Documentation

Linglib.Phenomena.Directives.Studies.RuytenbeekEtAl2017

@cite{ruytenbeek-etal-2017}: Indirect Request Processing, Sentence Types #

and Illocutionary Forces

Journal of Pragmatics 119 (2017) 46–62.

Two French eye-tracking experiments testing the literalist view that sentence types encode illocutionary force at the semantic level. Both studies support non-literalist theories: directive force in non-imperative constructions arises from semantic features they share with imperatives, not from sentence type per se.

Three mechanisms for non-imperative directive force #

The paper invokes three (overlapping) routes by which a non-imperative construction can carry directive force:

  1. Shared deontic semantics (@cite{kaufmann-2012}; paper §3.4 Discussion). A construction whose modal flavor matches the imperative's deontic semantics is directive-compatible. Vous devez VP and the permission reading of Vous pouvez VP (paper §3.4) instantiate this route. Formalised as SentType.deonticMatch.
  2. Preparatory-condition questioning (@cite{clark-1979}; paper §1 Introduction). An interrogative that questions an addressee preparatory condition (canonically .ability) is directive-compatible. Pouvez-vous VP? and Est-il possible de VP? instantiate this route. Formalised as SentType.prepConditionQueried; the deeper characterisation interrogative + circumstantial-modal is proved equivalent in prepConditionQueried_iff_interrog_circumstantial.
  3. Force-dynamic enablement (@cite{johnson-1987}, @cite{sweetser-1990}, @cite{talmy-2000}; paper §4 General Discussion, p. 61). The four constructions that semantically encode the addressee's enablement to perform the action — both interrogative IRs and both ability/possibility declaratives — pattern together. Formalised as SentType.enablementEncoded, derived directly from modalForce = some .possibility in this paper's domain. This is the broader generalisation that distinguishes the corrected formalisation from earlier versions that categorically denied directive force to Il est possible de VP.

The three mechanisms overlap. Vous pouvez VP fires (1) and (3); the two interrogative IRs fire (2) and (3); Il est possible de VP fires (3) only. The joint SentType.isDirective diagnostic licenses directive force when any mechanism fires (or when the construction is itself an imperative).

A note on empirical numbers #

Hand-transcribed RT means, fixation durations, and move-response proportions read off Figs. 3/5/6/8 in the original version of this file were either reconstructions of percentages reported in the paper text, upper-bound stipulations, or eyeball reads of bar charts that are demonstrably off (per the 2026-04-24 audit and 2026-05-13 PDF re-audit). The paper's actual claims are β estimates with confidence intervals (Study 1 §2.3, Study 2 §3.3) and significance tests. This file records the directional ordering predictions and lets statistics live in docstrings — Lean is not the right place for hand-transcribed regression coefficients.

Cross-paper bridges #

Mood substrate choice #

This file uses Core.Mood.IllocutionaryMood rather than the Minimalist-derived SAPMood. The two are isomorphic on the declarative/interrogative/imperative cases used here, and IllocutionaryMood is the canonical substrate (SAPMood adds syntax-derivation machinery the paper does not invoke). This also keeps the file independent of the Minimalist substrate.

Sentence types and modal projections #

source
inductive RuytenbeekEtAl2017.SentType :
Type

The eight sentence types appearing across Studies 1 and 2.

Study 1 §2.1 (p. 51) uses imperative, canYouInterrog, possibleInterrog, and ctrlInterrog (24 trials = 4 × 6). Study 2 §3.1 (p. 56) uses imperative (3 You must + 3 control imperatives), mustDeclarative, canDeclarative, possibleDecl, and plainDeclarative (24 trials). The 3 control imperatives in Study 2 collapse to imperative here since they are imperatives without a polar response option.

  • imperative : SentType

    Mettez le cercle rouge… — Study 1 + Study 2 imperative.

  • canYouInterrog : SentType

    Pouvez-vous VP? — Study 1 conventionalised IR.

