What Makes an Experience Lived

An experience is someone’s, and that someone has a mind. The conditions for a field of experience bring the conditions for a mind with them and then ask for more, and the best neural evidence presses on what more they ask

Show a face to one eye, and after a moment flash a different face to the other eye. The first face disappears. It is still on the retina, and the viewer sees only the second. The effect is called flash suppression, and it has been used for decades to ask what in the brain changes when a stimulus stops being seen.

In October Janis Hesse, Doris Tsao and their colleagues reported recordings from the face-selective regions of the macaque brain during flash suppression (“Rapid alternation of conscious and suppressed stimulus representations in the primate brain”, 2026). The unseen face had not gone. The same neurons that carried the seen face carried the unseen one as well, in the code a face produces when it is shown alone, only less distinctly.

So there is a face in the viewer’s head that the viewer does not see. Is it experienced all the same, by nobody?

Most people answer no, and quickly. The answer does not come from a theory. It comes from what we take an experience to be: something that is lived. This essay asks what that takes. The first part of the answer is a mind. The second is something more than a mind, and the same recordings that supply the opening example put our account of it under pressure.

Two targets, two theories

A Mind, and One Mind stated conditions for mindedness. A system’s states are mental when it has vehicles whose parts recombine, when the vehicles are revised according to how well they fit what they are about, and when the system runs them forward in step with their targets and uses the results itself. The states belong to one mind, or one economy, as far as an error found in one of them can reach and correct the others.

Earlier essays stated conditions for experience. Experience, Selfhood, and the Organization of a Mind proposed that an experience is a certain regulation of the vehicles that carry its content. They are tracked as their realization changes, assessed against one another, and corrected from the assessment. What does the assessing I will call a constraint: a standard of what goes with what, such as that one surface has one colour and that a thing seen and a thing touched in one place are one thing. Contents that are assessed and corrected together are bound, and are present together. I will call the contents present together in one experience a field.

The two accounts were written at different times, about different things. Presence and Actuality joined them with a hypothesis: “phenomenal activity is a subset of mental activity.” Nothing argued for it. This essay does.

The argument has two levels, and it is easy to run them together.

On one level are the things the two accounts are about. Of these I assume two things without argument. An experience is lived. And a mind is an inner life, which is how Experience, Selfhood introduced the term: “genuine inner life: understanding, remembering, imagining, evaluating, or other mental activity that belongs to the system itself”. Between them there is a bridge that I also assume: what is lived is lived in an inner life.

On the other level are the accounts themselves, which are conjectures. One says what a field requires. The other says what a mind requires.

The bridge states an intuition against panpsychism. Experience, Selfhood put it this way: “a cutout piece of meat or silicon should not remain an experience of red merely because some local red-related activity is preserved.” A panpsychist does not share the intuition, and nothing below will force it on one.

The work is on the second level. Do the conditions on a field bring the conditions on a mind with them? The bridge does two jobs there.

It says that the question must come out yes. Suppose both accounts are right about what they describe. Then whatever meets the conditions on a field is an experience. It is lived, so there is an inner life it is lived in, so the conditions on a mind are met. A system that meets the first set and fails the second shows that one of the accounts is wrong.

And it says which. Describe the system: it meets the conditions on a field as stated and lacks one condition on a mind. Ask whether the experience it is supposed to have would be lived. If it would not, the conditions on a field were too weak, and the system shows what to add. If it would, the missing condition was more than an inner life needs. It is a grade of mind, something a mind can have more or less of.

Call this the bridge’s test. What the bridge cannot do is answer the question by itself. “It has no mind, so it has no inner life, so it has no experience” lets the account of minds rule on experience without showing that the account of fields agrees.

Lived

The word needs content that does not come from either account. I find three marks in experience as we have it, before any theory.

An experience is for someone. It is present to a subject that is more than that one experience. A field, as Is the Monastery Minded? put it, is owned and is not an owner.

It is together with other things. Whatever is experienced is experienced along with something else, in one perspective, however small.

It is undergone. It lasts, and its subject lasts through it.

