Neural systems mediating processing of sound units of language distinguish recovery versus persistence in stuttering
© Mohan and Weber. 2015
Received: 27 August 2014
Accepted: 6 August 2015
Published: 18 August 2015
Developmental stuttering is a multi-factorial disorder. Measures of neural activity while children processed the phonological (language sound unit) properties of words have revealed neurodevelopmental differences between fluent children and those who stutter. However, there is limited evidence to show whether the neural bases of phonological processing can be used to identify stuttering recovery status. As an initial step, we aimed to determine if differences in neural activity during phonological processing could aid in distinguishing children who had recovered from stuttering and those whose stuttering persisted.
We examined neural activity mediating phonological processing in forty-three 7-8 year old children. Groups included children who had recovered from stuttering (CWS-Rec), those whose stuttering persisted (CWS-Per), and children who did not stutter (CWNS). All children demonstrated normal non-verbal intelligence and language skills. Electroencephalograms were recorded as the children listened to pairs of pseudo-words (primes-targets) that either rhymed or did not. Behavioral rhyme judgments along with peak latency and mean amplitude of the N400s elicited by prime and target stimuli were examined.
All the groups were very accurate in their rhyme judgments and displayed a typical ERP rhyme effect, characterized by increased N400 amplitudes over central parietal sites for nonrhyming targets compared to rhyming targets. However, over anterior electrode sites, an earlier onset of the N400 for rhyming compared to non-rhyming targets, indexing phonological segmentation and rehearsal, was observed in the CWNS and CWS-Rec groups. This effect occurred bilaterally for the CWNS, was greater over the right hemisphere in the CWS-Rec, and was absent in the CWS-Per.
These results are the first to show that differences in ERPs reflecting phonological processing are marked by atypical lateralization in childhood even after stuttering recovery and more pronounced atypical neural patterns for the children whose stuttering persisted. Despite comparable language and phonological skills as revealed by standardized tests, the neural activity mediating phonological segmentation and rehearsal differentiated 7-8 year old children whose stuttering persisted from those who had recovered from stuttering and typically developing peers.
Approximately 5–8 % of preschool children display stuttering-like disfluencies characterized by sound and syllable repetitions, prolongations of speech sounds, dysrhythmic phonations, and silent blocks . While up to 80 % of the children who stutter eventually recover, we are unable to predict with a reasonable degree of certainty which children are at high risk for developing persistent stuttering and may benefit most from early intervention. Estimates of stuttering recovery depend, to some extent, on the age of the children at initial assessment. For example, recovery rates are estimated to be approximately 80 % when sampling children near the typical onset of stuttering, i.e., at ages 2–3 years [1, 2], and these fall to approximately 50 % for 4-year-old children .
With considerable incidence of stuttering in preschoolers, resources are not available to treat every child who begins to stutter. Often, treatment is delayed for those children who ultimately persist in stuttering. For these at-risk children, early intervention is beneficial for improving fluency as well as for providing educational and emotional support for the child and family [2, 3]. The current limitations in reliably predicting eventual recovery or persistence in stuttering also has implications for research: studies of young children who stutter may contain confounds or increased variability due to the eventual recovery or persistence status of the participants.
In order to address the limitations in early identification of at-risk children, recent studies have focused on identifying behavioral factors that predict the course of stuttering [4–6]. While these studies have diagnostic advantages, stuttering is a highly heterogeneous, neurodevelopmental disorder [7, 8], resulting from atypical interaction of neural regions and connections that support speech and language processes [9, 10]. Although the diagnostic value of neurophysiological measures is currently limited, it may be highly relevant for predicting stuttering chronicity in the future. Therefore, a crucial step towards predicting stuttering persistence and recovery using neurophysiological measures is to evaluate the neurodevelopmental differences between children whose stuttering persists and those who have already recovered from stuttering. Neuroimaging techniques  and electrophysiological correlates of language processing  have been used to characterize neurodevelopmental differences between stuttering recovery and persistence. In the current study, we compare neurophysiological functions underlying one aspect of language, the processing of sound units of language or phonological processing, for three groups of young school-age children: children who have recovered from stuttering (CWS-Rec), children whose stuttering persists (CWS-Per), and children who did not stutter (CWNS).
Language abilities in children who stutter
Over recent years, a coherent picture of the essential characteristics of developmental stuttering has emerged, laying the groundwork for identifying the factors that may play a role in recovery or persistence. Evidence from a multitude of studies utilizing a variety of methodological approaches indicates that the development of stuttering involves a variety of factors in the domains of speech motor control, language, and emotion/temperament (for review, please see [1, 2, 7, 13, 14]). For example, a meta-analysis of 22 studies revealed that children who stutter (CWS) exhibited lower mean scores on tests of receptive and expressive vocabulary as well as lower mean length of utterance as compared to their fluent peers, even though they scored within normal limits . A growing body of literature has emphasized neurodevelopmental differences in children who stutter compared to their typically developing peers, including differences in grey matter volumes in speech-related regions, white matter connectivity for speech motor control, and neural correlates of phonological processing [9, 11, 16–21].
It has been reported that phonological disorders occur at a higher rate among the stuttering population relative to the general population [22, 23]. To investigate the phonological proficiency in CWS, researchers have used a number of behavioral tasks, including priming, rhyme judgment, and non-word repetition [21, 24–28]. For example, findings of lower rhyme judgment accuracy  and lower accuracy on non-word repetition tests among CWS relative to CWNS [26, 28] indicate an association between phonological proficiency and stuttering. Providing further evidence for the role of phonology in stuttering, Spencer and Weber-Fox  found that performance on the Bankson and Bernthal Test of Phonology—consonant inventory subtest (BBTOP-CI; ) as well as the Dollaghan and Campbell non-word repetition test  was predictive of stuttering persistence in preschool children.
Neural correlates of phonological processing in children who stutter
Standardized tests of phonology, behavioral measures of phonological encoding and phonological working memory often distinguish CWS and their fluent peers [24, 26–28, 31, 32]. Recognizing the neurodevelopmental nature of stuttering, a few researchers explored the underlying neurophysiological mechanisms of phonological processing. Sato and colleagues  used near-infrared spectroscopy (NIRs) to assess cortical auditory speech processing using phonemic and prosodic contrast syllables in adults and children (3–5 years and 6–12 years) who stutter and age-matched controls. The age-matched controls (both adults and children) exhibited increased concentration of hemoglobin in the left hemisphere for phonemic contrasts and in the right hemisphere for prosodic contrasts. The adults and children who stutter, however, did not show a left-lateralized response for the phonemic contrasts, suggesting atypical auditory functional lateralization for phonological processing.
