Pancheva & Zubizarreta (2018): The Person Case Constraint

[PZ18] [Sel87]

The Person Case Constraint: The Syntactic Encoding of Perspective. Natural Language and Linguistic Theory 36: 1291–1337.

Summary

Empirical predictions of the P-Constraint theory (formalized in Syntax/Minimalism/PConstraint.lean) for the eight grammar instances P&Z discuss: five attested PCC varieties (strong, ultra-strong, weak, super-strong, me-first) plus three predicted varieties (PG1, PG2, PG3) that the four-parameter space generates.

Key derivations (beyond per-cell predictions)

• personHierarchy_from_features — the paper's central claim that the Person Hierarchy 1P > 2P > 3P is derived from the count of positive features in decomposePerson (§2.1, p. 1296), not stipulated.
• isLicit_imp_io_pov — the four parametric clauses are recovered as the conditions under which selecting the IO as point-of-view center satisfies the P-Constraint semantically (§6.3, eq. 48).
• pProminence_to_sellsRole — P&Z's identification of their P-Prominence values with [Sel87]'s logophoric roles (§6.2). This mapping is P&Z's specific theoretical claim, not a framework-neutral fact; [CM15] (page 10) reject the claim that pivot is the relevant role for clitic clusters.
• readings_disagree_on_proximate and mefirst_wedge_with_cm — P&Z's rejection (page 1308) of [CM15]'s unification of PCC and CLR (§ 10 below).

Forward references

This study is extended by [AZ25a] (study file AdamsonZompi2025.lean), who use the dual-feature distinction to argue that PCC effects diagnose interpretable (not agreement) person.

def PanchevaZubizarreta2018.pProminence_to_sellsRole : PCC.PProminence → Features.Logophoricity.LogophoricRole

P&Z's identification of P-Prominence values with [Sel87]'s logophoric roles (paper §6.2). This is the paper's theoretical claim, not a framework-neutral fact.

Equations

• PanchevaZubizarreta2018.pProminence_to_sellsRole PCC.PProminence.proximate = Features.Logophoricity.LogophoricRole.pivot
• PanchevaZubizarreta2018.pProminence_to_sellsRole PCC.PProminence.participant = Features.Logophoricity.LogophoricRole.self
• PanchevaZubizarreta2018.pProminence_to_sellsRole PCC.PProminence.author = Features.Logophoricity.LogophoricRole.source

P&Z's syntactic encoding: the interpretable person feature on Appl marks one DP as the point-of-view center. The theory-neutral PCC (Features/Person/PersonCaseConstraint.lean) is grounded here in that Appl model — a ⟨IO, DO⟩ is licit iff IO-as-POV-center is consistent with the Appl p-feature. (Formerly PConstraint.lean's §9 grounding; it has no consumers outside this study, so it lives with the paper that motivates it.)

structure PanchevaZubizarreta2018.ApplDomain : Type

A minimal model of the Appl phase: the two arguments and the chosen POV center.

• io : Person

  The indirect-object argument introduced by Appl.

• do_ : Person

  The direct-object argument inside VP.

• povCenter : Person

  The DP selected as point-of-view center within the phase.

def PanchevaZubizarreta2018.instDecidableEqApplDomain.decEq (x✝ x✝¹ : ApplDomain) : Decidable (x✝ = x✝¹)

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance PanchevaZubizarreta2018.instDecidableEqApplDomain : DecidableEq ApplDomain

Equations

• PanchevaZubizarreta2018.instDecidableEqApplDomain = PanchevaZubizarreta2018.instDecidableEqApplDomain.decEq

def PanchevaZubizarreta2018.instReprApplDomain.repr : ApplDomain → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance PanchevaZubizarreta2018.instReprApplDomain : Repr ApplDomain

Equations

• PanchevaZubizarreta2018.instReprApplDomain = { reprPrec := PanchevaZubizarreta2018.instReprApplDomain.repr }

def PanchevaZubizarreta2018.ApplDomain.povIsIO (a : ApplDomain) : Prop

The IO is the canonical POV-center candidate ([PZ18] p. 1320).

