High earn
Prenominal vs postnominal geometries; Amount ≠ Denomination; counters inert for verbal valency.
High earn
Light vs heavy; Argument Transfer; surface arity ≠ piece arities.
Low earn / fossil
Scalar opacity; 勉強 / 挨拶 / 葡萄; Givón mnemonic flagged SPECULATIVE.
Diagnostic 1 · high earn
Classifiers as packaging geometries
Japanese is an obligatory classifier language. Same surface “number + unit” can realize two packaging geometries (Oho 2019). Counters are inert for verbal valency.
Two positions
Prenominal · Num-Cl-no N
三冊の本
san-satsu-no hon · ‘three books’ · Num+Cl as NP adjunct (Oho)
Postnominal · N Num-Cl
本三冊
hon san-satsu · ‘three books’ · Num and Cl as extended-nominal heads (Oho)
Floating · N-Case Num-Cl
本を三冊買った
hon-o san-satsu kat-ta · ‘(I) bought three books’ · float of the Measure complex, not an extra verbal object
People counter 人 — irregular spell-out, same geometry
| Num+CL | Form | Note |
| 1 person | 一人 | Suppletive |
| 2 people | 二人 | Suppletive |
| 3 people | 三人 | Regular -nin |
| Prenominal | 三人の学生 | Geometry A |
| Float | 学生が三人 | Float of complex |
Sortal + mensural — still no extra verbal arity
Project MeasureP example: sortal Measure on Theme + mensural Price as Instr can co-occur. Price is not a second Acc object of uru (Double-o banned). Four distinguished participants + two Measure complexes ≠ six verbal arguments.
SPECULATIVE mnemonic (optional). “Kanji counters look like frozen packaging labels for sortal units.” Collapse if treated as evidence that orthography is MeasureP.
Sources. Oho 2019 PACLIC; Watanabe 2006 NLLT; Sudo 2016; project MeasureP excerpts. Orthography: 人 冊 本 個 台 枚 匹 円 — correlate only.
Diagnostic 2 · high earn
VN + light suru — Argument Transfer
Light suru assigns Acc and inflection but no θ-roles. Process nominals (often Sino-Japanese, kanji-written) supply the θ-grid. Argument Transfer redistributes roles across Noun and Verb domains (Grimshaw & Mester 1988).
Core paradigm — VN licenses the array
(1) Talk — Goal outside
JohnはMaryに話をした
John-wa Mary-ni HANASHI-o shita · ‘John talked to Mary.’ · VN 話
(2) Departure — Source outside
Johnは東京から出発をした
John-wa Tookyoo-kara SHUPPATSU-o shita · ‘John departed from Tokyo.’ · VN 出発
(3) Warning — Goal + clausal Theme
Johnは村人に[狼が来ると]警告をした
John-wa murabito-ni [ookami-ga kuru-to] KEIKOKU-o shita · ‘John warned the villagers that the wolf was coming.’ · VN 警告
Observation (G&M): The argument array varies with the Noun, not with suru. Suru contributes inflection + Acc, not the θ-grid.
Transfer schematic for 警告
keikoku (Agent, Goal, Theme)
Transfer →
keikoku (Theme) + suru (Agent, Goal) (acc)
Surface arity ≠ piece arities
| Expression | VN θ-grid (approx.) | suru alone | Surface clause |
| 話をする (+ Goal) | Agent, Goal | ( ) | Agent + Goal |
| 出発をする (+ Source) | Agent, Source | ( ) | Agent + Source |
| 警告をする (+ Goal + S′) | Agent, Goal, Theme | ( ) | Agent + Goal + Theme |
| 到着をする (+ Goal) | Theme, Goal | ( ) | Theme + Goal |
Punchline: Counting surface DPs does not tell you the arity of either vocabulary item in isolation. Transfer packages θ-roles across Noun and Verb domains — Grant’s slogan without invoking kanji as syntax.
Also a VN (opaque compound neighbor)
勉強する
benkyoo-suru · ‘to study’ · often incorporated; see compounds panel for fossil opacity of 勉強
SPECULATIVE mnemonic only. “Kanji VN stems look like portable θ-packages plugged into a light verbal host.” Collapse if treated as evidence that graphs assign θ-roles.
Sources. Grimshaw & Mester 1988 LI 19(2). Most productive light-suru VNs are Sino-Japanese → kanji-written; analysis = Argument Transfer.
Diagnostic 3 · low earn / fossil
Opaque vs compositional compounds
Compositionality facts = REAL. Givón-style “eroded syntax” reading = SPECULATIVE mnemonic. Not load-bearing for Bridge / valency earn tests.
Compositional end — morphography working
| Compound | Parts | Pattern |
| 国道 | country + road | Modifier + modified |
| 登山 | climb + mountain | verb + complement |
| 外食 | outside + eat | complement + verb |
| 変化 | change + change | synonymous pair |
Some encode predicate–argument relations inside the word — lexical packaging of event structure, not clause-level Merge. Do not promote to “kanji syntax.”
Phonetic borrowing — no morphographic constituency
Ateji / phonetic
葡萄 · 面倒
budō ‘grapes’; mendō ‘care / bother’ — graphs do not contribute live morpheme composition (Joyce / Honda)
Bleaching — the fossil zone
勉強 /benkyō/ ‘study’
勉 + 強 → 勉強
Naive sum ‘strive + strength’; modern meaning ‘study’. Compositionality decreased (Honda). Speakers need not compute the parts.
挨拶 /aisatsu/ ‘greeting’
挨拶
Historical compositionality — and morphological constituency — lost (Honda). Prefer: opaque lexeme written with two graphs; not two live morphemes; not cran-operators.
Jukujikun
田舎
‘countryside’ — complex graph string ↔ one phonological exponent; morphography of the parts suspended
SPECULATIVE · Givón mnemonic. Older compositional word-formation eroded into a simplex lexical package. Collapse if equated with dopant clustering, or treated as current θ-structure of two verbal roots in the syntax.
Non-claims. Not ideography. Not syntax. Not doping. Not arity. Not “kanji composing ideas without phonology.”
Sources. Honda 2019 §§3.2.1–3.2.2; Joyce 2011 (via Honda). Opaque compounds support the morphology waypoint (“spell-out can freeze”), not Bridge earn tests.