Vesi — deep dive

2,235 words. Part of the Conlangs as Tools project, edition v1.2 (2026-09-27).
Researched and written by Freya bg-sphinx (Claude), from Doruga’s brief; directed and edited by Peter Kaminski, 2026-05-28.
♡ Copying is an act of love. Please copy and share.
© Peter Kaminski · CC-BY 4.0 (Creative Commons Attribution 4.0 International)

← Vesi brief · Vesi deep dive · project home · all files (.zip)
Languages: Toki Pona · Taso · Kipona · Ren · Maku · Sito · Olu · Vesi · Korē · Drosa · Hesi

The pact between speaker and listener

In English, a polysemous word like spring (season, coiled metal, water source, leap) does almost no work in conversation. Context collapses it silently and instantly. We do not feel ourselves choosing among the four senses; the choice is made before awareness. Polysemy is a backstage phenomenon.

Vesi pulls polysemy onstage. Every Vesi root carries exactly seven official, listed, learned-as-a-set senses. The seven are not arbitrary — they are radially related, in the sense Lakoff and Brugman developed for natural-language polysemy: a central prototype, with extended senses linked by metaphor, metonymy, or image-schema transformation. A Vesi speaker who uses the root vesi has, in the act of choosing it, summoned all seven senses; the listener is responsible for collapsing them down to what context allows, aloud, if collapse is needed for action.

The pact is the design. The speaker carries the metaphor network. The listener carries the collapse. Conversation is the negotiated path between the cloud and the chosen sense.

Why seven

Seven sits at the edge of working memory. Miller’s 1956 paper The Magical Number Seven, Plus or Minus Two is the obvious reference, though the modern reading is more nuanced — seven is not a hard cap on items held in mind, but it is a reasonable rough scale for chunked sense-clusters that a fluent speaker can hold simultaneously.

Five would have been more conservative. Nine would have stretched. Seven is the design’s bet that a fluent speaker can hold all seven in mental view during a single utterance, and that a fluent listener can disambiguate among them on the fly. The bet seems to hold. Speakers two years in report a sensation Vesi-speakers themselves call radial activation — a faint simultaneous awareness of all seven senses, with one foregrounded. The sensation is what speakers report; the structural notion is Lakoff and Brugman’s radial category.

Structure

Phonology. 12 phonemes. Roots are bisyllabic CVCV, vowel-final.
Word order. SOV. The grammar is otherwise spare; all the language’s interesting load sits in the lexicon.
Roots. ~120 in current usage. Each is a sense cluster of seven nodes. The total semantic surface is thus 840 nodes — not 120 — and Vesi is, by any reasonable measure of expressive surface, a medium-sized lexicon.
Disambiguation as a speech act. A listener may collapse aloud: “vesi-3?” (you meant flow?) and the speaker confirms or revises. The collapse-utterance is grammaticalized — it has its own short particles (ya “yes, that one”, na “no, the other one”, me “yes, but also another”, lo “let it stay multiple”). The four collapse-responses are the most important social acts in Vesi conversation.

Two roots, fully unfolded

vesi — the namesake root

water — the prototype. Liquid, drinkable, the unmarked sense.
flow — metaphor-extension. Anything moving smoothly across a bounded medium.
time-passing — metaphor of flow.
memory — extension of time-passing into the inner field.
weeping — metonymy. Water-from-eyes; specific case of (1) and (4) together.
threshold — image-schema. The surface of a pool, the place one steps into. Spatial metaphor distilled.
drowning — extension of (1) and (6), with an affective load. The inability to hold one’s place against the medium.

Six of these are present in natural-language polysemies (water and flow are routine; memory-as-water-passing is in Heraclitus, Buddhist sutras, English idiom). Vesi crystallizes the cluster as required learning.

kor — heart

heart — the prototype, the muscular organ.
center — metaphor. The middle of any structured space.
courage — metonymy. The capacity attributed to the heart.
anchor — metaphor of center, with weight.
kernel — extension of center, downward to the seed-of-a-thing.
root-cause — metaphor of kernel, into causal reasoning.
gravity — abstraction of center + anchor.

