Charnavel & Mateu (2015): The Clitic Logophoric Restriction

@cite{charnavel-mateu-2015} @cite{sells-1987} @cite{kuno-1987}

The Clitic Binding Restriction Revisited: Evidence for Antilogophoricity. The Linguistic Review 32(4).

Summary

@cite{charnavel-mateu-2015} (C&M) reanalyze the restriction on accusative clitics in Romance clitic clusters. Earlier work (Bhatt & Šimík 2009) attributed the restriction to binding; C&M's grammaticality experiment (97 French + 35 Spanish speakers, 9 conditions) shows that the relevant factor is antilogophoricity, not binding. Their generalization (eq. 26):

Clitic Logophoric Restriction (CLR): When a third person dative clitic and an accusative clitic co-occur in a cluster, the accusative clitic cannot corefer with a logophoric center.

Local apparatus

C&M propose a three-level hierarchy of logophoric centers (eq. 53–54): discourse participant > empathy locus > attitude holder. Each center is characterised by a feature subset of {A, B, C} (eq. 63):

• discourse participant = {A, B}
• empathy locus = {B, C}
• attitude holder = {C}

Antilogophoric clash occurs when two centers in the same domain share a feature (equivalent to: two adjacent-or-identical positions on the hierarchy). PCC and CLR are both instances:

• PCC = discourse participant + empathy locus (share B)
• CLR = empathy locus + attitude holder (share C)

Disagreement with @cite{pancheva-zubizarreta-2018}

C&M unify CLR and PCC under one mechanism. P&Z (page 1308) explicitly disagree: "We do not think the CLR and the PCC should be unified along the lines suggested by Charnavel and Mateu (2015). The two phenomena are related but nevertheless distinct." The cross-paper bridge file Phenomena/Anaphora/Antilogophoricity.lean documents this disagreement explicitly.

inductive Studies.Anaphora.CharnavelMateu2015.LogoCenter : Type

Three types of logophoric center, ordered by degree of perspective integration in the discourse (paper eq. 54: discourse participant > empathy locus > attitude holder).

• discourseParticipant : LogoCenter

  Speaker / addressee — directly defining the discourse.

• empathyLocus : LogoCenter

  Event participant the speaker empathizes with (Kuno's empathy locus). In Romance, typically the 3rd-person dative clitic.

• attitudeHolder : LogoCenter

  Attitude holder whose thoughts/discourse are reported. In Romance, typically a 3rd-person accusative clitic read de se.

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidableEqLogoCenter : DecidableEq LogoCenter

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidableEqLogoCenter x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

def Studies.Anaphora.CharnavelMateu2015.instReprLogoCenter.repr : LogoCenter → ℕ → Std.Format

Equations

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

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instReprLogoCenter : Repr LogoCenter

Equations

• Studies.Anaphora.CharnavelMateu2015.instReprLogoCenter = { reprPrec := Studies.Anaphora.CharnavelMateu2015.instReprLogoCenter.repr }

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instFintypeLogoCenter : Fintype LogoCenter

Equations

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

inductive Studies.Anaphora.CharnavelMateu2015.LogoFeature : Type

The three abstract logophoric features. B expresses the speaker-component (shared by discourse participants and empathy loci); C expresses perspectival distance from the speaker (shared by empathy loci and attitude holders).

• A : LogoFeature
• B : LogoFeature
• C : LogoFeature

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidableEqLogoFeature : DecidableEq LogoFeature

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidableEqLogoFeature x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

def Studies.Anaphora.CharnavelMateu2015.instReprLogoFeature.repr : LogoFeature → ℕ → Std.Format

Equations

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

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instReprLogoFeature : Repr LogoFeature

Equations

• Studies.Anaphora.CharnavelMateu2015.instReprLogoFeature = { reprPrec := Studies.Anaphora.CharnavelMateu2015.instReprLogoFeature.repr }

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instFintypeLogoFeature : Fintype LogoFeature

Equations

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

def Studies.Anaphora.CharnavelMateu2015.LogoCenter.features : LogoCenter → Finset LogoFeature

Feature decomposition (paper eq. 63).

