The effect of single syllable silent reading and pantomime speech in varied syllable positions on stuttering frequency throughout utterance productions

Available online at www.sciencedirect.com

ScienceDirect Speech Communication 75 (2015) 76–83 www.elsevier.com/locate/specom

The effect of single syllable silent reading and pantomime speech in varied syllable positions on stuttering frequency throughout utterance productions Daniel Hudock a,⇑, Nicholas Altieri a, Lin Sun b, Andrew Bowers c, Christian Keil d, Joseph Kalinowski b a

Department of Communication Sciences and Disorders, Idaho State University, 921 S. 8th Ave. Stop 8116, Pocatello, ID 83209, USA Department of Communication Sciences and Disorders, East Carolina University, 1310 Health Sciences Dr, Greenville, NC 27858, USA c Department of Communication Sciences and Disorders, University of Arkansas, Epley Center 275, 606 N. Razorback Rd., Fayetteville, AR 72701, USA d Silver Hills Health Care Center, 3450 N Buffalo Dr., Las Vegas, NV 89129, USA b

Received 18 February 2015; received in revised form 9 September 2015; accepted 28 September 2015 Available online 9 October 2015

Abstract Background: Stuttering is an overt speech disorder with the majority of disruptions occurring during phrase and sentence initiations. Recent theories and models of stuttering often describe deficits in neuromotor processes for planning motor speech movements, especially those involved during motor initiation. Interestingly, stuttering-like behaviors are reduced by approximately by nearly 100% during silent articulations (i.e., pantomime speech). If stuttering is primarily a deficit in neuromotor planning for speech actions, disfluent behaviors should be significantly reduced throughout an utterance when people who stutter employ silent reading or pantomime strategies on one syllable of an audibly produced utterance. Aims and scope: The aim of the first study was to examine stuttering frequency during oral reading as participants who stutter produced the initial syllable under silent reading (SR), pantomime (P), and redacted (R) speech conditions. Similarly, the second study examined stuttering frequency during oral reading as a unique set of participants who stutter employed SR or P strategies on single syllable productions in initial, middle, and final syllable positions of an utterance. Methods and procedures: Two unique sets of participants who stutter audibly read sentences under baseline and experimental conditions. Experimental conditions for the first study consisted of (1) SR, (2) P, and (3) R on the initial syllable of audibly produced utterances. Experimental conditions for the second study consisted of participants performing SR and P on single syllables in initial, middle, and final syllable positions throughout an audibly produced utterance. Results: Stuttering was significantly reduced during all experimental conditions in the first study. All experimental conditions differed from all others in the first study with P, SR, and R progressing from most to least effective. Results from the second study revealed differences from both SR and P to baseline and differences in both the initial and final syllable positions to baseline, but not the middle position to others, which approached statistical significance (p = 0.1). In the second experiment, post hoc comparisons revealed that P in the initial position was the most effective and that P was significantly more effective than SR in both the initial and final positions, which supports the findings from the first study. Conclusions and implications: Results from the current studies demonstrate a reduction of stuttering throughout an utterance when participants employed SR and P strategies on single syllables within an utterance. The greatest reduction in stuttering frequency occurred when silent motor plans were enacted and not just read or omitted (i.e., during P conditions). Supporting current feed-forward models of

⇑ Corresponding author. Tel.: +1 (208) 282 4403.

E-mail addresses: [email protected] (D. Hudock), [email protected] (N. Altieri), [email protected] (L. Sun), [email protected] (A. Bowers), [email protected] (C. Keil), [email protected] (J. Kalinowski). http://dx.doi.org/10.1016/j.specom.2015.09.012 0167-6393/Ó 2015 Elsevier B.V. All rights reserved.

D. Hudock et al. / Speech Communication 75 (2015) 76–83

77

stuttering, strategies were most effective when employed in initial syllable positions. It can be hypothesized that behavioral strategies such as SR and P speech alter predictive neuromotor planning via feedback mechanisms or enhancing output gain of neuromotor planning regions. Ó 2015 Elsevier B.V. All rights reserved.

Keywords: Stutter; Fluency; Pantomime; Silent reading; Models of speech; Feed-forward

