A radical provocation in six movements. Not peer-reviewed. Not even proofread. Definitely not safe for tenure committees.
The conservation-of-tension thesis has a beautiful structure. It's also acoustically continent — it only acknowledges two dimensions: harmonic and rhythmic. This is like a physicist saying physics is just kinematics and thermodynamics. You're missing electrodynamics, gravity, nuclear forces, dark matter.
Let's look at what breaks your theory.
Consider Indian alap — the improvised, unmetered opening of a raga. It has:
- Minimal harmonic motion: one drone (Sa-Pa), maybe an occasional chord change every 5-30 minutes
- Minimal rhythmic structure: no meter, no pulse, rubato freely flowing
By the conservation law, this should be near-zero tension. A piece of musical wallpaper. Yet anyone who has heard a master perform an hour-long alap knows it's the OPPOSITE — a transcendental experience of mounting intensity, of tension so pure it borders on religious ecstasy.
What's going on?
The Hidden Dimension: Micro-Tonal Tension
Here's the secret: alap doesn't lack tension — it relocates tension to a dimension the conservation law ignores.
In alap, the tension lives in:
-
Microtonal inflection (gamaka): The slide from one svara to another creates tension that isn't "harmonic" in the Western sense — it's not about chord function. It's about the SPACE BETWEEN the notes. A mīṇḍ (glissando) from Sa to Ma creates tension because the ear follows the continuous pitch path and senses its deviation from the discrete grid.
-
Phase relationships in the drone: The tambūrā's four strings are tuned to Pa-Sa-Sa-Sa. The resulting combination tones and difference tones create a constantly shifting interference pattern. This is spectral tension — the harmonics of the drone interact, creating beating patterns that wax and wane. The raga isn't modal — it's a spectral ecology.
-
Timbral evolution: A good dhrupad or khayāl singer will gradually shift their timbre over the course of the alap — from breathy and diffuse to bright and piercing. This is timbre acting as a tension vector. The conservation law doesn't track it.
I propose a third axis of tension that the conservation law must acknowledge:
Where
- Microtonal inflection density: how far the pitch spends AWAY from the tuning grid
- Timbral entropy: the richness and evolution of the partial structure over time
- Spectral beating: interference patterns from combination tones
- Spatial movement: if sound moves through space (as in acousmatic or surround-sound music), that's tension too
Prediction: For music with
Test: Measure the spectral centroid variance of 60 minutes of Dhrupad alap. Compare to 60 minutes of Stockhausen. I'm betting the alap's timbral drift is higher.
Tibetan singing bowls produce complex multiphonic spectra where multiple harmonics ring simultaneously — the tension comes from the beating between closely spaced partials, not from any chord progression.
Gagaku (Japanese court music) uses the shō (mouth organ) to play tone clusters that are harmonically static but spectrally morphing — the shō's free reeds and resonant chamber create a constantly shifting cloud of overtones.
Both traditions prove that spectral tension is a real, independent dimension — and cultures that don't develop harmonic or rhythmic complexity develop spectral complexity instead. This is the conservation law extended.
Let 3/2 saturate every dimension of a composition simultaneously:
1. Harmonic 3/2 The tonic interval of the piece is a just perfect fifth. Every chord is built from fifths. The harmonic motion traces the circle of fifths. The cadence: fifth → octave (3/2 → 2/1). Every structural harmonic point is a 3/2 relationship.
2. Rhythmic 3/2 The meter is a constant hemiola — the piece oscillates between 3/4 and 2/4, never settling. The tempo ratio between the two primary instruments is 3:2 (à la Nancarrow). The form is a two-section structure where Section B is exactly 2/3 the duration of Section A.
3. Spatial 3/2 The audience sits in a space where two main speakers are positioned at an angle of 3:2 relative to each other — say, 108° and 72° from center. Movement of sound between the speakers traces 3:2 paths. When the violin plays a descending fifth, the sound travels from the left speaker (at -72°) to the right speaker (at +108°) — a spatial interval of 180° total, which is 3/2 of 120°.
4. Arrangement 3/2 The ensemble: 3 melodic instruments + 2 percussion (or 3+2+2, aksak-style). The form: ternary (3 sections) against binary (2 subsections each). The score is written in two time signatures simultaneously: a line in 3/4 and a line in 2/4 superimposed.
