Sande, Clem & Dąbkowski (2026): Discontinuous vowel harmony in Guébie #
@cite{sande-clem-dabkowski-2026}
@cite{sande-clem-dabkowski-2026} introduce the phenomenon of discontinuous harmony: a long-distance ATR-harmony pattern in Guébie (a Kru language of Côte d'Ivoire) where the trigger and target are separated by intervening non-transparent (harmony-eligible) segments. The pattern arises in particle-verb focus-fronting constructions; the analysis is local harmony followed by syntactic movement that pulls the harmonized element away from its trigger.
Empirical core (the table at §6.1, (44)) #
| word order | particle-verb harmony? | verb spelled out in vP? |
|---|---|---|
SVOPart | × | × |
SAuxOPartV | ✓ | ✓ |
PartSVO | × | × |
PartSAuxOV | ✓ | ✓ |
Harmony applies iff V is spelled out inside vP — independent of the
surface adjacency between V and the particle. In PartSAuxOV,
predicate fronting separates them on the surface, but V was inside
vP at the moment of vP-spell-out (when local harmony applied).
The analysis #
Predicate fronting is remnant-VP movement (extending @cite{koopman-1997}'s Vata/Nweh analysis). The particle is the only overt element left in the remnant after V has moved through v to T and the object has shifted out (where applicable).
vP and CP are spell-out phases. Following the cyclic spell-out of @cite{newell-2008} / @cite{chomsky-2001}, vP is spelled out upon Merge of C; CP is spelled out at the end.
Local harmony in vP. ATR harmony is a vP-domain cophonology (per @cite{sande-jenks-inkelas-2020}'s Cophonologies-by-Phrase architecture, formalized in
CophonologyByPhrase.lean): when V and Part are both spelled out in vP, harmony applies; when only Part is in the vP-spell-out, no trigger is present and Part surfaces with its default −ATR value.Frozen ATR survives later movement. Once the particle has harmonized to V's ATR value at vP-spell-out, the value persists through subsequent A′-fronting of the remnant VP into Spec,CP (Cyclic-Linearization-style preservation —
Minimalist.Linearization.frozenValue).Strict PIC is rejected. For step 4 to work, already-spelled-out material must remain accessible to later syntactic operations (here, A′-movement of the remnant containing the particle). SCD 2026 §6.2 rejects both PIC₁ (@cite{chomsky-2000}) and PIC₂ (@cite{chomsky-2001}) in favor of a Cyclic-Linearization-bounded regime —
PICStrength.linearizationBoundinPhase.lean.
What this study file establishes #
For each of the four word orders, the file:
- proves the derivation is consistently linearizable
(
Minimalist.Linearization.SpelloutAndCheck), - computes the predicted harmony status (true iff V is in vP-content),
- decides the (44) table by
decide, - registers Guébie predicate fronting as a positive instance of
HarizanovGribanova2019.Amalgamation.VerbDoublingIsSyntactic(refactored from a universal axiom in light of this paper — seeHarizanovGribanova2019Amalgamation.lean).
The Wolof relative-clause parallel (§7) is added as a sister construction with the same architectural shape; the Atchan nasal-harmony case (§7) is recorded as an open question with sparse data — Russell, p.c.
Verified preliminaries (rejected alternatives) #
@cite{sande-clem-dabkowski-2026} §5 argues that purely phonological
theories of harmony predict strict locality and so cannot derive the
discontinuous pattern. The file does not formally simulate each
rejected theory (autosegmental @cite{clements-1985}, gestural
@cite{gafos-1998}, ABC @cite{rose-walker-2004}), but cross-references
their formalizations elsewhere in linglib —
Phenomena/Phonology/Studies/RoseWalker2004.lean,
Phenomena/Phonology/Studies/Hansson2010.lean,
Phenomena/Phonology/Studies/Sagey1986.lean — and notes how each
predicts the observed locality bound that SCD 2026 demonstrates is
empirically violated.
