01 Concept
The whole project hangs from one inversion: position is the sound, not a label for it. On this map, where you stand is what you hear — the coordinates are acoustic, so geometry and sound are the same two axes seen two ways. A held vowel is a single steady spot in the plane; a word is a journey across it.
The name is literal: this is a space you navigate by ear, where the IPA chart stops being a diagram of sound and becomes the instrument.
02 The scale model
The organizing idea is not a flat “word space” but movement between three nested scales. Words are not the primitive — units are, and scale itself is the structure.
Feature
The continuous substrate — the axes themselves (F1, F2), the acoustic parameters being varied.
Unit
A named position in the substrate: a point (vowel) or a rim landmark (consonant) — a single steady phone.
Word
An ordered visit to several units: a path or polyline — a glide or sequence through the space.
Moving between scales is natural because the relationship is compositional, not arbitrary: zoom down and a word decomposes into its unit-stops; zoom up and units chain into a word. Units come first — the vowel field is the one place geometry literally equals sound with nothing muddying it, and learning the terrain by ear as units lets words then read as motion through a space you already know.
03 The space
The field — vowels
A continuous interior: vertical is open↔close (F1, the jaw), horizontal is back↔front (F2, the tongue). Drag across it and a formant synthesizer bends the vowel in real time — the empty middle is every vowel between the named ones.
The rim — consonants
A place-by-manner lattice around the edge — stops, fricatives, nasals — discrete landmarks, not terrain. A fluid interior and a gridded coastline.
Words as paths
A polyline threading vowel-points and touching consonant-landmarks in order. Diphthongs are literally line segments — /haʊ/ (“how”) sweeps up and to the right into the back vowel: the project’s ↗.
The gaps
Load the lexicon and the field becomes a constellation — and most of it stays dark. Those are the pronounceable words nobody says: the catalogue of the possible, the actual a thin scatter of lit points inside it.
04 Interaction
01
Roam
Move through the field and play it like an instrument — no words required.
02
Summon
Call up a word and watch its path draw itself and sound out.
03
Morph
Pick two words and interpolate the path between them, hearing one bend into the other through all the nonsense in between.
Neighborhoods. Snap to a word and its nearest neighbors are its minimal pairs — cat, cot, cut, cab, can, each one phoneme away — so you hop the dense lattice of the lexicon node by node.
The signature demo — the bilingual overlay. Drop Spanish’s five vowels onto the plane (a clean, wide pentagon), then bring up English’s eleven-ish, crowded and overlapping in the same space. You hear exactly where they collide — why beat and bit fold into a single vowel for a Spanish ear. The thing that makes someone feel the concept rather than nod at it.
05 Synthesis & data
Sound is source-filter formant synthesis through the Web Audio API: a glottal sawtooth source (~118 Hz) feeds three parallel band-pass filters tuned to the formants, and the cursor’s x/y drives F1 and F2 in real time, gliding smoothly so the vowel morphs as you move. Monophonic by design — one position, one sound, honest to the thesis.
F1 · 250–850 Hz (open↔close)F2 · 700–2400 Hz (back↔front)CMU dictionary → ARPAbetHillenbrand · Peterson & Barney F1/F2
06 Roadmap
v1
Playable vowel field — shipped
Press-and-drag field with real-time formant synthesis; named English + Spanish units; nearest-unit readout; the Units ⇄ Words scale toggle; word-path playback.
v2
Consonants, neighborhoods, overlay, morph
The consonant rim as landmarks; minimal-pair neighborhoods over the CMU dictionary; the full bilingual overlay with collision highlighting; morph-between-words.
v3
Generative & multi-scale
Generative phonotactically-legal walks; the unrealized-word gaps as a deliberate view; scale down to individual features; scale up to phrases and prosody.