Inside the headset
Ishaan is 27 and builds virtual reality games in Pune. One Saturday he volunteers for a study at a university laboratory. A researcher, Dr Aditi Kulkarni, fits him with a headset, places electrodes on his scalp and chest, and seats him on a stool. Through the headset he sees a live video of a man's back, filmed by a camera two metres behind him. The man is sitting on a stool. It takes Ishaan a moment to realise that the man is him.
Ishaan, Aditi and the laboratory are invented. I will use them throughout this paper, changing one detail at a time, so that each experiment has a person attached to it. The studies, numbers and patients I cite around them are real.
Aditi begins to stroke Ishaan's back with a stick. He feels the stick and, in the headset, sees it touch the back of the man in front of him at the same moment. After a minute, something odd happens. The man in the video starts to feel like him, as if he were standing a little forward of where his body really sits. When Aditi later asks him to step back to where he thinks he was, he overshoots towards the figure he saw.
This is the full-body illusion, and it is one of the cleanest ways to move a person's sense of self. It works because the brain combines touch and vision into a single best guess about where the body is and which body is its own. When the two signals agree in time, the guess shifts towards the body that is seen.
But Aditi is interested in something less obvious. Every time Ishaan's heart beats, a small wave passes through the electrical activity of his brain about a quarter of a second later. This wave is the heartbeat-evoked potential, or HEP. Aditi wants to know whether that wave changes when the illusion takes hold, and whether it is merely a passenger or part of what holds Ishaan's point of view in place.
That is the question this paper takes from the series list. Do cardiac-evoked potentials in the anterior insula causally anchor the first-person perspective during rubber-hand and full-body illusions? The anterior insula is a region folded deep inside each side of the brain, long associated with the sense of the body's interior. The first-person perspective is the sense of experiencing the world from here, from inside this body.
The question is a test of a larger theory. Anil Seth, a neuroscientist at the University of Sussex, has argued that we are beast machines: living systems whose consciousness and selfhood are grounded in the brain's continuous prediction and regulation of its own body. If that is right, the heartbeat is not a background noise but one of the signals the self is built on. The HEP studies are among the most direct attempts to see that foundation at work.
My answer, stated early so the rest of the paper can test it, is that the evidence shows heartbeat responses tracking the illusion, not anchoring the self. The best scalp study was carefully controlled, but it localised the effect mainly to the posterior cingulate cortex. The insula evidence rests on two patients. And no study has yet manipulated the heartbeat signal and shown the point of view move. I will set out the theory, the illusions, the heartbeat studies, a toy analysis pipeline, the causal gap, and a reading through the layers of the Recursive Tattva Model, before giving the strongest case against my view.
Why this question matters
The question is not only of academic interest. If the first-person perspective rests on the brain's response to the heartbeat, then conditions that change cardiac signalling might change the self. That would matter for people with heart disease, for people fitted with devices that alter the pulse, and for theories that link anxiety and depersonalisation to disturbed interoception. It would also give consciousness science something it rarely has: a bodily signal that can be measured, timed and, in principle, manipulated.
If the answer is no, the question still teaches something. It shows how far a well-controlled correlation can travel before it needs a causal test, and where the beast-machine theory places its strongest bets.
The beast-machine idea
The beast-machine theory starts from a view of the brain as a prediction engine. On this view, the brain does not passively receive signals from the senses. It constantly generates predictions about what those signals will be, and uses the differences between prediction and signal, called prediction errors, to update its guesses. What we perceive is the brain's best current guess, not the raw input.
Most work on this idea concerns vision and hearing. Seth's contribution was to apply it to the inside of the body. In a 2013 paper in Trends in Cognitive Sciences, he described "interoceptive inference": the brain's predictive modelling of signals from the heart, lungs, gut and blood vessels. He proposed that emotions depend on these interoceptive predictions, and that the experience of body ownership is shaped by predictive integration of interoceptive and exteroceptive signals.
Interoception is the sense of the body's internal state. It includes the heartbeat, breathing, hunger, temperature and the state of the gut. Exteroception is the sense of the outside world through sight, hearing and touch. Proprioception, the sense of where the limbs are, sits between them.
From prediction to regulation
The theory took a further step in a 2018 review by Seth and Manos Tsakiris, titled "Being a Beast Machine: The Somatic Basis of Selfhood". The central idea is that the brain's predictions about the body are not mainly for knowing the body. They are for controlling it. The brain predicts what the heart rate and blood pressure should be, and acts to keep them within the range that keeps the organism alive.
On this view, the most basic layer of selfhood is not a story or a memory. It is the felt sense of being a living body that the brain is working to keep alive. Seth developed the idea for a general audience in his 2021 book Being You, which argues that our brains are prediction machines and that our sense of self derives from our body.
Why the heart, of all signals
The heart is not the only interoceptive signal, but it is the one researchers reach for first. There are practical reasons. The heartbeat is regular, it has a sharp electrical marker in the ECG, and its timing can be recorded to the millisecond with a few electrodes on the chest. Breathing is slower and harder to time precisely. Signals from the gut are slower still and harder to record without invasive methods.
The heart also has a theoretical appeal. Each beat sends a wave of pressure through the arteries, and receptors in the major vessels report it to the brainstem. The brain therefore receives a steady stream of signals that it did not cause by moving, and that it must predict if it is to regulate blood pressure. That makes the heartbeat a natural test case for interoceptive prediction.
The convenience of the heartbeat is also a risk. Because it is easy to measure, it may receive more weight in the theory than it deserves. A finding about the heartbeat is a finding about one signal among many. When this paper concludes that heartbeat responses do not anchor the self, it leaves open whether interoception as a whole does.
The anterior insula as a candidate hub
The theory does not require a single brain region to carry the self. But one region has attracted particular attention. In a much-cited 2009 review in Nature Reviews Neuroscience, the neuroanatomist A. D. (Bud) Craig argued that the anterior insular cortex re-represents interoceptive signals and may provide a basis for all subjective feelings. He suggested that it needs to be considered as a potential neural correlate of consciousness.
Craig's view made the anterior insula the natural place to look for a bodily anchor of the self. If interoceptive signals reach awareness through the anterior insula, then heartbeat-related activity there might be where the beast machine's predictions about the heart meet the sense of being someone. The question in this paper names that region for that reason.
What the theory predicts about illusions
If the self rests on integrating interoceptive and exteroceptive signals, then manipulating either kind of signal should change the self. Bodily illusions manipulate exteroceptive signals, such as the timing of touch and vision. The beast-machine theory predicts that interoceptive signals should matter too: that a heartbeat made visible, or a change in how the brain processes each heartbeat, should shift bodily self-identification.
