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 examines patterns of argument optionality in English ditransitive constructions and discovers four systematic asymmetries:
- Sluicing asymmetry: Implicit second objects and PPs license sluicing, but implicit first objects never do.
- Interpretation asymmetry: Implicit second objects and PPs can be definite or indefinite (depending on the verb), but implicit first objects are uniformly definite.
- Base transitivity constraint: A simple transitive that allows an implicit object does NOT allow it when used in the DOC (We're baking ✓ → *We're baking them with implicit second obj).
- Frame-dependent licensing: An implicit direct object is licensed only in the DOC for some verbs, and only in the PP frame for others.
Theoretical Analysis #
Bruening adopts the ApplP analysis (@cite{marantz-1993}): in the 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 make them implicit.
Formalization #
This file verifies Bruening's empirical generalizations against the English verb fragment. Each generalization is stated as a theorem over the set of ditransitive verbs with implicit argument properties.
Coverage #
Table (56) lists ~40 verbs across 10 cells. This formalization covers 31 of them. The remaining verbs (ask, promise, wish, leave, afford, lose, guarantee, rent, save) are either already in the Fragment with different complement types (attitude/question-embedding senses) or require new entries. The three-frame limitation (VerbCore has complementType + altComplementType, so max two frames) prevents fully encoding verbs like send/tell/throw that have NP, NP_NP, and NP_PP frames.
The set of ditransitive verbs from the English fragment that are relevant to Bruening's classification (Table (56)).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Each theorem verifies one verb's implicit argument properties against Bruening's classification (Table (56)). Changing a field on the Fragment entry breaks exactly one theorem.
DOC-only verbs: those that appear in the double object construction but NOT the prepositional dative (Table (56), "First Object" columns).
This is an explicit list because the two-field complement type model cannot always distinguish DOC-only from alternating verbs (verbs like envy have complementType = .np with altComplementType = .np_np, which makes them alternating in the data model, but DOC-only in Bruening's classification).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Bruening's Generalization 2 (§2.2): Among DOC-only verbs, every implicit first object is definite. No DOC-only verb allows an indefinite implicit first object.
This follows from the ApplP analysis: implicit first objects are licensed by ApplPass, which (like Pass for Voice) only produces definite implicit arguments.
Implicit themes (second objects) can be either definite or indefinite, depending on the verb. Both interpretations are attested.
Bruening's Generalization 3 (§2.3.1): Base transitives that allow an implicit object (bake, melt, build) do NOT allow it when used in the DOC.
Formalized as: these verbs have complementType = .np (base transitive)
with implicitObj = some _, but their complementType is NOT .np_np.
Bruening's explanation: the implicit ∃/ι heads adjoin to V and close off V's selectional feature. When Appl adds the first object above VP, V already has a complete argument structure — V's implicit-licensing head blocks the DOC from licensing an implicit second object.
Verbs with implicit themes.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Verbs with implicit goals.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Bruening vs Larson (@cite{larson-1988}): both analyses predict IO > DO c-command in DOC, but make different predictions about implicit arguments.
Larson's analysis (both objects selected by V) wrongly predicts that implicit first objects should behave like implicit second objects. In fact they differ: among DOC-only verbs, implicit goals are uniformly definite, while implicit themes vary.
This asymmetry is Bruening's primary argument against Larson's VP-shell analysis and in favor of the ApplP analysis (@cite{marantz-1993}).
The DOC-only verbs in our fragment all have NP_NP complement type (the double object frame), confirming that the DOC projects the first object structurally (via Appl) rather than lexically.
@cite{pylkkanen-2008}'s high applicative analysis predicts exactly this: the first object position is available independent of the lexical verb's complement type.
Bruening refines the general claim in DativeAlternation that ditransitives require both arguments. While give is standardly cited as requiring both a theme and a recipient, Bruening shows the goal CAN be implicit (definite): "She gave $5" (contextually determined recipient).
This is not a contradiction: DativeAlternation captures the GENERAL pattern, while Bruening's implicit argument data captures the EXCEPTIONS that specific verbs allow under specific pragmatic conditions.