You are here
State of the Science Conference: Developmental Stuttering
National Institute on Deafness and Other Communication Disorders
National Institutes of Health
U.S. Department of Health and Human Services
March 21-23, 2005
The Watergate Hotel
On this page:
James Battey, M.D., Ph.D., and Lana Shekim, Ph.D., National Institute on Deafness and Other Communication Disorders
Dr. Shekim welcomed participants and thanked them for their attendance and enthusiasm in addressing the complex topic of developmental stuttering. Noting that the meeting is structured to allow discussion, she encouraged participants to engage in conversation when possible. She noted that the meeting is co-sponsored by the National Institute on Deafness and Other Communication Disorders (NIDCD), the American Institute for Stuttering Treatment and Professional Training, the National Stuttering Association, and the Stuttering Foundation of America.
Dr. Battey noted that approximately 1% of Americans will stutter at some point. Although many of these cases resolve during childhood, much remains to be learned regarding the mechanisms that govern the stuttering process. Dr. Battey also commented that this meeting will enable the cross-disciplinary discussion that will help to focus NIDCD initiatives for stuttering research and treatment.
Sarah Bottjer, Ph.D., Department of Biology, University of Southern California
Dr. Bottjer began by noting that vocal learning is a combination of innate predispositions in perception and production and social factors that may interact with/override these innate tendencies. The vocal learning process in the zebra finch makes this songbird an applicable model system to study the neural and behavioral mechanisms by which humans learn to produce sounds. A juvenile male zebra finch learns to imitate sounds of a “tutor” during a sensitive period of its development, and these song patterns become so highly stereotyped that the bird is unable to change its pattern or to learn new patterns under normal circumstances. Incipient sub-song utterances are usually present by 20 days from birth, evolving into plastic songs as the bird becomes a juvenile adult (approximately 40 days). Syllables are the smallest units of motor production of bird song (Franz M and Goller F. J Neurobiology 2002;51:129-141). Each syllable corresponds to an increase in expiratory pressure, and the bird interrupts his song only at the end of an expiratory pulse. When birds copy acoustic syllables from a tutor, they also copy the tutor’s respiratory pattern.
Do motor constraints on sound production also constrain perception? Expiratory respiratory muscles set the basic temporal structure of the song, and the rate of expiratory pulses during song is five-fold higher in pressure than under non-song conditions (Suthers RA and Zollinger SA., Ann NY Acad Sci 2004;1016:109-129). The main vocal organ in the zebra finch is the syrinx. Two labia modulate the airflow through the syrinx, and the left and right syringeal muscles are controlled independently. The labia generate the sound, dorsal muscles time the sound production, and ventral muscles control the sound frequency.
During bird song, sound production arises from a combination of behavioral and neural aspects. Research indicates that sound structure emerges concurrently with a steep increase in the diversity of sounds (Tchernichovski O, et al. Science 2001; 291:2564-2569). Exposure to a tutor’s song causes a rapid increase in diversity of sound production, and the sounds that the bird produces become more structured. A later phase of the vocal imitation process is characterized by a shift from repetitions of early prototype sounds to in situ transformations, which yield imitations of tutor syllables. Thus, imitations of dissimilar sounds emerge from successive renditions of the same prototype.
With regard to the neural basis of bird song, neurons in the high vocal center (HVC) of the cortex process auditory input. The pathway from HVC to the robust nucleus of the archistriatum (RA) controls the motor production of song and is formed during the period of learning. A pathway that connects Area X to the medial portion of the dorsolateral nucleus of the thalamus (DLM), the lateral magnocellular nucleus of the anterior neostriatum (LMAN), and the RA is formed earlier in development and is also necessary for normal vocal learning in juvenile birds. All vocal-control brain regions receive auditory inputs, and neurons become tuned to the bird’s own song through a significant but poorly understood “gating” influence. Birth order and testosterone levels allow for some divergence in bird songs over time. Because zebra finches nest in small groups, there is a high priority placed on disparate songs. Neurons are selectively tuned to the bird’s own song during the course of the sensitive period of vocal learning, although it is not known whether the tutor’s song causes the same response. Once a certain age is reached, the tutor song and the bird’s own song are quite similar.
Lesions of LMAN made during early song development induce profound behavioral disruption of bird song, causing premature stereotypy of an abnormal song. Similar lesions of LMAN made during adulthood have little or no effect on the bird’s song, suggesting that LMAN plays an integral role in the development of acquired song in juvenile male zebra finches, possibly acting as an “error detector” during song development. Although humans have no known analog of the LMAN region, the mechanism by which LMAN confers learning is of great interest, as humans may also learn speech production by similar neural and behavioral mechanisms. For example, if the match between the motor output of the young bird and his tutor is poor, the young bird’s vocal output is modified. Deafening adult birds makes their songs gradually deteriorate, yet lesions in LMAN in fully-adult deafened birds result in stable song over time, possibly due to a short-circuit in the error message process. An age-dependent “practice effect” also plays a role in deafening-induced song deterioration; deaf full adults usually show deterioration in song 2-8 weeks following insult, whereas the song of young adults deteriorates much more quickly.
Preventing auditory experience with song during the sensitive period for vocal learning prevents the normal emergence of a topographic map within the LMAN → RA circuit (Iyengar S and Bottjer SW. J Neurosci 2002;22:946-958). Thus, there is experience-dependent remodeling of axonal connectivity early in the sensitive period for vocal learning. These changes are genetically programmed and timed to occur during a period of learning, indicating that they may be necessary prerequisites for learning. Axonal connections from LMAN to RA are topographically organized in older birds, but not in juvenile birds at the onset of vocal learning. A lack of auditory experience has no effect on the growth of RA, although it prevents normal remodeling of the LMAN axons that project to RA. This suggests that vocal experience regulates the formation of precise connections during vocal learning, a form of “neural commitment.”
How can songbird research address issues in human stuttering? Forced reduction in phonation (muting) leads to gradual modification of motor patterns of song in adult birds. However, such muting produces a much greater modification of motor patterns of song in young adults. Delayed auditory feedback causes syllable deletions, insertions, distortions, and stuttering; young adults insert novel motor gestures, in the form of short expiratory pulses, into the song motif. “Stuttering” was pronounced whenever such novel expiratory pulses were introduced. The songs of younger adult birds are also much more susceptible to deterioration following deafening, suggesting that song may be ingrained in older birds with more practice. This process may have some similarities to human speech acquisition, as infants lose the ability to perceive phonetic contrasts that are not expressed in their native language and increase their perceptual skills for native-language sounds.
One participant inquired whether the bird’s axons send numerous messages, which ultimately become pruned, or whether they initially send the correct specific signals. Dr. Bottjer noted that his was not known at this time, although pruning may be a likely mechanism.
Another attendee asked about the temporal relationship between a tutor and a young songbird. Social interaction occurs between the parent and child for up to 40 days, and vocalizations happen almost simultaneously. The father often sings while feeding and grooming the young bird; hearing and repeating the vocalizations are thus overlapping phenomena. In other species, even if this overlap between hearing the tutor and reproducing the sounds is broken, a good copy of the song is retained, suggesting a circuit of internal remembering.
Attendees then discussed stuttering in songbirds. When birds stutter in adulthood, they do not use the basal ganglia in Area X, suggesting that dopamine blockers are not useful. Motor plasticity can be induced by muting birds, yet lesions in the basal ganglia have no effect in vocal production in normal birds.
One participant inquired whether the syllable functioned as the coordinating unit for respiratory and glottal patterns in birds in an analogous manner to humans. Dr. Bottjer responded that manipulations such as muting or strobing interrupt vocal and syngringal programs in tandem. Single RA-projected neurons fire at individual points in the song, correlating sub-syllabic elements with notes. It is not known whether silencing a bird during development also silences its motor function.
