Documentation

Studies.HartshorneEtAl2016

Hartshorne et al. (2016): Psych verbs, the linking problem, and the acquisition of language #

This file formalizes [HODS+16]'s resolution of the psych-verb linking problem. Fear-type verbs (Agnes feared Bartholomew) and frighten-type verbs (Bartholomew frightened Agnes) describe the same emotions with reversed linking, the strongest apparent case against systematic mappings from meaning to form. The paper's proposal is that they lexicalize two conceptualizations of emotion, a habitual attitude of an experiencer directed at a target and an episode in which a stimulus causes an experiencer to be in an emotional state, and nine experiments show the distinction to be systematic (fear-type states are rated longer-lasting and the subject of a frighten-type verb is judged causally responsible, in Mandarin and Korean as in English), productive (adults in English, Japanese and Russian give novel verbs with attitude meanings experiencer-subject syntax and those with episode meanings experiencer-object syntax) and early (four- and five-year-olds do the same). The linking itself follows from one principle over the semantic structures of Fig. 11, prominence preservation: Sem.subject is the least embedded argument of such a structure, and since the causer of CAUSE and the holder of BE are highest whatever they embed (subject_cause), the experiencer heads the attitude and the stimulus the episode (attitude_subject, episode_subject), while the causation judgments answer to whether the structure contains CAUSE at all. The linking agrees with [BR88]'s Class I and Class II as recorded in Pesetsky1995 and with [Lev93]'s admire and amuse classes (agrees_with_levin).

Implementation notes #

The structures follow the prose of §5.4.1 rather than the figures: BE takes the entity in a state and the state, the state of an attitude is the root directed at a target, and CAUSE takes the causer and the caused BE, with no BECOME. Duration is not encoded in the structures, and the experiments' rates stay in prose: fear-type verbs were rated longer-lasting than frighten-type verbs, with 42 percent of the former above every one of the latter; causal responsibility fell on the subject for 214 of 216 frighten-type verbs and on no consistent participant for fear-type verbs; and frighten-type syntax was chosen for novel episode verbs over novel attitude verbs by adults (68 against 38 percent in English) and by children (66 against 33 percent at ages four to five). The paper's examples are the rows of Data/Examples/HartshorneEtAl2016.json.

TODO #

References #

The argument positions of Fig. 11: the experiencer, the target an attitude is directed at, and the stimulus that causes an episode.

Instances For
    @[instance_reducible]
    Equations
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      inductive HartshorneEtAl2016.Sem (α : Type) :

      A semantic structure (§5.4.1, Figs. 10–11): primitive predicates embedding one another, variables marking argument positions and the verbal root modifying a state.

      • var {α : Type} (x : α) : Sem α

        An argument position.

      • root {α : Type} : Sem α

        The verbal root: the kind of emotion.

      • be {α : Type} (x s : Sem α) : Sem α

        BE(x, s): x is in the state s.

      • cause {α : Type} (x e : Sem α) : Sem α

        CAUSE(x, e): x brings about e.

      • about {α : Type} (s y : Sem α) : Sem α

        The state s directed at y.

      Instances For
        @[instance_reducible]
        instance HartshorneEtAl2016.instDecidableEqSem {α✝ : Type} [DecidableEq α✝] :
        DecidableEq (Sem α✝)
        Equations
        def HartshorneEtAl2016.instDecidableEqSem.decEq {α✝ : Type} [DecidableEq α✝] (x✝ x✝¹ : Sem α✝) :
        Decidable (x✝ = x✝¹)
        Equations
        Instances For
          def HartshorneEtAl2016.instReprSem.repr {α✝ : Type} [Repr α✝] :
          Sem α✝ → ℕ → Std.Format
          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            @[instance_reducible]
            instance HartshorneEtAl2016.instReprSem {α✝ : Type} [Repr α✝] :
            Repr (Sem α✝)
            Equations
            def HartshorneEtAl2016.Sem.vars {α : Type} :
            Sem α → List (α × ℕ)

            The argument positions of a structure with their depth of embedding, left to right.

            Equations
            Instances For
              def HartshorneEtAl2016.Sem.subject {α : Type} (s : Sem α) :
              Option α

              Prominence preservation: the least embedded argument position becomes the subject.

              Equations
              • s.subject = Option.map Prod.fst (List.argmin Prod.snd s.vars)
              Instances For

                Whether the structure contains CAUSE.

                Equations
                Instances For
                  @[instance_reducible]
                  Equations
                  theorem HartshorneEtAl2016.Sem.one_le_snd_of_mem_vars_map {α : Type} {l : List (α × ℕ)} {q : α × ℕ} (hq : q ∈ List.map (fun (p : α × ℕ) => (p.1, p.2 + 1)) l) :
                  1 ≤ q.2
                  theorem HartshorneEtAl2016.Sem.subject_cause {α : Type} (x : α) (e : Sem α) :
                  ((var x).cause e).subject = some x

                  The causer of CAUSE is the subject whatever it brings about.

                  theorem HartshorneEtAl2016.Sem.subject_be {α : Type} (x : α) (s : Sem α) :
                  ((var x).be s).subject = some x

                  The holder of BE is the subject whatever state it is in.

                  Fig. 11a, the habitual attitude: the experiencer is in the emotional state the root names, directed at the target.

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

                    Fig. 11b, the caused emotional episode: the stimulus causes the experiencer to be in the emotional state.

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

                      Fear-type verbs map the experiencer onto the subject.

                      Frighten-type verbs map the stimulus onto the subject.

                      The episode encodes the stimulus as a cause, the attitude encodes no cause: the structural content behind the causation judgments of their Experiments 2–4.

                      §5.4.3: the structures admit a target in the episode, still headed by the stimulus, which predicts the unattested (7).

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

                        Agreement with the classifications the paper builds on #