Anterior Cingulate

Spiegel, White, and Poldrack (2024) produced the largest and most controlled neuroimaging study of hypnosis to date, using fMRI on highly hypnotisable subjects. Their principal findings centred on two changes that characterise the hypnotic state across subjects: a marked decrease in activity and functional connectivity in the Default Mode Network (particularly the posterior cingulate cortex and precuneus) and an increase in functional connectivity between the dorsolateral prefrontal cortex (DLPFC) and the insula.

The DMN reduction is consistent with the phenomenology of hypnosis. Subjects in trance report a quieting of the self-referential inner monologue, a reduction in the automatic evaluative commentary that normally accompanies experience. The DMN, as established in Artifact I, is the neural substrate of that self-referential processing. Its suppression under hypnosis corresponds precisely to the phenomenological report of the narrator going quiet.

The increase in DLPFC-insula connectivity is more puzzling and more interesting. The DLPFC is associated with executive control and working memory. The insula is associated with interoceptive processing, the sense of the body's internal states. The increased connection between the two in hypnosis may reflect the mechanism by which hypnotic suggestion can directly modulate bodily experience: the executive system and the body-sensing system come into closer alignment, allowing deliberate suggestion to have more direct influence on felt somatic states.

Hypnosis and the Anterior Cingulate Cortex: The Conflict Monitor Goes Offline

The anterior cingulate cortex (ACC) plays a central role in conflict monitoring, detecting when competing responses, beliefs, or perceptions are simultaneously active and signalling the need for increased cognitive control. It is the neural watchdog that notices, for instance, that your verbal report ("I feel calm") conflicts with your somatic state ("elevated heart rate"), or that a suggestion ("there is a cat on the table") conflicts with your current sensory experience ("I see no cat").

Several neuroimaging studies have found reduced conflict-monitoring activity in the ACC during the hypnotic state. This provides a mechanism for one of hypnosis's most distinctive features: the simultaneous holding of contradictory beliefs without distress. A subject can simultaneously know that no cat is present and experience the vivid perceptual reality of a cat, not because they are confused but because the monitoring system that would normally register and flag this contradiction is operating at reduced gain.

The Hidden Observer:
Hilgard's Challenge
to Unity

Ernest Hilgard was one of the most rigorous experimental psychologists of the twentieth century, and his accidental discovery in 1973 of what he called the "hidden observer" represents one of the most philosophically challenging findings in the history of consciousness research. It emerged not from a planned experiment but from an aside in a classroom demonstration.

Hilgard was demonstrating hypnotic deafness to a class at Stanford, a highly hypnotisable subject had been given the suggestion that he would be deaf to all sounds until told otherwise. A student in the audience asked whether some part of the subject might remain aware of the sounds even while the hypnotic deafness was maintained. Hilgard, on impulse, told the subject that while he would remain consciously deaf, some part of him that was "a hidden part, knowing and feeling everything" could communicate by allowing his finger to rise when asked a question. The finger rose. Asked whether he heard anything, the finger rose again. When the suggestion was cancelled and the subject could hear normally again, he reported hearing nothing during the demonstration, but could describe, accurately, the sounds that had occurred, attributing the knowledge to "an observer part of me" that had been present throughout.

Hilgard developed this finding systematically. In hypnotic analgesia experiments, subjects who reported feeling no pain in their conscious experience could, using the automatic writing or ideomotor signalling technique, report the presence of pain through a separate channel (a channel that bypassed the conscious self but remained active. The hidden observer seemed to have access to information that was being actively withheld from conscious experience) and to be able to report it without the subject's awareness that the report was being made.

Hilgard's Neodissociation Theory

To explain the hidden observer, Hilgard developed what he called neodissociation theory. The traditional view of the mind assumes a unified executive, a single control centre that directs all behaviour and has access to all significant mental contents. Hilgard's hidden observer challenged this. His model proposed that the mind is organised as a set of partially independent subsystems (what he called cognitive control structures) each with its own processing capacity, memory, and capacity for limited executive function.

In normal consciousness, these subsystems are coordinated under a central executive that integrates their outputs into unified experience. Hypnosis, on this model, creates what Hilgard called an "amnesic barrier", a functional disconnection between certain subsystems and the central executive. The dissociated system continues to process, register, and potentially act on information. But that information does not cross the barrier into the executive's domain, and therefore does not appear in conscious experience.