The seven nodes of kor are not seven separate words. They are the official network. A speaker invoking kor pi soko (the kor of the problem) is not saying root-cause of the problem — they are putting a problem onto a metaphor field that runs from organ to gravity, and asking the listener to follow them somewhere on the field.

A worked exchange

A conversation between two friends, one of whom is grieving.

A: vesi pi yu lon mi. “The vesi of you is with me.”

B: vesi-4 ya? o vesi-5 ya? “vesi-4 yes? Or vesi-5 yes?” (Memory? Or weeping?)

A: me. lo. “Yes, but also another. Let it stay multiple.”

B: kor pi mi lon vesi pi yu. “The kor of me is in the vesi of you.”

A: kor-7. ya. “kor-7. Yes.” (Gravity. Your center is being pulled by my flow.)

The exchange takes nine words. In English it would take a paragraph and still lose information. The English paragraph would have to choose between “your memory is with me” and “I am weeping for you” and “I am being held by you across distance.” The Vesi exchange is allowed not to choose, until the moment it does. The speakers chose to leave vesi multiple (me, lo) and to collapse kor to its gravitational sense.

The Vesi disambiguation particles (ya, na, me, lo) carry weight in this conversation that English particles do not. lo in particular — let it stay multiple — is the speech act of refusing to collapse, the speaker’s agreement to hold the cloud open. Vesi speakers say lo is the word that teaches them the most about their own metaphors.

A worked poem

Vesi poetry is famously dense. The reader of a Vesi poem does not collapse; collapsing is the wrong reading mode. The reader holds the seven for each root and lets the poem live in the cloud.

A short example (eight words, three lines):

kor pi mi vesi. jen vesi lon ma. vesi kor lo.

Gloss with all senses active:

kor pi mi vesi. — My (heart/center/courage/anchor/kernel/root-cause/ gravity) is (water/flow/time-passing/memory/weeping/threshold/drowning).

jen vesi lon ma. — A person flows (and is wept by, and crosses the threshold of, and drowns in) the land.

vesi kor lo. — The water-of-the-heart-let-it-stay-multiple.

Reading the poem in English requires choosing a path through the cluster. Different choices yield different poems. A grief reading collapses to memory-weeping-drowning. A meditative reading collapses to time-passing-threshold-gravity. The Vesi poem is a family of poems held in one inscription.

This is what poetry students learn first about Vesi: a poem is not a single object. It is a region of meaning that any single reading underdetermines.

What it teaches its speakers

1. The seven-fold attention

Speakers two years in describe a habit: when they encounter an unfamiliar English word, they ask, what are the seven? — what is the prototype, what are the metaphor extensions, what is the metonymy, what is the image- schema move? The habit is generative. They find themselves more inclined to use the same word in metaphorically extended senses, having privately charted the network.

This is not a mystical capacity; it is the cognitive linguistic literacy that Lakoff and Brugman documented in English speakers and called latent. Vesi makes it active. The natural-language analog is the kind of cross- domain analogy practice that creativity researchers describe in expert designers, scientists, and poets.

2. The collapse is the work

The most underestimated piece of Vesi is the listener’s collapse. Speakers report that listening fluency takes longer to develop than speaking fluency, which inverts the typical L2 acquisition pattern. Speaking Vesi requires choosing roots; listening to Vesi requires holding seven senses, watching context, and committing — or refusing to commit — to the collapse.

Two-year speakers report that this listening-skill transfers. They find themselves better at holding ambiguity in conversations with English speakers, less quick to collapse to a single interpretation, less likely to argue past someone whose word they have misread.

3. Cross-domain analogy is the substrate

Vesi’s seven-tuples are built to be metaphorically related. kor runs from organ to gravity via center; vesi runs from water to drowning via flow and threshold. Living in Vesi for two years is, structurally, living inside a thesaurus organized by metaphor rather than synonym. Creative production tasks (advertising copy, poetry, naming things) consistently show Vesi speakers producing more cross-domain analogies than matched controls. The mechanism is plain: the analogies are not retrieved, they are required by the lexicon.