  • possibleInterrog : SentType

    Est-il possible de VP? — Study 1 non-conventionalised IR.

  • ctrlInterrog : SentType

    Le cercle rouge est-il…? — Study 1 control interrogative.

  • mustDeclarative : SentType

    Vous devez VP — Study 2 deontic-necessity declarative.

  • canDeclarative : SentType

    Vous pouvez VP — Study 2 modal declarative. The paper's §3.4 Discussion attributes the directive readings to the permission sense of pouvoir (deontic possibility, not circumstantial ability); the paper's §4 General Discussion offers the broader force-dynamic enablement analysis (mechanism 3) as the unified pattern.

  • possibleDecl : SentType

    Il est possible de VP — Study 2 existential-possibility declarative. Per paper §2.1 the modal base is unrestricted (in the sense of @cite{kratzer-1991}); we choose .circumstantial as the canonical flavor since Pouvez-vous VP? (semantically close per p. 50) is restricted to the ability reading in the experimental contexts.

  • plainDeclarative : SentType

    Le cercle rouge est… — Study 2 control declarative.

Instances For
    source
    @[implicit_reducible]
    instance RuytenbeekEtAl2017.instDecidableEqSentType :
    DecidableEq SentType
    Equations
    source
    @[implicit_reducible]
    instance RuytenbeekEtAl2017.instReprSentType :
    Repr SentType
    Equations
    source
    def RuytenbeekEtAl2017.instReprSentType.repr :
    SentType → ℕ → Std.Format
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      source
      def RuytenbeekEtAl2017.SentType.mood :
      SentType → Core.Mood.IllocutionaryMood

      Morphosyntactic mood of each sentence type.

      Equations
      Instances For
        source
        def RuytenbeekEtAl2017.SentType.modalFlavor :
        SentType → Option Core.Modality.ModalFlavor

        Contextually salient modal flavor for each sentence type.

        The imperative branch is derived from Semantics.Modality.Assert.primaryFlavor rather than restipulated (layered grounding — see imperative_modalFlavor_eq_assert below).

        canDeclarative = .deontic follows the paper's §3.4 Discussion explanation of the Vous pouvez VP directive readings via the permission sense of pouvoir. canYouInterrog = .circumstantial follows the paper's §2.1 (p. 50) statement that the experimental contexts force the ability reading of pouvez/possible.

        ctrlInterrog and plainDeclarative carry no modal, hence the Option codomain.

        Equations
        Instances For
          source
          def RuytenbeekEtAl2017.SentType.modalForce :
          SentType → Option Core.Modality.ModalForce

          Modal force (necessity / possibility) for each sentence type.

          Equations
          Instances For
            source
            theorem RuytenbeekEtAl2017.imperative_modalFlavor_eq_assert :
            SentType.imperative.modalFlavor = some (Semantics.Modality.Assert.primaryFlavor Core.Mood.IllocutionaryMood.imperative)

            The imperative's flavor matches the one Hacquard's SAP architecture assigns to imperative speech acts — derived, not coincidentally equal. Layered Grounding for the headline @cite{kaufmann-2012} commitment.

            Mechanism 1 — shared deontic semantics (@cite{kaufmann-2012}) #

            A construction is mechanism-1-compatible with directive force when its modal flavor matches the imperative's. The check is a single equality on ModalFlavor. Mechanism 1 does not distinguish necessity from possibility — that ranking is a quantitative finding (paper §3.3: You must yields more directive interpretations than You can, z = -8.11, p < 0.001) which mechanism 1 alone underdetermines.

            source
            def RuytenbeekEtAl2017.SentType.deonticMatch (s : SentType) :
            Prop

            A sentence type is mechanism-1-compatible iff its modal flavor matches the imperative's. Encodes @cite{kaufmann-2012}'s deontic-modal account of imperatives: any construction with the same modal flavor is a candidate directive.

