THEORY-058: Mental disorders are maintained by self-reinforcing causal loops rather than expressing a latent common cause, which is how they can outlast their triggers

Source

• Borsboom (2017), LIT-547, read in full in NOTE-435: the network theory’s five principles, of which the fifth is hysteresis.
• Salkovskis, Sighvatsson and Sigurdsson (2024), LIT-538, read in full in NOTE-428: maintenance by a threat meaning (“Theory A”) and the behaviour that protects it.
• Kendler (2008), LIT-539, read in full in NOTE-432: non-aggregative, loop-ridden causation across levels.

What was actually shown

Borsboom (LIT-547) states the account as theory. In general medicine a tumour is a common cause: remove it and the symptoms go together. For most mental disorders no such cause has been found (NOTE-435, C1, cited to Kendler’s later papers). Borsboom’s answer is that the symptoms cause one another, often prosaically (“If you do not sleep, you get tired”), and that a strongly connected network has two stable states. Principle 5: “symptoms continue to activate each other, even after the triggering cause of the disorder has disappeared”. Health is the stable state of a weakly connected network, and disorder the alternative stable state of a strongly connected one. The evidence for direct symptom–symptom causation is indirect by the paper’s own account: conditional associations, and patients’ and clinicians’ reports (C2). Hysteresis rests on cited simulations of “realistically parameterized” networks (C5), which the record has not read. Experience-sampling work on critical slowing down is cited and not read.

Salkovskis et al. (LIT-538) reach the loop from Beckian CBT. Problems are sustained “not by their original causes but by maintaining factors”, because 50 years have produced no real advance on causes (NOTE-428, C3). The central maintaining factor is a meaning. The person who reads palpitations as a heart attack avoids, neutralises or seeks safety, and that behaviour keeps the meaning from being disconfirmed. They call these “vicious circles which trap them”. Their evidence that cutting the link helps is that exposure framed as a test of the threat belief beat exposure alone in three cited trials (C5). All three are co-authored by Salkovskis.

Kendler (LIT-539) finds loops across levels. In his alcohol-dependence sketch, tolerance raises consumption, temperament selects risky peers, and offspring of heavy drinkers abstain more often. He also cites a fearful child’s avoidance preventing habituation (NOTE-432, C3). Loops are one of his three reasons that psychiatric causation cannot be decomposed and simply added up.

What the three share, and where they differ. Each puts a self-reinforcing loop where a single latent cause would be, and each says the loop, not the start, is what keeps the person ill. They differ on what the loop connects. Borsboom’s nodes must be the manuals’ symptoms (Principle 2), so appraisals and safety behaviours have to enter as constituents of a symptom or of a connection. Salkovskis’s loop runs through exactly those. Kendler’s runs across genes, persons and cultures. That the three converge without citing one another for it is the account’s strength. That none of them measures a loop in an individual patient is why it is Proposed.

How it bears on causation, composition and emergence

• Why “the disorder” is not a cause of its symptoms on this account. On Borsboom’s definitions a disorder is a state of the symptom network, so it is composed of the symptoms, not something behind them. Hazelwood (LIT-156, NOTE-137) argues, from Woodward’s independent-fixability condition, that a whole cannot stand in a causal relation to its own parts: an intervention on one is already an intervention on the other. Applied here, which is the record’s application and not either author’s, the network theory’s refusal to let the disorder cause its symptoms is not just an empirical bet. It follows from treating the disorder as composed of them. The empirical bet is the other half: that the symptoms, which are independently fixable, cause one another.
• The causation is interventionist. Borsboom defines a connection by what an intervention on one symptom would do to the probability of the other (Woodward, Pearl). Symptom-to-symptom links are within one level, so the exclusion problem Pernu (LIT-147) analyses for mental causes does not arise for them. It does arise for Kendler’s cross-level loops; the companion account THEORY-048 says why his examples survive it.
• Hysteresis is borrowed from the physics of phase transitions without an order parameter or a measured transition (NOTE-435, on Rizi, LIT-150). That is the gap promote_when asks to close.

What this does not say

• Nothing about onset. Loops explain persistence after a trigger. Borsboom himself scopes the theory to episodic disorders with a clear onset and sets aside autism, personality disorders and parts of schizophrenia (NOTE-435, C9).
• Not that there are no common causes anywhere. A disorder with a found common cause would fall outside the account, not refute it elsewhere.
• Not which loop matters. Whether the important links are among symptoms, through meanings, or across levels is open. So is whether CBT works by weakening a link (Borsboom’s network intervention) or by moving a symptom.
• Not that the loops are psychological rather than biological. Borsboom is agnostic about how a connection is realised. Kendler’s include pharmacological tolerance.

Connections

• Psychological inflexibility (THEORY-050) is compatible, not a rival. NOTE-431 notes that ACT puts one process behind many syndromes while Borsboom denies most disorders a common cause. But ACT’s process is itself a loop: avoidance makes the avoided event more salient and so narrows behaviour further (LIT-555). On Borsboom’s terms inflexibility would be a constituent of many connections, not a latent disease. The two conflict only if inflexibility is read as a cause behind the symptoms that sustains them whatever their connections, and neither source reads it that way.
• A measured transition elsewhere: HKB coordination (THEORY-069). It is the record’s measured case of the signatures this account borrows, in human behaviour and with an order parameter: relative phase between the hands switches from anti-phase to in-phase at a critical movement frequency, critical fluctuations and critical slowing are reported before the switch, and hysteresis follows from the model’s two stable states. It shows that the evidence promote_when asks for can be obtained from people. It is not evidence that disorders behave this way, and no relation is declared.
• Process-based therapy (LIT-544) treats complex networks as “an extension of functional analysis” and an alternative to the latent disease model, which is this account’s negative half.
• Engel’s biopsychosocial model (LIT-548) and Kendler 2005 (LIT-552) are the unread background both Borsboom and Kendler descend from.