1. Background Stuttering is a fluency disorder in which a majority of the speech disruptions occur during phrase and sentence initiation (Bloodstein and Bernstein-Ratner, 2007). Interestingly, stuttering has also been shown to exhibit a marked reduction during whispered speech, and most germane to this paper, a near elimination during silently articulated pantomime speech (P) (Brenner et al., 1972; Commodore, 1980; Commodore and Cooper, 1978; Perkins et al., 1976). Researchers typically report 80–90% reductions in stuttering frequency during whispering (Brenner et al., 1972; Commodore, 1980; Commodore and Cooper, 1978; Perkins et al., 1976; Rami et al., 2005) and nearly 100% fluent speech productions during P (Commodore, 1980; Commodore and Cooper, 1978; Perkins et al., 1976).1 It should be noted that due to participants using whispered and P strategies throughout the entirety of the passages, Commodore (1980), Commodore and Cooper (1978), and Perkins et al. (1976) used visual-only presentations of client recordings during analysis of stuttering. This was done to maintain consistency across all conditions (control, whispered, and P). Type and frequency of stuttering was therefore classified using visually evident behaviors only. Pantomiming procedures in relation to studies in fluency disorders have primarily been utilized to test hypotheses regarding hierarchical discoordination speech systems in people who stutter (PWS) (Commodore, 1980; Commodore and Cooper, 1978; Perkins et al., 1976) and to examine the adaptation effect (i.e., reduction in stuttering after repeated productions of the same content) (Brenner et al., 1972). More specifically, a contribution from many of these studies was initial support for what became known as the ‘‘discoordination hypothesis”. This suggested that simplifying the speech process contributes to an easier and more efficient coordination of the articulatory system, and hence, increased fluency. For example, Commodore (1980) and Commodore and Cooper (1978) required participants who stutter to read passages during 1 These findings are consistently reported, with the exception of Van Riper (1971) who reports that Deal (1982) presented a case study of a client that had attempted suicide and did not exhibit benefit from many well- known fluency enhancing conditions (e.g., choral speech, signing, pantomiming, or whispering). This is also true for participants who have had laryngectomies and produce alayngeal speech, as reported in Van Riper (1971). Use of an electrolarynx, essentially pantomime speech, maintains similar fluency enhancements, as reported in Van Riper (1971).

voiced (i.e., baseline), whispered, and silently articulated (i.e., P) stressed and non-stressed conditions. Results from these studies indicated differences in stuttering frequency between stressed (i.e., visible video recording) to nonstressed conditions (i.e., hidden camera). More relevant to the current manuscript, differences were found between whispered and silently articulated experimental conditions to baseline. Significantly less stuttering was reported during speech that was whispered and silently articulated compared to during baseline. No differences were reported between the two experimental speech conditions. Similarly, in testing theoretical notions of fluency enhancements being caused by a slowed speech rate, Perkins et al. (1976) analyzed speech characteristics of participants who stutter that verbally read under voiced (i.e., baseline), whispered, and pantomimed conditions. Stuttering was reduced during experimental conditions with pantomime, whispered, and voiced progressing from greatest reduction to least. They also reported that as the speech hierarchy was reduced (i.e., reducing voicing, then airflow components) participants increased articulatory rate. Lastly, researchers have also utilized whispering, P, and silent reading (SR) procedures to examine the adaptation effect in stuttering. Brenner et al. (1972) had participants read the same passage three times during silent rehearsal (i.e., silent reading ‘‘SR”), silent rehearsal with lip movement (i.e., P), whispered rehearsal, aloud rehearsal (i.e., audible production), and no rehearsal conditions before audibly reading the passage a fourth time during which stuttering was analyzed. Results revealed that stuttering frequency was only significantly reduced during the audible rehearsal condition. Although previous studies of the P and SR effect on stuttering frequency have demonstrated that the two conditions reliably reduced stuttering frequency, all of the aforementioned studies employed fluency enhancing strategies (P and SR) throughout the productions. As recent theoretical frameworks posit that stuttering is most likely to occur on the initial sound or syllable, it may be important to examine how P and SR strategies reduce stuttering frequency when only implemented on single syllables within sentence productions. In summary, in all of these previously described studies researchers had participants use P and SR throughout the entirety of the passage then per the observed results, inferred about theoretical notions of hierarchies in the speech production systems of PWS. Given recent advances in neuro-based theories and models of stuttering and

78

D. Hudock et al. / Speech Communication 75 (2015) 76–83

speech production that offer not only hierarchies but also sequential neural location and functioning predictions, current models allow for behavioral paradigms to be applied and test certain predictions.

voicing begins the PWS may relay on the activated afferent signal to provide online correction.

