Bubnov (2026): Not all coexpressions are syncretisms #
@cite{bubnov-2026}
Limiting Nanosyntax. Glossa 11(1), 1–15.
Argues against the universal applicability of nanosyntactic feature decomposition to coexpression phenomena, using indefinite pronouns as a test case. Key claims:
@cite{dekier-2021}'s nanosyntactic analysis of indefinites (a containment hierarchy F₁ ⊂ F₂ ⊂ F₃) fails empirically: no morphological containment is attested in indefinite paradigms.
The semantic account of @cite{degano-aloni-2025}, based on variation and constancy from team semantics (@cite{hodges-1997}, @cite{vaananen-2007}), provides a better typology: 7 indefinite types from Boolean combinations of
var(y,x)anddep(y,x).The semantic account correctly predicts:
- Which indefinite type is unattested (type vi: SK+NS)
- Bidirectional diachronic change, unlike nanosyntax which predicts only unidirectional change
- The relative frequency of indefinite types (conjunctive requirements = rarer)
The broader implication: some coexpression patterns arise from semantic underspecification at LF, not structural containment at PF.
Connection to linglib #
DependenceLogic:variationandconstancypredicates formalize D&A'svar(y,x)anddep(y,x).type_vi_contradictoryderives the gap.Nanosyntax.Core:spelloutandabaViolationdemonstrate the negative result — nanosyntax predicts containment that indefinites lack.Typology.Indefinite:IndefiniteEntry(consensus function-coverage- morphological-basis data) and
classifyTriplefor syncretism patterns.
- morphological-basis data) and
Phenomena.Indefinites.Studies.DeganoAloni2025:DATypeandsurfaceDAType/consistentWithprojections from entries to D&A types.Fragments.{Russian,English,German,Latin,Yakut,Kannada}.Indefinites: per-language indefinite paradigms witnessing the typology.Fragments.German.ModalIndefinites: bridge connecting D&A's typology to Alonso-Ovalle & Royer's modal-indefinite typology for irgend-.
@cite{bubnov-2026}'s key objection: nanosyntax predicts the ABC pattern
should show morphological containment. This is NEVER attested for
indefinites. The Russian forms are surface-level prefixed/suffixed to
interrogative bases (kto-nibud', kto-to, koe-kto); the indefinite
morphemes themselves (-nibud', -to, koe-) share no material.
In case morphology, containment IS attested. In indefinites, NOT. This asymmetry supports @cite{bubnov-2026}'s claim that nanosyntax is the right tool for case but not for indefinites.
Type (vi) (SK+NS) is predicted unattested because its semantic
requirements are contradictory. dep(∅,x) requires x constant across
all assignments; var(v,x) requires x to vary among v-agreeing
assignments. By variation_monotone, this lifts to var(∅,x),
contradicting dep(∅,x) via constancy_excludes_variation.
Profile-level verification: type (vi)'s D&A profile is the
non-contiguous {SK, NS} — the contradiction surfaces structurally
as a non-Haspelmath-adjacent function set.
Type (i) unmarked: no restriction → all three SK/SU/NS functions.
Type (iii) non-specific: var(v,x) → only NS.
Type (iv) epistemic: var(∅,x) → SU + NS.
Type (v) specific known: dep(∅,x) → only SK.
Type (vii) specific unknown: dep(v,x) ∧ var(∅,x) → only SU.
Rare conjunctive type; Kannada yāru-oo is canonical.
Weakening from non-specific (iii) to epistemic (iv): dropping the within-world parameter makes variation global. The epistemic profile properly contains the non-specific profile, so the form gains SU while keeping NS.
Weakening from epistemic (iv) to unmarked (i): dropping the variation restriction. Unmarked profile properly contains epistemic.
Weakening from specific known (v) to specific (ii): broadening from cross-world constancy to within-world constancy. Specific profile properly contains specific-known.
The fundamental monotonicity underlying diachronic weakening:
variation w.r.t. a finer parameter (within-world v) implies variation
w.r.t. a coarser parameter (across-worlds ∅). This is
variation_monotone from team semantics.
Nanosyntax + Dekier: losing entry A (rank 0, NS-only) makes entry B (rank 1, SU) cover NS too via Superset Principle. Predicts SU → epistemic (AB → BB), but NOT NS → epistemic.