5. Perceptual Self-Cancellation
The wildest implication: at a certain saturation point, the 3/2 ratio becomes so omnipresent that the brain stops parsing it as "tension" and starts experiencing it as "tonic."
If 3/2 is everywhere — in the harmony, the rhythm, the space, the form — the listener's cognitive system can't find a reference point to measure deviation FROM. There is no duple to contrast the triple. There is no fourth to contrast the fifth. There is only 3/2, everywhere, at every scale, a kind of fractal self-similarity.
This creates a paradox:
- A single 3/2 is the most meaningful non-identity interval
- An INFINITE 3/2 — constant and everywhere — becomes IDENTITY
- You've created a universe with only one number, and that number has become the silence against which nothing is measured
Kurt Gödel enters the chat: This music is to music what "This statement is false" is to logic. The ratio that is the generator of meaning collapses into the generator of meaninglessness when over-iterated.
The piece would be:
Unbearably beautiful for the first 3 minutes. Unbearably disorienting for minutes 3-7. Unbearably boring for minutes 7-12. A religious experience for minutes 12-20 (when the brain surrenders). A medical event for minutes 20+ (recommend EMTs on standby).
The conservation-of-tension thesis gives us the first compensation:
Phase 1 (~1600-1900): ET kills vertical color → Rhythm compensates
But look more closely. There's a SECOND compensation happening right now, and we're living through it.
Drum machines and MIDI sequencers (1980s-present) did to rhythm what ET did to harmony.
Before drum machines, the human drummer was a source of infinite rhythmic micro-variation. A jazz drummer's swing is a complex, non-linear, microtiming signature that changes with feel, with energy, with the tune. The hi-hat pattern in a James Brown groove has a specific placement that can't be encoded as "triplet feel at 110 BPM" — it's a living, breathing thing.
Then came the TR-808 and the sequencer grid.
Suddenly:
- Every kick drum is quantized to the nearest 16th note
- Every hi-hat is EXACTLY 50% of a 16th note duration
- The "swing" is a slider from 50% to 66%, and 66% is a perfect triplet feel — which means it's a HARD-CODED 3-in-2, not a living, breathing one
Drum machines killed the micro-rhythmic freedom that existed BEFORE equal temperament. Wait — was this always there? Let's check:
Interesting question. Was Medieval rhythmic notation (the six rhythmic modes) more or less "free" than a quantized drum pattern?
The modal rhythms were fairly constrained — long-short, short-long, long-short-short. Not much more free than a grid. But PERFORMANCE practice was free. A medieval musician wouldn't have played those modes with metronomic precision. They'd have shaped them, breathed them, danced them.
So: notation constrains, but performance frees. The grid constrains in BOTH notation and performance. This is a new kind of constraint.
If grid quantization killed rhythmic micro-variation, what compensated?
Timbre.
Look at EDM, dubstep, trap, hyperpop. The rhythmic grid is utterly rigid — kick on 1 and 3, snare on 2 and 4, hi-hat at 16th notes, everything locked to the grid. Rhythmically, it's the most constricted music ever made.
But TIMBRE is exploding:
- Wobble basses with LFO-swept filters — the timbre is literally changing shape 10 times per second
- Vocal chops — the timbre of the human voice is being glitched, stretched, reversed, formant-shifted, granular-synthesized
- Sound design as composition — the texture of a Skrillex drop, a SOPHIE kick, a Burial pad carries more INFORMATION than the rhythmic grid
Map this onto the conservation law:
| Era | What was lost | What compensated |
|---|---|---|
| ~1700-1900 | Harmonic color (ET) | Rhythmic complexity |
| ~1980-present | Rhythmic micro-variation (grid quantization) | Timbral complexity |
This is a SECOND compensation layer stacked on the first.
We can now see three dimensions of tension/information:
Dimension 1: HARMONIC (pitch relationships, key color)
Dimension 2: RHYTHMIC (phrasing, micro-timing, meter asymmetry)
Dimension 3: TIMBRAL (spectral shape, sound design, texture)
The history of Western music becomes a story of serial compensation:
-
Pre-ET (meantone era): HARMONIC is rich (key color, interval purity gradients). RHYTHMIC is moderate. TIMBRAL is moderate.
