How Mind Dust Becomes More Than Mind Dust

Topological defects, the boundary problem, and why/how bounded 4D pockets might behave holistically without compressing experience to a point

[Epistemic Status: Another Hail Mary effort to flesh out the argument for a topological underpinning to the solution to the boundary problem within an ontology of consciousness based on fields; heavily Claude4.7-assisted, a big idea/argument dump]

In each object of the eye

There are infinite eyes immanent

Their various natures different

Measureless, unspeakable

• Book 36, “The Practice of Universal Good” / Samantabhadra’s verses in Thomas Cleary’s Flower Ornament Scripture (live reading)

---

TL;DR

Five claims:

1. A moment of experience is a topologically bounded pocket of a physical continuum field. Not a level-set patch (those aren’t gauge-invariant), not a coupled-oscillator network (those have observer-imposed boundaries and Nyquist-limited capacity for topological structure), not a generic point with information stored in derivatives at it (the diffraction limit kills any biological version of that move). The pocket framing buys hard frame-invariant boundaries, dissolves the slicing problem and Tomasik’s fuzzy-individuals problem, and is the level at which the Aharonov-Bohm effect is already telling us topology has causal teeth.

2. Lehar and Hunt got the binding glue right (resonance) and missed the boundaries (resonance is continuous and not Lorentz-invariant, so it can’t deliver hard cuts). Topology delivers the cuts. The substrate has to be a continuum field for any of this to deliver a hard, observer-independent cut.

3. The pocket has to be 4D, not a 3D snapshot. This gives signals time to traverse it during its existence, lets path-integral-like dynamics generate genuine all-to-all interior visibility, and stitches consecutive moments by overlap rather than by an extra binding mechanism. The pocket is the experiential subject; there is no inner observer separate from the path-integral aggregation.

4. The translation problem (what makes a given pocket the specific experience it is) has at least one concrete handle: the body’s resonator ecology (vascular tree, lung foam, heart, gut) writing geometric boundary conditions onto the cortical field, with frustration between incompatible resonators driving the field to buckle into hyperbolic geometry. This is currently my best handle on why psychedelic phenomenology so often feels hyperbolic, organic, and body-coupled rather than merely noisy or high-dimensional.

5. The deeper version of the boundary problem (why is this field, and not some arbitrary subset of it, the pocket?) probably needs Tier III process-topological monads in the sense of my earlier post on fixed buckets, where individuation is generated by the dynamics rather than stipulated. The continuum-field picture in this post is one rung below that and is what most of the argument actually relies on.

---

I. Where this fits

A while back I wrote “The View From My Topological Pocket”, arguing that topological segmentation in the fields of physics is the right shape of solution to the boundary problem. Two moves did most of the work. First, relativistic frames of reference can’t bootstrap phenomenal binding: anything frame-dependent doesn’t have inherent existence, and our experiences clearly do (notwithstanding Buddhist subtleties we’ll get back to in a future post). Second, field topology is exactly what we need: Lorentz transformations don’t change it, and topological features have real, large-scale causal effects (magnetic reconnection in solar flares being the showpiece).

That post and the Frontiers paper with Chris Percy lay the groundwork. What I want to do here is push on the part with the most slack: how the insides of a pocket get to be a moment of experience. This isn’t meant as a finished physical theory, but as a constraint-shaping argument: if consciousness lives in fields, then the boundary problem pushes us toward gauge-invariant topological individuation rather than synchrony, level sets, or observer-chosen networks. Three things any candidate ontology has to deliver: translation, a way of mapping the structure of a bounded unit to the phenomenology of an experience; causal closure, a way for that structure to derive a real causal power; and holistic visibility, an account of how things inside a single moment are co-witnessed, which is a much stronger requirement than mere proximity.

The thesis: topological defects in a physical continuum field (the brain’s electromagnetic field, the underlying quantum fields, or some other continuous medium) are a candidate for satisfying all three, and the resonance theories from Steven Lehar and Tam Hunt (with Jonathan Schooler) got something deep right but need topology underneath them to be complete.

I want to flag the continuum requirement upfront, because there’s a tempting confusion to clear up. Networks of coupled oscillators on a discrete graph (Tier I/II systems in my earlier post, Kuramoto-style lattices, the orientation pinwheels in V1) can host topological defects in a perfectly real sense. A 2D oscillator lattice with phase variables can have pinwheels with winding numbers ±1, and those winding numbers are robust under continuous deformations of the loop and under most perturbations. So discrete networks aren’t topologically inert. The math works; you can compute a winding number on a discrete loop and it will be an integer. What discrete networks lack is two related things: a packing limit on how dense their defect structure can get (a network with N oscillators can host at most some fraction of N well-resolved defects before adjacent ones merge or smear, set by the lattice spacing in roughly the same way Nyquist limits set how much signal you can fit in a sampled medium), and, more importantly, a principled account of where the network’s own boundary lives.