It predicts something stronger as well. If interoceptive prediction is the deepest layer of selfhood, then the first-person perspective should depend on it. Disrupting heartbeat processing should disrupt the point of view, not merely the sense of which hand is one's own. This stronger prediction is the one the series question asks about, and it is the one with least evidence.
Bodies the brain borrows
The rubber-hand illusion became the standard way to study body ownership. It needs only a rubber hand, two paintbrushes and a screen. When the strokes are synchronous, people report that the rubber hand feels like theirs, and they misjudge the position of their real hand towards it, an effect called proprioceptive drift. When the strokes are out of step, the illusion is weaker or absent.
In 2007, Bigna Lenggenhager, Olaf Blanke and colleagues extended the idea to the whole body. In a study in Science titled "Video ergo sum", they used conflicting visual and touch signals in virtual reality. During synchronous stroking, participants felt as if a virtual body seen in front of them was their own and mislocalised themselves towards it, to a position outside their bodily borders. This is the illusion Ishaan experienced.
That case matters for this paper because it is causal. Stimulating a particular region produced a change in the first-person perspective. It is the kind of evidence the series question asks for. But the region was the angular gyrus, near the junction of the temporal and parietal lobes, not the insula. And the signals implicated were touch and balance, not the heartbeat.
Making the heartbeat visible
The next step was to put the heartbeat into the illusion. In a 2013 study in Neuropsychologia, Keisuke Suzuki, Sarah Garfinkel, Hugo Critchley and Seth presented a virtual hand in augmented reality that pulsed visually either in time with the participant's own heartbeat or out of time. Subjective and objective measures of ownership of the virtual hand were enhanced when the feedback was in time with the heartbeat.
In the same year, Jane Aspell, Blanke and colleagues reported a cardiac version of the full-body illusion in Psychological Science. Participants saw a video of their virtual body outlined by a silhouette that flashed in synchrony with their heartbeat. Synchronous flashing increased self-identification with the virtual body, shifted self-location towards it, and altered the perception of touch on the participants' backs.
Predict before reading on
?> Before reading on, make a guess. If a virtual hand or body flashing with your heartbeat makes it feel more like yours, what happens when your own face, morphed with a stranger's, flashes with your heartbeat?
In a series of three studies, Giuseppina Porciello and colleagues showed participants ambiguous faces, morphed between their own face and a stranger's, flashing either with or against their heartbeat. In the published report, there was no difference between synchronous and asynchronous stimulation on self-face attribution in any of the three studies. The authors concluded that visuo-cardiac synchrony does not boost self-attribution of the face as it does for the hand and full body.
That null result is informative. It suggests that the cardiac effect is not a general stamp of selfhood applied to anything that pulses with the heart. It depends on the kind of body representation involved. Hands and bodies, which the brain localises in space using touch and position, are affected. Faces, which are recognised mainly by appearance, may not be.
The metronome thought experiment sharpens what the cardiac illusions show. They show that the brain integrates the timing of the heartbeat with visual signals when deciding which body is its own. That is consistent with the beast-machine theory. It does not yet show that heartbeat processing is the anchor of the self rather than one more timing cue that the brain combines, alongside touch.
What synchrony means to the brain
It helps to ask why synchrony matters at all. A body is a source of many signals at once. When something touches the skin, the touch is felt, and if the eyes are open, it is seen at the same moment and in the same place. The brain has learned that signals which arrive together usually come from the same object. Synchrony is evidence of a common cause.
The illusions exploit that inference. When a rubber hand is seen being stroked at the same moment as the hidden real hand is felt being stroked, the most economical explanation available to the brain is that the rubber hand is the source of the feeling. The cardiac versions add another signal to the same logic. A flash that coincides with the heartbeat is evidence that the flashing thing and the beating heart belong to one body.
On this reading, the heartbeat enters the illusion as a timing cue, like touch. That is fully consistent with Seth's predictive framework, which treats every signal as evidence to be weighed. But it also shows why the cardiac illusions cannot, on their own, establish that interoception is special. They would work the same way for any signal the brain reliably associates with its own body.
Does interoceptive skill predict the illusion?
If interoception matters for body ownership, then people who are better at sensing their heartbeat might be less easily fooled by a rubber hand. In a 2011 study in Proceedings of the Royal Society B, Tsakiris and colleagues reported exactly that. People with low interoceptive sensitivity, measured with a heartbeat-monitoring task, experienced a stronger illusion of ownership over the rubber hand.
The finding was influential, but it has not held up well. A 2024 study in Frontiers in Psychology notes that several later studies were unsuccessful in replicating the association, and its own results for the classical rubber-hand illusion supported the lack of association. Part of the problem may be the measure. In a 2018 analysis of 572 participants, Giorgia Zamariola and colleagues found that heartbeat-counting scores overwhelmingly reflected under-reports and that the correlation between actual and reported heartbeats was low overall, at 0.16.
Are the illusions what they seem?
A more radical challenge came from Peter Lush and colleagues, including Seth. In a 2020 paper in Nature Communications, they reported, in samples of 156, 404 and 353 participants, substantial relationships between hypnotisability and measures of the rubber-hand illusion. They argued that people may control their experience to meet expectations arising from the perceived requirements of a task.
This was a serious claim. If the illusion is partly produced by participants' expectations, then any brain response that tracks it might be tracking expectation rather than body ownership. A reply in the same journal reanalysed the 353-participant dataset using the difference between synchronous and asynchronous conditions. It found no significant relationship between hypnotic suggestibility and the condition-specific illusion ratings, and argued that the illusion is primarily a perceptual effect arising from multisensory mechanisms.
The dispute is not settled, but it changes what counts as good evidence. Studies that compare synchronous with asynchronous conditions, and that measure something participants cannot easily control, are more convincing than studies that rely on ratings in a single condition.
The heartbeat in the brain
The heartbeat-evoked potential is the brain's electrical response to each heartbeat. It is measured by recording the electrocardiogram, the ECG, alongside the electroencephalogram, the EEG, or magnetoencephalogram, the MEG. The researcher finds each R-peak, the sharp spike in the ECG that marks each beat, and averages the brain signal in a window locked to those peaks. What remains after averaging is the part of the brain signal that reliably follows the heart.
Each heartbeat sends signals to the brain through several routes. Pressure sensors in the major arteries, called baroreceptors, fire when blood is ejected from the heart. The heart's own movement stimulates receptors in the chest wall. Each route has its own delay, and the HEP is usually measured in a window from about 200 to 600 milliseconds after the R-peak.
