A heartbeat is unusually convenient for science.
It repeats. It can be recorded. People can be asked whether they felt it. Their confidence can be measured. Their answers can be compared with the underlying cardiac event.
That apparent simplicity made heartbeat tasks central to interoception research.
It also made them easy to overinterpret.
Sarah Garfinkel's work has been important partly because it helped make the measurement problem harder in exactly the right way.
Her research asks how signals from the body, especially the heart, interact with brain processes to shape emotion, memory, attention and clinical symptoms. A central contribution has been to separate kinds of interoceptive performance that older language often bundled together.
That separation changed the field's vocabulary.
Current role and research programme
As of 23 September 2026, UCL lists Professor Sarah Garfinkel as Professor of Cognitive Neuroscience and leader of the Clinical and Affective Neuroscience group in the Faculty of Brain Sciences.
Her current institutional profile identifies research interests in emotion, interoception, autism, anxiety and psychosis. The UCL group describes a programme focused on how emotion is expressed in body and brain, with particular attention to the heart, heart-brain interactions and interoception.
The group combines peripheral physiological measures such as ECG with central measures including fMRI, and works across basic and clinical questions.
Those current roles should be distinguished from the University of Sussex and Brighton and Sussex Medical School affiliations appearing in much of Garfinkel's earlier influential work.
A research profile has to be dated because affiliations move while papers remain fixed.
The three-dimensional distinction
Garfinkel and colleagues' 2015 paper, "Knowing your own heart," became a key reference because it separated three dimensions:
- Interoceptive accuracy: performance on objective behavioral tests, in that paper centered on heartbeat detection.
- Interoceptive sensibility: self-evaluated or questionnaire/interview-based beliefs about one's interoceptive experience.
- Interoceptive awareness: metacognitive correspondence between objective performance and confidence.
In a normative sample of 80 participants, the three dimensions were dissociable.
That result has consequences far beyond terminology.
A person can believe they are highly body-aware without performing accurately on a particular heartbeat task. Another person can perform relatively well but have poor confidence calibration. Self-report, task performance and metacognition answer different questions.
The paper therefore challenged the habit of speaking about "interoceptive ability" as if it were one scalar trait.
The terminology later became more complicated
The field has continued to debate names and measures.
Some researchers use "interoceptive awareness" broadly for self-reported attentiveness, while Garfinkel's 2015 framework used it specifically for metacognitive insight into objective performance. Later taxonomies have proposed additional distinctions among attention, accuracy, sensibility and metacognition.
This is not a minor naming dispute.
If two studies use the same word for different operations, apparent disagreements may be partly semantic.
A careful Knowledge Library therefore records the operational definition alongside the term.
Garfinkel's enduring contribution is not that one taxonomy must be final. It is the insistence that dimensions should not be assumed to coincide.
Heart-brain interactions and fear
Garfinkel's programme also helped make the timing of cardiac signals experimentally relevant.
The heart does not send a continuous, uniform stream. Baroreceptor signaling varies across the cardiac cycle. Work by Garfinkel, Critchley and collaborators examined how cardiac timing can modulate perception of threat and other salient stimuli.
Their 2016 Trends in Cognitive Sciences review summarized evidence that fear processing can be selectively influenced by phasic cardiac afferent signals.
This work is often rendered publicly as "the heart controls fear."
That is too coarse.
The studies concern specific physiological phases, perceptual tasks and neural mechanisms. They show that brain and body timing interact. They do not show that cardiac state determines an emotion or that heart signals contain a hidden judgment about the environment.
The precision is the finding.
Interoception and emotion are coupled, not interchangeable
Garfinkel's work sits at a boundary between internal sensation and emotion.
Emotion has bodily components, and bodily signals can change emotional processing. But emotion also includes appraisal, context, learning, memory and action tendency.
Interoception is therefore neither an optional side channel nor a complete theory of emotion.
This matters clinically.
People with anxiety may experience strong bodily sensations, heightened attention to those sensations, inaccurate interpretations, altered physiological reactivity or some combination. A single label such as "poor interoception" cannot specify which process is involved.
Dimensional measurement is particularly valuable when symptom states can arise from different combinations of sensitivity, belief and metacognitive calibration.
Autism research made dissociation visible
Garfinkel and colleagues' 2016 study of autism examined discrepancies among interoceptive dimensions and their relations to emotion and anxiety.
The paper is important because it resists a simplistic group-deficit story.
Different components of interoception can diverge. A person may report strong interoceptive sensibility while objective performance shows another pattern. Clinical meaning can lie partly in the mismatch rather than in a uniformly low or high score.
