Documentation

Linglib.Phenomena.ArgumentStructure.Studies.HaddicanEtAl2026

PVC–DOC Structural Priming #

@cite{haddican-tamminga-dendikken-wade-2026} @cite{dendikken-1995} @cite{halle-marantz-1993} @cite{johnson-1991} @cite{aarts-1989} @cite{bruening-2010a}

English Particle Verbs Prime Double Object Constructions in Production. Linguistic Inquiry. doi:10.1162/LING.a.558

Production priming experiment (N=238) testing whether PVCs prime DOCs.

Design #

Sentence completion task. Two subdesigns (Table 1, p.7):

PVC primes used particle-object order ("put down the vase") to control for surface string similarity with DOC targets (p.5).

Results #

PVCs prime DOCs (β=0.296, p=.005). PVC and DOC primes do not differ in priming magnitude (β=−0.069, p=.503). Consistent with identical structural representations under the SC analysis.

Cross-references #

A priming contrast between two prime conditions.

  • primeA : String
  • primeB : String
  • target : String
  • aFavorsTarget : Bool
  • significant : Bool
Instances For
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      Equations
      Instances For

        DOC production rate by prime condition. Table 1, p.7. Percentages are integers (e.g., 59 = 59%).

        • condition : String
        • docPct :
        • pdPct :
        Instances For
          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            Equations
            Instances For

              Table 1 cell rates #

              Equations
              Instances For
                Equations
                Instances For
                  Equations
                  Instances For
                    Equations
                    Instances For

                      Regression contrasts #

                      Baseline replication: DOC primes boost DOC production relative to PD primes (β=0.569, SE=0.114, p<.001).

                      Equations
                      Instances For

                        Key finding: PVC primes boost DOC production relative to non-PVC control primes (β=0.296, SE=0.105, p=.005).

                        Equations
                        Instances For

                          PVC and DOC primes do not differ in their priming of DOCs (β=−0.069, SE=0.104, p=.503; combined 2×4 model, n.9).

                          Equations
                          Instances For

                            Verification theorems #

                            DOC priming is strictly stronger than PD non-priming (baseline effect).

                            PVC primes DO boost DOC production.

                            PVC and DOC primes yield equivalent magnitude — no significant difference.

                            Lexical items #

                            Structural analyses #

                            DOC, Applicative (@cite{halle-marantz-1993}; @cite{bruening-2010a}): [ApplP DP_goal [Appl' Appl [VP V DP_theme]]]

                            Equations
                            Instances For

                              PVC, Complex predicate: [VP [V lift+up] DP]

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

                                @cite{dendikken-1995} SC family #

                                Explicit shapes #

                                DOC small clause: [VP V [SC DP DP]] — three leaves in a right-branching shape.

                                Structural isomorphism #

                                structurallyIsomorphic x y is Prop-valued (substrate change at 0.230.865; revived as x.shape = y.shape); previously Bool-valued on planar TraceTree. Decidable, so decide works.

                                The non-PVC transitive control has a different shape from SC-DOC.

                                The non-PVC control has the SAME shape as the complex predicate PVC.

                                Den Dikken SC family isomorphism #

                                SC family categorization #

                                DOC and PVC share SC predicate category P.

                                Nested SC for DOC #

                                @cite{bruening-2021}: process-level isomorphism #

                                doc_bruening below is a SyntacticObject witness of Bruening's V+P amalgam analysis. The lexical-fragment side of the same paper — Bruening's classification of implicit-argument behavior across ~43 ditransitive verbs (Table 56) — is formalized in Phenomena/ArgumentStructure/Studies/Bruening2021.lean. The bruening_give_field_consistent theorem below ties this structural witness to that lexical-fragment side.

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

                                  Both Bruening structures use a complex (incorporated) head leaf. The original theorem (planar substrate) projected the head via match doc_bruening with | .node (.leaf tok) _ => tok.item.isComplex, which doesn't reduce under MCB nonplanar SOs (children of .mul are unordered).

                                  The claim itself remains true and verifiable: both DOC and PVC structures contain a complex (head-incorporated) LIToken among their subtrees. Reformulate as a Multiset-membership claim:

                                  (∃ tok ∈ doc_bruening.subtrees.filterMap getLIToken, tok.item.isComplex) ∧
                                  (∃ tok ∈ pvc_bruening.subtrees.filterMap getLIToken, tok.item.isComplex)
                                  

                                  TODO Phase 2 / polish: prove the Multiset-version directly, or re-derive from a head-function-aware headLIToken : SO → Option LIToken once Phase 2 substrate lands.

                                  Bridge to Bruening 2021 lexical fragment #

                                  doc_bruening above is a SyntacticObject witness; this theorem ties it to the corresponding lexical-fragment entry in Bruening2021.lean, ensuring the verb we structurally treat as "give-in-DOC" is also the verb whose implicit-argument profile licenses Bruening's Table (56) classification. If Verbal.lean ever moves give to a different complement type or implicit-arg profile, this bridge fails — alerting both files.

                                  Bridge to experimental data #

                                  The SC analysis predicts DOC/PVC isomorphism and DOC/PD non-isomorphism.

                                  The ApplP + ComplexPred combination predicts DOC/PVC non-isomorphism.

                                  PVC priming magnitude equals DOC priming magnitude, as SC predicts.

                                  The complex predicate PVC analysis cannot explain the priming asymmetry.

                                  IsSmallClause companion-predicate witnesses #

                                  The flat encodings (pvc_sc, doc_sc, resultative_sc, causative_sc) name the whole [VP V SC] constituent — the SC itself is the right child. We characterise the inner SCs against the IsSmallClause companion predicate (SmallClause.lean).

                                  Three of the four families satisfy the predicate; DOC's flat DP–DP encoding does not. This surfaces a real subtlety: Haddican et al. (2026) explicitly say (p.2) "we set aside details of the internal structure of the small clause", and the flat DP–DP shape is the deliberate simplification. The richer DOC encoding (with BE+P decomposition / Predicate Inversion) in Phenomena/ArgumentStructure/Studies/Dendikken1995 does satisfy IsSmallClause at every nested SC layer.

                                  Phase 1.0 caveat: IsSmallClause is noncomputable because it routes through outerCat/headCat, which are Phase 1.0 placeholders via Quot.out on the FreeCommMagma carrier. Concrete-instance checks via decide fail at the kernel-reduction step. TODO Phase 2: once LCA-based head selection lands, restore by decide.

                                  Diagnostic: the flat DP–DP DOC encoding does NOT satisfy IsSmallClause — neither child is in the SC predicate set {P,A,V,N}. Both children are DPs (head category D). The companion predicate surfaces the simplification — the encoding is correct for the priming argument but incomplete as a structural SC analysis; den Dikken's BE+P decomposition supplies the missing predicate.