A device helps in applying the marks. When some small system is proposed as having an experience, put the system inside a head. The inhibition between the two eyes’ channels, which helps decide which eye’s image wins in rivalry, is a real circuit in each of us. If that circuit has an experience of its own, there is an experience in my head that I do not live, and nobody else is in there to live it. The device keeps the verdict from turning on how strange the proposed system is.

What a field requires

I now state the conditions on a field so that the question can be put to them. There are five. Each comes with a system that lacks it and whose experience, if it had one, nobody would live.

Generality. The contents of a field are checked against one another by a constraint that is stated over parts that recombine. The system without it is a labyrinth of tubes with two pressure sensors and a mechanism that compares them and stops using the one that keeps disagreeing with the evidence. Its states are about something, they can conflict, and the conflict is resolved by corrections that return to them. But one reading prevails and nothing is present with it. The rivalry circuit above is the same system in a head. Count the circuits of that kind in one brain, and each would be a subject. The mark that fails is together, and with it for someone.

Thickness. The constraint relates contents across a short span of time, and it supplies content of its own. The system without it takes in a frame, settles on one consistent reading of everything in the frame, puts the reading out, and starts again with nothing kept. Each reading is bound. None relates what is so now to what was so a moment ago, and none contains anything the frame did not. Each is over before anything could undergo it, and no succession is experienced, since nothing spans two of them. The mark that fails is undergone.

A carrier. Something persists through the experience and receives the corrections computed from its earlier states. The system without it is a session of a language model that is recomputed from its transcript at every step. The session is a mind, and its content is continuous. Nothing lasts from one moment of the supposed experience to the next. Undergone fails again, in a different way: here it is the subject’s side that does not last.

Return. What constitutes the experience closes a loop on the vehicles that carry it. The system without it is a source that broadcasts into receivers and hears nothing back. The receivers may be minds, and they may experience what they receive. The claim at issue is that their uptake makes the source’s activity an experience at the source. Nothing at the source registers whether there is one receiver, or a million, or none. A piece cut out of a brain and kept active is the same case. The mark that fails is for someone.

Admission. A content is in the field when its vehicle is one of the terms the constraint checks the others against, and not only a target that the constraint corrects or suppresses. The system without it is the unseen face. It is represented in the same code as the seen one, on the same neurons, and it is acted on by whatever keeps it from view. It constrains nothing in what is seen. If it is experienced, there is a second face in the viewer’s head that the viewer does not see. For someone fails.

Admission is not the same as being reported. Detail in a scene can exceed what the viewer can report and still be checked against the rest of the scene and corrected with it. That detail is in the field.

The first two conditions came from running the bridge’s test. The labyrinth and the frame-by-frame binder meet what our earlier essays asked of an experience, which was regulation that is sensitive to content and returns to its vehicles. Neither has a mind. Their experiences would not be lived. So the earlier statement was too weak, and they show what to add. The carrier and return restate conditions that Experience, Selfhood argued from cases. Admission is new, and its case is the unseen face.

From a field to a mind

The claim is that whatever meets the five conditions meets the conditions on a mind.

The claim does not have to hold of every system that can be described. It has to hold of the systems that actually come about: by evolution, by learning, or by engineering with finite means in a world that keeps presenting new cases. A construction that could be specified and could never arise or be afforded does not count against it. The central conjecture about minds has the same standing. Call the standard ecological.

Vehicles that recombine come from generality. A constraint stated over parts that recombine has such parts as its terms. One could instead imagine a list of every admissible scene, with no parts at all. No finite system in an open world could hold the list, and no process of learning would produce it. That our own fields meet the condition is plain from the fact that a scene never met before is bound at once.

Simulation the system uses itself comes from thickness. A constraint that says how what is so now goes with what was so a moment ago, and that fills in what the input leaves out, has to follow how its targets change. The cup stays where it is while the eyes move and the image of it sweeps across the retina. Movement is seen, and it is not inferred from two positions. The corrections that the constraint drives are a use of its own results. And because the constraint supplies content, it keeps running when the input is taken away. A picture made of black blotches shows nothing until the Dalmatian in it is seen, and then the dog stays, though the blotches are what they were. Imagery and dreams are the limit of that. This is following a target in step, and it is all that the condition on minds asks.