The hidden observer is the dissociated subsystem communicating around the amnesic barrier through an alternative channel, the automatic ideomotor signal that bypasses the executive's usual output mechanisms. It is not a separate person or a homunculus. It is a subsystem that retains access to information that the hypnotically-modified executive has been instructed not to register.

The sociocognitive challenge (advanced most forcefully by Nicholas Spanos) deserves serious engagement. Spanos demonstrated that the characteristics of the hidden observer can be shaped by the wording of the suggestion: when subjects are told the hidden observer has more pain than the hypnotised part, the hidden observer reports more pain; when told it has less, it reports less. This suggests the hidden observer's reports are not simply veridical readouts from a dissociated processing system but are partly constructed by suggestion, just as hypnotic experiences generally are.

The resolution is probably not an either/or. Neuroimaging provides objective evidence that hypnosis produces genuine changes in sensory processing, changes that are not simply produced by non-hypnotic role-playing. But those genuine changes are always implemented within a social and narrative context that shapes their specific content. The hidden observer may be genuinely dissociated in structure while being shaped in content by the hypnotic context. Both things can be true.

The Architecture
of Suggestion

How does a suggestion (a cluster of spoken words) produce a measurable change in neural processing, a genuine alteration in sensory experience? This is the mechanistic question at the heart of hypnosis research, and the answer that is emerging connects directly to the predictive processing architecture described in Artifact II.

The key insight is that suggestion does not work by producing belief through argument. It works by supplying the generative model with a high-precision input that is then processed as if it were sensory data. When an expert hypnotist tells a highly susceptible subject "your arm is becoming too heavy to lift," the words do not convince the subject's rational mind that the arm is heavy. They supply the generative model with a prediction ("arm is heavy, lifting is impossible") that the motor system then attempts to verify. The resulting failure to lift the arm is not a performance. It is the motor system failing to overcome a top-down prediction that it was not expecting to contradict.

The Four-Stage Model of Hypnotic Response

• 01 Absorption and attentional narrowing. The induction procedure (typically involving relaxation instructions, focused attention on a single point, and the establishment of an expectational frame) produces a narrowing of the attentional field and an increase in the precision assigned to the hypnotist's voice as a source of information. External stimuli become less salient. The internal commentator's credibility as an evaluating authority begins to reduce. The critical faculty's prior is loosened.
• 02 Reduction of critical evaluation. The hypnotic state involves reduced activity in the conflict-monitoring regions (particularly the dorsal ACC) that would normally flag the implausibility of suggestions. The generative model's evaluation of incoming suggestions against a reality-testing prior is impaired. Suggestions are processed as predictions rather than as propositions to be assessed. This is the functional equivalent of temporarily lowering the precision assigned to the reality-checking prior.
• 03 Top-down prediction override. With the evaluative gate reduced in precision, the suggestion generates a top-down prediction that propagates down the cortical hierarchy with unusual force. In visual suggestions, this reaches the visual cortex. In pain suggestions, it modulates somatosensory and cingulate processing. In motor suggestions, it conflicts with proprioceptive reality-testing in ways that produce the characteristic "involuntary" quality, the subject's report that something is happening to them rather than being done by them.
• 04 Response and the involuntary quality. The completed hypnotic response is experienced as involuntary because the initiation of the response is not tracked by the normal volitional monitoring system. The arm floats upward, but the experience is not "I am lifting my arm." It is "my arm is rising." This is the same dissociation between initiation and conscious experience described in the Libet experiments. The unconscious motor preparation precedes (and operates independently of) the conscious narrative of agency. In hypnosis, this structure is made phenomenologically explicit rather than being inferred from reaction times.

Ideomotor Action: The Body as Signal

William B. Carpenter identified ideomotor action in 1852: the principle that imagining an action tends to produce small, unconscious movements in the relevant body part (movements that are the preliminary stages of the action without the inhibitory control that ordinarily prevents their completion. The Chevreul pendulum effect) where a pendulum held in the hand begins to move "on its own" when the subject concentrates on imagining it moving in a particular direction, is ideomotor action made visible.