Heartbeats and perception
The idea that the HEP matters for consciousness predates the illusion studies. In a 2014 study in Nature Neuroscience, Hyeong-Dong Park, Catherine Tallon-Baudry and colleagues used MEG to show that heartbeat-locked neural events before a stimulus predicted whether people would detect a faint visual grating. The effects appeared in the posterior right inferior parietal lobule and the ventral anterior cingulate cortex.
In 2016, Mariana Babo-Rebelo and colleagues in the same group extended the idea to self-related thought. In a study in the Journal of Neuroscience, 16 participants let their minds wander and were interrupted at random to rate their current thought. Heartbeat-evoked responses in the left ventral precuneus distinguished thoughts rated high or low on an "I" scale, the self as subject. Responses in the left ventromedial prefrontal cortex distinguished thoughts rated high or low on a "Me" scale, the self as the object of thought.
| Study | Measure | Region reported |
|---|---|---|
| Park 2014 | Faint grating detected | Right inferior parietal; ventral anterior cingulate |
| Babo-Rebelo 2016 | "I" in mind-wandering | Left ventral precuneus |
| Babo-Rebelo 2016 | "Me" in mind-wandering | Left ventromedial prefrontal |
| Park 2016 | Full-body illusion | Posterior cingulate; insula did not survive correction |
| Park 2018 | Full-body illusion | Left insula, two patients |
Neither of those regions is the anterior insula. Both are parts of the default network, a set of midline regions active during self-related thought. This is a recurring pattern in the HEP literature. The heartbeat response linked to self and perception appears across several regions, and the insula is only one of them.
The key study
The study closest to the series question is a 2016 paper in the Journal of Neuroscience by Park, Blanke and colleagues. Seventeen participants experienced the full-body illusion exactly as Ishaan did, viewing their own back from cameras two metres behind them. In the synchronous condition the video was live; in the asynchronous condition it was delayed by half a second. One participant was excluded for a poor ECG, leaving 16.
During stroking, HEP amplitude differed between synchronous and asynchronous conditions in a window 250 to 305 milliseconds after the R-peak, over frontocentral scalp regions. Source analysis placed the difference in the bilateral posterior cingulate cortex, extending into the supplementary motor area. At a lower threshold, a cluster appeared in the left insula, but it did not survive correction for multiple comparisons.
The authors were careful about artefacts. They created 200 sets of surrogate R-peaks by shifting the real ones at random while preserving the intervals between beats, and found that the effect disappeared, which shows that it was truly locked to the heartbeat. The ECG itself did not differ between conditions in the HEP window, nor did heart rate, heart-rate variability or heartbeat-locked breathing. Within participants, HEP amplitude correlated positively with a combined rating of self-identification and illusory touch.
This is a well-controlled study, and its result is real. But it does not show what the series question asks. The effect was localised mainly to the posterior cingulate, not the anterior insula. It concerned self-identification and illusory touch, not the first-person perspective or self-location, which the study did not measure. And it was correlational.
Inside the insula
Two years later, Park and colleagues recorded directly from inside the brain. In a 2018 intracranial study in Cerebral Cortex, eight patients with epilepsy had depth electrodes implanted for clinical reasons. Of 474 recording sites, 48, or 10 per cent, showed heartbeat-related phase concentration. About a third of those were in the insula or operculum.
Four of the patients then performed the full-body illusion. Two of them reported much stronger self-identification in the synchronous condition, rating it 7 on a 7-point scale against 1 in the asynchronous condition. In both, a recording site in the left insula showed a condition-related HEP difference during stroking but not during rest. The authors stated plainly that the methods were correlative in nature, and suggested that a follow-up study using transcranial magnetic stimulation could reveal the causal link.
That is the strongest direct evidence for the insula's role. It comes from two people, and the authors themselves called for the causal test that would answer the series question.
Why the posterior cingulate matters
The localisation of the 2016 effect deserves more attention than it usually receives. The posterior cingulate cortex is a hub of the default network and has been linked to self-related processing, spatial orientation and the integration of bodily signals with a sense of location. It is plausible that a region concerned with where the self is in space would respond differently to heartbeats when the self seems to move.
That is a reasonable hypothesis, but it is not the anterior insula hypothesis. It suggests a network in which heartbeat signals reach several regions, each using them for a different purpose. The insula may relate them to feeling, the posterior cingulate to self-location, and the ventromedial prefrontal cortex to self-related thought. On that picture, there is no single anchor, and the question of which region holds the self in place may have no single answer.
How reliable is the HEP?
A 2021 systematic review and meta-analysis by Michel-Pierre Coll and colleagues assessed the HEP as a marker of interoception. It included 45 studies in the review and 42 in the meta-analyses. It found moderate-to-large effects of attention, arousal and clinical status on HEP amplitude, and a moderate association between HEP amplitude and behavioural measures of interoception.
The review also found that studies varied in how they addressed key confounds, especially the cardiac field artefact, and that measurement protocols were not standardised. The authors concluded that the reliability of the meta-analytic effects remains unknown and that effects may be driven by cardiac or somatosensory factors. The cardiac field artefact is the heart's own electrical field, which reaches the scalp and can contaminate EEG recordings in exactly the window used to measure the HEP.
The timing of fear
One more line of evidence shows that the heartbeat shapes experience in a way that is clearly not artefact. In a 2014 study in the Journal of Neuroscience, Garfinkel, Seth, Critchley and colleagues presented faces at systole, when blood is ejected and the baroreceptors signal each heartbeat, or at diastole, between beats. Fearful faces were detected more easily and rated as more intense at systole, and amygdala responses to them were larger.
This is a causal effect of cardiac timing on perception, achieved by choosing when stimuli appear. It suggests a design for the series question: present the synchronous stroking of a full-body illusion only at systole or only at diastole, and see whether self-identification or self-location changes. As far as I can find, that experiment has not been published.
A toy analysis pipeline
The HEP literature turns on a practical question: when a study reports that the heartbeat response differs between two conditions, how sure can we be that the difference is neural and locked to the heart? I built a small instrument to make that question concrete. It is a toy that simulates the analysis, not a model of real brain signals, and its numbers are illustrative.
How it works
The instrument simulates 16 participants, the sample size of the 2016 study, each tested in a synchronous and an asynchronous condition. For each participant and condition it produces an HEP amplitude made of four parts. The first is a true neural response locked to the heartbeat, which may or may not differ between conditions. The second is a cardiac field artefact, proportional to the amplitude of the ECG. The third is a shift in brain activity between conditions that is not locked to the heartbeat at all, such as a general change in attention. The fourth is noise.
