Name: Therapy Mechanisms
Creator: Realised
Published: 2026-06-01

Summary

Psychotherapy genuinely works, but most of what makes it work is shared across every credible method — the relationship, a believable rationale, your own expectation of getting better — not the branded technique on the door; the honest exception is that for fear-based problems (phobia, OCD, PTSD), one specific ingredient, exposure, does carry its own weight, so "all therapy is the same" is as wrong as "you must find the one right method."

Why Strong

Tier 1 for: psychotherapy outperforms no-treatment, and bona-fide therapies broadly converge for common conditions. This is one of the most replicated findings in clinical psychology across decades of trials and meta-analyses.

Tier 2 for: (a) the alliance as a robust predictor of outcome (r≈.28, ~300 studies); (b) therapist effects exceeding technique effects in variance decomposition; (c) exposure as a dismantling-proven specific ingredient for fear disorders.
• NOT Tier 1 for these because: alliance/therapist findings are observational; the exposure-superiority case, while strong, is condition-specific and sits against a general literature that struggles to isolate specific ingredients.

Tier 3 for: the causal claim that any single common factor drives improvement.
• Tier 3 not higher because: per Cuijpers 2019, no common (or specific) factor meets the criteria for an empirically validated working mechanism — the evidence is correlational with documented reverse causation (early improvement → alliance) and unexcluded third variables.
• NOT lower than Tier 3 because: there is a coherent theoretical model (contextual/Frank), convergent correlational data, and partial experimental support (warmth/competence manipulations move alliance and expectations) — it is a well-motivated hypothesis, not speculation.

Net: a genuinely mixed entry. The practical guidance (start, optimise for fit and engagement, seek exposure for fears) rests on Tier 1-2 ground; the causal story is held honestly at Tier 3.

Practical takeaway

This entry is decision-support, not a self-administered protocol. "Working" here means: making a good-enough choice quickly, then judging fit early instead of agonising over the brand.

Choosing a therapist — what actually predicts a good outcome:
1. Pick a credible method you can believe in, then stop optimising the brand. For depression / general anxiety / stress, CBT, ACT, IPT, psychodynamic and person-centred are all defensible. Your belief that the approach makes sense to you is itself an active ingredient — so pick one whose rationale you find plausible rather than the one with the best marketing.
2. Optimise for the person, not the school. The therapist matters more than the modality. You are looking for someone you feel understood by, who gives you a clear, coherent account of what you'll do together and why.
3. For a specific fear — phobia, OCD, panic, PTSD — actively seek exposure-based treatment (exposure therapy, ERP for OCD, prolonged exposure / trauma-focused CBT for PTSD). Here the technique is not interchangeable; a purely supportive or insight-only approach is the wrong tool. This is the one case where "find the right method" is correct advice.
4. Discount experience-as-years and brand prestige. Years in practice barely predict outcome. A trademarked or "proprietary" modality is not, on that basis alone, better than a generic evidence-based one.

What "working" looks like — track these:
• An early sense of being understood and of having a plan — a workable alliance usually forms in the first few sessions, not after months. A persistent feeling of not being understood, or no agreed plan, is a real signal — not something to endure indefinitely.
• Some movement by around 6–8 sessions for common conditions. Many people show measurable early response in the first month-plus of regular sessions. Lack of any movement by ~8 sessions is a reasonable trigger to discuss the plan, consider a rupture-repair conversation, or switch therapists — not to switch schools of therapy.
• For exposure work: "working" specifically looks like your anxiety rising during a planned exposure and then coming down on its own without escape, and the next exposure starting lower. If treatment for a fear problem never involves approaching the feared thing, that is a flag.

Response windows. Acute distress can lift early (remoralisation). Structural change (entrenched patterns, personality-level work, complex trauma) is a longer horizon — months, sometimes longer — and slower movement there is expected, not failure.

The alliance-repair point. Ruptures (feeling misunderstood, a session that lands badly) are normal and are not automatically a reason to quit. Whether they get named and repaired is more informative than whether they happen. A good therapist can take "that didn't land for me" without defensiveness.

Where therapy sits relative to the rest of Realised. Therapy is one lever among several. Exercise, sleep, social connection and reduced chronic stress load have their own substantial effects on mood and anxiety (see Cross-Pillar) and are often the right first or concurrent moves, not a replacement for therapy when therapy is indicated.