It was also noted that human stuttering typically occurs within one expiratory cycle, whereas bird song stuttering appears to involve the repetition and deletion of the entire respiratory cycle.
Uwe Juergens, Ph.D., Deutsches Primatenzentrum, Goettingen
Dr. Juergens noted that monkeys do not stutter, although their brains are structurally most similar to those of humans. For ethical reasons, anatomical studies in humans are limited. Anatomical mapping of the monkey brain has been enabled by injection of tracers into select brain regions, followed by histological analysis.
In the monkey, the internal laryngeal muscles are innervated by neurons in the nucleus ambiguus. The articulatory muscles are innervated by neurons located in the trigeminal motor nucleus, facial nucleus, hypoglossal nucleus, and rostral part of the nucleus ambiguus. The respiratory muscles involved in phonation are innervated by motor neurons in the ventral horn of the thoracic and upper lumbar cord. The question arises of how these widely-dispersed motor neuron pools are coordinated to accomplish a specific vocal pattern. Anatomical studies show that parts of the reticular formation of the lower brainstem have direct connections with all phonatory motor neuron pools conjointly. Theoretically, the reticular formation is thus in a position to coordinate the activity of the various muscles involved in phonation. Furthermore, electrophysiological studies in the monkey show that the reticular formation contains numerous neurons that are active during and immediately before vocalization. Some of these neurons code specific acoustic features of the vocalization, such as fundamental frequency, intensity and duration, suggesting that the reticular formation of the lower brainstem is indeed involved in vocal motor coordination.
The reticular formation depends on a facilatory input from the periaqueductal gray of the midbrain (PAG) in its vocal motor-coordinating function. If the PAG is destroyed, mutism occurs. PAG lesions, however, do not cause a paralysis of the vocal folds: vocal fold movements during quiet respiration remain normal. The animals, however, no longer react vocally to external stimuli. On the other hand, activation of the PAG by electrical brain stimulation or injection of glutamate agonists induces species-specific vocalization. Recording the electrical activity in the PAG during spontaneous vocalization reveals vocalization-correlated activity. The type of activity obtained suggests that the PAG is involved in triggering or gating vocal reactions rather than in their motor coordination.
The PAG represents a crucial station of the descending limbic pathways that control vocalization. It is not part of the voluntary motor control pathway that arises in the primary motor cortex. If the latter is activated by electrical stimulation of its most rostroinferior part, vocal fold adduction is obtained. If the PAG is inactivated pharmacologically during motor cortical stimulation, vocal fold adduction can still be obtained. This is in contrast to vocal fold movements elicited from limbic forebrain structures, which can be blocked completely by PAG inactivation. The motor cortex represents part of a system that is responsible for the production of learned vocal patterns. Other parts of this system are the ventrolateral thalamus, cerebellum, putamen, and substantia nigra. Lesions in all of these structures have been reported to affect speech production.
Apart from structures responsible for vocal motor coordination (e.g., motor cortex, ventrolateral thalamus, cerebellum, putamen, substantia nigra, parvocellular reticular formation), other brain structures of potential relevance for stuttering are those involved in the planning and preparation of longer motor sequences (e.g., premotor cortex, supplementary motor area, lateral prefrontal cortex). This follows from the fact that stuttering is more likely to occur in long verbal sequences than in short ones. Furthermore, the strong influence of the emotional state on stuttering indicates that the limbic system (e.g., anterior cingulate cortex, amygdala, hypothalamus, and PAG) is also involved in stuttering control. In addition, stuttering may be strongly influenced by manipulating auditory feedback, suggesting that the auditory system also plays a major role. Finally, the observations that (1) stutterers show reversed functional hemispheric asymmetries in speech-relevant brain areas, (2) stutterers have a smaller corpus callosum than non-stutterers, (3) males have a smaller posterior corpus callosum than females, and (4) males are three times more likely to stutter than females, suggest that the special anatomy of the interhemispheric commissures might also play a role in the genesis of stuttering.
One attendee inquired about comparability of brain regions in monkeys and humans. Dr. Juergens noted that the gross anatomical connections are virtually identical in monkeys and humans. There are, however, clear functional differences. For instance, a bilateral lesion in the inferior motor cortex completely eliminates the capacity to speak in humans but has no effect on vocal communication in monkeys. This discrepancy can be explained by the fact that speech consists of learned vocal patterns, while monkey calls represent genetically programmed vocal patterns. The motor cortex is needed only for the production of learned, but not that of innate, motor patterns. This also explains why patients unable to speak are sometimes still able to produce nonverbal emotional utterances, such as laughing, crying, whimpering, or moaning (which are homologous to monkey calls). On the other hand, the destruction of certain structures produces identical effects on vocal behavior in monkeys and humans. The destruction of the periaqueductal gray, for instance, causes mutism in monkeys and humans.
One participant inquired whether there were mechanisms in monkeys that allow the suppression of vocalization. Dr. Juergens reported an experiment in which vocalization was elicited repetitively by electrical stimulation of the PAG. Whenever the ventral raphe in the midbrain/pons region was stimulated during vocalization, vocalization was suppressed. Another participant proposed that the projection from the globus pallidus internus via the ventrolateral thalamus to the supplementary motor area may gate the subsequent action in the motor cortex. Dr. Juergens agreed, but added that the actual role of this pathway in stuttering is still unknown. Also, the question of which specific limbic structures account for the stutterers’ fluency in highly-emotional utterances, such as swearing, cannot be answered yet.
Steven Petersen, Ph.D., Department of Neurology, Washington University Medical School
Issues in developmental imaging of stuttering include:
Issues of speech production in the scanner
Issues in “developmental imaging” of children versus adults
Further implications of “event-related” functional magnetic resonance imaging (fMRI)
Because speaking accompanies head movements and conformational changes in air spaces near the brain, speaking while inside an MRI scanner causes susceptibility artifacts. Two types of designs are common to imaging studies: (1) Blocked designs that feature a control period followed by series of trials measured in blocks, and (2) event-related trial designs that enable extraction of hemodynamic time courses. Blocked designs often produce susceptibility artifacts, but event-related designs can measure blood oxygenation level-determined (BOLD) response. The BOLD response is relatively slow (on the order of 14 seconds) related to the rapid neurological response associated with speech-language planning and production. Birn and colleagues reasoned that a speech artifact should be isolated to the first five seconds, allowing measurement of the remainder of the hemodynamic response with minimal interference from susceptibility artifacts. Moreover, Palmer and colleagues (Neuroimage 2001;14:182-193) showed that overt studies do not produce excessive noise, suggesting that event-related designs may be used to study verbal response paradigms in fMRI. With these designs, speech produces susceptibility artifacts that are small and relatively isolated to specific regions of the brain. fMRI noise can be filtered, and speech can be heard. Thus, for single-word paradigms, people may speak into the scanner, and data on reaction times may be collected.
Developmental imaging issues include anatomical differences and variability across stages of development. The functional imaging of children is not trivial, even for children who are 7-8 years old. The lore in developmental imaging is that children’s brains are shaped differently and are more variable than adult brains, thus making children’s brains not amenable to fMRI analysis. However, contour and sulci measurements have indicated that the brain of a 7-year-old is 90-95% of the size of the adult brain, with differences on the order of 3 mm (Burgund ED, et al. Neuroimage 2002;17:184-200). This conclusion also holds true for inter-brain variability. The shape differences are thus small relative to the fMRI resolution, and the variability differences are small and are not systematic.
