Farb et al. (2012) — Abnormal network connectivity in frontotemporal dementia: evidence for prefrontal isolation
The wiki’s only clinical neurodegeneration paper, and its most direct “natural lesion” test of what the interoceptive/frontolimbic system does by watching what breaks when it fails. Chronologically the earliest of the wiki’s Farb data papers (2012, from his Rotman Research Institute / Baycrest postdoctoral period with Cheryl Grady and Tiffany Chow), and methodologically the odd one out: no mindfulness, no sad film, no breath — resting-state fMRI in frontotemporal dementia (FTD), read across ICA, fALFF, REHO, seed-based connectivity and PLS to make sure the story did not depend on one analysis pipeline.
The claim: FTD is prefrontal isolation, not just prefrontal loss
FTD is clinically defined by disinhibition, apathy and stereotypy (behavioural-variant bvFTD) and by semantic/language collapse (semantic dementia SD), with atrophy foci in ventral PFC, anterior temporal lobe and anterolimbic regions. The standard account is disconnection — the anterior insula and its frontolimbic partners degrade, and behaviour dysregulates because affective/visceral information stops reaching the PFC. Farb et al. reproduce that disconnection but add a second, less expected half: the isolated PFC does not go quiet — it becomes hyperconnected to itself. Both bvFTD and SD showed elevated local prefrontal connectivity (a medial-PFC cluster across the executive and salience templates, plus posterior-DMN elevation), on top of the reduced limbic connectivity. The paper’s thesis is that FTD is a disorder of frontolimbic disconnection leading to unconstrained prefrontal connectivity — “prefrontal isolation” — not simply of prefrontal damage.
Why this belongs on an interoception wiki
The load-bearing citation is the group’s own earlier work: the anterior insula “appears to integrate emotional and visceral information into representations of present moment context that guide socially appropriate behavior” — cited to Farb et al. (2007), Seeley et al. (2007) and Craig (2009). FTD is then the experiment nobody would run on purpose: degrade that integration and watch the behavioural cost. The right anterior insula emerges here as an early, convergent site of failure — reduced low-frequency power, reduced local coherence (REHO), and severed PFC connectivity — exactly the region the rest of the wiki treats as the apex of the salience hierarchy. So the paper supplies the clinical mirror image of the Farb contemplative corpus: where Farb (2010) shows training recruiting the insula under emotional challenge, this shows disease disconnecting it, and names the two symptoms that follow.
The two symptoms map onto the two halves of the signature
The most specific result is the double dissociation between the signature’s two parts and FTD’s two cardinal symptoms:
| network change | behavioural correlate | reading |
|---|---|---|
| PFC hyperconnectivity (medial/dorsal PFC, ↑local connectivity) | apathy (r=.68 with medial-PFC executive score) | prefrontal control without motivational input from limbic structures is insufficient to drive goal-directed action → withdrawal |
| frontolimbic disconnection (↓insula/thalamus/striatum connectivity) | lower disinhibition | disinhibition requires limbic communication; sever it and emotional impulses can no longer override prefrontal control |
Read the disinhibition row carefully, because it inverts the naïve expectation. One might assume that losing prefrontal inhibitory control causes disinhibition. The data say the opposite about the limbic side: reduced limbic connectivity predicts lower disinhibition — the positive, disinhibited symptoms need an intact affective channel to be produced at all. Apathy, by contrast, is what an isolated PFC produces on its own. Both associations held after controlling for atrophy, which is what lets the authors call them connectivity effects rather than lesion-location effects.
The bvFTD-specific default-mode finding
Unique to bvFTD was elevated default-mode connectivity in the right angular gyrus, tracking stereotypy (and apathy). The authors read this through the salience-network / DMN antagonism: as the salience network (anterior insula, ACC) dissolves, it stops regulating transitions in and out of the DMN, and DMN activity rises unchecked — “abnormally heightened DMN connectivity may in turn promote maladaptive behaviors driven by self-generated narratives and habit rather than what is emotionally salient and appropriate to environmental context.” Stereotypy — rigid, repetitive behaviour — is read as automaticity freed from motivational context: the DMN’s habit machinery running without the salience network’s veto. This is the same salience↔DMN switching architecture the wiki meets in Weng et al. (2021) and Barrett (2017), observed here in its failure mode.
Compensatory or maladaptive? The interpretive knot
The paper’s boldest move is to call prefrontal hyperconnectivity both compensatory in origin and maladaptive in effect: increased reliance on prefrontal processes “in isolation from subcortical structures appears to be maladaptive and may drive behavioral withdrawal.” Because PFC hyperconnectivity scaled with dementia severity (and with apathy/disinhibition), the authors argue it is unlikely to be benignly compensatory — it looks more like a marker of progression than a successful workaround. But this is an inference from cross-sectional correlations; whether the hyperconnectivity is the brain’s attempt to marshal executive control in the absence of affective feedback, or simply structurally-spared fibres firing without competition, cannot be settled here. The authors are explicit that a longitudinal model is the missing piece.
Method note: the local analyses caught what the network analysis missed
Worth recording for the wiki’s methods thread. The whole-brain group ICA did not flag the right anterior insula, but the fALFF and REHO local-signal analyses did — reduced low-frequency power and reduced regional homogeneity there in both groups, converging with a seed-based finding of lost insula–PFC connectivity. The authors present this as a case for local resting-state metrics as sensitive early markers of dementia pathophysiology that network-decomposition methods can miss. A useful caution against treating ICA network maps as the whole of what resting-state data contain.
Provenance
Norman Farb is this wiki’s author; recorded as provenance, not weighted as authority, per the convention across the Farb papers (farb-2010-neural-expression-sadness, farb-2011-relapse-prediction, farb-2007-modes-of-self-reference). This is his dementia/aging-lab work (senior collaborators Cheryl Grady, Tiffany Chow, Morris Freedman; Baycrest/Rotman), distinct from the Segal/Anderson contemplative line, and it shows the frontolimbic-integration framework being applied to a clinical population rather than a meditation sample. Funded by CIHR, NIA and the Women of Baycrest; the authors declare no role of funders in design or analysis.