The study helped motivate work on interoceptive discrepancy as a potentially important feature in autism and anxiety.
But the result should remain population- and task-bound.
One heartbeat paradigm does not define autistic embodiment. Group-level differences do not diagnose an individual. Autism is heterogeneous, and measurement validity remains an active concern.
Heartbeat tasks are both useful and contested
The heartbeat has generated some of the field's most familiar measures.
It has also generated extensive criticism.
Heartbeat counting can be influenced by beliefs about heart rate, time estimation and prior knowledge. Discrimination tasks can be difficult and may mix perception with decision criteria. Performance can depend on signal strength and physiology.
Contemporary systematic reviews of interoceptive-accuracy measures therefore recommend greater attention to task validity and modality.
This does not erase Garfinkel's dimensional contribution.
It changes how the original operationalizations should be used.
The conceptual distinction between performance, self-report and metacognition remains valuable even as the field develops better ways to measure each component.
The research programme is translational
UCL describes Garfinkel's work as spanning basic mechanisms and clinical conditions including anxiety, PTSD, autism and psychosis.
Translation is a demanding word.
A mechanism observed in a psychophysiology experiment does not become a treatment merely because it involves a clinical population. Intervention evidence requires separate trials, appropriate outcomes and safety assessment.
This boundary matters because interoception has become commercially attractive. Terms such as "nervous-system regulation" can move from laboratory findings to therapeutic claims with little friction.
Garfinkel's research provides mechanisms and candidate targets.
Those targets should not be presented as validated interventions without direct evidence.
Interoceptive training illustrates both promise and caution
Feedback-based interoceptive training has been investigated in Garfinkel's broader programme and in the field more generally. The possibility is compelling: if a person can improve calibration between internal signals and confidence, perhaps some forms of anxiety or emotional dysregulation could change.
But "training interoception" is not one intervention.
Training heartbeat detection, changing attention style, reducing catastrophic interpretation and increasing metacognitive accuracy are different targets.
The more interoception becomes an intervention language, the more important the dimensional framework becomes.
A programme should name what it is trying to change.
What this work contributes to epistemology
The 2015 framework can be read as a small theory of self-knowledge.
There is the signal.
There is performance in detecting the signal.
There is what the person believes about their ability.
There is confidence in a specific judgment.
Those layers can disagree.
The same architecture appears beyond heartbeat perception. People can be confident about memory and be wrong. They can underestimate expertise and perform well. They can feel certain about an emotion label while misreading context.
Garfinkel's work does not prove that all self-knowledge follows the same structure.
It offers a disciplined example of why subjective certainty should not be equated with accuracy.
Accuracy itself depends on how the task is built
The word accuracy can sound more objective than the measurement warrants. In heartbeat research, performance depends on the task. Counting one's heartbeats without feeling a pulse can be influenced by prior beliefs about typical heart rate and by time estimation. Synchrony or discrimination paradigms reduce some confounds but introduce others, including difficulty matching external stimuli to internal events.
This matters for interpreting the 2015 three-dimensional framework. The distinction among objective task performance, self-report and metacognition is conceptually important even if the field later improves the task used to instantiate the "accuracy" dimension.
A framework can remain useful while one measurement method is revised.
That is often how science matures: constructs survive by becoming more demanding about their operationalization.
Metacognition is the quiet center of the contribution
The most consequential part of the framework may be the confidence-accuracy relationship. A person who performs well but has no sense of when they are right has a different form of self-knowledge from someone whose confidence tracks performance trial by trial.
This is not unique to interoception. Metacognitive calibration matters in perception, memory, learning and professional judgment. What Garfinkel's work did was make the same architecture visible inside bodily perception.
That has methodological consequences. If an intervention claims to improve "body awareness," researchers can ask whether task performance changed, whether self-belief changed, whether confidence became better calibrated, or whether all three moved together. Those outcomes should not be treated as synonyms.
Clinical discrepancy is not automatically dysfunction
A mismatch between subjective sensibility and performance can be clinically interesting, but mismatch should not be moralized.
Self-report captures experiences that laboratory tasks may miss. A person can be highly attentive to internal state while performing poorly on a particular heartbeat paradigm because that paradigm samples a narrow channel under artificial conditions. Conversely, strong task performance need not imply useful body awareness in daily life.
Discrepancy therefore becomes a research question rather than a deficit label. Which dimension predicts symptoms, functioning or treatment response? Does the mismatch persist across modalities? Does it change with state?
The answer may differ across anxiety, autism, psychosis and nonclinical samples.
