Documentation

Linglib.Studies.Ferreira2023

Ferreira (2023): A square of necessities #

[Fer23]

A square of necessities: X-marking weak and strong necessity modals. Semantics and Pragmatics 16, Article 8: 1–54.

Core Contributions #

  1. Portuguese has a tripartite modal system (poder < dever < ter que) where weak necessity is lexicalized as a distinct root, unlike Spanish (deber = strong necessity) or English (ought = ambiguous).

  2. Both weak and strong necessity modals can be X-marked via past imperfect morphology (devia, tinha que), but X-marking does NOT weaken modal force — it shifts modal parameters (modal base or ordering source).

  3. Two independent X-marking operations generate a 2×2 square of necessities: Xf (modal base revision) and Xg (ordering source revision). Portuguese instantiates all four vertices.

  4. WN ≡ SN_Xg: weak necessity is strong necessity with X-marked ordering source — the secondary ordering favors the prejacent among best worlds.

Square Instantiation & Entailment Diamond #

    tem que ──Xf──→ tinha que
       │                │
       Xg               Xg
       │                │
     deve ────Xf──→ devia

Entailment flows downward through both paths (SN → SN_Xf → SN_Xfg and SN → SN_Xg → SN_Xfg), forming a diamond. No reverse entailments hold.

X-marking substrate #

Star-revision: X-marking on modal bases (Xf) #

Property: f' is a ∗-revision of f for p ([Fer23]). A ∗-revision widens the modal domain by adding p-worlds: (1) every world accessible under f remains accessible under f'; (2) every newly accessible world satisfies p.

Instances For

    Double-star-revision: X-marking on ordering sources (Xg) #

    Xg: X-marking targeting the ordering source (∗∗-revision). Adds a secondary ordering that favors p-worlds among the best worlds.

    Equations
    Instances For

      The four vertices of the square #

      @[reducible, inline]
      Equations
      Instances For
        @[reducible]
        Equations
        Instances For

          Key equation: WN ≡ SN_Xg #

          Entailment: SN → SN_Xg (must → ought) #

          theorem Ferreira2023.snXg_not_entails_sn :
          ¬∀ (W : Type) (f : Semantics.Modality.Kratzer.ModalBase W) (g : Semantics.Modality.Kratzer.OrderingSource W) (p : WProp) (w : W), snXg f g p wsn f g p w

          The converse fails: SN_Xg ⊭ SN.

          Counterexample: W = Bool, f = universal access, g = trivial ordering, p = (· = true). Then snXg holds (xMarkOrdering favors true so best = {true}), but sn fails (all accessible worlds best under empty ordering, p false is false).

          Forward entailment: SN → SN_Xf under star-revision #

          theorem Ferreira2023.sn_entails_snXf {W : Type u_1} (f f' : Semantics.Modality.Kratzer.ModalBase W) (g : Semantics.Modality.Kratzer.OrderingSource W) (p : WProp) (w : W) (hRev : IsStarRevision f f' p) (hSN : sn f g p w) :
          snXf f' g p w

          Forward entailments along square edges #

          theorem Ferreira2023.snXg_entails_snXfg {W : Type u_1} (f f' : Semantics.Modality.Kratzer.ModalBase W) (g : Semantics.Modality.Kratzer.OrderingSource W) (p : WProp) (w : W) (hRev : IsStarRevision f f' p) (h : snXg f g p w) :
          snXfg f' g p w

          Non-entailment: reverse arrows fail #

          Xf preserves the quantifier: SN_Xf is still ∀ over best worlds.

          @[reducible, inline]
          Equations
          Instances For

            Portuguese modal typology #

            The six Portuguese modal forms: three roots × two tense markings.

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

                Ascending scale of modal force (§2) #

                X-marking preserves force (§3) #

                X-marking does not change modal force: each pair shares force.