Electrophysiological measures have also provided valuable information about the neurodevelopmental differences in phonological processing in school-age CWS. Weber-Fox et al.  compared 9–13-year-old CWS and CWNS on a visual rhyme judgment task, using real-word prime-target pairs. Peak latency of the N400 elicited by target words was earlier over the left hemisphere for CWNS, while it was earlier over the right hemisphere in CWS. This indicated that the relative timing of hemispheric neural processing for the rhyming/non-rhyming stimuli distinguished CWS from CWNS.
Researchers comparing phonological processing in both older school-age children  and adults who stutter  with their fluent peers showed that for the same task, behavioral and electrophysiological responses were qualitatively different in adults and children who stutter. In a visual rhyme judgment task where orthographic and phonologic congruence was altered, adults who stutter did not differ from their fluent peers in overall accuracy or reaction time for rhyme judgments  but 9–12-year-old CWS displayed lower accuracy as compared to their typically fluent peers . Furthermore, compared to CWNS, the CWS displayed reduced mean amplitude of the contingent negative cariation (CNV), an event-related potential (ERP) component that has been thought to index silent rehearsal of the prime [21, 34]. Another interpretation from ERP studies of rhyming in children indicate that the CNV may index resource allocation for phonological processing of both written words  and auditorily presented non-words .The CNV and N400 elicited for rhyming in adults who stutter were similar to those of their fluent peers. The difference, however, was that the adults who stutter displayed a greater right hemisphere asymmetry of the N400 elicited by the rhyme/non-rhyme targets . Taken together, the two studies indicate that factors such as maturation and experience with stuttering are likely to affect neural functions during the course of development. Because of this, studying phonological processing in younger CWS closer to the age of onset will provide needed information regarding how or if neural networks related to phonological processing play a role during earlier stages of development of the disorder.
Recovery and persistence of stuttering
Phonological proficiency has also been associated with eventual recovery or persistence in stuttering [1, 2, 4–6, 37]. Converging evidence from a variety of studies indicates that stuttering persistence, on average, is associated with lower phonological and articulation proficiencies compared to those of preschool children who would eventually recover [4, 6]. However, authors from these studies note that there were wide ranges of scores within the groups as well as significant overlap of scores between groups and suggest that additional factors are necessary for reliable prediction. The large variability in behavioral scores across tests of expressive and receptive language  as well as tests of phonology [4–6] indicates that differences between children who persist and those recovered from stuttering may be too subtle to be uniformly characterized by standardized tests . Characterizing crucial neurophysiological processes in CWS-Rec and CWS-Per will shed light on possible underlying neurodevelopmental differences between stuttering recovery and persistence.
To date, only one study has assessed potential neuroanatomical differences in speech-related areas among children who had recovered from stuttering and those who persisted. Chang et al.  utilized measures of fractional anisotropy to study brain structure differences in CWNS, CWS-Rec, and CWS-Per between the ages of 9 and 12 years. Their findings indicated that white matter tract integrity in left hemisphere speech areas, including the arcuate fasciculus, was reduced in CWS-Per. The arcuate fasciculus connects regions involved in phonological operations such as rhyming  and measures of diffusion in the arcuate fasciculus have also been correlated with phonological awareness skills in children 7–11 years old . CWS-Rec, on the other hand, exhibited normal fractional anisotropy in the left hemisphere but a trend towards greater right fractional anisotropy compared to the other two groups. Chang and colleagues concluded that the reduction in white matter tract integrity in regions involved in the speech production network may indicate a risk for persistent stuttering and suggested that differences in neural connectivity may differentiate children who had recovered from those whose stuttering persisted.
The current study
In the current study, we focus on how underlying brain functions related to phonological processing may be associated with the recovery or persistence of stuttering. We compared the ERPs elicited in a rhyming paradigm for 7–8-year-old CWNS, CWS-Rec, and CWS-Per. Such rhyme judgments are known to rely on phonological awareness [40, 41].
Rhyming paradigms employing ERP measures, such as in our study, have elicited a reliable ERP index, the N400 [21, 35, 36, 42]. An increase in amplitude of the negative peak at approximately 400-ms post-stimulus onset (N400) for non-rhyming words relative to rhyming words has been observed across rhyming studies. The difference between the N400 elicited by the non-rhyming and rhyming words is called the “rhyme effect.” The rhyme effect is a broadly distributed negative component observed over bilateral central-parietal electrode sites . It is considered to be a part of the family of N400-like potentials that are sensitive to contextual information and underlies the cognitive processes mediating comparison of phonological representation of words . Coch et al.  also reported a reversal of the rhyme effect at bilateral anterior electrode sites, the anterior reverse rhyme effect, signaled by a larger amplitude N400 elicited by rhyming targets compared to non-rhyming targets. Spatiotemporal analysis in adults and children performing visual rhyme judgment suggests that the anterior reverse rhyme effect may be specific to the rhyming stimuli and is thought to reflect a cognitive process separate from that elicited by the non-rhyming stimuli at central-parietal sites . We hypothesized that CWNS, CWS-Rec, and CWS-Per would elicit the classic central-parietal rhyme effect, based on previous findings that the N400 for rhyme target processing did not distinguish CWNS and CWS in a more cognitively challenging rhyming task than the one used in the present study . Due to limited discussion regarding the anterior reverse rhyme effect in previous work, we could not make specific predictions about how the distribution of the N400 in anterior electrode sites might distinguish the groups. But based on the findings from Chang et al.  and Weber-Fox et al. , we predicted that there would be differences in hemispheric contributions for rhyme processing between the CWS-Per and the other two groups.
