Norman A. S. Farb

Cognitive neuroscientist at the University of Toronto Mississauga, working at the intersection of mindfulness/contemplative practice and predictive-coding models of interoception. The 2015 paper anchoring this wiki’s second ingest (see farb-2015-interoception-contemplative-health) grew out of an April 2013 Mind and Life Institute meeting Farb organized with co-authors spanning neuroscience (Paulus), clinical/complementary medicine (Daubenmier, Price, Mehling, Kerr), and Buddhist/contemplative studies (Klein).

Central theoretical contribution

The simulation-map model: extending Seth’s interoceptive-inference into a dual-route regulatory framework — active-inference (act to change sensation, mapped onto suppression/distraction/reappraisal) vs. perceptual-inference (update priors to match sensation, mapped onto contemplative acceptance/equanimity). Also extends Seth’s presence-and-agency constructs, arguing presence is reachable via perceptual inference and not only active inference as originally proposed.

Clinical/applied focus

Distinct from the more purely theoretical Seth (2013), Farb’s work consistently pushes the predictive-coding framework toward testable clinical and wellbeing applications — see mindfulness-interoceptive-training and interoceptive-training-clinical — while being notably cautious about overgeneralizing benefits (e.g., the explicit warning that interoceptive-awareness training can worsen symptoms in anorexia nervosa absent skillful guidance).

The interoceptive anchor: Farb et al. (2010)

Chronologically the earliest Farb data paper on the wiki (met third, after the 2011 and 2022 relapse papers). Farb, Anderson, Mayberg, Bean, McKeon & Segal (2010) compared MBSR completers with waitlisted controls during a sad-film provocation and found that, at equivalent felt sadness, trained participants recruit a right-lateralized viscerosomatic network — the right insula and subgenual ACC — where controls deactivate it, while both groups’ midline self-referential and left-language elaboration is reduced by training. The load-bearing correlation: right insula reactivity tracks lower depression (BDI r=-.465), and left language (Wernicke’s) reactivity tracks higher depression, with the two anticorrelated. This is the wiki’s only first-hand Farb evidence in which the protective sensory pole is genuinely interoceptive rather than exteroceptive — the exception to the caveat the 2011/2022 pages carry — bought at the price of a weaker, cross-sectional, between-groups design. See perceptual-inference, decentering.

The relapse anchor: Farb et al. (2011)

The wiki’s first-met Farb data paper (the 2015 anchor is a theory piece). Farb, Anderson, Bloch & Segal (2011) followed remitted depressed patients for 18 months and found that reactivity to sad films in elaborative/self-referential medial prefrontal cortex predicts relapse, while reactivity in sensory (visual) cortex protects against it — an inverse trade-off, with felt sadness itself predicting nothing. This is the clinical proof-of-concept of the sensory-vs-elaborative distinction Farb later carries into interoception: the perceptual-inference (sensory, accepting) vs. active-inference (elaborative, regulating) contrast, with a hard outcome attached. It also seats his collaboration with Segal and MBCT, and the cognitive-reactivity/decentering pair that runs through his contemplative work. The one caveat this wiki keeps visible: the protective sensory pole there is exteroceptive (vision), so the interoceptive reading is Farb’s generalization, not that paper’s finding.

The RCT sequel Farb et al. (2022) scans 85 patients before and after prophylactic therapy over a 2-year follow-up, and turns the 2011 correlation into the two-factor model: a static sensory-deactivation vulnerability that treatment does not move, and a dynamic prefrontal reactivity that it does. Two things there matter for the wiki’s reading of Farb: the sensory pole shifts from vision toward somatosensory/insular cortex (closer to interoception, still not landing on it), and the prefrontal prophylactic mechanism is trans-therapeutic — MBCT and Well-Being CT produced the same effect and the same null relapse difference — which tempers any claim that mindfulness/interoceptive training is the privileged route.

The fourth relapse paper, Wu et al. (2026) (Farb senior author), scans the same trial cohort during a self-referential encoding task rather than the sad film, and is the one that moves the two-factor model off its prefrontal centre. The static somatosensory-deactivation vulnerability replicates during self-processing; the treatment-responsive marker is now somatosensory (supramarginal activation rising in non-relapsers) rather than prefrontal; and in the combined survival model the somatosensory markers are the sole survivors while the subgenual prefrontal marker drops out. Its lesson for reading the corpus: the protective “sensory” pole Farb has tracked since 2007 is not incidental to the elaborative account but the part that carries the predictive weight — relapse vulnerability as “insufficient recruitment of somatosensory systems supporting embodied self-experience.” As with the whole arc, the sensory pole measured is somatosensory rather than interoceptive proper, so the interoceptive reading stays a generalization.