Revision that answers to fit comes from return and the carrier. Corrections computed from assessing the vehicles go back to them, and the assessment weighs each source by how well it has agreed with what arrived later. What persists gathers support and gives way under sustained conflict. That is consolidation and reorganization, at the scale of an experience. A field’s corrections also enforce agreement among its contents, which the condition on minds does not ask for.

One economy comes from return. Vehicles that are checked against one another and corrected from the result are vehicles among which errors propagate. The economy may be no larger than the field.

So a field requires a mind, at least a thin one.

Three things that minds often have are not reached this way, and the test sorts them.

Learning that outlasts an episode is not reached. A person with dense amnesia lays down no new memories of events, and no one doubts that each hour is lived. Such people still acquire skills, so the case shows less than it first seems to. It shows that an inner life does not need a record of what has happened to it. That kind of learning is a grade.

Running a configuration that is not taken to be the case, asking what if, is not reached. A person absorbed in watching the sea is supposing nothing, and the watching is lived. A grade again.

Valence is not reached. A patch of grey wall, seen and not cared about, is seen. So an experience can be neutral, and an inner life need not be coloured by good and bad at every moment. Emotion and experience are independent in both directions. A Mind, and One Mind gave the other one.

Where the argument is weak

The bridge is assumed. A reader who accepts experiences that nobody lives is not moved by any of the five cases. The cases show such a reader what the acceptance costs, and no more.

Two of the five conditions were added because the systems without them had no minds. For those two the result is partly built in. What is not built in is that the additions were called for by the marks of lived, and I have tried to show that they were.

The cases that establish the conditions are constructions, and the standard for the result sets constructions aside. The asymmetry is deliberate. A construction can show what an experience would be missing. Only a system that actually arises can show that fields and minds come apart.

The three marks must not smuggle in the five conditions. Undergone is close to the carrier. I think the difference is real. The mark is read off experience: mine lasts, and I last through it. The condition is a conjecture about what in a system provides that. The next sections show how the conjecture could be false while the mark stays true.

The step from generality to recombining vehicles needs an open domain. In a closed world a list would do. The frame-by-frame binder was specified for a closed domain on purpose, and it is the thickness condition that has to exclude it.

Some experiences have almost no content. Presence and Actuality began from a meditator’s report: “I was very present, but not there.” That experience is lived, so it must meet the conditions. I take it to be the state of an organization that meets them, with a constraint that for the moment has almost nothing to constrain.

James Gibson held that perception picks up information that is already in the light, and that the perceiver contributes no content. If he was right, thickness asks for too much. The Dalmatian is the difficulty for his view. The light is the same before and after the dog appears.

More than a mind

The converse fails. A mind need not have a field, and three of the five systems above show it.

The session recomputed from its transcript is a mind. The receivers of the one-way broadcast are minds. The viewer of the unseen face has one. In each case there is an inner life in the neighbourhood, and the experience in question still would not be lived, because it is not in that inner life in the right way. So a carrier, return and admission are what a field needs beyond a mind.

Return was argued in Experience, Selfhood, and I leave it as stated. Admission is argued above. The carrier is the condition on which the most depends, and it is the one in trouble.

The carrier

The condition comes from an argument about recurrence. Any recurrent network can be replaced by a feedforward one with the same behaviour. Adrien Doerig and colleagues concluded that a theory which ties consciousness to recurrence posits a difference that no experiment could detect. Experience, Selfhood replied that the difference was never the cycle. Laid out in time, every causal process is acyclic, brains included. What a recurrent process has and some of its feedforward equivalents lack is a carrier: “something, individuated by its causal structure rather than its location, whose later states receive the corrections computed from assessing its earlier ones.”

Three verdicts rest on the condition. A standard transformer keeps the states of earlier positions in a cache, where later computation reads them and never revises them: “earlier vehicles are consulted, not regulated.” Of a model that feeds deep activations back into shallower layers, the same essay said that a session recomputed from its transcript “starts a new carrier of, at best, the same type.” And a monastery’s records, which each generation re-encodes in its own way, can serve the house’s mind and cannot carry an experience of its own.