Architectural notes #
Substrate consumed (additions landed in this same overhaul):
Theories/Phonology/OptimalityTheory/CophonologyByPhrase.lean— @cite{sande-jenks-inkelas-2020}'s phrasal extension of @cite{sande-jenks-2017}'s VI-anchored cophonology.Theories/Syntax/Minimalist/Movement/Remnant.lean— @cite{koopman-1997} predicate-cleft remnant-XP movement.Theories/Syntax/Minimalist/Phase.lean— addedPICStrength.linearizationBoundandadmitsExtraction.Theories/Syntax/Minimalist/Linearization/Cyclic.lean— addedFrozenFeature/frozenValuefor cross-phase feature preservation.Phenomena/WordOrder/Studies/HarizanovGribanova2019Amalgamation.lean— refactoredaxiom verb_doubling_implies_syntacticto a per-constructionProp, motivated by SCD 2026's stance that the universal version is too strong (Landau 2006 Hebrew counterexample).
Per-language Guébie data lives entirely in this file (no
Fragments/Guebie/), per CLAUDE.md "per-language paper-specific data
lives in Studies, not Fragments".
Guébie vowel inventory (@cite{sande-clem-dabkowski-2026} (1)) #
+ATR: ə e i o u
-ATR: a ɛ ɪ ɔ ʊ
Vowels harmonize as a binary +/−ATR feature within a morpheme; affixes agree with the root (@cite{sande-2017}, @cite{sande-2019}, @cite{sande-2022}). Particles in particle-verb constructions inherit the ATR value of the verb root when both are inside the same spell-out domain — this is the discontinuous-harmony observation.
We do not enumerate the full vowel inventory here; for the Table 44 result we only need a per-terminal ATR value.
The ATR feature value carried by a phonological terminal.
+ATR = true, −ATR = false.
Equations
Instances For
The default ATR value for the particle when no vowel-harmony trigger is present in its spell-out domain. @cite{sande-clem-dabkowski-2026} §2.5 — particles surface with their lexical (typically -ATR) value when not local to a harmony trigger.
Instances For
The four particle-verb word orders that diagnose discontinuous harmony (@cite{sande-clem-dabkowski-2026} (44), §6.1).
- SVOPart : WordOrder
S V O Part: V moves to T; Part stays in vP; clause-final. - SAuxOPartV : WordOrder
S Aux O Part V: V stays in v (Aux occupies T); both V and Part in vP at vP-spell-out. - PartSVO : WordOrder
Part S V O: V moves to T; Part fronts to Spec,CP from vP. - PartSAuxOV : WordOrder
Part S Aux O V: V stays in v; remnant VP (containing Part) fronts to Spec,CP.
Instances For
Equations
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.instDecidableEqWordOrder x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
The terminals spelled out within vP for each word order
(@cite{sande-clem-dabkowski-2026} (45)/(48)). The decisive
difference: SAuxOPartV and PartSAuxOV keep V inside vP at
spell-out (V hasn't raised past v), while SVOPart and PartSVO
have V already in T at vP-spell-out, leaving only Part inside vP.
The object has independently shifted out of vP in all four derivations (per @cite{koopman-1997}-style remnant-VP analysis, extended in §4 of the paper), so it does not appear here.
Equations
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.vPSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.SVOPart = ["Part"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.vPSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.SAuxOPartV = ["Part", "V"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.vPSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.PartSVO = ["Part"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.vPSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.PartSAuxOV = ["Part", "V"]
Instances For
The terminals introduced or first linearized at the CP-phase spell-out. Each terminal is spelled out at exactly one phase per Cyclic Linearization, unless it has moved between phases — in which case the higher phase's linearization positions it again, consistently with the lower phase's statements (Order Preservation).
- In
SVOPartandSAuxOPartV, the Part stays inside vP and appears only invPSpellOut, not here. - In
PartSVOandPartSAuxOV, the Part has fronted to Spec,CP and so appears at the left edge ofcpSpellOut(in addition to its base position invPSpellOut). - The verb V appears here only when it has moved to T (i.e., in
SVOPartandPartSVO); when V stays in v (the SAuxOV cases), it remains inside vP.
