Ostrove (2026): Obligatorily Overt PRO in San Martín Peras Mixtec #
Linguistic Inquiry 57(1): 1–48.
San Martín Peras Mixtec (SMPM), an Oto-Manguean language (ISO: jmx), has obligatory control constructions where the controlled subject must be an overt clitic pronoun — null PRO is strongly ungrammatical. This is analyzed via the minimal pronoun framework ([Kra09b], [Saf14], [Lan15]): SMPM simply lacks a null vocabulary item for controlled subject position. The elsewhere item (→ pronoun) applies.
Core Contributions #
- Three-way clause typology: finite embedded, tensed subjunctive, untensed subjunctive — distinguished by TAM, noncoreferential subjects, and restructuring (26)
- OC with overt pronouns: untensed subjunctives show the full OC signature despite having an overt clitic pronoun, not null PRO
- Against movement, for base-generation: exempt anaphor distribution shows the controlled pronoun is base-generated (§6)
- Morphological analysis: overt PRO derived by lacking a null vocabulary item; cross-linguistic syncretism typology (92)
- Copy control typology: four types of copy control distinguished cross-linguistically (§5)
- Implicational universal: overt PRO → non-pro-drop (54)
Landau (2004) Bridge #
SMPM's subjunctive types map onto [Lan04a]'s finiteness scale, connecting clause-level tense properties to control. The paper explicitly discusses this connection (p.8), following Landau's distinction between "C-subjunctives" (untensed, OC) and "F-subjunctives" (tensed, non-OC).
Wurmbrand (2014) Partial Bridge #
[Wur14]'s three-way classification of infinitival tense (futureIrrealis, restructuring, propositional) maps partially to SMPM's subjunctive types. The mapping applies only to subjunctives — SMPM's finite embedded clauses are genuinely finite and fall outside Wurmbrand's infinitival classification.
| Wurmbrand class | SMPM clause type | OC? |
|---|---|---|
| futureIrrealis | tensed subjunctive | No |
| restructuring | untensed subjunctive | Yes |
| (not applicable) | finite embedded | No |
The control signature of each SMPM clause type, from
clauseProperties.noncoreferentialSubject via
Landau2013.ofNoncoreferential — the same derivation
Allotey2021.gaProfile uses for Gã.
Untensed subjunctives show the full OC signature (§4):
- Sloppy-only under VPE (33)
- Exhaustive binding — no partial control (37)
- Local c-commanding antecedent required (40, 44)
Tensed subjunctives and finite embedded clauses show none of these properties: they allow strict readings under VPE (30, 32), nonexhaustive binding (tensed subj., fn. 16), and non-local antecedents (43, 45).
Equations
Instances For
SMPM's two subjunctive types correspond to [Wur14]'s infinitival tense classes. The mapping applies only to subjunctives — SMPM's finite embedded clauses are genuinely finite and fall outside Wurmbrand's infinitival classification.
- futureIrrealis ↔ tensed subjunctive: future-oriented, non-OC
- restructuring ↔ untensed subjunctive: dependent tense, OC
Note: Wurmbrand's propositional class (ECM/attitude infinitives
like "believe Julia to be smart") has no SMPM correspondent. SMPM's
finite embedded clauses have full TAM morphology and freely
noncoreferential subjects — they are not infinitival.
Equations
- Ostrove2026.wurmbrandToSubjunctive Minimalist.InfinitivalTenseClass.futureIrrealis = some Mixtec.SMPM.EmbeddedClauseType.tensedSubjunctive
- Ostrove2026.wurmbrandToSubjunctive Minimalist.InfinitivalTenseClass.restructuring = some Mixtec.SMPM.EmbeddedClauseType.untensedSubjunctive
- Ostrove2026.wurmbrandToSubjunctive Minimalist.InfinitivalTenseClass.propositional = none
Instances For
Whether a Wurmbrand class involves obligatory control.
Equations
Instances For
For the Wurmbrand classes with SMPM correspondents, the mapping correctly predicts control properties.