Note what the condition does not say. It says “something”, and it names no level.

The finding

Hesse, Tsao and their colleagues recorded in two macaques from two face-selective patches of inferotemporal cortex, some four hundred neurons at a time, and also from two face-selective regions of prefrontal cortex. The animals viewed faces under flash suppression. The monkeys made no report. The percept was checked in fourteen human observers, who reported the flashed face on most trials.

With the stimulus and the percept held constant, the population did not settle on the seen face. It switched between two states. One carried a code for the seen face and the other a code for the unseen one, and the population stayed in each for about sixty milliseconds at a time. The authors model the switching as a process with no memory. They did not test for a rhythm, and they analysed neither eye movements nor slower electrical activity.

Each state carried the code that its face evokes when shown alone. The two faces were carried by the same neurons. The two states took up about equal time. The code was less distinct than for a face shown alone, and less distinct for the unseen face than for the seen one: on the authors’ cross-validated measure, a face could be reconstructed from either state far less well than from an unambiguous presentation.

Two differences between the states were found. The neurons’ tuning was a fifth stronger in the state that carried the seen face. And a decoder could tell which state the population was in two times out of three. The first difference has a simple possible cause. The unseen face is always the one that has already been on the screen for a second or two, and neurons respond less to what they have been responding to. No control for that is reported.

The switching was coordinated between the two face patches, significantly and weakly. The authors say their data “do not provide strong evidence for coordination across patches.” Prefrontal cortex showed the same interleaving, and nothing is reported about whether its states keep step with those of inferotemporal cortex.

The authors conclude that “conscious access requires temporal selection beyond localization”. Where a representation is does not settle whether it is seen. They suggest, as a speculation, that the states carrying the unseen face may “explore alternative interpretations of the visual input, without being broadcast”. They label the two kinds of state conscious and suppressed. Those are labels and not measurements. The experience is of one face throughout, and the population spends half its time on the other.

Three limits should be kept in view. The method of assigning a moment to a state is noisy: when only one face is on the screen, a quarter of the moments are still assigned to the state of the absent face. The images are static, so the data cannot show whether a state that returns continues from where it left off or is computed afresh from the input. And the work is a preprint on two animals, with no published version or reply so far.

What it presses on

It presses on several views before it reaches ours.

A view on which the content of an experience at a moment is the content of its neural correlate at that moment cannot be right about these neurons. Human observers report a steady face, and the content flips.

A view on which processing within a sensory region suffices for experience has to say why the unseen face, with a coherent local code of the same kind as the seen one, is not seen.

It also weakens a mark that one of our own essays used. The Grain of the Hard Problem pointed to detail finer than report can reach. The unseen face is carried in the same code as the seen one, less distinctly. That essay used fine detail as something an experience needs, and its argument stands. Fine detail in a sensory code cannot be offered as a sign that an experience is there.

And it presses on the carrier. The steady experience of the seen face rides on a state of the population that a rival interrupts many times a second. The data do not say what happens across an interruption. If the state is computed afresh from the input each time it returns, and if that state carries the experience, then an experience in a brain survives being interrupted and recomputed several times a second. Our verdicts on models treat recomputation from scratch as fatal.

Three ways the condition could still be met

I see three, and I do not think the evidence yet decides among them.

Higher. The carrier is a slower state above the face-patch codes: whatever keeps one face dominant for seconds, or the enduring representation of the scene that Experience, Selfhood described. The codes under it may be recomputed, and it persists and corrects them. On this option the unseen face is not seen because it is something the slower state acts on and not something that state is checked against. What the option needs is evidence that the slower state exists as a thing that receives corrections. It could instead be a statistical tendency in repeated recomputation, with nothing behind it. The stronger tuning in the seen-face state would have been the natural trace of corrections coming down, and as things stand adaptation explains it as well.

