This study analyzes approximately 4,000 hours of parliamentary speech across Croatian, Czech, Polish, and Serbian to investigate filled pause usage using transformer-based automatic detection.

  • The research models filled pause rates using Generalised Estimating Equations with Mundlak correction to distinguish within- from between-speaker effects.
  • It replicates the negative association of age and speech rate with filled pause occurrence.
  • Gender effects are found to be language-specific, contradicting prior literature.
  • Sentiment shows a consistent positive association with filled pause rates, while political orientation and power status modulate this effect within specific parliaments.

The findings highlight the importance of large-scale, multi-language corpora for generalizing insights into spontaneous speech features.