Evidence detail

Why This Entry Exists

A Realised user weighing therapy is drowning in marketing. One clinic sells CBT as "the evidence-based gold standard." Another sells a proprietary trademarked modality (EMDR, a named "neuro" protocol, a branded app programme) as the breakthrough the others lack. A friend swears by their psychodynamic analyst; a podcast says ACT is the future. The user reasonably concludes that choosing the wrong method means choosing failure — and either freezes, or spends months chasing a brand instead of getting started.

This entry exists to defuse that paralysis with the actual evidence, which is both reassuring and counter-intuitive: across the conditions most people seek help for (depression, anxiety, general distress), bona-fide therapies delivered competently produce broadly similar outcomes. The biggest measurable levers are not the technique but the relationship between you and the therapist, your belief that the approach makes sense and will help, and — large but rarely advertised — which individual therapist you see. That means the practical question is not "which school of therapy?" but "is this a person I can work with, who has a coherent plan I find credible, and am I willing to do the work?"

It exists equally to stop the opposite error. The reassuring "it's all just the relationship" story gets over-sold too, and it is wrong in two important ways. First, for fear-based disorders the technique genuinely matters: exposure is a real active ingredient and a warm chat without it underperforms. Second, "the alliance is what heals you" is stated as established fact far more often than the evidence allows — the alliance reliably predicts good outcomes, but predicting is not the same as causing, and the field has not yet shown it is the cause.

What bad advice does this protect against?
• "You picked the wrong type of therapy, that's why it didn't work" (usually false — fit and engagement matter more than school).
• "This trademarked method is uniquely powerful and the others are obsolete" (the evidence for across-the-board superiority is weak and inflated by who ran the trial).
• "Therapy is just paying someone to be your friend / it's all placebo" (false — it produces real, measured effects beyond no-treatment and beyond pill placebo).
• "The alliance is literally what cures you, so technique is irrelevant" (over-stated — alliance is a strong correlate, not a proven cause, and technique is not irrelevant for fear disorders).

Evidence

1. Therapy works, and bona-fide therapies broadly converge (Tier 1).
The single most replicated finding in psychotherapy research is that established, credible therapies intended to be therapeutic produce roughly comparable outcomes for the common presenting problems, and all clearly beat no-treatment and waitlist. This is the empirical core of the "dodo-bird verdict" (everyone has won and all must have prizes). For depression specifically, large bodies of comparative work (Cuijpers and colleagues' programme of meta-analyses) find that CBT, interpersonal therapy, problem-solving therapy, behavioural activation and others differ little in effect once methodological confounds are accounted for. (Funding: this comparative literature is dominated by academic / government / nonprofit-funded groups; Cuijpers' work is academic, VU Amsterdam.)

2. The relationship is the most reliable correlate of outcome (Tier 2 as prediction, Tier 3 as cause).
The therapeutic alliance — the bond plus agreement on goals and tasks — is the most-studied common factor. The definitive synthesis (Flückiger, Del Re, Wampold & Horvath, 2018, Psychotherapy; ~300 studies, >30,000 patients) puts the alliance–outcome correlation at r ≈ 0.28, i.e. roughly 8% of outcome variance, stable across disorders, measures and therapy types. The authors argue the association is not merely an artefact of intake severity. (Academic / independent.) But — see the critique below — a correlation of this kind does not establish that the alliance causes the improvement.

3. Variance decomposition: technique is a small slice; the therapist is a bigger one (Tier 2).
Wampold's contextual-model decomposition (synthesised in The Great Psychotherapy Debate, Wampold & Imel) attributes only a small share of total outcome variance to the treatment package at all — on the order of ~13% — with the bulk attributed to patient and extra-therapeutic factors. Within the treatment-attributable slice, the specific technique is consistently estimated at 0–10% of variance, while therapist effects (who you see, not what school they belong to) account for roughly 5% of total variance — larger than the technique effect. Translated: which individual sits in the chair matters more than which manual is on the shelf. (Academic / independent; Wampold is the leading common-factors advocate — note his allegiance when reading his numbers.)

