# blog.jaye.ch $ cat summary.md▊

Meditation can harm — and it does so invisibly

two series: the mechanism (measured), and the implications (arguments).

The mechanism

// a measured model, then a falsifiable prediction · 4 of 4 published
01

The Meditation Failure Mode

measured · The mechanism end to end — the runaway, the cliff, and why the state it lands in cannot see itself.

02

The Damping Dial

prediction · Who is at risk: anxiety-proneness read as low damping, and the cheap test — tolerance, not settling time — that would settle it.

03

What Actually Works

measured · The rescue result: a cheap fixed floor beats an elaborate one — and re-checking your support is what destroys it.

04

Recovery Is Not Immunity

measured · A full rescue buys the recovered state nothing against the next trigger — the threshold restored is the threshold the next episode drains.

The implications

// arguments built on cited literature — not measurements · 6 of 6 published
01

Manufacturing the Collapse

argument · The same collapse run deliberately — Schein, Lifton, Stein — and the inversion: the same three guardrails, two of them flipped.

02

The Sacred Science

argument · Why “the observer collapses reality” and “union with God” are one move — and how the frame steers the collapse to an owned destination.

03

The Empty Leader

argument · The person at the top: leader and follower share one dissociation pointing opposite ways — the group as his externalized self.

04

What the Traditions Knew

argument · Seven contemplative traditions, with no contact with each other, named this failure and prescribed the same five safeguards — primary-source passages.

05

The Safeguards

argument · What the traditions that survive the collapse wrote down — ethics first, metered pace, plural community, a teacher who points past himself, discernment as procedure — and why the clinical field is re-deriving the same list.

06

The Label

argument · "Cult" is a category that does harm — to the people it is meant to protect — and the sources this series leans on helped justify kidnapping. Why it names mechanisms instead.

The short version

$ cat summary.md

Summary — the full post is at /meditation-harm/.

The underlying paper is published: Marshall (2026), doi:10.5281/zenodo.22943641 · record.


Meditation-related harm is not fringe. The Varieties of Contemplative Experience study built a taxonomy of meditation-related challenges and framed them as experiences “typically underreported.” An independent assessor in a controlled mindfulness trial found adverse effects in 58% of participants and negative impacts on functioning in 37% — with the lasting effects (6–14%) “associated with signs of dysregulated arousal (hyperarousal and dissociation).” A widely-cited estimate puts an adverse effect lasting more than a month at roughly one in ten people who have meditated, including after a single practice.

And then the strange part. In a recent survey of 121 people with meditation-related difficulties, 61.7% of the meditation-triggered group scored above the clinical cutoff for depersonalization — yet almost none carried a diagnosis, and the same state was described as “more welcome, pleasant, and spiritually meaningful” than the comparison group’s identical phenomenology.

The phenomenology is common. The label is rare. The state is often experienced as good. And the self-report instruments cannot reliably see it (the standard dissociation taxon false-positives at 54% in non-clinical samples).

Why does a failure mode this common stay this invisible? That is the question this post tries to answer — with a measured control-theoretic model of self-regulation.

The short version:

  • Inward attention consumes self-content. Hold it and the generator falls; the falling generator raises an error term that arms a feedback switch, which pulls attention further inward. A runaway, one direction, no built-in brake.
  • The transition is a cliff, not a slope — and it is a specific kind of cliff (border-collision fold) with no critical slowing down. Recovery time stays constant to the edge; what shrinks is tolerance to perturbation. Which means a protocol that measures only settling time will miss it entirely.
  • It does not self-reverse: reduce the drive to zero and the state persists.
  • And the mechanism for the invisibility: at the collapsed point the reasoning faculty sits at exactly its healthy baseline while the self-content generator is annihilated. The part of the system that would report the failure is the part that survives it — so “I’m fine, this is freedom” is accurate from the only vantage left. Meanwhile the act of self-monitoring is the failure driver. The sensor is destroyed by the failure it exists to detect.

That last point is the useful one. If it is right, the reporting gap will not close with better questionnaires. It closes with relocated measurement — the body, behavior, an external observer — and with the provision the traditions built and modern delivery dropped: a frame before you arrive, community around you, and pace metered, plus a holder chosen in advance who can act when the state cannot act for itself.


Read the full mechanics →

(The model is a deterministic ODE from a single case; the mapping to practice is an interpretation, not a measurement. It is offered as one falsifiable hypothesis to the people who measure these things — not as a clinical claim.)