Propositional infinitives have no SMPM correspondent — SMPM finite embedded clauses are genuinely finite, not infinitival.
SMPM clause types map onto [Lan04a]'s finiteness scale. This is the framework the paper explicitly uses (p.8).
| Landau class | SMPM clause type | OC? |
|---|---|---|
| C-subjunctive | untensed subjunctive | Yes |
| F-subjunctive | tensed subjunctive | No |
| finite | finite embedded | No |
Equations
- Ostrove2026.landauToSMPM Control.ClauseClass.cSubjunctive = Mixtec.SMPM.EmbeddedClauseType.untensedSubjunctive
- Ostrove2026.landauToSMPM Control.ClauseClass.fSubjunctive = Mixtec.SMPM.EmbeddedClauseType.tensedSubjunctive
- Ostrove2026.landauToSMPM Control.ClauseClass.finite = Mixtec.SMPM.EmbeddedClauseType.finiteEmbedded
Instances For
The forward direction, derived from the fragment's TAM observables
via ClauseClass.ofFiniteness (the same derivation as
Allotey2021.gaToLandau).
Equations
Instances For
The stipulated table landauToSMPM is a section of the derived
classification: SMPM realizes every position of [Lan04a]'s
scale (contrast Allotey2021.ga_no_fSubjunctive).
SMPM Agr status for each Landau clause class.
- C-subjunctive (untensed): [−Agr] — no independent subject agreement
- F-subjunctive (tensed): [+Agr] — allows noncoreferential subjects, which indicates independent agreement capability
- Finite: [+Agr] — full agreement
Under the TTC's OC-NC generalization ((70) in [Lan15]), [+Agr] blocks logophoric control. This is why SMPM tensed subjunctives (F-subjunctives with [+Agr]) show no OC despite structurally permitting logophoric control.
Equations
Instances For
The Landau classification predicts control properties for all three SMPM clause types, taking Agr status into account.
- C-subjunctive [−Agr]: predicative OC (Agr-independent) → OC ✓
- F-subjunctive [+Agr]: logophoric OC blocked by Agr → no OC ✓
- Finite [+Agr]: no control tier → no OC ✓
English vocabulary items (94a–c).
Three items: null for controlled, reflexive for locally bound, pronoun elsewhere. English distinguishes all three non-free BVA contexts morphologically.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Haitian vocabulary items (96a–b).
Two items: null for controlled, pronoun elsewhere. Crucially LACKS a reflexive allomorph — reflexives and bound variables are both realized as pronouns ([DM94]).
Equations
- One or more equations did not get rendered due to their size.
Instances For
SMPM vocabulary items (98a–b).
Two items: reflexive for locally bound, pronoun elsewhere. Crucially LACKS a null allomorph — controlled subjects and bound variables are both realized as overt clitic pronouns (=rà, =ñá, etc.).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Quiegolani Zapotec: no context-specific items at all ([Bla94a]).
Everything — reflexives, controlled subjects, bound variables — surfaces as a single pronoun form (men). Total syncretism.
Equations
- Ostrove2026.quiegolaniInventory = { items := [], elsewhere := Minimalist.MinimalPronoun.PronForm.pronoun }
Instances For
Büli vocabulary items (inferred from [Sul21]).
Like SMPM and Gã: lacks a null allomorph for controlled subjects. Overt PRO in nonfinite complementation.
Equations
- Ostrove2026.buliInventory = { items := [], elsewhere := Minimalist.MinimalPronoun.PronForm.pronoun }
Instances For
English: controlled subjects are null (= silent PRO).
SMPM: controlled subjects are overt pronouns (= overt PRO). This is the paper's central empirical observation.
Haitian: controlled subjects are null.
Quiegolani Zapotec: controlled subjects are overt pronouns.
Büli: controlled subjects are overt pronouns.
English has reflexives distinct from pronouns.
SMPM has reflexives distinct from pronouns (mí + clitic).
Haitian lacks distinct reflexives — reflexives surface as pronouns.