Is the BOLD response different between adults and children-- e.g., is it more variable or localized differently? Kang and coworkers (Neuroimage 2003;20:1162-1170) demonstrated that locations of motor output regions do not differ statistically. Moreover, visual regions are similarly localized. Hemodynamic time courses are also similar, and there are no significant differences in peak responses. Thus, at the resolution typical of current fMRI studies, anatomical and physiological differences between adults and children do not preclude direct comparison of images. The greatest practical issue thus becomes getting the children to sit in the scanner. Currently, our lab has conducted fMRI analysis on more than 300 children, ages 7-19, using three types of verbal tasks--verb-, opposite-, and rhyme-generation. The two most commonly used types of images are: (1) main effect of time (e.g., Is there a reliable time course in a specific region of the brain? Do regions have reliable time courses across subjects?) and (2) interaction between group and time (e.g., Do regions show different time courses between groups?). In main-effect-of-time studies, the left frontal cortex has been found to be a region of difference. While it increases with age, the left extrastriatal cortex decreases with age. These observations suggest that children tend to use all of the ventral object processing system, which is then less apparent in adulthood. These studies offer proof-of-concept that a combination of extant methods can be employed to perform fMRI studies of verbal response tasks in adults and children age 6 and older.
One issue involved in the developmental imaging of stuttering is that stuttering does not occur reliably. However, event-related fMRI allows recoding of data based on accuracy or reaction time, as long as the data are collected in an event-related manner. Also, stuttering occurs most often in continuous speech output. Maccotta, et al. (Neuroimage 2001;14:1105-1121) showed that an important aspect of event-related designs is that the trial types are “jittered” with respect to one another. Rather than simply repeating tasks at preset intervals, subjects must be allowed to perform a task naturally. The natural variability in self-paced performance produces sufficient jitter for fMRI analysis. When this natural jitter is used, good imaging data result.
Zacks and colleagues have investigated ways that people naturally segment daily tasks, such as washing the dishes (Nat Neurosci 2001;4:651-655). When people were shown videotapes of such an activity and then asked actively to segment the activity into its main components, several brain regions showed activity at “event boundaries,” even in the passive condition. It is possible that similar approaches could be used for epochs of continuous speech that include stutters. However, continuous speech used in such a design presents a potential source of fMRI artifacts, and methods and validation would be necessary to carry out such a study.
In summary, when using specific techniques, fMRI studies of verbal production tasks are possible. With relatively minor adaptations, fMRI studies of development from age 6 to adulthood are straightforward. The use of tailored event-related designs should allow more specific identification of fMRI signatures of stuttering behavior.
Participants discussed several issues related to scanning and imaging, noting that the same regions of the brain are activated when people stutter and when they imagine that they are stuttering, suggesting possibilities for event-related fMRI. It was also noted that positron emission tomography (PET) imaging could be implemented in conjunction with fMRI for stuttering studies. However, logistics such as the work required outside of the scanner and confounding state variables must be considered in such experimental designs.
One attendee noted that auditory cortical areas show reduced activation during stuttering, and noise masking increases fluency. Both processes are modified during fMRI. To address these issues, clustered techniques may be used when collecting data scans. It was observed that any such techniques must be validated. While headphones can dampen the noise heard within the scanner, it is not known whether the wearing of headphones affects stuttering, although it affects auditory cortical response. Stuttering produces a decrement in event-related auditory cortical response, and noise inside the scanner may increase the response. The complexity of the scanning experiment provides some considerations for experimental design; the autonomic arousal sparked by speaking or getting into the scanner suggests the difficulties of determining isolated blood flow changes related solely to stuttering. Also, fMRI cannot image the differences between making an error and subsequently correcting it, since both actions produce a measurable response. Nonetheless, fMRI may offer advantages over other methods to measure speech, such as arterial spin tagging.
Participants discussed several issues related to scanning, noting that the temporal resolution with fMRI is not known but may be less than one second. It was also reiterated that thoughtful consideration of the work needed outside of (i.e., prior to the use of) the scanner is crucial to reduce the variability of the task.
Attendees also discussed the song-learning system in zebra finches. Persons who stutter may do so by failing to make a shift in semantic monitoring, e.g., Am I saying what I meant to say? While this has not been tested in songbirds, the idea can be tested. In birds, the difference between stuttering and severation of a motor response is not known. However, only those birds that insert novel expiratory pulses will stutter, and they stutter only on those pulses. This observation suggests a conflict between an inherent motor sequence plan and the novel expiratory burst. Within an expiratory burst, some pattern changes are observed, including the retention, deletion, and insertion of syllables. In birds that have been muted, the time between repetitions becomes shorter as the utterance proceeds. It was observed that, in humans, vocal paralysis is modified by extra inspiration.
Participants also addressed the neural components that are possibly involved in stuttering. It is known that inputs to motor neuron pools are disrupted in stuttering, leading to a competing controller hypothesis--the act of speaking requires suppression of emotional and other systems. Could persons who stutter have incomplete suppression of competing input systems? Different motor drivers may be in conflict, and neuromodular influences (e.g., feedback loops) may be involved. Several participants pointed to a need for computational neuromodeling systems to address these questions. Regarding the possible role of language planning models, it was noted that monkey calls result from genetically pre-programmed motor patterns. Accordingly, language planning models are not applicable to monkey calls. Moreover, monkey calls are produced in highly emotional states. Human stuttering, in contrast, is often ameliorated under specific emotions. Thus, monkey calls are not good models for stuttering, although the gross neuroanatomy between humans and monkeys is similar.
One attendee suggested that stuttering be considered as a constellation of factors--perhaps conflicting inputs to the vocal system--motor repetitions and emotional reactions to those repetitions. The mirror neuron system in monkeys is thought to be involved in the evolution of language, analogous to Broca’s area in the human brain. Songbirds, however, do not have mirror neurons, although some motor neurons display song-driven activity. Stuttering should also be considered in relation to motor repetitive tasks (e.g., musical instruments, American Sign Language (ASL)). From a genetic perspective, the FoxP2 gene, associated with speech disorders, is expressed more highly in juvenile birds who are involved in the learning process. However, there is currently little evidence from speech and learning studies to link the gene directly to stuttering.
Ehud Yairi, Ph.D., Department of Speech/Hearing Science, The University of Illinois at Urbana-Champaign
With respect to etiology of stuttering, major trends in the research indicate a shift from learning to genetics. From an epidemiologic perspective, focus has moved from uniformity to a diversity of concepts about onset, and from ascending to descending models in development. In studies of symptomatology, concepts of overlap have shifted toward those of distinction. Regarding the mechanisms thought to underlie stuttering, the emphasis has shifted from motor peripheral bases to a central planning basis. Therapeutic trends have shifted from modifying stuttering and emotional reactions to generating fluency. For children’s therapy, indirect treatment as the main, if not only, approach has given way to direct speech modification.
Our research concerning predisposing factors for stuttering relates to several of these mega-trends and includes incidence and prevalence, genetics, gender, age, concomitant disorders, stressors, natural recovery and persistency, and subtypes. The incidence of stuttering has been estimated at about 5% of the population, with a population prevalence somewhat under 1%. The prevalence in preschool aged children, however, is almost 2.5% (Craig A. J Speech Lang Hear Res 2002;45:1097-1105). Thus, in this age range, the problem is much greater, with important implications concerning the needs for research and the training of speech-language clinicians.