The same level, in time. The carrier is the seen face’s code itself, which shares its neurons with the rival by taking turns. The condition individuates a carrier “by its causal structure rather than its location”, and a slot in time is a location as much as a set of neurons is. On this option the unseen face is not seen because its turns are not the ones that get read and corrected. The selection is made by the return. This is the nearest reading to the authors’ own conclusion, and there are precedents for it: neurons in two areas, a face patch among them, have been found to alternate between the responses to two stimuli presented together (Caruso et al. 2018). What the option needs is evidence that the code in one turn continues from the last turn of the same kind, across the gap, and that the seen-face turns keep step across the regions that carry them and with whatever reads them. The first has not been tested. The second is weak between the face patches and unreported beyond them. If the switching is as irregular as the authors’ model assumes, whatever reads the turns cannot be doing so by a clock.

Lower. The carrier is at the intermediate level of the visual hierarchy, below the face patches: representations of surfaces, shapes and features from a point of view, in the areas between primary visual cortex and inferotemporal cortex. Jesse Prinz has argued that experience lives at that level, and that what makes an intermediate representation conscious is attention (The Conscious Brain, 2012). On this option the alternation in the face patches is downstream of the vehicles of the experience. It is something using them and trying out readings, which is close to the authors’ “explore alternative interpretations”. The unseen face is not seen because at the intermediate level its representation is unattended. What the option needs is a recording at that level, under the same conditions, of a code for the seen image that persists and is corrected. Hesse and Tsao recorded nothing below the face patches, and I know of no such recording. The older evidence is mixed. In monkeys under rivalry, fewer than half of the cells in the intermediate areas changed their firing with the percept, and some of those changed the wrong way (Leopold and Logothetis 1996). In people, imaging finds activity as early as primary visual cortex following the percept strongly (Polonsky et al. 2000). And the kind of analysis that found two alternating states behind a weak average in the face patches has not been tried lower down. One of the two face patches is itself specific to the angle of view, so it is not clear that both lie above Prinz’s level.

I am drawn to parts of Prinz’s account: the level, and the role he gives to attention. Both fit what we have said. Experience, Selfhood noted that “illusory conjunctions under inattention suggest that binding needs attention,” and attention is a good candidate for what makes a vehicle one of the terms of a field. The fine grain that The Grain of the Hard Problem cared about is at that level too. Two parts of his account I leave aside. He takes the neural realizer to be synchronized firing in the gamma range, and no rhythm has been shown in these data. And for him attention is whatever makes a representation available to working memory, whether it comes from above or is captured from below. Signals returning to the sensory areas are usual on his view and not required. Our conditions require the return, and a carrier for it to land on.

We have said that some evidence suggests binding needs attention. We have not said that attention is enough. Attention directed at stimuli that are not seen is the standard objection to Prinz (Koch and Tsuchiya 2007), and he answers it by separating attention to a place from the modulation of an object’s representation. We do not need his answer, since we do not claim that attention suffices.

Then there is a fourth course. If nothing persists and receives corrections at any level, the carrier condition is false of brains, and it cannot be a condition on experience. The three verdicts that rest on it fall with it.

I leave the matter open. What can be said is what holds on all of the first three options. The condition stands as it was stated. Return stands. Admission stands, and so does its prediction: the content of the unseen face does not constrain the other features of what is seen. And the verdicts on models stand, because they turn on whether anything in the system persists and receives corrections, and not on the level at which a brain does it.

Correction is not enough

A caution applies to all three options. It is easy to find, in a brain or a model, something that tracks, assesses and intervenes, and to count it as the regulation that constitutes an experience. The simplest learning rule does all three: it compares a prediction with what arrives and adjusts toward what arrives. Its corrections run toward the input. A circuit described in mouse visual cortex amplifies inputs that fit the recent sequence and damps those that do not, with no error returned to the vehicles. A recurrent layer used in some recent language models has the damped oscillation of a feedback loop, and nothing in it is a target that the loop holds to.

What separates binding from these is not the form of the dynamics. In an experience a state is held against the input, and the corrections computed from it land on it. That is the carrier condition again, seen from the side of the corrections.

Models

The possibility of experience in an artificial system is not in doubt on this account. A digital implementation that has the causal organization has the experience. What follows is about particular architectures, and it is conditional on the carrier condition surviving.