Equations
Instances For
The paper's four-way typology of copy control (§5, its own coinage of
all four labels). [PP06] supply the definition of
copy control — an overt element in the controlled position — and attest
three cases (Assamese, San Lucas Quiaviní Zapotec, Tongan), noting that
the Zapotec and Assamese patterns are themselves obligatory control.
The types split two ways: full-copy vs. pronominal copy control
(CopyControlType.isPronominal), the latter three-way by trigger. What
distinguishes the obligatory-pronominal type is unconditional
overtness (CopyControlType.unconditionallyOvert) — the
scope-sensitive type alternates with null PRO in the same OC
configurations ((60)).
- fullCopy : CopyControlType
Full copy: the controlled subject is a full DP copy of the controller (San Lucas Quiaviní Zapotec; Copala Triqui).
- logophoricPronominal : CopyControlType
Logophoric pronominal: an overt pronoun, in attitude reports only (Gengbe, Mandarin — whose PRO status Li 2021 doubts, per the paper).
- scopeSensitivePronominal : CopyControlType
Scope-sensitive pronominal: an overt pronoun triggered by scope-taking operators, alternating with null PRO otherwise (Italian, Hungarian, European Portuguese).
- obligatoryPronominal : CopyControlType
Instances For
Equations
- Ostrove2026.instDecidableEqCopyControlType x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Ostrove2026.instReprCopyControlType = { reprPrec := Ostrove2026.instReprCopyControlType.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
The typology's first cut: pronominal vs. full-copy copy control.
Equations
- Ostrove2026.CopyControlType.fullCopy.isPronominal = false
- x✝.isPronominal = true
Instances For
The obligatory-pronominal type's distinguishing property: overtness is unconditional, not contingent on attitudehood or a scope trigger.
Equations
Instances For
SMPM instantiates obligatory pronominal copy control: the controlled
subject is always an overt clitic pronoun showing the full OC
signature. Controlled subjects cannot bear focus — they must be clitic
pronouns, and clitics cannot bear focus ((65)/(67),
smpm_controlled_must_be_clitic) — which separates SMPM from the
scope-sensitive type.
Instances For
The clitic requirement, derived from the fragment and routed through the Cardinaletti–Starke deficiency order: the required controlled-subject class is strictly more deficient than every entry of the non-clitic series ((67)).
Exempt anaphors (reflexive forms used as possessors, outside the Condition A domain) are available in SMPM (74).
SMPM exempt anaphors cannot have quantified antecedents (75, 78).
Equations
- Ostrove2026.instDecidableEqEx86Item x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Ostrove2026.instReprEx86Item = { reprPrec := Ostrove2026.instReprEx86Item.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
The (86)–(87) control dependency: quantified controller position
0 to embedded clitic position 1.
Equations
- Ostrove2026.ex86Dependency = {(0, 1)}
Instances For
The attested occupants: exempt anaphors reject quantified antecedents ((78)), yet they are available in untensed subjunctives with quantified controllers ((86)–(87)) — so the embedded position holds a genuine referential pronoun, not a quantifier copy.
Equations
- Ostrove2026.ex86Occupant p = if p = 0 then Ostrove2026.Ex86Item.quantifierDP else Ostrove2026.Ex86Item.pronoun
Instances For
Movement is token identity, and the (86)–(87) occupants differ across the dependency — movement is refuted (§6, pp. 26–31): SMPM control is base-generated.
SMPM instantiates the universal: overt PRO and no pro-drop (the fragment's flag).
English satisfies the universal vacuously, whatever its pro-drop status: its PRO is null.
Büli instantiates the universal: overt PRO and no pro-drop.
SMPM clause types map to [Noo07]'s complement typology.
- Finite embedded → indicative complement (unrestricted TAM)
- Tensed subjunctive → subjunctive complement (irrealis only)
- Untensed subjunctive → subjunctive complement (irrealis, with equi-deletion / obligatory coreference)
All three are "balanced" in Noonan's terms — SMPM lacks morphologically nonfinite predicates entirely.