The percent of stuttering children with stuttering relatives is relatively high (Ambrose NG. J Speech Hear Res 1993;36:701-706), with a strong sex factor (e.g., brothers and fathers who stutter). Current data suggest a combination model of multifactorial polygenic and single major locus components is the most applicable (Ambrose NG et al. J Speech Lang Hear Res 1997;40:567-580). These findings justified biological genetics studies, one of which was recently completed by our team (Riaz N. Am J Hum Genet 2005;76:647-651). Transmissible factors that underlie stuttering may include biochemical pathways, structural anomalies, neural processing pathways, motor skills, and temperament. Ongoing and future behavioral genetic studies should explore familial patterns of expression, with a focus on high-risk relatives who do not stutter but have transmitted stuttering. Also, genetic research should look toward subpopulations with high (e.g., Down’s syndrome) and low incidence (e.g., deafness and cleft palate).
Gender distribution of stuttering indicates a male/female ratio of nearly 2:1 near the time of onset (Yairi and Ambrose 2005) and 4:1 in adolescence (Craig A. J Speech Lang Hear Res 2002;45:1097-1105). It appears that the sex ratio in stuttering is influenced by genetic factors, including natural recovery. Whereas the male/female ratio in recovered stutterers is 2.33, it is 3.75 in persistent stutterers. Thus, females are more likely to recover.
Future research should consider these overlaps in linguistic and motor development. Our research using structural MRI in school-age children has revealed differences between stutterers and non-stutterers in the white and gray matter associated with speech processes. Initial findings with aged persons who stutter indicate callosal area measurements that are below norms (Sullivan E, et al. Neurobiology of Aging 2001;22:603-611).
Concomitant disorders, including those affecting language, phonology/articulation, reading, emotional development, ADHD, voice, and brain damage, may be related to stuttering. These should be researched more thoroughly. There are also indications for many stressors associated with onset that may facilitate stuttering, including emotional upset, behavioral stress, illness, and rapid language development. Also, awareness of stuttering in children that increases sharply around age 4 should be explored more fully.
Investigators have also sought to determine whether recovery from stuttering is natural or the result of various environmental influences. After 4 years post-onset, roughly 75% of those who stutter have recovered. Some of those who have not recovered by this point may still recover, albeit at a much slower rate. Persistence and recovery run in families; individuals whose families have cases of recovered stuttering have a greater chance of recovery. Thus, research that investigates early predictive factors of these developmental paths is a top priority.
Participants inquired about methods for measurement of various stuttering-related indices. To assess self-awareness of stuttering in young children, puppets and other methods are used. Assessments of syntactic and morphologic development include indicators of grammatical as well as vocabulary skills (e.g., index of productive syntax, vocabulary anagrams), which have been quantitative. It was also noted that during the critical period of language development, the speech system undergoes large and rapid anatomical development.
Nancy Cox, Ph.D., Department of Human Genetics, The University of Chicago
Stuttering is highly familial, yet it is a complex disorder that does not follow a simple pattern of transmission within families. However, the risk for sibling stuttering is five times greater than that of the general population, and linkage mapping studies have been conducted to look for co-segregation of stuttering among families through genetic markers. Of 105 families investigated (481 individuals genotyped) by the Illinois International Genetics of Stuttering Project, 9,144 genetic markers have been considered. The strongest evidence for linkage is on chromosome 9; for persistent stutterers, linkages are also seen on chromosomes 13, 2, and 7. When looking for interactions across the genome, researchers have observed a high linkage between chromosomes 2 and 9 for different families, suggesting a genetic basis for “ever stuttered” phenotypes. For males, linkages appear on chromosomes 20 and 7. Conditioning on chromosome 7 gives evidence for linkage on chromosomes 12 and 18 (Riaz N. Am J Hum Genet 2005;76:647-651), and a linkage for females only is observed on chromosome 21.
These data suggest that the genetic component of stuttering displays a great level of overlap among early genome screens, although it appears to be much less complex than the overlap seen with other diseases, such as diabetes. Moreover, the data suggest that several genes contribute to risk factors for stuttering. Next steps include prioritizing candidate genes in the key chromosomal regions and identifying overlap between stuttering and diseases with a language component, such as autism. Bioinformatics network approaches will be useful to these efforts.
One attendee noted that FoxP2 is located on chromosome 7, and Dr. Cox commented that a slight change in signal intensity is observed on chromosome 7. Another attendee inquired about evidence of genetic overlap between stuttering and Tourette syndrome, of which none has been observed to date. Similarly, there is no overlap between stuttering and studies that examine the genomic basis of habit-forming behaviors, such as obsessive-compulsive disorders.
Dr. Cox commented that genomic analysis allows for a great deal of confidence for identifying persons who stutter. Regarding the percentage of carriers who stutter, she noted that the penetrance appears to be lower in females than males. She noted also that while linkages are suggestive, precise localization of markers will be necessary to identify points of convergence between stuttering and language impairment.
Christine Weber-Fox, Ph.D., Department of Speech, Language & Hearing Sciences, Purdue University
There is a wealth of evidence to suggest that stuttering should be considered within a theoretical framework that incorporates motor, linguistic, cognitive, psychosocial, and genetic factors. The occurrence of stuttering has a predictable relationship to linguistic constructs (Au-Yeung J et al. J Speech Lang Hear Res 1998;41:1019-1030), and online reaction time (RT) measures suggest that lexical access may be slower for adults who stutter (Bosshardt HG and Fransen H. J Speech Hear Res 1996;39:785-797). Moreover, a number of researchers have reported sub-clinical language deficits in areas of phonology and syntax. Current knowledge reveals a gap between models of language processing/production and those of speech motor control. A bidirectional influence between these areas that changes during the processes of development is hypothesized.
Linguistic complexity and the resultant speech motor output have been examined using kinematic recordings of upper lip, lower lip, and jaw movements. Adults who stutter show increased variability in movement patterns with increased linguistic and length demands (Kleinow J and Smith A. J Speech Lang Hear Res 2000;43:548-559). Electroencephalogram (EEG) recordings have been employed to assess whether brain activity for language processing and neural functions differ among individuals who stutter, even in the absence of speech production demands (Weber-Fox CM. J Speech Lang Hear Res 2001;44:814-825). Results indicate that in the absence of production demands, functional brain organization for language processing (as characterized by reduced negative brain potentials) is different in adults who stutter. Effects were related to processes underlying closed and open-class words and semantic abnormalities. With respect to grammatical processing, on-line grammaticality judgments were less accurate for adults who stutter (Cuadrado EM and Weber-Fox CM. J Speech Lang Hear Res 2003;46:960-976), although off-line judgments did not differ between adults who stutter and those with fluency. Consistent results were seen with the EEG studies; processing differences were observed in the absence of speech production demands.
Recent studies that examine phonologic processing indicate that phonological encoding systems operate similarly for adults who stutter and those who are fluent, at least when no overt speech is required. However, adults who stutter show a greater slowing for the most difficult rhyme judgment conditions, suggesting that adults who stutter are more vulnerable to increased cognitive loads and display greater right hemispheric involvement in late cognitive processing. These processing differences in the auditory modality parallel many of the findings for visual language processing. However, behavioral effects may be reduced in comparison to the visual modality.
Do children who stutter exhibit similar processing differences compared to their typically-developing peers? In a silent rhyming task, neural activity over the left hemisphere occurred later for children who stutter but earlier over the right hemisphere when compared to their normally-fluent peers. Results for children who stutter mirror those observed in adults who stutter; both are characterized by atypical functions in the relative contributions between the left and right hemispheres. These results support the hypothesis that functions of language-processing systems, including phonological encoding, may be important factors in the development of stuttering for some individuals. Furthermore, differences in language-related neural functions that distinguish normally fluent speakers from those who stutter may change over the course of development. Moreover, rhyme judgment accuracy and event-related potential (ERP) measures indicate that processes related to the visual phonological encoding that mediates lexical access operate atypically for at least some school-age children who stutter. Stuttering involves processes of both decision-making and lexical access.