4. Expectations / belief carry real weight (Tier 2).
A credible treatment rationale and the patient's positive outcome expectation are reliable predictors of improvement, and experimental manipulations of therapist warmth/competence shift both expectations and alliance. This is the legitimate, non-derogatory sense of "placebo": belief and expectancy are active components of psychological treatment, not noise to subtract. (This connects directly to belief_effects_and_honest_framing.) (Academic / independent.)

5. The honest exception — exposure is a genuine specific ingredient for fear-based disorders (Tier 1-2).
This is where "all therapy is equal" breaks, and Realised states it plainly. Dismantling studies (remove one component, see if outcome drops) consistently show that for specific phobia, panic, OCD and PTSD, exposure is the active ingredient: relaxation-only or supportive conditions underperform, and removing relaxation from an exposure package does not hurt outcomes while removing exposure does. Specific phobia responds to exposure in a large majority of completers, often with exposure as the only treatment needed. Bell, Marcus & Goodlad (2013, J. Consulting & Clinical Psychology) and Ahn & Wampold (2001, J. Counseling Psychology) — the two canonical component meta-analyses — found, overall, little evidence of specific-ingredient advantages across psychotherapy as a whole, yet exposure for anxiety/fear disorders is the clearest, replicated case where a specific component does carry the load. Both things are true at once: in general, components are hard to distinguish; for fear, exposure is the exception. (Academic / independent.)

6. ACT, psychodynamic, humanistic — all bona-fide, broadly comparable (Tier 1-2).
ACT shows medium effects vs treatment-as-usual and large effects vs waitlist across a wide range of conditions, working (per mediation studies) through increased psychological flexibility. Modern psychodynamic therapy and humanistic/person-centred therapy also outperform no-treatment and, in head-to-head comparisons, sit within the same broad band as CBT for common conditions. No single school has earned a general claim of superiority over the others for everyday depression and anxiety. (Academic; ACT's evidence base carries some developer-allegiance — see Industry Bias.)

7. The confound that inflates apparent winners — researcher allegiance (Tier 2).
When one therapy appears to beat another, a large share of the gap is explained by researcher allegiance: trials run by advocates of a method tend to find that method superior. Munder et al. (2013) estimate the allegiance–outcome association at roughly r ≈ 0.26 on average, stronger in lower-quality trials. This is the single most important reason to distrust "Method X beats Method Y" headlines, and it cuts against every brand equally.

Mechanism

There is no one mechanism, and pretending there is would itself be the error this entry guards against. The honest model has three layers.

Layer 1 — the shared machinery (common factors). Frank & Frank's classic account, which the modern contextual model formalises, holds that all credible therapies share four elements: (a) an emotionally charged, confiding relationship with a helper; (b) a healing setting that raises expectation of help; (c) a rationale / explanatory myth that makes the person's distress intelligible; and (d) a ritual / procedure the person actively participates in that follows from that rationale. On this model the specific content of the rationale and ritual matters far less than that they are coherent, believed, and acted on. Improvement flows from remoralisation (restored hope), a new and more workable understanding of the problem, and active mastery experiences — not from the branded technique per se. This is why a competent CBT therapist and a competent psychodynamic therapist, doing very different things, can land patients in the same place.

Layer 2 — the specific machinery (where technique earns its keep). For fear-based disorders the mechanism is more concrete and the technique is not interchangeable. Exposure works by extinction learning / inhibitory learning: repeated, structured contact with the feared situation without the predicted catastrophe builds new, competing safety associations that inhibit the old fear response. Relaxation or insight alone does not reliably create that new learning; the encounter with the feared thing, done in a graded and tolerable way, is the mechanism. This is a genuine active ingredient, and it is why "any warm conversation" is not adequate treatment for a phobia or for OCD.

Layer 3 — the therapist as the variable. Therapist effects exceed technique effects in the variance decomposition, and — importantly — years of experience do not predict better outcomes (effectiveness is flat or even slightly declines with time-in-career unless the therapist deliberately works to improve). What seems to distinguish more effective therapists is interpersonal skill under pressure, the ability to form and repair the alliance, and engagement in deliberate practice. The mechanism here is not a technique at all; it is a person-level capacity to keep the working relationship intact while the difficult work gets done.