Ideomotor action is the underlying mechanism of Erickson's arm levitation inductions, of the Ouija board phenomenon, of dowsing, and of certain forms of automatic writing. In each case, the imagination generates weak motor activation that, under the right conditions, propagates to actual movement, while the conscious narrator fails to recognise itself as the cause, experiencing the movement as external or autonomous.

This is not trivial. Ideomotor action demonstrates that the motor system has a direct channel to imagination, one that bypasses the intentional command structure that the conscious mind believes itself to be operating. The thought produces the movement. The narrator does not know it did so. This is the micro-scale version of the Libet result, visible in real time, without laboratory equipment.

Hypnosis and the
Predictive
Processing Bridge

The predictive processing framework described in Artifact II provides the most coherent mechanistic account of hypnosis currently available. The bridge between the two is precise and illuminating in both directions, hypnosis makes predictions about what predictive processing should produce, and predictive processing explains why hypnotic phenomena take the specific forms they do.

Hypnosis as Precision Manipulation

Anil Seth and colleagues have proposed that hypnotic induction can be understood as a systematic manipulation of precision weighting. The induction procedure does several things simultaneously that all act on the precision structure of the generative model. Relaxation and focused attention reduce the precision of signals from the external sensory environment (external distractions are assigned less weight. The establishment of an authoritative relationship with the hypnotist increases the precision assigned to that voice as a source of predictions. The suggestion of trance) the expectation that something special is going to happen, reduces the precision of the reality-testing prior that would ordinarily evaluate incoming suggestions against current experience.

This model makes specific, testable predictions that have largely been confirmed. It predicts that hypnotic phenomena should be modulated by the same neuromodulatory systems that control precision weighting generally (particularly acetylcholine, which is strongly implicated in attentional gain control, and dopamine. It predicts that conditions that naturally reduce sensory precision) monotonous environments, sensory deprivation, the hypnagogic state between waking and sleep (should facilitate hypnotic induction. Both predictions are supported. It predicts that the effectiveness of hypnotic suggestion should scale with the suggested experience's position in the cortical hierarchy) suggestions targeting lower-level processing (primary sensory cortex) should require higher hypnotic depth than suggestions targeting higher-level processing (emotional or motivational systems). This prediction is less thoroughly tested but is consistent with clinical experience.

Why Highly Susceptible Subjects Are Different

The predictive processing account offers an explanation for why hypnotisability varies. People with high hypnotic susceptibility appear to have a generative model that, at baseline, assigns higher precision to internally-generated signals relative to incoming sensory data. Their top-down predictions are more vivid, more strongly felt, more capable of overriding sensory input. This is the same parameter that underlies vivid visual imagery, the tendency toward absorption, and the phenomenon of strong synesthesia.

In highly susceptible subjects, the gap between the "normal" precision configuration and the "hypnotic" configuration is smaller (less work is required to shift the balance. The induction is rapid. The phenomena are dramatic. Low-susceptibility individuals have a generative model more firmly anchored in sensory input) it is harder to raise the precision of internally-generated predictions to the level required for vivid hypnotic phenomena.

Clinical Evidence:
What Actually Works

The transition of hypnosis from stage entertainment and theoretical curiosity to evidence-based clinical tool has accelerated considerably in the past two decades. The evidence base is now sufficient to support specific claims, and it is also honest enough to acknowledge significant gaps.

Pain: The Strongest Evidence

The clinical evidence for hypnotic analgesia is the strongest in the field. A 2000 meta-analysis by Montgomery, DuHamel, and Redd, covering 18 studies, found a moderate-to-large effect size (d = 0.67) for hypnosis in reducing clinical pain, with highly hypnotisable subjects showing larger effects. A 2016 Cochrane-style review by Adachi and colleagues confirmed these findings across acute and chronic pain conditions. For highly hypnotisable subjects specifically, the effect sizes rival those of moderate-dose opioid analgesia.

Clinical applications in pain management include: procedural pain in burn patients (where hypnosis has been combined with virtual reality to produce substantial reductions in pain medication use), labour and childbirth pain, chronic pain management, and dental procedures in patients with needle phobia or anaesthetic contraindications. The evidence for these applications is not uniformly strong, effect sizes are larger in the laboratory than in the clinic, and the heterogeneity of hypnotic depth across patients is a persistent methodological problem.

