Planning when to say: Dissociating cue use in utterance initiation using cross-validation.

Area of Science:

• Psycholinguistics
• Cognitive Science
• Speech Communication

Background:

• Conversational turns typically follow each other with minimal gaps, requiring speakers to anticipate the end of their partner's utterance.
• Launching coordinated speech involves predicting partner utterance length and content to minimize conversational delays.

Purpose of the Study:

• To investigate the relative importance of auditory cues (utterance length) versus visual cues (stimulus corepresentation) in initiating coordinated speech.
• To determine how predictability of utterance length influences the timing of speech production in conversational turn-taking.

Main Methods:

• Three experiments with Dutch adult participants producing prepared utterances in response to recorded confederate prompts.
• Manipulation of visual access to the confederate's stimulus and predictability of the confederate prompt's length.
• Measurement of inter-turn gaps and participants' visual attention allocation using machine learning for model selection (k-fold cross-validation).

Main Results:

• Inter-turn gaps were primarily predicted by auditory cues from the confederate's speech signal.
• Visual access to the stimulus provided some benefit but was less influential than auditory cues.
• Predictability of prompt length also played a role in gap prediction.

Conclusions:

• In simple laboratory tasks, speakers prioritize auditory cues from a partner's speech signal over visual information about the content for timing coordinated utterances.
• This finding highlights the dominant role of auditory processing in managing conversational flow and turn-taking dynamics.