The (un)reliability of item-level semantic priming effects

Area of Science:

• Cognitive Psychology
• Psycholinguistics
• Computational Linguistics

Background:

• Predicting semantic priming effects is crucial for understanding word association.
• Existing methods often rely on variables like associative strength and co-occurrence frequency.
• Reliability of item-level priming effects is a prerequisite for meaningful predictions.

Purpose of the Study:

• To assess the reliability of item-level semantic priming effects.
• To investigate the impact of strategic processing on priming reliability.
• To re-evaluate reliability in previously published semantic priming datasets.

Main Methods:

• Examined split-half and test-retest reliabilities of item-level priming effects.
• Controlled experimental conditions to minimize strategic influences.
• Reanalyzed data from three published semantic priming studies.

Main Results:

• Item-level priming effects demonstrated extremely low reliability.
• Reanalyses confirmed low reliability across multiple datasets.
• One large-scale study showed improved reliability, suggesting a data-size effect.

Conclusions:

• Previous attempts to predict item-level semantic priming are unlikely to succeed due to low reliability.
• Semantic priming research requires larger sample sizes (both items and participants) for robust findings.
• Future research should focus on "big data" approaches to enhance reliability and predictive power.