Volume 27 · Interleaf on auditory models
The Ground Bass
Vision flattens. Hearing does not. If a person is going to hold several dimensions in mind at once, the ear is the sense already trained for it.
Why dimension is hard to picture
Every figure in this volume is a projection. A metric tensor with ten independent components arrives on the page as a surface with a dent in it. A braided worldline arrives as two crossing lines. The reader is asked to accept a picture that has already discarded most of what the object carries, and then asked not to reason from what the picture appears to say. This is the flattening problem the volume has named elsewhere, and it is not a failure of the illustrator. Vision resolves a scene onto a two-dimensional retina. Depth is inferred, and only ever one extra dimension of it.
So the difficulty is not that higher-dimensional structures are obscure. It is that the sense we default to for modelling them cannot hold them, and no amount of care in the diagram fixes that.
What the ear already does
Hearing works differently. The cochlea decomposes an incoming pressure wave into frequency components along its length, and the listener perceives those components at once, as separable, without deciding to. Two people speaking, a bass line under a melody, a room's reverberation and the sound making it — these are held simultaneously and distinguished without collapsing into a single front-facing image. A trained listener can follow four independent voices in a fugue and report on each one.
That is the capacity worth borrowing. Not the physics of sound; the habit of attending to several independent quantities that coexist, interact, and remain individually legible. It is the closest thing ordinary experience offers to holding a many-component object without projecting it.
The Chaconne as instrument
A chaconne is built on a ground: a short harmonic pattern that repeats without interruption while everything above it is rewritten. Bach's Chaconne, the final movement of the Partita No. 2 in D minor, runs some sixty-four variations over one such pattern, on a single violin. The bass is often not even sounded — it is implied, held in the listener's memory, and the variations are heard against it.
This is why it works as a training instrument rather than an illustration. The listener is required to do three things at once: hold a structure that is not currently being played, follow a surface that keeps changing, and recognise the moment the surface returns to the structure. That is the exact cognitive posture a reader needs in front of a coordinate change — and unlike the diagram, the piece does not let the reader off with a picture.
The mapping, stated plainly
Because this volume's rule is to name the proxy out loud, here is the correspondence and its status, line by line. Two of the rows are pedagogical, one is an observation about attention, and one is declared not to map at all.
| In the music | What it models | Status |
|---|---|---|
| The repeating bass pattern (the ground) | The fixed structure the variations are expressed in — the coordinate scaffold that does not move while everything above it does. | Pedagogical mapping |
| Successive variations over the same ground | Different coordinate presentations of one object: the surface changes, the underlying relations are preserved. | Pedagogical mapping |
| Simultaneous voices on one instrument | Independent components carried at once — the experience closest to holding several degrees of freedom in attention without collapsing them into a picture. | Cognitive observation |
| The return of the ground after a distant variation | Recognition of an invariant: what survives the transformation is what the structure actually was. | Pedagogical mapping |
| Harmonic tension and release | No mathematical counterpart. Affect is what makes the exercise bearable, not what makes it true. | Explicitly not a mapping |
A listening exercise
- Listen once for the bass pattern alone. Do not follow the melody. Learn the length of the cycle until you can feel where it turns over.
- Listen again and count cycles rather than following the tune. Notice that the pattern continues even in passages where no bass note is played. You are now maintaining a structure from memory while receiving something else.
- Listen a third time and hold both: the invariant cycle underneath and the variation on top. Note the discomfort of the split attention. That discomfort is the same one a reader meets when a metric is written in two coordinate systems.
- Finally, mark the places where a variation strays furthest and still lands on the ground. Those returns are the audible form of an invariant, and they are the only part of this exercise that transfers directly to reading mathematics.
The catalogue of forms
Once the ear is admitted as a modelling instrument the catalogue grows quickly, and the Chaconne stops being special. Each form below trains a slightly different aspect of the same three capacities: holding a structure that is not always sounding, tracking change against invariance, and refusing to collapse simultaneous streams into a single image.
| Form | What it trains |
|---|---|
| Ground bass, chaconne, passacaglia | Invariant structure under continuous variation. The scaffold is often unsounded and must be carried in memory. |
| Fugue | Independent voices held at once and passed between registers. Trains against the collapse of simultaneous streams into one line. |
| Isorhythm | A fixed rhythmic pattern (talea) repeating beneath changing pitch material — one invariant axis while another varies freely. |
| Theme and variations | Recognition of return after distance: what survives the excursion is what the theme actually was. |
| Polyphony generally | Multiple lines interacting while remaining individually legible — interaction without loss of identity. |
| A sustained drone under changing melody | The minimal case: one fixed reference against which everything else is measured. The cheapest available exercise in invariance. |
The abundance is real, and the list can keep growing — spectral writing, isorhythmic motets, algorithmic and process pieces all belong here. The abundance changes nothing about the limits already stated. Every one of these forms is a perceptual exercise carried on a one-dimensional physical signal. None of them becomes a metric, a manifold, or evidence. The discipline is unchanged: name the proxy, take the training it offers, and leave the geometric object outside the song.