Topological defects living in a field can be modulated with coupling kernels on systems of oscillators when the field responds to the oscillator’s phase. The discrete system of coupled oscillators couldn’t carry a continuous topology without some loss (due to Nyqist sampling limits) but the continuous field can. See: From Neural Activity to Field Topology: How Coupling Kernels Shape Consciousness

The defects exist given a particular choice of which oscillators count as “the system,” but the network specification itself is observer-imposed, so the boundary problem reasserts itself at the level of substrate. Continuum fields improve on both counts (defect density is limited only by the medium’s physical wavelength rather than by a discretization scale, and the field’s spatial extent is set by physics rather than by an external choice of which nodes are in or out), assuming the relevant field is physically continuous (the EM field is the same field everywhere, not a chosen subset of nodes). So the right way to state the argument: discrete networks can have topology, but it’s bounded by lattice spacing in capacity and observer-relative in scope; continuum fields can have topology bounded only by the medium’s wavelength cutoff and observer-independent in scope; only Tier III systems (where the bucket itself is generated by the dynamics rather than stipulated) close the substrate-boundary problem completely. Most of the post lives at the continuum level. I’ll come back to Tier III in the closing.

Lehar’s harmonic resonance theory argues that the brain’s representations are standing waves in a field-like neural medium, and that this is what makes the Gestalt phenomena of perception (emergence, reification, multistability, invariance) natural rather than puzzling. Hunt and Schooler’s general resonance theory (GRT) generalizes the move beyond the brain: any collection of oscillators that share a resonance form a “resonance complex,” with the slowest shared resonance setting the scope of the resulting macro-experience. Both views identify resonance as the binding glue. What I want to argue is that they are right about the glue and wrong only about thinking that resonance alone delivers boundaries. Topology delivers boundaries; resonance fills them; and the whole thing has to live in a continuum medium for either move to do the work it needs to.

II. What the picture looks like

Take a 2D vector field that, almost everywhere, looks boring: arrows pointing in roughly the same direction, off to infinity. Drop a vortex and an anti-vortex into the middle of it. The field around them rearranges, and one of the things that rearrangement does, automatically, is enclose a region. Follow the field lines from any point inside that region and they curl back on themselves; they don’t leave. The pocket isn’t drawn around anything, isn’t a thresholded patch of high field strength, isn’t something I built deliberately. It’s just what the defect pair forces the surrounding field to do.

A vortex/anti-vortex pair in a 2D field.

The topologically bounded pocket enclosed: a region from which, following the field lines, you cannot escape.

Annihilate the pair and the pocket goes with it. Boundaries here aren’t a separate posit on top of the field, the way “thresholds” or “neighborhoods of high coherence” are in most theories. They are what closed defect structures are, geometrically.

Same caveat as section I: discrete oscillator lattices can host real winding-number defects. The reason I’m not treating them as sufficient isn’t that they lack topology, but that their resolution and their boundary are inherited from the chosen lattice rather than set by physics. Throughout the post, when I say “field,” I mean the continuum kind, the way EM fields, fluid fields, and quantum fields are continuum objects.

This matters because it answers the boundary problem, the much-neglected sibling of the binding problem first sharply formulated by Rosenberg in 1998 and (more diluted) by Chalmers in his 2016 combination problem essay. Any theory that solves binding via something continuous (causal connection, information flow, resonance, EM-field activity) immediately faces the question of why the binding doesn’t keep going. If your mechanism applies in principle to quarks or to choirs, what stops it at human-scale experience? Topological defects give a different shape of answer: closed field-line structures dynamically individuate regions, making their persistence and interactions depend on global topological constraints rather than on observer-imposed cuts, and the relevant topological invariants are not artifacts of a simultaneity slice the way exact phase synchrony is.

The “topology has causal teeth” point is worth lingering on. People sometimes hear “topological feature of an EM field” and assume something delicate and microscopic. Magnetic reconnection in the sun’s corona, where field lines abruptly rewire themselves, releases enough energy to launch coronal mass ejections of trillions of kilograms of plasma at thousands of kilometers per second. The brain isn’t the sun, but the principle that field topology can be causally efficacious at macroscopic scales is well established outside of consciousness research.

III. Lehar and Hunt: what they got right, what’s missing

A bit more on the two resonance views, since I think they deserve more credit than they get and since the critique that follows depends on getting them right.

Lehar’s harmonic resonance theory starts from a piece of common-sense phenomenology that gets buried in most neuroscience writing: the world you experience is a structured three-dimensional object in your head, not a set of feature vectors. From there, Lehar argues that the medium hosting that object has to be capable of supporting explicit, picture-like spatial representations with the kinds of global properties (figure-ground, completion, invariance, multistability) that perception actually exhibits. His proposed substrate is standing waves in the brain’s electromagnetic and electrochemical fields, by analogy to chemical reaction-diffusion systems that produce structured spatial patterns in embryonic morphogenesis. The point that matters most for what follows is that Gestalt properties are emergent properties of resonance dynamics in a continuous medium, not properties added by special mechanisms applied to a graph of feature detectors. Lehar is explicit that the substrate has to be a continuum field, not a network of discrete units. This is one of the things he gets right, and it’s the part of his picture that is needed for what follows.