Irritable Bowel Syndrome: An Unexpected Success

One of the most robustly replicated clinical applications of hypnosis is gut-directed hypnotherapy for irritable bowel syndrome (IBS). The original trials by Peter Whorwell and colleagues at Manchester Royal Infirmary in the 1980s showed response rates of over 70% in refractory IBS patients, patients who had failed other treatments. Multiple replications, including a large randomised controlled trial published in The Lancet Gastroenterology & Hepatology in 2016, have confirmed these findings. The effect appears to be durable, follow-up studies show sustained benefit at 5 years in the majority of responders.

IBS is a disorder of the gut-brain axis, of the bidirectional communication between the central nervous system and the enteric nervous system. Its symptoms are driven partly by altered central processing of visceral signals: abnormal amplification of gut sensations that are not meaningfully painful in most people. Hypnotic intervention that reduces the precision assigned to those afferent signals (reducing the brain's prediction that gut sensations are dangerous) fits precisely into the predictive processing account of what hypnotic analgesia does.

Psychotherapy Integration: EMDR, Parts Work, and the Erickson Legacy

Ericksonian hypnotherapy has been indirectly absorbed into several evidence-based psychotherapy modalities. EMDR (Eye Movement Desensitisation and Reprocessing) incorporates the dual-attention structure (the conscious witnessing of past material while maintaining present-moment grounding) that Erickson used in trauma processing. Internal Family Systems therapy's "parts" model maps directly onto the neodissociation architecture. Acceptance and Commitment Therapy's techniques for "defusion" (observing thoughts rather than identifying with them) structurally parallels Erickson's instruction to the conscious narrator to step back and allow the unconscious to work.

The Narrator
Loses the Pen

We arrive at what hypnosis actually proves. Not through argument (through demonstration. A person sits in a chair. They are awake and intelligent. They are told their arm will rise without their willing it) and it rises. They are told they will feel no pain (and they feel no pain, verifiably, in their anterior cingulate cortex. They are told they will not remember the session) and they do not remember, yet could, through an alternative channel, demonstrate that the memory exists. They are told they will perceive a colour that is not there, and their visual cortex processes it as if it were there.

In each case, the conscious narrator (the "I" who believes itself to be the author of experience) is watching something happen that it did not initiate, does not understand the mechanism of, and cannot prevent. The experience is vivid and first-person. The narrator is present throughout. It did not produce what it is experiencing.

This is not the unconscious doing something mystical. It is the normal architecture of the mind, made temporarily legible. The arm rises every time you reach for a cup, by the same motor machinery, with the same absence of real volitional initiation. The conscious narrator claims credit. Hypnosis removes the claim and leaves the mechanism visible.

The Significance for Self-Knowledge

Hypnosis forces a specific revision of the self-concept. The intuition that "I am the author of my experience" is not just philosophically unjustified, as the Libet experiments suggest. It is demonstrably wrong in real time, in waking experience, without any measurement equipment. The subject under hypnosis is not in an exotic state that doesn't apply to normal life. They are in a state where the normal architecture is being operated on deliberately, making its structure visible.

The practical implication is one of significant epistemological humility. If the narrator can be so thoroughly separated from authorship under hypnosis (if the experience of willing, of feeling, of remembering can be so cleanly decoupled from the narrator's claims) then the narrator's claims in ordinary life deserve to be treated with corresponding scepticism. When you believe you chose something freely, when you believe you feel a certain way for a certain reason, when you believe you remember something accurately, you are working from the narrator's report. That report is generated by a system that does not have transparent access to its own processes.

Erickson's lifelong insistence that the unconscious is not an enemy but a collaborator acquires new force here. The appropriate response to the discovery that the narrator is not in charge is not dismay. It is a reorientation of who needs to be addressed. Talking to yourself is rarely effective precisely because the conscious narrator is talking to itself, to the system that has the least access to what is actually driving behaviour. The more powerful intervention is to talk to the unconscious directly: through the body, through imagery, through narrative, through the structures of emotion and expectation that bypass the evaluating mind entirely.