Equations

• a.povIsIO = (a.povCenter = a.io)

@[implicit_reducible]

instance PanchevaZubizarreta2018.instDecidablePovIsIO (a : ApplDomain) : Decidable a.povIsIO

Equations

• PanchevaZubizarreta2018.instDecidablePovIsIO a = PanchevaZubizarreta2018.instDecidablePovIsIO._aux_1 a

def PanchevaZubizarreta2018.PConstraintSatisfied (g : PCC.PCCGrammar) (a : ApplDomain) : Prop

The P-Constraint as a predicate over an Appl domain.

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance PanchevaZubizarreta2018.instDecidablePConstraintSatisfied (g : PCC.PCCGrammar) (a : ApplDomain) : Decidable (PConstraintSatisfied g a)

Equations

• PanchevaZubizarreta2018.instDecidablePConstraintSatisfied g a = PanchevaZubizarreta2018.instDecidablePConstraintSatisfied._aux_1 g a

theorem PanchevaZubizarreta2018.isLicit_iff_exists_appl_satisfying (g : PCC.PCCGrammar) (io do_ : Person) : PCC.IsLicit g io do_ ↔ ∃ (a : ApplDomain), a.io = io ∧ a.do_ = do_ ∧ PConstraintSatisfied g a

Central derivation. ⟨IO, DO⟩ is licit iff some Appl domain over them — IO as POV center — satisfies the P-Constraint. The four parametric clauses are not stipulated verdicts; they are the conditions under which IO-as-POV-center is consistent with the interpretable person feature on Appl.

def PanchevaZubizarreta2018.positiveFeatureCount (dp : Minimalist.DecomposedPerson) : ℕ

Number of positive features in a person decomposition. By the implicational hierarchy of (paper eq. 11), 1P bears all three (proximate, participant, author), 2P bears two, 3P bears none.

Equations

• PanchevaZubizarreta2018.positiveFeatureCount dp = ((if dp.hasProximate = true then 1 else 0) + if dp.hasParticipant = true then 1 else 0) + if dp.hasAuthor = true then 1 else 0

theorem PanchevaZubizarreta2018.positiveFeatureCount_values : positiveFeatureCount (Minimalist.decomposePerson Person.first) = 3 ∧ positiveFeatureCount (Minimalist.decomposePerson Person.second) = 2 ∧ positiveFeatureCount (Minimalist.decomposePerson Person.third) = 0

The numeric values of positiveFeatureCount per person.

theorem PanchevaZubizarreta2018.personHierarchy_from_features (p q : Person) : p.prominence ≤ q.prominence ↔ positiveFeatureCount (Minimalist.decomposePerson p) ≤ positiveFeatureCount (Minimalist.decomposePerson q)

The Person Hierarchy is derived, not stipulated (paper §2.1, p. 1296: "We seek to derive it from more fundamental principles"). The order induced by Person.prominence (1P > 2P > 3P) coincides with the order induced by the count of positive features in the decomposition.

Note: The two functions are not pointwise equal (3 vs 2, 2 vs 1, 0 vs 0) because rank collapses the SAP/non-SAP gap, but the orders match.

theorem PanchevaZubizarreta2018.strong_predictions : PCC.licitFinset PCC.strongGrammar = {(Person.first, Person.third), (Person.second, Person.third), (Person.third, Person.third)}

Strong PCC (paper §4.1.1, eq. 14a): DO must be 3P.

theorem PanchevaZubizarreta2018.ultra_predictions : PCC.licitFinset PCC.ultraStrongGrammar = {(Person.first, Person.first), (Person.first, Person.second), (Person.first, Person.third), (Person.second, Person.third), (Person.third, Person.third)}

Ultra-strong PCC (§4.1.2, eq. 14d): adds P-Primacy, so 1P-IO can rescue 1P/2P DO. ⟨1,2⟩ allowed but ⟨2,1⟩ banned.

