Conversations
The corpus includes multi-turn exchanges — not just isolated sentences. These conversations are where Tonesu stops being a reference grammar and starts being a thing people do.
Each walkthrough below traces the pragmatic choices: why a speaker picks this epistemic verb, what ze is pointing at, where register shifts happen, how disagreement gets structured.
C003 — Travel plans (casual social)
Scene: Two non-specialists in an everyday context. A asks about travel plans. Four turns, no conflict — but the grammar still has to work.
What this tests: First polar question in the corpus, echo-confirmation strategy, ta-ti-be proximate-future time reference, purpose clause in casual register.
The exchange
A1 — Are you going to the city soon?
| Element | Parse |
|---|---|
to-si — |
polar question marker (fronted, pre-dash) |
la-tu |
agent: you |
ki |
motion / change |
pa-li-pu |
city — place (pa) + person (li) + group (pu) |
ta-ti-be |
soon — time (ta-ti) + approaching (be) |
The to-si — position is what makes this a yes/no question rather than a content question. If to-si were inside the clause (occupying an argument slot), it would ask what or who.
B1 — I'm going to the city soon.
This is an echo confirmation. B doesn't say ru (yes) — B restates A's proposition from their own perspective, swapping la-tu for la-mi. The effect: agreement by demonstration rather than assertion.
Why not just ru? Echo confirmation carries more social weight in informal speech — it says "I'm not just acknowledging your question, I'm inhabiting the claim."
A2 — A new water facility is there now.
| Element | Parse |
|---|---|
lo-ma-su |
patient: water-structure (facility) |
be-now |
currently growing / newly built |
pa-ze |
location: there (place + referent) |
A shifts to offering context — the reason for asking. Note lo-ma-su (patient) rather than la-ma-su (agent): the facility is a thing being described, not an actor.
B2 — I intend to go see it.
| Element | Parse |
|---|---|
la-mi |
agent: I |
wi [...] |
purpose frame |
ka-se |
action: perceive / go see |
lo-ze |
patient: it (back-reference to the facility) |
The wi purpose clause wraps B's intention. ze picks up ma-su (the facility) as most-recent salient referent. No ambiguity — there's only one candidate.
What this conversation shows
- Polar questions use
to-si —(fronted, pre-dash). Content questions useto-siin-slot. - Echo confirmation is a pragmatically richer alternative to bare
ru. zeworks cleanly when the referent set is small and unambiguous.- Four turns of ordinary casual speech — no epistemic machinery, no conflict — and the grammar handles it without strain.
C005 — Theological dispute (epistemic conflict)
Scene: A ra-ki-li (pilot) and a to-su-li (doctrine-keeper / archivist) dispute whether the community's guardian machine has zo-to (soul / identity-pattern). The pilot recently experienced ne-ra-ki (attunement) with the machine. This is a fight.
What this tests: Epistemic verb escalation (si → to → to-fe-ka), ze as propositional back-reference, citation of institutional doctrine, the moment a conversation crosses from disagreement into accusation.
The exchange
A1 — I hold as hypothesis: the machine has a soul.
A opens with si — the weakest epistemic stance. "I hypothesize." Not claiming certainty, not claiming evidence. This is a careful opening: the pilot knows this is contested territory.
B1 — I hold as established: what you experienced is raw perception.
B responds with to — the strongest epistemic verb. "I hold as established." And the content of B's claim: lo-ze se. ze here is propositional — it refers back to A's entire claim, not to a person. se is raw perception (the lowest epistemic category).
The move: B doesn't just disagree — B downgrades A's evidence. "What you experienced isn't even knowledge. It's sensation."
A2 — The canonical doctrine holds: the machine has a soul.
A escalates by citing authority. la-to-re-su = agent: the canonical doctrine system. ko = containment particle (the doctrine holds within itself this claim). A is no longer speaking from personal experience — A is invoking the institution.
B2 — I hold as established: the doctrine has not crossed the threshold.
B's counter: the doctrine itself hasn't reached to-fe (epistemic threshold). no-to-fe = has not crossed the boundary of established knowledge. ta-ti-now = as of right now.
This is devastating: B isn't saying the doctrine is wrong. B is saying the doctrine never rose to the level where it could be right or wrong.
A3 — I hold as established: what you said is deliberate epistemic fraud.
A fires the nuclear option. to-fe-ka (W060) = deliberate epistemic fraud — intentionally misrepresenting the status of a knowledge-claim. ze is again propositional — it points back to B's entire challenge.
B3 — I hold as not-established: that is not fraud. Submit this to the standards body.
B does two things:
no-to— negated epistemic verb. "I hold as not-established." B doesn't say A is wrong; B says A hasn't met the standard to make that accusation.- Procedural redirect:
la-tu ka-si lo-to-fe-su= "submit this toto-fe-su" (the epistemic standards body). B moves the dispute out of interpersonal argument and into institutional process.
