Cortical dynamics of disfluency in adults who stutter

Abstract

Stuttering is a disorder of speech production whose origins have been traced to the central nervous system. One of the factors that may underlie stuttering is aberrant neural miscommunication within the speech motor network. It is thus argued that disfluency (any interruption in the forward flow of speech) in adults who stutter (AWS) could be associated with anomalous cortical dynamics. Aberrant brain activity has been demonstrated in AWS in the absence of overt disfluency, but recording neural activity during disfluency is more challenging. The paradigm adopted here took an important step that involved overt reading of long and complex speech tokens under continuous EEG recording. Anomalies in cortical dynamics preceding disfluency were assessed by subtracting out neural activity for fluent utterances from their disfluent counterparts. Differences in EEG spectral power involving alpha, beta, and gamma bands, as well as anomalies in phase-coherence involving the gamma band, were observed prior to the production of the disfluent utterances. These findings provide novel evidence for compromised cortical dynamics that directly precede disfluency in AWS.

Keywords: Disfluent speech; neural oscillations; phase coherence; stuttering.

Figure 1

Behavioral paradigm for eliciting disfluency. (A) Speech motor task involved display of target utterances for 2 sec. After a 0.5 sec delay participants were prompted to read aloud the displayed utterances within a 2 sec long window. Speech waveforms corresponding to a fluent and disfluent version of an example target utterance, “clegtisprodup,” is shown below. (B). Disfluency score for all 80 target utterances used. Its range varied between 0 and 40%. 10 target utterances that did not elicit any disfluency are marked with asterisk.

Figure 2

Disfluency and the acoustics of spoken utterances. (A) Significant effect of disfluency was observed for duration and production onset time after the appearance of the prompt, but not for peak loudness. (B) Disfluency score was comparable to stuttering severity.

Figure 3

Neural activity shows anomaly preceding disfluency. (A) Scalp electrode locations used to record EEG from during the speech motor task are shown in gray. (B) Representative mean power traces for theta, alpha, beta, and gamma band are shown for fluent and disfluent utterances matched for words and participants. Only the portion of the power trace from the start of the display of a target utterance to the appearance of the production prompt is shown. Production prompt is at 0 sec. (C) Scalp electrode locations showing significant effect of disfluency on EEG power bands. Effects were observed for alpha band at electrode location Af3, for beta band at C2, C5, and Cz and for gamma band at Af4, Fc3, and Fc1. White circles denote electrode locations that showed significance rise in power preceding disfluency. At other electrode locations, power decreased significantly. (D) Phase coherence spectrograms based on t‐scores (color scale ranging from −6 to 6). Only phase coherence involving gamma band showed significant effect of disfluency. The ordinate of the spectrogram represents frequency range for the gamma band and the abscissa represent time, with 0 marking the appearance of the production prompt. The time axis includes the duration from the start of the display of a target utterance to the appearance of the production prompt. Alpha‐gamma phase coherence differences were observed at electrode locations Fc1 and Fc3, while theta‐gamma differences were found at Af4 and Fc3.