Equations
- Phenomena.Indefinites.Studies.Bubnov2026.dekierInitial = [{ rank := 0, exponent := "A" }, { rank := 1, exponent := "B" }]
Instances For
Equations
- Phenomena.Indefinites.Studies.Bubnov2026.dekierAfterLoss = [{ rank := 1, exponent := "B" }]
Instances For
kto-nibud' surface-classifies as type iii non-specific (actual = profile).
kto-to is classified as epistemic (var(∅,x)) per @cite{bubnov-2026} §7,
BUT its actual distribution (SU only) is narrower than the epistemic
profile (SU + NS) because -nibud' (type iii) blocks it for NS.
Surface classifier returns type vii (specificUnknown) — the type whose
profile exactly matches {SU}. Bubnov's manual type-iv classification
is the consistentWith claim: actual ⊊ profile. The two layers
are simultaneously asserted here.
koe-kto surface-classifies as type v specific-known.
Latin aliquis surface-classifies as type iv epistemic. Unlike Russian -to, no competition: quidam only covers SK, so aliquis fills both SU + NS unblocked, matching the epistemic profile exactly.
Yakut kim ere surface-classifies as type ii specific (SK + SU).
Kannada yāru-oo is the canonical type vii specific unknown:
dep(v,x) ∧ var(∅,x), profile {SU}.
English some- surface-classifies as type i unmarked (all 3 functions).
Yakut kim eme surface-classifies as type iii non-specific.
Latin quidam surface-classifies as type v specific-known.
Kannada yāru-aadaruu surface-classifies as type iii non-specific.
German irgend- is classified as type iv epistemic in D&A's typology AND as not-at-issue epistemic in Alonso-Ovalle & Royer's modal-indefinite typology. Compatible perspectives: the modal analysis describes WHAT irgend- does (domain widening); the team-semantic analysis describes its DISTRIBUTIONAL restriction (varying across epistemic alternatives).
@cite{bubnov-2026} §6: German irgend- instantiates the diachronic path (iii) → (iv) (@cite{aloni-port-2015}).
Russian kto-to: within-D&A interpretive disagreement. Both encodings
apply D&A 2025's classification, but reach different functions sets:
the Polarity-side file (Polarity/Studies/Haspelmath1997.lean:226)
encodes the bare D&A type-iv profile {SU, irrealis}; the Fragment
(Fragments/Slavic/Russian/Indefinites.lean:54) encodes Bubnov 2026
§7's narrowing argument {SU} only, since -nibud' paradigmatically
blocks -to from the irrealis function.
The disagreement is not cross-framework but interpretive within D&A: apply the bare type-iv profile, or apply it net of paradigmatic competition. Both are defensible readings of D&A's framework.
English some-: genuine cross-framework disagreement. The Fragment
(Fragments/English/Indefinites.lean:27) encodes some- as D&A type-i
unmarked, profile {SK, SU, irrealis} — some- covers everything D&A's
type-i permits. The Polarity-side file
(Polarity/Studies/Haspelmath1997.lean:175) encodes some- with
Haspelmath's competition-among-forms approach: some- covers {SK, SU}
only because any- takes the irrealis-through-indirectNeg territory.
This is the strongest of the three §11 theorems: it contrasts D&A's "one type per form, full coverage" methodology against Haspelmath's "one form per function, paradigm-internal division" methodology. The contrast is methodological, not just data-interpretation.
German irgend-/irgendein-: cross-framework consistency check.
Fragments/German/Indefinites.lean:33's irgendEntry.functions ({SU,
irrealis}, matching D&A type-iv epistemic) and Fragments/German/ ModalIndefinites.lean:27's irgendeinEntry.flavors (which includes
.epistemic, matching Aloni-BSML's epistemic-modal classification) line
up: both attribute epistemic semantics to the same morphological root.
This is a consistency check across two formalizations, not necessarily a substantive cross-framework agreement: both Fragment authors knew irgend- is canonically epistemic and encoded it accordingly. The theorem is a regression test for that consistency — if either Fragment changes, this breaks. Real value: catches drift between the two Fragments.