Documentation

Linglib.Phenomena.Numerals.Studies.HuangSpelkeSnedeker2013

@cite{huang-snedeker-2013} — Covered-Box Paradigm #

@cite{huang-snedeker-2013} @cite{musolino-2004} @cite{wynn-1992}

"What Exactly do Numbers Mean?" Language Learning and Development 9(2): 105-129

Design #

The covered-box task cancels scalar implicatures by embedding the target term in a definite description ("the box with two fish") and providing three options: a visible mismatch, a visible lower-bounded match, and a covered (hidden) box. When no exact match is visible, choosing the covered box implies the participant requires an exact match (rejecting the LB-compatible visible option).

The task is validated by a scalar control: "some" in the same paradigm yields lower-bounded readings (adults accept ALL as "some"), confirming that implicatures are successfully cancelled.

Key Empirical Findings #

Across 4 experiments with adults and two-knower children (ages 2;6–3;7):

  1. "some" control: lower-bounded when implicature cancelled — adults accept the total set (87% Exp 1, 60% Exp 4), children accept it (83–90%)
  2. "two" critical: exact under the same conditions — both adults (92–100%) and two-knowers (80–95%) reject the 3-fish box, choose covered
  3. Scalar–numeral divergence: "some" and "two" pattern differently in the same implicature-cancelling context; scalars go LB, numerals stay exact
  4. Two-knower specificity: children who know "one" and "two" but NOT "three" still give exact readings for "two"

Choice in the covered-box paradigm.

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

      A critical trial result.

      • experiment : Nat
      • population : String
      • term : String
      • visibleCounts : List Nat

        Cardinalities in the visible boxes

      • coveredPct : Nat

        Percentage selecting the covered box

      • lbPct : Nat

        Percentage selecting the lower-bounded visible match

      • note : String
      Instances For
        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
            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
                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
                    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

                        Knower level in the @cite{wynn-1992} acquisition sequence.

                        An empirical classification: children master numeral meanings one at a time. A two-knower gives 1 for "one", 2 for "two", and random handfuls for higher numbers in the Give-N task.

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

                            Two-knowers know "two" but not "three". This is the empirical fact that makes their exact "two" readings informative: they cannot have derived exactness via scalar implicature against "three".