High- and low-threshold models of the relationship between response time and confidence

Area of Science:

• Cognitive Psychology
• Psychology
• Neuroscience

Background:

• Recognition memory models are crucial for understanding how individuals retrieve information.
• Existing models differ in their assumptions about memory thresholds and retrieval processes.
• Accounting for both response times and confidence ratings provides a more rigorous test of model adequacy.

Purpose of the Study:

• To compare the explanatory power of high- and low-threshold discrete-state models of recognition memory.
• To evaluate model performance using confidence and response time (RT) data.
• To determine which model best accounts for the observed inverted-U pattern in RT-confidence relationships.

Main Methods:

• Comparison of three discrete-state models: 2-high threshold (2HT), 1-low threshold (1LT), and 2-low threshold (2LT).
• Analysis of response time (RT) and confidence rating data, focusing on the inverted-U pattern.
• Quantitative model fits to assess performance across different memory strength conditions.

Main Results:

• The 2-high threshold (2HT) model explained RT-confidence for correct responses but not errors.
• The 1-low threshold (1LT) model failed to explain the RT-confidence relationship for both correct responses and errors.
• Only the 2-low threshold (2LT) model successfully matched the full RT-confidence pattern for both correct and error responses.
• The 1LT model also performed poorly in accounting for confidence rating distributions across memory strength conditions.

Conclusions:

• The 2-low threshold (2LT) model provides a superior account of recognition memory data compared to 2HT and 1LT models.
• The ability to account for misleading detection is critical for models to explain the full RT-confidence relationship.
• Discrete-state models with two thresholds offer a more robust framework for understanding recognition memory.