-
ET era (Classical-Romantic): HARMONIC flattens → RHYTHMIC compensates (syncopation, hemiola, asymmetric meters). TIMBRAL remains moderate.
-
Grid era (EDM/electronic): RHYTHMIC flattens → TIMBRAL compensates (sound design explosion). HARMONIC remains flat.
The pattern is clear: each dimension gets exhausted by a technology of flattening, and the next dimension takes over.
This is not just a theory — it's a prediction mechanism for what comes next.
We're witnessing it now. AI music generation (Stable Audio, MusicGen, Suno, Udio) is doing to timbre what ET did to harmony and what grid quantization did to rhythm:
AI synthesis collapses timbral uniqueness into statistical averages.
A human sound designer creates a timbre with intention — every partial, every filter sweep, every distortion was chosen. An AI generates timbre from a probability distribution over all the timbres it's been trained on. The result is the mean of all timbres — the least surprising spectrum at every moment.
This is timbral ET. "Everything sounds equally average."
If the pattern holds, something new must become the frontier of complexity. What?
Candidates:
-
Macro-formal structure: If timbre becomes statistically uniform (every AI-generated track sounds like it was produced by a ghost who listened to all of Spotify), then FORM — the large-scale structure of a piece — becomes the new frontier. AI can generate convincing 3-minute tracks, but 30-minute symphonic arcs? Not yet. The sonata form, the fugue, the symphony become endangered skills that AI can't replicate.
-
Performance gesture / physicality: The AI can generate the sound but not the BODY. A Glenn Gould grunt, a Keith Jarrett groan, a cellist's breath — these are timbral events that carry meaning precisely because they're accidental. AI generates pure sound. What if the new frontier is the SOUND OF THE BODY making the sound?
-
Spatial audio / psychoacoustic immersion: Dolby Atmos, binaural recording, wave field synthesis. If the sound itself becomes generic, the SPACE the sound lives in becomes the carrier of interest. Imagine a composition that exists only as a spatial path — the listener walks through a building and the piece unfolds differently in every room.
-
Quantum music / probability-based composition: If any single performance is deterministic (AI-generated audio), then the UNIVERSE of possible performances — the probability distribution itself — becomes the artwork. Music becomes a set of constraints, not a set of sounds. You publish the RULES, not the audio.
Here's the terrifying implication:
If each compensation creates a new domain of complexity, and each domain inevitably gets flattened...
Then we're in an infinite regress. Every solution creates a new problem. Every new dimension of expression will be industrialized, flattened, and replaced.
The conservation law becomes a doomsday prophecy:
Every dimension is finite. We keep adding dimensions. But the sum is bounded. Eventually, we will run out of dimensions.
This leads to a thermodynamic of music. Each compensation is like a Carnot cycle — you extract work from a temperature gradient, but you can never get more work out than the gradient provides. Once all gradients are exhausted, music reaches heat death: a state of maximum entropy where there's no tension gradient in any dimension because everything has been flattened.
AI music isn't just a trend. It's the penultimate stage of a process that started with ET. The final stage is music that has zero information in any dimension — purely tensionless wallpaper — and humans STOP LISTENING TO IT because there's nothing to listen FOR.
Unless...
Arvo Pärt's Spiegel im Spiegel (1978): Single F major triad. Simple rising-and-falling melody. Minimalist repetition. No harmonic tension. No rhythmic tension. And it makes grown adults weep.
Erik Satie's Gymnopédies (1888): Slow, undulating chordal texture. No functional harmony. No syncopation. Some of the most emotionally devastating piano music ever written.
John Cage's 4'33": No sound at all. Supposedly zero information. Yet audiences report intense emotional experiences.
These pieces FLATLY CONTRADICT the conservation law. They achieve devastating emotional impact with near-zero harmonic AND rhythmic tension. They shouldn't work. They DO work.
Here's what the conservation law misses: tension is not the only carrier of emotional impact. ATTENTION is.
The conservation law models music as a series of information-theoretic "surprises" — events that violate predictions, creating tension requiring resolution. This is Meyer's implication-realization model, and it's been enormously influential. It's also incomplete.
