Schema-guided symptom perception
interoceptive-inference in clinical-psychology vocabulary, proposed thirteen years before Seth, by people who had never heard of a generative model.
The problem it was invented to solve
Van der Does et al. (2000) pooled 709 participants and found that more than 95% report perceiving their heart rate while roughly 80% are wrong by about 30%. That leaves a question the interoception literature does not usually have to face:
When more than 95% of all participants report that they perceive their heart rate, and when 80% of them appear to be wrong (by approximately 30%), what in fact do these individuals feel?
Not nothing — only 3.5% reported feeling no heartbeat at all. Everyone else felt something and counted it.
And the something has a shape. Participants do not report losing track of beats. They report feeling a regular rhythm, somewhat slower than their actual heart rate. A steady, wrong rhythm is not a degraded percept of a real one; it looks like a percept of something else.
The proposal
Following Pennebaker (1982), symptoms are organized in cognitive schemata formed from previous experience, and these schemata guide the perception and processing of current symptoms. Van der Does et al. apply it to panic:
Once a patient perceives a situation as threatening, an anxiety schema activates — bundling attentional shift, selective perception, expected high heart rate, other arousal symptoms, and anxiety itself. Perception of symptoms then becomes “more guided by the schema (that is, by past information) than based upon present physiological status.” Voluntarily shifting attention to the body — which is exactly what the heartbeat task instructs — may itself activate the schema.
So the count is a reading of the schema’s expected rhythm, not of the heart. And the schema, being built from past experience rather than present physiology, can be wrong in a stable, regular, confident way — which is what the self-reports describe.
Why this is interoceptive inference
Strip the vocabulary and compare:
| Van der Does et al. (2000) | Seth (2013) |
|---|---|
| symptom schema formed from past experience | prior / generative model |
| perception “guided by past information rather than present physiological status” | perception dominated by the prior when sensory precision is low |
| schema activated by perceived threat and by attentional shift | context-dependent precision-weighting |
| research programme: find “the conditions under which schema-guided versus physiology-based processing occur” | research programme: characterize precision-weighting of interoceptive prediction error |
The last row is the striking one. Van der Does et al. close by proposing that research “shift to investigating the conditions under which schema-guided versus physiology-based processing occur.” That is the predictive-coding research programme, stated as a next step, in 2000, in Behaviour Research and Therapy.
And their data supply a condition. The exercise result — accuracy appears above ~100 bpm and vanishes by ~95 — is, in this vocabulary, precision doing exactly what the theory says: when the afferent signal is loud, physiology-based processing wins; when it is quiet, the schema wins. Nobody in that experiment learned anything. The balance between prior and evidence shifted because the evidence got louder.
That is a cleaner demonstration of precision-weighting than most of what the predictive camp offers, and it was produced by people arguing about panic disorder. Recorded on feedforward-vs-predictive-interoception, where it is the first clinical evidence on the page and does not come from either camp.
What the wiki should not overclaim
The convergence is real but it is not endorsement. Van der Does et al. propose schema-guided processing as a “hypothetical” account of a finding they could not otherwise explain, in a discussion section, with no formal model, no test, and one piece of evidence that goes the wrong way — the somatosensory amplification scale (SSAS) failed to correlate with counted beats, which a straightforward “misinterpreting vague sensations” story predicts it should. They report the failure. Confidence ratings staying low even under strict instructions (60.4%) fits, weakly.
And it is not a mechanism. “A schema guides perception” names the phenomenon rather than explaining it; the predictive-coding vocabulary at least specifies a quantity (precision) that determines the balance. This is the reverse of the usual complaint about Seth — that predictive coding re-describes rather than explains — and it is worth noticing that the older, vaguer version has the same problem more severely.
Neither Pennebaker (1982) nor the surrounding psychosomatic literature is in raw/. The concept is recorded here for what it does to the heartbeat task’s interpretation. Its own evidence base is unread.
Consequences for the wiki
It changes what the heartbeat task’s failures mean. The wiki’s heartbeat-detection-task page treats inaccuracy as measurement error — noise around a true score. If the majority are reading a schema rather than a heart, their counts are not noisy estimates of their heart rate; they are accurate estimates of a different quantity. That is the mixture argument on is-the-heartbeat-counting-task-valid, and it is why the two positions there imply different statistics rather than different error bars.
It gives the predictive account a clinical foothold the wiki lacked. feedforward-vs-predictive-interoception has been a dispute among theorists reading anatomy and reviews. Here is a large behavioural dataset whose central puzzle — confident, regular, wrong percepts — is what a prior-dominated system produces and is awkward for any read-out account. Craig gestures at this (“distorted interoceptive predictions” in anxiety and functional somatic disorders) without formalizing it; this is the phenomenon he is gesturing at, measured.
And it suggests the task’s clinical use is upside down. If patients who report cardiac symptoms without cardiac events are schema-guided perceivers, then — as Van der Does et al. argue — interoceptive training might help them, by reconnecting perception to physiology or simply “by being informed that what they feel is not their heart.” See interoceptive-training-clinical, cognitive-model-of-panic.