Farb et al. (2011) — Mood-linked mPFC responses predict depressive relapse
The wiki’s first first-hand Farb empirical paper (the 2015 anchor is a theory paper), and the one that supplies the clinical evidence underneath the dual-mode framework the rest of the wiki’s interoception theory generalizes: an elaborative/ruminative mode of reacting to emotional challenge that predicts relapse, and a sensory mode that protects against it. Ingested from Week 11 The Absence of Emotion — and the folder title is the finding: the patients had no depression and felt sadness normally; what forecast their relapse was invisible in their feeling and visible only in how their cortex processed it.
Design
Sixteen fully remitted, recurrent-unipolar-depressed patients (≥3 prior episodes, all medicated, no psychotherapy) and sixteen healthy controls watched sad and neutral film clips during fMRI (film-based mood provocation; sad = “The Champ”, “Terms of Endearment”; neutral = gardening/woodworking programmes). Emotional reactivity was defined per voxel as BOLD(sad) − BOLD(neutral). Remitted patients were then followed for 18 months; relapse status was regressed onto reactivity to find the neural predictors. Ten of sixteen relapsed.
Two trait measures were taken at scan: rumination (Response Style Questionnaire) and acceptance (Acceptance and Action Questionnaire–Revised) — the maladaptive and adaptive cognitive modes in the face of dysphoria.
The core result: two modes, opposite prognoses
| reactivity to sad films | region | predicts | tracks trait |
|---|---|---|---|
| elaborative / self-referential | mPFC (BA 24/32, ventral; also dorsal BA 9/32) | relapse (r = .68, p < .005) | rumination (r = .56) |
| sensory | left calcarine / visual cortex (BA 17) | sustained remission (r = −.73, p < .001) | acceptance (r = .55) |
The two were inversely correlated within patients (r = −.55): a patient high on one was low on the other. Farb reads this as a trade-off — attention to an emotionally challenging stimulus is allocated either to elaborative (prefrontal, self-evaluative) or to sensory (visual, perceptual) cortical systems. This is the neural form of the sensory-vs-narrative distinction Farb had drawn behaviorally in 2007 (narrative vs. experiential self-reference) and would carry into interoception in 2013 and 2015.
Three things that make the result more than a correlation
1. It is not residual depression. Relapsers and sustained remitters did not differ in scan-time symptoms, prior-episode count, or medication dose. Entering all three as covariates increased the mPFC and visual prediction (partial r(11) = .79 and −.79). So the mPFC signal is not a scar of lingering illness — it is a latent reactivity revealed only by challenge.
2. It is not the trait questionnaires in disguise — the causal arrow runs brain→behavior. Rumination and acceptance only marginally predicted relapse (r = .46, p = .07; r = −.45, p = .08) and, crucially, did not mediate the brain→relapse link, whereas brain activity did mediate the behavior→relapse link. Brain reactivity out-predicted the questionnaires (ROC AUC .90 vs. .75 for mPFC/rumination; .93 vs. .75 for calcarine/acceptance). The neural mode is the more proximal cause; the trait report is a partial readout of it.
3. Remission has a signature, and it is two-part. In a control/remitted/relapse ANOVA: the relapse group’s mPFC reactivity differed from controls (p = .007) but the sustained-remission group’s did not (p = .164) — remission is normalization of mPFC toward healthy levels. Meanwhile the sustained-remission group’s visual reactivity exceeded controls (p < .001) — a compensatory sensory response. Enduring recovery is not just the quieting of the ruminative mode but the amplification of the sensory one.
Why “marker of risk, not successful regulation” matters
The intuitive reading of frontal engagement during a sad film is effortful emotion regulation — the patient working to down-regulate. Farb’s data invert this: mPFC reactivity does not protect, it forecasts collapse. The dorsal (task-oriented) and ventral (self-evaluative) mPFC, normally anticorrelated in healthy people, were positively coupled in relapsers (r = .49) and anticorrelated in controls (r = −.53). The signature of relapse risk is the failure to disengage self-referential processing from an external stimulus — an inability to just watch the film without turning it into a referendum on the self. This is the DMN/self-reference story told as a clinical predictor: the mPFC here is doing reactivated dysfunctional self-evaluation, not regulation.
The honest limit for an interoception wiki: the sensory pole is exteroceptive
The protective “sensory” reactivity in this paper is in visual cortex — it is exteroceptive perception of a film, not interoceptive perception of the body. So the paper does not directly show that interoceptive sensory processing is prophylactic. Its value to this wiki is structural and by analogy, in three registers, each flagged rather than smoothed:
- It is the clinical proof-of-concept of the dual-mode framework. perceptual-inference (change your relationship to sensation, stay perceptual) vs. active-inference (elaborate, regulate, problem-solve) is exactly the elaborative/sensory trade-off measured here — with a hard outcome (relapse) attached. The interoceptive generalization is Farb’s own subsequent move, not this paper’s claim.
- It grounds decentering neurally. Acceptance (AAQ-R) tracks the protective sensory pole; rumination tracks the harmful elaborative pole. Decentering — viewing a thought as a transient mental event rather than a cue to elaborate — is the trainable version of shifting toward the sensory pole, and this is the paper that gives it a prognostic brain correlate.
- It complicates “more sensory is better.” Read carelessly, this paper says more sensory reactivity = good — a clean vote for the sensory side in is-more-interoceptive-awareness-better. But the sensory it measures is visual, and the harmful pole is self-referential elaboration, not felt-body intensity. What generalizes is not “feel your body more” but “elaborate less” — which is the regulation construct (relationship to signal), not the intensity construct. See the debate page.
Replicated and extended: Farb et al. (2022)
The sequel is in raw/ and now on the wiki: Farb, Desormeau, Anderson & Segal (2022) ran the same sad-film mood induction on 85 remitted patients, scanned before and after eight weeks of prophylactic therapy (MBCT or Well-Being CT), with a 2-year follow-up. It replicates the core finding — greater sadness-evoked sensory deactivation predicts relapse — and extends it into a formal two-factor model by separating a static (treatment-invariant) somatosensory-deactivation biomarker from a dynamic (treatment-responsive) left-DLPFC-reactivity biomarker; combined they reach C = 0.86. Two upgrades to the 2011 story worth flagging here: (1) the protective/harmful sensory pole moves from visual cortex toward somatosensory + insular cortex — a step closer to interoception, though still not landing on the insula as the top predictor; and (2) neither MBCT nor WB-CT moved the sensory factor, and both moved the prefrontal factor equally, so the prefrontal mechanism is trans-therapeutic, not mindfulness-specific.
Provenance
Norman Farb is this wiki’s author (see norman-farb); this paper predates and underwrites his contemplative-interoception programme. Recorded as provenance, not weighted as authority — the same convention used for payne-2015-somatic-experiencing. Senior author Zindel Segal is the co-developer of MBCT and the source (via Segal et al. 2010) of the patient sample; co-author Adam K. Anderson recurs across Farb’s imaging work but is not given a page here.