Documentation

Linglib.Phenomena.ArgumentStructure.Studies.Newman2024

Newman (2024) — When Arguments Merge @cite{newman-2024} #

Elise Newman (2024). When Arguments Merge. MIT Press (Linguistic Inquiry Monographs 88). ISBN 9780262379960.

Overview #

The Categorial Merge Hypothesis (CMH) proposes that all elements of category V share the same Merge features, with only three argument-introducing features: [·D·] (DP-specific), [·X·] (category-unspecified), and [·V·] (VP-merge, v only). Combined with Feature Maximality, Weak Economy, and Generalized Tucking In (@cite{paille-2020}), the CMH derives:

  1. Non-DP First: XPs merge before DPs when V bears both features
  2. VP-as-specifier: When v bears [·X·], VP is forced to be a specifier of v (not complement)
  3. Symmetric passives: In "high IO" ditransitives, neither internal argument c-commands the other, so either can passivize
  4. DOMA: Weak Economy forces IO to be passive subject when IO is a wh-phrase (checks more features)
  5. No IO passives with inherent case: KPs can't check [·D·], so they can't A-move to subject (Greek, Tamil, German, Turkish, Spanish)
  6. Agent Focus in Mayan: Anti-redundancy deletes lower agreement copies when both probes target the same extracted argument

Formalized Predictions #

This file formalizes the verb type space, structural predictions about VP-specifierhood and its consequences, the KP/DP distinction for passive accessibility, and anti-redundancy for agreement.

CMH Apparatus (relocated from Minimalism/CMH.lean) #

The Categorial Merge Hypothesis primitives below are paper-specific to @cite{newman-2024} (with @cite{paille-2020} for Generalized Tucking In) and not consumed elsewhere in the library, so they live here under namespace Minimalist.CMH for symmetry with other Minimalist apparatus and to support qualified lookup if a future paper picks them up.

The three argument-introducing Merge features for verbal heads.

These are the only features that determine how V and v combine with their arguments. All other selectional requirements (c-selection for CPs, PPs, etc.) reduce to [·X·] — category-unspecified Merge.

  • D: Merges with DPs only (subjects, objects)
  • X: Merges with any non-DP phrase (PPs, CPs, APs, etc.)
  • V: Merges with VP (only on v; v selects V)
Instances For
    @[implicit_reducible]
    Equations
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      A bundle of Merge features on a single head. Represented as counts (0 or 1) of each feature type, reflecting Feature Maximality (a head checks at most one instance of each Merge feature type).

      • hasD : Bool
      • hasX : Bool
      • hasV : Bool
      Instances For
        def Minimalist.CMH.instDecidableEqMergeBundle.decEq (x✝ x✝¹ : MergeBundle) :
        Decidable (x✝ = x✝¹)
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            Equations
            Instances For

              Number of Merge features in a bundle.

              Equations
              Instances For

                Convert a bundle to a list of features.

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

                  Whether a category is a verbal head in the CMH sense.

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

                    A CMH-compliant verbal head: a V or v with a Merge feature bundle.

                    Constraint: [·V·] may only appear on v (not V), because V does not select VP — v does.

                    Instances For
                      def Minimalist.CMH.instReprCMHHead.repr :
                      CMHHeadStd.Format
                      Equations
                      • One or more equations did not get rendered due to their size.
                      Instances For
                        def Minimalist.CMH.mkVHead (hasD hasX : Bool) :

                        A V head: introduces internal arguments (complements and specifiers).

                        Equations
                        Instances For
                          def Minimalist.CMH.mkVLittleHead (hasD hasX : Bool) :

                          A v head: introduces external arguments and selects VP.

                          Equations
                          Instances For

                            Profile of a verb phrase: how many DP and XP arguments V and v collectively introduce. This determines the verb type.

                            The 4×4 matrix (V's args × v's args) produces all attested verb types, from weather verbs (0+0) to betting verbs (2DP + 2XP).

                            • vDPArgs :

                              DPs from V (0, 1, or 2)

                            • vXPArgs :

                              XPs from V (0 or 1)

                            • littleVDPArgs :

                              DPs from v (0 or 1)

                            • littleVXPArgs :

                              XPs from v (0 or 1)

                            Instances For
                              def Minimalist.CMH.instDecidableEqVPProfile.decEq (x✝ x✝¹ : VPProfile) :
                              Decidable (x✝ = x✝¹)
                              Equations
                              • One or more equations did not get rendered due to their size.
                              Instances For
                                Equations
                                • One or more equations did not get rendered due to their size.
                                Instances For

                                  Total number of DP arguments.

