This page exists so that nothing written elsewhere on this site has to be taken on trust. It sets out what body-oriented work can currently show, how strong each piece of it is, and where the honest answer is that we do not know.

Two grades are used. Documented — replicated, with a citable source. Contested — both sides have real evidence and the question is open. Claims that would grade lower than that are left out entirely rather than softened.

The calibration point, and it is not a somatic one

What holds

  • EMDR is the best-evidenced treatment for PTSD: repeated meta-analyses, first-line in both NICE and APA guidance, comparable to trauma-focused CBT. Documented.
  • Dismantling studies — ten of them, comparing full EMDR against EMDR with the eye movements removed — give a pooled effect of g ≈ −0.04 with no heterogeneity. Removing the supposedly active ingredient does not measurably change the outcome.
  • So the best-evidenced trauma treatment in the field works, while its own account of why it works is probably wrong.

Why it stands first on this page

  • “Real effect, contested mechanism” is not a special weakness of body-oriented work. It occurs at the top of the evidence hierarchy.
  • Anyone using that pattern as an argument against this kind of practice has to aim the same argument at EMDR.

What is measurable after trauma

What holds

  • Traumatised people show repeatedly replicated differences in autonomic reactivity, stress-hormone regulation and startle habituation. Documented.

What it does not say

  • It does not establish that traumatic memory is stored somatically, separately from ordinary memory, and must therefore be somatically released. That is a further claim, and it is the subject of the next section.

“The body keeps the score”

What holds

  • As a clinical metaphor it is usable, and both clients and therapists work with it productively.
  • As a neuroscientific mechanism it is not demonstrated. In popular writing it is dressed in the language of settled science. Contested.

What it rests on

  • Fitzgerald et al., BJPsych Bulletin, 2025/26 — a content analysis coding 122 discrete claims from the book (42 neurobiological, 51 on treatment efficacy, 29 other). Support for the strongest claims is inconsistent relative to the confidence with which they are stated. The paper’s frame is that the public narrative outran the science, not that the phenomena are absent.
  • McNally, Remembering Trauma (Harvard University Press, 2003), on the 1994 theory as “a concept in search of a phenomenon”. Sharp criticism, but from inside trauma research.
  • Otgaar, Howe, Patihis, Merckelbach, Lynn, Lilienfeld, Loftus, “The Return of the Repressed”, Perspectives on Psychological Science 14(6), 2019.

The other side, at full strength

  • For some difference in encoding: traumatic intrusions are disproportionately sensory and fragmentary and poorly anchored in time, which is compatible with hippocampal suppression and amygdala-driven encoding.
  • Against the strong version: all emotionally intense memories show some fragmentation, and body-memory claims have historically been used to justify recovery techniques that produced false memories.
  • This belongs on the page as a live scientific dispute, not as a settled debunking.

Where the evidence is strong — and it is the category, not the brand

What holds

  • Body psychotherapy as a category: 18 randomised controlled trials across diagnoses, medium effect sizes on psychopathology and distress against controls. Documented. Rosendahl, Sattel & Lahmann, Frontiers in Psychology, 2021.
  • Trauma-sensitive yoga: one well-designed trial with an active control, which is methodologically better than a waiting list. van der Kolk et al., Journal of Clinical Psychiatry 75(6), 2014 — N=64 women with chronic treatment-resistant PTSD, 10 sessions; 52% no longer met criteria against 21% of controls; d = 1.07 against 0.66.

What that does not license

  • Somatic Experiencing: two randomised trials against waiting-list controls (Brom et al., Journal of Traumatic Stress 30(3), 2017, N=63, d ≈ 0.94–1.26), plus a scoping review from within its own field — Kuhfuß et al., European Journal of Psychotraumatology 12(1), 2021: 16 usable papers out of 83, only two quality trials. That self-criticism is worth more than any outside attack.
  • Sensorimotor Psychotherapy, Hakomi and the Reichian bioenergetic lineage have effectively no direct trial base.
  • “Body-oriented psychotherapy has measurable effects” and “this named method is evidence-based” are different sentences. This site keeps them apart.

Interoception — the one case here that can be checked, and what checking it showed

What holds

  • Self-reported attunement to the body and measured performance at reading it come apart. Documented — and it does not hang on any one instrument. They are uncorrelated where both measures are temporally stable (Ferentzi et al., 2017: r = .06, with test–retest of .60 and .73); they dissociate inside a single sample; and they dissociate even where performance genuinely is higher.
  • The sharpest single case, and it is in our own population: experienced meditators rated their performance as superior and found the task easier than controls did, with no accuracy advantage at all. Khalsa et al., 2008.
  • Disturbed interoception is repeatedly documented in PTSD, anxiety, depression, autism, eating disorders and somatisation.

The trap, and it is ours

  • “I am very attuned to my body” is a self-rating. “I detect my own heartbeat” is a performance. Somatic training cultures routinely take the first as evidence of the second, and the evidence says that inference does not hold.
  • Same shape as everything else on this page — a category that looks like a measurement and is a self-rating. The difference is that here it can be checked.

