Documentation

Linglib.Studies.Benz2025

Benz (2025): Structure and Interpretation Across Categories #

Three case studies from [Ben25] (PhD dissertation, University of Pennsylvania) on the syntax–LF interface in German, all running on the contextual-allosemy substrate of DistributedMorphology.Allosemy. Nominalizations like Beobachtung carry event, referential, and content readings from a single syntactic structure, with the variation located in the allosemes of v and n (Ch. 3). The co-occurrence restrictions among prefixes, particles, and resultative secondary predicates follow from the conjunction of a phrase-structural factor and an event-structural one, [Ten94]'s Single Delimiting Constraint (Ch. 4, Table 3). The three nominalization types resolve the particle "structure problem" ([Lud01b]) in three different ways — phrasal inputs, particles-as-heads, outer attachment — from which the distribution of preverbal elements across them follows (Ch. 5).

The (32), (87)–(89), and (115) stimuli live in Data.Examples.Benz2025. availableReadings derives the reading inventory from the exponence engine's licensed allosemes, with adopted_unique characterizing the adopted analysis as the unique economical derivation of each reading and denoteN grounding the reading table in typed denotations (readingFromAllosemes_isSome_iff_denote); blocked_sound and blocked_complete show the two Ch. 4 principles exactly generate the co-occurrence paradigm; peAcceptable_from_solutions derives the Ch. 5 distribution from the structure-problem solutions.

Content nominalizations (Ch. 3) #

The three readings of Beobachtung #

The (32) stimulus rows, in margin-label order Event, RN, Content.

Equations
Instances For

    The reading a stimulus row exemplifies, from its reading feature. The paper's loose "RN" label is rendered as the simple entity reading.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      theorem Benz2025.beobachtung_diagnostics :
      List.map (fun (e : Data.Examples.LinguisticExample) => (e.feature? "duration_predicate", e.feature? "plural", e.feature? "cp_complement")) beobachtungRows = [(some "yes", some "no", some "no"), (some "no", some "yes", some "no"), (some "no", some "no", some "yes")]

      Each (32) example exhibits exactly its characteristic diagnostic — duration modification for the event reading, pluralization for the RN reading, CP complementation for the content reading ([Gri90]-style diagnostics).

      Readings from allosemes #

      The alloseme pair deriving each reading on the adopted analysis, where v is semantically vacuous on all readings but the CEN (§3.5, following [Woo23]). By adopted_roundtrip and adopted_unique, this is the unique derivation of each reading in which the two heads are not both contentful.

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

        Among derivations in which the two heads are not both contentful, the adopted pair is the only one deriving the reading — economy of interpretation pins the analysis.

        Alloseme selection in the nominalization structure #

        The selection side is DM's own engine, not a further table: allosemes are Vocabulary Items over neighborhoods, applicability is the Subset Principle, and canonical defaults are Elsewhere competition (winner?). The readings available in the (66)/(68) structure [n [v √]] are whatever the licensed allosemes compose to.

        v's context in the nominalization structure [n [v √]] — its complement is the root, event-entailing or not, and it is embedded under n.

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

          n's context in [n [v √]] — a verbal complement, eventive or not.

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

            The readings derivable in the nominalization structure: any licensed v alloseme composed with any licensed n alloseme.

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

              The canonical v alloseme of the root typology is the engine's Elsewhere winner: the more specified eventive entry beats vacuous v exactly when the root entails an event, so Verbalizer.Alloseme.fromRootType is derived, not stipulated.

              Every attested Beobachtung reading is available in the single structure: the engine licenses the allosemes and composition delivers the readings (Ch. 3's chapter claim).

              Without an event-entailing root neither the complex event nor the result reading is derivable; the content reading survives, since simple content nouns like Gerücht 'rumor' need no verbal source (Table 2).

