Docs / Studies / Studies.ErlewineSommerlot2025
Voice and Extraction in Malayic
Erlewine, M. Y. & Sommerlot, C. J. (2025). Language, 101(4), 700–750
This file formalizes the account of voice and Ā-extraction in Malayic in Erlewine & Sommerlot 2025. The verbal domain has two heads, after Nomoto: v, which introduces the agent and licenses one nominal, and Voice, which heads a phase with exactly one nominal specifier. Every nominal is licensed by v or by T, so the nominal that v leaves unlicensed moves through Spec,VoiceP to the subject position. Each clause's Minimalist derivation is built from its parameters, and its VoiceP and CP Spell-outs, read off the derivation as Fox and Pesetsky's cyclic linearization reads them, are the paper's snapshots. Vocabulary items realize Voice and v by their adjacency at VoiceP Spell-out, a null Voice is pruned from the ordering statements, and a clause is grammatical when it is licensed, meets the EPP and the vocabulary, affixes an overt Voice to the verb, and linearizes.
The subject-only restriction on Ā-extraction follows, and so does its exception: a nominal other than the agent can be extracted while the agent raises or stays low only when Voice is null, since an overt Voice is ordered before the agent at VoiceP and after it at CP (57); the agent of a bare passive cannot be extracted across its theme (60); N- on v rather than Voice allows object extraction (§5.2); and the registers of Madurese differ in whether Voice has a null allomorph (§5.3). The six grammars predict the judgments of the paper's examples.
Implementation notes
- A clause is its syntactic parameters, and the exponents of Voice and v are chosen after VoiceP
Spell-out. Each phase, of Voice or C, is spelled out when the next head projects over it
(
Minimalist.Derivation.phaseSchedule). v+V is one terminal of category v, the paper not showing V-to-v movement, and one terminal stands for T with its auxiliaries. - The active v always projects its agent (30). The ban on an agent licensed by the passive v moving to Spec,VoiceP is the antilocality the paper tentatively suggests (§3.3).
- Desa's T licenses a nominal in situ instead of attracting one under nominal Ā-extraction (§3.3). Kuching Malay's Voice is always null, and Jakarta Indonesian's di- passive is taken from Standard Indonesian (§5.2).
- Ditransitive, possessor and embedded-argument extraction (§4.4), ellipsis (§4.5) and the Kendayan passives (§5.2) are not modelled.
References
- Erlewine & Sommerlot 2025, Voice and Extraction in Malayic
- Nomoto 2015, Person Restriction on Passive Agents in Malay and Givenness
- Nomoto 2021, Bare Passive Agent Hierarchy
- Fox & Pesetsky 2005, Cyclic Linearization of Syntactic Structure formalized
- Embick & Noyer 2001, Movement Operations after Syntax
- Jeoung 2017, On the Number of Voices in Madurese
Declarations
A bivalent verb's nominal arguments are its agent and its theme.
Derives DecidableEq, Repr, Fintype.
Used by Extracted, Term, Clause, Clause.nominals, Clause.LicensedByV, Clause.instDecidableLicensedByV, Clause.aBar, Grammar.TLicenses, Grammar.instDecidableTLicenses, subjectTable, surfaceOrder
C attracts a nominal or a PP to Spec,CP.
Derives DecidableEq, Repr, Fintype.
Built from
Used by Extracted.term, Clause, Grammar.Generates, Grammar.instDecidableGenerates, extractedTable, surfaceOrder, Row
Voice and v are realized as me-, N-, di- or e- across the languages.
Derives DecidableEq, Repr, Fintype.
Used by Exponence, Exponence.affix, Grammar, Grammar.Generates, Grammar.instDecidableGenerates, affixTable, Row
A Spell-out orders nominals, the nonnominal specifier, Voice, the complex head v+V, and T with its auxiliaries.
Derives DecidableEq, Repr, Fintype.
Built from
Used by Term.token, Term.ofToken?, Extracted.term, Clause.spellouts, Clause.voiceP, Clause.moved, Clause.cP, Clause.surface, Exponence.prune, Clause.phases, Grammar.Generates, Grammar.instDecidableGenerates, surfaceOrder, Row
Each terminal spells out its own lexical item, v+V that of the light verb it projects, which selects the theme; Voice selects vP and T selects VoiceP.
Equations
- (Term.dp Nominal.agent).token = { item := Minimalist.LexicalItem.of Minimalist.Cat.D [], id := 1 }
- (Term.dp Nominal.theme).token = { item := Minimalist.LexicalItem.of Minimalist.Cat.D [], id := 2 }
- Term.pp.token = { item := Minimalist.LexicalItem.of Minimalist.Cat.P [], id := 3 }
- Term.voice.token = { item := Minimalist.LexicalItem.of Minimalist.Cat.Voice [Minimalist.Cat.v], id := 4 }
- Term.verb.token = { item := Minimalist.LexicalItem.of Minimalist.Cat.v [Minimalist.Cat.D], id := 5 }
- Term.aux.token = { item := Minimalist.LexicalItem.of Minimalist.Cat.T [Minimalist.Cat.Voice], id := 6 }
Built from
Used by IsPhaseHead, instDecidablePredTokenIsPhaseHead, Clause.derivation
A lexical item spells out at most one terminal.
