Background. Disfluencies (filled pauses, repetitions, self-repairs) provide a real-time window on speech planning and monitoring. Objective. To compare how speech rate (fast vs. slow) relates to disfluency types, positional distribution, and production stages in spontaneous English and Uzbek speech.
Methods. We analyzed ≈100 minutes per language from televised interviews (The Graham Norton Show; Darakchi.uz), ≈10 speakers per language (≥18 y.o.). Disfluencies were coded by type (filled pause, repetition, self-repair), position (initial/medial/final), and stage (conceptual planning, formulation/encoding, articulation, self-monitoring). A pause threshold of ≥200 ms was used; speech rate buckets (fast/slow) were assigned from observed WPM.
Results. English: 78 filled pauses, 36 repetitions, 3 self-repairs. Uzbek: 151 filled pauses, 13 repetitions, 7 self-repairs. Fast speakers produced more errors but fewer fillers; slow speakers produced fewer errors but more fillers. Disfluencies clustered utterance-medially in both languages. Gender patterns showed small asymmetries (e.g., more exchanges/perseverations among Uzbek males).
Conclusion. Disfluency profiles co-vary with speech rate and sociolinguistic style, reflecting universal psycholinguistic mechanisms modulated by community-specific communicative practices.