1.2. Rationale and hypothesis 1.1. Implications for feed-forward models of stuttering Recent theories and models of stuttering attribute one major cause of overt stuttering to be deficits in feedforward neuromotor networks involved with speech articulation (see; Max et al., 2004; McQuaide et al., 2014; Salmelin et al., 2000; Venkatagiri, 2004). Max et al. (2004) tested predictions of Guenther’s (1994) Directions into Velocities of Articulators (DIVA) model of speech production by applying syntax for acquired stuttering. Guenther (1994) proposed an integrative internal mechanism of speech production that utilizes both feed-forward and feedback networks. Feed-forward systems use motor commands prepared in premotor and speech motor cortical areas before the initiation of articulatory movements. Feedback systems do not use preparation prior to articulation: Instead, motor commands are generated simultaneously with the execution of top-down control mechanisms with adjustments occurring in real-time. As differential afferent and efferent signals are received according to the model, neural circuitry makes real-time adjustments and modifies the internal model of speech (Guenther, 1994). Max et al.’s implementation of Guenther’s model of speech production integrates information from both feed-forward and feedback systems for its application to stuttering. The important point is that Max et al. (2004) proposed that stuttering occurs because of an over-reliance on feedback from the auditory consequences of speech movements due to deficits in the internal modeling process. Max hypothesized that during auditory or visual feedback from speech signals (Delayed Auditory Feedback or Delayed Visual Feedback) activation of the auditory cortex allows for monitoring of predicted plans, called ‘‘efference copies”, which would otherwise go uncorrected. A second explanation suggests that the constant altered feedback essentially shuts down the feedback system, thereby increasing the output gain of the feed-forward system and reducing overreliance on feedback while aligning internal models resulting in increased fluency and accuracy of feed-forward plans.2 Additionally, the model hypothesizes that stuttering primarily occurs during initiation of phrases because at the time of speech initiation the afferent feedback has not arrived at the motor plan for comparison. By contrast, when speech has already been initiated and 2 Bohland et al. (2014) have recently proposed an update to the DIVA model, entitled the gradient order DIVA (GODIVA), which predicts greater involvement from basal ganglia circuitry. Their model predicts similar disruptive feed-forward and feedback mechanisms involved with stuttering, however, it has yet to be empirically tested with perturbations for fluency disorders.

One prediction from these modeling frameworks, and the purpose of this paper, is to test the hypothesis that when P and SR are employed during initial syllable productions it should increase fluent speech outputs. This should lead to an increased activation of pre-planning (i.e., the feed-forward system) and as a result, should alter articulatory representations creating enhanced fluency. The following two studies therefore examined the effect of P and SR applied to different positions in the speech sample on stuttering frequency. Specifically, the first study examined P, SR, and redacted speech (R) on the first syllable of the utterance. As participants employed P and SR on the initial syllables, they audibly produced speech on the second syllable, unlike in the baseline condition in which speech was audibly produced on the first syllable. By blacking out, or redacting, the first syllable in the R condition, participants began producing speech on the second syllable just like in the P and SR conditions. Therefore the R condition in the first experiment served as a secondary baseline condition to compare speech that was audibly produced on the second syllable of the phrase. If efferent predictive plan copies are enacted and subsequently transmitted to motor planning areas, then implementation of P and SR should enhance the plan and allow adequate time for comparison of the plan within feed-forward and feed-back systems. This may also allow for correction to variable, incomplete, or insufficient efferent copies. Implementing such strategies on single syllable productions should increase fluency throughout an utterance. Next, due to the extent to which these procedures differ in regard to activation of cortical auditory regions and neural motor systems, it was hypothesized that P, representing a more complete efferent plan with likely increased neural motor planning, would be more effective than SR at reducing stuttering. In regards to the second study initial, medial, and final syllable positions with P and SR were examined. If increased stuttering during initiations is due to afferent signals not being received by motor planning areas before speech occurs, instituting strategies in the initial syllable position would delay the signal production. This might thereby allow time for the afferent signals to be received and comparison/correction to occur. It is likely that planned changes to the middle and final syllables of the utterance would still alter efferent copies and correction of motor plans, however the initial position would be more effective than middle or final positions due to the output gain of the feed-forward system being highest during initiation. Lastly, it was hypothesized that strategies used on the final syllable would be more effective than those used on the middle syllable. This prediction was made due to the fact that as participants plan for alteration of the final

D. Hudock et al. / Speech Communication 75 (2015) 76–83

syllable they are primarily reliant on feedback systems as no additional planning for speech is occurring. 2. Methods All participants were verbally briefed before reading then signing approved informed consent documents. The first study had Institutional Review Board approval from both Idaho State University and East Carolina University, while the second study only had approval from Idaho State University, as all data for that study was collected and analyzed at that institution. Participants from the first study were run at East Carolina University. All participants were instructed not to use additional fluency enhancing strategies during any portion of the studies. 2.1. Participants Participants included individuals who stutter with no self-reported history of concomitant speech, language, reading, cognitive, emotional, hearing, or uncorrected visual deficits. Eight individuals who stutter with an average age of 33 years, with a standard deviation of 15 years participated in the first study. Nine individuals who stutter with an average age of 28 years with a standard deviation of 12 years participated in the second study. 2.2. Instrumentation A 13-inch Macbook Pro or 13-inch HP Pavilion laptop was positioned approximately 24 inches directly in front of participants who audibly read phrases presented via PowerPoint 2010 and used experimental procedures when indicated. A Samsung SCMX-10 video recorder was positioned approximately 27 inches directly in front of participants at eye level for audiovisual recording. 2.3. Stimuli Stimuli texts were retrieved from Kalinowski et al. (2000) who segmented similar seventh and eighth grade level 300-syllable length passages into approximately 28 8–12 syllable length phrases. Phrases were presented in black 28-point Calibri font with a white background. During the first study participants were tested under baseline and three experimental conditions: (1) redacted (R); (2) silent reading (SR); and (3) pantomime (P). In the first study, participants always performed procedures on the first syllable of the phrase and completed a condition in its entirety before starting another condition/procedure. For example in the phrase ‘‘The radio was developed”, the word ‘‘the” was italicized for the SR condition but was italicized and underlined in the P condition. Because use of SR and P strategies had the participants begin typically producing speech on the second syllable, the R condition served as a second baseline by having participants