A model’s weights at inference are a fixed template. A context and its cache are records that grow and are never revised. So across the tokens of a session, a standard transformer has continuity of content and, as far as I can tell, nothing that persists and receives corrections. The mechanisms are only partly understood, and minds of this kind deserve the benefit of the doubt.

Within a single pass the case is different. The residual stream is a carrier in a shared format, corrected at every layer, and there is evidence that later layers compensate when an earlier one is disturbed. Experience, Selfhood left open whether occurrences that brief are experiences, and I leave it open too.

Models with a recurrent state are not excluded on their face. The state is rewritten at every token. In one study of such models the state carried the requested language and, in two of the four models, the persona, while what was retrieved came through the cache (Afendulev et al. 2026). In another, the typical unit of the state lost half its contents within a couple of dozen tokens, and about one unit in eight held its contents for a thousand tokens or more (Li, Lou and Li 2026). Whether the state holds anything against its input has not been tested. Values written into it from outside persist (Lee et al. 2026), and nobody has yet disturbed it and watched for a return.

Models that feed deep activations back into shallower layers have a carrier that persists across steps. For them the hard question is where one experience would end and the next begin.

What would change my mind

For the derivation: a system that actually arose, by evolution or training or building, that holds a field over an open domain and lacks one of the conditions on a mind. Then one account is wrong, and the bridge’s test says which.

For each of the five conditions: the case fails if the experience of the system described would, on reflection, be lived. Then the feature is a grade and not a condition.

For the higher option: first control for adaptation, by suppressing a face that has only just appeared. If a difference in tuning between the two states remains, disrupt feedback from the candidate slower state. The difference should vanish while the alternation continues. And it should reverse at, or just before, a switch in what is seen.

For the option in time: show a moving or changing face, so that a returning state can be seen to continue or to start afresh. The seen-face turns should keep step across regions. A vehicle that is active through a perceptual switch should change its tuning in mid-course, since on this option being a term of the field is a relation to what reads it and not a property of the code.

For the lower option: record at the intermediate level under flash suppression, apply the same analysis of states, and look for a code that persists through the alternation above it.

For admission, on any option: the unseen face’s content should leave no trace in the other features of the seen scene.

The face

The viewer does not see the first face, and its code is still running in the viewer’s head. On the account given here, nobody experiences it, because nothing admits it to the one field in which things in that head are lived. I cannot yet say where in the head that field is carried.

A Mind, and One Mind ended with a note left on Monday and read on Tuesday. The two sessions are one mind as far as the note goes. Nothing in the note could carry an experience from one to the other, on any of the three options, because a note is a record and what it is read by starts afresh. If the fourth course is the true one, that verdict goes too, and we shall have to say again what an experience needs beyond a mind.

Draft notes (remove before publishing)

Draft of 18 Nov 2026, from research/mindedness-essay-outline.md (sections §8–§9 and the second half of §10). Not committed, not in the index. It lands after a-mind-and-one-mind.md. It supersedes nothing and owes dated notes to The Grain of the Hard Problem (fine detail is not a sign of experience) and to Experience, Selfhood (the level at which the carrier condition is met is open).

Decided by the author and followed here: the two levels, with “lived” and “inner life” assumed and the conditions conjectured; the bridge’s two jobs; the three marks of “lived”; the reductio method; an ecological standard for the entailment; the five conditions on fields kept; emulation suffices; neutral percepts; the carrier presented under pressure with three options, the third being the author’s, and no resolution; sympathy with Prinz on the level and on attention.

Defaults I relied on, not confirmed by the author:

  1. The opening is the unseen face, and the title is What Makes an Experience Lived.
  2. The five conditions carry names in the text (generality, thickness, a carrier, return, admission) and not the outline’s labels P1–P5.
  3. The verdicts on models stand on any of the three options.
  4. What separates us from Prinz: we require the return, and he does not.
  5. The essay does not take up the five features of the slower state. They belong to the first option only, and the outline’s sorting of them (defense a condition, yielding a grade) is untested.