Subject characteristics of CWNS, CWS-Rec, and CWS-Per and standard scores on cognitive and language tests
(n = 22)
(n = 12)
(n = 9)
Of the 21 children originally classified as CWS, 12 had recovered from stuttering (CWS-Rec) at the time of testing for the current study. Recovery was defined as receiving a rating of 0 or 1 by both parent and clinician on the eight-point stuttering severity measure and displaying fewer than three SLDs per 100 syllables during spontaneous speech. The remaining nine CWS persisted in stuttering (CWS-Per) at the time of testing. The three groups, CWNS, CWS-Rec, and CWS-Per, did not differ in age, F (2, 40) = 1.52, p = 0.24 or socio-economic status as defined by mother’s highest level of education ; F (2, 40) = 0.85, p = 0.44).
All the participants were native English speakers with no reported neurological, language, reading, visual, or hearing impairments. Their hearing was screened at 250, 500, 1000, 2000, 4000, 6000, and 8000 Hz at 20dBHL and found to be bilaterally normal. None of the children displayed impairments of social interaction or restriction of activities as indexed by the Childhood Autism Rating Scale .
All the children were also tested on a comprehensive battery to measure non-verbal intelligence and language and phonological abilities. The behavioral test protocol included measurement of non-verbal intelligence using the Columbia Mental Maturity Scale (CMMS) , the Test for Auditory Comprehension of Language—third edition (TACL-3)  to assess language comprehension, the Structured Photographic Expressive Language Test—third edition (SPELT-3)  to examine expressive language abilities including use of morphemes, and the Bankson and Bernthal Test of Phonology—consonant inventory subtest (BBTOP-CI)  to measure articulation abilities. All the children displayed normal non-verbal intelligence (CMMS > 94) and normal receptive language (TACL-3 > 91). Standard scores for expressive language (SPELT-3) and phonological abilities (BBTOP-CI) were all within one standard deviation of the mean (see Table 1 for a summary of performance for the three groups).
Stimuli for rhyme judgment task
The pseudo-word auditory stimuli for eliciting ERPs were taken from the rhyming study by Coch and colleagues . These included 44 pairs of rhyming pseudo-words (e.g., feap-neap, trum-pum) and 44 pairs of non-rhyming pseudo-words (e.g., bry-pag, mag-yare). Each pseudo-word followed phonological rules of English and had no semantic content. The first pseudo-word of the pair was the “prime” and the second pseudo-word was the “target.” The prime-target pairs in the non-rhyming list were made by associating the prime of one rhyming pair with the target of another rhyming pair. As in Coch et al. , there were two lists of pseudo-word pairs. A pseudo-word occurred as a target only in one list. The two lists were counterbalanced across participants. A female, native American-English speaker from the authors’ lab produced the pseudo-word stimuli, which were digitized using PRAAT  software at a rate of 22.5 kHz. The average pseudo-word duration was 582.4 ms (SD = 72.1).
An elastic electrode cap (Quick-cap) containing 32 Ag-Cl electrodes was used to record electrical activity from the scalp. The cap was fitted snugly over the head with the electrodes positioned in homologous locations over the left and right hemispheres, consistent with the international 10-10 system (American Electroencephalographic Society, 1994). The electrode channels included lateral sites F7/F8, FT7/FT8, TP7/TP8, P7/P8; medial sites FP1/FP2, F3/F4, FC3/FC4, CP3/CP4, P3/P4, O1/O2; and midline sites FZ, FCZ, CZ, CPZ, PZ, and OZ. Electrodes were also placed on the left superior and inferior orbital ridge to monitor vertical eye movements. Horizontal movements were monitored by electrodes on the left and right outer canthi. Electrodes on left and right mastoids served as linked references. The electroencephalogram (EEG) signals were amplified using a band pass filter between 0.1 and 100 Hz and digitized online at the rate of 500 samples per second.
Now you will listen to pretend words. Sometimes the words will rhyme, like “zoof” and “noof”. Sometimes they won’t rhyme, like “jat” and “misk”. Listen to the pairs of words and try to tell if they rhyme or if they don’t rhyme. It is important that you sit very still. While you are listening to the pretend words, look at the mark on the screen in front of you. If you think the words rhyme, press the green button and if they don’t, press the red button. Every once in a while you will see a picture pop up. This is the time when you can move! Every time you see a picture, you can play a turn of our game. When you finish the game, you will get a surprise! We will start now. Remember to listen carefully to the words that rhyme and don’t rhyme. Let’s practice listening to a few now. Are you ready?
The child sat in a sound-treated booth 160 cm away from a 47.5-cm computer monitor. A researcher was present in the booth throughout the experiment to provide support and reinforcement. The stimuli were presented using Presentation® Version 14.9 via a speaker placed directly above the monitor at a level of 70–75 dB SPL. The child was asked to start the presentation of stimuli by pressing any button on a response pad. After a delay of 1080 ms, the prime was presented, followed by an inter-stimulus interval of 1070 ms and then the target. The child was required to make rhyming judgments using the response pad. The use of the left and right buttons for the “yes” and “no” responses were counterbalanced across subjects.
The stimuli were presented in 11 blocks of eight pairs each, with breaks in between. The breaks were indicated to the child by a picture on the computer monitor. During these breaks, a single turn of games such as “Connect 4,” “Tic Tac Toe,” or “Guess Who” was played by the child and researcher.
EEG data analysis
The three groups, CWNS, CWS-Rec, and CWS-Per, were compared for measures of neural activity in the rhyme judgment task. The data were analyzed using EEGLAB and ERPLAB , toolboxes of MATLAB®, (MathWorks, Natick, MA, USA). Only ERP data corresponding to correct responses were analyzed. Eye artifacts were removed from the EEG signals using independent component analysis (ICA; EEGLAB). ICA is an EEGLAB tool that separates independent sources of EEG signals. Artifacts to be removed such as blinks, horizontal eye movements and voltage drifts were identified independently by two trained research assistants. Any discrepancies were resolved by a third independent research assistant. Such components were removed from the continuous EEG data. The EEG signals were low-pass filtered at 30 Hz with a 12-dB roll-off to eliminate high-frequency noise. The continuous EEG records were epoched from 100 ms prior to stimulus onset and continued until 1000 ms post-stimulus for purposes of averaging and ERP component measures. Automatic voltage-dependent artifact removal was then performed on all the EEG channels. The EEG epochs were averaged for the ERPs elicited by the prime, rhyming targets, and non-rhyming targets. The average percentage of usable trials following artifact rejection across the conditions was 80.08 % (SD 9.46) for the primes, 82.27 % (SD 9.7) for the rhyme targets, and 82.05 % (SD 9.99) for the non-rhyme targets. There was no significant difference between groups in the number of trials rejected for any of the conditions, F (2, 31) = 0.97, p = 0.38.