What this conversation shows
- Epistemic escalation has a grammar:
si(hypothesis) →to(established) →to-fe-ka(accusation of fraud). zeshifts referent type without morphological marking. When the predicate isse(perception),zereads as propositional. When the agent isla-ze, it reads as a person. Context does the work.- Institutional citation (
la-to-re-su ko [...]) is a specific move — it shifts authority from the speaker to the doctrine. no-tovsno: B could have saidno(flat denial). Instead,no-toframes the denial as epistemic — "the accusation hasn't been established." More precise, more damaging.- The conversation goes from careful hypothesis → evidence downgrading → institutional citation → threshold challenge → accusation → procedural redirect. Six turns, five distinct pragmatic strategies.
C008 — Arbitration hearing (ze collision)
Scene: A to-fe-li (epistemic arbiter, B) conducts a compliance hearing. A ra-ki-li (pilot, A) testifies that a to-su-li (archivist, absent third party) suppressed a signal record. The exchange manufactures four distinct lo-ze uses — and then shows what happens when it breaks.
What this tests: ze as person vs propositional back-reference, predicate-type disambiguation, the moment where ze becomes genuinely ambiguous, and the canonical repair strategy.
The exchange
A1 — The archivist suppressed the record. (person-reading)
la-ze = agent: the archivist (most-recent salient person). de = decay / suppression. lo-si-de = patient: the signal record. ta-ti-de = in the past.
Here ze is clearly a person — it's in agent position with an action verb.
B1 — I hypothesize: ze did not suppress the record. (competing proposition enters)
B introduces a counter-proposition. Now we have two things ze could refer to: (1) the archivist (person), (2) the proposition that the archivist suppressed the record.
Inside B's embedded clause, la-ze is still clearly the person (agent position + action verb). But the existence of the embedded proposition creates a second candidate referent for any subsequent ze.
A2 — I hold as established: that claim is raw perception. (propositional ze, forced by predicate)
Here lo-ze can't be the person — a person isn't "raw perception." The predicate se forces a propositional reading: ze = B's hypothesis. This is predicate-type disambiguation — the grammar doesn't mark the shift, but the semantics make it unambiguous.
B2 — I hold as established: [ze / that claim] is deliberate fraud. (genuinely ambiguous)
Now the system breaks. to-fe-ka (epistemic fraud) can apply to either a person ("ze is a fraud") or a proposition ("that claim is fraudulent"). Both readings are grammatically valid. The arbiter may intend either one. This is a real parse failure — a listener cannot resolve ze without extralinguistic context.
A3 — Ze committed the fraud — not me. (disambiguation by restatement)
A's repair strategy: restate with an explicit NP in a structure that forces the person reading. la-ze in agent position + intra-clause contrast (no la-mi = "not me") pins ze to the archivist. The ambiguity from B2 is resolved retroactively.
B3 — I note: ze suppressed a record. Submit this to the standards body. (arbiter's resolution)
The arbiter drops back to se (perception) instead of to (established) — they've witnessed testimony, not established fact. And they use la-ze with an explicit action verb (de), restoring the person reading cleanly.
What this conversation shows
zehas a failure mode. When the predicate type admits both person and proposition readings, ambiguity is real, not resolvable by grammar alone.- Predicate-type disambiguation works when it works —
seforces propositional, agent-position + action verb forces person. Butto-fe-kais ambiguous between the two. - Restatement is the canonical repair. When
zegoes ambiguous, speakers don't add morphology — they re-say what they mean with explicit NPs. - Institutional register matters. After the ambiguity event, the arbiter prefers explicit NPs over
zefor the rest of the formal record. This is documented practice: in proceedings,zeambiguity triggers a reset to full NPs. - The language chose not to fix this. A morphological distinction (person-
zevs propositional-ze) would prevent the ambiguity. Tonesu doesn't add it. The design choice: discourse repair is cheaper than morphological overhead, and the failure mode is productive — it surfaces exactly the cases where meaning is genuinely contested.
What conversations reveal
Single sentences prove that grammar rules exist. Conversations prove they work — or show exactly where they don't.
The three exchanges above demonstrate:
- C003: The grammar handles everyday casual speech with no strain — questions, confirmations, back-references all compose cleanly.
- C005: Epistemic conflict has a grammar. Each move in the argument corresponds to a specific particle choice. The escalation from
sitotototo-fe-kais visible in the syntax. - C008:
zeworks until it doesn't. The language has a diagnosed failure mode, a known repair strategy, and a principled reason for not fixing it.
A language that only showed its successes would be suspicious. These conversations include the breakdowns.