Documentation

Studies.Allotey2021

Allotey (2021): overt pronouns of infinitival predicates of Gã #

This file formalizes [All21]. Obligatory control into Gã irrealis ni-clauses requires an overt subject proclitic: null PRO is ungrammatical, a lexical subject is ungrammatical, and the proclitic shows the whole OC signature (Table 2). The controlled clause is non-finite and irrealis rather than subjunctive — it bars tense and aspect, focus fronting and free reference, licenses NPIs across its boundary, negates preverbally, and carries the irrealis marker only as a high tone on its subject (Table 4). The pronoun is overt because that tone needs a segmental host. The subject markers are pronouns, not agreement: the verb is invariant across subjects (§4.4) and a marker cannot double a lexical subject (§6.1), against the affixal analysis of [Cam17]; Gã is therefore a non-pro-drop language (§2.1). [DH13b]'s 101A coding of Gã, made from [AO68], is affixal too.

Everything is read off the paper's example rows and the Fragment. The control profile of a clause type is [Lan13]'s signature as the rows attest it (Control.ofAttested): no row attests a criterial configuration of the ni-clause, so it is obligatory control, while the finite akɛ-clause attests free reference. Complementizer selection is Verb.Takes over the Fragment's frames; every finiteness diagnostic is a row theorem conditioned on the complementizer's finiteness alone, the paper's convergence argument; Table 4 is the exponents the rows show in each irrealis context, and the tone-hosting requirement then derives the overt pronoun from the minimal-pronoun inventory.

Implementation notes #

The paper's pro-drop status is its own reading of its 101A cell, obligatory subject pronouns (§4.2), and is stated here rather than in the Fragment because [DH13b] codes the same markers as affixes; wals_codes_affixes records the discrepancy and no_agreement_rows the paper's evidence against it. The bound-variable row of Table 2 rests on the de se example (53); the paper has no only test, so Control.Diagnostic.strictUnderOnly is never attested and the bound-variable clause of the signature holds unrefuted rather than tested. The paper tests reference only in the ni- and akɛ-clauses, so the comparison with [Lan04a]'s scale is stated for those two.

References #

Pronouns (Table 3) #

theorem Allotey2021.objective_form_iff (p : Person) (n : Number) :
(∃ q ∈ Ga.Pronouns.pronouns, q.person = some p ∧ q.number = some n ∧ q.case_ = some Case.acc) ↔ n = Number.singular ∧ (p = Person.second ∨ p = Person.third)

Only second and third person singular have a dedicated objective form.

The paradigm distinguishes case in the second and third person singular alone: every other referential category has one form for all three columns of Table 3.

Every referential category has a form: the one form of first person plural wɔ covers clusivity and the number of others alike.

theorem Allotey2021.card_subjectForms (q : PersonalPronoun) :
q ∈ Ga.Pronouns.pronouns → ∀ p ∈ q.person, ∀ n ∈ q.number, (Ga.Pronouns.subjectForms p n).card = 1

Each cell of Table 3 has one subject form.

Subject pronouns are not agreement (§4.4, §6.1) #

theorem Allotey2021.no_agreement_rows :
(∀ row ∈ Examples.all, row.feature? "diagnostic" = some "agreement" → (row.feature? "person").isSome = true → ("tee", "went") ∈ row.glossedTokens) ∧ ∃ r₁ ∈ Examples.all, ∃ r₂ ∈ Examples.all, r₁.feature? "diagnostic" = some "agreement" ∧ r₂.feature? "diagnostic" = some "agreement" ∧ r₁.feature? "person" ≠ r₂.feature? "person"

The verb of exx 79–81 is the one form tee 'went' whatever the person of its subject: Gã marks no subject agreement on the verb.

[DH13b] codes the same markers as subject affixes on the verb, not pronominalSubjects: the analysis of [Cam17] too, against which an affix would co-occur with a lexical subject (ex 123) and could not be separated from its verb by negation (ex 125). Gã's WALS code is ga; gaa is Gaagudju.

Complementizer selection (§5.5.1) #

The three-way clause typology is the selection relation: each clause type's frame takes exactly its own complementizer.

theorem Allotey2021.takes_ni_iff_control (v : Verb) :
v ∈ Ga.verbs → (v.Takes Ga.ni ↔ ∃ r ∈ v.readings, r.control.isSome = true)

A verb takes ni exactly when some frame of it is controlled.

theorem Allotey2021.proposition_iff_finite (v : Verb) :
v ∈ Ga.verbs → ∀ r ∈ v.readings, ∀ s ∈ r.size, s = Clause.Size.proposition ↔ ∃ z ∈ Ga.complementizers, r.frame.Takes z ∧ z.IsFinite

[WL23]'s hierarchy at Gã's granularity: a proposition complement is exactly one whose frame takes a finite complementizer; situations and events share the infinitival ni-frame.

theorem Allotey2021.control_rows :
(∃ v ∈ Ga.verbs, ∃ r ∈ v.readings, r.control = some ControlType.subjectControl) ∧ ∃ v ∈ Ga.verbs, ∃ r ∈ v.readings, r.control = some ControlType.objectControl

Subject and object control are both in the inventory (Table 2).

