Anticipating words during spoken discourse comprehension: A large-scale, pre-registered replication study using brain...

Area of Science:

• Psycholinguistics
• Cognitive Neuroscience
• Computational Linguistics

Background:

• Studies suggest the brain anticipates upcoming words during language processing, evidenced by neural responses to mismatches.
• However, replicability issues and data-contingent analyses raise questions about the reliability of these findings.
• The original Van Berkum et al. (2005) study demonstrated an early positivity effect related to grammatical gender prediction.

Purpose of the Study:

• To conduct a close replication of the Van Berkum et al. (2005) ERP study on word anticipation.
• To investigate whether Dutch listeners reliably use adjectival inflection for noun prediction.
• To address concerns about data-contingent analyses and small sample sizes in psycholinguistic research.

Main Methods:

• A pre-registered close replication of Van Berkum et al. (2005), Experiment 1, using Dutch spoken mini-stories.
• Employed a grammatical gender mismatch manipulation with pre-nominal adjectives and highly predictable nouns.
• Utilized a larger dataset (N=187) and Bayesian mixed-effects models for analysis.

Main Results:

• A small, N400-like effect (more negative voltage for mismatch) was observed, but lacked strong statistical support (Bayes Factors).
• The specific early positivity effect reported by Van Berkum et al. (2005) was not replicated.
• Effects related to noun processing from the original study were successfully replicated.

Conclusions:

• The findings do not support the original Van Berkum et al. prediction effect, suggesting potential issues with data-contingent analyses and small sample sizes.
• The results question the consistency and reliability with which Dutch listeners use adjectival inflection for noun prediction.
• This replication underscores the importance of robust methodologies and larger sample sizes in psycholinguistic research.