Translational work needs target engagement
If interoceptive training is intended to change a mechanism, a strong trial should show that the proposed mechanism changed, not only that symptoms improved.
For example, a feedback intervention might improve metacognitive calibration while leaving raw detection unchanged. Another intervention might reduce catastrophic interpretation without changing either detection or confidence. Both could be useful, but they would work through different targets.
This target-engagement logic prevents "interoception" from becoming a decorative explanation attached to any body-focused intervention after the fact.
What the evidence does not show
Garfinkel's research does not show that heartbeat perception is a complete measure of interoception.
It does not show that interoception is one stable ability across organs.
It does not show that stronger bodily sensitivity is always healthier.
It does not show that cardiac phase determines what a person will feel.
It does not show that autistic or anxious people share one interoceptive profile.
It does not show that self-report is invalid simply because it can diverge from task performance.
It does not show that improving heartbeat-task performance treats anxiety, autism, PTSD or psychosis.
It does not show that heart-brain coupling provides direct access to truth about external events.
And it does not justify commercial claims that a generic body-awareness practice is "based on" this research without testing the specific intervention.
Current relevance to human-AI systems
Why does a psychophysiology research programme matter in an AI-age knowledge library?
Because AI systems increasingly produce interpretations of human state from behavior and biosignals.
Garfinkel's dimensional distinctions are a warning against collapsing layers.
A wearable signal is not subjective experience. A questionnaire is not objective accuracy. Confidence is not performance. Performance on one modality is not global self-knowledge.
Systems that infer anxiety, engagement or readiness from sparse physiological data risk committing exactly the category errors this literature helps expose.
The more machine systems claim to read people, the more valuable construct separation becomes.
Institute synthesis: self-knowledge has layers
The Institute uses layered self-knowledge as a synthesis phrase for keeping signal, task performance, subjective belief and metacognitive calibration distinct.
This phrase is not Garfinkel's formal construct and should not be attributed to her.
The synthesis matters because inner development is often discussed as if increased awareness were enough.
Awareness of what?
Measured how?
With what calibration?
Against what external criterion?
Garfinkel's work makes those questions harder to avoid.
Why this profile belongs independently
Garfinkel's programme connects interoception, emotion, clinical neuroscience and measurement in a way that has shaped the field's conceptual vocabulary.
A profile is justified not because the researcher is prominent, but because following the programme makes several Knowledge Library entities more legible: interoceptive accuracy, sensibility, awareness, heartbeat perception, emotion, anxiety and the interpretation of bodily signals.
The profile should remain a map into the work.
It is not a substitute for reading the primary papers or the debates that followed them.
Editorial limitation note
This profile maps a research programme rather than endorsing every interpretation associated with it. Heartbeat tasks have known validity limits, group-level findings do not diagnose individuals, and academic affiliation is not evidence for a scientific claim.
Research references
- University College London Faculty of Brain Sciences. Sarah Garfinkel: current institutional profile. Accessed 23 September 2026. https://www.ucl.ac.uk/brain-sciences/sarah-garfinkel
- University College London. Clinical and Affective Neuroscience research group. Accessed 23 September 2026. https://www.ucl.ac.uk/brain-sciences/icn/research/research-groups/clinical-and-affective-neuroscience
- Garfinkel SN, Seth AK, Barrett AB, Suzuki K, Critchley HD. Knowing your own heart: distinguishing interoceptive accuracy from interoceptive awareness. Biological Psychology. 2015;104:65-74. DOI: 10.1016/j.biopsycho.2014.11.004. https://pubmed.ncbi.nlm.nih.gov/25451381/
- Garfinkel SN, Tiley C, O'Keeffe S, Harrison NA, Seth AK, Critchley HD. Discrepancies between dimensions of interoception in autism: Implications for emotion and anxiety. Biological Psychology. 2016;114:117-126. DOI: 10.1016/j.biopsycho.2015.12.003. https://pubmed.ncbi.nlm.nih.gov/26724504/
- Garfinkel SN, Critchley HD. Threat and the Body: How the Heart Supports Fear Processing. Trends in Cognitive Sciences. 2016;20(1):34-46. DOI: 10.1016/j.tics.2015.10.005. https://pubmed.ncbi.nlm.nih.gov/26628111/
- Desmedt O, Luminet O, Walentynowicz M, Corneille O. The new measures of interoceptive accuracy: A systematic review and assessment. Neuroscience & Biobehavioral Reviews. 2023;153:105388. DOI: 10.1016/j.neubiorev.2023.105388. https://pubmed.ncbi.nlm.nih.gov/37708919/