And the checking instrument is itself contested — which this page says rather than skips

  • The standard heartbeat counting task carries a serious validity critique: over 95% of scores are under-counts, and the score is structurally bound to the participant’s own heart rate (Zamariola et al., 2018, N=572). Tightening the instruction to “count only the beats you actually feel” roughly halves the scores (Desmedt et al., 2018).
  • Two published comments argue that part of that critique is an artefact of scoring a ratio variable (Zimprich et al., 2020; Ainley et al., 2020). Both sides accept the under-counting. The contamination-by-instruction charge has not been answered.
  • Harder still: heartbeat counting and heartbeat detection are each internally reliable and yet uncorrelated with each other (Ring & Brener, 2018). There is no single quantity called objective interoceptive accuracy.
  • A systematic review of the field’s own measurement situation (Desmedt et al., 2023) calls high validity “potentially elusive”. The field has not converged on a standard, and says so itself.

What this page therefore does not say

  • That interoception has three validated dimensions. The three-part model — accuracy, sensibility, awareness — is not a confirmed structure. Its only direct empirical test confirmed the hierarchy only in part and found the result depended on which task was used (Forkmann et al., 2016); its critics hold that revision is necessary (Murphy, Catmur & Bird, 2019); and its own senior author moved to a broader multi-axis framework in 2022 (Suksasilp & Garfinkel), saying that assessing the dimensions in isolation fails to capture what interoception is. To cite the 2015 trio as settled taxonomy is to cite an intermediate step.
  • That improving interoceptive accuracy causally improves emotion regulation. A systematic review of trials (Psychiatry and Clinical Neurosciences, 2023) found around two thirds beat controls on interoceptive measures — which is not the same as symptom improvement.

The other side, at full strength

  • Dose may be the issue rather than the claim. A nine-month contemplative training programme did raise objective accuracy — small, but real, and growing with time on task: d = 0.17 at six months, d = 0.27 at nine. Its authors say explicitly that eight-week courses are too short. Bornemann & Singer, 2017.
  • Channel may be the issue too. Cardiac accuracy does not generalise across bodily channels, so a cardiac null may be a mismatch of channel rather than an absent trained skill. That is a hypothesis, not a finding.

And the uncomfortable one

  • The clearest objective advantages found so far belong to training that is not contemplative at all — dancers and musicians. The authors of one of those studies note that attempts to find the same advantage in yoga and meditation groups were not successful.

Polyvagal theory

What holds

  • The neurophysiological and evolutionary foundations are under live dispute. The dispute is not primarily about clinical usefulness, and that distinction matters. Contested.

What it rests on

  • Porges, “Polyvagal Theory: Current Status, Clinical Applications, and Future Directions”, Clinical Neuropsychiatry 22(3), 2025.
  • Grossman and 38 co-authors, “Why the Polyvagal Theory Is Untenable”, same journal, 2026 — among them researchers Porges had previously cited in support. Points at issue: the evolutionary claim; the anatomical separability of the pathways, since freezing in mammals is mediated primarily via the nucleus ambiguus — the “ventral” system — and the deep bradycardia the dorsal account predicts is typically not observed in dissociative freezing; and the measurement claim about RSA. Porges replied in the same issue that the critique targets a reconstruction of his theory.
  • Earlier: Grossman & Taylor, Biological Psychology 74(2), 2007.
  • These exchanges are recent enough that secondary commentary is still settling.

What survives even the hardest reading — and what does not

  • Survives: people continuously and below awareness read safety and threat in another’s face, voice and posture, and this changes physiological state. That does not require the specific anatomy to hold. So does the framework as descriptive language for a client’s experience — mobilised, shut down, calm and in contact.
  • Does not: telling a client they are “in a dorsal vagal state” as a physiological diagnosis, or resting a treatment rationale on the literal truth of the anatomical claims.

What the practitioner sees and measurement misses

What holds

  • Differences in outcome between bona fide psychotherapies are small. Therapist effects, alliance quality, client expectation and empathy explain substantially more variance than the specific technique. Documented. Wampold, World Psychiatry 14(3), 2015; Wampold & Imel, The Great Psychotherapy Debate, 2nd ed., 2015.
  • Which means any method may show a real effect largely because it is a competently and warmly delivered structured intervention giving the client a coherent rationale and a working relationship — not necessarily because its own technique is mechanistically right.

The other side, and it is serious

  • Cuijpers and colleagues: a meta-analysis of dismantling studies found that adding the specific component gives a small but real effect which is larger at follow-up, not smaller — which does not sit well with a purely common-factors reading. A variance decomposition in depression puts roughly a third with extratherapeutic factors, about half with non-specific ones, and about 17% specifically with technique. Real but modest; not zero. Contested.
  • And a methodological point that is not a rhetorical shield: trials require manualised, time-limited, single-diagnosis treatment with adherence control in homogeneous samples, which differs substantially from how therapy is actually practised, especially with complex and comorbid presentations. “The evidence for X is thin” therefore does not automatically mean “X does not work”. Thin evidence is often a fact about what is expensive and difficult to standardise and fund.

The honest summary

Something real happens when body-oriented work helps traumatised clients. Autonomic dysregulation is measurable, interoceptive processing is genuinely different, and several imperfect studies across different methods point the same way. But the explanatory stories offered to justify them — literal storage in the body, a cleanly separable dorsal–ventral switch — are weaker than the clinical observations they are meant to explain.

Many indications, weak to moderate models, and practitioner expectations stronger than current evidence licenses. That gap is better named for the reader than settled artificially in either direction — especially since the same gap appears at EMDR, which is the best-evidenced treatment the field has.

Drawn from two internal research dossiers of 13 September 2026 — one on somatic evidence, one on the measurement of interoception (v0.1). Claims graded unverified in those dossiers do not appear on this page. Where a source is recent enough that commentary is still settling, the page says so.