Voice and Extraction in Malayic

@cite{erlewine-sommerlot-2025}

@cite{erlewine-sommerlot-2025} propose a new analysis of Malayic clausal morphosyntax that derives the interaction between voice morphology and nominal A'-extraction from cyclic linearization (@cite{fox-pesetsky-2005}).

Core claims

1. The verbal domain involves two functional heads: Voice (higher, phase head) and v (lower, introduces external argument). The active prefix meN- reflects both: me- on Voice, N- on v.

2. VoiceP is a phase with exactly one nominal specifier. Only the subject — the nominal in Spec,VoiceP — can A'-extract from the basic clause types (active, di- passive, bare passive).

3. Object extraction involves the theme moving to Spec,VoiceP (the single nominal specifier) and then A'-moving to Spec,CP, while the agent stays low in Spec,vP. This is possible only when Voice is realized as a null allomorph.

4. meN-deletion derived: overt Voice (me-) in object extraction creates an ordering paradox at cyclic linearization. At VoiceP Spell-out, Voice precedes the agent; at CP Spell-out, the agent (having moved to Spec,TP or staying high) precedes Voice. The contradiction forces Voice to be null.

5. Cross-linguistic variation (Desa, SI/SM, polite/familiar Madurese) reduces to parametric differences in vocabulary items for Voice and v.

PIC mode

@cite{erlewine-sommerlot-2025}'s analysis derives the meN-deletion constraint from cyclic linearization alone, without invoking the Phase Impenetrability Condition. This positions them at PICStrength.linearizationBound (per Theories/Syntax/Minimalist/Phase.lean): no opacity constraint per se, only ordering constraints from SpelloutAndCheck. Same regime adopted by @cite{sande-clem-dabkowski-2026} for Guébie discontinuous harmony, @cite{branan-davis-2019} for agreement-edge unlocking, and others. The structural diagnostic that the meN-deletion derivation is ruled out is precisely not a PIC-violation but a Cyclic Linearization contradiction — see men_deletion below.

Formalization strategy

We model each derivation as a sequence of Spell-out domains, where each domain is the left-to-right sequence of overt terminals. The cyclic linearization machinery from Linearization/Cyclic.lean checks for ordering contradictions. The key theorem is that object extraction with overt Voice creates a contradiction, while null Voice does not.

Overt terminals at VoiceP Spell-out

At VoiceP Spell-out, the phase is:

[VoiceP  DP_spec  Voice  [vP  (DP_agent)  v+V  (DP_theme)]]

The specifier of VoiceP is the nominal that has moved there to satisfy the single-specifier requirement. In active clauses, this is the agent; in passives, the theme; in object extraction, the theme.

We record the left-to-right sequence of overt terminals. Null heads (Voice = ∅, v_PASS = ∅) are omitted from the sequence.

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_active : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_active = ["agent", "me-", "NV", "theme"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_active_short : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_active_short = ["agent", "NV", "theme"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_diPassive : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_diPassive = ["theme", "di-", "V"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_barePassive : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_barePassive = ["theme", "agent", "V"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_objExtr_null : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_objExtr_null = ["theme", "agent", "NV"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_objExtr_null_sism : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_objExtr_null_sism = ["theme", "agent", "V"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_objExtr_overt : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voiceP_objExtr_overt = ["theme", "me-", "agent", "NV"]

Overt terminals at CP Spell-out

At CP Spell-out, material that has moved out of VoiceP is ordered relative to VoiceP-internal material and any new CP-level material (complementizer, auxiliaries). The key principle: moved material precedes VoiceP-internal material in the CP Spell-out domain.

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_active : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_active = ["agent", "Aux", "me-", "NV", "theme"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_active_short : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_active_short = ["agent", "Aux", "NV", "theme"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_subjExtr : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_subjExtr = ["agent", "Aux", "me-", "NV", "theme"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_diPassive : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_diPassive = ["theme", "Aux", "di-", "V"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_barePassive : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_barePassive = ["theme", "Aux", "agent", "V"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_objExtr_sism : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_objExtr_sism = ["theme", "agent", "Aux", "V"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_objExtr_desa : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_objExtr_desa = ["theme", "agent", "NV"]

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_objExtr_overt : List String

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.cp_objExtr_overt = ["theme", "agent", "Aux", "me-", "NV"]

Grammatical derivations

Each grammatical derivation produces consistent ordering across phases.

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.active_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_active, cp_active]

Active clause is consistently linearizable. @cite{erlewine-sommerlot-2025} (36).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.active_short_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_active_short, cp_active_short]

Active clause with short N- prefix (Desa free variation).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.subject_extraction_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_active, cp_subjExtr]

Subject extraction from active is consistent. The subject (agent) was already leftmost in VoiceP and moves further left through Spec,TP to Spec,CP — classic edge movement (Scenario 1 of @cite{fox-pesetsky-2005}).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.di_passive_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_diPassive, cp_diPassive]

di- passive is consistently linearizable. @cite{erlewine-sommerlot-2025} (37a).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.bare_passive_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_barePassive, cp_barePassive]

Bare passive is consistently linearizable. @cite{erlewine-sommerlot-2025} (37b)/(39c).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.obj_extraction_desa_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_objExtr_null, cp_objExtr_desa]

Object extraction with null Voice (Desa) is consistent. @cite{erlewine-sommerlot-2025} (44).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.obj_extraction_sism_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_objExtr_null_sism, cp_objExtr_sism]

Object extraction with null Voice (SI/SM) is consistent. @cite{erlewine-sommerlot-2025} (54)–(56).

