Documentation

Linglib.Phenomena.Focus.Studies.Rooth1992

@cite{rooth-1992} Bridge — Focus Interpretation @cite{rooth-1992} #

Bridges the empirical data in Focus/Basic.lean to the formal theory in Focus/Interpretation.lean (FIP, Q-A congruence), with a full compositional derivational chain through Montague semantics and connection to English fragment entries.

Pipeline #

Fragments/English/Nouns ──▷ Montague Lexicon ──▷ Tree
                                                        │
                                                    interp
                                                        │
                                                        ▼
                              propositions (Set QAWorld)
                                                        │
                                                   set literals
                                                        │
                                                        ▼
                                              PropFocusValue = Set (Set QAWorld)
                                                        │
                                                   FIP / qaCongruent
                                                        │
                                                        ▼
                                              Bridge theorems ↔ Data

Model #

What's exercised #

Worlds crossing subject (Fred/Mary) × object (beans/rice). Sufficient to distinguish subject-focus from object-focus alternative sets.

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

      Focused "FRED" in "FRED ate the beans" (Rooth §2.4, ex. 23a): O-value = "Fred"; A-value = {"Fred", "Mary"}.

      Equations
      Instances For

        Non-focused "ate the beans": singleton A-value (no alternatives). Exercises AltMeaning.unfeatured.

        Equations
        Instances For

          Unfeatured O-value equals the input.

          Unfeatured A-value is a singleton containing the O-value. Non-focused expressions evoke no alternatives (@cite{rooth-1992} §1).

          Focus partition of "FRED ate the beans": Fred is focused, evoking {Fred, Mary} as alternatives (Rooth §2.4, ex. 25a).

          Equations
          Instances For

            Background of "FRED ate the beans": the non-focused material.

            Equations
            Instances For

              Theme: the QUD presupposition "_ ate the beans" (λ-abstract). Rooth §2.4: in a Q-A pair, the theme corresponds to the question's content.

              Equations
              Instances For

                Rheme: the answer "Fred".

                Equations
                Instances For

                  Full information structure of "FRED ate the beans" in response to "Who ate the beans?"

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

                    Theme carries the presupposed content.

                    Rheme carries the asserted answer.

                    Focus list matches the focused element.

                    Background list matches the background elements.

                    Minimal derivation type for exercising the IS partition. Pairs a focused constituent with background material.

                    • focusedWord : String
                    • backgroundWords : List String
                    Instances For
                      Equations
                      • One or more equations did not get rendered due to their size.
                      Instances For

                        A FocusedSentence determines an InfoStructure.

                        (Previously a HasInfoStructure FocusedSentence String instance — the typeclass shape was deleted in the 0.230.489 cleanup since no caller dispatched on it.)

                        Equations
                        Instances For
                          Equations
                          Instances For

                            The extractor correctly puts the focused word in foci.

                            The extractor correctly puts background words in background.

                            @cite{rooth-1992} §2.4, constraint (26d): In a Q-A pair ⟨ψ, α⟩, ⟦ψ⟧° ⊆ ⟦α⟧f. The ordinary semantic value of the question is a subset of the focus semantic value of the answer.

                            "Who ate the beans?" — Hamblin question with subject alternatives. ⟦Q⟧° = {fredAteBeans, maryAteBeans} (Rooth §2.4, ex. 24).

                            Equations
                            Instances For

                              Focus value of "[FRED]_F ate the beans" — same subject alternatives. ⟦A⟧f = {fredAteBeans, maryAteBeans} (Rooth §2.4, ex. 25a).

                              Equations
                              Instances For

                                Focus value of "Fred ate the [BEANS]_F" — object alternatives. ⟦A⟧f = {fredAteBeans, fredAteRice} (varies object, not subject).

                                Equations
                                Instances For

                                  Q-A congruence: subject focus value = question denotation. ⟦[FRED]_F ate the beans⟧f = ⟦Who ate the beans?⟧ (Rooth §2.4).

                                  FIP (27) holds: question alternatives ⊆ focus value. Trivially satisfied when the sets are equal.

                                  "maryAteBeans" is in the question alternatives...

                                  ...but it is NOT in the object-focus alternative set...

                                  ...so FIP fails for object focus, explaining why "#Fred ate the BEANS" is not a congruent answer to "Who ate the beans?"

                                  Rooth §2.1, constraint (26a): the domain of quantification C of a focusing adverb is a subset of the focus semantic value ⟦α⟧f.

                                  Rooth's formalization (30b): only(C) introduces
                                    ∀P[P ∈ C ∧ P(m) → P = VP']
                                  where C is constrained by the FIP to be a set of properties
                                  matching the focus semantic value. 
                                  