Percent accuracy of response in judging rhyming from non-rhyming pairs was obtained from the button press responses and a repeated measures ANOVA was used to compare the percentage correct responses for the three groups (CWNS, CWS-Rec, and CWS-Per) across two conditions (rhyme and non-rhyme).
ERP measurements considered for statistical comparison were the peak latencies and mean amplitudes of the N400 ERP component, consistent with previous studies . The peak latency and mean amplitude of the N400 elicited by the prime (N400P) and the N400 elicited by the target stimuli (N400T) were measured within a 450–700-ms time window. From visual inspection of the grand average ERPs, differences in the onset of the N400T elicited by the rhyming and non-rhyming targets were apparent, particularly over anterior electrode sites. Therefore, the mean amplitude of N400T was also measured within an earlier temporal window of 300–500 ms to capture the onset interval of the component.
Similar to previous ERP studies, e.g., [12, 20, 21, 33, 35, 42], mixed-effects repeated measures analysis of variance (ANOVA) was used for the ERP measures separately for the primes and targets, which included a between-subject factor of group (CWNS, CWS-Rec, and CWS-Per). Within-subject factors for primes were hemisphere (left, right), anterior-posterior (AP) distribution (anterior: frontal and frontal-central; posterior: central, central-parietal, parietal), and laterality (lateral, medial). Within-subject factors for targets included condition (rhyme targets, non-rhyme targets), hemisphere, AP distribution, and laterality. For analysis of the midline electrode sites, a separate mixed-effects repeated measures ANOVA was calculated for primes by AP distribution (anterior: FZ and FCZ; posterior: CZ, CPZ, and PZ) and for targets with condition and AP distribution as within-subject factors. Hyunh-Feldt (H-F) adjusted p values were used to determine significance with a criterion of p < 0.05. The effect sizes determining strength of associations, indexed by the partial eta squared statistic (ƞ 2 p ), are reported for all significant effects. Step-down repeated measures ANOVAs were performed to further explore significant interactions. All the EEG recording, ERP analyses, and statistical measures were consistent with the guidelines specified by Picton et al. .
Response judgment accuracy
Mean rhyme judgment accuracy for CWNS, CWS-Rec, and CWS-Per
ERPs elicited by prime pseudo-words
N400P mean amplitude
The overall mean amplitudes of the N400P did not differentiate groups, F (2, 40) = 0.356, H-F p = 0.70, nor were there any interactions with group, F (8, 160) < 1, H-F p > 0.42. A significant anterior-posterior effect at the lateral/medial sites, F (4, 160) = 29.48, H-F p < 0.001, ƞ 2 p = 0.42 and midlines, F (5, 200) = 25.73, H-F p < 0.001, ƞ 2 p = 0.391, along with a significant anterior-posterior × laterality effect, F (4, 160) = 17.72, H-F p = 0.01, ƞ 2 p = 0.09, revealed that the mean amplitude of the N400P elicited by the primes was largest primarily at the anterior medial sites.
N400P peak latency
Based on the distribution of N400P, the peak latency of the N400P was measured at six anterior electrode sites: F3/4, FZ/FCZ, and FC3/4. The peak latency measure of the N400P did not differentiate groups, F (2, 40) = 0.86, H-F p = 0.43, nor were there any group interactions, F (10, 200) < 1, H-F p > 0.40.
ERPs elicited by target pseudo-words
Onset interval of N400T
Mixed-model results for CWNS (88 observations for 22 participants)
Condition (rhyming, non-rhyming)
Electrode (frontal, frontal-temporal)
95 % CI
95 % CI
Mixed-model results for CWS-Rec (48 observations for 12 participants)
Condition (rhyming, non-rhyming)
Electrode (frontal, frontal-temporal)
95 % CI
95 % CI
Mixed-model results for CWS-Per (36 observations for 9 participants)
Condition (rhyming, non-rhyming)
Electrode (frontal, frontal-temporal)
95 % CI
95 % CI
Mean estimates were the same using the mixed-effects approach as with the repeated measures ANOVA approach. However, the p values differed slightly due to the fact that the mixed model did not make strict assumptions about the homogeneity of variance, compound symmetry, or sphericity because the covariance structure was explicitly modeled. Using this approach, our results were replicated, except that p values were lower for all the tests. The result of the mixed-effects model indicated a difference between experimental conditions for the CWS-Rec group over the right hemisphere with the p < 0.001.
As displayed in Table 1, the BBTOP-CI revealed higher scores for the CWNS as compared to the other two groups. Because the CWS-Rec and CWS-Per performed similarly, it is unlikely that scores on the BBTOP-CI influenced our main finding that the anterior onset rhyme effect distinguished CWS-Per from the other two groups. In order to confirm this assumption, a Pearson correlation was performed to assess whether the differences between groups for the mean amplitude onset interval N400T at the anterior sites were associated with differences in the BBTOP-CI scores. The BBTOP-CI scores for the participants were not correlated with the mean amplitude onset interval of the N400T for the left hemisphere, r = −2.16, p = 0.16, or for the right hemisphere, r = 0.03, p = 0.84.
N400T mean amplitude
The N400T mean amplitude for the non-rhyming targets was more negative as compared to the N400T amplitude for the rhyming targets over central-parietal electrode sites (CP3, CPZ, CP4, P3, PZ, P4; classic rhyme effect) for all the three groups, F (4, 160) = 4.89, H-F p < 0.01, ƞ 2 p = 0.16 (lower panel of Fig. 1). There were no interactions with condition, F (1, 40) < 2.33, H-F p > 0.13 or group interactions, F (8, 160) < 1, H-F p > 0.58.
N400T peak latency
There were no condition effects across lateral/medial electrode sites, F (1, 40) < 3.51, H-F p > 0.07 or across midlines, F (1, 40) < 1, H-F p > 0.35, nor were there any significant interactions with group at lateral/medial sites, F (8, 160) < 1, H-F p > 0.48 or midlines, F (10, 200) < 1, H-F p > 0.34.