Equations
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.cpSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.SVOPart = ["S", "V", "O"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.cpSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.SAuxOPartV = ["S", "Aux", "O"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.cpSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.PartSVO = ["Part", "S", "V", "O"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.cpSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WordOrder.PartSAuxOV = ["Part", "S", "Aux", "O"]
Instances For
Cyclic-Linearization derivation: the two-phase spell-out for a given word order.
Equations
- One or more equations did not get rendered due to their size.
Instances For
SVOPart: V in T, Part in clause-final position. Consistent.
SAuxOPartV: V in v, both V and Part inside vP. Consistent.
PartSVO: V in T, Part fronts. Consistent under
Cyclic-Linearization (no contradiction with vP-internal order).
PartSAuxOV: V in v, remnant VP fronts. Consistent. The crucial
derivation for discontinuous harmony — Part is local to V at
vP-spell-out, and the linearization remains coherent after
fronting.
Harmony applies iff the trigger V is spelled out within vP. Independent of the linear adjacency between V and Part at the surface. The target Part is by stipulation always inside vP at vP-spell-out (it's the Part of a particle-verb construction; even when it later fronts, it starts in vP), so the trigger presence is the only varying condition.
Equations
Instances For
The decisive empirical theorem: SCD 2026 Table 44, decided structurally. Harmony applies in exactly the two SAuxOV-shape derivations — the ones where V hasn't raised past v.
Discontinuous harmony at one glance: in PartSAuxOV, Part is
surface-non-local to V (the subject, auxiliary, and object
intervene at CP-spell-out), yet harmony applies.
The ATR value frozen at vP-spell-out for PartSAuxOV: when V is
+ATR (e.g. /joku/), Part inherits +ATR via local vP-internal
harmony. This is the freezing event.
Equations
- One or more equations did not get rendered due to their size.
Instances For
After remnant-VP fronting at the CP-phase spell-out, no new ATR assignments are issued for Part — the value is preserved.
@cite{sande-clem-dabkowski-2026}'s commitment: the PIC mode for
Guébie discontinuous harmony is linearizationBound — already
spelled-out material remains accessible to later syntax. Strict
PIC₁ or PIC₂ would block the remnant-VP movement of the particle
after its ATR value has been frozen.
Equations
Instances For
Under the Guébie PIC mode, every phase admits extraction of any
goal at the phasehood layer; concrete crashes come from
SpelloutAndCheck instead. The four-derivation theorem above
confirms no derivation crashes.
@cite{sande-clem-dabkowski-2026} §3 establishes via island
sensitivity (their (27)–(30)) that Guébie predicate-fronting verb
doubling is narrow-syntactic, not PF dislocation. This positions
Guébie alongside @cite{harizanov-gribanova-2019}'s Russian as a
positive instance of VerbDoublingIsSyntactic — and against
@cite{landau-2006}'s Hebrew analysis (which makes verb doubling
PF-driven).
The previous formulation as a universal axiom was inconsistent
with Landau's Hebrew analysis; the SCD 2026 paper is the trigger
for the substrate refactor in
`HarizanovGribanova2019Amalgamation.lean`.
We register Guébie as a positive instance via a `RemnantFronting`
witness with a `PredicateDoubling` extension. The structural data
is schematic (we use `Minimalist.SyntacticObject.leaf` placeholders
for V and the fronted XP rather than a concrete tree) — this is
consistent with the rest of the file's spell-out-list abstraction.
What it does establish is that the substrate is *usable*: the
`properRemnant` predicate is decidable on the witness.
The Guébie predicate-fronting witness: V evacuates the VP, and the remnant VP fronts to Spec,CP. The trace is pronounced — verb doubling.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The witness is a proper remnant: the evacuated verb originally
sat inside the fronted VP (true by construction here, since
frontedVP = node verb verb).
