The neuropsychological impact of insular cortex lesions (Jones, Ward & Critchley 2010)
The wiki’s evidence about the insula — the structure nearly every page here runs through — is overwhelmingly correlational: activation, connectivity, grey-matter volume, cytoarchitecture. Lesion evidence has arrived twice, both times as a paragraph inside a much larger review (Bonaz et al. 2021; Quadt et al. 2018). This is the dedicated treatment, eleven years earlier, and it is explicit about its purpose: to appraise models of insula function built from neuroimaging against what happens when the structure is damaged.
Its answer is that the models do not survive the comparison intact.
A note on the year. The file is Jones 2009.pdf; the paper was received 8 September 2009 and accepted 27 November 2009, and published in the June 2010 volume. Filed as 2010 by journal volume — the same convention used for wiens-2000-heartbeat-detection-emotion, nummenmaa-2014-bodily-maps and van-der-does-2000-heartbeat-perception-reanalysis.
Why lesion evidence is different, and why it is harder than it looks
See lesion-symptom-mapping, created with this ingest, for the method. The short version matters here because it shapes every result below.
Activation tells you a region participates. A lesion tells you whether the function survives without it — which is the only evidence in this wiki that speaks to necessity. But the insula is unusually badly served by the design. Isolated insular infarcts are uncommon. The insula’s blood supply is the middle cerebral artery, and MCA occlusion produces diffuse damage across internal capsule and basal ganglia as well. So “insula lesion” in this literature almost always means “large stroke that happened to include the insula.”
The review contains the cleanest available demonstration of what that costs. Apraxia of speech localized to left anterior insula (Dronkers 1996, in Nature) is one of the most cited lesion findings about this structure. Hillis et al. (2004), using diffusion- and perfusion-weighted imaging to separate infarcted from merely hypoperfused tissue, found apraxia of speech correlated with damage to Broca’s area and not the insula — and concluded the insula “is most commonly damaged because it is most vulnerable to disruption of the middle cerebral artery,” so the symptom and the insular damage “may be independent manifestations of large stroke.”
A canonical localization, standing for eight years, produced by the vascular anatomy of the sampling rather than by the function. Hold that alongside every other row in the review’s Table 1.
The anterior/posterior split, and where it contradicts the wiki
The review’s headline structural claim maps onto this wiki’s gradient — and then breaks from it at one point that matters.
What matches. Anterior-insula lesions disrupt affective and language functions: amusia (loss of the emotional content of music), aphasia, apraxia of speech, especially after left-hemisphere damage. Posterior-insula lesions disrupt the representation of interoceptive information: pain, temperature, tactile perception. That is roughly the posterior-primary / anterior-integrative gradient the insular-cortex page is built on, arriving from a causal design.
What does not. The wiki, following Craig, treats awareness as the anterior insula’s business. Disorders of bodily unawareness — anosognosia for hemiplegia, anosognosia for hemianaesthesia, somatoparaphrenia — would on that reading be anterior deficits, and the authors note that “some previous reviews imply” exactly this. The lesion evidence says otherwise: Karnath et al.’s (2005) lesion-overlay analysis found the posterior insula the structure most commonly damaged in stroke patients with anosognosia for hemiplegia, against a matched control group. Cereda et al. (2002) report a focal right posterior insular lesion sufficient to produce somatoparaphrenia.
So the region that maps the body is also the region whose loss makes a patient unable to register that the body has stopped working. See anosognosia, where this is developed, and note that it does not sit comfortably with an architecture in which awareness is a re-representation happening further forward.
Disgust: the one emotion-specific lesion deficit, and it is left
“In terms of specific emotional deficits following insula damage, neuropsychological evidence is sparse, with the exception of disgust recognition.”
Three convergent sources: patient NK, whose focal left insula/basal ganglia lesion produced highly selective impairment in recognising disgust from faces and voices and a diminished subjective experience of it (Calder et al. 2000); patient B, with impaired recognition of dynamic disgust expressions after bilateral insular lesions (Adolphs et al. 2005); and Kipps et al.’s (2007) voxel-based morphometry in preclinical Huntington’s disease, where left anteroventral insular grey-matter volume correlated with disgust recognition and with no other emotion, in mutation carriers but not controls.
