Pancheva and Zubizarreta (2018): The Person Case Constraint

This file formalizes the predictions of Pancheva and Zubizarreta's P-Constraint, the proposal that Person Case Constraint effects are the syntactic encoding of perspective: an interpretable person feature on the applicative head marks the indirect object as the point-of-view center of its domain, and a clitic combination is licit exactly when selecting the indirect object as that center satisfies the constraint (ApplDomain, PConstraintSatisfied, isLicit_iff_pConstraint). The Person Hierarchy 1 > 2 > 3 is derived from the count of positive person features rather than stipulated (personHierarchy_from_features). The constraint's four parameters, the prominence threshold, the domain of application, P-Uniqueness, and P-Primacy, generate the five attested varieties, strong, ultra-strong, weak, super-strong, and me-first, together with three predicted ones, each checked against the paper's examples (strong_predictions and its siblings, french_strong_examples and its siblings), and the attested varieties are ordered by containment of their licit sets (strong_le_ultra, ultra_le_weak, super_le_strong). The three prominence values correspond to Sells's logophoric roles pivot, self, and source (prominenceToSellsRole).

Implementation notes

The grammars and the licitness predicate live in Syntax/Agreement/PersonCaseConstraint. The Italian and Spanish clitic combinations are read off the Fragments, the French, Catalan, Kambera, and Bulgarian ones off the parameter settings, since no clitic Fragment exists for those languages. The paper's separation of the Clitic Logophoric Restriction from the PCC, against the unification Charnavel and Mateu propose, the two phenomena sharing only the point-of-view marking on the applicative head, is not represented: the restriction is a semantic constraint outside this model.

TODO

The substrate's P-Primacy checks only the indirect object's [+author], so the ultra-strong grammar licenses the ⟨1, 1⟩ combination that the descriptive statement (14d) bans (ultra_one_one_licensed); the paper does not walk the mechanism through that cell.

References

• [PZ18]
• [Sel87]
• [CM15b]

The applicative domain

structure PanchevaZubizarreta2018.ApplDomain : Type

The applicative phase holds the indirect object introduced by Appl, the direct object inside VP, and the argument selected as point-of-view center.

• io : Person
• do_ : Person
• povCenter : Person

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.

@[instance_reducible]

instance PanchevaZubizarreta2018.instDecidableEqApplDomain : DecidableEq ApplDomain

Equations

• PanchevaZubizarreta2018.instDecidableEqApplDomain = PanchevaZubizarreta2018.instDecidableEqApplDomain.decEq

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

The P-Constraint holds over an Appl domain when the domain is exempt, or the indirect object is the point-of-view center, satisfies the prominence threshold, and P-Uniqueness holds or P-Primacy rescues it.

Equations

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

@[instance_reducible]

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

Equations

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

theorem PanchevaZubizarreta2018.isLicit_iff_pConstraint (g : PCC.Grammar) (s : Clause.Scenario Person) : PCC.IsLicit g s ↔ PConstraintSatisfied g { io := s.high, do_ := s.low, povCenter := s.high }

A clitic combination is licit exactly when the domain with the indirect object as point-of-view center satisfies the P-Constraint, the parametric clauses being the conditions on that selection, the only felicitous derivation being the one in which Appl agrees with the argument it introduces (§4.5).

The Person Hierarchy

def PanchevaZubizarreta2018.positiveFeatures (p : Person) : Finset PCC.ProminenceThreshold

The positive person features of a person are the prominence values it meets by its own features, first person bearing proximate, participant, and author, second person the first two, and third person none (11).

Equations

• PanchevaZubizarreta2018.positiveFeatures p = {x : PCC.ProminenceThreshold | x.InherentlyMetBy p}

theorem PanchevaZubizarreta2018.personHierarchy_from_features (p q : Person) : p.prominence ≤ q.prominence ↔ (positiveFeatures p).card ≤ (positiveFeatures q).card

The Person Hierarchy is derived, the prominence order coinciding with the order by count of positive features (§2.1).

The varieties (§4)

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

Under the strong PCC (14a) the direct object must be third person.