            Equations
            Instances For
              source
              @[implicit_reducible]
              instance RuytenbeekEtAl2017.instDecidablePredSentTypeDeonticMatch :
              DecidablePred SentType.deonticMatch
              Equations

              Mechanism 2 — preparatory-condition questioning (@cite{clark-1979}) #

              Per @cite{clark-1979}, asking about a preparatory condition for a request licenses the directive interpretation without sharing the imperative's modal semantics. The substrate Core.Discourse.PreparatoryCondition (Searle's hierarchy: ability / knowledge / memory / perception / permission / willingness) is the target type; the projection SentType.queriedPrep mirrors Phenomena/Politeness/Studies/FrancikClark1985.lean's RequestForm.queriedCondition.

              source
              def RuytenbeekEtAl2017.SentType.queriedPrep :
              SentType → Option Core.Discourse.PreparatoryCondition

              The preparatory condition queried by each sentence type, when one is queried. The two interrogative IRs both query .ability (per paper §2.1 p. 50, the experimental contexts force the ability reading). The empty .queriedPrep cases are the imperative, declaratives, and the control interrogative — none of which raise a polar question over an addressee precondition.

              Equations
              Instances For
                source
                def RuytenbeekEtAl2017.SentType.prepConditionQueried (s : SentType) :
                Prop

                Mechanism 2 directive licensing: the construction queries an addressee preparatory condition.

                Equations
                Instances For
                  source
                  @[implicit_reducible]
                  instance RuytenbeekEtAl2017.instDecidablePredSentTypePrepConditionQueried :
                  DecidablePred SentType.prepConditionQueried
                  Equations
                  source
                  theorem RuytenbeekEtAl2017.queriedPrep_eq_ability_iff_interrog_circumstantial (s : SentType) :
                  s.queriedPrep = some Core.Discourse.PreparatoryCondition.ability ↔ s.mood = Core.Mood.IllocutionaryMood.interrogative ∧ s.modalFlavor = some Core.Modality.ModalFlavor.circumstantial

                  Deeper characterisation of mechanism 2 in this paper's domain: queriedPrep = some .ability iff the construction is an interrogative with circumstantial modal flavor. This is the structural content of @cite{clark-1979}'s "convention of means" applied to the paper's specific stimulus set, where circumstantial modal force in an interrogative form picks out exactly the ability-questioning indirect requests.

                  Mechanism 3 — force-dynamic enablement (@cite{johnson-1987}, #

                  @cite{sweetser-1990}, @cite{talmy-2000})

                  Per the paper's §4 General Discussion (p. 61), all four constructions that semantically encode the addressee's enablement to perform the action — Pouvez-vous VP?, Vous pouvez VP, Est-il possible de VP?, and Il est possible de VP — pattern together as candidate directives. This is the broader generalisation that mechanism 2 (questioning only) misses: declarative ability/possibility constructions also encode the enablement-to-act semantic content, and the paper's data show that Il est possible de VP indeed receives directive interpretations (though fewer than Vous pouvez VP; z = -3.29, p = 0.0028, §3.3).

                  Force-dynamic enablement in this paper's domain reduces to "the construction has possibility modal force": the four enablement-encoding constructions are precisely the four with modalForce = some .possibility. The derivation is captured directly in the definition; the explicit four-construction list (matching the paper's prose at p. 61) appears in mechanism_attribution.

                  source
                  def RuytenbeekEtAl2017.SentType.enablementEncoded (s : SentType) :
                  Prop

                  Mechanism 3 directive licensing: the construction encodes a force-dynamic enablement pattern, characterised in this paper's domain by possibility modal force. The four ability/possibility constructions in the paper instantiate this.