                                  Worlds for the "only" model: who Mary introduced to Sue.

                                  Instances For
                                    @[implicit_reducible]
                                    Equations
                                    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

                                        Focus on BILL (Rooth §2.1, ex. 3a): O-value = introBill; A-value = {introBill, introJohn}. Focus determines the domain of "only".

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

                                          "Only" with focus on BILL: O-value holds and all non-actual alternatives are excluded (Rooth §2.1, (30b)).

                                          "Only" with focus on JOHN: symmetric case.

                                          Different focus → different "only" meaning. Focus on BILL excludes John; focus on JOHN excludes Bill (Rooth §2.1, exs. 3a vs 3b).

                                          Focus in the congruent answer matches the data.

                                          Focus in the incongruent answer matches the data.

                                          The data (Basic.lean) says "FRED ate the beans" is congruent and "#Fred ate the BEANS" is incongruent with "Who ate the beans?". The theory (FIP, §6) explains:

                                          - Subject focus produces a focus value equal to the question
                                            denotation (§6a), so FIP is satisfied.
                                          - Object focus produces a focus value that differs (§6b):
                                            maryAteBeans ∈ ⟦Q⟧° but maryAteBeans ∉ ⟦A⟧f, so FIP fails.
                                          
                                          For "only" (§7), the data says focus determines what "only"
                                          excludes. The theory confirms: the FIP constrains the domain C
                                          of "only" to be a subset of the focus value, so different focus
                                          positions yield different exclusion domains. 
                                          

                                          The propositions in §2 were hand-defined. Here we derive them compositionally: entity denotations + a world-indexed verb meaning are combined via direct function application and Heim & Kratzer's interp to produce the same truth conditions.

                                          The derivational chain is:
                                            Fragment entry → Montague LexEntry → Tree → interp → Bool
                                          run once per world to yield a world-indexed proposition. 
                                          

                                          Entity domain for the focus model.

                                          • fred : E
                                          • mary : E
                                          • beans : E
                                          • rice : E
                                          Instances For
                                            @[implicit_reducible]
                                            Equations
                                            Equations
                                            • One or more equations did not get rendered due to their size.
                                            Instances For

                                              Syntax tree: [S [NP Fred] [VP [V ate] [NP beans]]]

                                              Equations
                                              Instances For

                                                Syntax tree: [S [NP Mary] [VP [V ate] [NP beans]]]

                                                Equations
                                                Instances For

                                                  Syntax tree: [S [NP Fred] [VP [V ate] [NP rice]]]

                                                  Equations
                                                  Instances For

                                                    Extract the Prop truth value from a tree interpretation. Returns none if the tree is uninterpretable or has non-t type.

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

                                                      The propositions from §2 were stipulated directly. Here we show they are derivable: running interp at each world produces the same truth values.

                                                      Compositionally derived "Fred ate beans" proposition.

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

                                                        Compositionally derived "Mary ate beans" proposition.

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

                                                          Compositionally derived "Fred ate rice" proposition.

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

                                                            Direct function application matches tree interpretation.

                                                            Grounding: compositional "Fred ate beans" agrees with hand-defined proposition at each world.

                                                            Grounding: compositional "Mary ate beans" = direct function application.

                                                            Grounding: compositional "Fred ate rice" = direct function application.

                                                            Connect the model's lexicon to English fragment entries. Fragment entries provide morphological and syntactic properties; the bridge verifies that these properties are consistent with the model and that fragment surface forms feed the compositional lexicon.

                                                            Fred is a proper name in the English fragment.

                                                            Mary is a proper name in the English fragment.

                                                            Fragment surface forms feed the Montague lexicon. The form field of each fragment entry matches a lexicon key.

                                                            "eat" has past tense "ate" matching the lexicon entry.

                                                            The complete derivational chain from fragments to FIP:

                                                            1. Fragment entries (§14) provide surface forms and properties
                                                            2. Surface forms feed the Montague lexicon (§10)
                                                            3. Tree derivations compose meanings via interp (§11)
                                                            4. Running at each world yields propositions grounding §2 (§12)
                                                            5. Propositions build Hamblin questions and focus values (§6)
                                                            6. FIP/qaCongruent proves congruence (§6a) or incongruence (§6b)
                                                            7. Theoretical predictions match empirical judgments (§9) 
                                                            

                                                            End-to-end: the compositional derivation produces the same truth values as the hand-defined propositions used to build the Hamblin question. At each world, the tree interp matches the hand-defined proposition.