Two things this does to the wiki, developed on disgust:
- It is the best lesion case for functional selectivity in the Lindquist sense — and, as the insular-cortex page already notes, selectivity-without-specificity predicts exactly this pattern (a disgust deficit among several others), so it is not the locationist result it can look like.
- It cuts against Craig’s forebrain emotional asymmetry. The authors say so directly: “The lesion data discussed thus far do not fully support this view, in that impairment in disgust processing results particularly from left insula damage.” Disgust is a withdrawal-type negative emotion; the asymmetry hypothesis puts those on the right.
Pain: the insula is not required for it
The most consequential single result for Craig’s reclassification of pain as an interoceptive modality represented in the insula.
The lesion record is already inconsistent. Posterior-insula lesions (but not anterior) have been reported to raise pain thresholds (Greenspan & Winfield 1992), or to produce pain asymbolia — pain is perceived, but its aversive character is abolished (Greenspan et al. 1999; Berthier et al. 1988). Both are losses.
Starr et al. (2009) found the opposite. Two patients with large left MCA ischaemic strokes involving insular cortex rated noxious stimuli to their affected (right) side as significantly more intense than healthy controls did, with no insula activation in either hemisphere during those stimuli — while stimulation of the unaffected side did activate right insula. And their unpleasantness ratings did not differ from controls’, dissociating the two dimensions of pain from each other. The authors’ conclusion is the sentence to keep:
subjective awareness of noxious stimuli involves multiple, distinct patterns of brain activity where insular cortex is not a prerequisite.
Not a refutation of the anatomy — the ascending lamina I pathway is unaffected by any of this — but a hard constraint on the inference from the insula represents pain to the insula is where pain is felt. Two patients. See chronic-pain, where-are-feelings-constituted.
Decision-making: three lesion studies, two directions, one common factor
The review’s interoception section is largely a summary of somatic marker theory with the insula named as the substrate for the represented bodily state. The evidence it then assembles is more interesting than the framing.
| study | patients | task | result |
|---|---|---|---|
| Bar-On et al. (2003) | non-focal vmPFC / amygdala / insular-somatosensory lesions vs. lesions outside limbic and viscerosensory regions | IGT | experimental group made significantly more disadvantageous choices, net loss; also impaired social functioning and emotional intelligence; no difference in general cognitive performance |
| Clark et al. (2008) | focal insula lesions (greatest overlap anterior, n=13) vs. focal vmPFC vs. dorsolateral-PFC control vs. healthy | Cambridge Gambling Task | insula patients bet at a high level even as odds became unfavourable — no risk adjustment |
| Weller et al. (2009) | insula lesions, n=10 | risky gains and losses | insula patients were more cautious, taking significantly less risk than healthy controls even where risk was advantageous |
Clark and Weller point in opposite behavioural directions. The authors’ reconciliation is that both show insensitivity to the expected value of the outcome — Weller’s “emotional bluntness towards risk” — with the direction of the resulting bias underdetermined by the deficit.
That reading is available and it is also unfalsifiable as stated: a deficit that predicts both more and less risk-taking predicts nothing about risk-taking. The wiki should record the datum (insula lesions degrade the calibration of risky choice) and hold the interpretation loosely. Bar-On’s group had non-focal lesions spanning vmPFC, amygdala and insula, and the authors say plainly that the small groups made site-by-site comparison impossible — so it attributes nothing to the insula specifically. See does-somatic-feedback-guide-decisions, where this is added as a field position, and iowa-gambling-task.
Craving, and the wiki’s provenance improves
The craving page records “loss of drug-craving after insular damage” as the sharpest causal evidence in this wiki that a felt bodily state constitutes a drive rather than accompanying it — held at one remove, via a clause in Bonaz et al.