One attendee asked if increases in right hemispherical activity are compensatory, and Dr. Weber-Fox noted that some of the activity may be compensatory. It is not fully known whether these differences are specific to language processing, although it is hypothesized that the auditory paradigm may have similar processing modalities to the visual paradigm. However, preliminary results indicate that the auditory modality does not differ as greatly between adults who stutter and those who are fluent as does the visual modality. It was also noted that behavioral measures could possibly be used to predict neurological signals and vice versa.
Another attendee asked if differences were observed in ERP patterns for auditory data, even in the absence of observed performance differences. Dr. Weber-Fox noted that in the case of reduced semantic expectation, behavioral accuracy may not link to ERP, although this relationship is not certain in verb agreement violations.
One attendee suggested the possible value of assessing, by means of EEG and related methodologies, the relation between developing reading abilities--particularly so-called “phonic skills--and phonological development in pre-school-age children who do and do not stutter.
Workshop participants discussed genetic influences and biological triggers that may affect stuttering. Limited data exist on the early prediction of stuttering based on familial traits, although many genetic disorders do not demonstrate physiological predictors (e.g., biochemical measures). Moreover, many factors, such as environmental stressors, are not predictable, even though evidence of genetic predisposition may be present. Causal relationships between stuttering and various biological triggers or environmental factors are difficult to establish, although some consistent data are becoming available. Therefore, a simple model is not likely to explain stuttering--a systems model that allows for perturbations will be necessary. Also, acquired neurogenic stuttering has been associated with many lesion sites, suggesting that diminished fluency may be a response to numerous interruptions of neural circuits.
It has been observed that children who recover from stuttering have slower rates of speech than their fluent counterparts who never stuttered. One participant asked whether data exist on speech rates of children who recover versus those who control their stuttering and those who persist. As with all variables, the speaking rate must be compared with all other influences, suggesting that a multi-faceted approach will be most useful.
The literature of stuttering indicates differences that are not limited to language; coping behaviors may play a role. Recovered stutterers may therefore provide insight into neural plasticity. It was also noted that awareness and concern are discrete issues; individuals who acquire stuttering during adulthood are often less concerned (many are simply annoyed) by the disruption that stuttering causes than are children who stutter. While an “adaptation” effect is not present in Down’s and Tourette syndromes, children who stutter as young as age 6 show a spectrum of coping behaviors (some of which may be avoidance strategies). However, there is no current complete developmental model of speech motor learning.
It was also observed that at 2.5 to 3 years of age, the density of dopamine D2 receptors peaks, suggesting possible neural bases for the onset of stuttering. However, developmental and chronological years do not correlate directly. Alternately, the time course that follows the onset of stuttering may create neurological changes that are more important. Because treatment effects drop off substantially at 18 months, this suggests that the system is losing plasticity in a key facet. Some pruning may be taking place after a period of time with this behavior. The new behavior in turn modifies neural systems. Neural plasticity is elevated at young ages, as seen in disorders such as childhood aphasia.
John Bates, Ph.D., Department of Psychology, Indiana University
Initial social development models emphasized parental effects on children, producing findings that suggested that parental warmth and control may be associated with child development problems. Recently, developmental model systems have posited reciprocal influences in socialization; non-linear interactions and organized, chaotic patterns of change may occur during development. Such theoretical frameworks necessitate a concept such as “temperament” to understand how children may shape their own development. Sameroff has used the transactional approach, hypothesizing that behavioral phenotypes are influenced by genes and environment and in turn interact with each. Lerner has suggested a multidimensional system in which each dimension influences all others. A further complication, not represented in these models, is the effect of moderators such as child temperament and parenting.
What is temperament? Rothbart has defined temperament as individual differences in reactivity and self-regulation traits (negative emotionality such as fear and anger, positive emotionality, impulsivity, executive control of attention) that are fairly stable and consistent over time. Marvin Zuckerman has equated temperament with basic personality, citing the “Big Five” (e.g., neuroticism, extraversion, agreeableness, conscientiousness, openness) or the “Big Three” (e.g., negative emotionality, positive emotionality, constraint) personality traits. However, temperament has the capacity to develop, especially at the level of phenotype. Effortful control is one of the better examples of temperamental development, and this construct becomes meaningful as the brain develops.
At present, there is no single universal measurement of temperament. Temperament may be measured via observation, either structured or naturalistic, and caregiver reports (e.g., Infant Characteristics and Infant Behavior Questionnaires). Infant Characteristics Questionnaire dimensions include unadaptability (predisposition to fearfulness in a novel situation; provides insight into the behavioral inhibition system), difficultness (behavioral inhibition and activation systems), and resistance/unmanageability (less executive control of attention). Measurement challenges exist with both observational and caregiver reports. If possible, combination of both measurements is most desirable, but if resources are limited, at early stages of research, the parental report may be acceptable.
Parental temperament ratings have objective components. Cries of difficult infants include long pauses between sounds and a higher pitch at peak intensity. When confronted with such patterns, unrelated women often say that such children sound annoyed or spoiled. Parent reports on various temperament dimensions correlate with those from home observations and between parents, suggesting some degree of objectivity. Subjective components are not dominant, and mother characteristics (e.g., social desirability) account for some variance in temperament ratings. Highly objective ratings of child verbal development are also explained in part by the mother’s personality. While subjective factors must be considered, they are not sufficient grounds to discard parental report as a source of data.
Several theoretical models of interaction between temperament and environment have been hypothesized. Chess and Thomas have proposed a “goodness of fit” model that suggests that a child’s temperament becomes problematic only if it does not fit well with the environment. Wachs has proposed an organismic specificity model, in which a given environment affects different children differently. Kochanska has observed that gentle control is most effective with inhibited children, and Arcus and colleagues have suggested that stern parenting of reactive children is less likely to promote behavioral inhibition. Bates and colleagues have found that strict parental control of unmanageable children may lead to lower levels of externalizing behavior than would have been predicted by the temperamental unmanageability alone. Moreover, temperament-temperament interactions are theoretically likely, although empirical examples are limited.
Variables that may be relevant when applying temperament to the study of developmental stuttering include fear, inhibition, and disposition to anxiety. Currently, relevant research applying these models to stuttering is limited, and there is some non-convergence on issues, e.g., whether the anxiety associated with stuttering is primary or secondary. Research also remains unclear on the etiological bases of stuttering. Attentional self-regulation--diverting attention from one stimulus to another--may also play a role in stuttering. It has been postulated that children who are oversensitive to discrepancies de-automize speech too easily. It has also been speculated that a lower ability to direct attention to non-stressful elements may increase fear and distract from efficiently and rapidly forming speech and language. If, as some have suggested, stuttering involves approach/avoidance conflict, then there must be some motivation to approach, or there is no conflict. Under this assumption, predisposition to both fear and approach social situations would be necessary, even though insufficient to develop stuttering. Environmental processes that may add to or interact with temperament include the criticality of parents of children who are developing stuttering problems, the level of parental concern over stuttering, and the possible roles of older siblings and distal stressors (e.g., family stress, sleep disruption stress) that lead to poor self-regulation. Other biological factors may also be involved in speech-language production, such as abnormalities in the areas of the brain that control timing of speech production.
Temperament will be a useful conceptual tool in developmental studies of stuttering if:
Theoretically-relevant dimensions of temperament are well-measured
Temperament dimensions are considered in relation to one another and in relation to other biological factors, including stress, development of speech-language areas of brain, and social environment factors
One attendee inquired about the predictive power of temperamental variables for later behavior. Dr. Bates noted that coefficients of correlation vary, but there is some evidence to suggest that certain behavioral characteristics may be linked to temperament.
One participant asked about the roles of the brain hemispheres with regard to temperament, and Dr. Bates noted that parallels exist between adults and children who are negatively emotional. Both show dominance of the right hemisphere.