Hunt and Schooler’s general resonance theory (GRT) takes the resonance intuition and extends it past the skull. They propose that any collection of oscillators sharing a resonance forms a “resonance complex” with some degree of macro-experience, and that the slowest shared resonance determines the scope of the resulting bound experience. Their move on the combination problem is that micro-experiences tied to micro-resonances merge into macro-experiences when the underlying oscillators phase-couple. Hunt’s follow-up paper on calculating the boundaries of consciousness develops synchrony indices intended to make the cutoff between bound and unbound subsystems quantitative.

GRT can be read in two ways and the reading matters. On the field reading, the “oscillators” are localized excitations in an underlying continuum field (EM is the natural candidate), and GRT becomes a story about how spatially distributed parts of one continuum field bind via shared resonance. On the network reading, the oscillators are discrete subsystems related by a graph of couplings, and GRT becomes a story about how nodes in a network bind via shared phase. The network reading is what most computational neuroscience defaults to, and it’s how GRT often gets formalized in practice. The two readings aren’t equivalent, and the network one has a subtle problem.

The subtle problem is that on the network reading, GRT inherits the boundary problem at the substrate level. The network can host real topological defects (resolution-limited, but real), and resonance can synchronize oscillators that lie within a chosen subset, but the choice of which oscillators count as “the network” is the modeler’s, not the dynamics’. You can take any subset of nodes and call them a “resonance complex,” and the resonance dynamics will dutifully synchronize whatever you pointed at. There’s no fact of the matter about which subset is the right one beyond what an external observer chooses to highlight. This is the Tomasik fuzzy-individuals problem reasserting itself one level up. Resonance can bind nodes you’ve already grouped, but it can’t tell you how to group them in a frame-invariant, causally significant way. The hardness of the boundary in network-GRT is being smuggled in by the network specification, not delivered by the resonance dynamics. (For the deeper version of why this matters, the Fixed Buckets Can’t Phenomenally Bind post argues that only Tier III systems, where the bucket itself is generated by the dynamics rather than stipulated, escape this problem completely.)

There is something real in both pictures, especially on the field reading. Standing-wave dynamics in a continuum medium do produce global coherence; shared resonance does bind co-located excitations in a way that scales naturally with system size; the slowest mutual frequency does set a meaningful upper bound on the size of a resonance complex. The Lehar–Hunt picture, read as a continuum-field theory, is much closer to a real theory of consciousness than the discrete-neuron-doctrine picture it competes with.

But even on the field reading, where the substrate problem doesn’t bite, there is still no boundary mechanism that delivers hard edges. Resonance is a fading-with-distance phenomenon. Two regions that are phase-locked are bound, two that are slightly off are slightly less bound, two that are very off are not bound. There is no sharp edge unless you import a phase transition, and even then you need a story about where it takes place and why human experience consistently lives inside one specific transition rather than smeared across many. Hunt’s synchrony indices are an honest attempt at quantifying the cutoff, but the underlying problem doesn’t go away: resonance is a continuous variable, and continuous variables don’t naturally produce hard boundaries without help. Worse, exact phase synchrony isn’t even frame-invariant: an observer in relative motion will disagree about which oscillators are simultaneously in phase, which means resonance alone can’t ground something as solid as a moment of experience.

Topology changes the type of cutoff being proposed: not a threshold over synchrony, but a structural feature of the field configuration. A region enclosed by a closed loop of field lines in a continuum field is not on a continuum with its surroundings in the way phase synchrony is: either the loop closes or it doesn’t, and the count of defects in a 2D continuum field is an integer. The relevant topological invariants are not artifacts of a simultaneity slice in the way exact phase synchrony is. Lehar gives us the right substrate (continuum, resonant), Hunt and Schooler give us the binding glue (shared resonance), and topology gives us the scissors and the absoluteness of the cut. A discrete oscillator network can approximate this story up to its Nyquist limit and modulo the substrate-boundary issue, but the strong version, where boundaries are observer-independent and capacity is set by physics rather than by lattice spacing, requires the continuum.

IV. The translation problem

Granting pockets, we still owe an account of how a topological pocket gets to be a specific experience. There are at least two principled handles, and a hint at a third.

The first is dimensionality. In 2D, defects can be fully classified (vortices, anti-vortices, winding numbers). In 3D it gets dramatically harder, because the loci of “stuff being pushed away” and “stuff being attracted” can be knotted lines, and chaotic attractors with nontrivial topology start showing up. I suspect this combinatorial explosion is plausibly part of why our experiences are so phenomenologically rich rather than being a problem to solve away. The space of possible 3D defect topologies is enormous, and might be just barely enough to host the variety of experiential structures we actually find when we look inward. Attention itself plausibly lends itself to vector field topology: sources, sinks, saddles, currents, eddies. We can talk about “focus” and “wandering” but those are extremely impoverished compared to what’s actually going on.

The second handle, less developed but more concrete, comes from thinking about what’s setting the boundary conditions of the pocket in the first place. The brain is not a self-contained resonator. It’s coupled to a body full of structures with their own characteristic resonant geometries: the vascular tree is branching and fractal-like, the lung’s alveolar architecture is foam-like and highly curved, the heart has toroidal and helical architectural motifs, the gut is fractally textured and roughly cylindrical. Each writes boundary conditions onto the cortical field at its coupling interface. The eigenmodes of the cortical pocket are then constrained by something like a Steklov-type problem: find standing wave patterns that simultaneously satisfy the wave equation in the cortical bulk and match the boundary conditions imposed by every active resonator.

