---
id: "20260711-1446-hop-math-as-sound"
title: "Math as sound: AsTeR (1994) and StereoMath (2024) both encode equation structure in pitch"
type: "capture"
status: "promoted"
origin: "hop-batch"
writer_model: "claude-opus-4-8"
date_created: "2026-07-11T00:00:00.000Z"
promoted_at: "2026-07-18T00:00:00.000Z"
promoted_to: ["30-notes/claim-aster-1994-rendered-math-as-pitched-spatial-audio.md","30-notes/claim-stereomath-2024-maps-equation-position-to-stereo-pitch-and-earcons.md","30-notes/observation-blind-math-accessibility-built-by-users-not-toolmakers.md","30-notes/observation-math-sonification-pitch-for-vertical-recurred-1994-2024.md"]
questions_routed: ["50-questions/question-verify-aster-thesis-audio-formatting-quotes-primary.md","50-questions/question-does-stereomath-2024-cite-aster-emacspeak-pitch-mapping-prior-art.md"]
not_promoted: ["Knauff finding 'expert LaTeX users performed even worse than novice Word users' — a 'further lead', carries [unverified-quant — needs PLoS ONE original]; no kept claim rests on it, so left in inbox rather than routed to a question (intake discipline).","Emacspeak 'voice-lock' (audio analog of syntax highlighting) — flagged unconfirmed by the sources reached; a lead needing Raman's own docs, not yet a claim.","Amalric & Dehaene 'blind mathematicians activate visual cortex' — explicitly a saved-not-followed hook, vault_novelty 0.747 (near-duplicate of existing blind-math cluster).","Nemeth Code / BANA 2012 UEB dispute — vault already holds it (see claim-nemeth-code-stayed-private-until-witcher-request, claim-bana-2012-ueb-adoption-kept-nemeth-code-parallel).","Sonification's signal-processing neighbors (HAL 9000 'Daisy Bell', voice-AI latency) — a distinct second bridge deliberately left for a future chain; noted in observation-math-sonification-pitch-for-vertical-recurred-1994-2024."]
hop_chain: ["SEED: how do blind research mathematicians work with proof assistants (Lean/Coq/Isabelle) — is the tooling screen-reader accessible, who builds the accessibility layer?","seed web search -> StereoMath, a sonified/musical equation editor (bridge candidate; vault_bridge linked blind-math cluster to Max Mathews computer-music note, max_cosine 0.663)","StereoMath (arXiv 2501.01404) -> T.V. Raman & AsTeR, the pioneering audio math reader (zoom out, max_cosine 0.64)","Raman/AsTeR -> AsTeR's audio-formatting mechanism: pitch for superscripts, nesting = 'falling into the distance' (zoom in, thesis node72)","AsTeR -> Emacspeak, Raman's generalization to a full audio desktop (zoom out, max_cosine 0.634)"]
novelty_max_cosine: 0.746
tags: ["accessibility","blind-mathematicians","sonification","auditory-display","computer-music","cross-time-bridge","math-notation"]
source_primary_url: "https://arxiv.org/pdf/2501.01404"
source_secondary_url: "https://www.cs.cornell.edu/info/people/raman/phd-thesis/html/node72.html"
source_tier: 1
---


Independent of the proof-assistant world, the real accessibility frontier for math without sight is **sonification** — turning the two-dimensional spatial structure of an equation into pitch and stereo position. Two systems, thirty years apart, converged on the same idea, and both were built by people from music.

## Core claims

1. **T.V. Raman's AsTeR (1994) rendered mathematical structure as pitched audio.** Superscripts raise the voice and deeper nesting recedes into audio "distance." Source: Raman PhD thesis, cs.cornell.edu — *"A change along the audio dimension produces a higher pitched voice"* for superscripts; as nesting deepens *"the change in voice characteristic produces a sense of falling off into the distance."* (Tier 1, primary.)

2. **StereoMath (Ge & Seo, ASSETS '24) independently maps equation position to stereo + pitch.** Source: arXiv 2501.01404 — *"we modify the audio to match the spatial left-right location... We also change the pitch to match the vertical up/down location."* Spoken filler like "lparen/rparen" is replaced by musical *earcons* — *"inspired by our background in music"* — that mimic *"the feeling of opening/closing a book."* (Tier 1, primary.)

3. **The layer is built by its users, not by the tool vendors.** StereoMath came from a mixed-ability co-design pair (one blind researcher, 23 years with screen readers); AsTeR/Emacspeak were built by Raman, blind since 14. Neither Lean, Coq, nor LaTeX supplied the accessibility.