                Entailment judgments (§2) #

                ter que p ⊨ dever p: strong necessity entails weak. Follows from Directive.strong_entails_weak — the Xg-refined best worlds are a subset of the unrefined best worlds.

                dever p ⊭ ter que p: weak necessity does not entail strong.

                dever p ⊨ poder p: weak necessity entails possibility, completing the ascending scale poder p < dever p < ter que p. Requires seriality (nonempty best worlds) — the D axiom.

                Consistency judgments (§2) #

                dever p ∧ ¬p is consistent: weak necessity is compatible with the prejacent being false ("Este homem deve ter sido assassinado, mas ele pode não ter sido").

                ter que p ∧ ¬p is contradictory when the base is realistic: if w ∈ ∩f(w) and all best worlds satisfy p, then w satisfies p (by the T axiom).

                Non-entailment between present and past forms (§3) #

                devia p ⊬ deve p: X-marking the modal base (via Xf) widens the domain, and the new p-worlds can change the best set under the refined ordering.

                Note: the reverse direction (deve p ⊨ devia p) DOES hold — see PortugueseSquare.deve_entails_devia. This follows from sn_entails_snXf applied to the refined ordering: ∗-revision only adds p-worlds, which cannot worsen the truth of the prejacent among best worlds.

                Square instantiation: Portuguese occupies all four vertices #

                The square of necessities applied to Portuguese modal verbs.

                Each field maps to a vertex of the square:

                • sn = tem que (strong necessity, unmarked)
                • snXf = tinha que (strong necessity, X-marked modal base)
                • snXg = deve (= weak necessity, X-marked ordering source)
                • snXfg = devia (weak necessity, X-marked modal base)
                Instances For

                  tem que: top-left vertex (SN).

                  Equations
                  Instances For

                    deve: bottom-left vertex (SN_Xg = WN).

                    Equations
                    Instances For

                      tinha que: top-right vertex (SN_Xf).

                      Equations
                      Instances For

                        devia: bottom-right vertex (SN_{Xf,g}).

                        Equations
                        Instances For

                          tem que ⊨ deve: top-left entails bottom-left (SN → SN_Xg).

                          Forward entailment under star-revision (§3) #

                          tem que ⊨ tinha que: SN entails SN_Xf under ∗-revision. Follows from sn_entails_snXf: best worlds in the wider domain either (a) were already best in the narrower domain, or (b) are new p-worlds.

                          deve ⊨ devia: SN_Xg entails SN_{Xf,g} under ∗-revision (bottom-left → bottom-right). Follows from snXg_entails_snXfg.

                          tinha que ⊨ devia: SN_Xf entails SN_{Xf,g} (top-right → bottom-right). Follows from snXf_entails_snXfg.

                          Entailment diamond #

                          The four vertices form a Hasse diagram — entailment flows from SN downward to SN_Xfg through both intermediate vertices:

                                 tem que (SN)
                                  ╱        ╲
                            tinha que     deve
                              (SN_Xf)    (SN_Xg)
                                  ╲        ╱
                                 devia (SN_Xfg)
                          

                          English ambiguity (§3) #

                          English ought/should is ambiguous between two square vertices: non-X-marked WN (SN_Xg) and X-marked WN (SN_{Xf,g}).

                          Portuguese disambiguates overtly: deve vs devia. English collapses them into one form.

                          English fragment bridge (§3) #

                          English should and ought (from FunctionWords) are both classified as .weakNecessity — the SN_Xg vertex of the square. But unlike Portuguese, English lacks overt X-marking morphology (deve vs devia), so the SN_Xfg reading (counterfactual should) is available but not distinguished.

                          Note: should carries tense := .Past (morphological past = X-marking), while ought carries no tense marking. Both are semantically present-tense weak necessity in their default readings.

                          English should has morphological past tense (X-marking), but ought does not. This reflects Iatridou's generalization: X-marking in English is realized as past morphology. Portuguese makes this overt: deve (unmarked) vs devia (past imperfect = X-marked).