Using a rhyming paradigm, we asked whether the neural indices of phonological processing would differentiate 7–8-year-old typically fluent children, those who had recovered from stuttering and those whose stuttering persisted. Results indicated that although all three groups made accurate overt rhyme judgments, there were differences in the underlying neural activity that mediated the processing of the rhyming and non-rhyming targets, as evidenced by the ERPs. All three groups demonstrated the classic central-parietal rhyme effect. However, the anterior onset rhyme effect was absent in the children with persistent stuttering. Further, the distribution of the anterior effect, while bilateral in the CWNS group, was most pronounced over the right hemisphere for the CWS-Rec.
The use of strict inclusionary criteria ensured that the groups were matched for age, gender distribution, handedness, socio-economic status, non-verbal IQ (CMMS), language comprehension (TACL-3), and language expression (SPELT-3). The CWS-Rec and CWS-Per performed similarly on the test of articulation abilities, the BBTOP-CI. But the CWNS had scores that were slightly higher than the CWS on this test. Follow-up analyses, however, revealed that the differences in BBTOP-CI scores across groups did not account for the group differences in the mean amplitudes of the onset interval of the N400 elicited by the targets.
Children in all three groups were more accurate in making non-rhyme decisions compared to rhyme decisions, consistent with results from Coch et al. . Although the typically fluent children performed with slightly greater accuracy in the rhyme judgment task, it was evident that the children in all three groups were highly proficient in making rhyme judgments. In a previous study, CWS and CWNS in this age range also showed high proficiency for other phonological tasks such as rhyme monitoring . However, Weber-Fox et al.  found that older school-age CWS (9–13 years) had significantly lower rhyme judgment accuracy compared to the typically fluent children in a visual rhyming task that involved orthographic constraints. The additional demand of orthographic incongruence as well as the requirement to respond as quickly as possible likely enhanced the differences in behavioral accuracy between the groups. The task in our study was significantly different—it was in the auditory modality, relatively simple, and resulted in significant but smaller differences in rhyme judgment accuracy between the CWS (Rec and Per) and CWNS. It is important to note, however, that the CWS-Rec and CWS-Per did not differ in their behavioral accuracy, so this cannot account for any differences in the neural indices of phonological processing observed between these two groups.
N400T rhyme effect
The well-known central-parietal rhyme effect, indexed by larger N400T mean amplitude elicited by non-rhyming stimuli relative to rhyming stimuli, was observed in all three groups as predicted. Weber-Fox et al.  also observed a rhyme effect in their visual rhyming study in 9–13 year old CWS and fluent peers. The rhyme effect is thought to index neural processes mediating the comparison of phonological representations of words [34, 35]. Thus, in the current study, it seems likely that the neural functions that mediate the comparison and ease of integration of phonological properties of the prime and target were similar for CWNS, CWS-Rec, and CWS-Per.
N400T onset interval
The anterior effect observed in our study refers to a measure of the mean amplitude of the onset interval of the N400T measured in the 300–500 ms temporal window. In earlier studies, the rhyme and reverse rhyme effects were elicited concurrently [36, 42], and the anterior effect was considered to be a “reversal” of the central-parietal rhyme effect. We believe that this interpretation may be inadequate. The anterior onset rhyme effect and central-parietal rhyme effect may function independently to some degree. Findings from Khateb and colleagues  indicated that adults’ ERPs elicited by the rhyming condition relative to non-rhyming condition started before 300 ms post-stimulus onset and peaked at 350 ms at left frontal-temporal sites, while the central-parietal rhyme effect (non-rhyme > rhyme) maximum amplitude occurred at 400 ms. Khateb and colleagues suggested that the anterior reverse effect and the central-parietal rhyme effect involve distinct neural processes. Our findings indicated group differences for the anterior, but not central-parietal rhyme effect, providing further evidence that these component indices may be distinct. Taken together, with earlier findings , our results suggest that the N400T onset over anterior lateral electrodes may index an earlier phonological processing stage separate from the phonological integration stage indexed by the N400T elicited over central-parietal sites.
The rhyme judgment task using prime-target pseudo-word pairs utilized in the present study involved automatic processes including segmenting the prime into its rime component, holding it briefly in verbal working memory, and then comparing the rime with the upcoming target. This task involved both phonological segmentation and rehearsal, functions that have been found to involve Broca’s area and the precentral gyrus . In such a task, the phonologically related prime pre-activates some of the targets’ segments ([36, 57], Expt 2) and hence, target processing in the rhyming condition occurs with greater efficiency. While efficient processing of rhyming targets is represented as smaller amplitude N400T as compared to non-rhyming targets over central-parietal electrode sites during phonological integration, a different pattern was exhibited over anterior lateral sites. We speculate that over anterior lateral sites, prime facilitation and subsequently, more efficient processing of rhyming targets may have been indexed by earlier onset of N400T for rhyming relative to non-rhyming targets.
Comparable peak latencies of the N400 elicited by the primes indicated that processing of the phonological components of the primes was similar across all three groups. However, the anterior onset rhyme effect distinguished the groups, as demonstrated by the significant group × condition × hemisphere interaction. Based on converging evidence from three prior studies [36, 42, 43] and the current findings for the anterior rhyme effect, we interpreted the presence of an anterior onset rhyme effect in the CWNS and CWS-Rec groups as reflecting a facilitation for processing the rhyming targets relative to the non-rhyming targets. These findings suggest that the representation of the prime for the CWNS and CWS-Rec groups was likely more salient and robust in these groups compared to the CWS-Per. For the CWS-Per, it may be that the prime did not facilitate phonological access to the rhyming targets in this early window, exhibited by the absence of the anterior onset rhyme effect. However, a robust central-parietal rhyme effect observed in all the groups indicated sensitivity to phonological mismatch of the rime in the non-rhyming condition. Thus, the rhyming task in the present study elicited group differences in the stage of neural processing involving phonological segmentation and rehearsal but not at the stage indexing phonological integration.