It then runs two analyses. The naive analysis asks only whether the HEP differs between conditions, using a paired test at the 5 per cent level. The full pipeline adds two controls that mirror those in the 2016 study. It requires the difference with real heartbeats to exceed the difference with surrogate heartbeats, shifted at random, and it requires the ECG amplitude not to differ between conditions.
Four scenarios
I ran four scenarios, each 500 times. In the null scenario, nothing differs between conditions. The naive test called 6 per cent of studies positive, and the full pipeline 7 per cent, close to the expected 5 per cent false-positive rate.
In the true-effect scenario, the neural heartbeat response really differs between conditions. The naive test detected it in 61 per cent of studies, and the full pipeline in 36 per cent. The controls cost power, because comparing real with surrogate triggers adds noise.
In the artefact scenario, the neural response does not differ, but the ECG amplitude is larger in one condition, perhaps because the participant sits differently. The naive test called 56 per cent of these studies positive. The full pipeline called none of them positive, because the ECG check caught the difference every time.
In the offset scenario, brain activity shifts between conditions, but not in a way locked to the heartbeat. The naive test called 61 per cent positive, exactly as often as in the true-effect scenario. The full pipeline called 7 per cent positive, the false-positive level, because the surrogate triggers carry the same offset and the comparison removes it.
A naive HEP analysis cannot tell a true heartbeat effect from a shift that has nothing to do with the heart. In the toy, the two produce identical positive rates. Only the surrogate comparison separates them.
Sample size
I also asked how the full pipeline's sensitivity depends on sample size. For the strong effect, it rose from 26 per cent with 8 participants to 36 per cent with 16 and 58 per cent with 32. For a weaker effect of half the size, it reached only 14, 13 and 19 per cent, barely above the false-positive level.
Strong effect (solid)Weak effect (dashed)5% false-positive level (dotted)
Reading the 2016 study through the instrument
The 2016 study is the positive case for this instrument. It used both controls: surrogate R-peaks and an ECG comparison, and also checked heart rate, heart-rate variability and breathing. Its effect survived all of them. On the instrument's logic, that result is not an artefact or a non-locked offset. It is a heartbeat-locked difference in brain activity between conditions that differ in how strongly people identify with the virtual body.
The negative control is the offset scenario. A study that reported only a naive HEP difference, without surrogate triggers, would look the same as a true effect, and the instrument shows that it could be produced entirely by a change unrelated to the heart. Not every HEP study in the literature used surrogate controls, which is part of what the 2021 meta-analysis meant when it said reliability remains unknown.
What the instrument cannot do
The instrument can tell whether an effect is locked to the heartbeat. It cannot tell whether the effect causes anything. A heartbeat-locked difference that tracks self-identification could be a cause of the illusion, a consequence of it, or a parallel effect of something else, such as attention to the body. Every scenario in the toy is correlational, because every real study is.
It also cannot say where in the brain an effect arises. Scalp EEG mixes signals from many regions, and source localisation adds its own uncertainty. That is why the 2016 study's localisation to the posterior cingulate, and the 2018 intracranial finding in the insula, carry different kinds of weight.
What a good HEP report contains
The instrument suggests a short list of what any HEP study of selfhood should report. It should show the ECG in both conditions, in the same time window, with a test of difference. It should report heart rate and heart-rate variability. It should run the same analysis with surrogate heartbeat times and show that the effect is larger with real ones. It should state the sample size and the number of heartbeats per condition. And it should say whether the analysis window was chosen in advance.
The 2016 study meets almost every item on that list. Many studies in the wider HEP literature do not, which is why the 2021 meta-analysis could not judge the reliability of the pooled effects. A reader who meets a new claim about heartbeats and the self can use the list as a filter before taking the claim seriously.
What the toy cannot show
The toy pipeline has clear limits, and they should be stated. It models each participant with a handful of numbers, not with real brain signals spread across many electrodes and time points. It assumes the artefact and the neural response add together in a simple way. It ignores the problem of choosing a time window after looking at the data, which in real studies can inflate false positives far beyond the levels shown here.
It also says nothing about where in the brain an effect comes from. Localisation requires source modelling or intracranial recording, and each brings its own assumptions. The toy is a teaching device for one question: can a given analysis tell a heartbeat-locked effect from a general shift and from artefact? On that question it gives a clear answer. On every other question it is silent.
The causal gap
The series question asks whether heartbeat responses in the anterior insula causally anchor the first-person perspective. Three separate things would need to be shown. First, that the heartbeat response in the anterior insula changes with the first-person perspective. Second, that changing that response changes the perspective. Third, that without it there is no stable perspective. This is what anchoring would mean.
The evidence so far meets part of the first requirement for self-identification, and none of the others. The 2016 scalp study localised its effect mainly outside the insula. The 2018 intracranial study found insula effects in two patients, without specifying the anterior part as the locus. Neither measured the first-person perspective or self-location directly. No study, as far as I can find, has manipulated insular heartbeat processing and measured the effect on the point of view.
The patient without insulae
That case does not concern the full-body illusion. But it tests the strong form of the insula hypothesis directly, and the hypothesis fails. A man without insulae still had feelings and, as the paper's title puts it, sentience. A 2021 study in Scientific Reports notes, citing earlier work, that a patient with extensive damage to the insula and anterior cingulate showed preserved heartbeat perception after a drug that raised his heart rate.
Hearts without a beat
The pulseless case is not a clean experiment. The native heart usually still beats, the ECG is still present, and the brain receives other interoceptive signals. But it shows that the first-person perspective survives a drastic reduction in the arterial heartbeat signal. If the heartbeat is an anchor, it is one the self can lose without drifting.
Two ways to get the same result
The 2021 meta-analysis found that attention has moderate-to-large effects on HEP amplitude. That makes the second Ishaan a live possibility. Synchronous stroking is more engaging than asynchronous stroking, and engagement with the body could alter the heartbeat response without the heartbeat response doing any work.
The only way to separate the two Ishaans is to intervene. Change the heartbeat signal, by timing the stroking to systole or diastole, or by stimulating the relevant region, and see whether the illusion changes. Or change attention without changing synchrony, and see whether the heartbeat response follows attention or the illusion.
Designing the causal test
Three experiments could close the causal gap, and each is within reach of existing laboratories. The first is the timing design already described: deliver synchronous stroking only at systole or only at diastole, and measure self-location by asking participants to walk back to where they feel they were. Because participants cannot easily tell when their heart is in systole, this measure would be hard to shape to expectation.
The second is stimulation. The 2018 intracranial study suggested transcranial magnetic stimulation with a coil designed to reach deep regions. A more direct route would use the depth electrodes already implanted in epilepsy patients, which allow brief electrical stimulation for clinical mapping. If stimulating an insular site during the illusion changed self-identification, while stimulating a nearby control site did not, that would be causal evidence for the insula specifically.