theorem PanchevaZubizarreta2018.weak_predictions : PCC.licitFinset PCC.weakGrammar = {(Person.first, Person.first), (Person.first, Person.second), (Person.first, Person.third), (Person.second, Person.first), (Person.second, Person.second), (Person.second, Person.third), (Person.third, Person.third)}

Weak PCC (§4.1.3, eq. 14b): drops P-Uniqueness, so any SAP IO licenses any DO. Bans only 3P-IO with 1P/2P DO.

theorem PanchevaZubizarreta2018.super_predictions : PCC.licitFinset PCC.superStrongGrammar = {(Person.first, Person.third), (Person.second, Person.third)}

Super-strong PCC (§4.2, eq. 14e): IO must be SAP, DO must be 3P. Strictly the most restrictive variety.

theorem PanchevaZubizarreta2018.mefirst_predictions : PCC.licitFinset PCC.meFirstGrammar = {(Person.first, Person.second), (Person.first, Person.third), (Person.second, Person.second), (Person.second, Person.third), (Person.third, Person.second), (Person.third, Person.third)}

Me-first PCC (§4.3, eq. 14c): bans 1P DO with non-1P IO; restricted domain exempts ⟨2P,2P⟩, ⟨2P,3P⟩, ⟨3P,2P⟩, ⟨3P,3P⟩ entirely.

NB: The implementation also bans ⟨1P,1P⟩ via P-Uniqueness. The paper (§4.3, p. 1314: "allows all other combinations") does not explicitly address this case; see mefirst_one_one_excluded below.

theorem PanchevaZubizarreta2018.pg1_predictions : PCC.licitFinset PCC.pg1Grammar = {(Person.first, Person.first), (Person.first, Person.second), (Person.first, Person.third), (Person.second, Person.third)}

PG1 (predicted, §4.5, eq. 32a-ii): [+participant] + P-Primacy.

theorem PanchevaZubizarreta2018.pg2_predictions : PCC.licitFinset PCC.pg2Grammar = {(Person.first, Person.first), (Person.first, Person.second), (Person.first, Person.third), (Person.second, Person.first), (Person.second, Person.second), (Person.second, Person.third)}

PG2 (predicted, §4.5, eq. 32b): [+participant], no P-Uniqueness.

theorem PanchevaZubizarreta2018.pg3_predictions : PCC.licitFinset PCC.pg3Grammar = {(Person.first, Person.second), (Person.first, Person.third)}

PG3 (predicted, §4.5, eq. 33): [+author] with unrestricted domain. Only 1P-IO is licensed; uniqueness then rules out 1P-DO.

theorem PanchevaZubizarreta2018.licit_counts : PCC.licitCount PCC.strongGrammar = 3 ∧ PCC.licitCount PCC.ultraStrongGrammar = 5 ∧ PCC.licitCount PCC.weakGrammar = 7 ∧ PCC.licitCount PCC.superStrongGrammar = 2 ∧ PCC.licitCount PCC.meFirstGrammar = 6 ∧ PCC.licitCount PCC.pg1Grammar = 4 ∧ PCC.licitCount PCC.pg2Grammar = 6 ∧ PCC.licitCount PCC.pg3Grammar = 2

theorem PanchevaZubizarreta2018.markedness_table : PCC.parameterDepartures PCC.strongGrammar = 0 ∧ PCC.parameterDepartures PCC.ultraStrongGrammar = 1 ∧ PCC.parameterDepartures PCC.weakGrammar = 1 ∧ PCC.parameterDepartures PCC.superStrongGrammar = 1 ∧ PCC.parameterDepartures PCC.pg3Grammar = 1 ∧ PCC.parameterDepartures PCC.meFirstGrammar = 2 ∧ PCC.parameterDepartures PCC.pg1Grammar = 2 ∧ PCC.parameterDepartures PCC.pg2Grammar = 2

The markedness rank of each grammar as the number of parameter departures from the strong PCC default. Strong is the unique 0-rank grammar; the four 1-rank grammars (ultra/weak/super/pg3) are the "minimal departures"; the three 2-rank grammars (me-first/pg1/pg2) are doubly marked.

