Documentation

Linglib.Phenomena.Imprecision.Studies.BeltramaSchwarz2024

@cite{beltrama-schwarz-2024}: Empirical Data #

@cite{beltrama-schwarz-2024}

Experimental data from @cite{beltrama-schwarz-2024}, showing that speaker persona stereotypes affect imprecision resolution for round numerals.

Experiments #

  1. Experiment 1 (Covered Screen Task): Participants choose which of two phone screens (COVERED vs VISIBLE) a speaker was looking at when making a round-numeral statement. Both Nerdy and Chill personae significantly affect choice rates.

  2. Experiment 2 (Truth-Value Judgment Task): Participants judge whether a speaker's round-numeral statement is RIGHT or WRONG given a visible phone screen. Only the Chill persona significantly affects judgments.

Key empirical generalization #

Speaker persona stereotypes modulate imprecision resolution bi-directionally in inference tasks (Exp 1) but asymmetrically in judgment tasks (Exp 2): Chill speakers increase tolerance for imprecision in both tasks, but Nerdy speakers only increase exactness demands in the inference task.

Speaker persona condition (between-subjects). Participants read a short vignette describing the speaker (§3.1).

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

      Screen fit condition (within-subjects, §3.1). Determines the relationship between the speaker's round-numeral statement and the amount shown on the VISIBLE phone screen.

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

          Experimental task type.

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

              The trait descriptors used for each persona condition. Each participant read: "Ryan is [descriptors]."

              Equations
              Instances For

                A coefficient from a mixed-effects logistic regression model. Models fitted via glmer in R (lme4; @cite{bates-etal-2015}).

                • predictor : String
                • beta :

                  Coefficient estimate (log-odds)

                • se :

                  Standard error

                • significant : Bool

                  Significant at p < 0.05?

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

                    Experiment 1 sample size after exclusions (§4.1). 134 participants from 151 recruited.

                    Equations
                    Instances For

                      Experiment 1, Model 1 (main effects only). DV: COVERED (1) vs VISIBLE (0). Reference levels: NoPersona, ScreenFit[Imprecise]. Table 1, Model 1 (§4.2).

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

                                Experiment 1, Model 2 interaction terms (Table 1, Model 2). All Persona × ScreenFit interactions are non-significant (§4.2.2): persona effects are comparable across screen fit conditions.

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

                                    Experiment 2 sample size after exclusions (§5.1). 132 participants from 150 recruited.

                                    Equations
                                    Instances For

                                      Experiment 2, Model 1 (main effects only). DV: RIGHT (1) vs WRONG (0). Reference levels: NoPersona, ScreenFit[Imprecise]. Table 2, Model 1 (§5.2).

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

                                                All critical items use round dollar amounts. Example from §3.1: "The phone costs $200."

                                                Equations
                                                Instances For

                                                  The imprecise screen shows a close-but-not-exact amount. Example from §3.1: visible screen shows $193 when statement says $200.

                                                  Equations
                                                  Instances For

                                                    Core finding: both persona conditions significantly affect Exp 1 (inference task).

                                                    Key asymmetry: in Exp 2 (judgment task), only Chill is significant.

                                                    Chill decreases COVERED rate (negative β): Chill persona leads listeners to favor the VISIBLE screen, treating the round numeral as imprecise (§4.2.2).

                                                    Nerdy increases COVERED rate (positive β): Nerdy persona leads listeners to favor the COVERED screen, treating the round numeral as exact (§4.2.2).

                                                    Chill increases RIGHT rate (positive β on RIGHT DV), i.e., increases tolerance for imprecision in truth-value judgments (§5.2.2).

                                                    Nerdy has no significant effect on truth-value judgments (§5.2.2).

                                                    No significant Persona × ScreenFit interactions in Exp 1 (§4.2.2): persona effects are comparable across all screen fit conditions.

                                                    This absence (while main effects ARE significant) implies that persona modulates a global interpretive parameter — the precision threshold — rather than a local disambiguation strategy that operates only in ambiguous contexts.

                                                    The Chill effect is larger in Exp 2 (|β| = 1.04) than in Exp 1 (|β| = 0.67).

                                                    Precision variants for numeral use.

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

                                                        The indexical field for numeral precision.

                                                        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

                                                                Whether imprecise readings are available for a numeral.

                                                                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.

                                                                  Multiples of 10 always have imprecise readings.