                  Equations
                  Instances For
                    source
                    @[implicit_reducible]
                    instance RuytenbeekEtAl2017.instDecidablePredSentTypeEnablementEncoded :
                    DecidablePred SentType.enablementEncoded
                    Equations

                    Joint diagnostic #

                    The paper's non-literalist model: a sentence type is directive-compatible iff at least one mechanism fires (or the construction is itself an imperative). The imperative is direct (no indirection); Vous devez VP fires (1); Vous pouvez VP fires (1) and (3); the two interrogative IRs fire (2) and (3); Il est possible de VP fires (3) only; the two controls fire none.

                    source
                    def RuytenbeekEtAl2017.SentType.isDirective (s : SentType) :
                    Prop

                    Joint diagnostic: directive force is licensed by mechanism 1, 2, or 3 (or by being an imperative outright).

                    Equations
                    Instances For
                      source
                      @[implicit_reducible]
                      instance RuytenbeekEtAl2017.instDecidablePredSentTypeIsDirective :
                      DecidablePred SentType.isDirective
                      Equations

                      Predictions #

                      Six of the eight sentence types are directive (imperative + the five modal sentences); the two non-modal controls are not. The joint isDirective diagnostic produces this partition. The mechanism-attribution table records which mechanism licenses each directive sentence type — the empirical content of the paper's non-literalist analysis.

                      source
                      theorem RuytenbeekEtAl2017.joint_diagnostic_partitions_sentence_types :
                      SentType.imperative.isDirective ∧ SentType.mustDeclarative.isDirective ∧ SentType.canDeclarative.isDirective ∧ SentType.canYouInterrog.isDirective ∧ SentType.possibleInterrog.isDirective ∧ SentType.possibleDecl.isDirective ∧ ¬SentType.ctrlInterrog.isDirective ∧ ¬SentType.plainDeclarative.isDirective

                      The paper's headline qualitative finding (§4 General Discussion): the joint diagnostic correctly partitions the eight sentence types into directive (imperative + 5 modal sentences) and non-directive (2 controls).

                      source
                      theorem RuytenbeekEtAl2017.mechanism_attribution :
                      SentType.imperative.deonticMatch ∧ SentType.mustDeclarative.deonticMatch ∧ SentType.canDeclarative.deonticMatch ∧ ¬SentType.canYouInterrog.deonticMatch ∧ ¬SentType.possibleInterrog.deonticMatch ∧ ¬SentType.possibleDecl.deonticMatch ∧ SentType.canYouInterrog.prepConditionQueried ∧ SentType.possibleInterrog.prepConditionQueried ∧ ¬SentType.mustDeclarative.prepConditionQueried ∧ ¬SentType.canDeclarative.prepConditionQueried ∧ ¬SentType.possibleDecl.prepConditionQueried ∧ SentType.canYouInterrog.enablementEncoded ∧ SentType.possibleInterrog.enablementEncoded ∧ SentType.canDeclarative.enablementEncoded ∧ SentType.possibleDecl.enablementEncoded ∧ ¬SentType.mustDeclarative.enablementEncoded

                      Mechanism-attribution table: which sentence types each mechanism licenses, and which it rejects. Documents the empirical content of the three-mechanism analysis: mechanism 1 (deontic match) fires on the imperative, Vous devez VP, and Vous pouvez VP; mechanism 2 (preparatory-condition questioning) fires only on the two interrogative IRs; mechanism 3 (force-dynamic enablement) fires on all four ability/possibility constructions.

                      Force–type mismatch (the headline anti-literalist claim) #

                      The paper's core theoretical claim (§4 General Discussion) is that sentence type does NOT encode illocutionary force: declarative sentence types can carry directive force (via mechanisms 1/3 for Vous devez VP, Vous pouvez VP, and Il est possible de VP), and interrogative sentence types can carry directive force (via mechanisms 2/3 for the two IRs).

                      The literalist default is the Searle taxonomy: declarative mood maps to assertive force, interrogative mood maps to question-class directives (asking, not commanding). Both defaults are falsified by the existence of force–type mismatch witnesses below.

                      source
                      @[reducible, inline]
                      abbrev RuytenbeekEtAl2017.SentType.forceTypeMismatch (s : SentType) (m : Core.Mood.IllocutionaryMood) :
                      Prop

                      A sentence-type carries a force–type mismatch when its morphosyntactic mood is m but it nonetheless licenses directive interpretation. The five mismatch cases (mustDeclarative, canDeclarative, possibleDecl, canYouInterrog, possibleInterrog) all instantiate this.

