Accounting for item-level variance in recognition memory: Comparing word frequency and contextual diversity

Area of Science:

• Cognitive Psychology
• Psycholinguistics
• Computational Linguistics

Background:

• Word frequency is a traditional predictor of lexical processing.
• Contextual diversity and semantic diversity offer more nuanced measures of word strength.
• Previous research shows these diversity measures improve lexical organization models.

Purpose of the Study:

• To test if contextual and semantic diversity generalize to word-level episodic recognition.
• To compare the predictive power of diversity measures against word frequency for recognition data.
• To explore the utility of large-scale behavioral datasets for developing new memory models.

Main Methods:

• Reanalyzed existing episodic recognition data (Cortese et al., 2010, 2015).
• Applied models incorporating contextual diversity and word frequency.
• Assessed the variance accounted for by different predictive variables.

Main Results:

• Contextual diversity significantly improved predictions of episodic recognition rates.
• The best contextual diversity model explained substantially more variance than word frequency alone.
• Findings align with previous research on lexical organization.

Conclusions:

• Contextual and semantic diversity are crucial for understanding word-level episodic recognition.
• Large-scale behavioral data facilitate the development of theoretically grounded memory models.
• Diversity measures offer a more robust framework for predicting word recognition performance.