              Typed alloseme denotations (Ch. 3) #

              The denotations of the deverbal allosemes, after [Woo23] as taken over in Ch. 3: nominalization semantics runs over a domain in which eventualities are entities, and each n alloseme builds an entity predicate from what v hands it. The reading typology is then derived, not tabulated: readings exist exactly where the composition is defined (readingFromAllosemes_isSome_iff_denote), and the analytical options for the result and content readings compose to identical entity predicates (result_options_pred_agree, content_options_agree).

              structure Benz2025.NominalizationModel (E : Type u_3) (S : Type u_4) :
              Type (max u_3 u_4)

              A model for nominalization denotations: eventualities embed into the entity domain (a nominal can describe an event as an entity), split into stative and dynamic, with result relating an entity to the eventuality that produced it and hasContent picking out the entities with propositional content.

              • ev : SE

                Eventualities as entities.

              • ev_injective : Function.Injective self.ev
              • stative : SProp

                Stative eventualities.

              • result : ESProp

                The entity is the result of the eventuality.

              • hasContent : EProp

                The entity has propositional content (rumor, idea, claim).

              Instances For
                structure Benz2025.RootMeaning (E : Type u_3) (S : Type u_4) :
                Type (max u_3 u_4)

                A root's contribution to nominalization semantics: what it says of entities and of eventualities, and its Theme relation — which entity an eventuality of the root's kind is predicated of.

                • onEntities : EProp
                • onEvents : SProp
                • theme : ESProp
                Instances For
                  inductive Benz2025.VerbalDenotation (E : Type u_3) (S : Type u_4) :
                  Type (max u_3 u_4)

                  What v hands to n: under the eventive alloseme, verbal event content together with the Theme position v introduces (§2.2); under the zero alloseme, the untouched root — and no argument position, since none is introduced by v (§3.5).

                  Instances For
                    structure Benz2025.NominalDenotation (E : Type u_3) (S : Type u_4) :
                    Type u_3

                    A nominal denotation: the entity predicate, together with the internal-argument relation when the nominal retains one. The relation is present exactly when v introduced the Theme position — no such position is part of the denotation unless v contributes it (§2.2, §3.5).

                    • pred : EProp

                      What the nominal describes.

                    • internalArg : Option (EEProp)

                      The internal-argument relation: internalArg y x holds when y saturates the nominal x's Theme position (the observation of the sky).

                    Instances For

                      The n alloseme applied to v's output, none where the combination is uninterpretable. The CEN describes the events the verb describes and retains the Theme position v introduced; the SEN predicates the root's entity content of an event-entity; the result alloseme picks out what an event of the root's kind produced ([Woo23]'s denotation); the content alloseme ignores the verbal layer entirely. The non-deverbal allosemes have their semantics in Semantics/Possessive/Relational.lean.

                      Equations
                      Instances For

                        The event and result readings are mirror images at the entity-predicate level: the two analytical options for the result reading — eventive v with n's result alloseme, or vacuous v with the same — agree on what the nominal describes (§3.5, crediting [Woo23]).

                        ...but not on argument structure: on the both-heads-interpreted option the result nominal retains the internal-argument position v introduced, on the v-vacuous option it has none. Since result nominals cannot saturate an internal argument, this is the reason for adopting the vacuous option for the RN reading (§3.5, following [Woo23]).

                        CENs retain argument structure: the complex event nominal carries the Theme position v introduced (the observation of the sky), which is what separates it from every zero-v reading (zero_v_no_argument_structure).

                        No zero-v reading has an internal-argument position: none is introduced by v, so none is part of the denotation (§3.5).

                        The content reading likewise ignores the verbal layer: both v options compose to hasContent, which is how simple content nouns can have the reading with no verbal source at all (§3.5).

                        The reading typology tracks denotational definedness: a (v, n) pair has a reading exactly when its composed denotation is defined.

                        A complex event nominal holds only of event-entities: the ground of its event reading (temporal modification, aspectual behavior).

                        Prefixes, particles, and resultatives (Ch. 4) #

                        The three preverbal-element types #

                        The three types of German preverbal elements (Ch. 4).

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

                            A morphological element is either a head (X⁰) or a phrase (XP).