Equations

• Studies.Anaphora.CharnavelMateu2015.LogoCenter.discourseParticipant.features = {Studies.Anaphora.CharnavelMateu2015.LogoFeature.A, Studies.Anaphora.CharnavelMateu2015.LogoFeature.B}
• Studies.Anaphora.CharnavelMateu2015.LogoCenter.empathyLocus.features = {Studies.Anaphora.CharnavelMateu2015.LogoFeature.B, Studies.Anaphora.CharnavelMateu2015.LogoFeature.C}
• Studies.Anaphora.CharnavelMateu2015.LogoCenter.attitudeHolder.features = {Studies.Anaphora.CharnavelMateu2015.LogoFeature.C}

def Studies.Anaphora.CharnavelMateu2015.Clash (x y : LogoCenter) : Prop

Two centers clash iff their feature sets share at least one feature. Equivalent to "identical or adjacent on the hierarchy" (paper eq. 54).

Equations

• Studies.Anaphora.CharnavelMateu2015.Clash x y = (x.features ∩ y.features).Nonempty

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidableClash (x y : LogoCenter) : Decidable (Clash x y)

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidableClash x y = Studies.Anaphora.CharnavelMateu2015.instDecidableClash._aux_1 x y

def Studies.Anaphora.CharnavelMateu2015.Antilogophoric (centers : List LogoCenter) : Prop

C&M's antilogophoric intervention (paper §3.5.2, generalising eq. 64): a configuration of logophoric centers in a single domain is antilogophoric iff some pair of distinct centers clash.

Note: a single center never clashes with itself in this formulation — the "identical centers" case of (54) corresponds to multiple positions bearing the same center type, not the abstract type clashing with itself. We model the multi-position case with a List.

Equations

• Studies.Anaphora.CharnavelMateu2015.Antilogophoric centers = ∃ (i : Fin centers.length) (j : Fin centers.length), i ≠ j ∧ Studies.Anaphora.CharnavelMateu2015.Clash centers[i] centers[j]

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidableAntilogophoric (centers : List LogoCenter) : Decidable (Antilogophoric centers)

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidableAntilogophoric centers = Studies.Anaphora.CharnavelMateu2015.instDecidableAntilogophoric._aux_1 centers

theorem Studies.Anaphora.CharnavelMateu2015.pcc_clash : Clash LogoCenter.discourseParticipant LogoCenter.empathyLocus

PCC clash: discourse participant (1/2 clitic) + empathy locus (3.dat clitic) share feature B.

theorem Studies.Anaphora.CharnavelMateu2015.clr_clash : Clash LogoCenter.empathyLocus LogoCenter.attitudeHolder

CLR clash: empathy locus (3.dat) + attitude holder (3.acc de se) share feature C.

theorem Studies.Anaphora.CharnavelMateu2015.dp_ah_no_clash : ¬Clash LogoCenter.discourseParticipant LogoCenter.attitudeHolder

Discourse participant + attitude holder do not clash — no shared feature. This is the licit configuration (Table 3 final row).

theorem Studies.Anaphora.CharnavelMateu2015.table3_predictions : Clash LogoCenter.discourseParticipant LogoCenter.discourseParticipant ∧ Clash LogoCenter.discourseParticipant LogoCenter.empathyLocus ∧ Clash LogoCenter.empathyLocus LogoCenter.empathyLocus ∧ Clash LogoCenter.attitudeHolder LogoCenter.empathyLocus ∧ Clash LogoCenter.attitudeHolder LogoCenter.attitudeHolder ∧ ¬Clash LogoCenter.discourseParticipant LogoCenter.attitudeHolder

The full Table 3 prediction: clash iff the pair shares a feature.

structure Studies.Anaphora.CharnavelMateu2015.Condition : Type

A test condition in C&M's grammaticality experiment, parameterised by the three crossed factors (paper Table 1). The 9 conditions enumerate {c-command, no c-command} × {logophoric centre as antecedent, not} × {3.dat dative clitic, 1/2.dat dative clitic}, dropping the "bound 3" sub-case (their condition 3) for which they collapse the same prediction as condition 1.