Additionally, there were hemispheric differences between the groups for processing the rhyming target at the anterior electrode sites, as verified by the mixed-effects model. The CWNS displayed prime facilitation for processing the rhyming targets equally in both hemispheres, consistent with the finding of bilateral anterior rhyme effect observed in typically developing 7- and 8-year-olds in Coch et al. , while the CWS-Rec did not show this “fluent” pattern. The anterior onset rhyme effect was more predominant over the right hemisphere for the CWS-Rec. Despite the reduced power of the effect, the larger amplitude of the ERPs elicited by the rhyming compared to the non-rhyming targets in the onset interval over the right hemisphere was seen in nine of the 12 CWS-Rec and indicated that for this particular stage of phonological processing, many CWS-Rec recruited hemispheric functions differently as compared to the CWNS. To our knowledge, this is the first study that examines neural activation during phonological processing in children who have recovered from stuttering. Using neuroanatomical imaging, Chang et al.  indicated a non-significant trend towards greater white matter fractional anisotropy in certain regions of the brain, including the right rolandic operculum and right supramarginal gyrus in the recovered group. The left and right supramarginal gyri have been shown to be involved in phonological processing, as observed using transcranial magnetic stimulation . However, the link between our electrophysiological findings and the structural imaging data from Chang et al.  is unclear, given that EEGs have poor spatial resolution and identifying the location and orientation of neural generators is difficult. Therefore, more neurophysiological evidence is required to evaluate the role of the right hemisphere in the neurodevelopmental course of stuttering recovery.
In contrast to the CWNS and CWS-Rec, the CWS-Per group did not show an anterior onset rhyme effect—an increased amplitude N400T elicited by rhyming target relative to non-rhyming target at anterior sites—in either hemisphere. They did, however, exhibit a reversal of the expected pattern over the left hemisphere. While the significance of the reversed pattern is difficult to interpret at this point in time, the absence of prime facilitation at the anterior electrode sites for the CWS-Per suggests a difference in phonological processing as compared to the other two groups. These findings argue against the domain-specific theories of stuttering in favor of the dynamic interactions model . Stuttering is not just a series of “stutter events”—linguistic, emotional, and motor processes interact in specific ways that influence the probability of speech motor breakdown. Phonological processing and encoding interact very closely with speech production processes. Atypical development of any of these processes could likely be related to breakdown in another domain. Although many children who stutter typically score within normal limits on phonological, semantic, or morphosyntactic tests, it may be conceivable that the presence of subtle communicative difficulties would tax the weaker or less stable speech production system, contributing to a greater likelihood of repeated or prolonged speech sounds. For example, recent ERP evidence indicates that young children (age 6–7 years) whose stuttering is persisting display less mature neural systems for processing of phrase structure violations when presented in Jabberwocky sentences compared to age-matched children who have recovered from stuttering and typically developing peers .
Our results indicate that despite comparable non-verbal intelligence, language skills, phonological skills, and behavioral rhyme judgment accuracy, neural activity mediating phonological processing necessary for rhyme judgments may distinguish CWS-Rec from CWS-Per. The children whose stuttering persisted did not show the anterior onset rhyme effect, indicating less efficient rime access and prime facilitation during the stage of phonological segmentation and rehearsal. While heterogeneity of neural correlates of language processing in CWS is typical, an absence of the anterior onset rhyme effect in 78 % (7/9) of the CWS-Per suggested that this may be a valuable indicator of the neurodevelopmental characteristic of stuttering persistence. Additionally, although the target stimuli elicited the anterior onset rhyme effect for the CWS-Rec and the CWNS, hemispheric differences in the neural indices of phonological processing suggested that the CWS-Rec did not process the rhyming target in the same way as the CWNS. The children who had recovered from stuttering showed a tendency for greater right hemisphere involvement during target processing as compared to the typically fluent children who showed bilateral processing. Therefore, the children who had recovered from stuttering still exhibited some traces of a history of stuttering during phonological processing, as evidenced by the electrophysiological measures. A crucial next step is to determine if these neurophysiological differences for phonological processing are evident earlier in development when all the children stutter. These findings also have important implications for how neurodevelopmental trajectories differ for children who stutter compared to typically fluent peers.
Efforts were taken to ensure that the groups were matched for various factors in order to eliminate potential confounding variables. However, some variables that were not controlled were time since stuttering onset, duration of recovery status, and the effect of stuttering therapy. Future studies should examine the possible role these variables may have on the differences in neural functions among children who have recovered from stuttering and those who persist.
In the current study, we only describe the neural indices of phonological processing in groups of children who have already recovered and whose stuttering persisted at the time of the study. It is possible that some of the children who were classified as CWS-Per will recover in the future. However, the likelihood of recovery reduces with increase in years since stuttering onset . Seven out of the nine children in the CWS-Per group had been diagnosed 3 years prior to testing, while two had been diagnosed 2 years prior. Therefore, given the children’s ages and time since stuttering onset, the chances of recovery among CWS-Per in our study are minimal. While we examined young school-aged children, longitudinal assessment of children nearer the age of stuttering onset may prove valuable in identifying an early, predictive neural marker of stuttering recovery and persistence.
The interpretation of the current ERP findings for the prime facilitation (anterior onset rhyme effect) would have been enhanced with reaction time measures. However, because it is difficult to get reliable reaction time measures from 7–8-year-old children, we did not instruct them to respond as quickly as possible during rhyme judgment. On the other hand, the ERP measures provide a means of examining phonological processing and possible priming in young children when behavioral responses may be unreliable in this population.
Our findings are the first to show that neural activity mediating phonological processing necessary for rhyme judgment may distinguish 7–8-year-old children who have recovered from those whose stuttering persists. Specifically, we suggest the onset interval of the anterior rhyme effect, which likely indexes a different, earlier stage of phonological processing (phonological segmentation and rehearsal) as compared to the central-parietal rhyme effect (phonological integration), is the stage of processing that distinguishes the neural patterns for the children who recovered from those whose stuttering persists. Further investigation in younger children who stutter is necessary to determine if the absence of anterior prime facilitation is an early marker that may help predict which children who stutter are likely to persist.
