Interoceptive attention task

Method at a glance

• Modality: multimodal (cardiac, gastric, respiratory, bladder), most often cardiac
• Measures: Which brain regions engage when a person deliberately directs attention to sensations from a named internal organ, relative to attention directed at an external stimulus. It measures interoceptive ATTENTION — the taxonomies’ attention-tendency construct — and not detection, accuracy, or any quantity with a ground truth.
• Procedure: Block-design fMRI. A cue word naming an organ (‘HEART’, ‘STOMACH’) is displayed for ~10 s while the participant attends to sensations from it; an exteroceptive control block (a visual target requiring sustained attention, e.g. a word whose colour fades imperceptibly) is matched for cue format, duration and attentional demand; rest blocks with fixation interleave. Conditions repeat over several runs. Compliance is checked by practice trials and by a post-block or post-run rating of how intensely the sensation was felt (often a 1-7 Likert). The contrast of interest is organ-attention minus visual-attention; where more than one organ is cued, the organ-vs-organ contrast is available too.
• Also called: interoceptive attention paradigm, visceral attention task, heartbeat attention task, body-focus fMRI task, attention-to-body paradigm

Strengths

• Non-invasive and repeatable. The alternatives for mapping visceral cortex — barostat distension, intravesical infusion, isoproterenol bolus — require instrumentation, carry discomfort, and cannot be combined in one session, so no single study could compare organs with them.
• Because it perturbs nothing, the activation it produces is not confounded by the discomfort, anxiety and arousal that a visceral challenge induces. This matters specifically for comparing organs: a distension protocol maps aversiveness as much as viscerotopy.
• Yields a robust and cross-culturally replicable map — insula, frontal and parietal operculum, middle cingulate, SMA — consistent across labs in Europe, the US and Japan (Caseras et al. 2013; Simmons et al. 2013; Tan et al. 2018; DeVille et al. 2020; Haruki & Ogawa 2021, 2023).
• Applies to organs with no perceptual task attached. There is no gastric or bladder analogue of the heartbeat-counting task, but there is a gastric attention block — which is currently the main way those channels enter human neuroimaging at all.
• Clinically portable: the same 10-second block works in anorexia (Kerr et al. 2016), depression (Wiebking et al. 2015; and the count-heartbeats-vs-count-tones variant of Wiebking et al. 2010) and suicide attempters (DeVille et al. 2020).

Limitations

• NO GROUND TRUTH. Nothing in the paradigm establishes that the participant attended to the organ named, or attended to anything bodily at all. Compliance rests on instruction, a practice trial, and a self-rating — the weakest verification of any measure in this wiki.
• It measures attention and is routinely reported as measuring awareness. The inferential step — that attending to an organ generates awareness of it — is assumed by the paradigm rather than tested within it.
• Task difficulty is rarely equated. Organs differ in how vivid they are: stomach sensations rate as less intense than heartbeat, which rates as less intense than a visual target. Difficulty differences can be covaried out post hoc, but they cannot be designed away.
• The exteroceptive control is doing a lot of work and is not standardized across labs. Everything reported is a difference from whatever that control happened to be.
• Visual imagery is an unexcluded alternative. Participants lie with eyes open, told to attend to an interior they may not feel; imagining the body region would produce activation that the paradigm cannot separate from interoceptive attention.
• Physiological recording is often omitted entirely, so studies frequently cannot say whether the attended signal was present, or whether attending changed it.

Connections

• Key studies: Haruki & Ogawa (2023), Farb et al. (2013), Farb et al. (2010), Weng et al. (2020), Weng et al. (2021), Quadt et al. (2018), Wiebking et al. (2010)
• Related methods: Heartbeat detection task, Electrogastrography (EGG), Respiratory tracking task with visuomotor perturbation (respyra), Multidimensional Assessment of Interoceptive Awareness (MAIA), body-focused contemplative training, Multidimensional Experience Sampling (MDES)

The dominant neuroimaging paradigm in interoception research, and — until this page — the wiki’s most cited unpaged method. It has been sitting behind claims on Farb et al. (2015), body-focused contemplative training, Interoceptive training in clinical practice, Helen Y. Weng and a dozen other pages, doing the work of “the study that found insula activation during interoception.”

Created with the Haruki & Ogawa (2023) ingest, which is the first study in raw/ to run the paradigm on two organs at once and therefore the first to expose what it can and cannot deliver.

The design in one line

Show a word naming an organ; ask the person to feel it; subtract a matched exteroceptive attention block.

That is nearly all of it. The sophistication is in the control condition and in what gets contrasted, not in the task.

