A Bayesian Approach to German Personal and Demonstrative Pronouns

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Cognitive Science

Background:

• Pronoun interpretation relies on identifying referents, often modeled using Bayes' Rule.
• Previous research successfully applied Bayesian models to English and Mandarin Chinese.

Purpose of the Study:

• To extend cross-linguistic evidence for the Bayesian pronoun interpretation model.
• To apply and quantitatively assess the model using German personal and demonstrative pronouns.
• To provide conservative uncertainty estimates via a hierarchical Bayesian framework.

Main Methods:

• Two story-continuation experiments were conducted with German speakers.
• Model performance was evaluated within a hierarchical Bayesian statistical framework.
• Predictions were compared against production and next-mention biases.

Main Results:

• The Bayesian model demonstrated superior accuracy in predicting pronoun interpretation compared to alternative biases.
• The model successfully accounted for both personal and demonstrative pronoun resolution in German.
• Demonstrative pronoun resolution preferences, though rigid, were captured by the model.

Conclusions:

• The Bayesian framework provides a robust and accurate account of pronoun interpretation across languages.
• The study strengthens the cross-linguistic validity of the Bayesian approach to psycholinguistics.
• Hierarchical Bayesian modeling offers a conservative method for estimating uncertainty in cognitive models.