Related Experiment Video
Updated: Aug 19, 2025
08:12
A Psychophysics Paradigm for the Collection and Analysis of Similarity Judgments
Published on: March 1, 2022
2.6K
A comparative investigation of integral- and separable-dimension stimulus-sorting behavior
Charles A Doan1,2, Ronaldo Vigo3,4
1Department of Psychology, Marietta College, 406 Mills Hall, Marietta, OH, 45750, USA. cad007@marietta.edu.
Psychological Research
|November 26, 2022
Summary
This study reveals that using integral stimuli in unsupervised sorting tasks reduces rule-based strategies (SD, XOR) while increasing family resemblance (FR) strategies. Generalized representational information theory (GRIT) better explains these sorting behaviors than the simplicity model.
Area of Science:
- Cognitive Psychology
- Computational Neuroscience
- Machine Learning
Background:
- Unsupervised categorization research traditionally focuses on a few key strategies: one-dimensional (SD), family resemblance (FR), and exclusive-or (XOR).
- Previous studies utilized multidimensional stimuli contrasting these strategies, manipulating tasks and stimulus properties to observe their impact on sorting behavior.
Purpose of the Study:
- To investigate how integral-dimension stimuli affect sorting strategies within constrained unsupervised categorization tasks.
- To compare the prevalence of observed strategies with those found using separable stimuli in similar tasks.
- To evaluate the explanatory power of generalized representational information theory (GRIT) and the simplicity model for unsupervised sorting behavior.
Main Methods:
- Participants sorted integral-dimension stimuli across two novel constrained unsupervised categorization tasks.
- The tasks employed three-dimensional Boolean category structures, encompassing six distinct sorting strategies (including SD, FR, and XOR).
- Strategy prevalence was analyzed and compared to prior studies using separable stimuli and the same underlying structures.
Main Results:
- A significant decrease in rule-like sorting behavior (SD or XOR) was observed with integral stimuli across various structures.
- This reduction in SD and XOR behavior correlated with an increase in FR and FR-related sorting behaviors.
- The GRIT model demonstrated superior performance in accounting for the observed sorting strategies compared to the simplicity model across both separable and integral paradigms.
Conclusions:
- Integral stimuli shift unsupervised sorting behavior away from rule-based strategies towards family resemblance.
- The findings highlight the influence of stimulus dimensionality on categorization strategy selection.
- Generalized representational information theory provides a robust framework for understanding unsupervised sorting across different stimulus types and tasks.
