Intensity generalization: physiology and modelling of a neglected topic.

Area of Science:

• Behavioral Science
• Computational Neuroscience
• Psychology

Background:

• Generalization of learned behavior to new stimuli is a key concept in understanding learning.
• Existing models often struggle to explain how variations in stimulus intensity affect this generalization.

Purpose of the Study:

• To review empirical data on behavioral generalization, focusing on intensity variations.
• To evaluate computational models of generalization for their ability to explain intensity-dependent effects.
• To propose necessary components for accurate generalization models.

Main Methods:

• Review of empirical data on stimulus generalization.
• Evaluation of computational models against behavioral data, particularly intensity generalization.
• Analysis of stimulus representation in relation to sensory organ function.

Main Results:

• Stimulus intensity affects behavior differently than other stimulus characteristics (e.g., shape, frequency).
• Models require realistic stimulus representations, informed by sensory organ physiology, to capture intensity effects.
• The intuitive similarity-based generalization model is insufficient for intensity generalization.

Conclusions:

• Accurate modeling of generalization necessitates incorporating how sense organs process stimulus intensity.
• Simple models combined with realistic sensory representations can explain diverse generalization patterns.
• New conceptual tools are introduced for evaluating and comparing behavioral models.