Documentation

Linglib.Studies.Kriz2016

Križ (2016): homogeneity, non-maximality, and all #

This file verifies [Kri16]'s predictions against a finite model, using the homogeneity substrate in Semantics.Homogeneity and its plural instantiation (barePlural, allPlural — both originating with this paper). A five-world model checks the predictions end-to-end, including the §4.2 sensitivity to what an exception does instead and the §4.1 unmentionability of exceptions. Closing sections connect the analysis to the typed data in Data.Examples.Kriz2015 and to [Mag14]'s rival gap derivation.

Implementation notes #

QUD W is the substrate partition type, not the [Rob96] QUD-stack: §4.5 of the paper argues (examples (39)-(40)) that the current issue is an overarching, not directly manipulable, property of the discourse, so coarseQ/fineQ below are pedagogical constructions. Following §4.4, the gap is trivalent but not presuppositional (contra [Gaj05]). The §4.6 numeral puzzle (the ten professors smiled resists non-maximality) is left unaddressed, as in the paper.

References #

Finite model #

Three professors attend Sue's talk; the predicate is "smiled".

WorldSmithJonesLeeBare pluralAll
allSmiledTRUEtrue
smithNeutralGAPfalse
smithAngryGAPfalse
onlyLeeSmiledGAPfalse
noneSmiledFALSEfalse

smithNeutral and smithAngry agree on who smiled but differ in what Smith does instead, which the coarse QUD ("Was the talk well-received?") is sensitive to (§4.2); the fine QUD ("Did every professor smile?") separates all worlds.

Worlds of the five-world model. In both smithNeutral and smithAngry Smith fails to smile; the worlds differ in whether his demeanour is relevant to the coarse issue (§4.2).

Instances For
    @[instance_reducible]
    Equations
    @[instance_reducible]
    Equations
    def Kriz2016.instReprProfWorld.repr :
    ProfWorldStd.Format
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      inductive Kriz2016.Prof :
      Instances For
        @[instance_reducible]
        instance Kriz2016.instDecidableEqProf :
        DecidableEq Prof
        Equations
        @[instance_reducible]
        Equations
        def Kriz2016.instReprProf.repr :
        ProfStd.Format
        Equations
        Instances For
          @[instance_reducible]
          instance Kriz2016.instFintypeProf :
          Fintype Prof
          Equations
          @[instance_reducible]
          instance Kriz2016.smiled.instDecidable (p : Prof) (w : ProfWorld) :
          Decidable (smiled p w)
          Equations
          • One or more equations did not get rendered due to their size.
          def Kriz2016.profs :
          Finset Prof

          All three professors.

          Equations
          Instances For

            Reception grades for the coarse QUD: Smith's anger pulls reception down to mixed, his neutrality leaves it positive (§4.2).

            Instances For
              @[instance_reducible]
              Equations

              Coarse QUD: "Was Sue's talk well-received?"

              Equations
              Instances For

                Fine QUD: "Did every professor smile?"

                Equations
                Instances For

                  Trivalent values at each world #

                  End-to-end predictions #

                  Unmentionability of exceptions (§4.1) #

                  "#The professors smiled, but one of them didn't" is infelicitous even where the non-maximal reading is licensed — the paper's (25a-b), with the although diagnostic traced to [Kro74] via [Las99]. The derivation is pure Addressing (the substrate's exception_unaddressable): non-maximal use requires a cell containing both a true-world and the gap-world, and the exception-mentioning continuation straddles it.

                  "Smith didn't smile", the exception-mentioning continuation.

                  Equations
                  Instances For

                    Under the coarse issue licensing the non-maximal use at smithNeutral, "…but Smith didn't" cannot address the issue (§4.1).

                    What exceptions do (§4.2) #

                    Whether an exception is tolerated depends on what it does instead: Smith looking neutral is irrelevant to reception, Smith looking angry is not. The model places smithNeutral and smithAngry in different coarseQ cells, so the same sentence under the same QUD is usable at one gap-world and not the other — a contrast unavailable to accounts without an issue parameter (restricted reference, alternative geometry).

                    The bare plural is not usable at smithAngry under the coarse QUD: smithAngry shares its cell with onlyLeeSmiled, and neither is in the positive extension.

                    The typed switches data #

                    The switches items of Data.Examples.Kriz2015 show the model's pattern in the wild: "Oh no, the switches are on!" is acceptable under the existential issue and unacceptable under the universal one (cf. smithNeutral_usable_coarse vs smithNeutral_not_usable_fine), and the all variant is unacceptable even in the permissive context (cf. all_not_usable_smithNeutral). The gap rows lift to .indet observations in the pooled Generalizations.HomogeneityGap data.

                    The model matches the data: the bare plural's value at the gap-world is the value the positive gap row observes.

                    Conjunction overgeneration (§6.2) #

                    Conjunctions of proper names are homogeneous (Szabolcsi & Haddican 2004, [Mag14]) yet generally resist non-maximal readings. Modelled as a plural over its conjunct atoms, the machinery predicts non-maximal use at a gap-world (conj_modeled_as_plural_predicts_nonmax); the paper's informal response is that mentioning an individual prompts accommodation of a finer issue on which no non-maximal reading survives, an accommodation step not formalized here or in the paper.

                    Instances For
                      @[instance_reducible]
                      Equations
                      @[instance_reducible]
                      Equations
                      def Kriz2016.instReprConjAtom.repr :
                      ConjAtomStd.Format
                      Equations
                      Instances For
                        Instances For
                          @[instance_reducible]
                          Equations
                          @[instance_reducible]
                          Equations
                          def Kriz2016.instReprConjWorld.repr :
                          ConjWorldStd.Format
                          Equations
                          Instances For
                            @[instance_reducible]
                            instance Kriz2016.wentThere.instDecidable (a : ConjAtom) (w : ConjWorld) :
                            Decidable (wentThere a w)
                            Equations
                            • One or more equations did not get rendered due to their size.
                            Equations
                            Instances For

                              Cells of the coarse "did anyone go?" issue.

                              Instances For
                                @[instance_reducible]
                                Equations

                                Modelled as a plural over {Bert, Claire, Dora}, the conjunction is predicted usable at the gap-world dorasMissing under the coarse issue — the overgenerated non-maximal reading.

                                Križ vs Magri on the gap's value #

                                [Mag14] derives homogeneity from double exhaustification over alternative geometry: on a gap scenario, doubleExh .mystery is bivalent-false (Magri2014.gap_positive_false) — the gap collapses inside the semantics. For Križ the same input is .indet, and the gap is pragmatically recoverable under a coarse issue (gap_enables_nonmax). A false sentence is unutterable on standard Gricean terms, so Magri's account needs additional pragmatic machinery to license the non-maximal uses the finite model exhibits. Magri2014.fromPredicate translates the model's (smiled, profs, smithNeutral) into Magri's count abstraction, so both operators run on the same input.

                                A 3-atom Magri scenario where 2 of 3 atoms satisfy the predicate.

                                Equations
                                Instances For

                                  Counting satisfiers of smiled at smithNeutral yields the 2-of-3 scenario: the divergence below is a same-input comparison.

                                  Magri's doubleExh .mystery is false on a 2-of-3 gap.

                                  On the same input, Magri's operator returns false while Križ's returns .indet and the sentence is usable under the coarse issue: the two accounts assign the gap incompatible statuses.