Inefficient speech-motor control affects predictive speech comprehension: atypical electrophysiological correlates in stuttering

Abstract

Listeners predict upcoming information during language comprehension. However, how this ability is implemented is still largely unknown. Here, we tested the hypothesis proposing that language production mechanisms have a role in prediction. We studied 2 electroencephalographic correlates of predictability during speech comprehension—pre-target alpha–beta (8–30 Hz) power decrease and the post-target N400 event-related potential effect—in a population with impaired speech-motor control, i.e. adults who stutter (AWS), compared to typically fluent adults (TFA). Participants listened to sentences that could either constrain towards a target word or not, modulating its predictability. As a complementary task, participants also performed context-driven word production. Compared to TFA, AWS not only displayed atypical neural responses in production, but, critically, they showed a different pattern also in comprehension. Specifically, while TFA showed the expected pre-target power decrease, AWS showed a power increase in frontal regions, associated with speech-motor control. In addition, the post-target N400 effect was reduced for AWS with respect to TFA. Finally, we found that production and comprehension power changes were positively correlated in TFA, but not in AWS. Overall, the results support the idea that processes and neural structures prominently devoted to speech planning also support prediction during speech comprehension.

Cortical tracking of speech is reduced in adults who stutter when listening for speaking

“…For naming latencies (response times, RT), we took the data from (Gastaldon et al, 2023), also available here: https://osf.io/5jkur/. Here we summarize how latencies were derived in the original study, but we refer the reader to the original article for additional details.…”

“…For both listening conditions, sentences could either be highly or low constraining towards a specific word completing the sentence. In this study, in order to allow for a better estimation of CTS and increase statistical power, we have not divided the unfinished sentence into high vs low constraining as in the original study (Gastaldon et al, 2023). Instead, we focused on the manipulation of the task, which implied two different listening conditions: listen with the aim of comprehending and answer to a true/false judgment question that appeared on the screen in the 20% of the trials (listening only), and listen in order to be ready to complete the sentence as quickly as possible by naming a picture (listening-for-speaking).…”

Trial‐by‐trial fluctuations of pre‐stimulus alpha power predict language ERPs

“…However, precision levels fluctuate spontaneously during the task even in presence of equally reliable cues (Feldman & Friston, 2010) and there is evidence that higher precision levels are associated with lower pre-stimulus power (Bauer et al, 2014;Cao et al, 2017). In Section 1.1, we reviewed evidence stating that in sentence comprehension paradigms featuring different levels of contextual constraint, more constraining contexts were associated with lower levels of pre-stimulus alpha power (Gastaldon et al, 2020;León-Cabrera et al, 2022;Rommers et al, 2017;Terporten et al, 2019;Wang et al, 2018), while similar studies found that contexts eliciting lower pre-stimulus alpha power were associated with more negative N400s (Gastaldon et al, 2023;Rommers et al, 2017;Terporten et al, 2019;Wang et al, 2018). This body of evidence helped us generate the hypothesis that lower pre-stimulus alpha might be associated with more precise predictions.…”

“…Additional supporting information can be found online in the Supporting Information section at the end of this article. Data S1: Supporting Information How to cite this article: Lago, S., Pezzetta, R., Gastaldon, S., Peressotti, F., & Arcara, G. (2023). Trial-by-trial fluctuations of pre-stimulus alpha power predict language ERPs.…”

“…Contrasts between different levels of contextual constraint indicated that expectations modulate alpha power in the pre‐stimulus interval: stronger predictions are associated with lower levels of pre‐stimulus alpha power than sentences leading to weaker predictions (Gastaldon et al, 2020; León‐Cabrera et al, 2022; Rommers et al, 2017; Terporten et al, 2019; Wang et al, 2018). Some of these studies also investigated N400 amplitude in relation to the preceding context and found that contexts eliciting lower pre‐stimulus alpha power are associated with more negative poststimulus N400s (Gastaldon et al, 2023; Rommers et al, 2017; Terporten et al, 2019; Wang et al, 2018). Even though the current evidence points to a link between pre‐stimulus alpha and N400 amplitude, to our knowledge post‐N400 positivities, such as the late posterior positivity/P600, have not been investigated yet in relation to pre‐stimulus dynamics.…”