                            Instances For
                              @[instance_reducible]
                              Equations
                              def Benz2025.instReprSynLevel.repr :
                              SynLevelStd.Format
                              Equations
                              Instances For
                                @[instance_reducible]
                                Equations

                                Prefixes are heads forming a complex head with the root (inseparable under V2 movement), while particles and RSPs are phrasal (stranded under V2). Following [Wur98] and [Zel01], particles are phrasal complements of the verb not dominated by further functional material.

                                Equations
                                Instances For

                                  Whether an element obligatorily introduces a result-state specification. Prefixes and RSPs always specify a result state, while particles can have non-delimiting (directional, completive) readings (§4.4).

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

                                      The two compatibility factors #

                                      def Benz2025.incorporationAllowed (outer inner : SynLevel) :
                                      Bool

                                      Structural combinability of an outer and an inner element, the inner one closer to the root. A phrasal inner element is impossible, since a head outside it cannot form a complex head with the root, and a phrasal outer element competes with it for the verb's single complement position (§4.4).

                                      Equations
                                      Instances For

                                        Structural combinability depends only on the inner element's level — derived, not stipulated.

                                        Two elements that both obligatorily specify a result state cannot co-occur, since the end state of a complex event can only be specified once — [Ten94]'s Single Delimiting Constraint (p. 79, quoted at (159)).

                                        Equations
                                        Instances For

                                          A combination is predicted possible iff both factors permit it (§4.4).

                                          Equations
                                          Instances For

                                            Prefix and particle inventory (Table 4) #

                                            German inseparable prefixes (Table 4). ge- is the rare non-participial prefix (ge-bären, ge-denken, ge-fallen).

                                            Equations
                                            Instances For

                                              German prepositional separable particles (Table 4, which additionally lists nominal and adjectival particles like klavier-, rad-, leicht-, whose classification is controversial).

                                              Equations
                                              Instances For

                                                Elements occurring both as prefix and as particle (Table 4).

                                                Equations
                                                Instances For
                                                  theorem Benz2025.ambiguous_not_pure :
                                                  (ambiguousElements.all fun (e : String) => !inseparablePrefixes.contains e && !separableParticles.contains e) = true

                                                  The ambiguous elements appear in neither pure inventory.

                                                  The co-occurrence paradigm (Table 3) #

                                                  A cell of Table 3's factor columns, recording whether the factor predicts the combination to be possible. particleDependent renders the table's parenthesized check mark, a prediction that depends on the specific particles involved (§4.1).

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

                                                      Boolean reading of a factor cell, on which particleDependent counts as possible since the generic classification treats particles as result-neutral.

                                                      Equations
                                                      Instances For

                                                        A row of Table 3 (repeated as Table 5). The printed table's final row merges pfx-RSP and PRT-RSP into one row "pfx/PRT-RSP", rendered here by a two-element outers list.

                                                        • outers : List PreverbalElement

                                                          The outer element(s) of the printed row.

                                                        • The inner element, closer to the root.

                                                        • allowed : Bool

                                                          The Allowed column.

                                                        • structureVerdict : FactorVerdict

                                                          The Structure Predicts column.

                                                        • interpretationVerdict : FactorVerdict

                                                          The Interpretation Predicts column.

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

                                                              Table 3, cell for cell.

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

                                                                The conjunction of the two factors reproduces the Allowed column at every cell of Table 3.

                                                                The structural factor reproduces the Structure column at every cell.

                                                                The interpretive factor reproduces the Interpretation column at every cell except the merged row's particle half (see prt_rsp_interpretation_particle_dependent).

                                                                The printed table marks the merged pfx/PRT-RSP row's interpretation cell ✗, but on the account's own classification only the prefix half is interpretively excluded. For PRT-RSP the structural factor does the work, while only "some RSP-particle verbs are additionally also ruled out semantically" (§4.4).

                                                                PRT-pfx is the unique allowed combination ((84) aus-er-wählen, an-ver-trauen, vor-ent-halten, um-ent-scheiden, ab-er-kennen).

                                                                Neither factor alone predicts the paradigm — structure alone misses the double-delimitation rows (pfx-pfx, RSP-pfx) and interpretation alone misses the phrase-structure rows (pfx-PRT, PRT-PRT, RSP-PRT).