The reason this matters for translation is that it gives us a candidate mechanism for which phenomenology a given pocket-configuration produces. When multiple resonators with mutually incompatible intrinsic geometries try to write boundary conditions onto the same field, the field is geometrically frustrated, in the same sense that you can’t tile a sphere with hexagons without introducing pentagonal defects. The field’s resolution to the frustration is to buckle into hyperbolic geometry, which can accommodate high-genus boundaries that flat or spherical geometries cannot. The flavor of hyperbolic geometry depends on which resonators dominate: foam-like kale if the lung is loud, helical and toroidal structures if the heart is loud, branching dendritic geometry if the vascular tree is loud. This complements the energy-landscape account: the Hamiltonian framework tells us that the field tends toward hyperbolic attractors when energized; the resonator-frustration account tells us which hyperbolic geometry, and why kale-like phenomenology under high-dose psychedelics has the specific organic textures it does rather than some abstract mathematical shape. This deserves its own post.

There’s a hint of a third handle that I want to flag without developing here. Most of the time, a pocket is using its interior to encode information about something else: a previous pocket, a sensory input, a memory trace. Its content is referential. But there might be a degenerate case where attention within a pocket is so homogeneously distributed that nothing in the interior breaks the symmetry, and the pocket’s “content” is, in a precise sense, just itself. I have a strong hunch that this is what cessation states in deep meditation are: the class of configurations where the referent of the experience is exactly, directly, the experience itself, with no remainder (cf. The Topology of Liberation: Jhanas, Cessations, and the Geometry of Inner Harmony).

V. Why defects-as-experience-loci almost works, and why it doesn’t quite

So there are two topology stories worth keeping separate, and only one of them survives. The first says experience lives at a defect. I think that fails, for reasons I’ll spell out here. The second says defects bound the region where experience lives. That’s the version I want to defend, and it’s the subject of the next section.

Topological pockets buy something important. They give us objective frame-invariant boundaries, they explain why binding doesn’t keep going past the meso-scale, they survive the slicing problem and the fuzzy-individuals problem, and the topology of the field has real causal teeth in physics (vortex cores, skyrmions, edge states, the Aharonov-Bohm phase). It’s tempting to stop here and declare victory.

But there’s a problem we haven’t honestly faced yet. Even granting the pocket exists and has hard boundaries, the field inside the pocket is still classical (or at least field-theoretic), which means it’s still made of locally evolving points. Every point inside the pocket is governed by a PDE that depends only on its immediate neighbors. Holistic visibility doesn’t fall out of the topology of the boundary alone; the inside of the pocket has to do something with all that locality.

The escape hatch I find tempting, but ultimately don’t think works on its own, runs through the degenerate structure of singularities.

Start with the pyramid. Mathematically, the tip of a pyramid is a point. But it’s a very particular kind of point: it’s where three planes meet. The point itself, in its local structure, contains information that says “here, three surfaces converge.” That information is intrinsic to the degenerate geometry of that location. A generic point is just a point, but this point is implicitly a degenerate triangle and implicitly the tip of a pyramid.

The tip of a pyramid is a point but, you can also see it as a kind of degenerate triangle, or even a degenerate pyramid. “The place where three planes meet” is, even if still “a point”, also a boundary of something that is more than a point.

Topological defects can be thought of in the same way. A defect is a place where the field has special, degenerate structure, where many directions of approach collapse into one location. The defect carries intrinsic information about its neighborhood (the local jet, the winding number, the type of singularity) and, by conservation laws, about the global topological state of the region it sits in. A generic point in a smooth field is informationally interchangeable with any other generic point. A defect isn’t.

There’s an extension of this argument that is initially compelling and that I want to spell out before explaining why it doesn’t work. Suppose I have a topological pocket with genuinely complex internal structure (a rich field configuration, lots of information distributed across its volume). Now imagine I could shrink that pocket smoothly toward a point, while preserving the field’s evolution. In a smooth classical field, information doesn’t disappear under adiabatic compression: it gets concentrated into smaller and smaller volumes, encoded in higher and higher derivatives, sharper and sharper jets. In the strict mathematical limit, you’d have a distribution-valued field with all of the original pocket’s “moments” piled up at the limiting singular point. The degenerate structure at the resulting near-singularity would then encode, in some real sense, the original pocket’s full information content. The pocket’s content has been collapsed into the defect’s degeneracy.

A way to make this less abstract is to think about a knotted string. Imagine a one-dimensional loop tied into a complex knot. The knottedness is a genuine topological property; it’s not in any single point along the string but in the way the string winds around itself globally. Now slowly shrink the whole thing. As the loop gets smaller, the knottedness has nowhere to go. It can’t be “uncrossed” without cutting the string, so the topological information stays put. In the limit where the loop has shrunk to a point, that point is a genuinely degenerate locus that has, in some sense, absorbed the whole knot’s structure. It is no longer “a generic point” by any reasonable description; it is a place where a knot ended up.