The ordering paradox: overt Voice in object extraction

This is the paper's central formal result. When Voice is overt (me-), VoiceP Spell-out establishes me- < agent (Voice precedes agent in Spec,vP). But at CP Spell-out, the agent has moved past Voice, establishing agent < me-. The two are contradictory.

@cite{erlewine-sommerlot-2025} (57).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.men_deletion : ¬Minimalist.Linearization.SpelloutAndCheck [voiceP_objExtr_overt, cp_objExtr_overt]

meN-deletion theorem: object extraction with overt Voice creates an ordering paradox.

VoiceP: [theme, me-, agent, NV] → me- < agent CP: [theme, agent, Aux, me-, NV] → agent < me-

These two statements contradict: me- < agent ∧ agent < me-.

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.men_deletion_witness : ((Minimalist.Linearization.allPrecs voiceP_objExtr_overt).any fun (s : Minimalist.Linearization.Prec) => s.before == "me-" && s.after == "agent") = true ∧ ((Minimalist.Linearization.allPrecs cp_objExtr_overt).any fun (s : Minimalist.Linearization.Prec) => s.before == "agent" && s.after == "me-") = true ∧ Minimalist.Linearization.HasContradiction (Minimalist.Linearization.allPrecs voiceP_objExtr_overt ++ Minimalist.Linearization.allPrecs cp_objExtr_overt)

The specific contradiction: VoiceP says me- before agent; CP says agent before me-.

Desa vs SI/SM: the N- contrast

In Desa, the short nasal prefix N- (realization of v_ACT) survives in object extraction because it is on the lower head v, which is not implicated in the ordering paradox. In SI/SM, N- is lost because v_ACT only realizes as N- when Voice is linearly adjacent — and in object extraction, Voice is null (pruned), so the adjacency condition fails.

@cite{erlewine-sommerlot-2025} §2.3, (3) vs (22), (25).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.desa_n_survives : Fragments.Malayic.VoiceSystem.desa.vExponent Fragments.Malayic.VoiceSystem.VoiceConstruction.objectExtraction = some "N-" ∧ Fragments.Malayic.VoiceSystem.desa.voiceExponent Fragments.Malayic.VoiceSystem.VoiceConstruction.objectExtraction = none

Desa object extraction: verb bears short N- but not me-. Desa (3): Opai yang inya m-ewa' 'What did s/he bring?' m- = N- (v_ACT), no me- (Voice).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.sism_full_deletion : Fragments.Malayic.VoiceSystem.standardSISM.vExponent Fragments.Malayic.VoiceSystem.VoiceConstruction.objectExtraction = none ∧ Fragments.Malayic.VoiceSystem.standardSISM.voiceExponent Fragments.Malayic.VoiceSystem.VoiceConstruction.objectExtraction = none

SI/SM object extraction: verb bears NO prefix at all. SI/SM (22): baju-baju yang Ali tidak basuh No meN-, no N-, bare stem basuh.

Polite vs familiar Madurese

@cite{jeoung-2017} documents that polite Madurese has three voices (active, e- passive, bare passive) plus object extraction, while familiar Madurese has only two voices (active, e- passive). The contrast reduces to whether Voice has a null elsewhere allomorph.

@cite{erlewine-sommerlot-2025} §5.3, (76)–(83).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.polite_madurese_obj_extr : Fragments.Malayic.VoiceSystem.politeMadurese.clauseAvailable Fragments.Malayic.VoiceSystem.VoiceConstruction.objectExtraction = true

Polite Madurese allows object extraction (null Voice available).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.familiar_madurese_no_obj_extr : Fragments.Malayic.VoiceSystem.familiarMadurese.clauseAvailable Fragments.Malayic.VoiceSystem.VoiceConstruction.objectExtraction = false

Familiar Madurese blocks object extraction (no null Voice).

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.madurese_minimal_pair : Fragments.Malayic.VoiceSystem.politeMadurese.hasNullVoice ≠ Fragments.Malayic.VoiceSystem.familiarMadurese.hasNullVoice

The register contrast is exactly the null-Voice parameter.

PP A'-movement

@cite{erlewine-sommerlot-2025} correctly predict that nonnominal constituents (PPs) can A'-extract freely, regardless of voice, because VoiceP hosts nonnominal specifiers in addition to its one nominal specifier ((35b), (42)). PP extraction from active clauses retains meN- ((40a), (41a)).

We model this as: a PP in an additional specifier of VoiceP moves to Spec,CP. It is ordered before Voice at VoiceP Spell-out and remains before Voice at CP Spell-out — no contradiction regardless of whether Voice is overt.

