Documentation

Studies.Keine2019

Keine (2019): Selective Opacity #

This file formalizes the horizons account of selective opacity of [Kei19]: a syntactic domain may be opaque to some operations and transparent to others (their (1)), and the constraint is on the searches of probes, each terminated by a category of its own, its horizon ((38)), rather than on domains or on the moving element. With category inheritance within an extended projection ((43)) a clause's label collects the categories it projects, so a clause is opaque to a probe exactly when its label contains the horizon, and Upward Entailment ((40)), that larger clauses are at least as opaque, is the Horizon Inheritance Theorem ((45), (46)), transparency antitone in the extension order of clause sizes (sizes_le). The Hindi probes of (48) and (57), φ-agreement and A-movement on T⁰ with horizon T, wh-licensing on C⁰ with horizon C and Ā-movement on C⁰ without one, derive the transparency table (58) with its three locality types (transparency_table) and with it the generalizations (21) and (23) on long-distance agreement; English hyperraising is A-movement with horizon C (Section 4.2.3), and its availability elsewhere is the absence of that horizon (hyperraising_iff).

The Height-Locality Connection ((33), (62)), that higher probes search into more kinds of structure, is derived rather than stipulated: a probe whose horizon lies below its position finds it in its sister and is vacuous ((63)), so for a nonvacuous probe the position bounds the horizon and the horizon bounds the position, the Height-Locality Theorem ((65)) instantiated for Hindi in (66) and (67).

Implementation notes #

Labels are the bilateral labels of the substrate's ClauseSpines and transparency is Probe.Profile.TransparentTo; the article's probes are defined here, and vacuity is Probe.Profile.IsVacuous, which takes the sister of a probe on C⁰ to be TP and of one on T⁰ to be vP.

References #

The label of a finite clause, the categories inherited up its extended projection ((44)).

Equations
Instances For

    The label of a large nonfinite clause.

    Equations
    Instances For

      The label of a small nonfinite clause.

      Equations
      Instances For

        Horizons and Upward Entailment (Section 4.1) #

        (45), (46): Upward Entailment follows from category inheritance. The three clause sizes stand in the extension order, TP extending vP and CP extending TP, and transparency is antitone in that order for every probe (Probe.Profile.transparentTo_label_antitone), so a probe blocked by a clause is blocked by every larger one.

        The Hindi probes (Section 4.2) #

        The φ-agreement probe [∗φ∗] of (48b), on T⁰ with horizon T.

        Equations
        Instances For

          The A-movement probe [•A•] of (48a), on T⁰ with horizon T.

          Equations
          Instances For

            The wh-licensing probe [∗wh∗] of (57), on C⁰ with horizon C.

            Equations
            Instances For

              The Ā-movement probe [•Ā•] of (48c), on C⁰ without horizon.

              Equations
              Instances For

                A probe's row of the transparency table over the finite, large nonfinite and small nonfinite clauses.

                Equations
                Instances For
                  theorem Keine2019.transparency_table :
                  row phiProbe = [false, false, true] ∧ row aProbe = [false, false, true] ∧ row whLicensing = [false, true, true] ∧ row ābarProbe = [true, true, true]

                  (58): φ-agreement and A-movement search only into vP clauses, wh-licensing into TP and vP clauses, Ā-movement into all three.

                  Selective opacity is not binary, since the table has three locality types.

                  Finite clauses, edge included, are opaque to A-movement and φ-agreement but not to Ā-movement, (23) for the finite clause embedding (49).

                  The two probes on T⁰ share their horizon, so a nonfinite clause small enough for A-extraction is the vP structure and is transparent to φ-agreement, which makes long-distance agreement obligatory, while Ā-movement enters the TP structure too and has no such effect, (21) for the nonfinite embeddings (50) and (51).

                  Hyperraising (Section 4.2.3) #

                  The English A-probe, on T⁰ with horizon C.

                  Equations
                  Instances For

                    The English extraposition probe, on T⁰ with horizon T.

                    Equations
                    Instances For

                      No A-probe search enters a finite clause in English, while Ā-extraction is unaffected, and extraposition, with horizon T, cannot leave even a nonfinite clause (59).

                      theorem Keine2019.hyperraising_iff (h : Option Minimalist.Cat) :
                      { probeHead := Minimalist.Cat.T, horizon := h }.TransparentTo cpLabel ↔ ∀ (c : Minimalist.Cat), c ∈ cpLabel → h ≠ some c

                      Hyperraising is a horizon parameter. An A-probe on T⁰ enters finite clauses exactly when its horizon is none of the categories a finite clause inherits, as in the languages that allow it.

                      The Height-Locality Connection (Section 5) #

                      A probe on C⁰ with horizon T finds its horizon in its sister and has no search space (63).

                      theorem Keine2019.height_to_locality (h : Minimalist.Cat) :
                      (¬{ probeHead := Minimalist.Cat.C, horizon := some h }.IsVacuous → ¬h ∈ tpLabel) ∧ (¬{ probeHead := Minimalist.Cat.T, horizon := some h }.IsVacuous → ¬h ∈ vpLabel)

                      Height to locality (65a). A nonvacuous probe on C⁰ has no horizon among the categories of its sister TP, and one on T⁰ none among those of vP, so those clauses are necessarily transparent to it.

                      theorem Keine2019.locality_to_height :
                      { probeHead := Minimalist.Cat.C, horizon := some Minimalist.Cat.T }.IsVacuous ∧ { probeHead := Minimalist.Cat.T, horizon := some Minimalist.Cat.v }.IsVacuous ∧ { probeHead := Minimalist.Cat.C, horizon := some Minimalist.Cat.v }.IsVacuous

                      Locality to height (65b). A probe with horizon T is vacuous on C⁰, and one with horizon v on T⁰ and C⁰, so a nonvacuous probe's horizon bounds its position from below.

                      theorem Keine2019.nonvacuous_transparent (h : Option Minimalist.Cat) :
                      (¬{ probeHead := Minimalist.Cat.C, horizon := h }.IsVacuous → { probeHead := Minimalist.Cat.C, horizon := h }.TransparentTo tpLabel ∧ { probeHead := Minimalist.Cat.C, horizon := h }.TransparentTo vpLabel) ∧ (¬{ probeHead := Minimalist.Cat.T, horizon := h }.IsVacuous → { probeHead := Minimalist.Cat.T, horizon := h }.TransparentTo vpLabel)

                      A nonvacuous probe on C⁰ searches into TP and vP clauses, and one on T⁰ into vP clauses, whatever their horizons, which is the Height-Locality Connection for the two positions.

                      The four Hindi probes are nonvacuous, so the two on C⁰ search into TP and vP clauses and the two on T⁰ into vP clauses, which is why nonfinite clauses are no islands for Ā-movement or wh-licensing and why only these interact with long-distance agreement as (21) says; and the A-probe's horizon T would make it vacuous on C⁰, so A-movement lands inside nonfinite clauses, (66) and (67).