There was also some discussion of trait (e.g., temperament) versus state (e.g., emotional reactivity) aspects of affect and the need to study both aspects separately as well as jointly, relative to developmental stuttering.
Cynthia Stifter, Ph.D., Department of Human Development and Family Studies, The Pennsylvania State University
Regarding the development of emotions and behavioral control, there is consent about the timetable of emotional onset. Newborn babies display distress, disgust, and surprise; at 2-3 months, joy and anger appear; at 6-9 months, fear and sadness are added; at 18-24 months, toddlers display embarrassment, envy, and empathy; at 30-36 months, pride, shame, and guilt are present. Anger results when the developing child is blocked from a desired goal. Fear arises from a desire to maintain the integrity of self and avoid danger. Regulation of these and other emotions involves the extrinsic and intrinsic processes responsible for monitoring, evaluating, and modifying emotional reactions, especially on their intensive and temporal features, to accomplish one’s goals (Thompson RA. Monogr Soc Res Child Dev 1994;59:25-52). Even though infants have the regulatory ability to react, extrinsic factors are quite important. Under conditions of high stress, the caregiver is central to reduce and regulate stress and turn this into positive affect. Parents may induce emotion, restrict opportunities, intervene directly, selectively reinforce a particular emotion, and interpret emotionally-charged situations. Maternal regulation of infant distress has been measured during inoculation procedures (Jahromi LB and Putnam. Dev Psychol 2004;40:477-487), and parental vocalizing and holding/rocking are much more effective at reducing stress than are distraction techniques. Moreover, mothers who are more effective at reducing their children’s distress at two months have less distressed babies at six months.
While anger emerges early in infancy, the emotion is initially relatively benign. Typically, there are no negative consequences for an angry infant; parents try to resolve the frustration rather than to punish the infant. Anger occurs with much greater frequency than other negative emotions, thereby providing the infant the opportunity to practice regulatory behaviors and for parents to teach and/or support an infant’s regulatory skills. Anger is a context for regulatory strategies. Early self-regulatory skills include attentional control, locomotion, and self-comforting behaviors such as thumb sucking, and, later, nonverbal communication.
Fear differs from anger in its development. The only behavior effective against fear is withdrawal, in contrast to anger, for which distraction, reaching for a toy, and looking toward the mother are all effective regulatory behaviors. Awareness of self must occur to develop embarrassment, envy, empathy, and other complex emotions. At 30-36 months, parents request behavioral control by setting rules and standards, although parental demands of their children change over time. At 13-24 months, demands focus on safety, preserving property, and manners. By 24-36 months, parents ask children to exhibit self-care and to contribute to family routines. At 36-42 months, elaborations of family and social norms occur. Parent-child conflicts peak at 21-30 months, followed by a decrease resulting from increases in the child’s self-control and ability to negotiate.
Compliance includes delaying gratification, resisting a tempting but forbidden impulse, and moderating frustration. Does regulation of anger lead to compliance? Noncompliance codes include assertion, avoidance, defiance, and passive noncompliance. Babies who show regulatory behavior to arm restraint also demonstrate high regulation linked to compliance at 30 months. The parent has tremendous influence on the success of compliance requests. Extroverted children are more likely to correlate with anger reactivity and externalization of behavior problems. In introverted children, fear reactivity is an outcome, leading to internalizing behavior problems.
To identify emotional precursors of stuttering, self-regulation of emotion as well as caregiver regulation of emotion (particularly anger) in first year of life should be examined. To identify behavioral control correlates of stuttering, parental demands for rules/standards, parental behavior, and child behavioral regulation (e.g., compliance) at the end of the second year of life should be examined. While conjectural at present, it is possible that stuttering may be linked to emotional trait profiles. For example, persons who stutter may be outgoing by nature, yet have enough inhibition to feel anxious about their stuttering.
One attendee commented that there are gaps within the literature of behavior and personality. Many of the experiments traditionally used to investigate “personality” were actually attempts to understand psychopathology. Distinguishing between a trait seen in early development and its phenotypic manifestation later in development is essential. Another participant noted that the concept of self-regulation, important in psychology, medicine, and education, may provide a model useful in studies of the nature and treatment of stuttering.
Another participant noted that a limited body of literature exists regarding social anxiety and inquired whether stuttering could be a reasonable by-product of behavior that would produce mild anxiety. Many procedures focused on anxiety reduction fail, although some circumstantial evidence exists to support the notion of invoking anxiety.
One attendee asked whether models exist that demonstrate ways that shifts in life events may impact and influence stuttering. Is the environment a factor in transforming stuttering into a cause of anxiety? It was noted that parental control remains relatively consistent, even if stuttering behavior is being observed 2-3 months after onset. Also, temperament should not change, in theory, during those two months. However, environment encourages sociability, which in turn induces certain coping behaviors. Thus, behavioral strategies may change even if temperamental characteristics remain relatively stable.
Another participant asked whether internalizing disorders lead to psychosomatic problems. A child may exhibit both externalizing and internalizing behaviors; both dimensions occur to differing degrees in each individual. With respect to the role of validating a child’s emotion, studies have investigated parental warmth, sensitivity, and awareness of emotion. Identifying and validating emotion is a positive parental role in emotional development.
Participants also discussed temperament research in children, possible correlations to adults, and the extent of genetic studies of the basis of temperament. Considerable correlations have been noted between behavior and genetic factors; for example, childhood temperament has been linked to adult “Big Five” traits. However, a longitudinal study is necessary to identify specific correlates and provide theoretical modeling for these linkages. Moreover, it is unclear how extrapolation of personality measures links to stuttering; people may adopt phenotypic traits as a result of their life histories that differ greatly from early predispositions.
Participants also discussed children’s strategies to recover successfully from stuttering. Since a large number of those who stutter recover between ages 2 and 5, this suggests that parental reaction to stuttering plays a role in recovery. What types of parental strategies or children’s behavior could reduce the incidence of stuttering? Those who recover may be those who are unconcerned about stuttering and who eventually develop developmental control. Strict parental control seems to reduce stuttering dramatically. For example, by gently or non-adversively telling the child to stop stuttering, the parent may be temporarily taking over behavioral control. The parental rebuke does not need to be overtly punitive to be successful; it was conjectured that, by highlighting the stuttering, the parent may be temporarily preventing the child’s tendency to “turn down” the auditory system.
It was also observed that there is great variance in the rate of language development between 24 and 36 months, which is a peak period for conflict between the parent and child. There are two types of compliance, willing and situational. In terms of treatment, compliance depends on the child’s temperament, suggesting a need to examine the parent’s style and how the style fits with compliance and internalization.
Anita Miller Sostek, Ph.D., Clinical and Population Studies, Center for Scientific Review, NIH
Dr. Sostek provided participants with an overview of the NIH review process, the details of which are not reproduced here. Information about the NIH review process is available at www.csr.nih.gov.
Barry Guitar, Ph.D., University of Vermont
Issues in treatment assessment include natural recovery (5% of those who stutter have a lifetime incidence; 75-80% in the population appear to recover naturally by 6 years of age), adequate sample size, variability of stuttering, reproducibility of treatment effects, and the need for appropriate measures of change. While single-subject study designs may be useful for pilot studies, larger control groups are necessary to provide the statistical power to observe treatment effects and to address dropouts. Variability of stuttering requires repeated measures and long-term follow-up (at least one year is necessary; two or more are preferable). Also, measures conducted in the real-world environment (e.g., the child’s home) are necessary to measure treatment efficacy. Moreover, it is essential that studies be reproducible to ascertain the reliability of specific treatment effects. Appropriate measures of change require an understanding of the impairment, disability, and handicap. For preschool children, behavioral aspects of stuttering, including the frequency and severity, may be measured at home and outside of the home. For school-age children, evaluated at home and at school, measures should include behavioral, cognitive, and emotional aspects of stuttering, including frequency and severity, naturalness and communicative competence, thoughts and beliefs about speech and listeners, and emotional concerns (e.g., attitudes and anxiety about communication). Adolescents and adults can be measured in social and occupational situations. Behavioral characteristics, cognitive issues (including self-rating of stuttering), and emotional attitudes (e.g., frustration, fear, shame) can all be measured in these adolescent/adult populations.