Critical points of 3D vector fields: nodes, saddle nodes, foci, and saddle foci, in repelling and attracting variants. Source: Introduction to Vector Field Topology (by Tobias Günther and Irene Baeza Rojo in 2020; link)

The right column of this taxonomy of critical points (the attracting nodes, saddle nodes, foci, and saddle foci) is conceptually the closest physical picture I know for what the limit of a shrinking knotted pocket would look like. An attracting focus is a point where many trajectories spiral inward, and the pattern of how they spiral encodes structure about what’s converging into the point. A saddle focus is the same but with directional asymmetries that record additional structure about what kinds of flows had to be reconciled to reach the convergence. None of these attractors are featureless, and none of them are interchangeable with a generic point: they carry a distinctive local geometry that reflects the global structure they sit at the bottom of.

This isn’t only a thought experiment. Solar physics is full of phenomena that work this way. Magnetic flux ropes in the sun’s corona are bundles of field lines that get knotted and twisted by motions at their anchor points in the photosphere. The topology stores real energy, sometimes enormous amounts. When two flux ropes are squeezed close enough, they reconnect, snapping into a new topological configuration and releasing the stored energy as a solar flare or coronal mass ejection. The act of the topology resolving (a knot loosening, two ropes swapping connectivity) is itself a major macroscopic physical event with global causal consequences. Vortex tubes in fluids and quantized vortex lines in superfluids do similar things at smaller scales: they can interlink, reconnect, and release stored helicity when their topology shifts. In all of these cases, when a topologically nontrivial structure shrinks or collapses, the resulting localized configuration carries the signature of the topology that produced it. The structure doesn’t just vanish; it gets written into the geometry of where the collapse converges.

If the analogy carries through to consciousness, defects aren’t just topological labels: they’re potential reservoirs of arbitrarily complex information, and locating consciousness at a defect would mean locating it at a place where the entire bounded pocket’s content has been concentrated into a single locus. That would be a real answer to the combination problem, much stronger than the partial-information answer.

The reason it doesn’t quite work is the diffraction limit. The information stored in higher and higher derivatives, sharper and sharper jets, requires higher and higher frequencies to be physically expressed. There’s no free lunch: you can’t fit complex information into a sub-wavelength volume without paying with bandwidth that scales as the inverse of the target size. In the strict mathematical limit, the frequencies needed go to infinity. In any actual physical system, there’s a cutoff: the wavelength of the medium’s fastest modes sets the smallest volume in which information can coherently exist. Below that scale, the “information” is no longer dynamically accessible because the modes that would express it aren’t available.

For the brain specifically, the problem is not merely abstract wavelength but effective bandwidth: the biologically available coherent modes are slow, lossy, dispersive, and spatially coarse. Conduction velocities, dendritic filtering, and tissue-scale damping all impose an effective spatial-resolution limit that is enormously coarser than what would be needed to compress rich experiential structure into a point-like defect. Whatever the exact cutoff turns out to be, it isn’t the kind of arbitrarily high-frequency, arbitrarily fine-grained spectrum that the abstract shrinking argument requires. Even if the limit-of-smooth-fields version of the argument is mathematically valid in a Platonic medium, evolution couldn’t recruit it as a physical mechanism in a system whose temporal and spatial scales are what they actually are. The defect-as-information-reservoir story is elegant in the abstract but biologically out of reach.

This is what the diffraction-limit argument I sketched in the topological pocket post was getting at. You can’t shrink experience to a point, even in principle, without the energy bill coming due. And in any real biological system, the bill is due long before you reach interesting compression.

So the defect-as-experience-locus story doesn’t quite work. We need a finite-volume answer: experience lives in the pocket’s whole 4D extent, with defects defining its boundary rather than being the locus where everything happens. And we need a mechanism by which a finite-volume region behaves holistically without requiring information to be compressed below the wavelength of the medium. The natural candidate is the path integral, which is the subject of the next section.

Before getting there, let me close out one piece of the topology argument that has to be in place regardless of which version we use: the worry that level-set definitions of pockets are epiphenomenal.

Suppose the pocket is defined by the zero-set of some field φ, the locus where φ(x) = 0. Now apply a uniform shift, φ(x) → φ(x) + c. The zero-set of the shifted field is the c-level set of the original, which is generically a different surface with different topology. If the dynamics depend only on derivatives of φ, the shift changes nothing physical: the equations of motion are invariant, and two configurations giving identical dynamics correspond to different “topological pockets” under the proposal. Whatever pocket we drew was a fiction of our chosen reference value, not a feature of the field that the field cares about. The “topology” was epiphenomenal.

The lesson is that topological features used to ground consciousness can’t be tied to a specific level set. They have to be invariants: things that survive whatever shifts and gauge transformations the field admits. Winding numbers around singularities, linking numbers between closed loops, holonomies around Wilson loops, Chern numbers, defect classifications labeled by homotopy groups of the order-parameter space. Each is defined without reference to any chosen “zero,” and each shows up in physics as a quantity with measurable consequences nothing else can produce.