                                                                Blocking derivations #

                                                                A proof that a preverbal-element combination violates one of the two principles. Soundness and completeness against predictedAllowed show the two principles exactly generate the paradigm.

                                                                Instances For
                                                                  theorem Benz2025.blocked_sound {o i : PreverbalElement} (h : Blocked o i) :
                                                                  predictedAllowed o i = false

                                                                  Every blocking derivation corresponds to a predicted-blocked combination — the theory does not over-generate.

                                                                  Every blocked combination has a derivation — the two principles account for all restrictions.

                                                                  The allowed combination has no derivation, since the prefix is a head and the particle is result-neutral.

                                                                  pfx-pfx is blocked only by the Single Delimiting Constraint (both are heads), so the interpretive rule is not redundant.

                                                                  pfx-PRT is blocked only structurally (particles are result-neutral), so the structural rule is not redundant.

                                                                  Cross-framework contrast: the phrase-in-word-slot cell #

                                                                  ConstructionGrammar.Slot.IsPhraseInWordSlot — a phrasal filler in a zero-level position — is the configuration of phrasal compounds and the PAL construction ([GS25]; contemporaneous with this dissertation, neither cites the other). In this file's terms that configuration is exactly the banned phrasal-inner cell: the structural principle rejects what the constructionist analysis licenses (cf. GoldbergShirtz2025.pal_load_bearing).

                                                                  theorem Benz2025.phraseInWordSlot_incorporation_banned (s : ConstructionGrammar.Slot String) (h : s.IsPhraseInWordSlot) :
                                                                  ∃ (site : SynLevel) (filler : SynLevel), Option.map SynLevel.ofBarLevel s.level = some site s.filler.synLevel = some filler incorporationAllowed site filler = false

                                                                  A phrase in a word-level slot occupies the banned cell — its site is a head and its filler a phrase, so incorporationAllowed rejects the pair. The standing counterexample class is the PAL construction, which licenses exactly this configuration.

                                                                  German resultative data (§4.2) #

                                                                  Complex predicate semantics after [Wil15], adopted at (158): ⟦v⟧ = λx λe₁ ∃e₂ ∃s. Means(e₁,e₂) & Pred₁(e₂) & Theme(e₁,x) & End(e₁,s) & Pred₂(s), where the M(eans) predicate is the verb, the R(esult) predicate the RSP, and the End Theme Postulate (108) links the complex event's Theme to the end state. See Causation.Resultatives for the complementary causal-dynamics analysis.

                                                                  theorem Benz2025.german_allows_non_unergative_M :
                                                                  List.map (fun (x : Data.Examples.LinguisticExample) => x.feature? "verb_class") [Examples.ex115a, Examples.ex115e, Examples.ex115f] = [some "obligatorily transitive", some "unaccusative", some "inherently reflexive"] (rspRows.all fun (x : Data.Examples.LinguisticExample) => x.judgment == Features.Judgment.acceptable) = true

                                                                  German allows non-unergative M predicates in resultatives ((115)), against weak-resultative reanalyses of the whole class.

                                                                  RSP co-occurrence contrasts (§4.1) #

                                                                  RSPs are incompatible with prefixed verbs, while the same RSP with the simplex verb is fine ((87), from [Cre20]). Each row's grammatical baseline carries its ungrammatical prefixed alternative, and the ge- of the baselines is participial, not a prefix in the relevant sense.

                                                                  theorem Benz2025.rsp_prt_contrast_pattern :
                                                                  ([Examples.ex88ab, Examples.ex88cd].all fun (e : Data.Examples.LinguisticExample) => e.feature? "blocker_type" == some "particle" && e.judgment == Features.Judgment.acceptable && !e.alternatives.isEmpty && e.alternatives.all fun (x : String × Features.Judgment) => x.2 == Features.Judgment.ungrammatical) = true

                                                                  RSPs are likewise incompatible with particle verbs ((88)), including particles not characterizable as resultative, like an- in (88d).

                                                                  Interpretive transparency #

                                                                  Whether the element can receive a non-transparent interpretation with the verb ((161) an-fangen 'start' vs. (162) platt klopfen 'pound flat'). RSPs are always interpreted transparently since they sit outside the locality domain for allosemy, while particles are bare phrasal complements local enough for it.