                                  Equations
                                  Instances For

                                    Total number of XP arguments.

                                    Equations
                                    Instances For

                                      Total argument count.

                                      Equations
                                      Instances For

                                        Construct a VPProfile from a V head and v head.

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

                                          Weather verbs: no arguments (V: ∅, v: [·V·])

                                          Equations
                                          Instances For

                                            Unergative: one DP subject (V: ∅, v: [·D·][·V·])

                                            Equations
                                            Instances For

                                              Unaccusative: one DP object (V: [·D·], v: [·V·])

                                              Equations
                                              Instances For

                                                Transitive: subject + object (V: [·D·], v: [·D·][·V·])

                                                Equations
                                                Instances For

                                                  Ditransitive with XP: e.g. "put the book on the table" (V: [·D·][·X·], v: [·D·][·V·])

                                                  Equations
                                                  Instances For

                                                    Ditransitive DOC: double object (V: [·D·], v: [·D·][·X·][·V·]) where the indirect object is a non-DP introduced by v

                                                    Equations
                                                    Instances For

                                                      Raising: V selects non-DP complement only (e.g., "seem [to VP]") (V: [·X·], v: [·V·])

                                                      Equations
                                                      Instances For

                                                        Maximal verb: both V and v bear both [·D·] and [·X·] (V: [·D·][·X·], v: [·D·][·X·][·V·]) — e.g., "bet"

                                                        Equations
                                                        Instances For

                                                          When V bears both [·D·] and [·X·], the XP merges first (as complement) and the DP second (as specifier).

                                                          This follows from Feature Maximality: V can check at most one Merge feature per step. If V first merges with an XP (checking [·X·]), the resulting V' can then merge with a DP (checking [·D·]). The reverse order is blocked because a DP cannot check [·X·].

                                                          Represented as an ordering constraint on the features.

                                                          Instances For
                                                            @[implicit_reducible]
                                                            Equations
                                                            Equations
                                                            • One or more equations did not get rendered due to their size.
                                                            Instances For
                                                              def Minimalist.CMH.nonDPFirst (_vHead : CMHHead) (_hD : _vHead.features.hasD = true) (_hX : _vHead.features.hasX = true) :

                                                              For a V head with both [·D·] and [·X·], the XP is the complement and the DP is the specifier.

                                                              Equations
                                                              Instances For

                                                                When v introduces an XP (v bears [·X·]), VP is forced to become a specifier of v rather than its complement.

                                                                This is because v's [·X·] must be checked before [·V·]: the XP merges as v's complement, and then VP merges as v's specifier.

                                                                Equations
                                                                Instances For

                                                                  When VP is a specifier of v, V and v do not form a complex head (they are not in a head-complement relation). This has consequences for verb movement and ellipsis.

                                                                  Equations
                                                                  Instances For

                                                                    A pending operation: a possible Merge at a given derivational step. Tracks which features would be checked and whether the operation bleeds any other pending operation.

                                                                    • featuresChecked :

                                                                      How many features this operation checks

                                                                    • bleeds : List

                                                                      Which other operations this one would bleed (make impossible)

                                                                    Instances For
                                                                      Equations
                                                                      • One or more equations did not get rendered due to their size.
                                                                      Instances For
                                                                        def Minimalist.CMH.weakEconomyValid (ops : List PendingOp) (chosenIdx : ) :
                                                                        Bool
                                                                        Equations
                                                                        Instances For
                                                                          def Minimalist.CMH.mutuallyBleeding (op1 op2 : PendingOp) (i j : ) :
                                                                          Bool

                                                                          When two operations stand in a bleeding relation (one bleeds the other), Weak Economy does not enforce a preference — either may apply. This captures the optionality in cases like Greek clitic doubling.

                                                                          Equations
                                                                          Instances For

                                                                            Specifier position in a tree with multiple specifiers. Under Generalized Tucking In, new specifiers merge below existing ones — as close to the head as possible.

                                                                            Instances For
                                                                              Equations
                                                                              • One or more equations did not get rendered due to their size.
                                                                              Instances For
                                                                                def Minimalist.CMH.tuckIn (_existingSpecs : ) :

                                                                                Tucking in: a new specifier merges below all existing specifiers. Given n existing specifiers, the new one gets position innermost and all existing specifiers shift outward by one.