The third is a dissociation of attention from illusion. Ask participants to attend closely to their body during asynchronous stroking, and loosely during synchronous stroking. If the heartbeat response follows attention rather than the illusion, the second Ishaan is the right model. If it follows the illusion regardless of attention, the first Ishaan gains support.
None of these experiments would be decisive alone. Together they would answer the question in this paper's title more directly than the past decade of correlational work.
Reading the result in five layers
The Recursive Tattva Model, or RTM, is the framework this series uses to organise claims about consciousness. It distinguishes five layers. L1, spanda, is the dynamic activity of a system. L2, prakāśa, is manifestation: the fact that anything appears at all. L3, vimarśa, is reflexive grasp: the capacity to take what appears as one's own and act on it. L4, pratyabhijñā, is recognition: the sense of being the same self over time. L5, cit, is consciousness as ground. RTM is a hypothesis, and I use it here to separate questions that the beast-machine debate tends to run together.
The heartbeat is the most literal example of L1 one could ask for: a rhythmic pulsation that continues from before birth to death. It is tempting to read spanda as the heartbeat and to conclude that the self rests on it. That would be a mistake of level. L1 in RTM refers to the dynamics of the whole system, and the heartbeat is one rhythm among many, including breathing, the gut and the brain's own oscillations.
The full-body illusion concerns L3. To identify with a body is to take it as one's own, which is reflexive grasp. The 2016 study shows that a heartbeat-locked signal covaries with that grasp. On the RTM reading, this is evidence that L1 dynamics feed L3, which is unsurprising. It is not evidence that L1 anchors L3.
The first-person perspective, in the sense of experiencing from here, sits closer to L2, because it concerns the point from which things appear. The studies reviewed here do not reach L2. They measure self-identification and self-location, which are L3 judgements about a body, not the fact of appearing from somewhere.
The Damasio case and the pulseless patients fit this layering. They show that L4, the narrative self, can change, and that L3 judgements about the body can be altered, while the first-person point of view persists. If anything anchors that point of view, the evidence suggests it is distributed and partly subcortical, not tied to one cortical region's response to one bodily rhythm.
A contemplative experiment
The contemplative experiment is not evidence about the insula. It illustrates the distinction the layered reading relies on. Interoceptive signals can strengthen the sense of being a body, which is L3, without becoming the point from which experience happens. The beast-machine theory may be right that bodily regulation shapes the self, and still not have shown that it is the self's anchor.
A rival camp cuts against the layering
The 2022 reply weakens this objection for the rubber-hand illusion, because it found no specific link between suggestibility and the synchronous-minus-asynchronous difference. But it does not remove it. The full-body illusion relies heavily on ratings, and the 2016 study's correlation was with ratings. A measure that participants cannot shape, such as a physiological response to a threat to the virtual body, would make the evidence much harder to explain away.
What the theory means for machines
The beast-machine theory has an implication that Seth has argued for directly. In a target article in Behavioral and Brain Sciences, he makes the case that consciousness depends on our nature as living organisms. He concludes that real artificial consciousness is unlikely along current trajectories, but becomes more plausible as AI becomes more brain-like or life-like. The article drew 50 published commentaries, many of them critical. I take no position on whether any machine has experience, and nothing in this paper bears on machine L5.
The narrower point is methodological. The HEP work shows how hard it is to establish that one signal anchors the self even in human beings, where we can record from the brain, manipulate the body and ask for reports. Claims that a machine does or does not have a self because it lacks or has a body-like control loop face the same problem in a harder form. They need to say what result would count against them.
| Analogy | What it maps to | Status |
|---|---|---|
| Thermostat predicting the house | Interoceptive regulation | ✓ holds Prediction for control. |
| A box with no stake in the house | An organism staying alive | ✗ breaks The beast machine has a stake. |
| Drum on the recording | Cardiac field artefact | ✓ holds Louder drum, changed recording. |
| Switch the drum off | Remove the heartbeat | ✗ breaks Cannot; use ECG and surrogate checks. |
| Anchor and boat | Heartbeat and perspective | ~ strained The self may hold by many lines. |
Holding the layers apart
Why insist on separating L2 from L3 at all? Because the two make different predictions. If the point of view and self-identification are the same thing, then anything that weakens self-identification should weaken the sense of seeing from here. In the full-body illusion that does not seem to happen. Participants report that the virtual body feels more like theirs, and some report drifting towards it, but they do not report losing the perspective from which these things are felt.
That observation is informal, and the studies did not ask about it directly. A simple addition would help: a question, asked in every condition, about whether the participant felt they were experiencing the scene from a location at all. If the answer never changes while self-identification does, the layers come apart in practice as well as in theory.
Two cases and a control
Worked case one: Ishaan
Aditi runs the full protocol on Ishaan and 15 other volunteers. She records EEG and ECG, uses surrogate R-peaks, compares ECG amplitude between conditions, and records heart rate and breathing. These results are invented but follow the pattern of the 2016 study.
She finds a heartbeat-locked difference between synchronous and asynchronous conditions about 270 milliseconds after the R-peak. It survives the surrogate test, and the ECG does not differ between conditions. Within participants, the HEP amplitude correlates with self-identification ratings. Source analysis places the effect mainly in the posterior cingulate.
What can Aditi conclude? She can say that a heartbeat-locked brain response covaries with self-identification during the illusion, and that it is not an artefact or a non-locked offset. She cannot say that the response is in the anterior insula, that it causes the illusion, or that it anchors the first-person perspective. If she writes carefully, she will say that cardiac signals are integrated into bodily self-identification, which is what the evidence supports.
| Check | Result | What it rules out |
|---|---|---|
| HEP differs, sync vs async | ✓ Yes, ~270 ms | Nothing yet |
| ECG same in both | ✓ Yes | Cardiac field artefact |
| Real beats beat shifted beats | ✓ Yes | Shift not locked to heart |
| Effect in anterior insula | ✗ Posterior cingulate | Insular anchor |
| Causal manipulation | ✗ None | Passenger account |
Worked case two: Ishaan with the timing manipulation
Change one thing. Aditi now delivers the synchronous strokes either at systole or at diastole, using the ECG to time each touch, following the logic of the 2014 fear study. If the heartbeat signal helps anchor the self, the illusion should differ between the two timings. Because the stroking is synchronous in both, touch and vision agree equally in each condition.
Suppose self-identification is stronger when strokes land at systole. That would be the first causal evidence that cardiac timing shapes the full-body illusion. It still would not show that the anterior insula carries the effect, nor that the first-person perspective depends on it. Suppose instead there is no difference. That would suggest the cardiac effect on bodily self is small compared with touch and vision. Either result would be more informative than another correlational study.