The one manipulation: Eusebio et al. (2022)

Every other Farb paper here is observational, quasi-experimental, or a comparison of two full therapies. Eusebio et al. (2022) is the only one that manipulates interoceptive emphasis itself, holding format, dose, instructor training and social contact constant across two yoga curricula. It is therefore the closest test in this wiki of Farb’s central applied claim.

The claim did not survive in the form it is usually stated. Interoceptive instruction did not raise self-reported interoception more than instruction that avoided mentioning the body, and it did not reduce symptoms more. What it did produce was behavioural transfer — better sustained attention on a task with no interoceptive content — moderated by interoceptive sensibility and by none of the six other traits measured.

Read alongside Farb et al. (2022), a pattern in this group’s two-armed trials becomes visible and is worth naming here rather than on either study page: when Farb’s interoceptive/mindfulness arm is compared against a serious active comparator rather than a waitlist, it does not win on the nominated outcome, while a mechanistic measure moves. In 2022 that was equal relapse rates with a shared prefrontal biomarker; in Eusebio et al. it is equal MAIA and symptom change with a differential attention transfer. This is not a criticism of the work — the trials are well designed and the nulls are reported plainly — but it is the wiki’s strongest reason to treat “interoceptive training works through interoception” as a hypothesis this group has repeatedly tested and not yet confirmed. See does-mindfulness-enhance-interoceptive-accuracy, yoga-as-interoceptive-training.

The third instance, and the one that measured the mechanism: Rai et al. (2026)

Rai et al. (2026) extends the pattern above from two trials to three, and does so on the mechanism point most directly. A smartphone mindfulness therapeutic (the research predecessor of Mobio Interactive’s AmDTx) was tested against an active cognitive-training control for high-intensity cycling. As in 2022 and Eusebio, the comparator is serious rather than a waitlist; as in both, self-reported interoception (MAIA) moved as advertised. The nominated outcome — endurance — moved only slightly (β=0.21, p=0.041), and affective wellbeing during exercise did not move at all.

What makes this the cleanest of the three is that it measured a mediator and reported its failure. No preregistered candidate mechanism (mindfulness, attentional focus, mental toughness, mental/physical health) showed a group×time effect, so mediation could not run; and the raised interoceptive awareness was uncorrelated with the raised performance across individuals (MAIA↔TVE r=−0.02, ns). The factor that did track performance was effort and distress tolerance. So the reading Eusebio and 2022 support by inference — the benefit is not visibly the interoceptive component — Rai et al. support by direct test: interoceptive awareness rose, and it was inert with respect to the outcome. Note the standing conflict-of-interest here (a coauthor co-founds and holds ~21% of the app’s maker; Farb is an advisor with <1% stock), which the paper declares.

The clinical natural-lesion paper: Farb et al. (2012)

The wiki’s outlier among the Farb corpus — no mindfulness, no sad film, no breath. Farb et al. (2012) comes from his postdoctoral period at Baycrest’s Rotman Research Institute (senior collaborators Cheryl Grady, Tiffany Chow, Morris Freedman), and applies the same frontolimbic-integration framework to frontotemporal dementia via resting-state fMRI. Its value here is as the clinical mirror of the contemplative work: where the meditation papers show the anterior insula being recruited under emotional challenge, this shows disease disconnecting it, and names the behavioural cost — apathy tracks prefrontal hyperconnectivity, disinhibition needs the (now-severed) limbic channel. It is the wiki’s cleanest illustration that the anterior insula’s job of “integrating emotional and visceral information into representations of present-moment context that guide behaviour” (his own 2007 framing) is load-bearing: remove it and behaviour dysregulates in two dissociable directions. Read as a bridge between the interoception corpus and the DMN/salience network machinery, observed in its failure mode.

A provenance note

Farb was one of the three named reviewers of Payne, Levine & Crane-Godreau (2015), the wiki’s Somatic Experiencing paper — published the same year and in the same journal (Frontiers in Psychology) as his own contemplative-interoception framework, with which it shares a bottom-up, interoception-first orientation toward trauma and meditation. Recorded as a fact about the wiki’s provenance (its own author reviewed a source it holds), not as an endorsement of SE, which the wiki files as an evidence-free theory.