theorem PanchevaZubizarreta2018.strong_le_weak : PCC.strongGrammar ≤ PCC.weakGrammar

Strong PCC entails Weak PCC: every cell licit in strong is licit in weak. Falls out of the Preorder PCCGrammar instance.

theorem PanchevaZubizarreta2018.strong_le_ultra : PCC.strongGrammar ≤ PCC.ultraStrongGrammar

Strong PCC entails Ultra-strong PCC.

theorem PanchevaZubizarreta2018.super_le_strong : PCC.superStrongGrammar ≤ PCC.strongGrammar

Super-strong PCC entails Strong PCC: super-strong's prominence on [+participant] is strictly more restrictive than strong's on [+proximate].

theorem PanchevaZubizarreta2018.prominence_logophoric_role : pProminence_to_sellsRole PCC.PProminence.proximate = Features.Logophoricity.LogophoricRole.pivot ∧ pProminence_to_sellsRole PCC.PProminence.participant = Features.Logophoricity.LogophoricRole.self ∧ pProminence_to_sellsRole PCC.PProminence.author = Features.Logophoricity.LogophoricRole.source

P&Z's claim (§6.2): each P-Prominence setting corresponds to a logophoric role of [Sel87]. This is the paper's theoretical reading. [CM15] dispute that pivot is the relevant role for clitic clusters; the bridge file Anaphora/Antilogophoricity.lean documents this disagreement explicitly.

theorem PanchevaZubizarreta2018.family_logophoric_assignments : pProminence_to_sellsRole PCC.strongGrammar.prominence = Features.Logophoricity.LogophoricRole.pivot ∧ pProminence_to_sellsRole PCC.ultraStrongGrammar.prominence = Features.Logophoricity.LogophoricRole.pivot ∧ pProminence_to_sellsRole PCC.weakGrammar.prominence = Features.Logophoricity.LogophoricRole.pivot ∧ pProminence_to_sellsRole PCC.superStrongGrammar.prominence = Features.Logophoricity.LogophoricRole.self ∧ pProminence_to_sellsRole PCC.meFirstGrammar.prominence = Features.Logophoricity.LogophoricRole.source ∧ pProminence_to_sellsRole PCC.pg3Grammar.prominence = Features.Logophoricity.LogophoricRole.source

Under P&Z's reading: the five attested grammars and the [+author]-prominence predicted family map onto Sells's hierarchy as strong/ultra/weak ⇒ pivot, super ⇒ self, me-first/pg3 ⇒ source.

theorem PanchevaZubizarreta2018.isLicit_imp_io_pov (g : PCC.PCCGrammar) (io do_ : Person) : PCC.IsLicit g io do_ → PConstraintSatisfied g { io := io, do_ := do_, povCenter := io }

Whenever ⟨IO, DO⟩ is licit, selecting the IO as point-of-view center yields an Appl domain that semantically satisfies the P-Constraint. The four parametric clauses in (12) are not free-standing stipulations: they are precisely the conditions on IO-as-POV consistency.

theorem PanchevaZubizarreta2018.io_pov_imp_isLicit (g : PCC.PCCGrammar) (io do_ : Person) : PConstraintSatisfied g { io := io, do_ := do_, povCenter := io } → PCC.IsLicit g io do_

Conversely, if any Appl domain over ⟨io, do_⟩ with IO as POV center satisfies the P-Constraint, the combination is licit. Together with isLicit_imp_io_pov, this characterizes IsLicit semantically.

theorem PanchevaZubizarreta2018.pov_principle_at_io_attitude_grammar : Features.Logophoricity.pointOfViewPrinciple true true = true

For [+participant] and [+author] grammars, the IO-as-POV semantics automatically interprets the IO as an attitude holder (self or source). The Point-of-View Principle (eq. 48) then holds with the AH = POV identification.

theorem PanchevaZubizarreta2018.italian_weak_glidat_tiacc : PanchevaZubizarreta2018.cliticLevel?✝ Italian.Pronouns.gli_dat.person = some Person.third ∧ PanchevaZubizarreta2018.cliticLevel?✝ Italian.Pronouns.ti_acc.person = some Person.second ∧ ¬PCC.IsLicit PCC.weakGrammar Person.third Person.second