The negative control
A negative control is a condition where the answer is known. In the illusion studies, it is the asynchronous condition. In the HEP analysis, it is the surrogate heartbeat, which keeps everything except locking to the heart. In the instrument, it is the offset scenario, which the full pipeline correctly rejects.
A good study of the series question would add one more control: a non-bodily object, such as a box, filmed and stroked in the same way. If the heartbeat response differs between synchronous and asynchronous stroking of a box, the effect concerns visuo-tactile synchrony in general, not the self.
A counterexample
The face null is the clearest counterexample to a general cardiac anchor. Porciello's three studies found no effect of visuo-cardiac synchrony on self-face attribution. If heartbeat signals anchored the self broadly, they should affect the most self-specific image there is, one's own face. They did not. The effect seems confined to body representations that the brain locates in space.
The Damasio patient is a second counterexample, at the level of the insula rather than the heartbeat. A brain without insulae still produced feelings and a point of view. Whatever the anterior insula does for interoceptive awareness, it is not necessary for having a perspective.
What the cases teach together
Read side by side, the two worked cases and the control make one point. The first case shows that a real, heart-locked effect can exist without telling us anything about causes. The second shows what a causal design would look like and how small its effect might be. The negative control shows that a naive analysis would have reported an effect where none existed. The counterexample shows that the strong version of the claim is already in trouble.
None of these cases needs exotic equipment. Each could be run, or has been run, in a standard laboratory with a headset, a camera, an ECG and a scalp recording. That is part of what makes the gap so striking. The missing evidence is not out of reach; it has simply not been gathered, or not been published.
A reader might ask why the timing study has not been done, given how cheap it would be. One plausible reason is that a null result would be hard to publish and hard to interpret, because a small effect could be missed by a modest sample. My toy pipeline suggests that a weak effect would be detected less than a fifth of the time with 32 people. A useful study would need to be preregistered and larger than most in this field.
The failure boundary
The argument has limits. It concerns the first-person perspective and bodily self-identification in healthy adults and a few patients. It does not address whether interoceptive signals matter more in development, when the self is being formed. It does not address psychiatric conditions such as depersonalisation, in which people report feeling detached from their body and self, and in which interoceptive processing may differ.
It also depends on what one means by anchor. If anchoring means being one of the necessary inputs to a continuously updated model of the bodily self, then the evidence is compatible with heartbeat signals as a minor anchor among many. If it means being the foundation on which the perspective rests, the evidence is against it.
| Condition | Does the argument hold? |
|---|---|
| Healthy adults, hand and body illusions | ✓ Holds |
| Self-face recognition | ✓ Holds: no cardiac effect |
| Development, infancy | ~ Unknown |
| Depersonalisation | ~ Unknown |
| Interoception as a whole, not heartbeat | ✗ Not tested here |
At the office party
The joke has a point for the series question. The classic illusions work without any interoceptive manipulation. The cardiac versions show that adding heartbeat timing can strengthen them. That is evidence of integration, not of foundation.
The strongest case against
The strongest case against my conclusion has four parts.
First, absence of causal evidence is not evidence of absence. The intracranial recordings in 2018 found insula effects in both patients who experienced the illusion strongly. That is a small sample, but a consistent one. The causal test has not failed; it has not been done. A defender of the beast-machine view can say that the theory predicts exactly what has been found so far, and that the missing experiment is a gap in the literature, not a problem for the theory.
Second, the beast-machine theory does not claim that heartbeats alone anchor the self. It claims that the brain's predictive regulation of the whole body does. The heartbeat is the most measurable interoceptive signal, which is why the studies use it. Showing that the heartbeat is not necessary is compatible with interoception as a whole being necessary. The Damasio patient's preserved feelings fit Damasio's own view that the substrate is subcortical and bodily, which is close to Seth's.
Third, the first-person perspective may not be separable from self-identification in the way I have assumed. If the sense of experiencing from here is built from the same multisensory and interoceptive integration that produces self-identification, then evidence about one is evidence about the other. My RTM layering separates them; the theory may not.
Fourth, the pulseless patients still have rich interoceptive signals and a native heart that usually beats. The case shows only that the arterial pulse is not necessary, not that cardiac signals are irrelevant.
What the critics of the illusions add
A different line of criticism targets the illusions themselves. Peter Lush and colleagues, in work cited earlier, argued that rubber-hand ratings track how readily people respond to imaginative suggestion. If that is right, some part of every illusion result reflects what participants think is expected. The cardiac versions would inherit the problem, because participants can often guess which condition is supposed to produce the stronger feeling.
The reply from illusion researchers, also cited earlier, did not find the same link in a large sample. The dispute is not settled. For this paper the point is narrower. A HEP difference is not a questionnaire answer, so it is harder to shape to expectation. But attention is also shaped by expectation, and the HEP is known to change with attention. The objective measure does not escape the problem; it moves it one step back.
What the defence gets right
It is worth stating plainly what I accept from the case against. The beast-machine theory is a theory of interoception broadly, not of heartbeats. The causal test has not been run, so the anchor claim has not failed it. And the line between self-identification and the first-person perspective may be less sharp than my layering assumes.
What I do not accept is that these points rescue the specific claim. If the anchor is interoception as a whole, the evidence about the insular heartbeat response is no longer the evidence that matters. If the causal test has not been run, the claim is untested, not supported. And if self-identification and perspective cannot be separated, then the face null and the lesion case count against both at once.
Where the objection still bites
The second objection is the strongest. It shifts the claim from heartbeats in the insula to interoception in general, which is harder to test but also harder to refute. I accept that move for the theory as a whole. But the series question is specific: cardiac-evoked potentials in the anterior insula, during illusions, as a causal anchor of the first-person perspective. For that specific claim, the evidence is correlational, localised mainly elsewhere, and contradicted by the face null and the patient without insulae.
What would change my mind
Three kinds of result would move me towards accepting the claim. First, a study that timed synchronous stroking to systole or diastole and found a reliable difference in self-location, measured by a method participants cannot easily shape, such as walking back to where they feel they were.
Second, an intracranial or stimulation study showing that disrupting heartbeat processing in the anterior insula specifically, and not elsewhere, abolishes or shifts the full-body illusion. Disruption elsewhere, such as in the posterior cingulate, would support a broader network view rather than an insular anchor.
Third, a replicated scalp study with a non-bodily control object, showing that the heartbeat-locked effect occurs for one's own body and not for an object stroked in the same way.