The provenance is now better, though still not first-hand. Naqvi et al. (2007) is here with its design: 19 patients with strokes involving the insula (right n=6, left n=13), who “reported greater ease in giving up smoking, without relapse or persistent urge to smoke, in contrast to smokers with strokes affecting other areas.” Jones et al. offer two candidate mechanisms and decline between them: a reduced ability to detect interoceptive states linked to craving, or a reduction in the hedonic feelings induced by smoking. Those are different theories — one about perceiving the drive, one about the reward itself — and nothing here separates them.
The review also flags the obvious next question, still open sixteen years later: how general the effect is across other addictions.
What the review says models cannot do
The summary is unusually direct for a paper co-authored by one of the field’s principal modellers, and it is the reason to keep this page.
At present, models of insula function fall short of adequately explaining in detail the range of neuropsychological effects reviewed above. From lesion and neuroimaging data, we can conclude broadly that the insula supports an integration of body and mind, yet the mechanisms through which these interoceptive signals are perceived, how they are translated into perceptual sensation (‘feeling’), and how disparity between predicted and actual feeling states is detected, is less obvious in the presence of marked insula damage.
Two specific difficulties the review leaves standing:
- The unitary-account problem. The insula is active in language, audition, risk, sensorimotor processing, pain, taste and flavour, and autonomic control. The authors set out the options — one integrative function, or a core visceral function from which specialised functions evolved (taste, disgust, limb awareness, speech articulation) — and reach no verdict. Craig’s global emotional moment is presented as the leading unitary attempt and is not endorsed.
- The recovery problem. Duffau et al.’s (2006) 42 insular glioma resections produced substantial immediate deficits — 21 transient hemiparesis, 10 articulatory disorders, 7 with abulia (inertia with loss of interest and affect) — and at three months, no visceral, cardiovascular, sensorimotor, gustatory, auditory-vestibular or language disorders in any patient. Whatever the insula does, other tissue can largely do it within a quarter. This is the review’s own best evidence against reading any of its localizations strongly.
The comparator, in 2010
Worth recording for feedforward-vs-predictive-interoception because of its date. Three years before Seth (2013) and seven before Barrett (2017), Critchley is already describing the insula as a comparator detecting incongruence between predicted and actual emotional states, with the anterior insula storing “an error representation when the states are mismatched” usable in reinforcement learning to adjust future behaviour (Preuschoff et al. 2008; Pessiglione et al. 2006), and citing the Singer, Critchley & Preuschoff (2009) integrative model in which concurrent and prediction signals are integrated into a dominant motivational feeling state. The false-feedback result is offered as the demonstration: induce interoceptive mismatch experimentally and anterior insula activity tracks the increased emotional salience attributed to previously neutral faces (Gray et al. 2007).
This is prediction-error language without the generative-model apparatus — closer to Craig’s ascending-comparator architecture than to Seth’s descending one, and a useful datum on how the Sussex group’s position developed. See hugo-critchley.
And note the tension the same paragraph creates with everything above it: the mechanism proposed is a mismatch detector, and the review’s own verdict is that this mechanism “is less obvious in the presence of marked insula damage.” The model is offered in the same section that reports the lesion data failing to show it.
Provenance and standing
Critchley is senior author; this is his fourth first-hand paper in this wiki and, at 2010, the earliest — it predates khalsa-2018-interoception-roadmap, quadt-2018-interoception-health-disease and bonaz-2021-diseases-of-interoception by eight to eleven years. First author Catherine Jones was an MRC-funded doctoral researcher at Brighton and Sussex Medical School; Jamie Ward is a Sussex cognitive neuroscientist (synaesthesia, body representation). The contributor note is unusual — “JW and HDC contributed equally.” Funding: MRC and BSMS (CLJ), a Wellcome Trust programme grant (HDC). No declared competing interests. Externally peer reviewed, not commissioned.
No researcher pages created for Jones or Ward: single appearance in raw/, per the co-author convention used for Jia, Wiens, Amedi, Koch, Lane, Kanoski, Grill, Feldman and Napadow.