79

begin speaking on the second syllable of the utterance. In this condition, text was highlighted in black. For example, ‘‘The radio was developed” thereby controlling for positioning within the phrase. In the second study the same identifiers (i.e., italics or italics with underlying) were used for SR and P, but no R condition was employed due to the examination of the varied syllable positions. The second study presented SR and P in initial, middle, and final syllable positions throughout the phrase. If an uneven number of syllables were in a given phrase, researchers attempted to designate the middle syllable with the experimental marker. When a multisyllabic word was in the middle syllable position of the phrase the experimental syllable was moved to the beginning of that word or to the beginning of the next word. Researchers attempted to move the experimental syllable as minimally as possible, but always presented it at the beginning of a word. When there was an even number of syllables in a phrase researchers designated the first middle position as the experimental position, if at the beginning of a word, or if a monosyllabic word. Throughout all conditions the most that the middle syllable position was moved was one syllable position in distance. Unlike the middle syllable position, the final position occasionally was the final syllable of a multisyllabic word. 2.4. Procedures A researcher demonstrated experimental procedures during a training phase prior to the start of each experimental condition. Participants then practiced specific procedures from four to six times on text that was not used during the recorded experiment. Between each condition participants maintained a two-minute period of spontaneous conversation with the researcher to reduce any potential carry-over effects. Passages and condition sequences were randomly assigned to participants using numeric sequences via www.randomizer.org. Participants orally read phrases performing experimental strategies when designated. Each condition was completed in its entirety before performing other experimental procedures. During baseline conditions participants orally read the presented phrases. ‘‘Redacted speech” in the first study was introduced to include another type of baseline control condition since speech was actually initiated on the second syllable of the utterance in the SR and P conditions. ‘‘Silent Reading” procedures required participants to read experimental syllables without articulator movements. Pantomiming procedures required participants to silently articulate the experimental syllables without audible speech being produced. All non-experimental syllables throughout the utterances were typically produced. 2.5. Analysis Stuttering was defined as syllable or word repetitions, phoneme prolongations, and postural fixations (i.e., ‘‘silent

D. Hudock et al. / Speech Communication 75 (2015) 76–83

blocks”; Armson and Stuart, 1998). As there is often overlap when categorizing type of stuttering episodes, researchers for the current studies only denoted observation stuttering events and did not label the type of event. To maintain consistency the first author analyzed all recordings for total number of stuttered syllables by conditions, which were then divided by 300 (the total syllable length) to determine proportion of stuttering by participant per experimental condition. An intra-rater reliability analysis was conducted on a random sample of 10% of the data unique to each experiment. A student researcher was trained on identifying and denoting instances of stuttering by the first author using unrelated data. This training took a total of one hour and thirty minutes over three training sessions for the student to obtain higher than .75, or excellent, reliability via Cohen’s Kappa analysis (Cohen and Cohen, 1983). The student performed unique inter-rater reliability testing on a randomized 10% of the data for each of the experiments. For the first study Cohen’s Kappa (SPSS 21.0 for Mac) syllable-by-syllable agreement (Cohen and Cohen, 1983) revealed a Kappa value of 0.86 for the intra-rater and 0.69 for the inter-rater testing. Reliability testing for second study revealed Kappa values of 0.92 for intra-rater analysis and 0.74 for inter-rater analysis. Values above 0.41 represented moderate agreement and values greater than 0.75 represented excellent agreement above chance (Viera and Garrett, 2005) (see Figs. 1 and 2). 3. Results Prior to inferential statistical analyses, proportions of stuttering were transformed into arcsine units to reduce end point weighting of proportion values (see Viera and Garrett, 2005). This is a commonly used transformation when working with proportional values that are under .10 or over .90. As several of the participants exhibited proportion of stuttering within the lower (.10) range, this is an appropriate transformation.

0.14

Proporon of Stuering

0.12 0.1

0.12 0.1 0.08 0.06 0.04 0.02 0

Baseline

IP

ISR

MP

MSR

FP

FSR

Fig. 2. Stuttering frequency (proportion) for the second study averaged across the nine participants for each of the seven conditions with error bars representing plus one standard error of the mean. I = initial, M = middle, and F = final syllable positions in the phrase.

