Cognitive Mechanisms of (dis)fluency

Difficulties can arise at several stages of speech production, from conceptual preparation and lexical selection to phonological encoding and articulation. Through controlled connected-speech paradigms, I have examined whether difficulties at these different levels produce distinguishable patterns of disfluency. This work combines speech measures with eye movements, which provide complementary information about the timing of planning and the speaker’s allocation of visual attention. The findings show that disfluencies are informative about processing difficulty.

Where do disfluencies come from?

Difficulties can arise at several stages of speech production, from conceptual preparation and lexical selection to phonological encoding and articulation. Through controlled connected-speech paradigms, I have examined whether difficulties at these different levels produce distinguishable patterns of disfluency. This work combines speech measures with eye movements, which provide complementary information about the timing of planning and the speaker’s allocation of visual attention. The findings show that disfluencies are informative about processing difficulty.

Selected publications

Pistono, A., & Hartsuiker, R. J. (2023). Object identification difficulty can be predicted based on disfluencies and eye movements in connected speech. PLOS ONE, 18(3), e0281589.
Rapoeye, K., Hartsuiker, R. J., & Pistono, A. (2023). Semantic interference affects speech production by increasing disfluencies, not errors. Royal Society Open Science, 10(6), 230006.
Pistono, A., & Hartsuiker, R. J. (2022). Word-form-related disfluency versus lemma-related disfluency: An exploratory analysis of disfluency patterns in connected-speech production. In Proceedings of DiSS 2021.
Pistono, A., & Hartsuiker, R. J. (2021). Eye movements can help disentangle mechanisms underlying disfluency. Language, Cognition and Neuroscience, 36(8), 1038–1055.

A) Example of a network used in Experiment 1. Each circle represents a fixation. Lexical selection difficulty (low name agreement) is shown in red. B) Classification accuracy for each participant when predicting item name agreement from eye movements or disfluencies. The dashed line represents chance level. Each dot represents classification accuracy for a single participant.

Monitoring and Adaptive Control

Disfluencies are not a direct readout of processing difficulty: they also depend on how speakers detect and respond to that difficulty. My research examines how monitoring and control mechanisms regulate response outcomes. Using experimental manipulations and computational modelling, I investigate how competition between lexical candidates translates into different speech outcomes. I also study ADHD as a model of altered cognitive control, asking how differences in attention and inhibition affect speakers’ ability to regulate production difficulties during connected speech

Selected publications

Pistono, A., Senoussi, M., & Hartsuiker, R. J. (2025). Disfluencies reflect a… uh… competition between response options: Evidence from a drift-diffusion analysis. Psychonomic Bulletin & Review.
Rapoeye, K., Hartsuiker, R. J., & Pistono, A. (2023). Semantic interference affects speech production by increasing disfluencies, not errors. Royal Society Open Science, 10(6), 230006.
Bijnens, A., & Pistono, A. (2025). Disfluency production in children with attention-deficit/hyperactivity disorder during a narrative task. Speech Communication, 171, 103244.

Application of the drift diffusion model to language production. By jointly modelling naming latencies and accuracy, the model disentangles lexical evidence accumulation from decision thresholds and non-decision processes. In our study, semantic interference reduced the drift rate, indicating less efficient lexical selection. Disfluent responses were likewise associated with a lower drift rate and longer non-decision time, but not with changes in the response threshold. This suggests that disfluencies reflected difficulties in lexical evidence accumulation and in processes occurring before or after selection, rather than a more cautious response strategy.

On-going projects on monitoring and adaptive control in language production

Inner Speech

Another strand of my current research concerns inner speech (i.e., the “little voice” in our heads) and its contribution to cognitive and metacognitive regulation. This work reverses the perspective adopted in much of my previous research: rather than examining how non-linguistic cognitive functions influence speech, it asks how language itself can be used as a cognitive tool. An ongoing study with younger and older adults examines whether inner speech supports the resolution of tip-of-the-tongue states by helping speakers maintain, reorganize, or strengthen partially activated information, and whether its contribution changes with age.

Computational Modelling

Bram De keersmaecker’s doctoral research develops and tests MONET (MONitoring for Error and Time; Pistono & Hartsuiker, in preparation), a framework in which disfluencies emerge from the interaction between speech planning, execution, and monitoring. MONET proposes that activation conflict, interruption timing, and response thresholds jointly determine whether speakers pause, repeat, repair, or produce an error.

ADHD

In collaboration with Prof. Mélanie Planton, I am extending this research to adults with ADHD. The ongoing study examines fluency and coherence in narrative and autobiographical discourse. It aims to determine whether discourse characteristics reflect the linguistic consequences of attentional and executive-control difficulties.

Google Sites

Report abuse