Sadakane & Koizumi 1995 @cite{sadakane-koizumi-1995} @cite{martin-1975} #

On the nature of the "dative" particle ni in Japanese. Linguistics 33(1):5–33.

Headline claim #

The apparent ambiguity of Japanese particle ni between case marker and postposition is illusory (Conclusion, p. 23). What looks like a single particle with mixed properties is actually four homophonous lexemes:

Dative case marker ni (Martin (1975) categories A, O1)
Postposition ni (categories B, C1–C3, E, F, G, H1–H3, I, K, L1, M, O2, R, T, U)
ni of ni-insertion (J1, J2, L2; Takezawa 1987's Japanese analogue of English of-insertion — a last-resort default for caseless arguments)
Copula ni (P1, P2, Q, S, V; a form of the copula da/de aru)

Plus an ambiguous bucket (D, N1, N2) where speakers disagree on whether ni is a case marker or a postposition; encoded as Classification.classify _ = none.

S&K SUPPORT @cite{kuno-1987}'s and Miyagawa (1989)'s case-marker / postposition dichotomy — once the four homophones are split apart, the remainder respects the binary partition. This is in contrast to the "third-type" view (a single ni with both case-marker and postposition properties) widely held in Japanese linguistics.

Three operational tests (§2) #

S&K distinguish the four types by three syntactic diagnostics, summarised in tables 14, 27, 29, and 32:

Type	Floating NQ	Cleft + particle	Cleft − particle
Dative case marker	OK	*/??	OK
Postposition	*	OK	*/?/OK
Ambiguous (D,N1,N2)	OK	OK	OK
ni-insertion	*	*/??	OK
Copula ni	*/N.A.	*/??	*/??

The diagnostic acceptability scores are encoded in Features.Acceptability (the project canon): */?? reduces to unacceptable; */?/OK to variable (genuine speaker variation); */N.A. to unacceptable (per S&K fn. 10, the test fails for an independent non-referentiality reason PLUS a second structural reason — both yield *).

Affectedness criterion (§4, p. 18) #

"The case marker ni is attached to an NP whose referent is relatively more affected by the action denoted by the verb (predicate/sentence), and the postposition ni is attached to an NP whose referent is less affected." (p. 18)

Hierarchy (figure 45, p. 22) carries TWO orthogonal dimensions: a 4-rank position scale (NP-in-PP < dative NP < upper accusative NP < lower accusative NP) and an AffectedKind distinction (phenomenally vs. structurally affected). Examples 42–43: Tom-o korosita "killed Tom" structurally affects Tom (he ceases to exist as such); Bill-o hometa "praised Bill" only phenomenally affects Bill's psychological state.

Acquisition prediction (§5, pp. 23–24) #

If S&K's analysis is correct (case-marker and postposition ni are distinct lexemes), Japanese-learning children should acquire them independently. Morii (1993) confirms: case-marker ni (categories A, O1) is acquired between 2;0 and 2;11; postposition ni (categories B–U) is acquired only after 3;0.

Heine 2009 grammaticalization #

The four S&K classifications align partially with Heine's case grammaticalization cline (Core.CaseGramStage: lexical → adposition → caseAffix → lost). Both case-marker ni and postposition ni are at .adposition stage in modern Japanese (morphologically free), but case-marker ni is more grammaticalized within that stage (no inherent meaning, omissible in casual speech). The cline doesn't capture intra-adposition gradience; the projection Classification.gramStage is correspondingly coarse.

Layered grounding to linglib #

Diagnostic acceptability scores use Features.Acceptability (the project canon), not a per-paper Grammaticality enum.
Classification.marantz aligns S&K's 4-way with @cite{baker-2015}'s Syntax.Case.CaseSource from Theories.Syntax.Case.Dependent. The map is partial: copula ni lies outside Marantz's case-assignment domain. Note: .niInsertion → .unmarked (Marantz/Schütze's "default-case" fallback) rather than .agree — Takezawa's salvage operation is what happens when Agree fails, not Agree itself.
MartinCategory.InFragmentNi is derived as a Finset intersection of the per-category fragmentCases footprint with the Fragment's Fragments.Japanese.Case.ni.cases = {.dat, .loc, .all, .Tem}. The conflation theorem fragment_ni_predicts_inconsistent_signatures derives the Tsujimura/Fragment vs. S&K granularity disagreement as a real claim about set intersection plus diagnostic-signature mismatch.

Dialect parameter #

S&K's footnote 9 (p. 30) flags that judgments throughout the paper are from the innovating dialect (per @cite{kuno-1987}, Miyagawa 1989). The conservative dialect (Shibatani 1977) gives different judgments for some categories (notably K, Ohaio Ginkoo-ni 'work for Ohio Bank'). This Studies file's Classification.signature reflects the innovating dialect; a future Phenomena/Case/Studies/Shibatani1977.lean could formalise the conservative judgments and surface where they diverge. The Dialect enum is intentionally NOT introduced here (was dead code in the previous version) — it earns its keep when the conservative file lands.

Korean parallel (Sells 1995, not yet in linglib) #

Sells (1995, Journal of East Asian Linguistics) documents the parallel case-particle/postposition split for Korean -i/-eseo but does not engage S&K's homophony move for Korean equivalents of ni. This gap cannot be Lean-formalised until Fragments/Korean/Case.lean adopts Pattern B (rich marker structure); currently it's a Finset Core.Case stipulation only. Documented here as future work.

Sadakane & Koizumi 1995 @cite{sadakane-koizumi-1995} @cite{martin-1975} #

Headline claim #

Three operational tests (§2) #

Affectedness criterion (§4, p. 18) #

Acquisition prediction (§5, pp. 23–24) #

Heine 2009 grammaticalization #

Layered grounding to linglib #

Dialect parameter #

Korean parallel (Sells 1995, not yet in linglib) #

§1 Classification — S&K's four homophonous ni lexemes #

§2 Martin (1975) usage categories — 31 letter codes #

§3 Operational tests + diagnostic signature #

§4 Cardinality theorems — Finset.card decomposition #

§5 Conflation: Fragment's ni collapses S&K types #

§6 Alignment with Marantz dependent case #

§7 Affectedness hierarchy (§4, figure 45) #

§8 Acquisition prediction (Morii 1993, cited at §5 p. 23–24) #

§9 Heine grammaticalization stage projection #

§5 Conflation: Fragment's `ni` collapses S&K types #