We thank Anne Smith for her valuable comments on this project and the manuscript. We also wish to thank Dr. Patricia Zebrowski and her research team at the University of Iowa for their contribution to the data collection. We are grateful to Barb Brown for her assistance in recruiting subjects and clinical testing at Purdue. We thank Janna Berlin and Amanda Hampton Wray for their expert role in collecting ERP data. We also greatly appreciate Dr. Sharon Christ for her help with the statistical analyses and interpretation. This work was funded by a grant from the National Institutes on Deafness and other Communication Disorders (DC00559) awarded to Anne Smith and Christine Weber.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Yairi E, Ambrose N. Epidemiology of stuttering: 21st century advances. J Fluency Disord. 2013;38:66–87.PubMed CentralView ArticlePubMedGoogle Scholar
- Yairi E, Ambrose NG. Early childhood stuttering: for clinicians by clinicians. Austin: Pro-Ed; 2005.Google Scholar
- Subramanian A, Yairi E, Amir O. Second formant transitions in fluent speech of persistent and recovered preschool children who stutter. J Commun Disord. 2003;36:59–75.View ArticlePubMedGoogle Scholar
- Spencer C, Weber-Fox C. Preschool speech articulation and nonword repetition abilities may help predict eventual recovery or persistence of stuttering. J Fluency Disord. 2014;41:32–46.View ArticlePubMedGoogle Scholar
- Yairi E, Ambrose NG, Paden E, Throneburg RN. Predictive factors of persistence and recovery: pathways of childhood stuttering. J Commun Disord. 1996;29:51–77.View ArticlePubMedGoogle Scholar
- Paden EP, Yairi E. Phonological characteristics of children whose stuttering persisted or recovered. J Speech Lang Hear Res. 1996;39:981–91.View ArticleGoogle Scholar
- Smith A. Stuttering: a unified approach to a multifactorial, dynamic disorder. In: Bernstein Ratner N, Healey EC, editors. Stuttering research and practice: bridging the gap. Mahwah: Erlbaum; 1999. p. 27–44.Google Scholar
- Ludlow CL, Loucks T. Stuttering: a dynamic motor control disorder. J Fluency Disord. 2003;28:273–95.View ArticlePubMedGoogle Scholar
- Beal DS, Gracco VL, Brettschneider J, Kroll RM, De Nil F. A voxel-based morphometry (VBM) analysis of regional grey and white matter volume abnormalities within the speech production network of children who stutter. Cortex. 2013;49 Suppl 8:2151–61.PubMed CentralView ArticlePubMedGoogle Scholar
- Kell CA, Neumann K, von Kriegstein K, Posenenske C, von Gudenberg AW, Euler H, et al. How the brain repairs stuttering. Brain. 2009;132:2747–60.View ArticlePubMedGoogle Scholar
- Chang SE, Erickson KI, Ambrose NG, Hasegawa-Johnson MA, Ludlow CL. Brain anatomy differences in childhood stuttering. Neuroimage. 2008;39:1333–44.PubMed CentralView ArticlePubMedGoogle Scholar
- Usler E, Weber-Fox C. Neurodevelopment for syntactic processing distinguishes childhood recovery versus persistence. J Neurodev Disord. 2014;7:4.View ArticleGoogle Scholar
- Bloodstein O, Bernstein Ratner N. A handbook on stuttering. 6th ed. New York: Thomson Delmar Learning; 2008.Google Scholar
- Guitar B. Stuttering: an integrated approach to its nature and treatment. 3rd ed. Baltimore: Lippincott, Williams, & Wilkins; 2006.Google Scholar
- Ntourou K, Conture EG, Lipsey MW. Language abilities of children who stutter: a meta-analytical review. Am J Speech Lang Pathol. 2011;20:163–79.PubMed CentralView ArticlePubMedGoogle Scholar
- Beal DS, Quraan MA, Cheyne DO, Taylor MJ, Gracco VL, De Nil LF. Speech-induced suppression of evoked auditory fields in children who stutter. Neuroimage. 2011;54 Suppl 4:2994–3003.PubMed CentralView ArticlePubMedGoogle Scholar
- Chang SE, Zhu DC. Neural connectivity differences in children who stutter. Brain. 2013;137 Suppl 4:1–18.Google Scholar
- Kaganovich N, Hampton A, Weber-Fox C. Non-linguistic auditory processing and working memory update in pre-school children who stutter: an electrophysiological study. Dev Neuropsychol. 2010;35(6):712–36.PubMed CentralView ArticlePubMedGoogle Scholar
- Sato Y, Mori K, Koizumi T, Minagawa-Kawai Y, Tanaka A, Ozawa E, et al. Functional lateralization of speech processing in adults and children who stutter. Front Psychol. 2011;2:70.PubMed CentralView ArticlePubMedGoogle Scholar
- Weber-Fox C, Hampton Wray A, Arnold H. Early childhood stuttering and electrophysiological indices of language processing. J Fluency Disord. 2013;38:206–21.PubMed CentralView ArticlePubMedGoogle Scholar
- Weber-Fox CM, Spuill III JE, Spencer R, Smith A. Atypical neural functions underlying phonological processing and silent rehearsal in children who stutter. Dev Sci. 2008;11:321–37.PubMed CentralView ArticlePubMedGoogle Scholar
- Paden E, Yairi E, Ambrose N. Early childhood stuttering II: initial status of phonological abilities. J Speech Lang Hear Res. 1999;42:1113–24.View ArticlePubMedGoogle Scholar
- Wolk L. Intervention strategies for children who exhibit coexisting phonological and fluency disorders: a clinical note. Child Lang Teach Ther. 1998;14:69–82.View ArticleGoogle Scholar
- Byrd CT, Conture EG, Ohde RN. Phonological priming in young children who stutter: holistic versus incremental processing. Am J Speech Lang Pathol. 2007;16:43–53.View ArticlePubMedGoogle Scholar
- Arnold H, Conture E, Ohde R. Influence of phonological density on picture naming responses of young children who stutter. J Fluency Disord. 2005;30:125–48.View ArticlePubMedGoogle Scholar
