Farb, Segal & Anderson (2013) — Attentional modulation of primary interoceptive and exteroceptive cortices
The paper this wiki had been carrying secondhand and thought it could not reach. Its two headline findings — that the posterior insula is a respiratory-rate sensor “putatively considered primary interoceptive cortex,” and that interoception enhances insula connectivity with the posterior ventromedial thalamus — sat on insular-cortex and respiratory-interoception as claims sourced through Weng et al. (2021), flagged there as belonging to a “Farb et al. (2013)” paper that “sits in the excluded raw/papers/Farb/ archive.” The exclusion was lifted on 2026-07-19, and this is that paper, read first-hand at last.
One disambiguation up front, because the queue contains two Farb 2013 papers and the secondhand note conflated them. This is the Cerebral Cortex paper, whose subject is attention (interoceptive vs. exteroceptive), in a mixed sample where MBSR completion is only a covariate. The companion — [[farb-2013-training-interoceptive-cortex|Farb, Segal & Anderson (2013), SCAN]], “Mindfulness meditation training alters cortical representations of interoceptive attention” — is the training paper, now ingested. So the “mindfulness training increased posterior insula activation” phrasing the wiki inherited is the SCAN paper’s claim; everything below is an attention contrast (IA vs. EA), not a trained-vs-untrained one. (Read first-hand, the SCAN paper’s group-level effect is in anterior insula, its posterior-insula effect is a practice-dose correlation, and its VMpo-thalamus claim is actually this paper’s — see the correction filed on insular-cortex and respiratory-interoception.)
The design
Twenty-seven adults (20 women, mean age 45 ± 12.63) performed a block-design fMRI experiment alternating interoceptive attention (“Breathe”: attend to the sensory aspects of the breath without altering it) with exteroceptive attention on visual personality-trait words in two forms — Maintain (a 1-back working-memory task) and Suppress (read the word but “keep your mind blank”). Blocks were 36 s, cued for 10 s. Two things make the design unusually clean for this literature:
- Respiration was recorded throughout (belt, 40 Hz), so the study can check that attended and unattended breathing were mechanically the same — and they were (rate, volume and RVT all n.s. between conditions). Most interoceptive attention studies omit physiological recording entirely and so cannot say what the body was doing while it was attended to.
- The interoceptive signal is objective and observable. Rather than asking whether the insula lights up during breath-focus, the paper uses the measured respiratory rate as a regressor and asks where in the brain it is tracked — a localizer anchored to a real viscerosomatic signal, not to an instruction.
The core result: attention selects the cortex, not the signal
The finding worth leading with is not that interoception activates the insula — it is that the same respiratory signal is represented in different cortex depending on where attention is pointed. When participants attended to the breath, respiratory rate correlated with the right posterior insula (dorsal posterior long gyrus, the granular primary interoceptive cortex; IA r=0.63). When they attended to the visual words instead, the very same respiratory rate correlated with right somatosensory cortex (EA r=0.59) and the posterior-insula coupling fell to non-significance (r=0.23). Breathing did not change; its cortical destination did.
This is the wiki’s cleanest single demonstration that interoception is not a fixed feedforward readout of a channel but a representation that attention gates — the precision-weighting reading of attention made visible, though the paper frames it in the vocabulary of “gain” rather than of predictive coding. IA increased the gain of viscerosomatic responses in posterior granular and middle dysgranular insula; the study reads this as attention amplifying the response in the attended interoceptive receptive field.
The insula as a graded posterior→anterior tuning field
The anatomical ROI analysis is where this paper earns its place on insular-cortex. Drawing eight gray-matter ROIs along the insular gyri and scoring each for attentional tuning (IA − EA), the study found a graded shift: interoceptive responses dominate the posterior gyri and give way to exteroceptive responses in the anterior gyri (Fig. 5; dorsal main effect of seed location F(4,104)=25.36, p<1×10⁻⁶). The posterior→mid→anterior gradient this wiki builds on is usually stated as primary map → re-representation; here it acquires an orthogonal reading — posterior = interoceptively tuned, anterior = exteroceptively tuned — from a within-subject attention manipulation.
The sharp end of that gradient: the anterior insula was better predicted by exteroceptive than interoceptive attention, and the authors state the consequence bluntly — the anterior insula is “not an area of pure body awareness.” They propose instead that it binds outside-world representations to internal bodily state, which is why they reach for it as a substrate of emotional experience (feeling about something) rather than of bare interoception.
This lands directly on Haruki & Ogawa (2023), which a decade later found the right dorsal anterior insula preferring a demanding visual task to interoception altogether, and preferring the heart within interoception — the same deflation of the anterior insula’s status as the seat of body-awareness-as-such, reached by a different route. The two studies should be read together: Farb et al. show the anterior insula tuning toward exteroception under attention; Haruki & Ogawa show it tuning toward arousal/cardiac and toward the visual task. See interoception-exteroception-boundary, where the anterior insula’s exteroceptive lean is now a first-hand data point rather than an inference.
Two dissociable attention networks
Beyond the insula, the whole-brain and PPI analyses give interoceptive and exteroceptive attention distinct large-scale networks:
| interoceptive attention (IA) | exteroceptive attention (EA) | |
|---|---|---|
| primary cortex | posterior/mid granular insula | primary/secondary visual cortex |
| network | posterior limbic + medial parietal: posterior cingulate, precuneus, hippocampus/parahippocampus, midcingulate | lateral frontoparietal “executive”: DLPFC, dmPFC, inferior/superior parietal, caudate, cerebellum |
| enhanced connectivity (PPI) | insula ↔ posterior ventromedial thalamus (VMpo), insula ↔ contralateral posterior insula | visual cortex ↔ inferior parietal lobule, visual cortex ↔ visual (pulvinar) thalamus |
Two points matter for the wiki. First, the EA network is the frontoparietal “executive” attention network (Corbetta & Shulman; Seeley et al.), and the paper’s proposal is that this well-characterized network is specialized for exteroceptive monitoring — IA recruits a different, more limbic/medial system. Second, the IA-enhanced insula–VMpo thalamic connectivity is a claim about the afferent pipeline itself: VMpo is the lamina I spinothalamocortical relay, low in the interoceptive hierarchy, so attention is modulating transmission at the thalamic gate, not only at cortex. Whether that is best read as improved signal transmission or as top-down precision-weighting of a low-level channel is exactly the feedforward-vs-predictive-interoception question, and this single connectivity result does not settle it — but it is the wiki’s first first-hand instance of an attentional effect on that relay.
What the wiki gains by holding it first-hand
The insular-cortex Weng-sourced section and the respiratory-interoception “primary interoceptive cortex” line were accurate in substance but mis-framed the effect as a training increase. Read first-hand, the effect is an attention contrast, in a sample where training is a covariate. The correction is filed on both pages. Nothing is lost by the reframing — if anything the attention reading is stronger, because it shows the posterior-insula respiratory representation is gated by the direction of attention within subjects, which a between-groups training comparison could never demonstrate.
Provenance
Norman Farb is this wiki’s author (see norman-farb); recorded as provenance, not weighted as authority, per the convention used across the Farb papers (farb-2010-neural-expression-sadness, farb-2011-relapse-prediction). Senior collaborator Zindel Segal and co-author Adam K. Anderson recur across this group’s output. The paper declares no conflicts of interest; it was funded by CIHR, NIMH, the Ontario Mental Health Foundation and a Mind and Life Institute Varela grant.