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

Under the ultra-strong PCC (14d) P-Primacy lets a first-person indirect object rescue a local direct object, so ⟨1, 2⟩ is licit and ⟨2, 1⟩ is not.

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

Under the weak PCC (14b), without P-Uniqueness, any local indirect object licenses any direct object; only a third-person indirect object with a local direct object is banned.

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

Under the super-strong PCC (14e) the indirect object must be local and the direct object third person.

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

Under the me-first PCC (14c) a first-person direct object needs a first-person indirect object, and the restricted domain exempts every combination without a first-person argument; ⟨1, 1⟩ falls to P-Uniqueness.

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

The first predicted variety (32a) has [+participant] prominence with P-Primacy.

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

The second predicted variety (32b) has [+participant] prominence without P-Uniqueness.

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

The third predicted variety (33) has [+author] prominence over an unrestricted domain, so only a first-person indirect object licenses, and P-Uniqueness then excludes ⟨1, 1⟩.

theorem PanchevaZubizarreta2018.restricted_participant_surfaces_as_strong : PCC.licitFinset { prominence := PCC.ProminenceThreshold.participant, restrictedDomain := true, primacy_le_uniqueness := restricted_participant_surfaces_as_strong._proof_1 } = PCC.licitFinset PCC.strongGrammar

Restricting the domain to applicatives with a participant argument while setting the threshold to [+participant] yields a grammar whose licit set is the strong PCC's (§4.5); the paper locates the residual difference in Clitic Logophoric Restriction effects on third-person combinations, outside this model.

Containment among the attested varieties (§4.5)

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

Activating P-Primacy only enlarges the licit region.

theorem PanchevaZubizarreta2018.ultra_le_weak : PCC.ultraStrongGrammar ≤ PCC.weakGrammar

Dropping P-Uniqueness enlarges it further.

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

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

The [+participant] threshold of the super-strong variety is stricter than the [+proximate] threshold of the strong one.

Logophoric roles (§6.2)

def PanchevaZubizarreta2018.prominenceToSellsRole : PCC.ProminenceThreshold → Reference.LogophoricRole

The prominence values are the logophoric roles of [Sel87], proximate arguments being pivots, participants selves, and authors sources.

Equations

• PanchevaZubizarreta2018.prominenceToSellsRole PCC.ProminenceThreshold.proximate = Reference.LogophoricRole.pivot
• PanchevaZubizarreta2018.prominenceToSellsRole PCC.ProminenceThreshold.participant = Reference.LogophoricRole.self
• PanchevaZubizarreta2018.prominenceToSellsRole PCC.ProminenceThreshold.author = Reference.LogophoricRole.source

theorem PanchevaZubizarreta2018.family_logophoric_assignments : prominenceToSellsRole PCC.strongGrammar.prominence = Reference.LogophoricRole.pivot ∧ prominenceToSellsRole PCC.ultraStrongGrammar.prominence = Reference.LogophoricRole.pivot ∧ prominenceToSellsRole PCC.weakGrammar.prominence = Reference.LogophoricRole.pivot ∧ prominenceToSellsRole PCC.superStrongGrammar.prominence = Reference.LogophoricRole.self ∧ prominenceToSellsRole PCC.meFirstGrammar.prominence = Reference.LogophoricRole.source ∧ prominenceToSellsRole PCC.pg3Grammar.prominence = Reference.LogophoricRole.source

On the role hierarchy the strong family requires a pivot, the super-strong variety a self, and me-first and the [+author] predicted variety a source.

The paper's examples (§4)

theorem PanchevaZubizarreta2018.italian_weak_glidat_tiacc : ∃ io ∈ Italian.Pronouns.gli.person, ∃ do_ ∈ Italian.Pronouns.ti_acc.person, ¬PCC.IsLicit PCC.weakGrammar { high := io, low := do_ }

Italian dative gli is third person and accusative ti second, so the weak PCC bans the pair.

theorem PanchevaZubizarreta2018.italian_weak_tidat_lacl : ∃ io ∈ Italian.Pronouns.ti_dat.person, ∃ do_ ∈ Italian.Pronouns.la.person, PCC.IsLicit PCC.weakGrammar { high := io, low := do_ }

Italian ti la, second-person dative over third-person accusative, is licit.