Guébie's vP cophonology as a PhrasalCophonology instance #
The vP-domain ATR-harmony cophonology of @cite{sande-jenks-inkelas-2020}
applied to Guébie. The phase selector matches v heads (per the v*
phase head of Chomsky 2000); the constraint subranking is left
abstract here (it would be [ATRHARM ≫ IDENT-IO(ATR)] over the
candidate type the OT machinery uses, which we don't instantiate
inline).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The Guébie vP-cophonology bundle. The subranking is left as an
empty list of constraints over Unit candidates because the
ATR-harmony cophonology's actual constraints live in
Theories/Phonology/Process/Harmony/OT.lean and would require
threading the OT-candidate type through this study file — out of
scope. The substrate use is exhibited by the bundle's existence
and the matched-phase predicate appliesTo.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The Guébie vP-cophonology applies to a v head. (Witness: a leaf
SO whose token's category is .v.)
Guébie as a positive instance of VerbDoublingIsSyntactic #
@cite{sande-clem-dabkowski-2026} §3 provides three independent
diagnostics that the Guébie verb-doubling movement is narrow-syntactic:
successive cyclicity (their (25)–(26)), island sensitivity ((27)–(28)),
and creates-island effects ((29)–(30)). On the basis of this evidence,
we register the Guébie predicate-fronting case as a positive instance
of HarizanovGribanova2019.Amalgamation.VerbDoublingIsSyntactic.
The Lean statement uses the MCB-aligned VerbDoublingIsSyntacticIn
(see HarizanovGribanova2019Amalgamation.lean), which takes a
Derivation and asks whether the verb is among the movedItems of
the derivation. Per MCB §1.4.3.1, Internal Merge IS the syntactic
mechanism that produces surface verb doubling via PF copy/trace
pronunciation rules — the verb appears once in the syntactic tree
(deeper copy as mkTrace), but PF rules pronounce both.
A schematic Guébie Derivation: the verb undergoes Internal Merge in the predicate-fronting derivation (per SCD 2026 §3 island diagnostics establishing this is narrow-syntactic, not PF).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Guébie predicate-fronting verb doubling registered as a positive
instance of VerbDoublingIsSyntacticIn: the verb is among
guebieFrontingDerivation.movedItems.
Decidable from the Derivation structure (no sorry). This
discharges the audit's outstanding sorry for Guébie VDIS.
Wolof shows the same architectural shape as Guébie discontinuous harmony, in a relative-clause construction (@cite{sy-2005}, @cite{martinovic-2019}). The head noun starts local to the distal demonstrative inside DP; both spell out together with local ATR harmony applying; the head noun then A′-moves to the left edge of the relative clause; intervening stative-verb material does not harmonize.
- localDP : WolofRelClauseShape
Local DP, harmony applies (no relative clause):
[DP head dem]— head and dem in the same spell-out domain. - relClause : WolofRelClauseShape
Relative clause, harmony at distance:
head … [stative-V] … dem— head fronted out of DP, but its ATR is set inside the DP-phase before fronting.
Instances For
Equations
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.instDecidableEqWolofRelClauseShape x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.wolofVPSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WolofRelClauseShape.localDP = ["head", "dem"]
- Phenomena.Phonology.Studies.SandeClemDabkowski2026.wolofVPSpellOut Phenomena.Phonology.Studies.SandeClemDabkowski2026.WolofRelClauseShape.relClause = ["head", "dem"]
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Both Wolof shapes have the head and demonstrative spelled out together at the DP-phase spell-out — predicting harmony in both, consistent with @cite{sy-2005}'s data. The discontinuous appearance in the relative-clause case is post-spell-out movement, not a different harmony mechanism.
@cite{russell-2023} reports a parallel pattern in Atchan nasal harmony: a nasal singular subject pronoun triggers nasal harmony on auxiliaries and the verb; in verb-doubling focus constructions, the higher copy of the verb (in focus position) also surfaces nasal even though it is not surface-local to the nasal subject pronoun.
The data are sparse (a single personal-communication example,
SCD 2026 (51)) and the Atchan verb-doubling syntax is not yet
independently established as narrow-syntactic. Recorded here as
an open question.
TODO: Atchan formalization waits on independent syntactic
diagnostics for verb doubling (the analogue of SCD §3 for
Guébie).