A fourth result would also move me: evidence that people with severely reduced interoceptive signals, such as those with certain autonomic failures, show weaker or absent full-body illusions while their visuo-tactile integration is intact. That would suggest that interoception is a necessary input to bodily self-identification, even if it is not tied to one region.
Each of these results is specific enough that a reader could check whether it has happened. That is deliberate. A view that cannot say what would refute it is not much of a view, and the heartbeat hypothesis deserves a fair test, not a vague verdict.
Two kinds of result would move me further away. A well-powered timing study that found no effect of cardiac phase on the illusion. Or evidence that attention, manipulated directly, produces the same HEP changes as the illusion without producing the illusion.
What I still don't know
- Systole versus diastole stroking
- Box control object
- Stimulating insular sites during the illusion
- Attention versus illusion as the HEP driver
- Whether the point of view has one anchor at all
I do not know whether any group has run the systole-diastole version of the full-body illusion and not published it. I do not know whether the posterior cingulate and insular effects reflect one process or two. I do not know how strongly self-identification, self-location and the first-person perspective come apart in practice. Nor do I know how the beast-machine theory would treat a case in which the first-person perspective persists while interoceptive signals are severely reduced. The obvious reply appeals to signals that were not measured, which makes the theory harder to test.
[AUTHOR NOTE: if you have a view on how the RTM reading of L1 as spanda should relate to bodily rhythms such as the heartbeat, a sentence here on whether you see the heartbeat as an expression of L1 or merely an instance of it would help readers.]
Claim Ledger
| ID | Claim | Status | Evidence |
|---|---|---|---|
| MF44-C1 | Synchronous visuo-tactile stimulation shifts self-identification towards a seen body. | established | Botvinick and Cohen 1998; Lenggenhager 2007 |
| MF44-C2 | Visual feedback in time with the heartbeat strengthens ownership of a virtual hand and body. | likely | Suzuki 2013; Aspell 2013 |
| MF44-C3 | Visuo-cardiac synchrony does not strengthen self-face attribution. | likely | Porciello, three studies |
| MF44-C4 | HEP amplitude covaries with self-identification during the full-body illusion, after artefact controls. | likely | Park 2016 (n = 16) |
| MF44-C5 | The scalp HEP effect during the illusion localises mainly to the posterior cingulate, not the insula. | likely | Park 2016 |
| MF44-C6 | Insular HEPs track self-identification during the illusion. | speculative | Park 2018 (two patients) |
| MF44-C7 | No study has shown that changing insular heartbeat processing changes the first-person perspective. | likely | Literature search; Park 2018 authors' call |
| MF44-C8 | The insula is not necessary for feelings. | likely | Damasio, Damasio and Tranel 2013 (one patient) |
| MF44-C9 | A naive HEP test cannot separate heartbeat-locked effects from non-locked shifts. | likely | Toy instrument; Park 2016 surrogate method |
| MF44-C10 | Heartbeat signals modulate bodily self-identification rather than anchor the first-person perspective. | speculative | Synthesis of C4 to C8 |
Dated predictions
- MF44-P1 (by 2030): a peer-reviewed study will deliver full-body or rubber-hand illusion stimulation timed to cardiac systole versus diastole, and will report an effect on self-identification or self-location smaller than the effect of visuo-tactile synchrony itself.
- MF44-P2 (by 2032): no peer-reviewed study will report that disrupting the anterior insula alone abolishes the full-body illusion in healthy participants.
Established (solid)Likely (striped)Speculative (hatched)
Glossary
- Interoception: the sense of the body's internal state, including heartbeat, breathing and gut.
- Exteroception: the senses directed at the outside world, such as sight and touch.
- Interoceptive inference: Seth's term for the brain's predictive modelling of interoceptive signals.
- Heartbeat-evoked potential (HEP): the brain's averaged electrical response locked to each heartbeat.
- R-peak: the sharp spike in the ECG that marks each heartbeat.
- Cardiac field artefact: the heart's electrical field reaching scalp electrodes and contaminating EEG.
- Surrogate R-peaks: randomly shifted heartbeat times used to test whether an effect is truly heartbeat-locked.
- Systole and diastole: the phase of the cardiac cycle when blood is ejected, and the phase between beats.
- Anterior insula: the front part of a cortical region folded deep within each hemisphere.
- Posterior cingulate cortex: a midline region in the default network.
- Full-body illusion: an induced feeling that a seen virtual body is one's own.
- Proprioceptive drift: the shift in perceived hand position towards a rubber hand.
- spanda (स्पन्द): vibration or pulsation; RTM layer L1, the system's dynamic activity.
- prakāśa (प्रकाश): light or manifestation; RTM layer L2, the fact of appearing.
- vimarśa (विमर्श): reflexive awareness; RTM layer L3, grasping what appears as one's own.
- pratyabhijñā (प्रत्यभिज्ञा): recognition; RTM layer L4, the self known as continuous.
- cit (चित्): consciousness as ground; RTM layer L5.
A rhythm, not a root
Ishaan took off the headset and asked Aditi whether his heart had been part of the illusion. The evidence in this paper lets her give a careful answer. His brain's response to each heartbeat probably changed as the virtual body started to feel like his. That change is real, not artefact, and it is locked to his heart. But it was mainly in a region behind the midline, not deep in the insula, and nobody has yet shown that it caused the illusion rather than accompanying it.
The beast-machine theory remains one of the most interesting ideas in consciousness science. It moves the foundation of selfhood from the cortex's model of the world to the organism's work of staying alive, and it has generated real experiments. The cardiac illusions and HEP studies are evidence that interoceptive timing is integrated into the sense of which body is one's own.
What they are not is evidence of an anchor. The strong claim fails on two cases: a man without insulae who still had feelings, and people on pulseless pumps who still have a point of view. The specific claim fails for want of a causal test. My toy pipeline adds a methodological warning: without surrogate controls, a heartbeat analysis cannot tell a heartbeat effect from a shift that has nothing to do with the heart.
On the RTM reading, the heartbeat is an L1 rhythm that feeds L3 judgements about the body. The first-person perspective, closer to L2, is not reached by any of the studies reviewed. That is not a refutation of the beast-machine theory. It is a precise statement of what remains to be shown.
There is a wider lesson about method. Heartbeat studies are among the most carefully controlled in this field, and the 2016 study passed every check my toy pipeline could devise. Even so, it could not answer the causal question, because no correlational design can. Passing the controls makes a result real; it does not make it a cause. That distinction applies to every neural marker discussed in this series.
For Ishaan, the lesson is more personal. The feeling that the virtual body was his was built from touch, vision and, perhaps, the quiet timing of his own heart. Remove any one of them and the feeling might weaken. But the point of view from which he felt it was there before the headset went on and was still there when it came off.