Treatment studies may be grouped into several phases: Phase I studies demonstrate that a treatment does no harm, Phase II studies test long-term efficacy, and Phase III trials investigate large-scale application. The treatment studies reviewed are organized by type of treatment with three age groups: preschool, school age, and adolescent/adult. Most studies are treatment-outcome investigations rather than treatment-process studies. Preschool treatments reported in the literature include the Lidcombe Method, a parent-delivered operant conditioning to promote fluency in which parents are taught to use a casual, non-interruptive comment on fluent speech. Lidcombe is supported by numerous outcome studies that demonstrate the safety of the treatment, long-term effectiveness, and increased efficacy compared to controls (e.g., Harris V, et al. J Fluency Disord 2002;27:203-213). Another preschool treatment with supporting evidence is the parent-child interaction (PCI) approach, which involves changing the child’s environment to make it conducive to the development of fluency. Also available is the Fluency-Rules Program, in which the clinician teaches the child to speak slowly and to reduce repetitions and prolongations. Parent-Child Groups, in which parents are supported while they learn to change the environment, and children modify stutters in a group setting, have also been investigated. While each of these treatments has shown promise, long-term data collection and rigorous methodology associated with Lidcombe have set a standard for evaluating stuttering treatments.
Treatments assessed for school-age children who stutter include delayed auditory feedback (DAF) (Craig A, J Speech Hear Res 1996;39:818-826; Hancock K. J Speech Lang Hear Res 1998;41:1242-1252), which has demonstrated a significant reduction in state and trait anxiety and a significant improvement in stuttering relative to controls. Long-term follow-up indicates that improvements were maintained, as were reductions in trait anxiety.
A meta-analysis of treatments for adults and adolescents who stutter (Andrews, Guitar, and Howie.J Speech Hear Disord 1980;45:287-307) indicated that the most effective treatments employed prolonged speech and gentle onset of voicing. Studies of treatments that alter speech patterns via modification in the length of phonation intervals (Ingham RJ. J Speech Lang Hear Res 2001;44:1229-1244) suggest that fluency may be maintained for a year following treatment. By contrast, prolonged speech strategies (e.g., slow rate, gentle onset, loose contacts, pausing) using the Camperdown Program have proved somewhat tedious, with a high dropout rate. Also, the Comprehensive Stuttering Program, a 3-week intensive program of fluency-shaping and added cognitive-behavioral skills, showed that 80% of enrollees maintained a “clinically meaningful” change after five years (Langevin and Kully. J Fluency Disord 2003;28:219-235).
Various pharmacological approaches have been investigated; in the 1970s, haloperidol reduced stuttering in double-blind studies, although side effects were severe. These studies implicated dopamine as a factor in stuttering. Other drugs investigated for effectiveness in stuttering include clomipramine, risperidone, and the dopamine antagonist olanzapine.
Assistive devices, such as Fluency Master and Speech Easy, that typically provide auditory stimulation, such as masking or DAF (sometimes triggered by the speaker’s voice), have also been evaluated. These devices are generally relatively expensive compared to other treatment modalities and have yielded mixed results.
Issues to consider include:
Why is stuttering tractable before age 6 and so difficult to treat after elementary school?
Why is stuttering severity unrelated to outcome in preschool therapy and so predictive of outcome in adults?
Do cognitions and emotions change as a result of increased fluency? Is there a placebo effect of just “being in therapy”?
What are the treatment goals (e.g., spontaneous fluency, modified stuttering, improved communication effectiveness) and who decides these?
John Walkup, M.D., Division of Child and Adolescent Psychiatry, Johns Hopkins Hospital
Many lessons that are applicable to clinical trials for stuttering can be learned from large-scale clinical trials of other disorders that have enrolled children and adolescents. Early clinical trials for psychiatric disorders (i.e., ADHD, depression, anxiety) were small, single-site, idiosyncratic, and often never replicated. Clinical trials for stuttering treatments are at a similar stage. Real benefits may be had by moving to a large-scale, multi-site clinical trial for stuttering, as such a trial may move the field forward in a number of positive ways. Examples from recently published and currently active clinical trials were provided to demonstrate the following points.
Large scale, multi-site trials can raise the public’s awareness of a particular problem and have significant public health impact. Moreover, such trials may ultimately change clinical practice and, perhaps more importantly, change how the next generation of practitioners is trained.
Multi-site studies require investigators to come together and set a method for both intervention and assessment of outcome. Such a focus may help a field with widely different treatment and assessment strategies to focus on essential constituents of treatment and simple and useful definitions and ratings of outcome.
Comparative treatment trials of pharmacotherapy versus psychosocial treatment allow for the use of a pill placebo. The use of a medication arm and a pill placebo offers two powerful comparator groups to evaluate psychosocial treatments. Medication arms (active and placebo) in such a design include treatments that subjects believe may make them better. If psychosocial treatment proves more effective than both placebo and active medication, such a design provides powerful evidence to support its efficacy.
Psychosocial treatment controls are standard in stuttering, but may not be ecologically valid. If they are credible, then they are likely to be partially active and may make it difficult to demonstrate outcome of treatment.
- In comparative treatment trials, more complicated (i.e., comorbid and chronic) patient presentations often favor the combination treatment arm. For a disorder that waxes and wanes in severity and features a high spontaneous remission rate, it may be more useful to focus on severe and chronic subjects for a large-scale trial.
Moderator and mediator analyses often require very large samples to achieve adequate power. To demonstrate why and for whom a treatment works may necessitate a very large clinical trial. The best way to identify effective treatments and appropriate populations is to characterize subjects well.
NIH Program initiatives have been helpful in moving intervention research forward. Funding mechanisms have included contracts, cooperative agreements, and multi-site R01s.
Conference grants may be a useful way to begin such a process.
Selecting sites for large-scale trials is critical. The capacity to recruit and investment in the experiment (not just in the treatment approach or assessment strategy) are critical.
Large-scale trials also require cooperation among sites and good will. Without such spirit, trials may become mired.
Significant support must be included for data management and analysis to utilize large and comprehensive data sets.
Family agencies may be a resource for investigators who wish to advance these agendas. The NIH is particularly sensitive to such constituencies.
Participation in large-scale clinical trials may be seen as a threat to academic advancement. Therefore, it is often useful for senior investigators to lead the trails and for junior investigators to participate in apprentice roles.
Participants discussed clinical trials in stuttering. One attendee asked whether any clinical trials have assessed the impact of parents or home environment, and it was noted that such studies have been limited. Currently, no Phase III trials have been carried out for the study of stuttering. Currently, the groundwork is being created for Phase II studies that use well-controlled groups. One attendee commented that a successful treatment could be established as a baseline strategy for use in comparative studies.
Participants also elaborated on collaborative studies carried out with the Tourette Syndrome Association (TSA). These collaborations began as a series of independent studies to identify genetic bases for Tourette syndrome, ultimately leading to a large-scale clinical trial sponsored by a grant from the NINDS to the TSA. Based on the success of the trial, the TSA has formed an Imaging Consortium and a Behavioral Sciences Consortium to encourage collaborative, cross-disciplinary research. Two large treatment grants have been funded as a result of TSA support.