                                                                                Equations
                                                                                Instances For

                                                                                  Types of nominals, ordered by feature-checking capacity. The CMH establishes an entailment hierarchy among nominal types: wh-DP ⊇ DP ⊇ non-DP in which Merge features each can check.

                                                                                  This hierarchy drives multiple predictions:

                                                                                  • KPs (inherent case → non-DP) cannot A-move (no [·D·] checking)
                                                                                  • wh-DPs check more features than plain DPs (DOMA)
                                                                                  • DPs can satisfy both [·D·] and [·X·] probes
                                                                                  Instances For
                                                                                    @[implicit_reducible]
                                                                                    Equations
                                                                                    Equations
                                                                                    • One or more equations did not get rendered due to their size.
                                                                                    Instances For

                                                                                      wh-DPs can additionally check [·wh·] on C. This is NOT a Merge feature on V/v but appears on functional heads (C). It is what makes wh-DPs strictly more powerful than plain DPs in the feature-checking hierarchy.

                                                                                      Equations
                                                                                      Instances For

                                                                                        Feature-checking capacity for A-movement (to Spec,TP via [·D·]) plus Ā-features ([·wh·]). This count drives Weak Economy comparisons in DOMA configurations.

                                                                                        Equations
                                                                                        Instances For

                                                                                          The entailment hierarchy: wh-DP ⊇ DP ⊇ non-DP. Each type can check a superset of the features checkable by the types below it.

                                                                                          Equations
                                                                                          Instances For

                                                                                            Can this nominal type undergo A-movement? Requires checking [·D·] on T.

                                                                                            Equations
                                                                                            Instances For

                                                                                              When two adjacent φ-probes copy features of the same goal, the lower copy is deleted at PF. This is the anti-redundancy principle that derives Agent Focus morphology in Mayan languages.

                                                                                              Adjacency is required: only probes with no intervening probe between them trigger deletion of the lower copy.

                                                                                              • probePosition :
                                                                                              • goalId :
                                                                                              Instances For
                                                                                                def Minimalist.CMH.instDecidableEqAgreeCopy.decEq (x✝ x✝¹ : AgreeCopy) :
                                                                                                Decidable (x✝ = x✝¹)
                                                                                                Equations
                                                                                                • One or more equations did not get rendered due to their size.
                                                                                                Instances For
                                                                                                  Equations
                                                                                                  • One or more equations did not get rendered due to their size.
                                                                                                  Instances For
                                                                                                    def Minimalist.CMH.probesAdjacent (copies : List AgreeCopy) (c1 c2 : AgreeCopy) :
                                                                                                    Bool

                                                                                                    Two probes are adjacent if no other probe in the configuration has a position strictly between them.

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

                                                                                                      Identify redundant copies: lower copies where an adjacent higher probe targets the same goal.

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

                                                                                                        Delete redundant copies. Returns surviving copies after PF deletion of redundant agreement.

                                                                                                        Equations
                                                                                                        Instances For

                                                                                                          Weather verbs have zero arguments.

                                                                                                          Unergatives have exactly one argument (the subject from v).

                                                                                                          Unaccusatives have exactly one argument (the object from V).

                                                                                                          Transitives have exactly two DP arguments.

                                                                                                          Ditransitives with an XP argument have three arguments.

                                                                                                          DOC ditransitives also have three arguments.

                                                                                                          Raising verbs have exactly one argument (the XP complement).

                                                                                                          Maximal verbs ("bet") have four arguments.

                                                                                                          In a low-XP ditransitive (V: [·D·][·X·], v: [·D·][·V·]), VP is NOT a specifier of v because v lacks [·X·].

                                                                                                          In a high-XP / DOC ditransitive (v: [·D·][·X·][·V·]), VP IS a specifier of v because v bears [·X·].

                                                                                                          When VP is a specifier, V and v do NOT form a complex head. This predicts VP-ellipsis mismatches: Voice mismatches are tolerated in simple transitives (V-v complex) but not in DOC ditransitives (VP as specifier).

                                                                                                          wh-DPs additionally check [·wh·] on C.

                                                                                                          A-movement accessibility: DPs and wh-DPs can A-move; non-DPs cannot. This is the core of the no-IO-passive explanation for languages with inherent case on indirect objects.