The next experiment is clear and cheap: time the strokes to the heart and see whether the self moves. Until someone runs it, the heartbeat is best described as a rhythm the self listens to, not the root it grows from.
Sources
Theory
- Seth AK (2013). Interoceptive inference, emotion, and the embodied self. Trends in Cognitive Sciences. https://www.fil.ion.ucl.ac.uk/~karl/Interoceptive%20inference%20emotion%20and%20the%20embodied%20self..pdf
- Seth AK, Tsakiris M (2018). Being a beast machine: the somatic basis of selfhood. Trends in Cognitive Sciences. https://pure.royalholloway.ac.uk/en/publications/being-a-beast-machine-the-somatic-basis-of-selfhood/
- Seth AK (2021). Being You: A New Science of Consciousness. https://www.anilseth.com/being-you/
- Seth AK (2025). Conscious artificial intelligence and biological naturalism. Behavioral and Brain Sciences. https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/conscious-artificial-intelligence-and-biological-naturalism/C9912A5BE9D806012E3C8B3AF612E39A
- Craig AD (2009). How do you feel, now? The anterior insula and human awareness. Nature Reviews Neuroscience. https://pubmed.ncbi.nlm.nih.gov/19096369/
Bodily illusions
- Botvinick M, Cohen J (1998). Rubber hands 'feel' touch that eyes see. Nature. https://www.nature.com/articles/35784
- Blanke O, Ortigue S, Landis T, Seeck M (2002). Stimulating illusory own-body perceptions. Nature. https://www.nature.com/articles/419269a
- Lenggenhager B, Tadi T, Metzinger T, Blanke O (2007). Video ergo sum: manipulating bodily self-consciousness. Science. https://pubmed.ncbi.nlm.nih.gov/17717189/
- Suzuki K, Garfinkel SN, Critchley HD, Seth AK (2013). Multisensory integration across exteroceptive and interoceptive domains modulates self-experience in the rubber-hand illusion. Neuropsychologia. https://scispace.com/papers/multisensory-integration-across-exteroceptive-and-57chvyrhtm
- Aspell JE, Heydrich L, Marillier G, Lavanchy T, Herbelin B, Blanke O (2013). Turning body and self inside out: visualized heartbeats alter bodily self-consciousness and tactile perception. Psychological Science. https://pubmed.ncbi.nlm.nih.gov/24104506/
- Porciello G, et al. (2018). Visuo-cardiac synchrony does not boost self-face attribution. https://d-nb.info/1255608633/34
- Tsakiris M, Tajadura-Jiménez A, Costantini M (2011). Just a heartbeat away from one's body. Proceedings of the Royal Society B. https://pubmed.ncbi.nlm.nih.gov/21208964/
- Influence of interoception and body movement on the rubber hand illusion (2024). Frontiers in Psychology. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2024.1458726/full
- Zamariola G, Maurage P, Luminet O, Corneille O (2018). Interoceptive accuracy scores from the heartbeat counting task are problematic. Biological Psychology. https://sites.uclouvain.be/FOOD4GUT/wp-content/uploads/2018/10/Zamariola-et-al.-Biological-Psychology.-2018..pdf
- Lush P, Botan V, Scott RB, Seth AK, Ward J, Dienes Z (2020). Trait phenomenological control predicts experience of mirror synaesthesia and the rubber hand illusion. Nature Communications. https://pubmed.ncbi.nlm.nih.gov/32978377/
- No specific relationship between hypnotic suggestibility and the rubber hand illusion (2022). Nature Communications. https://www.nature.com/articles/s41467-022-28177-z
Heartbeat-evoked responses
- Park HD, Correia S, Ducorps A, Tallon-Baudry C (2014). Spontaneous fluctuations in neural responses to heartbeats predict visual detection. Nature Neuroscience. https://pubmed.ncbi.nlm.nih.gov/24609466/
- Babo-Rebelo M, Richter CG, Tallon-Baudry C (2016). Neural responses to heartbeats in the default network encode the self in spontaneous thoughts. Journal of Neuroscience. https://pmc.ncbi.nlm.nih.gov/articles/PMC4961773/
- Park HD, Bernasconi F, Bello-Ruiz J, Pfeiffer C, Salomon R, Blanke O (2016). Transient modulations of neural responses to heartbeats covary with bodily self-consciousness. Journal of Neuroscience. https://www.jneurosci.org/content/jneuro/36/32/8453.full.pdf
- Park HD, et al. (2018). Neural sources and underlying mechanisms of neural responses to heartbeats, and their role in bodily self-consciousness: an intracranial EEG study. Cerebral Cortex. https://academic.oup.com/cercor/article/28/7/2351/3862200
- Coll MP, Hobson H, Bird G, Murphy J (2021). Systematic review and meta-analysis of the relationship between the heartbeat-evoked potential and interoception. Neuroscience and Biobehavioral Reviews. https://research.birmingham.ac.uk/en/publications/systematic-review-and-meta-analysis-of-the-relationship-between-t/
- Garfinkel SN, Minati L, Gray MA, Seth AK, Dolan RJ, Critchley HD (2014). Fear from the heart: sensitivity to fear stimuli depends on individual heartbeats. Journal of Neuroscience. https://www.jneurosci.org/content/jneuro/34/19/6573.full.pdf
Lesions and devices
- Damasio A, Damasio H, Tranel D (2013). Persistence of feelings and sentience after bilateral damage of the insula. Cerebral Cortex. https://pubmed.ncbi.nlm.nih.gov/22473895/
- Evidence toward the potential absence of relationship between temporal and spatial heartbeats perception (2021). Scientific Reports. https://www.nature.com/articles/s41598-021-90334-z
- Medical management of patients with continuous-flow left ventricular assist devices. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC4158403/
- Lived experiences of patients using left ventricular assist devices: a systematic review (2026). https://link.springer.com/article/10.1186/s12982-026-01907-0
Cite this work
Mahapatra, R. R. The heartbeat in the mirror: does the brain's response to each heartbeat anchor the first-person point of view? Mind Frontier, paper 44. https://rajeshrmahapatra.com/papers/mind-frontier/U0876-seth-beast-machine-heartbeat-self. DOI: to be assigned.
@misc{mahapatra_mf44,
author = {Mahapatra, Rajesh Ranjan},
title = {The heartbeat in the mirror: does the brain's response to each heartbeat anchor the first-person point of view?},
howpublished = {Mind Frontier, paper 44},
url = {https://rajeshrmahapatra.com/papers/mind-frontier/U0876-seth-beast-machine-heartbeat-self},
note = {DOI: to be assigned}
}