The causality caveat — load-bearing. It is tempting to read "alliance correlates with outcome at r≈.28" as "build the alliance and you cause the cure." The evidence does not support that step. Cuijpers, Reijnders & Huibers (2019, Annual Review of Clinical Psychology) review the common-factors literature and conclude that no common factor — and no specific factor either — currently meets the criteria to be called an empirically validated working mechanism. The reasons are precise: (1) the research is almost entirely correlational; (2) there is strong evidence of reverse causation — early symptom improvement drives a better later alliance at least as much as the reverse, so a good alliance is partly a consequence of getting better, not only a cause; (3) third variables (a patient who is improving for unrelated reasons rates both their alliance and their outcome more highly) can manufacture the correlation. To call something a causal mechanism you need temporal precedence, a dose-response relationship, exclusion of third variables, supportive experimental manipulation, and a coherent theory — and the alliance literature does not yet clear that bar. So Realised's mechanism claim is: the alliance is the best-established signpost of therapy going well; it is not proven to be the engine.

Risks And Contraindications

• Therapy can have adverse effects. A minority of patients deteriorate during treatment. Common drivers: poor fit / persistently weak alliance, a therapist out of their depth with the presenting problem, or a poorly-paced intervention (e.g. exposure or trauma processing pushed too fast). Deterioration is a reason to change something — pace, plan, or therapist — not to assume therapy "doesn't work for you."
• Exposure done badly can sensitise rather than extinguish. Flooding someone faster than they can tolerate, or letting them escape mid-exposure, can reinforce fear. Exposure should be graded, collaborative, and within tolerance — this is why it is a clinician-delivered technique, not a self-help instruction in this entry.
• Trauma processing can be destabilising. Surfacing traumatic material without adequate stabilisation and support can worsen symptoms. Trauma-focused work belongs with a clinician competent in it.
• "It's all the relationship" can become an excuse to avoid effective technique. For a fear disorder, a warm therapist who never does exposure is providing support, not treatment. Comfort is not the same as recovery here.
• Crisis / safety override. Active suicidality, psychosis, or risk of harm needs urgent clinical / emergency care, not a deliberation about therapy modality. This entry is about elective help-seeking, not crisis triage.
• Not a substitute for medical assessment. Symptoms that could be physical (thyroid, anaemia, sleep disorder, medication effects) warrant medical work-up alongside or before psychological attribution.

Controversy

The dodo-bird verdict: do common factors or specific techniques drive change? This has been argued for fifty years and is not settled.

Position A — Common factors (contextual model; Wampold, Frank, the dodo-bird camp). Bona-fide therapies produce equivalent outcomes; the active agents are the relationship, expectation, rationale and ritual shared by all of them; specific techniques contribute little. Evidence: the convergence of outcomes across methods, the robust alliance–outcome correlation, therapist effects exceeding technique effects, and the collapse of apparent method-superiority once researcher allegiance is controlled.

Position B — Specific ingredients / medical model. Different disorders have different mechanisms and call for different, identifiable active treatments; treating therapy as interchangeable is clinically careless. Evidence: exposure is a dismantling-proven specific ingredient for fear disorders; some head-to-head reviews find a CBT advantage for certain anxiety/depressive presentations; "equivalence" can be an artefact of underpowered trials and outcome measures too blunt to detect real differences; and — a sharp methodological critique — common and specific factors are themselves correlated, so treating them as independent and partitioning variance between them is statistically unsound and biases the picture toward "common factors."

The funding / bias dimension (cui bono). This is not a neutral academic dispute; money sits on both sides.
• "Our specific branded method is superior" profits a large ecosystem: trademarked-modality certification and training, manualised-treatment licensing, brand-named therapy apps and digital programmes, and CME/workshop economies. Researcher allegiance (r≈.26) is the visible fingerprint of this incentive in the literature — advocates' trials favour advocates' methods.
• **"It's all common factors" is not incentive-free either.** It is convenient for generalist providers, low-cost / lay-counsellor delivery models, and insurers who would prefer to fund the cheapest interchangeable option. It can also be used to justify not training clinicians in harder specific skills like exposure — which would harm exactly the fear-disorder patients for whom technique matters.
• So both poles have a paymaster. The honest reading is not loyalty to either camp.