                                                                                                          The high-IO structure has VP as specifier.

                                                                                                          IO-to-subject in wh-passive: IO is a wh-DP. Feature count derived from NominalType.aMovementFeatures .whDP. Unlike Greek CD (§7), DOMA operations are pure alternatives at the same derivational step — not in a bleeding relation.

                                                                                                          Equations
                                                                                                          Instances For

                                                                                                            DO-to-subject in wh-passive: DO is a plain DP. Feature count derived from NominalType.aMovementFeatures .dp.

                                                                                                            Equations
                                                                                                            Instances For

                                                                                                              Weak Economy prefers IO-to-subject (checks more features).

                                                                                                              Weak Economy disprefers DO-to-subject when IO checks more.

                                                                                                              Two probes (C at height 3, v at height 1) agree with the same agent (goal 0).

                                                                                                              Equations
                                                                                                              Instances For

                                                                                                                Anti-redundancy identifies the lower copy as redundant.

                                                                                                                After anti-redundancy, only the higher copy (C's agreement) survives.

                                                                                                                theorem Newman2024.af_c_probe_survives :
                                                                                                                (Minimalist.CMH.applyAntiRedundancy mayanAFCopies).head? = some { probePosition := 3, goalId := 0 }

                                                                                                                The surviving copy is the one on C (height 3).

                                                                                                                In Greek active transitives, three operations compete after IO merges as XP:

                                                                                                                1. Clitic double the IO (checks [·D_weak·] + [uφ] = 2 features)
                                                                                                                2. Merge VP (checks [·V·] = 1 feature)
                                                                                                                3. Merge transitive subject DP (checks [·D·] = 1 feature)

                                                                                                                Weak Economy prefers (1) over (2) but NOT over (3): merging the subject checks [·D·] which also checks [·D_weak·], bleeding clitic doubling. Since (1) and (3) stand in a bleeding relation, Weak Economy allows either order — making clitic doubling optional.

                                                                                                                Equations
                                                                                                                Instances For
                                                                                                                  Equations
                                                                                                                  Instances For
                                                                                                                    Equations
                                                                                                                    Instances For

                                                                                                                      Clitic doubling is valid under Weak Economy.

                                                                                                                      Merging the subject is also valid (bleeding licenses either order).

                                                                                                                      CMH v with [·D·] introduces an external argument (thematic).

                                                                                                                      CMH v without [·D·] does not introduce an external argument.

                                                                                                                      Equations
                                                                                                                      Instances For

                                                                                                                        Voice's hasD (PF subcategorization) and assignsTheta (θ-role assignment) do NOT always coincide — confirming that these are distinct features. The CMH's [·D·] corresponds to assignsTheta, not to hasD. Anticausative and passive Voice have hasD = true (anticausative for SE marking, passive for by-phrase licensing) but assignsTheta = false (no θ-role).

                                                                                                                        The CMH adopts the obligatory-no-crash model for Merge features. This is @cite{newman-2024}'s Assumption 19c: features can go unchecked without crashing.

                                                                                                                        Equations
                                                                                                                        Instances For

                                                                                                                          Under Feature Failure, unchecked Merge features do not crash.

                                                                                                                          The crash model would reject derivations with unchecked features. Feature Failure is what separates the CMH from standard Minimalism on this point.

                                                                                                                          Weather verbs introduce zero DP arguments. In pro-drop languages without expletives, T's [·D·] goes unchecked — Feature Failure prevents a crash. Under the crash model, an expletive would be required to check T's [·D·].

                                                                                                                          Passive v lacks [·D·], so no external argument merges. If no internal argument is available for A-movement to Spec,TP (impersonal passives), T's [·D·] goes unchecked — tolerated by Feature Failure.

                                                                                                                          Low-XP: DO asymmetrically c-commands XP — binding asymmetry. DO (specifier of V, by Non-DP First) c-commands into the complement position.

                                                                                                                          DOC: neither internal argument c-commands the other — symmetric. DO is inside VP (outer specifier), IO is complement of v. They are in separate branches.

                                                                                                                          In the low-XP structure, the agent c-commands both internal arguments (VP is complement, fully inside agent's sister).

                                                                                                                          In the DOC structure with Tucking In, the agent (inner specifier) c-commands IO (complement of v) but NOT DO (inside VP, the outer specifier). The agent and VP are in separate branches — a direct consequence of VP being a specifier rather than a complement.