6/ Yet, they also highlight the importance of viewing communication as a continuum rather than a binary distinction.

5/ Our results align with prior work (Finlayson & Corley, 2012), reinforcing the idea that filled pauses are largely a byproduct of language production difficulties rather than deliberate communicative signals.

4/ We also examined whether autistic traits (AQ score) or stress levels were related to the proportion of filled pauses, but found no significant correlations.

3/ Our results showed that participants produced significantly more words per tangram when speaking to an interlocutor, suggesting they adapted their communication. However, filled pauses were not part of this adjustment.

2/ We compared self-directed and social speech: Participants described tangrams (thinking they were doing a memory test). If filled pauses are used as a communicative signal, we would expect more of them in social speech.

5/ These findings contribute to the development of computational models of disfluency and point to future research using neural networks to refine our understanding of competition and accumulation mechanisms in language production

4/ Key findings:
Disfluent answers had lower drift rates, meaning they reflect competition between response options, not a stalling strategy.
Despite time pressure, we found individual differences in how participants handled semantic interference, suggesting variation in speed–accuracy trade-offs.

3/ The use of a DDM approach allowed to determine the underlying nature of disfluencies: related to lexical-semantic processes (drift rate), postlexical processes (non-decision time) or related to speakers’ adaptation to task demand (decision threshold).

2/ DDM helps break down response processes into key components:
⚡ Drift rate = how fast information accumulates
🎯 Decision threshold = amount of evidence required/strategies of decision making
⏳ Non-decision time = time spent on other processes (e.g., motor prep)