A model of exact small-number representation

Area of Science:

• Cognitive Science
• Neuroscience
• Computational Modeling

Background:

• Existing models of the mental number line struggle to explain the absence of size effects in tasks other than numerical comparison.
• Current assumptions like magnitude coding, compressed scaling, and increasing variability face challenges with empirical data, particularly regarding symmetries in priming studies.

Purpose of the Study:

• To propose and validate a novel neural network model of the mental number line that overcomes limitations of existing theories.
• To account for the size effect in numerical comparison while explaining its absence in other tasks like number naming and parity judgment.

Main Methods:

• Development of a neural network model employing place coding, linear scaling, and constant variability for the mental number line.
• Training the model on numerical comparison, number naming, and parity judgment tasks.
• Analyzing the model's performance regarding the size effect and symmetries in primed tasks.

Main Results:

• The proposed model successfully reproduces the size effect in numerical comparison.
• The model demonstrates the absence of the size effect in number naming and parity judgment tasks, aligning with empirical observations.
• The model exhibits no asymmetries in primed naming or parity judgment, consistent with experimental findings.

Conclusions:

• The new model provides a more comprehensive account of mental number line processing than previous assumptions.
• Place coding, linear scaling, and constant variability offer a viable alternative framework for understanding numerical cognition.
• The findings have implications for theories of numerical representation and cognitive modeling.