Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The pleasure of rapid word segmentation in speech
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Iris van der Wulp1, Francesco Pupillo1, Laura Batterink2, Eleonore Smalle1; 1Tilburg University, The Netherlands, 2Western University, Canada
Statistical learning (SL) refers to the ability to extract regularities from the environment and is essential for speech segmentation and early word learning. Previous research has largely relied on offline, post-familiarization measures, limiting insight into the dynamics of the SL process. In a study with N = 208 adults, we used a reaction time-based task that tracks online SL of trisyllabic word structures in continuous speech. This task consisted of multiple speech streams in which participants detected syllables in either more predictable, less predictable, or random positions. Faster detection of syllables in more predictable positions relative to less predictable and random positions indicates successful SL. To assess participants’ intrinsic reward signals during the SL task, we incorporated a measure of subjective emotional responses related to intrinsic reward based on an adapted version of the Self-Assessment Manikin scales (SAM), presented to participants after each stream. Participants were asked to rate their momentary affective state using a digital version of the valence (pleasantness) and arousal subscales of the SAM. Furthermore, we included a manipulation of external feedback. This way, we examined whether online SL engages reward mechanisms, consistent with principles of reinforcement learning. Results indicated that participants indeed detected syllables more quickly when they occurred in more predictable positions, compared to both less predictable positions and random streams, indicating successful SL. External feedback did not affect the SL pattern or pleasantness ratings. Importantly, results showed that participants reported higher pleasantness after detecting syllables in more predictable positions, compared to detecting syllables in both less predictable positions as well as in random streams. Predictability did not affect arousal overall; instead, arousal was higher for predictable positions only among participants who received feedback. These findings suggest that speech segmentation engages intrinsic reward signals, aligning with prior work on other aspects of language learning. Overall, the results support the role of the striatum in SL and highlight a link between language learning and reinforcement learning mechanisms, which might originate early in development. As the current project made use of subjective ratings of pleasure and arousal, we will additionally propose a planned follow-up study using facial electromyography (EMG) as an objective measure of pleasure and arousal. We will investigate whether EMG measurements of pleasure and arousal within the streams correlate with the subjective SAM, and whether they yield similar results regarding the role of reward sensitivity in SL.
Topic Areas: Language Development/Acquisition,