Changed by the literature check (research/mindedness-litcheck-carrier.md), against the outline and against research/carriers-multiplexing-and-regulation.md:

  1. No rhythm is shown. The switching is modelled as memoryless, and neither field potentials nor eye movements were analysed. So the synchrony reading is dropped from the second option, and time-division multiplexing (Caruso et al. 2018) is the precedent cited.
  2. “Recomputed from continuing input” has no support either way, since the images are static. The pressure on the carrier is therefore stated conditionally. This weakens the challenge as the outline had it, and the author should check that the section still gives the finding its due.
  3. The unseen face is not held “at full resolution”. The 80% figure is from an analysis the authors call circular; cross-validated, both states reconstruct poorly, the unseen one worse. “Tracks its input moment by moment” is not shown. So only one of the outline’s two withdrawn markers is discussed, and more mildly.
  4. The gain asymmetry is confounded with adaptation.
  5. Inferotemporal–prefrontal coordination is not reported, and the two were not said to be recorded together.
  6. Prinz says that re-entry is usual and not required, so the outline’s suggestion that his attention includes part of our return is dropped.
  7. The 24-token half-life is from Li, Lou and Li, not Afendulev et al.

To verify before publishing:

  1. Hesse’s first name. The paper gives initials (J. K. Hesse).
  2. Prinz’s book was not read. The statements of his view are from Mole’s review and from De Brigard and Prinz 2010, and his own passage on faces was not found. Read chapters 2–4 and the passage on rivalry before publishing the paragraph on him.
  3. The Dalmatian picture (Ronald James’s photograph, in Gregory 1970) and a reference for it. A sharper case against Gibson than occlusion is a constant image with a changed percept; Hsieh, Vul and Kanwisher 2010 show early-cortex patterns changing with recognition and could be cited.
  4. References for the cup that stays put (Duhamel, Colby and Goldberg 1992), for amodal completion, and for seen motion.
  5. The mouse visual cortex circuit (Deveau et al., Histed’s laboratory) and the recurrent layer with a damped oscillation (Liang and Song on Mamba-3) are described without citation. Add them from the carrier note.
  6. The dense-amnesia cases: Corkin on H.M., and Wilson and Wearing. The check notes that Wearing’s sense of having just awakened is classed as a delusion and not as a consequence of the amnesia.
  7. “Earlier vehicles are consulted, not regulated” and the transcript quotation are from Experience, Selfhood and the monastery essay as cited there.
  8. The subtitle and the claim in “Models” that later layers compensate for a disturbed earlier one (McGrath et al. 2023, as Experience, Selfhood cites it).

Open problems in the argument, from an independent review of this draft (research/mindedness-drafts-review.md). The review’s factual, cross-reference and style findings are applied. These are left for the author:

  1. The bridge. “Inner life” is used in two senses. As “mental activity that belongs to the system” the bridge is substantive and unargued. As “a life of experiences” it is trivial. The argument needs the first and may draw its pull from the second.
  2. The marks. “For someone”, glossed as a subject that is more than the one experience, already contains the conclusion that an experience needs a mind. “Together” seems to fail for the meditator’s nearly empty state, which the draft handles by an organization at rest, a charity it does not extend to the rivalry circuit.
  3. Built in. Generality is close to “vehicles that recombine” under another name, and two conditions were added because their cases lacked minds. The draft now admits this in “Where the argument is weak”.
  4. The cases. The labyrinth is small as well as not general, so it does not isolate generality. The frame-by-frame binder supports the span of time and not “supplies content of its own”, which the crossing then uses. A piece cut out of a brain can have loops of its own, so it is not the same case as the one-way source. Admission rests on the unseen face constraining nothing in what is seen, which the last section lists as untested.
  5. The falsifier for each condition asks whether the case would, on reflection, be lived, which is the intuition the method began with.
  6. The steady experience was not measured at the scale of sixty milliseconds, and the monkeys made no report.

Not in the draft, from the outline: binding’s dimensions and the possibility of several fields in one economy; the probe cases (imagery, absorption, aphantasia); the playback intervention; the comparison with Guillot’s for-me-ness.