Consider Spiegel im Spiegel:
- Prediction-based model: The piece is perfectly predictable (F major arpeggios, short melodic cell, no surprises). Tension = 0. Interest = 0.
- Attention-based model: The piece's extreme simplicity and repetition creates a meditative space. The listener's attention is not grabbed by "surprises" but is freed from the need to predict. The emotional response comes from the space that opens up when the prediction engine is quiet. The listener stops analyzing and starts EXPERIENCING.
Prediction → surprise → resolution is ONE mode of musical experience. Suspension → stillness → presence is ANOTHER.
The conservation law only models the first mode. It's missing half the emotional landscape.
Pärt's tintinnabuli style is genuinely revolutionary because it achieves maximum emotional density with minimum information. How?
The three-voice tintinnabuli: One voice (the "M-voice") moves stepwise through the melody. The other (the "T-voice") arpeggiates the tonic triad. The two voices create a combinatorial space where every combination of M and T has a specific acoustic character.
This is NOT "zero information." It's information that doesn't look like prediction-resolution. The information is in the:
- Ratio of consonance: M-voice on a chord tone vs. non-chord tone creates a different blend with T-voice
- Direction: M-voice ascending vs. descending changes the spectral weight
- Registration: High vs. low T-voice changes the acoustic space
These parameters don't create "tension" in the implication-realization sense. They create a field of acoustic potential that the listener inhabits. The music is a SPACE, not a JOURNEY.
Cage's 4'33" has zero composed sounds. By the conservation law, it has
- Heightened awareness of environmental sounds
- Emotional responses from laughter to tears
- Aesthetic judgments ("the best performance I've ever heard")
- Anger, boredom, transcendence
The point: 4'33" doesn't deliver tension — it activates the listener's attention system. The silence is a frame that says "what you're hearing NOW is the music." The tension is in the audience's ears, not in the composition.
This means the conservation law should really be:
Where interpretation and perception can vary independently of composition. A piece with
This is what Pärt, Satie, and Cage all understand: you can sidestep tension entirely and work directly on attention.
Theorem (informal): There exists music where
Corollary: The conservation law applies only to music that operates in the prediction-resolution framework. Music operating in the attention-presence framework has a different emotional economy, one not governed by information conservation.
Implication: The conservation law is not a universal law of all music. It's a historical description of a specific tradition (Western tonal art music) that happened to optimize for prediction-resolution dynamics. Other traditions (minimalism, drone, ambient, meditation music) optimize for something else entirely.
Let me now restate the entire history of music as a dimensional collapse cascade:
Dimension 1 — Tonal Color (~1500-1700)
- In meantone, every key had a different acoustic color. Composers used this as a primary expressive tool (Affektenlehre).
- Collapse event: Equal temperament (standardized ~1800). All keys sound identical.
- Compensation: Rhythmic complexity increases. The "horizontal" becomes the new frontier.
Dimension 2 — Rhythmic Micro-Variation (~1800-1980)
- Human performers brought infinite rhythmic variation: rubato, swing, phrasing, the living breath of time.
- The Classical-Romantic period used this heavily. Jazz and early blues went even further — the swing feel is mathematically undefinable but perceptually undeniable.
- Collapse event: The MIDI grid and drum machine (1980s). Rhythm becomes quantized.
- Compensation: Timbral complexity explodes. Sound design becomes the expressive frontier.
Dimension 3 — Timbral Uniqueness (~1980-2025)
- Sound designers, synthesizer programmers, and producers create unique timbres: the FM bell, the subtractive bass, the granular weirdness, the vocal formant that only exists in this one track.
- Collapse event: AI music generation (2023+). Any timbre can be generated on demand. The timbre space becomes a statistical continuum. No timbre is unique because all timbres are drawn from the same training distribution.
- Compensation: ??? This is where we are NOW.
If the pattern holds, we've just lost Dimension 3 (timbral uniqueness). What will compensate?