Realised Position. Both extremes are wrong, and the truth is locatable. (1) Therapy works and is worth doing — Tier 1. (2) For most everyday distress, which credible method matters far less than which therapist, what relationship, and your own engagement and belief — so don't let brand-shopping delay starting. (3) For fear-based disorders, the specific technique (exposure) genuinely matters and should be sought — so don't let "it's all the relationship" talk you out of effective treatment. (4) The popular claim that the alliance causes recovery outruns the evidence; it is the best signpost we have, not a proven engine. Realised coaches the user toward fit, engagement, and the right tool for fear problems, while being honest that the causal mechanism of general talk therapy is not nailed down.

Cross-Pillar Connections

• Mental (mental_social_connection): the therapeutic relationship is, mechanistically, a specialised instance of the broader human-connection effect on mental health; the same relational machinery is at work, which is part of why supportive relationships outside therapy also move outcomes.
• Mental (belief_effects_and_honest_framing): expectation and a credible rationale are active ingredients of therapy, not noise — the legitimate, non-derogatory sense of "placebo." This entry's "believe in the method you pick" guidance grounds there.
• Mental (mental_mindfulness_meditation): mindfulness/acceptance practices are the technique core of ACT and third-wave CBT; the convergence-of-outcomes picture includes these modalities.
• Mental (autonomic_nervous_system_balance): exposure's extinction-learning mechanism interacts with autonomic regulation — the fear response that exposure retrains is an autonomic one, and downregulation skills are a common adjunct.
• Mental (depression_lifestyle_interventions, chronic_stress_management): therapy is one lever among several; lifestyle interventions and stress-load reduction have their own substantial effects and are often appropriate first or concurrent moves rather than alternatives.
• Mental (psilocybin_assisted_therapy_clinical): an instructive boundary case — psychedelic-assisted therapy is explicitly a drug-plus-psychotherapy package where the psychological support (set, setting, integration) and the pharmacology are entangled, sharpening the common-vs-specific question rather than resolving it.

What would change our mind

Falsifiability: explicit upgrade/downgrade criteria from source

**We would upgrade the causal status of a common factor (toward Tier 2) if:**
• Experimental studies that manipulate the alliance (not just measure it) showed a downstream outcome effect, with reverse causation and third variables excluded — i.e. it met Cuijpers' five mechanism criteria (temporality, dose-response, third-variable exclusion, experimental support, theory).

**We would upgrade the specific-technique position (toward "methods meaningfully differ in general") if:
• High-quality, allegiance-controlled, adequately powered head-to-head trials showed reliable, replicated method-superiority for common conditions beyond the fear-disorder/exposure case — with component network meta-analyses isolating which ingredients carry it.

We would downgrade / revise toward stronger common-factors equivalence if:
• The remaining specific-ingredient cases (including exposure) failed to replicate in pre-registered dismantling trials, or were shown to be explained by expectancy/credibility differences rather than the technique itself.

What would NOT change our mind:**
• A single allegiance-laden trial declaring one brand superior, or a popular-press claim that "the relationship is everything." Neither addresses the actual evidentiary gaps.

Industry bias note

Structural incentives the evidence base may reflect

Bias risk here is HIGH and bidirectional, which is unusual and worth naming explicitly.
• The branded-superiority incentive. Trademarked modalities, manualised-treatment licensing, certification/training pipelines, and therapy-app companies all have revenue tied to the belief that their specific technique outperforms generic care. Researcher allegiance (Munder et al. ~r≈.26, stronger in weaker trials) is the measurable signature of this incentive contaminating the outcome literature. Treat any "Method X beats the rest" claim — especially from the method's own developers or a company selling it — as allegiance-suspect until shown otherwise by independent, allegiance-controlled trials.
• The "it's all common factors" incentive. Conversely, the equivalence story is commercially convenient for low-cost, lay-delivered, or insurer-driven models, and for platforms (including, in principle, a coaching platform like Realised) that benefit from framing structured technical skill as optional. Realised flags its own potential bias here: a recovery-and-relationship framing is congenial to us, so we hold ourselves to the evidence that exposure-class technique is not optional for fear disorders.
• The underfunding signature on the right answer. The most useful, least monetisable finding — fit and engagement matter more than brand; pick a credible method and start — has no product attached, which is precisely why it is under-marketed relative to its evidential strength. Unbranded truths don't buy ad space.

Therapy Mechanisms