Why the field uses it

Not because it is a good measure of interoception — it is a fairly poor one — but because the good measures cannot be applied to most of the body.

Mapping visceral cortex properly would mean perturbing a viscus and recording the response: fundus distension by barostat (Lu et al. 2004), bladder infusion (Jarrahi et al. 2015), isoproterenol bolus (Khalsa et al.). Each requires its own apparatus, each is invasive, and no two can be combined in a single session — which means that until this paradigm, a within-subject comparison of two organs was not merely undone but undoable. Perturbation methods also introduce a confound that is fatal for exactly that comparison: the activation includes the participant’s discomfort and alarm, which differ by organ and by how nasty the manipulation is.

The attention paradigm gives that up in exchange for asking almost nothing of the participant and nothing at all of the body.

What it actually measures

In the taxonomies’ vocabulary, this instrument operationalizes attention — Farb et al.’s “attention tendency,” Khalsa et al.’s “attention,” Berntson & Khalsa’s “attention.” It does not measure detection, magnitude, discrimination, accuracy, sensitivity, insight or coherence, because it has no correct answer. There is nothing the participant can get right.

This is worth stating flatly because the paradigm’s papers, including its best ones, routinely title themselves about awareness. Haruki & Ogawa name their conditions “cardiac interoceptive awareness” and “gastric interoceptive awareness” on the strength of an assumption stated once in their introduction: that “interoceptive attention has been found to elicit interoceptive awareness.” Perhaps. The paradigm does not test it, and the wiki’s own evidence that attention and accuracy dissociate — meditators attend more without detecting better (contemplative training increase objective interoceptive accuracy?) — is a reason to keep the two words apart.

The control condition is the measurement

Since the reported map is always a difference, the exteroceptive control determines what the paradigm can see. Haruki & Ogawa’s control is unusually well matched: a word (“TARGET”) in the same position as the organ cues, whose colour fades from black to grey in five imperceptible steps — so all three conditions are “sustain attention to a faint, non-salient signal for ten seconds,” varying only in where the signal is.

Even so, the control turned out to be the most demanding condition: participants rated the visual target as more intense than either bodily sensation, and it produced the study’s largest activations. One consequence, recorded on Insular cortex, is that the right dorsal anterior insula — the wiki’s canonical interoceptive-awareness region — was more active during the exteroceptive control than during interoceptive attention.

A paradigm whose subtrahend engages the region of interest more than the minuend does is a paradigm to read carefully.

What one study got out of it that no other design could

The organ-vs-organ contrast. Because both interoceptive conditions run in the same session with the same instructions and the same control, Haruki & Ogawa could ask which regions prefer the heart to the stomach — and found that the right anterior insula does, that gastric attention recruits visual cortex overlapping the Gastric network, and (by MVPA over anatomical subdivisions) that the left dorsal middle insula carries organ identity in its multivoxel pattern.

None of those questions is answerable with a perturbation method, a heartbeat task, or a questionnaire. That is the case for the paradigm, and it is a good one — provided its output is labelled attention.

Variants worth distinguishing

• Single-organ vs multi-organ. Most studies cue only the heart. Kerr et al. (2016) and Simmons et al. (2013) included stomach blocks but did not contrast the organs against each other, which is why Haruki & Ogawa could claim the comparison as unmade.
• With vs without physiological recording. Farb et al.’s breath-focus work records respiration; Haruki & Ogawa record nothing. The difference determines whether the study can say what the body was doing while it was being attended to — and it does more than housekeeping. Because Farb et al. record the breath, they can use the measured respiratory rate as the regressor rather than the instruction, turning the paradigm’s central weakness (no ground truth) into a partial strength: the localizer is anchored to a real viscerosomatic signal, and the study can show that attention changed which cortex tracked that signal (posterior insula when attended, somatosensory when not) rather than merely that “the insula lit up during breath-focus.” Their respiration-equated design also rules out the mechanical-breathing confound that an unrecorded study cannot exclude.
• Decoding rather than contrasting. Weng et al.’s EMBODY framework (EMBODY framework (brain decoding of internal attention)) trains participant-specific classifiers to detect interoceptive attention from brain activity, which converts the paradigm from a localizer into a measure of attention that neither asks the participant nor counts heartbeats. This is the most promising direction out of the paradigm’s central weakness, since a decoder supplies the verification the instruction cannot — for the training task. Note the irony the decoder then inherits: applied across-task to an uninstructed meditation, it acquires a fresh no-ground-truth problem one level up (see Weng et al. (2020)).