3.1. Experiment 1 A one-factor repeated measures analysis of variance (ANOVA; SPSS Inc., SPSS 21.0 for Mac) was conducted to examine the effect of each condition on stuttering frequency. First, a significant main effect for condition was revealed [F(3, 105) = 117.67, p < 0.001, g2p = 0.77]. Next, to examine the source of the main effect across conditions, post hoc comparisons were used with Bonferroni adjustments. Six comparisons were carried out in total, as shown in Table 1 (Adjusted_a = .05/6 = .008). All conditions were significantly different from all other conditions (p < 0.001). From baseline, stuttering was reduced 15% during the redacted condition, 42% in SR, and 72% during P. 3.2. Experiment 2 A one-factor repeated measures ANOVA (SPSS Inc., SPSS 21.0 for Mac) was carried out to examine the effect of condition on stuttering frequency. A significant main effect for the conditions was revealed [F(3.237, 26.188) = 3.475, Greenhouse-Geisser p = 0.027) g2p = .303]. Posthoc comparisons with Bonferroni adjustments (resulting in an adjusted_a = .05/21 = .002) revealed a difference between baseline to initial pantomime (p = 0.005), with no other significant differences revealed. Researchers then used orthogonal single degree of freedom contrast comparTable 1 Post-hoc comparisons across conditions presented as mean percent differences of stuttering with standard error, significance level, and Cohen’s d effect size estimates. Means are based on estimated marginal means.

0.08 0.06 0.04 0.02 0

0.14

Proporon of Stuering

80

Baseline

SR

P

R

Fig. 1. Stuttering frequency (proportion) for the first study averaged across the eight participants for each of the four conditions with error bars representing plus one standard error of the mean.

Comparison

Mean difference % (SE)

Significance (p)

Effect size

Baseline vs. SR Baseline vs. P Baseline vs. R SR vs. P R vs. SR R vs. P

15.4 (1.7) 29.3 (1.5) 6.2 (1.4) 13.9 (1.6) 9.2 (2.0) 23.0 (1.7)

<.001 <.001 <.001 <.001 <.001 <.001

3.2 6.9 1.6 3.1 1.6 4.8

D. Hudock et al. / Speech Communication 75 (2015) 76–83

isons structured within the ANOVA for exploratory analysis. Differences were revealed between P to baseline (p = .008), SR to baseline (p = .017), initial position to baseline (p = .004), and final position to baseline (p = .010). No significant differences were revealed between the middle position to Baseline (p = .10), or between the middle position to either initial or final positions (p > .05). Lastly, no significant differences were observed between P and SR (p = .746); however, P resulted in significantly less stuttering than SR when used in the initial and final syllable positions (p < .05). This result indicates differential effects between P and SR that are diluted by the data from the middle syllable conditions. 4. Discussion Examining fluency-enhancing strategies for PWS is not a new line of investigation: Previous researchers have investigated the efficacy of P and SR procedures, and have inferred theoretical constructs regarding stuttering and speech production (Brenner et al., 1972; Commodore, 1980; Commodore and Cooper, 1978; Perkins et al., 1976). However, unlike any previous examination, this was the first study to vary the P and SR procedures for use on single syllables as opposed to consistent use throughout tasks. Additionally, this was the first study to examine the effects of such procedures across varied syllable positions throughout an entire utterance. Recent neurological evidence along with current model frameworks allows for application and interpretation of behavioral paradigms to theoretical constructs. The most pertinent finding from the current study is that P and SR strategies used on single syllables differentially affected stuttering frequency throughout the audible production of utterances. In both studies, P was more effective than SR; however, the magnitude of change was greater in the first study. The P versus SR comparison also showed a strong effect size (>3; Table 1); this points to the effectiveness of P in reducing stuttering as compared to a technique like SR that did not distinctly implement a lip-movement prior to speech initiation. This fluency enhancing effect hypothetically arose since P constitutes a motormovement that alters and likely recalibrates the previous motor-movement plan. Conversely, SR constitutes a rehearsal of phoneme-graphic constructs via reading that do not deal directly with physically implemented motormovements. Taken together, the alteration of feedforward mechanisms appeared weaker in the SR compared to P condition. Incidentally, the results from the first study did indicate that the R condition reduced stuttering. Baseline to R accounted for the smallest degree of change indicating that even shifting the initiation position enhances fluency and likely feed-forward motor planning. An interesting finding relating to the second study was that stuttering was not only reduced when strategies were used on the initial syllable – which was the most effective – but also reduced stuttering when used in the final posi-