Equations
- Ostrove2026.smpmToNoonan Mixtec.SMPM.EmbeddedClauseType.finiteEmbedded = Complement.Coding.indicative
- Ostrove2026.smpmToNoonan Mixtec.SMPM.EmbeddedClauseType.tensedSubjunctive = Complement.Coding.subjunctive
- Ostrove2026.smpmToNoonan Mixtec.SMPM.EmbeddedClauseType.untensedSubjunctive = Complement.Coding.subjunctive
Instances For
SMPM's CTP classes map to Noonan's CTP classes.
The paper's predicate lists (27a–c) align with Noonan's semantic classification:
Finite embedded (27a):
- utterance: say (kà'àn), said (káchi), chat (ntatǔ'un)
- propAttitude: think (ka'án), believe (nakanini)
- commentative: be happy (kusijǐ ini), be sad (ntsi'i ini), regret (ntsiko ini)
- knowledge: know (kòni), wonder (kuntàà ini)
Tensed subjunctive (27b):
- desiderative: want (kòni), hope (ntatu), pray (nakwatu), agree (xiinka), refuse (xǐunka). 'Hate' (sǐso ini), 'be afraid' (iyì'bí), 'be scared' (kuntasí) are emotive predicates but select irrealis complements in SMPM, functioning like desideratives. 'Get the idea' (chikàà ini) is cognitive.
Untensed subjunctive (27c):
- phasal: start (kìxà), finish (ntsi'i), stop (xikwīn), continue (kò xikwīn)
- achievement: try (ntùkú), remember (nàkú'ún), forget (nantōso)
- modal: need (xiniñu'u)
- desiderative: like to (kutō)
- knowledge: know how to (kòni xá kasa), learn how to (sakwā'a)
- negative: not bother (kò ntaa)
Equations
- Ostrove2026.smpmCTPClass Mixtec.SMPM.EmbeddedClauseType.finiteEmbedded = [CTPClass.utterance, CTPClass.propAttitude, CTPClass.commentative, CTPClass.knowledge]
- Ostrove2026.smpmCTPClass Mixtec.SMPM.EmbeddedClauseType.tensedSubjunctive = [CTPClass.desiderative]
- Ostrove2026.smpmCTPClass Mixtec.SMPM.EmbeddedClauseType.untensedSubjunctive = [CTPClass.phasal, CTPClass.achievement, CTPClass.modal, CTPClass.desiderative, CTPClass.knowledge, CTPClass.negative]
Instances For
Noonan's reality status predicts SMPM clause type selection.
Irrealis CTPs select subjunctive complements (tensed or untensed); realis CTPs select indicative (finite embedded) complements.
This holds for the core cases: desiderative (want, hope) is irrealis and selects tensed subjunctive; phasal (start, stop) is realis but selects untensed subjunctive — an apparent exception that reflects the restructuring/monoclausal nature of phasal predicates.
[Ost26] groups SMPM with Gã ([All21]) and Büli
([Sul21]) as obligatory-pronominal copy control languages.
The Gã fragment and study predate this paper, so the cross-language
bridges live here (chronology: the later paper draws the
comparison), consuming Studies/Allotey2021.lean.
Gã instantiates obligatory pronominal copy control: the controlled
subject of an irrealis ni-clause is always an overt subject
proclitic showing the full OC signature ([All21]).
Instances For
Gã and SMPM occupy the same copy-control cell.
Gã syncretism row, derived from the Allotey2021 inventory: reflexive ×, controlled =, bound variable = .
Equations
Instances For
Gã patterns with SMPM in the syncretism typology: distinct reflexive, but controlled subjects and bound variables syncretic with the referential pronoun.
Gã sits in the same cell of the overt-PRO / pro-drop typology as SMPM: identical controlled-subject realization, identical pro-drop status.
Where the two overt-PRO languages part ways in [Noo07]'s typology: SMPM's complements are all balanced (finite or subjunctive), while Gã's control complement is a deranked bare-root infinitive.