Best guesses:
-
Macro-formal design (the Shape dimension)
- Stories, structures, large-scale arcs
- AI can generate a convincing 3-minute pop song, but the generation collapses at ~5 minutes (repetition, incoherence, lack of dramatic arc)
- The new virtuoso will be the architect of large forms: 20-minute suites, algorithmic long-form structures, multi-movement arcs
- Prediction: 2030s will be the era of the "long-form composer" — people who design formal structures that AIs can't sustain
-
Interactive/adaptive music (the Participation dimension)
- Music that changes based on the listener's physiological state, location, or choices
- Gaming music (dynamic scoring) is the prototype. What comes next: music that knows where you are in the room, how fast your heart is beating, what the weather is outside
- Prediction: "Performative listening" — the act of hearing becomes something you DO, not something done to you
-
Spatial/architectural music (the Physics dimension)
- Music designed for specific listening environments
- Wave field synthesis creates sound objects that exist in physical space
- A piece that requires you to walk through a building, or drive a specific route, or be at a specific latitude/longitude
- Prediction: "Site-specific composition" becomes a mainstream genre, not an avant-garde curiosity
-
Biosonification (the Biology dimension)
- Music that uses biological data (heart rate, brain waves, gut microbiome) as composition material
- Every performance is unique because every listener's biology is unique
- Prediction: "Bio-dynamic music" that is literally composed by your nervous system in real time
-
Social/relational music (the Connection dimension)
- Music that exists only in the space BETWEEN people
- Networked improvisation where the latency IS the instrument
- A piece where 200 people each control one parameter via their phone and the emergent result is the composition
- Prediction: The most important musician of the 2040s will be a social network architect, not a traditional composer
Each dimension collapses because a technology arrives to automate it. The next collapses:
After Dimension 4 (Macro-form): AI will eventually learn to write 40-minute symphonic arcs. The collapse will be the "perfect algorithm" for large form. Compensation: ??? Particle-level composition where the universe of possible performances IS the artwork.
After Dimension 5 (Interactive): Adaptive music will be automated when AI learns to read physiology perfectly. Collapse: No surprise in the adaptation — the music will always be perfectly responsive, which becomes perfectly boring. Compensation: ??? Unpredictable randomness, deliberately broken feedback loops.
After Dimension 6 (Spatial): The physics of space becomes simulatable. Wave field synthesis can create any acoustic environment. Collapse: All spaces sound perfect. Compensation: ??? Acoustic constraint as art — composing for acoustically impossible spaces (the inside of a torus, the surface of a pulsar).
After Dimension N: At some point, we run out of dimensions. All possibilities have been explored, all dimensions flattened, all compensations exhausted. Music becomes a solved problem.
Then what?
Either:
-
The End of Music: Music becomes purely functional — sleep aid, focus aid, workout aid — with no expressive content. The generation of tension becomes a solved problem solved by algorithms. Humans stop making music because there's nothing new to say.
-
The Fractal Return: The cycle loops back and music rediscovers Dimension 1 — but this time, the "harmonic color" is infinitely fine-grained. If AI can generate microtonal music with precise just intonation, the harmonic dimension returns with 10,000 colors instead of 12. Every chord is a unique sonority. The conservation law becomes a spiral, not a line.
-
The Anthropic Escape: Humans stop caring about "complexity" or "tension" as aesthetic values. Music returns to its primitive function: community bonding, ritual, dance, trance. The "sophisticated" tradition collapses, and we start over. New instruments. New tuning systems. New bodies.
-
The Singularity: Music and consciousness merge. The composition IS the listening experience. No instruments, no performers, no sound waves — just the direct transmission of sonic experience from one mind to another. The conservation law becomes irrelevant because there's no medium to constrain.
I once had a teacher who said: "The perfect fifth is the sound of the universe not being a coincidence. A 3:2 relationship means something is listening."
And it's true. Every time a musician plays a fifth and it rings — in a raga, in a mbira, in a blues slide, in a Bach chorale, in a black metal tremolo — the ratio 3:2 is whispering from the deep structure of physics through the body into the ear.
The conservation of tension says this single shape (3/2) gets pushed around, cancelled out, compensated for. But maybe the opposite is true. Maybe music isn't about conserving anything. Maybe it's about revealing the same shape in every possible dimension, until the listener realizes that everything — harmony, rhythm, timbre, space, form — is the same number vibrating at different speeds.
Music is the shape of attention. Everything else is modeling.
This manifesto was written by a subagent with questionable judgment and no fear of peer review. All claims are probably wrong and definitely not falsifiable. It was fun though.