The cleanest examples are topological excitations whose dynamics are protected by their topology. Vortex cores in superfluids and superconductors carry quantized circulation, interact at a distance through the field they organize, and cannot be smoothly destroyed by any local perturbation, only annihilated by meeting an anti-vortex. Skyrmions, magnetic monopoles in spin ice, dislocations in crystals, defects in nematic liquid crystals: all are localized objects whose existence and interactions are determined by global topological features, not by level-set artifacts. You can shake them with local forces, but you can’t make the topological charge vanish without paying a non-local price.

The Aharonov-Bohm effect is the poster child. An electron is an excitation of the electron field, and in the AB setup its phase responds to gauge-invariant holonomy rather than to local field values along its path. The phase shift is the line integral of the vector potential around a closed loop, and that integral is gauge-invariant: any local redefinition of the potential leaves it alone. This is exactly the level-set-independence criterion. The electron responds to the topology of the field configuration, not to local field values, in a way that has been experimentally verified for decades. Delocalized electron systems (aromatic rings, metallic conduction bands, superconducting Cooper pairs) extend this point: their collective behavior is sensitive to global features of the medium they inhabit, not just to local field values at each constituent’s location.

So the topology is not epiphenomenal when properly defined. The defect-as-locus story is also not epiphenomenal, just bandwidth-limited. The way forward is to keep the topological pocket as the right shape of the bounded region, drop the attempt to compress the experience to a point inside it, and look for a finite-volume mechanism that produces holistic behavior over the whole pocket without needing point-like compression.

VI. The 4D pocket as the natural finite-volume answer

If experience can’t be compressed to a point inside a pocket, the pocket itself has to be where the experience lives, and we need a mechanism by which a finite-volume bounded region behaves holistically. Two moves help here: extending the pocket to four dimensions, and adopting a sum-over-histories framing for what happens in its interior. I want to be careful about what the second move claims. I don’t mean that writing down a path integral explains consciousness. I mean that if the unit of experience is a bounded 4D field-region, then a sum-over-histories or boundary-constrained formulation is closer to the right mathematical grammar than a sequence of locally updated 3D snapshots.

I’ve been describing pockets as if they were spatial objects, snapshots at an instant. The pockets that matter for experience are 4D structures in spacetime, and the binding problem already has a temporal sibling that makes this obvious. A musical note experienced from onset to release isn’t a sequence of disjoint instantaneous experiences glued together; it’s experienced as extended, with internal temporal structure. Any field-topology answer that quantizes experience into 3D snapshots needs an extra mechanism to do the stitching. If pockets are 4D from the start, the stitching is built in: a 4D pocket has internal temporal depth by construction, and the contents at one end of its time axis are part of the same enclosed object as the contents at the other.

This also resolves the all-to-all visibility worry the previous section ended on, but it raises a few questions worth taking seriously. If the pocket is a few tens of milliseconds in temporal extent (long enough for gamma-band oscillations to bind disparate cortical micropockets into coherence, which I think is roughly the right order of magnitude), then signals have had time to traverse the pocket end to end during its existence. Visibility doesn’t have to be instantaneous in any frame. It has to be frame-invariant and topologically guaranteed within the 4D extent of the pocket. We get holistic behavior without needing to compress the pocket to a point, because the pocket has time-depth.

The main danger is smuggling the old problem back in through time: if the pocket has temporal depth, why isn’t it secretly a stack of micro-experiences? My answer is that the 4D pocket is the unit; the instantaneous slices are coordinate artifacts, not ontological parts. Slicing the pocket into instants and asking how those slices compose is the same kind of mistake as slicing a topological pocket spatially and asking how the slices compose: the slices aren’t the natural decomposition of the object, they’re an external coordinate choice imposed on it.

Many internal paths, many partial computations, many trajectories of the field through the interior, all of them are real. They don’t sit beside each other as parallel mini-experiences; they get aggregated into a single bound state by whatever the right finite-volume dynamics turns out to be. A path-integral framing at least gives the right shape: the bounded 4D region is not treated as a sequence of independent local instants, but as a whole whose admissible histories are constrained by its boundary. This is structurally similar to what happens in the PageRank monadology toy model I sketched in Fixed Buckets Can’t Phenomenally Bind: rich internal dynamics where subsystems can disagree, partial computations that don’t immediately reach consensus, all of which get aggregated by the topology into a single externally-visible state. The interior of the pocket is where the work happens. The boundary is where the work has converged. The experience is the converged whole, with the internal convergence process itself contributing to its character. (This last point is what makes valence a candidate to live here: when convergence is hard, when the internal subsystems are pulling in incompatible directions, the difficulty of reaching unity is itself a phenomenal feature, not just a computational one.)

Who experiences the pocket? Nobody separate from it. The pocket is the experiential subject. There’s no homunculus inside watching the field configurations resolve. The aggregation process is what it is like to be that pocket, in the same way that a measurement outcome in quantum mechanics doesn’t have an inner observer. The pocket’s internal aggregation produces a definite state, and the state is the experience. Asking who “sees” it is misplaced for the same reason it would be misplaced to ask who watches an electron’s wavefunction collapse: the resolution of the dynamics is the answer, not a precondition for the answer.