Why an instrument this expensive
If these are only mental models, a fair objection follows: why would a mind have them at all? The capacity to hold several streams at once, to track an invariant through transformation, to stretch into an unfamiliar scaffold — these are metabolically and informationally costly. The question scales up without losing its force. Why would any system, in any universe, sustain something as expensive as a human if there were no return on it?
Four accounts remain live, and they are not mutually exclusive. Each is a hypothesis about function, not a finding, and each is kept inside its band rather than promoted into the volume's argument.
Free rider on survival machinery
Working hypothesisThe capacities that track a ground bass under variation are the ones that model predators, coalitions, tool trajectories, and seasonal pattern at a distance. Once such modelling exists it can be turned onto pure structure at little extra cost. The apparently useless exercise rides on machinery that already paid for itself.
Surplus of open-ended prediction
Working hypothesisA system forecasting an environment that never stops changing benefits from generating more models than it currently needs. The surplus shows up as play, art, and mathematics. On this reading the expensive mind is not a luxury but what a sufficiently general predictor looks like from inside.
Selection for non-local pattern
ContestedRelations, invariants, nested structure, and long-range correlation are real features of the world. A mind that registers them — even through proxies — can act on scales a reactive system cannot. Whether this pays in every possible universe is unknown; in this one it has been sufficient to keep the lineage going.
No external reason is required
Live alternativeThe universe is under no obligation to be economical by our standards. Complexity can arise because the dynamical rules permit it, not because a purpose demanded it. Humans would then be one of the things a universe with these laws and this history happens to produce, and the mental models simply what such a system does when it turns attention on itself.
None of these settles the question, and the volume does not need it settled. What can be said is narrower: the expense is real, the models keep being elaborated, shared, and stretched, and on any of the four accounts that continued elaboration is doing some work. No inevitability is claimed and no purpose is assigned. The honest position is that the accounting is open and the practice earns its keep for now.
Evolutionary psychology of cognition
The four hypotheses above can be read through the evolutionary-psychology literature, but the literature does not converge on a single verdict. Several well-known programmes each capture part of the picture and each has its own boundary.
Ecological intelligence
Working hypothesisThe capacity to model objects, forces, trajectories, and seasonal patterns at a distance pays inside the environments where it evolved. The same machinery that tracks a thrown stone or a predator's path can be turned toward abstract structure once the immediate survival problem is solved. On this reading the expensive mind is an overfitted tracker: the surplus capacity is real, but its later uses are side-effects, not the selection pressure that built it.
Social brain hypothesis
ContestedMuch of the human cognitive budget may have been selected for tracking alliances, reputations, deception, and cooperation in large groups. The skills that let a person hold several social narratives at once — who owes whom, who is lying, who is allied with whom — are not unlike the skills that let a listener hold several voices in a fugue. Whether this is the dominant driver of general intelligence remains disputed; the correlation between group size and brain size is suggestive but not decisive.
Cumulative culture
Working hypothesisOnce a species can store, transmit, and improve solutions across generations, the selective landscape changes. Minds that can learn from accumulated tools, languages, and notations gain access to problems no individual could invent alone. Mathematics and music become possible not because genes selected for them directly, but because cultural inheritance can build structures that biology alone would not pay for.
Exaptation
Live alternativeA trait built for one function can be co-opted for another. The auditory-scene analysis that lets a listener separate prey sounds from wind may be exapted into the ability to separate simultaneous voices in a fugue. The question then becomes whether the later use is a lucky free ride or whether it feeds back into fitness in ways that reshape the trait. The evidence is mixed, and the answer is likely different for different capacities.
The common thread is that none of these programmes needs the mental models to be directly selected for. They can arise as by-products, social tools, or cultural accumulations. The expense is still real, but the accounting is distributed across biology, social life, and history rather than falling on a single adaptive story.
Thermodynamics of information processing
There is also a thermodynamic way to ask why an expensive mind is sustained. Information processing is not free. Landauer's principle sets a minimum energy cost for erasing one bit of information at temperature T: kT ln 2 per irreversible operation. Neural computation, whatever else it is, must pay this floor and the much larger biological overhead that comes with keeping a brain alive.
Several frameworks try to make this cost look like a design pressure rather than a constraint. They are useful models, but each has its own validity band.
Predictive processing / free energy
ContestedThe brain is framed as an inference machine that minimizes surprise by updating internal models against sensory input. A more general mind can absorb more kinds of surprise, but it also spends more energy maintaining the models. Whether the brain literally minimizes a variational free-energy quantity, or whether this is a useful formal redescription, is still debated. The framework is productive; it is not yet established as a literal mechanism.