- Hakim HB, Bernstein Ratner N. Nonword repetition abilities of children who stutter: an exploratory study. J Fluency Disord. 2004;29:179–99.View ArticlePubMedGoogle Scholar
- Melnick K, Conture E, Ohde R. Phonological priming in picture-naming of young children who stutter. J Speech Lang Hear Res. 2003;46:1428–43.View ArticlePubMedGoogle Scholar
- Anderson JD, Wagovich SA, Hall NE. Nonword repetition skills in young children who do and do not stutter. J Fluency Disord. 2006;31 Suppl 3:177–99.PubMed CentralView ArticlePubMedGoogle Scholar
- Bankson NW, Bernthal JE. Bankson-Bernthal test of phonology. Austin: Pro-Ed; 1990.Google Scholar
- Dollaghan C, Campbell TF. Nonword repetition and child language impairment. J Speech Lang Hear Res. 1998;41:1136–46.View ArticlePubMedGoogle Scholar
- Anderson JD, Conture EG. Language abilities of children who stutter: a preliminary study. J Fluency Disord. 2000;25:283–304.View ArticleGoogle Scholar
- Sasisekaran J, De Nil FL, Smyth R, Johnson C. Phonological encoding in the silent speech of persons who stutter. J Fluency Disord. 2006;31:1–21.View ArticlePubMedGoogle Scholar
- Weber-Fox C, Spencer RMC, Spruill III JE, Smith A. Phonological processing in adults who stutter: electrophysiological and behavioral evidence. J Speech Lang Hear Res. 2004;47:1244–58.View ArticlePubMedGoogle Scholar
- Rugg MD. Event-related potentials in phonological matching tasks. Brain Lang. 1984;24:225–40.View ArticleGoogle Scholar
- Grossi G, Coch D, Coffey-Corina S, Holcomb PJ, Neville HJ. Phonological processing in visual rhyming: a developmental ERP study. J Cogn Neurosci. 2001;13:610–25.View ArticlePubMedGoogle Scholar
- Coch D, Grossi G, Skendzel W, Neville H. ERP nonword rhyming effects in children and adults. J Cogn Neurosci. 2005;17 Suppl 1:168–82.View ArticlePubMedGoogle Scholar
- Reilly S, Onslow M, Packman A, Wake M, Bavin EL, Prior M, et al. Predicting stuttering onset by the age of 3 years: a prospective, community cohort study. Pediatrics. 2009;123(1):270–7.View ArticlePubMedGoogle Scholar
- Ben-Shalom D, Poeppel D. Functional anatomic models of language: assembling the pieces. Neuroscientist. 2008;14:119–27.View ArticleGoogle Scholar
- Yeatman JD, Dougherty RF, Rykhlevskaia E, Sherbondy AJ, Deutsch GK, Wandell BA, et al. Anatomical properties of the arcuate fasciculus predict phonological and reading skills in children. J Cogn Neurosci. 2011;23(11):3304–17.PubMed CentralView ArticlePubMedGoogle Scholar
- Gathercole SE, Willis C, Baddeley AD. Dissociable influences of phonological memory and phonological awareness on reading and vocabulary development. Br J Psychol. 1991;82:387–406.View ArticleGoogle Scholar
- Hoien T, Lundberg I, Stanovic KE, Bjaalid IK. Components of phonological awareness. Read Writ. 1995;7:171–88.View ArticleGoogle Scholar
- Coch D, Grossi G, Coffey-Corina S, Holcomb P, Neville H. A developmental investigation of ERP auditory rhyming effects. Dev Sci. 2002;5 Suppl 4:467–89.View ArticleGoogle Scholar
- Khateb A, Pegna AJ, Landis T, Michel CM, Brunet D, Seghier ML, et al. Rhyme processing in the brain: an ERP mapping study. Int J Psychophysiol. 2007;63:240–50.View ArticlePubMedGoogle Scholar
- Yairi E, Ambrose NG. Early childhood stuttering I: persistency and recovery rates. J Speech Lang Hear Res. 1999;42:1097–112.View ArticlePubMedGoogle Scholar
- Hollingshead AB. Four factor index of social status. New Haven: Yale University Department of Sociology; 1975.Google Scholar
- Schopler E, Reichler R, Renner B. The Childhood Autism Rating Scale (CARS). Los Angeles: Western Psychological Services; 1988.Google Scholar
- Burgemeister B, Blum L, Lorge I. The Columbia Mental Maturity Scale. New York: Harcourt Brace Jovanovich; 1972.Google Scholar
- Carrow-Woolfolk E. Test for auditory comprehension of language. 3rd ed. Circle Pines: AGS; 1999.Google Scholar
- Dawson JI, Stout CE, Eyer JA. Structured photographic expressive language test: third edition. DeKalb: Janelle Publications; 2003.Google Scholar
- Boersma P, Weenink D. Praat. Dallas: Summer Institute of Linguistics; 2007.Google Scholar
- Lopez-Calderon J, Luck S. ERPLAB Toolbox (1.1.0); 2010. Retrieved from http://erpinfo.org/erplab.
- Picton TW, Bentin S, Berg P, Donchin E, Hillyard SA, Johnson Jr R, et al. Guidelines for using human event-related potentials to study cognition: recording standards and publication criteria. Psychophys. 2000;37:127–52.View ArticleGoogle Scholar
- Howell DC. Statistical methods for psychology (8th edition). Belmont: Thomson Wadsworth; 2013.Google Scholar
- Vossen H, van Breukelen G, Hermens H, van Os J, Lousberg R. More potential in statistical analyses of event-related potentials: a mixed regression approach. Int J Methods Psychiatr Res. 2011;20(3):e56–68.PubMedGoogle Scholar
- Sasisekaran J, Byrd CT. A preliminary investigation of segmentation and rhyme abilities of children who stutter. J Fluency Disord. 2013;2013(38):222–34.View ArticleGoogle Scholar
- Thierry G, Boulanouar K, Kherif F, Ranjeva JP, Demonet JF. Temporal sorting of neural components underlying phonological processing. Neuroreport. 1999;10(12):2599–603.View ArticlePubMedGoogle Scholar
- Praamstra P, Stegeman DF. Phonological effects on the auditory N400 event-related brain potential. Cogn Brain Res. 1993;1(2):73–86.View ArticleGoogle Scholar
- Hartwigsen G, Baumgaertner A, Price CJ, Koehnke M, Ulmer R, Siebner HR. Phonological decisions require both the left and right supramarginal gyri. Proc Nat Acad Sc. 2010;107(38):16494–99.View ArticleGoogle Scholar