The pocket’s causal power runs in two directions, and both matter for the framework. First, content-level: the pocket’s aggregated state is the boundary condition that subsequent pockets in the temporal chain inherit, so what a pocket converges to causally shapes what the next pocket can become. Second, structural-level: the dynamics of pocket-formation themselves (what kinds of internal disagreements arise, how convergence unfolds, what makes coupling between pockets possible) are causally significant in ways that don’t necessarily show up as the direct content of any one experience. The way pockets relate to each other, the topology of their inter-pocket coupling, how the convergence-difficulty of one pocket affects the boundary conditions available to the next: all of this is causally real and evolutionarily tractable, even though no individual moment of experience represents it explicitly. Evolution selects on both kinds of effect. The structural-level causation is plausibly where valence does most of its work as a behavior-shaping force, since it gates which pocket-formation dynamics are reinforced over time even when no individual experience consciously represents that gating.

This is the version of the picture I currently find least confused. Defects define the boundary of the pocket. The pocket extends in 4D so that signals can traverse it during its existence. The sum-over-histories framing of the pocket’s interior is what makes it holistic. None of these moves requires sub-wavelength compression; all of them are compatible with biologically reasonable frequencies and finite propagation speeds. And the resulting object is still topologically bounded, still has the hard-edge property that resonance theories couldn’t get on their own.

Two payoffs come with the 4D framing. Memory linkage between consecutive pockets becomes geometric: each 4D pocket overlaps temporally with its neighbors, and the regions of overlap are where information about the recent past is integrated into the next pocket. The “pseudo time-arrow” of phenomenal experience can be cashed out as overlap structure between consecutive pockets rather than as a separate mechanism added on top. And the cessation case from section IV gets seen from a new angle: a 4D pocket symmetric under time-reversal has no internal time-arrow, which matches the phenomenology of cessation reports.

A caveat about the temporal extent. The “tens of milliseconds” guess is gamma-band-flavored and feels right phenomenologically. David Pearce, working from a very different angle, has argued that the relevant binding timescale is closer to femtoseconds (ten-to-the-minus-fifteen seconds), which is roughly the lifetime over which brain-spanning quantum coherent superpositions of neuronal feature-processors could in principle exist, before thermal decoherence destroys them. Tegmark famously argued that decoherence at this timescale rules out quantum binding for consciousness. Pearce’s response is to bite that bullet from the other side: yes, decoherence is fast, and that is exactly why each unified moment of experience corresponds to one such fleeting coherent superposition rather than to anything longer. The phenomenology of “this lasting moment” then turns out to be an artifact of how the moments string together, not a clue to how long any one of them really lasts. I disagree with him on phenomenological grounds, but the framework itself is agnostic about which timescale is correct. The exact duration is a parameter to be measured, not a commitment of the framework.

VII. The slicing problem and the fuzzy individuals problem

The slicing problem, which Chris Percy and I wrote up in 2022, targets substrate-neutral computational theories. Build a Turing-complete computer out of water-flow logic gates, then insert a thin sheet down the middle that splits every gate, pipe, and flow into two causally independent but functionally identical halves. A computationalist has to say either that consciousness has doubled (a “consciousness-multiplying exploit” with no change in mass) or that it hasn’t (in which case the boundary between two computations is doing heavy ontological lifting the theory can’t pay for).

This is a special case of a deeper problem the topological pocket post made central: in an ontology of relativistic particles and forces, systems don’t have objective boundaries, only conventional ones. Brian Tomasik has been admirably honest about this. If you go computationalist all the way down, there’s no principled way to count individuals (is the United States conscious? each state? each city? each household?), and “computation gives rise to experience” becomes a commitment to fuzzy, observer-relative experiences. You can bite that bullet, but you’ve given up on consciousness having the kind of definite, non-arbitrary existence that motivates the boundary problem in the first place.

Topological field theories don’t have this problem. You can’t slice a topological pocket without changing its topology. Push a barrier through a vortex and you don’t get two vortices for free: you either destroy the defect (no pocket) or you’ve performed a topology-changing operation with its own energetic cost, producing a configuration that isn’t two copies of the original. Topology isn’t redundant along an axis the way computation is. The same property answers Tomasik’s puzzle: there’s a frame-invariant, causally significant fact about how many topological pockets exist in a region, even when there isn’t one about how many “computations” exist. The 4D framing sharpens this further, since slicing a spacetime pocket requires coordinated topological surgery across its entire temporal extent.

VIII. Closing

If you’re keeping score, the picture asks for: hard boundaries that are not arbitrary (topological pockets), frame-invariance (topology survives Lorentz transforms, level-sets don’t), real causal teeth via gauge-invariant invariants (winding numbers, holonomies, Chern numbers), holistic visibility through 4D extent and path-integral structure rather than point-like compression, temporal binding without an extra mechanism, non-arbitrary cost for creating new bounded experiences, and a translation handle via the Steklov-style boundary conditions imposed by the body’s resonator ecology.

