We studied 13 AWS of whom three were excluded from analysis for lack of clear right-handedness (Table 1). The AWS were primarily recruited from an intensive therapy course held in the Kassel Stuttering Therapy Centre, Bad Emstal, Germany. Inclusion was based on the participants’ consent, an absence of medical or self-reported neurological disease, and an absence of medication that influences the excitability of the central nervous system. Given the enrollment in a therapy course and the day-to-day variability of symptom severity, we did not use a minimum percent of syllables stuttered for inclusion in the AWS group. All participants in the AWS group had their diagnosis confirmed by a board-certified phoniatrician prior to enrollment in the therapy program. Sample size was estimated based on previous work using comparable methodologies (e.g., Sommer et al., 2003, 2009). In addition, we recruited 13 fluent speakers matched for age and sex, and carefully selected for the absence of a personal or family history of stuttering or any treatment by a speech-language pathologist. The participants’ speech fluency levels were assessed by a qualified speech-language pathologist who was blinded regarding group status. Stuttering severity was quantified using the German version of the stuttering severity instrument (SSI-3) (Sandrieser and Schneider, 2008), based on two video samples of spontaneous speech as well as reading. Musical practice refers to playing an instrument or singing and was quantified as 1 = never or rarely, 2 = regularly, 3 = professionally. Education was quantified ordinally (1 = school [mittlere Reife], 2 = high school [Abitur], 3 = less than 2 years college, 4 = 2 years college [Bachelor, Vordiplom], 5 = 4 years college [Studium], 6 = postgraduate); handedness was assessed using the Edinburgh Handedness questionnaire (Oldfield, 1971).

We used a Magstim 200² monophasic TMS device to quantify motor cortex excitability before, during and after the speech tasks. Participants sat in a comfortable reclining chair with armrests, while the experimenter stood behind and out of the participant’s visual field. To detect the optimal motor representation of the abductor digiti minimi muscles of each side, we delivered marginally suprathreshold pulses with a figure-of-eight coil with an outer diameter of 7 cm. The coil was held about 45° posteriolaterally, approximately perpendicular to the presumed location of the central sulcus. We delivered suprathreshold pulses to induce current flow from posterior to anterior in the brain, in order to detect the optimal representation, and then marked the scalp with a pen. We then reduced the stimulus intensity gradually to detect the resting motor threshold (RMT), defined as the minimal intensity where five out of ten consecutive trials elicited contralateral motor evoked potentials (MEP) larger than 50 μV in the target muscle (Rossini et al., 2015). We detected intensity yielding contralateral MEP amplitudes of about 0.5 mV, which were then used as test pulse intensity for the speech tasks (0.5 mV MEPs). These procedures were carried out for each hemisphere separately, in random order. We chose comparatively small baseline MEP amplitudes (Tokimura et al., 1996) because we did not expect any inhibition, so we aimed at increasing the yield with regard to MEP amplitude facilitation by choosing relatively low MEP amplitudes at baseline.

Motor evoked potentials were recorded bilaterally from the abductor digiti minimi muscle using silver-silver chloride cup electrodes in a belly tendon montage with a sampling frequency of 5 kHz, filtered between 20 and 2000 Hz and recorded using a CED 1401 amplifier and Signal 4.16 software (Cambridge Electronic Design, Cambridge, United Kingdom).

In this study, we explored a wide range of six tasks – a rest/null speech task (RE), solo reading (SR), chorus reading (CR), spontaneous speech (SP), singing (SI), and non-verbal orofacial movements (NM), CR and SI being known to be fluency-enhancing in AWS (Bloodstein and Ratner, 2008). Each of the six conditions were presented between a brief period of rest before and after each trial. For the rest condition, the participants were asked to remain silent and not to engage in conversation with the experimenter.

For the two reading tasks, Solo Reading and Chorus Reading, we chose a non-fictional text (annual report by the supervising board of the Deutsche Bahn) (Aufsichtsrat der Deutschen Bahn, 2009). This was printed in easily readable font size and displayed on a hold at about 30 cm before the participant’s eyes. Participants were instructed to read either alone (Solo Reading), or loud in chorus with the experimenter standing behind the participant (Chorus Reading). For the Spontaneous Speech task, the text display was removed and participants were encouraged to speak about their recent activities; the experimenter asked open questions to enhance communicative output when necessary. For Singing, participants were encouraged to sing familiar German nursery rhymes [“Alle meine Entchen” (Public domain, 2012a) or “Hänschen klein” (Public domain, 2012b)]. For the Non-verbal Orofacial Movement task, participants were encouraged to produce articulatory movements without verbal content; these included humming, lip smacking, kissing movements, or tongue clicking. To ensure appropriate performance, the different experimental procedures were briefly practiced before starting the actual recordings.

Before each task, we recorded 4 min of suprathreshold TMS pulses at 0.25 Hz, 30 pulses per hemisphere, with the respective test pulse intensity. Details about the forthcoming tasks were only revealed to the participant after this baseline; they were then given four to 5 min to perform that task, i.e., as long as necessary to stimulate either hemisphere 30 times, again at 0.25 Hz. After each task, we recorded another 4 min to obtain 30 MEPs for either side, also with test pulse intensity at 0.25 Hz. An interval of at least 10 min followed before starting the next speech task’s baseline. The order of speech tasks was randomized. The order of hemispheres was stable before, during and after each task, but randomized across tasks. Participants were instructed to refrain from hand and arm movements while performing the tasks.

Demographic data of all participants is shown in the Table 1.

In both groups, RMT was lower in the left hemisphere than in right hemisphere. The analysis of RMT yielded no main effect of group, but an effect of hemisphere (F_(1,21) = 13.98, p = 0.001), and no interaction of hemisphere by group (Figure 1). Following the same pattern, stimulus intensity was lower in the left hemisphere than in right hemisphere. It yielded no main effect of group, but an effect of hemisphere (F_(1,21) = 9.68, p = 0.005), and no interaction of hemisphere by group.

Motor evoked potentials amplitudes were small at baseline, increased to a variable extent during task performance, and returned to baseline levels or slightly above after the end of task performance (Figure 2).

At baseline before task onset, the raw contralateral MEP amplitudes did not yield any main effect, but an interaction of hemisphere of stimulation × group (F_(1,21) = 10.77, p = 0.004, Figure 3), and no other significant interaction. Post hoc t-tests showed higher amplitudes in ANS than in AWS for left hemispheric stimulation only. Apparently, a proper between-group matching of MEP amplitudes at baseline had not been successful. The coefficients of variance of these raw MEP amplitudes for the six baselines ranged from 0.30 to 0.54 in the control group and from 0.25 to 0.63 in the patient group. The F-tests involving group were all non-significant, excluding major differences in MEP amplitude variability between groups.

The MEP amplitudes during task performance normalized to the individual baseline revealed a main effect of speech task (F_(5,105) = 8.13, p < 0.0001, Figure 4), and no other main effect or interaction. Specifically, no main effect or interaction involving hemisphere of stimulation was observed, which is why we pooled hemispheres in Figure 4. Post hoc t-tests indicated that the Rest/null speech task differed from all other tasks except Solo Reading, and that Solo Reading differed from Non-verbal Orofacial Movements. Compared to ANS, AWS tended to show stronger MEP amplitude increases during Solo Reading, but tended to show smaller MEP amplitude increases during Non-verbal Orofacial Movements; but these between group comparisons failed to reach significance (Figure 4, see also Appendix).

Excluding the Rest/null speech task condition from the above-mentioned analysis did not reveal any effect or interaction involving group.

After the end of task performance, MEP amplitudes returned to baseline values, not yielding any main effects or interactions.