Documentation

Linglib.Studies.TieuEtAl2020

Tieu, Bill, Romoli & Crain (2020) [TBRC20] #

Testing theories of plural meanings. Cognition 205, 104307.

Three experiments comparing adults' and children's interpretations of bare plurals in upward- and downward-entailing environments. The results support an implicature approach to multiplicity inferences: children compute fewer multiplicity inferences than adults, in parallel with their behavior on standard scalar implicatures, and the two inference types are correlated within children.

Core Argument #

The paper adjudicates between three theories of why "Emily fed giraffes" means "more than one":

  1. Ambiguity ([FdS10]): plural is polysemous (inclusive/exclusive), Strongest Meaning Hypothesis selects the stronger reading.
  2. Implicature ([Sau03], [Spe07], [Zwe09]): plural literally means "one or more," the "more than one" inference is a scalar implicature with the singular as alternative.
  3. Homogeneity ([Kri15b]): multiplicity arises from homogeneity presupposition.

Key discriminating prediction (Uniformity Prediction): if multiplicity inferences are scalar implicatures, children should compute fewer of both, and rates should be correlated.

Main declarations #

The paper's three-way taxonomy of theories of the multiplicity inference: ambiguity ([FdS10]), implicature ([Sau03], [Spe07], [Zwe09]), homogeneity ([Kri15b]).

  • ambiguity : PluralTheory

    Plural is ambiguous; Strongest Meaning Hypothesis resolves.

  • implicature : PluralTheory

    Plural literally means "one or more"; multiplicity is implicature.

  • homogeneity : PluralTheory

    Plural interpretation via homogeneity presupposition.

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

      Does the theory analyze multiplicity as arising via the same mechanism as scalar implicatures? The paper's discriminating predictions — acquisition delay, within-child SI correlation, polarity asymmetry, singular-context truth-value asymmetry — all transfer from known SI properties via this property.

      Equations
      Instances For

        The implicature theory is uniquely identified by the SI mechanism, hence by any of the predictions that transfer through it.

        Look up a stimulus row in the paper's example pool by id.

        Equations
        Instances For
          @[reducible, inline]

          Core multiplicity datum, positive form: "Emily fed giraffes".

          Equations
          Instances For
            @[reducible, inline]

            Core multiplicity datum, negative form: "Emily didn't feed giraffes".

            Equations
            Instances For
              @[reducible, inline]

              Experiment 1, upward-entailing TVJ trial: "Emily fed pigs" after a story in which she fed exactly one pig (rejection indicates computing multiplicity).

              Equations
              Instances For
                @[reducible, inline]

                Experiment 1, downward-entailing TVJ trial: "Emily didn't feed giraffes" after she fed exactly one (acceptance indicates a local multiplicity reading under negation).

                Equations
                Instances For
                  @[reducible, inline]

                  Experiment 3, positive plural in a singular context: "Koala bought pears" when Koala bought exactly one pear.

                  Equations
                  Instances For
                    @[reducible, inline]

                    Experiment 3, negative plural in a singular context: "Koala didn't buy pears" when Koala bought exactly one pear.

                    Equations
                    Instances For
                      theorem TieuEtAl2020.consistent_with_monotonicity_data :
                      (fedGiraffesPos.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "multiplicity (>1)" x.readings) = some Features.Judgment.acceptable (fedGiraffesNeg.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "multiplicity (>1)" x.readings) = some Features.Judgment.unacceptable

                      The core monotonicity pattern: the multiplicity reading is available in the positive datum but unavailable under negation.

                      theorem TieuEtAl2020.exp1_multiplicity_reading_contrast :
                      (exp1_positive.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "multiplicity (>1)" x.readings) = some Features.Judgment.acceptable (exp1_negative.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "multiplicity (>1) local under negation" x.readings) = some Features.Judgment.marginal

                      Experiment 1 reading availability: the multiplicity reading is fully available in the UE trial; the local multiplicity reading under negation is only marginal (adults' negative-condition rate was moderate).

                      Inference rate for a group in a condition.

                      • group : String

                        Which group

                      • inferenceType : String

                        Inference type

                      • Polarity of context

                      • rate : String

                        Rate of inference-consistent responses (qualitative)

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

                          Experiment 1 key results (qualitative — no exact numbers cited).

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

                            Experiment 2 key results (qualitative).

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

                              The Uniformity Prediction: if multiplicity inferences are scalar implicatures, then the between-group pattern (children < adults) should be the same for both inference types.

                              The paper confirms this prediction and additionally finds that children's rates on the two types are significantly correlated.

                              • childrenFewerMultiplicity : Bool

                                Do children compute fewer multiplicity inferences than adults?

                              • childrenFewerSI : Bool

                                Do children compute fewer scalar implicatures than adults?

                              • correlatedInChildren : Bool

                                Are the two rates correlated within children?

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

                                    All three components of the Uniformity Prediction are confirmed.

                                    The singular/plural scale predicts multiplicity as a scalar implicature: using the plural (weaker) implicates the negation of the singular (stronger).

                                    In DE contexts, the scale reverses (weaker alternatives are relevant), so the multiplicity inference does not arise.

                                    Experiment 3 uses a ternary judgment task (small/medium/large strawberry for false/neither/true) with adults on Amazon Mechanical Turk.

                                    In singular contexts (exactly one object acted upon), only the SI mechanism predicts a positive/negative truth-value asymmetry: "Koala bought pears" is literally true with a false implicature (misleading), while "Koala didn't buy pears" is literally false. On lexical or presuppositional accounts both have the same status. Together with implicature_uniquely_supported, the observed asymmetry singles out PluralTheory.implicature.

                                    theorem TieuEtAl2020.exp3_confirms_asymmetry :
                                    (exp3_positive_plural.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "preferred_reward" x.paperFeatures) = some "intermediate" (exp3_negative_plural.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "preferred_reward" x.paperFeatures) = some "minimal"

                                    Adults assign different reward status to positive vs negative plurals in singular contexts: intermediate reward for the (literally true but misleading) positive, minimal for the (literally false) negative.

                                    theorem TieuEtAl2020.multiplicity_parallels_si_de_blocking :
                                    (fedGiraffesPos.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "multiplicity (>1)" x.readings) = some Features.Judgment.acceptable (fedGiraffesNeg.bind fun (x : Data.Examples.LinguisticExample) => List.lookup "multiplicity (>1)" x.readings) = some Features.Judgment.unacceptable

                                    The multiplicity inference exhibits DE blocking: available in the positive fed-giraffes row, unavailable under negation — the same UE-arises / DE-blocked pattern as the classical some/all scalar implicature ([Hor72]).