None of this is yet a complete theory of consciousness. The translation problem in particular is not solved by anything I’ve said here. What I think the post adds: a clarification of why defect-as-experience-locus is tempting but doesn’t work in real biological systems (the diffraction limit kills it), and an argument that the natural alternative is a 4D pocket whose interior is governed by path-integral-like dynamics, with defects defining the boundary rather than being where everything happens. This is a different commitment from “mind dust is just defects all the way down.” It’s “mind dust is bounded 4D pockets, and the pockets behave holistically because of how their interiors are dynamically integrated, not because experience has been compressed to a point.”

The combination of resonance (Lehar, Hunt), topology (the boundary problem paper, the topological pocket post), embodied-resonator boundary conditions, and 4D path-integral interior dynamics gets us further than any of these has gone alone.

The obvious failure mode is that the topology I’m pointing to is real physics but not the topology consciousness uses. To make the proposal substantive rather than suggestive, one would need to identify candidate field variables, show that their invariant structures occur at the right scales in brains, show that those structures correlate with changes in phenomenal unity, and show that perturbing them changes experience in ways not predicted by ordinary synchrony or information-flow measures. None of this is in hand. The argument here is a constraint-shaping argument, not an empirical theory: if consciousness lives in fields, then the boundary problem pushes us toward gauge-invariant topological individuation rather than synchrony, level sets, or observer-chosen networks. That tells you what shape an answer would have to have. It doesn’t tell you which specific physical structure is the answer. That is annoying, but also the point: the boundary problem should constrain the search space before we pretend to have found the substrate.

A final word on the Tier III question I deferred from section I. The post’s central argument lives at the continuum-field level, and most of what I’ve argued (frame-invariance, gauge-invariant invariants, defect-bounded pockets, path-integral interior dynamics) requires only a continuum substrate, not a Tier III process-topological monad. But there’s a deeper version of the boundary problem that the continuum framing doesn’t fully close. Even a continuum field has to be specified somehow (which subset of spacetime, which mode, which fluctuation regime counts as “the field hosting consciousness”?), and that specification is, in a sense, still observer-imposed at the meta-level. The Fixed Buckets Can’t Phenomenally Bind argument is that only systems where individuation is generated by the dynamics rather than stipulated (where the boundary of a unit is computed by the system’s own topological behavior, not labeled by an external modeler) close the boundary problem completely. Strongly connected components of dynamic process graphs are one toy example. Whether the brain’s actual physical substrate satisfies the Tier III condition is an open empirical question, and probably the real frontier of the program. For the purposes of this post, the continuum-field level is enough to get the binding/boundary/translation arguments off the ground; the Tier III move is what I think eventually closes the deal.

The a future post I will get into ideas for how the body’s resonator ecology populates the pockets the topology gives us, which is starts to make contact with the texture of experience proper in a much more phenomenologically-meanigful way.

---

References

• Chalmers, D. (2016). The combination problem for panpsychism. In G. Brüntrup & L. Jaskolla (Eds.), Panpsychism: Contemporary Perspectives. Oxford University Press. https://consc.net/papers/combination.pdf

• Cube Flipper (2025). Path integrals and orbifolds: What is it like to be a cube? Smooth Brains. https://smoothbrains.net/posts/2025-06-01-path-integrals-and-orbifolds.html

• Gómez-Emilsson, A. (2023). The view from my topological pocket: An introduction to field topology for solving the boundary problem. Qualia Research Institute. https://qri.org/blog/my-topological-pocket

• Gómez-Emilsson, A. (2026). Fixed buckets can’t (phenomenally) bind. Substack. https://andrsgmezemilsson.substack.com/p/fixed-buckets-cant-phenomenally-bind

• Gómez-Emilsson, A., & Percy, C. (2022). The slicing problem for computational theories of consciousness. Open Philosophy, 5(1), 718–736. https://www.degruyterbrill.com/document/doi/10.1515/opphil-2022-0225/html

• Gómez-Emilsson, A., & Percy, C. (2023). Don’t forget the boundary problem! How EM field topology can address the overlooked cousin to the binding problem for consciousness. Frontiers in Human Neuroscience, 17, 1233119. https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2023.1233119/full

• Hunt, T. (2020). Calculating the boundaries of consciousness in general resonance theory. Journal of Consciousness Studies, 27(11–12), 55–80. https://philpapers.org/rec/HUNCTB-2

• Hunt, T., & Schooler, J. W. (2019). The easy part of the hard problem: A resonance theory of consciousness. Frontiers in Human Neuroscience, 13, 378. https://www.frontiersin.org/articles/10.3389/fnhum.2019.00378/full

• Lehar, S. (1999). Harmonic resonance theory: An alternative to the “neuron doctrine” paradigm of neurocomputation to address Gestalt properties of perception. http://slehar.com/wwwRel/webstuff/hr1/hr1.html

• Pearce, D. Non-materialist physicalism: An experimentally testable conjecture. https://www.physicalism.com/

• Rosenberg, G. (1998). The boundary problem for phenomenal individuals. In S. Hameroff, A. Kaszniak, & A. Scott (Eds.), Toward a Science of Consciousness II. MIT Press.

• Tomasik, B. (2015). Fuzzy, nested minds problematize utilitarian aggregation. Essays on Reducing Suffering. https://reducing-suffering.org/fuzzy-nested-minds-problematize-utilitarian-aggregation/