81

tion, and approached statistical significance (p = 0.1) in the middle position. Stuttering was reduced to the smallest extent when strategies were used in the middle position. Interestingly no statistically significant differences were observed in contrast comparisons between syllable positioning, with the exception of post hoc comparisons revealing differences between IP to baseline, indicating that IP had the strongest effect. As phrase initiation creates the greatest amount of disfluency, usage of alteration strategies on initial syllables is efficacious and therefore common in clinical practice (Bloodstein and Bernstein-Ratner, 2007). Alterations to the middle and final syllable positions are likely more difficult to apply clinically; they also result in smaller degrees of change similarly to the effect found in the current study. It requires less effort to prepare and plan for changes to the initial syllable of a sentence rather than during ongoing productions in the middle or at the end of the sentences. Slightly increased effectiveness in the final position may be due to the fact that no additional planning is required after production of the final syllable. Whereas, the lowest degree of change found from the middle position might be due to the modification during production and need for additional speech planning after the alteration. 4.1. Model-based explanations There is a current lack of research specifically examining the use of speech alteration strategies in varied syllable positions within a phrase; however it could be hypothesized that anticipated and planned use of SR and P strategies in medial and final syllable positions still affected the feedforward mechanism, albeit not to the same extent as the initial position. Planned alterations could, as predicted by Max et al. (2004) and Bohland et al. (2014), reduce overreliance on feedback and alter initiation, maintenance, and termination processes of the basal ganglia in partial contribution to the efference copy. Overall, delays affecting the internal modeling process as described by Max et al. (2004) are supported by the common finding of decreased white matter tracts underlying the ventral primary motor cortex (vMC) (Sommer et al., 2002; Watkins et al., 2008), and reports of elevated dopamine levels affecting basal ganglia in adults who stutter (AWS) (Wu et al., 1996). In a simulation study using the GODIVA computational model, Civier et al. (2013) demonstrated that simulating elevated striatal dopamine levels resulted in greater disfluency in the initial syllable position. However, simulated white matter tract anomalies in the vMC resulted in dysfluencies in the middle and final positions and could not explain initial syllable dysfluencies. The explanation for those findings was that elevated dopamine levels introduce noise into the process of speechsound cell selection via interaction with the thalamus, creating a lower signal-to-noise ratio for the selection of the desired speech sound cell map (SSM). In this way, elevated dopamine levels may bias disfluency toward the initial syllable position because the selection of the SSM for subse-

82

D. Hudock et al. / Speech Communication 75 (2015) 76–83

quent production is noisy. By contrast, white-matter anomalies appear to only alter or delay the selection of the next syllable in the sequence, affecting middle and final syllable positions. Stated simply, the widely supported tendency for stuttering to occur on the initial sound or syllable is supported by alterations to BG activity in utterance planning. In the current study, the finding that stuttering was reduced to the greatest extent when strategies were used in the initial position, then progressed in effectiveness from final to middle positions, is consistent with basal ganglia involvement in altering motor planning. As suggested by Alm (2004, 2005) one potential neurophysiological explanation for this phenomenon is that motor planning for speech makes use of different pathways depending on whether the movement is overlearned and automatic or speech planning is performed with increased attention to auditory or sensory perceptions. In the former case, speech planning relies on a medial premotor system (i.e., supplemental motor area and BG) while in the later case it relies more heavily on a lateral premotor system (i.e., ventral premotor cortex and cerebellum). Increased attention to speech along with motoric alterations from habitual speaking is well-known to decrease stuttering moments along with externally cued pacing mechanisms, also known to increase activity in the lateral premotor system (e.g., choral speech; metronome speech) (Bloodstein and BernsteinRatner, 2007). As stated previously, P is an inherently unusual mode of speaking, most likely resulting in alterations in the motor planning process. As such, findings in the current studies for P in the initial syllable position may be interpreted as being consistent with compensation for medial premotor timing miscues by inducing timing cues from the lateral premotor system prior to speaking. Further, reduced effects on the middle and final syllables may reflect delays in syllable sequence timing due to white matter anomalies present in the lateral premotor system. Another potential explanation for our findings is that t halamo-cortical–striatal loops also interact with peripheral motor state. Prior to motor movements, the basal ganglia structures send a stop signal to muscles maintaining certain postures (for a review of circuitry, see Redgrave et al., 2010; Shepherd, 2013). Once the muscles maintaining the posture release, this allows voluntary motor commands to be initiated. This disinhibition is most present during motor initiations; as motor actions occur, the basal ganglia maintain components of fine motor control (Seidler et al., 2014). At the end of the motor movements, the basal ganglia reinitiate inhibitory controls for postural stability, likely increasing required activity. Therefore as predicted by current models of speech production with predictions for stuttering (Bohland et al., 2014; Max et al., 2004) and Alm (2004), it is likely that feed-forward systems are altered during planned use of strategies, while the differential effect of position may be due to subcortical involvement and feedback coherence. However, no theoretical accounts of speech motor control have provided predic-