Rows #

Gã has no context-specific vocabulary item for minimal pronouns, so the elsewhere pronoun realizes every context.

Equations
Instances For
    def Allotey2021.verbOf (row : Datum) :
    Option Verb

    The Fragment entry for a row's matrix verb.

    Equations
    Instances For

      The clause type a complementizer feature names.

      Equations
      Instances For

        A row's clause type is its complement's, or the finite type for a matrix clause the paper labels finite.

        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          def Allotey2021.reads (row : Datum) (r : String) (j : Judgment) :

          A row records reading r with judgment j.

          Equations
          Instances For
            @[instance_reducible]
            instance Allotey2021.instDecidableReads (row : Datum) (r : String) (j : Judgment) :
            Decidable (reads row r j)
            Equations

            The realized form of a row's embedded subject.

            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              theorem Allotey2021.controlled_subject_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "diagnostic" ∈ [none, some "nullSubject"] → ∀ f ∈ formOf row, row.judgment = Judgment.acceptable ↔ f = gaInventory.controlForm

              In the rows whose only point is the shape of the controlled subject, the control frame is grammatical exactly with the inventory's control form (exx 2–3, 34–44, 54–59 vs 40–41).

              theorem Allotey2021.lexical_subject_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "embeddedSubject" = some "lexical" → row.feature? "clauseContext" = none → row.judgment = Judgment.ungrammatical

              A lexical subject in the control frame is out (exx 42b–c, 64) — the copy the Movement Theory of Control ([Hor99]) would pronounce.

              theorem Allotey2021.c_selection_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "diagnostic" = some "cSelection" → ∀ v ∈ verbOf row, ∀ c ∈ clauseTypeOf row, row.judgment = Judgment.acceptable ↔ v.Takes c.complementizer

              A complementizer is grammatical exactly when the verb takes it (exx 104–106).

              theorem Allotey2021.tam_rows (row : Datum) :
              row ∈ Examples.all → ∀ c ∈ clauseTypeOf row, ∀ t ∈ row.feature? "embeddedTAM", t ≠ "none" → (row.judgment = Judgment.acceptable ↔ c.complementizer.IsFinite)

              Overt tense or aspect in the complement is grammatical exactly in the finite clause types (exx 101, 111).

              theorem Allotey2021.focus_npi_rows (row : Datum) :
              row ∈ Examples.all → ∀ c ∈ clauseTypeOf row, (row.feature? "diagnostic" = some "focus" → (row.judgment = Judgment.acceptable ↔ c.complementizer.IsFinite)) ∧ (row.feature? "diagnostic" = some "npi" → row.feature? "negation" = some "matrix" → (row.judgment = Judgment.acceptable ↔ ¬c.complementizer.IsFinite))

              Focus fronting, [Riz97]'s strong-CP diagnostic, is available exactly in the finite clauses (exx 107–108); matrix negation licenses an embedded NPI exactly across the non-finite one (exx 116–117).

              theorem Allotey2021.negation_rows (row : Datum) :
              row ∈ Examples.all → ∀ c ∈ clauseTypeOf row, ∀ p ∈ row.feature? "negationPosition", p = "preverbal" ↔ ¬c.complementizer.IsFinite

              Negation precedes the verb exactly in the non-finite clause (exx 121–122, 124).

              theorem Allotey2021.subject_tone_rows (row : Datum) :
              row ∈ Examples.all → row.judgment = Judgment.acceptable → ∀ c ∈ clauseTypeOf row, ∀ t ∈ row.feature? "subjectTone", (t = "high") = (c = Ga.EmbeddedClauseType.ni)

              The embedded subject bears the irrealis high tone exactly in the ni-clause (exx 110–112, 118).

              The irrealis marker #

              theorem Allotey2021.marker_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "clauseContext" ∈ [none, some "control"] → ∀ m ∈ row.feature? "irrealisMarker", m = "present" ↔ (row.feature? "control").isSome = true

              The marker appears on the subject of the control frame and nowhere else in the complement data (exx 34, 88–89, 92, 100–103, 106, 109, 112, 117–119, 122); it tracks the frame, not the verb — kai 'remember' takes it in its ni-frame (exx 43, 117a) and lacks it in its realis akɛ-frame (ex 89a).

              theorem Allotey2021.implicative_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "diagnostic" = some "implicative" → ∀ v ∈ verbOf row, row.feature? "irrealisMarker" = some "absent" ↔ v.implicative.positive = ↑Polarity.positive

              In the paper's implicative contrast (ex 89) the marker is absent exactly under the verbs of the Fragment whose affirmed use entails the complement realized ([Kar71a]).