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.pp_extraction_with_overt_voice : Minimalist.Linearization.SpelloutAndCheck [["PP", "agent", "me-", "NV", "theme"], ["PP", "agent", "Aux", "me-", "NV", "theme"]]

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.all_grammatical_derivations_consistent : Minimalist.Linearization.SpelloutAndCheck [voiceP_active, cp_active] ∧ Minimalist.Linearization.SpelloutAndCheck [voiceP_active_short, cp_active_short] ∧ Minimalist.Linearization.SpelloutAndCheck [voiceP_diPassive, cp_diPassive] ∧ Minimalist.Linearization.SpelloutAndCheck [voiceP_barePassive, cp_barePassive] ∧ Minimalist.Linearization.SpelloutAndCheck [voiceP_objExtr_null, cp_objExtr_desa] ∧ Minimalist.Linearization.SpelloutAndCheck [voiceP_objExtr_null_sism, cp_objExtr_sism] ∧ ¬Minimalist.Linearization.SpelloutAndCheck [voiceP_objExtr_overt, cp_objExtr_overt]

Summary: for every basic clause type, the derivation with appropriate null/overt Voice assignments is consistently linearizable. Object extraction requires null Voice (derived, not stipulated).

Connecting Malayic Voice/v to Core VoiceHead

@cite{erlewine-sommerlot-2025}'s two-head system (Voice + v) maps onto Core's single VoiceHead type (@cite{kratzer-1996}, @cite{schaefer-2008}).

• v_ACT (introduces agent) → agentive Voice flavor
• v_PASS (no θ) → passive Voice flavor
• VoiceP is ALWAYS a phase in E&S's system, diverging from Core's flavor-default tabulation in which passive Voice is non-phasal (@cite{collins-2005}, @cite{chomsky-2001}). This divergence is captured by the malayic_passive_phase_diverges theorem below.

The phase divergence is theoretically significant: E&S need VoiceP to be a phase in ALL clause types to trigger cyclic linearization at VoiceP Spell-out, including in passives. Phase.lean's Transfer corresponds to each Spell-out domain in Linearization/Cyclic.lean, but operates on SyntacticObject rather than terminal strings.

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.vFlavorToCore : Fragments.Malayic.VoiceSystem.LittleVFlavor → Minimalist.VoiceFlavor

Map Malayic v-flavor to Core VoiceFlavor. v_ACT introduces the external argument → agentive. v_PASS licenses EA without θ → passive.

Equations

• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.vFlavorToCore Fragments.Malayic.VoiceSystem.LittleVFlavor.act = Minimalist.VoiceFlavor.agentive
• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.vFlavorToCore Fragments.Malayic.VoiceSystem.LittleVFlavor.pass = Minimalist.VoiceFlavor.passive

def Phenomena.FillerGap.Studies.ErlewineSommerlot2025.clauseToVoiceHead : Fragments.Malayic.VoiceSystem.VoiceConstruction → Minimalist.VoiceHead

Map each Malayic clause type to its Core VoiceHead equivalent. Active and object-extraction Voice are agentive (flavor-default phasal). di-passive and bare passive use phaseOverride := some true to express E&S 2025's claim that VoiceP is universally a phase, diverging from the Core default for passive Voice (@cite{chomsky-2001}, @cite{collins-2005}, encoded in VoiceFlavor.defaultPhasal).

Equations

• One or more equations did not get rendered due to their size.
• Phenomena.FillerGap.Studies.ErlewineSommerlot2025.clauseToVoiceHead Fragments.Malayic.VoiceSystem.VoiceConstruction.active = { flavor := Minimalist.VoiceFlavor.agentive, hasD := true }

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voice_flavor_consistent (ct : Fragments.Malayic.VoiceSystem.VoiceConstruction) : (clauseToVoiceHead ct).flavor = vFlavorToCore ct.vFlavor

The VoiceFlavor component is consistent with the v-flavor mapping.

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.voice_always_phase (ct : Fragments.Malayic.VoiceSystem.VoiceConstruction) : (clauseToVoiceHead ct).IsPhasal

All Malayic clause types treat VoiceP as a phase head.

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.malayic_passive_phase_diverges : (clauseToVoiceHead Fragments.Malayic.VoiceSystem.VoiceConstruction.diPassive).IsPhasal ∧ ¬Minimalist.voicePassive.IsPhasal

Phase divergence: Malayic passives are phases, but Core's default passive Voice (following @cite{collins-2005}) is not.

theorem Phenomena.FillerGap.Studies.ErlewineSommerlot2025.active_consistent_with_core : (clauseToVoiceHead Fragments.Malayic.VoiceSystem.VoiceConstruction.active).flavor = Minimalist.VoiceFlavor.agentive ∧ ((clauseToVoiceHead Fragments.Malayic.VoiceSystem.VoiceConstruction.active).IsPhasal ↔ Minimalist.voiceAgent.IsPhasal)

Active clause type maps to Core's agentive Voice, which IS a phase head — consistent across both analyses.