                                                                  Equations
                                                                  Instances For

                                                                    Particles pattern with RSPs structurally (both phrasal) but with prefixes interpretively (both can be non-transparent) — the evidence that particles are phrasal yet local enough for allosemy (§4.3).

                                                                    Prefixes in nominalizations (Ch. 5) #

                                                                    Nominalization types and the structure problem #

                                                                    The three German nominalization types discussed in Ch. 5.

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

                                                                        A solution to the particle "structure problem" ([Lud01b]) — how a phrasal particle can end up inside a derived nominal. Ch. 5 argues the three nominalization types favor different solutions, reinforcing "the strange status of particles in the grammar".

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

                                                                            The solution each nominalization type favors (§5.2–5.4). Nominalized infinitives take phrasal inputs (even das Durch-den-Wald-Reiten), -ung favors particles-as-heads, and Ge-...-e favors outer attachment ((223)), since its no-internal-argument eventive semantics conflicts with prefix verbs' argument-structural demands.

                                                                            Equations
                                                                            Instances For

                                                                              Whether the element can attach as a non-phrasal head. Prefixes always do, particles can attach low as heads (§5.3), and RSPs cannot be attached non-phrasally ((205)–(206)).

                                                                              Equations
                                                                              Instances For

                                                                                The preverbal elements a structure-problem solution accommodates. Phrasal inputs admit everything, particles-as-heads admits the head-attachers, and outer attachment admits the phrasal elements, since prefixes are verbal heads and cannot attach outside a noun.

                                                                                Equations
                                                                                Instances For

                                                                                  The observed Ch. 5 distribution. -ung takes prefixes and particles but not RSPs ((197), (204) *Platt-hämmer-ung). Ge-...-e takes particles ((212)) and a subset of RSPs ((216) das Wach-ge-küss-e, restricted to non-obligatorily-transitive bases) but not prefixes ((218) *Ge-be-such-e, in either affix order). Nominalized infinitives take all three ((193)).

                                                                                  Equations
                                                                                  Instances For

                                                                                    Each nominalization type admits exactly the elements its structure-problem solution accommodates — the Ch. 5 distribution projects the same head/phrase classification that drives the Ch. 4 paradigm.

                                                                                    Prefixes and RSPs are in complementary distribution across -ung and Ge-...-e, since head attachment and outer attachment make opposite demands.

                                                                                    Particles occur in both -ung and Ge-...-e — the dual attachment options (head or phrase) that constitute the structure problem.

                                                                                    Nominalized infinitives accept all three element types ((193)).

                                                                                    -ung and event structure #

                                                                                    Whether a verb can undergo -ung nominalization, which requires complex ("bi-eventive") change-of-state event structure ([RK10], endorsed at §5.3.1). Over this fragment's entries the bi-eventive verbs are the accomplishments (the simplex deadjectival cases of (199) are not represented).

                                                                                    Equations
                                                                                    Instances For

                                                                                      In the (198c) minimal pair *Mal-ung vs Be-mal-ung, the prefix supplies the complex change-of-state structure -ung needs, derived here from the fragment entries' Vendler classes.

                                                                                      Simplex activity verbs cannot form -ung nominalizations ((198) *Mal-ung, *Arbeit-ung, *Schieß-ung; the other fragment activities pattern identically).

                                                                                      Prefix and particle verbs with complex event structure form -ung nominalizations ((197) Ein-führ-ung, Ver-bind-ung; Ch. 3 Beobacht-ung).

                                                                                      Fragment grounding #

                                                                                      Transitivity derived from the fragment entry's typed fields.

                                                                                      Equations
                                                                                      Instances For

                                                                                        The stimulus rows' verb_class labels are derivable from the fragment entries their m_predicate features name, so changing frieren's unaccusative field or hämmern's complementType would break this theorem without touching the rows.

                                                                                        A property-concept root selects vacuous v, which with entity n yields frieren's simple entity reading. Allosemy leaves the eventive v available too, since the canonical alloseme is a default rather than a constraint.