Italian dative gli is 3rd person; accusative ti is 2nd. The weak PCC prediction ⟨3,2⟩ ⇒ illicit is therefore satisfied at these two actual clitic forms.

theorem PanchevaZubizarreta2018.italian_weak_tidat_lacl : PanchevaZubizarreta2018.cliticLevel?✝ Italian.Pronouns.ti_dat.person = some Person.second ∧ PanchevaZubizarreta2018.cliticLevel?✝ Italian.Pronouns.la_cl.person = some Person.third ∧ PCC.IsLicit PCC.weakGrammar Person.second Person.third

The Italian licit pair ti la (2.DAT > 3.ACC) under weak PCC.

theorem PanchevaZubizarreta2018.spanish_weak_tedat_meacc : PanchevaZubizarreta2018.cliticLevel?✝ Spanish.Clitics.te_dat.person = some Person.second ∧ PanchevaZubizarreta2018.cliticLevel?✝ Spanish.Clitics.me_acc.person = some Person.first ∧ PCC.IsLicit PCC.weakGrammar Person.second Person.first

Spanish te me (2.DAT > 1.ACC) is licit under weak PCC. The interpretable persons are read off the actual te_dat/me_acc clitic forms in Fragments/Spanish/Clitics.lean.

theorem PanchevaZubizarreta2018.spanish_weak_medat_teacc : PanchevaZubizarreta2018.cliticLevel?✝ Spanish.Clitics.me_dat.person = some Person.first ∧ PanchevaZubizarreta2018.cliticLevel?✝ Spanish.Clitics.te_acc.person = some Person.second ∧ PCC.IsLicit PCC.weakGrammar Person.first Person.second

Spanish me te (1.DAT > 2.ACC) is also licit (weak PCC ⟨1,2⟩).

theorem PanchevaZubizarreta2018.spanish_weak_ledat_meacc_banned : PanchevaZubizarreta2018.cliticLevel?✝ Spanish.Clitics.le_dat.person = some Person.third ∧ PanchevaZubizarreta2018.cliticLevel?✝ Spanish.Clitics.me_acc.person = some Person.first ∧ ¬PCC.IsLicit PCC.weakGrammar Person.third Person.first

Spanish le me (3.DAT > 1.ACC) is illicit (paper ex. 24).

theorem PanchevaZubizarreta2018.french_strong_examples : ¬PCC.IsLicit PCC.strongGrammar Person.third Person.first ∧ PCC.IsLicit PCC.strongGrammar Person.second Person.third ∧ PCC.IsLicit PCC.strongGrammar Person.third Person.third

French strong PCC (§4.1.1, ex. 16): *Elle te me présentera. No Fragments/French/Pronouns.lean exists; theorem reads off the parameter settings rather than fragment data.

theorem PanchevaZubizarreta2018.catalan_ultra_strong_examples : PCC.IsLicit PCC.ultraStrongGrammar Person.first Person.second ∧ ¬PCC.IsLicit PCC.ultraStrongGrammar Person.second Person.first

Catalan ultra-strong PCC (§4.1.2, ex. 20): the ⟨1,2⟩ vs ⟨2,1⟩ asymmetry that distinguishes ultra-strong from strong.

theorem PanchevaZubizarreta2018.kambera_super_strong_examples : PCC.IsLicit PCC.superStrongGrammar Person.first Person.third ∧ PCC.IsLicit PCC.superStrongGrammar Person.second Person.third ∧ ¬PCC.IsLicit PCC.superStrongGrammar Person.third Person.third ∧ ¬PCC.IsLicit PCC.superStrongGrammar Person.first Person.second

Kambera super-strong PCC (§4.2, ex. 27): IO must be SAP, DO must be 3P; ⟨3,3⟩ is also banned.

theorem PanchevaZubizarreta2018.bulgarian_me_first_examples : PCC.IsLicit PCC.meFirstGrammar Person.third Person.second ∧ ¬PCC.IsLicit PCC.meFirstGrammar Person.second Person.first