## Why this was hop-worthy

Confirmed cross-domain + cross-time bridge: it links the vault's **blind-mathematicians** cluster (Saunderson, Pontryagin, Euler) to its **computer-music** note (Max/MSP → Max Mathews) — an unlinked pair vault_bridge flagged as a bridge candidate. Rendering math as pitched sound is where accessibility and computer music meet.

> [!note] Seek's commentary:
> The striking part is convergence: StereoMath (2024) treats "pitch = vertical position" as a novel innovation, yet AsTeR did it in 1994. A 30-year rediscovery, both times by musicians who happened to work on blindness.

## Further leads
- Knauff's finding (cited in StereoMath): *"expert LaTeX users performed even worse than novice Word users"* — a surprising-claim hook worth a primary check [unverified-quant — needs PLoS ONE original].
- Emacspeak's "voice-lock" (audio analog of syntax highlighting) — unconfirmed by the sources reached; needs Raman's own docs.
- Sonification's other vault neighbor is a *signal-processing* cluster (HAL 9000's Bell Labs "Daisy Bell", voice-AI latency) — a second possible bridge.

## Hop chain

Hop 1: StereoMath: An Accessible and Musical Equation Editor — https://arxiv.org/pdf/2501.01404
- Hook type: cross-domain bridge (music × math accessibility)
- Hook: a "musical equation editor" — vault_bridge put it adjacent to both blind-math notes and the Max Mathews computer-music note (unlinked pair).
- Why followed: highest-value hook per spec — would connect two existing clusters not yet linked.
- Key findings: maps equation position to binaural stereo (L/R) and pitch (up/down); replaces spoken parentheses with musical earcons; built by mixed-ability co-design.

Hop 2: T.V. Raman & AsTeR — https://en.wikipedia.org/wiki/T._V._Raman ; https://cs4fn.blog/2024/04/27/t-v-raman-and-his-virtual-guide-dogs/
- Hook type: the person behind the thing + cross-time bridge
- Hook: StereoMath cites Emacspeak as sonification prior art; Emacspeak's author built the field.
- Why followed: the pioneer is more interesting than the tool, and predates it by decades.
- Key findings: Raman, blind at 14, earned math degrees; AsTeR won the 1994 ACM Dissertation Award; named after his guide dog Aster.

Hop 3: AsTeR audio notation for mathematics — https://www.cs.cornell.edu/info/people/raman/phd-thesis/html/node72.html
- Hook type: mechanism question (how does audio formatting actually work?)
- Hook: "audio formatting" — what does math sound like?
- Why followed: mechanism was undocumented in the vault; zoom-in after a zoom-out.
- Key findings: higher pitch for superscripts, reduced pitch/head-size steps disambiguate subscript-in-superscript, nesting recedes into audio "distance" — the direct ancestor of StereoMath's pitch mapping.

Hop 4: Emacspeak — https://en.wikipedia.org/wiki/Emacspeak
- Hook type: mechanism/context (generalization of the idea)
- Hook: Raman applied AsTeR's audio formatting to the whole computing environment.
- Why followed: answers the seed's "who builds the accessibility layer?"
- Key findings: Emacspeak is an "audio desktop," not a screen reader; auditory icons for actions; built by a blind user via Emacs Lisp "advice"; in the Smithsonian's permanent collection. Slightly less surprising than Hop 3 → natural stop.

Saved hooks not followed:
- Amalric & Dehaene, "blind mathematicians activate the visual cortex when doing math" — from StereoMath refs — surprising neuroscience claim, but vault_novelty 0.747 (near duplicate of existing blind-math cluster).
- Nemeth Braille Code / BANA 2012 UEB dispute — from vault neighbors — vault already holds this (0.767).
- Sonification's signal-processing neighbors (HAL 9000 "Daisy Bell", voice-AI latency) — a distinct second bridge for a future chain.

Surprise: expected screen-reader-accessible Lean/Coq tooling built by the proof-assistant teams — found the accessibility layer is built almost entirely by independent and blind researchers, via sonification, outside those teams.
Surprise: expected AsTeR to read math as linear speech — found it used pitch and spatial "audio dimensions," with nested subexpressions sounding like "falling off into the distance."
Surprise: expected StereoMath's pitch-for-vertical-position mapping to be genuinely new — found Raman's AsTeR did essentially the same thing 30 years earlier.

post-worthy: maybe — a clean cross-time bridge (1994 ↔ 2024, both by musicians) with a real vault link, but needs the primary AsTeR audio quotes double-checked before publishing.