                      Equations
                      Instances For
                        source
                        theorem RuytenbeekEtAl2017.declarative_default_searleClass :
                        Core.Mood.IllocutionaryMood.declarative.searleClass = Core.Discourse.SearleClass.assertive

                        The literalist Searle default for declarative mood is .assertive, not .directive. The paper's data falsify the consequence "declarative-mood utterances cannot carry directive force" (witnessed by anti_literalism_for_declaratives).

                        source
                        theorem RuytenbeekEtAl2017.anti_literalism_for_declaratives :
                        ∃ (s : SentType), s.mood = Core.Mood.IllocutionaryMood.declarative ∧ s.isDirective

                        Anti-literalism witness: there is a sentence type whose morphosyntactic mood is declarative yet which licenses directive interpretation. (mustDeclarative, canDeclarative, and possibleDecl all witness this; mustDeclarative is the canonical paper Study 2 case.)

                        source
                        theorem RuytenbeekEtAl2017.anti_literalism_for_interrogatives :
                        ∃ (s : SentType), s.mood = Core.Mood.IllocutionaryMood.interrogative ∧ s.isDirective

                        Anti-literalism witness: there is a sentence type whose morphosyntactic mood is interrogative yet which licenses directive interpretation. (canYouInterrog and possibleInterrog both witness this; either is the canonical paper Study 1 case.)

                        source
                        theorem RuytenbeekEtAl2017.nonImperative_force_type_mismatches :
                        SentType.mustDeclarative.forceTypeMismatch Core.Mood.IllocutionaryMood.declarative ∧ SentType.canDeclarative.forceTypeMismatch Core.Mood.IllocutionaryMood.declarative ∧ SentType.possibleDecl.forceTypeMismatch Core.Mood.IllocutionaryMood.declarative ∧ SentType.canYouInterrog.forceTypeMismatch Core.Mood.IllocutionaryMood.interrogative ∧ SentType.possibleInterrog.forceTypeMismatch Core.Mood.IllocutionaryMood.interrogative

                        The five non-imperative sentence types that carry directive force — declarative-form mismatches and interrogative-form mismatches alike.

                        Bridge to French Fragment #

                        The SentType.modalFlavor and SentType.modalForce projections are consistent with the lexical entries in Fragments/French/Modals.lean: each sentence type's force–flavor pair appears in the corresponding modal verb's forceFlavors list. This is a derive-don't-duplicate consistency check — changing a Fragment entry's flavor inventory will break these theorems.

                        source
                        theorem RuytenbeekEtAl2017.mustDeclarative_forceFlavor_in_devoir :
                        { force := Core.Modality.ModalForce.necessity, flavor := Core.Modality.ModalFlavor.deontic } ∈ Fragments.French.Modals.devoir.forceFlavors

                        Vous devez VP = deontic necessity, present in devoir's force-flavor inventory.

                        source
                        theorem RuytenbeekEtAl2017.canDeclarative_forceFlavor_in_pouvoir :
                        { force := Core.Modality.ModalForce.possibility, flavor := Core.Modality.ModalFlavor.deontic } ∈ Fragments.French.Modals.pouvoir.forceFlavors

                        Vous pouvez VP = deontic possibility (permission), present in pouvoir's force-flavor inventory.

                        source
                        theorem RuytenbeekEtAl2017.possibleDecl_forceFlavor_in_ilEstPossible :
                        { force := Core.Modality.ModalForce.possibility, flavor := Core.Modality.ModalFlavor.circumstantial } ∈ Fragments.French.Modals.ilEstPossible.forceFlavors

                        Il est possible de VP = circumstantial possibility, present in the impersonal construction's force-flavor inventory.