tions for how peripheral motor state interacts with cortical internal model formulation in stuttering. More recent state feedback control models (SFC) of motor control in speech production suggest the current state of muscular effectors is critical for the formulation of dynamic predictions as to the future state of the effectors (Houde and Nagarajan, 2011). As such, sensory feedback from the effectors may be critical in CNS formulation of internal model predictions even prior to movement initiation. Differences in the state of muscular effectors as measured by acoustic and kinematic measures compared to controls have been widely reported in stuttering and appear to be highly dependent on speaking context, developmental history, severity, and treatment history (Armson and Kalinowski, 1994). We tentatively speculate that the difference between a motor control strategy requiring actual movement (e.g., P) relative to one requiring silent rehearsal, may be mediated by decreasing model prediction errors via interaction between CNS model formulation and sensory feedback prior to speaking. Perhaps a grand challenge for models of motor control in stuttering is to integrate CNS internal model formulation with the well-known motoric, fluency enhancing compensatory strategies, some of which are employed even prior to an utterance (e.g., preparatory set). 5. Summary and conclusion Findings from the current study first added converging evidence to prior research demonstrating that active motoric alterations or non-motor operations applied throughout an utterance are effective at reducing stuttering frequency when implemented on single syllable productions. Due to the fact that stuttering most often occurs during initiations of phrases and sentences, clients are often trained to implement speech modification techniques at those locations. Before sentence or phrase initiations occur there is typically extended pause time so, using strategies during initiations also has the added benefit of allowing increased time for neuromotor planning, as compared to during online sentence productions. Generally speaking, conversational speech is hierarchically more difficult than audible reading for PWS; therefore, in applying the current findings to conversational speech, one would likely find more benefit from implementing strategies during phrase and sentence initiations as well as during moments of anticipated stuttering. As PWS can briefly anticipate stuttering before it occurs, stuttering modification approaches train clients to use strategies multiple times throughout phrases and sentences as they anticipate stuttering. This brief anticipation of stuttering may be illustrated by the fact that PWS often use audible interjections (e.g., ‘‘and”, ‘‘like”, ‘‘although”, ‘‘um”) at the beginning of phrases and sentences and just prior to words in which they anticipate stuttering on (Bloodstein and Bernstein-Ratner, 2007). Using silent rehearsal strategies on the interjected words might have a positive effect on one’s overt fluency; however, this notion should be taken with caution, because use of interjections

D. Hudock et al. / Speech Communication 75 (2015) 76–83

and other avoidance-like behaviors can quickly become learned habits that may be overused and become difficult to stop. As with the use of other motor-based strategies, using silent rehearsal strategies multiple times throughout utterances would likely exhibit a positive cumulative effect on fluency. Lastly, P strategies with silent articulatory productions may also be similar to other motor-based modification techniques such as light articulatory contacts and preparatory sets. Light articulatory contacts occur by reducing the force and proximity of the articulators during speech production (e.g., approximating the lips for a bilabial /m/ production, but not forming complete closure). Preparatory sets occur when the speaker plans and executes a phoneme prolongation (e.g., purposefully expands the production the /m/ sound when saying the sentence: ‘‘Mmmmy, it’s a lovely day”). In fact, most behavioral speech modification strategies for PWS are motor based and likely act on similar neruomotor mechanisms as discussed in the manuscript. Results from the current study support the notion that stuttering may be a result of interference in feed-forward processing of efferent plans for motor-speech (Bohland et al., 2014; Max et al., 2004). Anticipation of and subsequent alteration of the syllable should create changes in the trajectory of typical motor productions by using actual or imagined motor production. Modification of the motorplan should shift the original motor plan for speech production to a more fluent form by increasing the preparatory time and therefore accuracy of the feed-forward process. Using P speech should especially engage the feed-forward system, thus allowing a fluent-motor plan to replace the previous plan. Such internal representation of motor rehearsals during phrase onset should act to engage the basal ganglia circuitry earlier in planning motor actions and aligning sensory, feedback, consequences. Future studies should examine the effectiveness of similar and other motor-based strategies on monosyllabic and multisyllabic productions. Additionally, studies should investigate the use of audible interjections along with P strategies on interjections for fluency enhancement. References Alm, P.A., 2004. Stuttering and the basal ganglia circuits: a critical review of possible relations. J. Commun. Disord. 37 (4), 325–369. Alm, P.A., 2005. On the Causal Mechanisms of Stuttering. Lund University. Armson, J., Kalinowski, J., 1994. Interpreting results of the fluent speech paradigm in stuttering research difficulties in separating cause from effect. J. Speech, Lang., Hear. Res. 37 (1), 69–82. Armson, J., Stuart, A., 1998. Effect of extended exposure to frequencyaltered feedback on stuttering during reading and monologue. J. Speech, Lang., Hear. Res. 41, 479–490. Bloodstein, O., Bernstein-Ratner, N., 2007. A Handbook on Stuttering, sixth ed. Thomson Delmar Learning, Clifton Park, NY. Bohland, J.W., Myers, E.M., Kim, E., 2014. An informatics approach to integrating genetic and neurological data in speech and language neuroscience. Neuroinformatics 12 (1), 39–62. http://dx.doi.org/ 10.1007/s12021-013-9201-6.