The definition is too large to print here; read it in the source.
Built from
Used by Clause.spellouts
The complementizer that attracts the Ā-extracted phrase is silent.
- C₀ = { item := Minimalist.LexicalItem.of Minimalist.Cat.C [Minimalist.Cat.T], id := 7 }
- instDecidablePredTokenIsPhaseHead x✝ = { decide := decide (x✝ = Term.voice.token) || decide (x✝ = C₀), reflects_decide := ⋯ }
Built from , ,
An extracted phrase is a nominal or the PP.
- (Extracted.nominal a).term = Term.dp a
- Extracted.pp.term = Term.pp
Built from ,
Used by Clause.moved
A bivalent clause is fixed by the flavour of v, whether v projects the agent, the nominal moved to Spec,VoiceP, the nominal T attracts to Spec,TP if any, and what C attracts if anything.
- flavor : VFlavor
- agentProjected : Bool
- spec : Nominal
- subject : Option Nominal
- extracted : Option Extracted
Derives DecidableEq, Repr, Fintype.
Built from , ,
Used by Clause.voicePSteps, Clause.cpSteps, Clause.derivation, Clause.schedule, Clause.spellouts, Clause.nominals, Clause.AgentInSitu, Clause.instDecidableAgentInSitu, Clause.voiceP, Clause.moved, Clause.cP, Clause.surface, Clause.Adjacent, Clause.instDecidableAdjacent, Clause.LicensedByV, Clause.instDecidableLicensedByV, Clause.aBar, Clause.WellFormed, Clause.instDecidableWellFormed, Clause.spellouts_eq, Clause.schedule_eq, Clause.adjacent_iff, Clause.nodup_cP, Clause.voiceP_subset_cP, Clause.phases, Clause.consistent_phases_iff, LocalDislocation, instDecidableLocalDislocation, Grammar.TLicenses, Grammar.instDecidableTLicenses, Grammar.Licensed, Grammar.EPP, Grammar.Vocabulary, Grammar.Grammatical, Grammar.instDecidableLicensed, Grammar.instDecidableEPP, Grammar.instDecidableVocabulary, Grammar.instDecidableGrammatical, Grammar.exponences, Grammar.Generates, overt_voice_paradox, bare_passive_agent_paradox
The derivation and its Spell-outs
VoiceP is built from the theme: v+V merges, then the PP that will be extracted, the agent in Spec,vP, and Voice; the nominal specifier raises, and the PP raises above it (30), (42).
The definition is too large to print here; read it in the source.
Built from
Used by Clause.derivation, Clause.schedule_eq
Above VoiceP, T with its auxiliaries merges and attracts the subject, and C merges and attracts the extracted phrase.
The definition is too large to print here; read it in the source.
Built from
Used by Clause.derivation
A clause's derivation starts from the theme.
- c.derivation = { initial := ↑(Term.dp Nominal.theme).token, steps := c.voicePSteps ++ c.cpSteps }
Built from , , ,
Used by Clause.schedule, Clause.spellouts, Clause.schedule_eq
Each phase is spelled out when it is complete.
Built from , ,
Used by Clause.spellouts, Clause.schedule_eq
The derivation's Spell-outs read as terminals, before vocabulary insertion.
- c.spellouts = List.map (fun (p : ℕ × ℕ) => List.filterMap Term.ofToken? (c.derivation.spellout p.1 p.2)) c.schedule
Built from , , , ,
Used by Clause.spellouts_eq
The clause projects the theme, and the agent when v projects it.
- c.nominals = if c.agentProjected = true then [Nominal.agent, Nominal.theme] else [Nominal.theme]
Built from ,
Used by Grammar.Licensed
The agent stays in Spec,vP at VoiceP Spell-out when it is projected and not the nominal specifier.
- c.AgentInSitu = (c.agentProjected = true ∧ c.spec ≠ Nominal.agent)
Built from
Used by Clause.instDecidableAgentInSitu, Clause.adjacent_iff, overt_voice_paradox
- c.instDecidableAgentInSitu = { decide := decide (c.agentProjected = true) && decide (c.spec ≠ Nominal.agent), reflects_decide := ⋯ }
Built from ,
At VoiceP Spell-out (55a) the PP precedes the nominal specifier, then Voice, the agent if it stays in Spec,vP, v+V, and the theme if it stays in situ.
The definition is too large to print here; read it in the source.
