Measuring binding effects in event-based episodic representations.

Area of Science:

• Cognitive Psychology
• Psychometrics
• Computational Neuroscience

Background:

• Coherent memory recall relies on binding event elements, creating stochastic dependencies during retrieval.
• Existing models for these dependencies include contingency-based, Yule's Q, and item response theory (IRT) approaches.

Purpose of the Study:

• To compare the empirical detection rates and performance of five different modeling approaches for memory binding effects.
• To evaluate the susceptibility of these models to varying levels of memory performance.

Main Methods:

• A simulation study was conducted to compare five distinct modeling approaches: contingency-based, Yule's Q, adjusted Yule's Q, IRT, and a nonparametric IRT variant.
• An empirical application was also performed to validate findings from the simulation.

Main Results:

• The IRT-based approach and its nonparametric variant demonstrated the highest power in detecting dependencies and differences in dependency.
• The nonparametric IRT variant showed increased Type I error rates with higher data dependency when testing for differences.
• Yule's Q approaches yielded biased estimates and were highly sensitive to memory performance, while other measures showed less sensitivity.

Conclusions:

• The IRT-based approach is recommended as the most suitable method for measuring binding effects in memory.
• Considerations for selecting a modeling approach, including performance under varying memory conditions, are discussed.