              The OC signature (Table 2) #

              theorem Allotey2021.cCommand_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "diagnostic" = some "cCommand" → (row.judgment = Judgment.acceptable ↔ row.feature? "antecedent" = some "cCommanding")

              The antecedent must c-command the controlled pronoun (exx 45–46): grammatical exactly under the paper's c-commanding coindexation.

              theorem Allotey2021.longDistance_rows (row : Datum) :
              row ∈ Examples.all → row.feature? "diagnostic" = some "longDistance" → (row.judgment = Judgment.acceptable ↔ row.feature? "antecedent" = some "local")

              No long-distance antecedent (exx 47–49): grammatical exactly under the local coindexation.

              theorem Allotey2021.sloppy_only :
              ∃ row ∈ Examples.all, row.feature? "diagnostic" = some "ellipsis" ∧ reads row "sloppy" Judgment.acceptable ∧ reads row "strict" Judgment.unacceptable

              Under ellipsis the controlled pronoun has the sloppy reading only (ex 52).

              theorem Allotey2021.free_reading_rows (row : Datum) :
              row ∈ Examples.all → ∀ c ∈ clauseTypeOf row, ∀ (j : Judgment), reads row "free" j → (j = Judgment.acceptable ↔ c.complementizer.IsFinite)

              A free reading of the embedded subject is available exactly in the finite complement (exx 110–111 vs 92, 112).

              theorem Allotey2021.deSe_witness :
              ∃ row ∈ Examples.all, row.feature? "diagnostic" = some "deSe" ∧ row.judgment = Judgment.unacceptable ∧ row.context ≠ ""

              Ex 53 witnesses the de se row: infelicitous in its context.

              The controlled form φ-covaries with its controller (exx 37–39), unlike [Sat19]'s form-invariant Ewe yè.

              An acceptable row attests one control diagnostic, a non-c-commanding or long-distance antecedent by the paper's coindexation, a free reading of the embedded subject, or a strict reading under ellipsis.

              Equations
              Instances For
                @[instance_reducible]
                instance Allotey2021.instDecidablePredDiagnosticAttests (row : Datum) :
                DecidablePred (attests row)
                Equations
                • One or more equations did not get rendered due to their size.

                The diagnostics the rows attest for a clause type.

                Equations
                Instances For
                  @[instance_reducible]
                  Equations

                  The control profile of a clause type in [Lan13]'s signature: the clauses no attested diagnostic refutes.

                  Equations
                  Instances For

                    No row attests a criterial configuration of the ni-clause.

                    The ni-clause is obligatory control: every clause of the signature holds, so it admits no criterial configuration. [Sza09]'s long-distance Agree reaches the embedded subject across this weak CP.

                    The finite akɛ-clause attests a free reading (exx 110–111).

                    So it is not obligatory control.

                    Landau's scale #

                    Gã clause types on [Lan04a]'s finiteness scale — a scale position, not a mood claim. Unrestricted TAM and independent tense both coincide with the complementizer's finiteness (exx 109–111, 118–119).

                    Equations
                    Instances For

                      No Gã clause type is a tensed-but-controlled F-subjunctive.

                      theorem Allotey2021.landau_predicts_control (c : Ga.EmbeddedClauseType) (hc : c ≠ Ga.EmbeddedClauseType.keji) (agr : Bool) :
                      gaProfile c = Set.univ ↔ (gaToLandau c).HasOC agr

                      The scale agrees with the rows where the paper tests reference: OC exactly on the C-subjunctive, at any Agr value — Gã has no φ-agreement (exx 79–81, 123).

                      In [Noo07]'s typology the non-finite clause is the reduced one: .infinitive, the paper's own term for the bare-root ni-complement.

                      Table 4 #

                      The five irrealis contexts of Table 4.

                      Instances For
                        @[instance_reducible]
                        Equations
                        Equations
                        • One or more equations did not get rendered due to their size.
                        Instances For

                          The irrealis marker has three exponents, high tone on the subject, high tone on the verb, and the vowel segment a.

                          Instances For
                            @[instance_reducible]
                            Equations
                            def Allotey2021.instReprExponent.repr :
                            Exponent → ℕ → Std.Format
                            Equations
                            • One or more equations did not get rendered due to their size.
                            Instances For

                              A row shows an exponent.

                              Equations
                              Instances For
                                @[instance_reducible]
                                instance Allotey2021.instDecidablePredExponentShows (row : Datum) :
                                DecidablePred (shows row)
                                Equations
                                • One or more equations did not get rendered due to their size.

                                The rows in an irrealis context with judgment j.

                                Equations
                                Instances For

                                  Table 4 is read off the rows, as the exponents some grammatical row shows in the context (exx 85–86, 93–97, 100–103).

                                  Equations
                                  Instances For

                                    Embedded control realizes the marker as the subject's high tone alone.

                                    The embedded-control realization is unique among the five contexts; in particular it lacks the subjunctive's doubled high tone (ex 88).

                                    The realizations are exact: adding the verb's tone to the control clause (ex 88) or the subject's to the future (ex 96), or dropping the verb's from the subjunctive (ex 87), is ungrammatical.

                                    Deriving the overt pronoun #

                                    A tonal exponent needs a segmental host; the null form has none.

                                    Equations
                                    Instances For

                                      Null PRO is impossible in Gã: a null controlled-subject form cannot host the irrealis tone.

                                      The Gã inventory meets the tone-hosting requirement.

                                      Controlled subjects surface as overt pronouns.