83

Brenner, N.C., Perkins, W.H., Soderberg, G.A., 1972. The effect of rehearsal on frequency of stuttering. J. Speech, Lang., Hear. Res. 15 (3), 483–486. Civier, O., Bullock, D., Max, L., Guenther, F.H., 2013. Computational modeling of stuttering caused by impairments in a basal ganglia thalamo-cortical circuit involved in syllable selection and initiation. Brain Lang. 126 (3), 263–278. Cohen, J., Cohen, P., 1983. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, second ed. Erlbaum, Hillsdale, NJ. Commodore, R.W., 1980. Communication stress and stuttering frequency during normal, whispered and articulation-without phonation speech: a further study. Hum. Commun. 5 (143), 150. Commodore, R.W., Cooper, E.B., 1978. Communicative stress and stuttering frequency during normal, whispered, and articulationwithout-phonation speech modes. J. Fluency Disord. 3 (1), 1–12. Guenther, F., 1994. A Neutral Network Model of Speech Acquisition and Motor Equivalent Speech Production. Boston University Center for Adaptive Systems Technical Report, 1-24. Houde, J.F., Nagarajan, S.S., 2011. Speech production as state feedback control. Front. Hum. Neurosci. 5. Kalinowski, J., Stuart, A., Rastatter, M.P., Snyder, G., Dayalu, V., 2000. Inducement of fluent speech in persons who stutter via visual choral speech. Neurosci. Lett. 281 (2), 198–200. Max, L., Guenther, F., Gracco, V., Ghosh, S., Wallace, M., 2004. Unstable or insufficiently activated internal models and feedbackbiased motor control as sources of dysfluency: a theoretical model of stuttering. Contemp. Iss. Commun. Sci. Disord. 31, 105–122. McQuaide, A., Akram, H., Zrinzo, L., Tripoliti, E., 2014. A review of brain circuitries involved in stuttering. Front. Hum. Neurosci. 8, 884. http://dx.doi.org/10.3389/fnhum.2014.00884. Perkins, W., Rudas, J., Johnson, L., Bell, J., 1976. Stuttering: discoordination of phonation with articulation and respiration. J. Speech Hear. Res. 19, 509–522. Rami, M.K., Kalinowski, J., Rastatter, M.P., Holbert, D., Allen, M., 2005. Choral reading with filtered speech: effect on stuttering. Percept. Mot. Skills 100 (2), 421–431. Redgrave, P., Rodriguez, M., Smith, Y., Rodriguez-Oroz, M.C., Lehericy, S., Bergman, H., et al., 2010. Goal-directed and habitual control in the basal ganglia: implications for Parkinson’s disease. Nat. Rev. Neurosci. 11 (11), 760–772. http://dx.doi.org/10.1038/nrn2915. Salmelin, R., Schnitzler, A., Schmitz, F., Freund, H.J., 2000. Single word reading in developmental stutterers and fluent speakers. Brain 123 (6), 1184–1202. Seidler, R.D., Benson, B.L., Boyden, N.B., Kwak, Y., 2014. Motor skill learning. In: Ochsner, Kevin N., Kosslyn, Stephen M. (Eds.), The Oxford Handbook of Cognitive Neuroscience, . In: Core Topics, vol. 1. Oxford University Press, New York, NY, US, pp. 416–435 (xvi, 621). Shepherd, G.M., 2013. Corticostriatal connectivity and its role in disease. Nat. Rev. Neurosci. 14 (4), 278–291. http://dx.doi.org/10.1038/ nrn3469. Sommer, M., Koch, M.A., Paulus, W., Weiller, C., Bu¨chel, C., 2002. Disconnection of speech-relevant brain areas in persistent developmental stuttering. The Lancet 360 (9330), 380–383. Van Riper, C., 1971. The Nature of Stuttering. Prentice Hall, Englewood Cliffs, New Jersey. Venkatagiri, H.S., 2004. Slower and incomplete retrieval of speech motor plans is the proximal source of stuttering: stutters occur when syllable motor plans stored in memory are concatenated to produce the utterance motor plan. Med. Hypotheses 62 (3), 401–405. Viera, A.J., Garrett, J.M., 2005. Understanding interobserver agreement: the kappa statistic. Fam. Med. 37, 360–363. Watkins, K.E., Smith, S.M., Davis, S., Howell, P., 2008. Structural and functional abnormalities of the motor system in developmental stuttering. Brain 131 (1), 50–59. Wu, J.C., Maguire, G., Riley, G., Fallon, J., LaCasse, L., Chin, S., et al., 1996. A positron emission tomography [18F] deoxyglucose study of developmental stuttering. NeuroReport 6 (3), 501–505.