theorem PanchevaZubizarreta2018.spanish_weak_tedat_meacc : ∃ io ∈ Spanish.Clitics.te_dat.person, ∃ do_ ∈ Spanish.Clitics.me_acc.person, PCC.IsLicit PCC.weakGrammar { high := io, low := do_ }

Spanish te me (23), second-person dative over first-person accusative, is licit in the weak variety.

theorem PanchevaZubizarreta2018.spanish_weak_medat_teacc : ∃ io ∈ Spanish.Clitics.me_dat.person, ∃ do_ ∈ Spanish.Clitics.te_acc.person, PCC.IsLicit PCC.weakGrammar { high := io, low := do_ }

Spanish me te, first-person dative over second-person accusative, is licit as well.

theorem PanchevaZubizarreta2018.spanish_weak_ledat_meacc_banned : ∃ io ∈ Spanish.Clitics.le.person, ∃ do_ ∈ Spanish.Clitics.me_acc.person, ¬PCC.IsLicit PCC.weakGrammar { high := io, low := do_ }

Spanish me le (24), third-person dative over first-person accusative, is banned.

theorem PanchevaZubizarreta2018.french_strong_examples : ¬PCC.IsLicit PCC.strongGrammar { high := Person.third, low := Person.first } ∧ PCC.IsLicit PCC.strongGrammar { high := Person.second, low := Person.third } ∧ PCC.IsLicit PCC.strongGrammar { high := Person.third, low := Person.third }

In French (16), a strong-PCC language, a third-person dative with a first-person accusative is out, while a second-person dative with a third-person accusative and two third persons are in.

theorem PanchevaZubizarreta2018.catalan_ultra_strong_examples : PCC.IsLicit PCC.ultraStrongGrammar { high := Person.first, low := Person.second } ∧ ¬PCC.IsLicit PCC.ultraStrongGrammar { high := Person.second, low := Person.first }

Catalan (20), an ultra-strong-PCC language, shows the ⟨1, 2⟩ against ⟨2, 1⟩ asymmetry that separates the ultra-strong from the strong variety.

theorem PanchevaZubizarreta2018.kambera_super_strong_examples : PCC.IsLicit PCC.superStrongGrammar { high := Person.first, low := Person.third } ∧ PCC.IsLicit PCC.superStrongGrammar { high := Person.second, low := Person.third } ∧ ¬PCC.IsLicit PCC.superStrongGrammar { high := Person.third, low := Person.third } ∧ ¬PCC.IsLicit PCC.superStrongGrammar { high := Person.first, low := Person.second }

In Kambera (27), a super-strong-PCC language, the indirect object must be local and the direct object third person, so ⟨3, 3⟩ is banned as well.

theorem PanchevaZubizarreta2018.bulgarian_me_first_examples : PCC.IsLicit PCC.meFirstGrammar { high := Person.third, low := Person.second } ∧ ¬PCC.IsLicit PCC.meFirstGrammar { high := Person.second, low := Person.first }

In Bulgarian (29), a me-first-PCC language, ⟨3, 2⟩ is licit, where every [+proximate] variety bans it, and ⟨2, 1⟩ is not.

The ⟨1, 1⟩ corner

theorem PanchevaZubizarreta2018.mefirst_one_one_excluded : ¬PCC.IsLicit PCC.meFirstGrammar { high := Person.first, low := Person.first }

Me-first bans two first-person arguments by P-Uniqueness on [+author], since neither is uniquely the perspectival source (§4.5).

theorem PanchevaZubizarreta2018.ultra_one_one_licensed : PCC.IsLicit PCC.ultraStrongGrammar { high := Person.first, low := Person.first }

The substrate's P-Primacy checks only the indirect object's [+author], so the ultra-strong grammar licenses ⟨1, 1⟩ where the descriptive statement (14d) bans it.

theorem PanchevaZubizarreta2018.mefirst_three_three_exempt : PCC.IsLicit PCC.meFirstGrammar { high := Person.third, low := Person.third }

Me-first shows no ⟨3, 3⟩ effect. Without a first-person argument the domain is exempt, so the spurious-se restriction of the [+proximate] varieties is unavailable (§4.4).