It was also noted that some components of stuttering treatment are well established, such as controlling the length of utterances, suggesting that the time may be appropriate to conduct a larger trial. Despite empirical support for this approach, there is some concern that clinicians may not be following research recommendations. Large-scale clinical trials enable a focus on the training of practitioners, as there is a strong credible empirical support.
It was observed that multi-institutional grants require detailed planning, strong leadership, and an emphasis on quality assurance.
It was noted that there is professional jeopardy for untenured assistant professors who become involved in collaborative trials, since these trials are not viewed as independent research. Collaborative trials may require several years to complete, costing assistant professors valuable time for professional evaluation. However, such issues are not unique to this field. Junior scientists should be trained for clinical trials, suggesting that alternate credited pathways to tenure and career development may be needed.
Participants also noted several current unmet needs in stuttering research, including:
Assessment protocols that include factors such as attitude
Well-developed theoretical models that provide good mediators that explain for whom the effect occurs
Study designs that allow for clinician modifications to established protocols while maintaining integrity of treatment
An agreed-upon ratings scale for clinical trials that includes a basic outcome measure and established fidelity of the program(s) being tested
Participants agreed that the field must decide on the evidence needed in order to move forward with clinical trials. Also, researchers must be made aware of collaborative opportunities available, which are listed in myriad sources, including NIH Institutes and the newsletters of the various professional and voluntary associations interested in stuttering. Moreover, the stuttering research community must set lofty standards and goals--well-conducted stuttering research should be seen as a force that will make a difference in the health outcomes for the country. While competition remains important, the rich history in this field provides a solid base for collaboration and multi-center studies.
It was also noted that functional subtypes of behaviors associated with stuttering are necessary to assign appropriate treatments. Models and computational strategies should be created to address the complexity of the behavior. Moreover, patients and end users have a wealth of experiential knowledge and information that can inform basic science.
Participants offered several recommendations regarding scientific directions for research and clinical trials, including:
Conduct studies that investigate the science that underlies self-regulation
Encourage clinicians to teach self-management to patients who stutter
Incorporate client characteristics into treatment design
Improve social validation of outcomes for treatments under consideration (e.g., the social validation that parents are required to pay attention to a child’s behavior in the Lidcombe method)
Encourage the teaching of established methods such as the Lidcombe method during academic training
Conduct a large-scale trial to assess the Lidcombe method
Dr. Shekim then thanked participants for their input and provocative discussions. She noted that the NIDCD will evaluate the topics discussed at the workshop to inform Institutional initiatives to stimulate applications for stuttering research.
Summary issues discussed at the subgroup meeting included the neural bases of stuttering, the role of infant affect in stuttering traits, and clinical trials related to stuttering. Dr. Shekim charged participants with identifying the opportunities and next steps for the scientific community with respect to these areas. Discussion and next steps are detailed below.
Neural Bases of Stuttering: While evidence suggests neuroanatomic and neurophysiologic differences in those who stutter, many models are potentially useful for study. The songbird model offers insight into the neural underpinnings of learning, pruning mechanisms, and axonal guidance. Also, examination of human brain development during periods critical for the onset of stuttering may provide clues. However, neural factors can change rapidly; plasticity confounds the understanding of neural patterns. Current research has not provided a working model of the day-to-day variances in plasticity; thus, multiple imaging modalities will provide differing resolutions. Because stuttering appears to involve both a biologic component as well as a reaction to the aberrant biology, stuttering research must be conducted across species. While an animal model gives insight into the underlying neural motor components of stuttering, language develops within a context not illuminated by animal models.
Specific suggestions for the NIDCD included:
Establish a National Registry for stuttering, perhaps coordinated through the NIH, the NSA, and patient support groups, to enable prospective studies
Support longitudinal studies in high-risk families that will allow data collection prior to the onset of stuttering and identify early predictive factors for persistence or recovery from stuttering* (see below for details)
Coordinate with other NIH Institutes (e.g., the National Institute oo Mental Health) to develop a brain tissue bank
Employ coordinated NIH efforts (e.g., the Neuroscience Blueprint) in functional imaging with initiatives in other relevant areas, such as autism and dyslexia
Support animal studies that will yield hypothesis-driven models of neural pathways and their connectivity to stuttering
Support the development of structure-equation models of stuttering
Support research on linguistic planning and language simulation
Focus research on brain development and characteristics during the key period for stuttering onset, as well as in confirmed, aged persons who stutter who may exhibit more pronounced differences
*Participants also discussed specifics of such longitudinal studies, noting that control groups must be selected to include groups other than just those who do not stutter (e.g., recovered stutterers, twin siblings of persons who stutter). Studies that have yet to be carried out include cross-sectional, longitudinal studies that collect data on a single person across confirmed lengths of stuttering time. Studies also needed are those that establish candidate genes and identify dominant “forms” of stuttering. Sibling pairs and families will be resources for these efforts. The NIDCD could also coordinate with other organizations to establish cell lines from high-density families and maintain these as a bank that can be tied back into the registry. A stuttering phenotype is needed to model against the genetic data; identifying relevant genes and proteins may provide insight into the heritable characteristics of stuttering. Such studies will allow connections to be elucidated between the interacting subsystems of stuttering: motor, language, affect, and auditory-perceptual.
The Role of Infant/Pre-School-/School-Age Affect: One participant noted that temperament resembles a season of the calendar (e.g., fall, a period of general cooling down), whereas emotional reactivity/regulation more closely resembles the temperature fluctuation of individual days within the season. Understanding the Behavioral Approach and Behavioral Inhibition Systems is crucial to understanding how both the trait and state aspects of emotion contribute to the onset, development, and maintenance of developmental stuttering. There is a wealth of experimental data that examines factors that affect stuttering on a situation-specific basis. Also, many standard measures of temperament, as well as emotional reactivity/regulation, have been used successfully (e.g., observational, experimental, parent reports, questionnaires). Longitudinal studies are needed to examine the factors that exacerbate stuttering and contribute to its persistence. Cross-sectional studies, both descriptive and experimental, of trait and trait aspects of affect at or near the onset of stuttering would also shed light on earlier affective contributions to developmental stuttering.
Specific suggestions for the NIDCD included:
Convene a regular, annual, research-based, interdisciplinary consortium to discuss the direction that the field should develop, publicize training opportunities, and discuss publication issues. The consortium could be modeled on the Speech Foundation of America (SFA) meetings, yet with a specific focus on research. This single-day meeting could be funded by an NIH R13 mechanism and rotate among hosts including the NIH, the SFA, and the NSA. A Steering Committee could be employed to plan threads that connect consecutive meetings.
Support studies of state and trait aspects of affect, their relation to changes in stuttering, and their influence on initial and long-term treatment outcome.
Consider partnership opportunities with institutions abroad, such as the Michael Palin Center in London.
Encourage an active culture change that will embrace a communal vision that includes junior faculty and supports multi-site clinical trials.
Clinical Trials of Stuttering: Participants discussed issues related to clinical trials for stuttering, noting that the majority of children who stutter ultimately remit, thus confounding trial design. The Lidcombe method, an approach with a considerable degree of validated assessment criteria, should be considered, as data on the time course of recovery are available using this method. Clinical trials using the Silverman voice training were also suggested as one model for clinical trials for developmental stuttering.
Specific suggestions for the NIDCD included:
Coordinate clinical trials with the American Speech-Language-Hearing Association (ASHA) clinical trials group, which may serve as a resource for numerous aspects of clinical trials (e.g., packaging, analysis).
Consider multi-site clinical trials with broad-based outcome measures in adults who stutter.