Built from ,
Used by Clause.cP, Clause.Adjacent, Clause.instDecidableAdjacent, Clause.spellouts_eq, Clause.voiceP_subset_cP, Clause.phases, Clause.consistent_phases_iff
At CP Spell-out (56b) the moved phrases precede the auxiliaries and the rest of VoiceP.
Built from , , ,
Used by Clause.surface, Clause.spellouts_eq, Clause.nodup_cP, Clause.voiceP_subset_cP, Clause.phases, Clause.consistent_phases_iff
The clause's word order is its CP Spell-out with Voice affixed to the verb.
- c.surface = List.filter (fun (x : Term) => decide (x ≠ Term.voice)) c.cP
Built from , ,
Used by Grammar.Generates
Voice and v are linearly adjacent at VoiceP Spell-out.
- c.Adjacent = ([Term.voice, Term.verb] <:+: c.voiceP)
Built from ,
Used by Clause.instDecidableAdjacent, Clause.adjacent_iff, LocalDislocation, Grammar.Vocabulary
- c.instDecidableAdjacent = [Term.voice, Term.verb].decidableInfix c.voiceP
Built from , ,
v licenses a nominal inside VoiceP: the active v the theme it c-commands, the passive v the agent in its specifier (§3.1).
- c.LicensedByV Nominal.theme = (c.flavor = VFlavor.act)
- c.LicensedByV Nominal.agent = (c.flavor = VFlavor.pass ∧ c.agentProjected = true)
Built from ,
Used by Clause.instDecidableLicensedByV, Grammar.TLicenses, Grammar.Licensed
The definition is too large to print here; read it in the source.
Built from , ,
The Ā-extracted nominal, if any.
- c.aBar = match c.extracted with | some (Extracted.nominal n) => some n | x => none
Built from ,
Used by Grammar.TLicenses, Grammar.EPP
A clause is well formed when the active v projects its agent (30), the specifier, the subject and the extracted nominal are projected, and no agent licensed by the passive v moves to Spec,VoiceP (§3.3).
The definition is too large to print here; read it in the source.
Built from
Used by Clause.instDecidableWellFormed, Clause.spellouts_eq, Clause.schedule_eq
Every well-formed clause's derivation spells out the paper's VoiceP and CP snapshots.
Built from , , , ,
The phase of Voice is complete when VoiceP is built, and that of C at the end.
(c : Clause) :
c.WellFormed → c.schedule = [(c.voicePSteps.length, c.voicePSteps.length), (c.derivation.length, c.derivation.length)]
Built from , , , ,
Built from , ,
Exponence
The exponents chosen for Voice and v at VoiceP Spell-out.
Derives DecidableEq, Repr.
Built from
Used by Exponence.affix, Exponence.prune, Clause.phases, Clause.consistent_phases_iff, LocalDislocation, instDecidableLocalDislocation, Grammar.Vocabulary, Grammar.Grammatical, Grammar.instDecidableVocabulary, Grammar.instDecidableGrammatical, Grammar.exponences, overt_voice_paradox, bare_passive_agent_paradox
A null Voice is pruned from the ordering statements (l.1351–1354).
- e.prune l = if e.voice.isSome = true then l else List.filter (fun (x : Term) => decide (x ≠ Term.voice)) l
Built from ,
The ordering statements of a clause under an exponence.
Built from , , , , ,
Used by Clause.consistent_phases_iff, overt_voice_paradox, bare_passive_agent_paradox
An overt Voice prefixes to v+V by local dislocation, which needs adjacency (§3.1).
- LocalDislocation c e = (e.voice.isSome = true → c.Adjacent)
Built from , ,
Used by instDecidableLocalDislocation
The definition is too large to print here; read it in the source.
Built from , ,
Grammars
A language's vocabulary items (§3.1, §4.1, §5) give the exponents Voice may take, by the flavour of v and whether the two heads are adjacent, likewise for v, and whether T licenses a nominal in situ under nominal Ā-extraction instead of attracting one.
- inSitu : Bool
Built from ,
Used by desa, standard, jakarta, kuching, politeMadurese, familiarMadurese, Grammar.TLicenses, Grammar.instDecidableTLicenses, Grammar.Licensed, Grammar.EPP, Grammar.Vocabulary, Grammar.Grammatical, Grammar.instDecidableLicensed, Grammar.instDecidableEPP, Grammar.instDecidableVocabulary, Grammar.instDecidableGrammatical, Grammar.exponences, Grammar.Generates, Grammar.instDecidableGenerates, grammarTable, Row
In Desa (32) me- is optional before the active v, di- appears before the passive v and Voice is null elsewhere, the active v is N-, and T licenses in situ under Ā-extraction.
The definition is too large to print here; read it in the source.
Built from
In Standard Indonesian and Malay (49) me- and N- each appear only next to the other.
