Documentation

Linglib.Phenomena.ArgumentStructure.Studies.Bruening2021

Bruening 2021 — Implicit Arguments in English Double Object Constructions #

@cite{bruening-2021}

Implicit arguments in English double object constructions. Natural Language and Linguistic Theory 39:1023–1085.

Core empirical findings #

Bruening identifies four asymmetries in English ditransitives (§2.4 summary, p. 1040):

  1. Sluicing asymmetry (G1): implicit second objects and PPs license sluicing, but implicit first objects never do.
  2. Interpretation asymmetry (G2): implicit second objects and PPs can be definite or indefinite (depending on the verb), but implicit first objects are uniformly definite.
  3. Base-transitivity constraint (G3): a simple transitive that allows an implicit object does NOT allow it when used in the DOC (*We're baking them with intended-recipient reading).
  4. Frame-dependent licensing (G4): an implicit direct object is licensed only in the DOC for some verbs (show, ask, pay, serve, teach, feed), only in the PP frame for others (pass, throw, give), and only write in both (§2.3.2 ex. (63)–(65)).

Theoretical analysis #

Bruening adopts the ApplP analysis of @cite{marantz-1993}, developed for English in @cite{bruening-2010a}: in DOC, the second object is selected by V while the first object is projected by Appl(icative) above VP. Implicit arguments of V are licensed by functional heads ∃ (indefinite) or ι (definite) that adjoin to V; implicit arguments of functional heads (Voice, Appl) require a higher functional head (Pass, ApplPass) to be implicit.

Bruening explicitly rejects the rival accounts that this file's contrastive theorems engage with: @cite{pesetsky-1995}'s "both objects selected by V" view (§3.1 p. 1041), @cite{larson-1988}'s VP-shell (§3.1), small-clause analyses including @cite{pylkkanen-2008}'s low applicative (§3.1 + fn. 10 p. 1042), and Landau 2010's null-NP implicit-argument view (§4.1).

Formalization scope #

This file verifies Bruening's classification (Table 56, p. 1037) against the English verb fragment via a single decide-checked drift sentry over a BruRow table. Two cross-framework contrastive theorems make Bruening's positions Lean-checkable against linglib's existing Pylkkänen 2008 (english_appl.classification = .lowRecipient) and Larson 1988 (docDativeShift) formalizations.

What is formalized: G2 directly; G3 with the implicational consequent (melt/build-class transitives don't license .np_np as their alt frame).

What is deferred:

G1 (sluicing asymmetry) is formalized as a sibling in Phenomena/Ellipsis/Studies/Bruening2021Sluicing.lean, deriving implicit-second-obj-licenses-sluicing-but-implicit-first-obj-doesn't from the maximal-projection identity condition (Bruening §5.5).

Coverage #

Table (56) lists ~43 verbs across 15 cells (11 populated, 4 empty by Bruening's footnote on p. 1037). This file covers 31 verbs whose Fragment encoding is unambiguous; the remainder (ask, promise, wish, leave, afford, lose, guarantee, rent, save) are listed in the Fragment with attitude/question-embedding senses or absent.

The BruRow.alternates field encodes Bruening's classification of DOC-vs-PP alternation. It is NOT checked by BruRow.matches because the Fragment's two-slot frame schema cannot always represent {.np_np, .np_pp} alternation — e.g. tell has complementType = .finiteClause, altComplementType = some .np_np, with no PP slot available. Treating alternation as Bruening's classification (not as a Fragment-derivable fact) keeps the drift sentry focused on the schema fields it can actually check.

See Phenomena/ArgumentStructure/Studies/HaddicanEtAl2026.lean (doc_bruening and bruening_give_field_consistent) for a SyntacticObject witness of Bruening's V+P amalgam structure that consumes this file's give Fragment entry — a complementary tree-shape angle on the same paper.

Bruening's Table (56) — encoded as a row table #

source
structure Bruening2021.BruRow :
Type

A row in Bruening (56). Carries the verb, the expected implicitObj/implicitGoal Fragment fields per Bruening's classification, and Bruening's claim about whether this verb alternates between the DOC and PP frames.

alternates is documentation of Bruening's classification, not a Fragment-checked claim — see the module docstring on the schema limitation.

Instances For
    source
    def Bruening2021.BruRow.matches (r : BruRow) :
    Bool

    A BruRow matches when the verb's Fragment fields agree with the expected projection. Returns false (rather than failing decide) so the diagnostic example below names the offending row directly.

    Equations
    Instances For
      source
      def Bruening2021.bruening2021Table :
      List BruRow

      Bruening (56), p. 1037, encoded against the English verb fragment. Row groupings reflect Bruening's cell taxonomy; comments name each cell.

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

        Drift sentry #

        A single decide-checked theorem replaces 31 per-verb ⟨rfl, rfl⟩ readouts. If a Fragment field changes or any verb's classification drifts, this theorem fails — and the diagnostic example below names the offender via Repr-printable output.

        source
        theorem Bruening2021.verbs_match_bruening_table :
        bruening2021Table.all BruRow.matches = true

        Derived verb subsets #

        source
        def Bruening2021.ditransitiveVerbs :
        List Fragments.English.Predicates.Verbal.VerbEntry

        All ditransitive verbs in Bruening (56). Derived from the table.

        Equations
        Instances For
          source
          def Bruening2021.docOnlyVerbs :
          List Fragments.English.Predicates.Verbal.VerbEntry

          DOC-only verbs (those NOT alternating with a PP frame, per Bruening's classification — see module docstring on alternation). Derived.