• cCommandingAntecedent : Bool

  Does the antecedent c-command the accusative clitic?

• logoCenterAntecedent : Bool

  Is the antecedent a logophoric centre (attitude holder)?

• dative3rdPerson : Bool

  Is the dative clitic a 3rd-person form (an empathy locus)?

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidableEqCondition : DecidableEq Condition

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidableEqCondition = Studies.Anaphora.CharnavelMateu2015.instDecidableEqCondition.decEq

def Studies.Anaphora.CharnavelMateu2015.instDecidableEqCondition.decEq (x✝ x✝¹ : Condition) : Decidable (x✝ = x✝¹)

Equations

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

def Studies.Anaphora.CharnavelMateu2015.instReprCondition.repr : Condition → ℕ → Std.Format

Equations

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

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instReprCondition : Repr Condition

Equations

• Studies.Anaphora.CharnavelMateu2015.instReprCondition = { reprPrec := Studies.Anaphora.CharnavelMateu2015.instReprCondition.repr }

def Studies.Anaphora.CharnavelMateu2015.predictsUngrammatical (c : Condition) : Prop

C&M's hypothesis (paper §2.1): a sentence is ungrammatical iff the antecedent of the accusative clitic is a logophoric centre AND the dative clitic is 3rd person. C-command is not the relevant factor (contra Bhatt & Šimík).

Equations

• Studies.Anaphora.CharnavelMateu2015.predictsUngrammatical c = (c.logoCenterAntecedent = true ∧ c.dative3rdPerson = true)

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidablePredictsUngrammatical (c : Condition) : Decidable (predictsUngrammatical c)

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidablePredictsUngrammatical c = Studies.Anaphora.CharnavelMateu2015.instDecidablePredictsUngrammatical._aux_1 c

theorem Studies.Anaphora.CharnavelMateu2015.table2_results : predictsUngrammatical { cCommandingAntecedent := true, logoCenterAntecedent := true, dative3rdPerson := true } ∧ ¬predictsUngrammatical { cCommandingAntecedent := true, logoCenterAntecedent := true, dative3rdPerson := false } ∧ ¬predictsUngrammatical { cCommandingAntecedent := true, logoCenterAntecedent := false, dative3rdPerson := true } ∧ ¬predictsUngrammatical { cCommandingAntecedent := true, logoCenterAntecedent := false, dative3rdPerson := false } ∧ predictsUngrammatical { cCommandingAntecedent := false, logoCenterAntecedent := true, dative3rdPerson := true } ∧ ¬predictsUngrammatical { cCommandingAntecedent := false, logoCenterAntecedent := true, dative3rdPerson := false } ∧ ¬predictsUngrammatical { cCommandingAntecedent := false, logoCenterAntecedent := false, dative3rdPerson := true } ∧ ¬predictsUngrammatical { cCommandingAntecedent := false, logoCenterAntecedent := false, dative3rdPerson := false }

Paper Table 2: the experimental result. Conditions 1, 3, 6 (the three rows with logo-centre antecedent + 3.dat) received significantly lower scores than controls; the other six conditions did not. We collapse Table 1's "bound 3" condition into the 3.dat case.

theorem Studies.Anaphora.CharnavelMateu2015.c_command_irrelevant (logoCenter dative3rd : Bool) : predictsUngrammatical { cCommandingAntecedent := true, logoCenterAntecedent := logoCenter, dative3rdPerson := dative3rd } ↔ predictsUngrammatical { cCommandingAntecedent := false, logoCenterAntecedent := logoCenter, dative3rdPerson := dative3rd }

The key empirical finding (paper §2.4, conditions 4 vs 6): c-command is irrelevant; only logophoric centrehood + 3rd-person dative matter.

def Studies.Anaphora.CharnavelMateu2015.ClusterConfig : Type

A clitic-cluster configuration as a list of two logophoric centres (one for each clitic). The dative is read first, then the accusative.