Efficient coding
Working hypothesisSensory systems appear to allocate limited resources to the statistical structure of the environment: more precision where signals vary more, less where they are predictable. This is a constrained optimization, and it explains why neural tissue is arranged as it is. It does not, by itself, explain why some minds go on to construct abstract geometries or ground-bass models; it only explains why the early stages of perception are shaped the way they are.
Thermodynamic depth
Live alternativeSome accounts measure the complexity of a system by how much historical information is required to produce its current state. A mind with a long cultural and developmental history is then thermodynamically deep in a literal sense: it carries many irreversible steps. This does not make the mind optimal; it only gives a measure of how much history is compressed into it.
The thermodynamic angle changes the question. It is no longer "why does a mind have these capacities?" but "given that information costs energy, how much of that cost is being recovered, and in what currency?" The currency may be survival, social coordination, prediction, or something less obvious. The accounting is open.
A profligate universe
Step back further. By any human measure of economy the universe is extravagantly profligate. It produces far more emergent entities than any obvious function appears to require: galaxies, stars, planets, chemical species, biological lineages, neural architectures, and, within some of those architectures, the capacity to hold ground basses, spectra, braids, and provisional geometries. It spends billions of years and cubic gigaparsecs on processes that never converge on a single optimal form. Variability is not rationed; it is the default.
A profligate universe does not ration its dimensions, its topologies, its stretches of time and space, or the imaginations that arise inside them. It generates form after form, function after function, and then generates minds capable of noticing the patterns, holding multiple streams at once, shifting perspective, and building provisional models of what they encounter.
| What the universe supplies | What it makes possible |
|---|---|
| Dimensions and topology | The possible relations among things: nearness, continuity, winding, enclosure, branching. |
| Time and space | The room in which those relations can unfold and vary, so that form is not frozen at birth. |
| Imagination | The capacity to register, extend, and sometimes re-express relations in new forms — trained first by the ear, later by geometry, later still by collaboration with parallel agents. |
That profligacy changes the force of the earlier question. If the background condition is abundance rather than scarcity, then the existence of expensive, open-ended mental models no longer looks like an anomaly that demands a special justification. It looks like one more expression of a system that routinely generates more structure, more variation, and more nested relation than any local purpose would have commissioned.
In that setting the human capacity to stretch into topological or auditory models is not a puzzle to be solved so much as a local instance of the same generative excess. The universe does not appear to be running a tight budget. The excess is the point: more structures than any single purpose requires, more viewpoints than any single observer can occupy, more possible mappings than any one mental model can exhaust. Form and function keep appearing because the underlying generative rules permit them, and because some of the emergent entities — ourselves included — have learned to pay attention.
It appears to be running an open exploration — and minds capable of noticing, holding, and extending patterns are simply one of the things that exploration has so far produced. That is the horizon that stays open.
The observation does not explain why the rules permit such profligacy. It only removes the pressure to treat every costly structure as something that must earn its place by a narrow standard of utility. In a profligate universe, the surprising thing would be the absence of such minds, not their presence.
Where the analogy stops
Music is one-dimensional in its physical variable: air pressure as a function of time. The many voices a listener separates are extracted from a single scalar signal, not carried on independent axes. So the multiplicity is perceptual, not dimensional, and nothing about hearing a fugue demonstrates anything about a four-dimensional manifold.
Nor is the ground bass a metric. It is a convention, chosen by a composer, with no invariance under transformation to defend. A coordinate change is constrained by the requirement that measurable quantities not change; a variation is constrained only by taste. Anyone who takes the resemblance further than mental modelling will be making claims the music cannot support.
What survives all of that is small and real: the ear is practised at holding several coexisting quantities without flattening them, and that practice can be lent to the reading of structures the eye cannot present.
Validity band and falsifier
Validity band
Comprehension aid only: the construction of a mental model for simultaneity, invariance, and re-presentation. It carries no weight in any physical or number-theoretic argument in this volume, and no result anywhere in the archive depends on it.
Falsifier
The pedagogical claim is testable: if readers given the listening exercise show no improvement over readers given the same material with diagrams alone — on tasks requiring recognition of an invariant across a coordinate change — the device has no teaching value and should be dropped. It is not, at present, tested.
Where this interleaf is weakest
It rests on the author's own experience of listening and reading. The auditory-scene literature supports the separability of simultaneous streams; it does not establish transfer to mathematical comprehension. Held as a working suggestion.
Companion interleaves: The Parallax Effect in Music and The Place of the Eigenvalues. Standing entries: Where This Volume Is Weak and The Fluid Still Carries the Sheet.