          Equations
          Instances For

            G2: definiteness asymmetry #

            Bruening's G2 (§2.4 summary point 2): implicit first objects in DOC are uniformly definite, never indefinite. This is the empirical core of his argument against accounts (Larson, Pesetsky) that treat both objects symmetrically as V's selected arguments.

            ANALOGUE: Bruening's ι-head (def-implicit licensor) and ApplPass (licensor for implicit first object) have no current substrate primitives in Theories/Syntax/Minimalist/. Voice.ExternalArgSemantics includes thematicExistential (the analogue of Bruening's ∃-head); a parallel thematicIota and a Pass-of-Appl head are substrate gaps.

            source
            theorem Bruening2021.g2_doc_only_implicit_first_obj_definite :
            (docOnlyVerbs.all fun (v : Fragments.English.Predicates.Verbal.VerbEntry) => decide (v.implicitGoal ≠ some Semantics.Lexical.ImplicitInterp.indef)) = true

            G3: base-transitivity constraint #

            Bruening's G3 (§2.3.1, summary point 3): a simple transitive that allows an implicit object does NOT allow it when used in the DOC.

            The encoded consequent: melt and build (Bruening p. 1025 ex. (7)–(8)) have complementType = .np (transitive) with implicitObj.isSome, AND have no .np_np alt — so the Fragment itself blocks the spurious "implicit-second-obj-in-DOC" reading for these verbs.

            (Bruening's prototypical example bake is not in the English fragment.)

            source
            def Bruening2021.baseTransitivesWithImplicit :
            List Fragments.English.Predicates.Verbal.VerbEntry
            Equations
            Instances For
              source
              theorem Bruening2021.g3_base_transitive_constraint :
              (baseTransitivesWithImplicit.all fun (v : Fragments.English.Predicates.Verbal.VerbEntry) => decide (v.complementType = Semantics.Lexical.ComplementType.np) && v.implicitObj.isSome && decide (v.altComplementType ≠ some Semantics.Lexical.ComplementType.np_np)) = true

              G1 and G4: deferred substrate gaps #

              -- G1 (sluicing asymmetry, §2.1): formalized as a sibling in
-- `Phenomena/Ellipsis/Studies/Bruening2021Sluicing.lean` via
-- `g1_sluicing_asymmetry`. Substrate: `bruening2021Identity` in
-- `Theories/Syntax/Minimalist/Ellipsis/FormalMatching.lean` § 7.
--
-- TODO(G4): Frame-conditioned implicit-arg licensing (§2.3.2 ex. 63-65).
-- The Fragment carries `implicitObj : Option ImplicitInterp` globally,
-- not per-frame. Bruening shows this distinction is real (e.g. *show*
-- is DOC-only, *pass* is PP-only, *write* is both). Schema extension
-- (`implicitObjByFrame : ComplementType → Option ImplicitInterp`) would
-- enable the theorem; deferred — out of scope for this refactor.

              Cross-framework contrasts #

              Bruening positions his analysis explicitly against several rivals (§3.1 + fn. 10 p. 1042 + §4.1). Two of those rivals — Pylkkänen 2008 and Larson 1988 — are formalized in this directory. Below we make Bruening's disagreements with them Lean-checkable.

              source
              theorem Bruening2021.bruening_vs_pylkkanen_low_recipient :
              Pylkkanen2008.english_appl.classification = Minimalist.ApplType.lowRecipient ∧ (docOnlyVerbs.all fun (v : Fragments.English.Predicates.Verbal.VerbEntry) => decide (v.implicitGoal ≠ some Semantics.Lexical.ImplicitInterp.indef)) = true

              Bruening vs Pylkkänen 2008.

              Both analyses agree that the first object of English DOC is in an Appl-projection above V (not selected by V). Pylkkänen's english_appl commits English DOC to .lowRecipient (Pylkkanen2008.lean:227); this classification correctly predicts the structural facts about c-command, binding, and quantifier scope.

              They diverge on what licenses implicit first objects. Bruening's ApplPass derivation predicts implicit first objects are uniformly definite (G2 above). Pylkkänen's .lowRecipient classification alone does not entail this — it requires Bruening's additional Pass/ApplPass machinery, which has no substrate analogue (see G2 ANALOGUE note).

              Bruening explicitly rejects Pylkkänen 2008's analysis as a "variety of small clause analysis" (fn. 10 p. 1042).

              source
              theorem Bruening2021.bruening_vs_larson_implicit_first_obj :
              Fragments.English.Predicates.Verbal.pay.implicitObj = some Semantics.Lexical.ImplicitInterp.indef ∧ Fragments.English.Predicates.Verbal.pay.implicitGoal = some Semantics.Lexical.ImplicitInterp.def ∧ Fragments.English.Predicates.Verbal.pay.implicitObj ≠ Fragments.English.Predicates.Verbal.pay.implicitGoal

              Bruening vs Larson 1988.

              Larson's VP-shell analysis (Larson1988.docDativeShift) treats both objects of DOC as selected arguments of V (lower V'-shell + Dative Shift movement). This predicts implicit first objects should pattern with implicit second objects — both arguments-of-V should behave alike.

              Bruening's data refute the prediction: pay has indef-implicit second object (some .indef) but def-implicit first object (some .def). The asymmetry is the load-bearing argument against Larson (§3.1 p. 1041), and against @cite{pesetsky-1995}'s "both selected by V" view.