The definition is too large to print here; read it in the source.
Built from
In Jakarta Indonesian (§5.2) the optional N- realizes Voice.
The definition is too large to print here; read it in the source.
Built from
Used by n_on_v_allows_object_extraction
In Kuching Malay (§5.2) the optional N- realizes v and Voice is always null.
The definition is too large to print here; read it in the source.
Built from
Used by n_on_v_allows_object_extraction
In polite Madurese (82) Voice is N- or e- next to v and null elsewhere.
The definition is too large to print here; read it in the source.
Built from
Used by madurese_registers
In familiar Madurese (83) Voice is N- or e- by the flavour of v alone, with no null allomorph.
The definition is too large to print here; read it in the source.
Built from
Used by madurese_registers
T licenses the nominal it attracts, or, where the EPP is relaxed under Ā-extraction, the one nominal it c-commands that v leaves unlicensed.
- g.TLicenses c n = match c.subject with | some s => n = s | none => g.inSitu = true ∧ c.aBar.isSome = true ∧ ¬c.LicensedByV n
Built from , , , ,
The definition is too large to print here; read it in the source.
Built from , , ,
Every nominal is licensed, by v or by T (§3.1).
- g.Licensed c = ∀ n ∈ c.nominals, c.LicensedByV n ∨ g.TLicenses c n
Built from , , , ,
Used by Grammar.instDecidableLicensed
The exponents are those the vocabulary items allow, given the adjacency of Voice and v at VoiceP Spell-out.
- g.Vocabulary c e = (e.voice ∈ g.voice c.flavor (decide c.Adjacent) ∧ e.v ∈ g.v c.flavor (decide c.Adjacent))
Built from , , ,
Used by Grammar.instDecidableVocabulary
A clause under an exponence is grammatical when it is well formed, licenses every nominal, meets the EPP and the vocabulary, affixes an overt Voice, and linearizes.
The definition is too large to print here; read it in the source.
Built from , ,
- g.instDecidableLicensed c = id inferInstance
Built from , ,
- g.instDecidableEPP c = id inferInstance
Built from , ,
- g.instDecidableVocabulary c e = id inferInstance
Built from , , ,
The definition is too large to print here; read it in the source.
Built from , , ,
The exponences the vocabulary items allow a clause.
The definition is too large to print here; read it in the source.
Built from , ,
Used by Grammar.Generates
The grammar generates a construction when some grammatical clause extracts the phrase, has the word order, and shows the prefix.
- g.Generates extracted order affix = ∃ (c : Clause), c.extracted = extracted ∧ c.surface = order ∧ ∃ e ∈ g.exponences c, e.affix = affix ∧ g.Grammatical c e
Built from , , , , , , , ,
Used by Grammar.instDecidableGenerates, rows_grammatical, n_on_v_allows_object_extraction, madurese_registers
The rows
The grammars as named in the rows.
The definition is too large to print here; read it in the source.
Built from
The extracted phrases as named in the rows.
The definition is too large to print here; read it in the source.
Built from
The subjects as named in the rows.
- subjectTable = [("none", none), ("agent", some Nominal.agent), ("theme", some Nominal.theme)]
Built from
Whether the agent is low, as named in the rows.
- lowAgentTable = [("yes", true), ("no", false)]
The prefixes as named in the rows.
The definition is too large to print here; read it in the source.
Built from
A row's labels describe a word order of the paper's schema DP Aux* (DPag) v+V (DPth).
The definition is too large to print here; read it in the source.
Built from , ,
A row gives a grammar, what is extracted, the word order, the prefix, and the judgment.
- grammar : Grammar
- extracted : Option Extracted
- order : List Term
- affix : List Exponent
- judgment : Judgment
Built from , , ,
Used by Row.ofDatum, rows, rows_grammatical
A row is read off an example's labels.
The definition is too large to print here; read it in the source.
Built from
Used by row_ofDatum_isSome, rows
Built from
The rows of the six grammars.
- rows = List.filterMap Row.ofDatum Examples.all
Built from ,
Used by rows_grammatical
Every example is acceptable exactly when its grammar generates its construction.
(r : Row) :
Built from , ,
The ordering paradoxes
With Voice overt and the agent in Spec,vP at VoiceP Spell-out, raising the agent to Spec,TP orders it before Voice at CP, against VoiceP (57).
(c : Clause) (e : Exponence) (hv : e.voice.isSome = true) (hi : c.AgentInSitu) (hs : c.subject = some Nominal.agent) :
Built from , , ,
(c : Clause) (e : Exponence) (ha : c.agentProjected = true) (hsp : c.spec = Nominal.theme) (he : c.extracted = some (Extracted.nominal Nominal.agent)) (hs : c.subject = some Nominal.theme) :
Built from , ,
The predictions of §5
Built from , ,
Built from , ,