Verifying verification strategies of quantifiers using Bayesian cognitive modeling and eye-tracking: A Registered Report

The quantifiers most and more than half, despite often being held to be truth-conditionally equivalent, display diverging representations and behaviors. This has led researchers to hypothesize differing ‘verification strategies’, whereby more than half is evaluated by a more precise, explicit, rule-based process than most. To date, however, the case for this has been built exclusively on behavioral results, offering limited insights into the potential mechanism. Our study will combine eye-tracking with a Bayesian hierarchical cognitive model (Sarafoglou et al. 2024), providing a psychophysiologically informed and computationally explicit investigation of verification strategies. Native Dutch speakers will evaluate 228 quantifier statements, including ‘de meeste’ (most) and ‘meer dan de helft’ (more than half), using dot-array stimuli varying in visual complexity (‘few’ vs ‘many dots’), to test the credibility of divergent verification strategies. We conducted a pilot study (n=9) to refine the experimental design. We predict increased vagueness (α), meaning threshold (β), and response noise (γ) for most in the Bayesian cognitive model, in line with previous behavioral findings; as well as larger saccades for most, and greater pupil constriction for more than half (especially under the ‘many dots’ condition), reflecting differences in visual search strategy and attentional breadth. Our results indicated that there was XXX evidence [supporting/against] the hypothesis that the three parameters were greater for most in the Bayesian cognitive model (BFrc = XXX). There was XXX evidence [supporting/against] the hypothesis that saccade amplitudes were larger for most (BF+c = XXX). Pupil constriction was increased for XXX, indicating XXX evidence [in favour of/against] our hypothesis (BF+c = XXX). Finally, there was XXX evidence [supporting/against] an interaction between dot count and quantifier on pupil size, such that the 'many dots' condition resulted in a greater decrease in pupil size for XXX compared to XXX (BF+c = XXX). These results [offer credibility to/offer mixed evidence for/challenge] the hypothesis that more than half is associated with a more precise, ‘counting to a criterion’ verification strategy.