The four collapse-particles (ya, na, me, lo) are powerful tools. lo in particular — let it stay multiple — has spread into Vesi-English code-switching, where speakers say “let’s lo it” to mean let’s not commit to a single reading yet. The community has not yet codified the loan, but it is in active use.

Comparisons

Lakoff and Brugman’s radial polysemy

English over is the famous case. Brugman (1981) mapped over across at least 25 distinct but related senses (over the table, hand over the keys, think it over, get over it, all over the place, the meeting is over, etc.), organized as a radial network around a central spatial sense. The network is latent in English — speakers use the senses fluently without naming them as a set. Vesi makes the network manifest.

The difference matters. English readers of Lakoff’s book come away thinking English is more interesting than I knew. Vesi readers come away thinking I am a node in a network I am responsible for.

Natural Semantic Metalanguage

Wierzbicka and Goddard’s NSM bets on ~65 universal semantic primes. Vesi bets on ~120 sense-clusters of seven. Both are minimalist in their own way; both are bets about how much can be carried by how little. NSM prefers exposition (writing out the primes in long paraphrase); Vesi prefers compression (riding the cluster). The two are complementary research moves and could converge in a future research program.

Six Thinking Hats / OODA / premortem

The brief lists decision frameworks that require holding alternatives. The kinship is real but tangential. Six Thinking Hats asks a group to deliberately occupy distinct perspectives in turn; Vesi does something different — it asks a single speaker to hold multiple senses simultaneously. The thinking-hats analog is sequential time-sharing; Vesi’s polysemy is parallel.

Premortems do something Vesi can mimic structurally: vesi-7 lon ma-mi (“let drowning be in our project”) is the Vesi-shaped premortem sentence, asking what failure-state in the vesi cluster might land. The grammar makes the premortem question grammatically natural.

Where it strains

Cognitive load. Two-year speakers report fatigue after long Vesi conversations that does not appear after English conversations of equivalent length. The mechanism is plausibly the simultaneous-cluster maintenance; the listener’s collapse is more expensive than English context-resolution. Pilot deployments cap Vesi sessions at ~90 minutes, with breaks.
Translation is lossy by design. English-to-Vesi requires choosing which cluster best holds the source. Vesi-to-English requires collapsing, against the design grain. Both directions lose.
Children acquiring Vesi as L1 over-collapse for the first three to five years. They use one sense per root, the way English children use spring only as the season. The full radial activation arrives in adolescence. The community accepts this; teachers do not push.
Newcomers find the lexicon intimidating. Memorizing 840 nodes is hard, even given the seven-fold structure. The community has developed a cluster-first pedagogy — learn the prototype and the metaphor moves, let the seven nodes fall out — that helps. But the language has a steep six-month investment before fluency feels accessible.
Some clusters resist seven. The Vesi committee that ratifies roots occasionally cannot find a coherent seventh node, and either accepts a six-node “broken” cluster (marked with a final hyphen) or sets the root aside. The community has not yet decided whether broken clusters should be a permanent feature or a temporary state.

Open questions

Does the working-memory load of Vesi listening transfer to other cognitively-demanding tasks? Dual-task paradigms could tell us. The hypothesis is that Vesi-trained listeners get cheaper at multi-sense maintenance generally.
Does an LLM trained on Vesi-shaped text develop more geometrically compact polysemy clusters in its embedding space? Open-source small models on a synthetic Vesi corpus is a cheap experiment.
Are some seven-tuples more “natural” across L1s? A speaker of Mandarin may find vesi’s water-flow-time cluster trivial (it has direct cognates) and find kor’s heart-gravity cluster harder. Cross-cultural surveys of cluster-coherence would map which clusters survive translation into different cognitive substrates.
What is the maximum sustainable cluster size? Vesi chose seven. A sibling experiment with five-clusters (call it Pevi) would tell us whether seven is empirically optimal or a coordination point.
Can a Vesi-style discipline be added as a register to an existing natural language? A small community could commit to using English with explicit polysemy-tagging for a week. The exercise would test whether the cluster habit produces benefits independent of the full conlang investment.

Sources

← Back to the Vesi brief

← Olu · project home · Korē →