Bias and noise in proportion estimation: A mixture psychophysical model.

Area of Science:

• Cognitive Psychology
• Psychophysics
• Developmental Psychology

Background:

• Proportional reasoning is crucial in psychology and education, but mental representations of proportions remain poorly understood.
• Existing models for proportion estimation have limitations in explaining observed behaviors and individual differences.

Purpose of the Study:

• To develop and validate a novel psychophysical model for proportion estimation.
• To investigate the internal representation of proportions, including bias and noise parameters.
• To account for contaminating behaviors like guessing and response reversals in estimation tasks.

Main Methods:

• Proposed a model where proportion stimuli (part and complement) are represented as Gaussian activations.
• Combined these representations to derive internal proportion representations, incorporating bias and noise.
• Utilized a mixture of components and a hierarchical framework to model guessing and response reversals.
• Empirically tested the model with 4th-grade children in a spatial proportion estimation task.

Main Results:

• The model's internal density successfully replicated estimation asymmetries (skewedness) observed in spatial tasks.
• The model accurately described significant inter-subject variability in behavioral data.
• Bias estimates were reduced compared to previous methods due to the model's handling of contaminating behaviors.
• Higher internal noise levels were found compared to spatial magnitude discrimination, with potential explanations discussed.

Conclusions:

• The developed psychophysical model provides a robust framework for understanding proportion estimation.
• The model's ability to absorb contaminating behaviors enhances its relevance for linking psychophysical measures with cognitive abilities.
• Quantitative models are valuable for studying scaling effects in proportional reasoning.