Equations

• Studies.Anaphora.CharnavelMateu2015.ClusterConfig = (Studies.Anaphora.CharnavelMateu2015.LogoCenter × Studies.Anaphora.CharnavelMateu2015.LogoCenter)

def Studies.Anaphora.CharnavelMateu2015.CLRViolated (cfg : ClusterConfig) : Prop

The Clitic Logophoric Restriction (paper eq. 26): a configuration is blocked when its two centres are antilogophoric.

Equations

• Studies.Anaphora.CharnavelMateu2015.CLRViolated cfg = Studies.Anaphora.CharnavelMateu2015.Clash cfg.1 cfg.2

@[implicit_reducible]

instance Studies.Anaphora.CharnavelMateu2015.instDecidableCLRViolated (cfg : ClusterConfig) : Decidable (CLRViolated cfg)

Equations

• Studies.Anaphora.CharnavelMateu2015.instDecidableCLRViolated cfg = Studies.Anaphora.CharnavelMateu2015.instDecidableCLRViolated._aux_1 cfg

theorem Studies.Anaphora.CharnavelMateu2015.canonical_clr : CLRViolated (LogoCenter.empathyLocus, LogoCenter.attitudeHolder)

The canonical CLR configuration: 3.dat (empathy locus) + 3.acc de se (attitude holder). Violation.

theorem Studies.Anaphora.CharnavelMateu2015.canonical_pcc : CLRViolated (LogoCenter.empathyLocus, LogoCenter.discourseParticipant)

The canonical PCC configuration in C&M's terms: 3.dat (empathy locus) + 1/2.acc (discourse participant). Violation under the same mechanism.

theorem Studies.Anaphora.CharnavelMateu2015.cm_unification : CLRViolated (LogoCenter.empathyLocus, LogoCenter.attitudeHolder) ∧ CLRViolated (LogoCenter.empathyLocus, LogoCenter.discourseParticipant)

C&M's unification claim (paper §3.4): both phenomena are instances of CLRViolated. The bridge file documents P&Z's dissent.

theorem Studies.Anaphora.CharnavelMateu2015.spanish_me_lo_dat_is_discourse : Fragments.Spanish.Clitics.me_dat.person = UD.Person.first ∧ Fragments.Spanish.Clitics.lo.person = UD.Person.third

Spanish me lo — 1.DAT.ACC. Under C&M's typology, the 1.dat clitic is a discourse participant; lo is the accusative. The cluster is licit when lo is not read with an attitude-holder antecedent (paper §1.1, ex. 6 type: licit accusative).

theorem Studies.Anaphora.CharnavelMateu2015.spanish_te_lo_dat_is_discourse : Fragments.Spanish.Clitics.te_dat.person = UD.Person.second ∧ Fragments.Spanish.Clitics.lo.person = UD.Person.third

Spanish te lo — 2.DAT + 3.ACC. Same pattern: 2.dat is a discourse participant; the cluster is licit on non-attitude-holder readings.

theorem Studies.Anaphora.CharnavelMateu2015.spanish_le_lo_is_clr_pair : Fragments.Spanish.Clitics.le_dat.person = UD.Person.third ∧ Fragments.Spanish.Clitics.lo.person = UD.Person.third

Spanish le lo (which surfaces as se lo by the spurious-se rule): 3.DAT + 3.ACC. The CLR-relevant configuration — empathy locus + attitude holder. Bad under de se readings of lo (paper §1, ex. 2b; §3.4 implementation).

theorem Studies.Anaphora.CharnavelMateu2015.spanish_le_lo_de_se_violates_clr : CLRViolated (LogoCenter.empathyLocus, LogoCenter.attitudeHolder)

The clash-yielding cluster for Spanish se lo / le lo under a de se reading: empathy + attitude holder.

theorem Studies.Anaphora.CharnavelMateu2015.spanish_le_me_pcc_violates : Fragments.Spanish.Clitics.le_dat.person = UD.Person.third ∧ Fragments.Spanish.Clitics.me_acc.person = UD.Person.first ∧ CLRViolated (LogoCenter.empathyLocus, LogoCenter.discourseParticipant)

The PCC pattern *le me / *le te in Spanish: 3.dat + 1/2.acc — empathy locus + discourse participant.