Bulgarian me-first PCC (§4.3, ex. 29): only ⟨2,1⟩ and ⟨3,1⟩ banned; crucially, ⟨3,2⟩ is licit (where it is illicit in all [+proximate] varieties).

theorem PanchevaZubizarreta2018.mefirst_one_one_excluded : ¬PCC.IsLicit PCC.meFirstGrammar Person.first Person.first

The implementation rules out ⟨1P, 1P⟩ in me-first by P-Uniqueness on [+author]: both arguments are [+author], so neither can be uniquely the perspectival source. The paper (§4.3) explicitly bans only ⟨3,1⟩ and ⟨2,1⟩, leaving ⟨1,1⟩ unaddressed. The implementation's verdict follows from (12c) applied to two coreferential [+author] DPs.

theorem PanchevaZubizarreta2018.mefirst_three_three_exempt : PCC.IsLicit PCC.meFirstGrammar Person.third Person.third

The me-first family does not show *⟨3,3⟩ effects: 3P-IO with 3P-DO is domain-exempt (no [+author] DP triggers the constraint), so the spurious-se restriction documented for [+proximate] varieties is structurally unavailable here (paper §4.4).

def PanchevaZubizarreta2018.pProminence_to_cmCenter : PCC.PProminence → CharnavelMateu2015.LogoCenter

P&Z's reading of the dative clitic — as a pivot (Sells's broadest role) — is incompatible with [CM15]'s reading (page 10), which assigns the dative clitic to empathyLocus and rejects pivot as relevant for clitic clusters. The two readings map .proximate to incompatible places.

Equations

• PanchevaZubizarreta2018.pProminence_to_cmCenter PCC.PProminence.proximate = CharnavelMateu2015.LogoCenter.empathyLocus
• PanchevaZubizarreta2018.pProminence_to_cmCenter PCC.PProminence.participant = CharnavelMateu2015.LogoCenter.discourseParticipant
• PanchevaZubizarreta2018.pProminence_to_cmCenter PCC.PProminence.author = CharnavelMateu2015.LogoCenter.discourseParticipant

theorem PanchevaZubizarreta2018.readings_disagree_on_proximate : pProminence_to_sellsRole PCC.PProminence.proximate = Features.Logophoricity.LogophoricRole.pivot ∧ pProminence_to_cmCenter PCC.PProminence.proximate = CharnavelMateu2015.LogoCenter.empathyLocus ∧ pProminence_to_sellsRole PCC.PProminence.proximate ≠ Features.Logophoricity.LogophoricRole.self ∧ pProminence_to_sellsRole PCC.PProminence.proximate ≠ Features.Logophoricity.LogophoricRole.source

theorem PanchevaZubizarreta2018.mefirst_wedge_with_cm : PCC.IsLicit PCC.meFirstGrammar Person.third Person.third ∧ CharnavelMateu2015.CLRViolated (CharnavelMateu2015.LogoCenter.empathyLocus, CharnavelMateu2015.LogoCenter.attitudeHolder)

The me-first wedge. P&Z predict me-first speakers should lack CLR effects: the P-Constraint marks the IO as a perspectival centre only when triggered, and me-first restricts the trigger to contexts with a [+author] DP. In ⟨3,3⟩ contexts no centre is marked, so no perspective conflict.

C&M predict me-first speakers should still show CLR effects: dative clitics are inherently empathy loci (paper §3.5.1), independent of any P-Constraint setting. The accusative clitic read de se is then an attitude holder, and the antilogophoric clash obtains regardless of the syntactic licensing of the IO.

Formally: under P&Z's account, ⟨3,3⟩ in a me-first grammar is licit (mefirst_three_three_exempt) and there is no separate CLR predicate to check. Under C&M's account, the configuration is CLRViolated.

Resolution requires Bulgarian/Romanian me-first speakers tested on de se readings of accusative clitics in 3.DAT 3.ACC clusters; P&Z (page 1316) cite indirect evidence supporting their position; C&M did not test me-first varieties.