Sensorimotor learning and control in individuals who stutter
Kim, Kwang S
MetadataShow full item record
Despite accumulating evidence that stuttering is associated with deficiencies in sensorimotor integration, the exact mechanisms underlying this disorder of speech fluency remain not only unknown but controversial. Since our research group’s first formulation of a broad theoretical framework and a set of testable hypotheses (Max, 2004; Max, Guenther, Gracco, Ghosh, & Wallace, 2004), numerous studies have produced mostly supporting but also some seemingly incompatible results. Here, I present new results from a series of closely related studies designed to fill several critical gaps in our understanding of the sensorimotor mechanisms underlying stuttering. Specifically, I examined in both children and adults who stutter several aspects of speech and limb sensorimotor learning by means of adaptation paradigms in which participants learn to update the planning of future movements in response to an experimentally introduced feedback perturbation. In addition, in adults who stutter, I also examined aspects of sensorimotor control by recording kinematic data for unperturbed articulatory movements and applying statistical methods that yield estimates of the relative contributions of feedforward and feedback control mechanisms. The experiments presented in Chapter I were directly motivated by recent work, from our own lab and others, demonstrating poor performance in auditory-motor learning among adults who stutter (AWS) as compared with adults who do not stutter (AWNS) (Daliri, Wieland, Cai, Guenther, & Chang, 2018; Daliri & Max, 2018; Sengupta, Shah, Gore, Loucks, & Nasir, 2016). However, it remained controversial whether such poor sensorimotor learning can be causally related to the mechanisms that are responsible for breakdowns in speech fluency. For example, in adult speakers, multiple years of coping with stuttering may have affected the functioning of neuromotor control processes, and this, in turn, may affect assessments of sensorimotor learning. In fact, Daliri et al. (2018) specifically reported not finding a difference in auditory-motor learning between children who stutter (CWS) and children who do not stutter (CWNS), suggesting that no learning deficit was present during the childhood years closer to the onset of stuttering. In addition, it remains unclear whether sensorimotor learning limitations in individuals who stutter are speech-specific or whether they also affecting nonspeech effector systems such as the upper limb during reaching and grasping. sensorimotor system. Therefore, the first study (Chapter I) in the series aimed to examine both speech auditory-motor and reach visuomotor learning in adults and children who stutter. Results suggest that both children and adults who stutter showed statistically significant limitations in auditory-motor adaptation to formant-shifted auditory feedback. In fact, this limitation was even more profound in children than in adults and in younger children (3-6 years of age) versus older children (7-9 years of age). Between-group differences in the adaptation of reaching movements performed with rotated visual feedback were subtle, but still statistically significant, for adult participants. In children, even nonstuttering children showed limited visuomotor adaptation, and, as a result, there was no difference between the stuttering and nonstuttering groups. From these results, I conclude that sensorimotor learning is deficient in individuals who stutter, and that at least substantial speech auditory-motor learning problems are already present at a young age near the onset of stuttering. Thus, atypical motor learning may play a critical role in the fundamental mechanisms underlying stuttering. In the first study described above, participants’ auditory feedback was manipulated with a commercially available vocal effects processor. Prior to the adoption of a different methodological approach that implements the feedback perturbation in user-adjustable Matlab (The Mathworks) software (i.e., the Audapter package, Cai, Boucek, Ghosh, Guenther, & Perkell, 2008; Cai, Ghosh, Guenther, & Perkell, 2010; Cai, Ghosh, Guenther, & Perkell, 2011), it became clear that the literature contained much implausible information about the minimum time latencies that can be achieved when combining this software with a personal computer and professional or consumer-grade audio interfaces. It is crucial, however, for researchers to be cognizant of the fact that the reproducibility of results from speech auditory-motor adaptation studies depends on accurately measuring and reporting the overall feedback loop latency that was in effect during the experiment. Therefore, in a second study (Chapter II), I measured the total feedback loop latencies (including both hardware and software latency) for various hardware and software combinations. Results showed that hardware-specific latencies were overlooked in several published reports, but that these latencies are not at all negligible for some of the tested audio interfaces (adding up to 15 ms delay). In addition, the measured total latencies were also generally larger than claimed in the literature. Therefore, the manuscript included as Chapter II emphasizes that the actual total latency (hardware plus software) needs to be correctly measured and described in all published reports. Furthermore, the study also demonstrated that the use of non-default parameter values can improve Audapter’s own processing latency without negative impact on formant tracking. The paper concludes with several recommendations to improve feedback latency in speech auditory-motor adaptation paradigms. The third study (Chapter III) examined which sub-processes of learning may be implicated in stuttering individuals’ auditory-motor learning difficulties. Recent studies in upper-limb motor control indicate that sensorimotor learning involves at least two distinct components: (a) an explicit component that includes intentional strategy use and presumably is driven by target error, and (b) an implicit component that updates an internal model without awareness of the learner and presumably is driven by sensory prediction error. The presented study constitutes an initial attempt at dissociating these components for speech auditory-motor learning in AWS versus adults who do not stutter (AWNS). I developed a novel paradigm to obtain information on participants’ awareness and intent related to adaptive articulatory behavior when speaking with formant-shifted feedback. First, results replicated previous findings that such auditory-motor learning is indeed limited in AWS. Second, in neither group did participants report any awareness at all of changing their productions in response to the perceived auditory feedback. These findings suggest that speech auditory-motor adaptation relies exclusively on implicit learning, and, therefore, that the limited adaptation found in AWS is likely due to poor implicit learning mechanisms. This conclusion is in direct agreement with other recent lines of evidence implying sensory prediction deficits in stuttering. Chapter IV includes a study investigating sensorimotor control rather than learning in AWS versus AWNS. The foundation for this work lies in the well-established observation that the central nervous system (CNS) has the remarkable ability to both pre-plan a desired movement prior to movement onset (i.e., feedforward control) and quickly and efficiently adjust the ongoing movement based on sensory input (i.e., feedback control). Our laboratory’s theoretical perspective on stuttering suggests that an atypically large weight on online feedback-based control may render the sensorimotor system unstable—potentially leading to sustained or repetitive articulatory movements and, thus, speech dysfluencies (Max, 2004; Max et al., 2004). My previous work already demonstrated the feasibility of using kinematic data to examine feedforward versus feedback mechanisms in the control of unperturbed speech movements in nonstuttering adults (Kim & Max, 2014). For the new study presented here, I used the same approach to analyze the kinematic data of jaw and tongue movements in AWS and AWNS. Specifically, the study determined how well initial kinematic landmarks (peak acceleration, peak velocity) could predict final movement kinematics (movement extent). Results show generally large correspondence between initial and final kinematics in both groups, suggesting that speech movements are mostly under feedforward control. However, this relationship was statistically significantly weaker for stuttering speakers. In addition, estimates of feedback-driven adjustments in movement duration were significantly larger for AWS than for AWNS in the second half of the movement, suggesting that stuttering individuals’ regulate duration online in order to compensate for less efficient planning of the final kinematics. Thus, the findings add support for the hypothesis that stuttering individuals may over-rely on online feedback control and less on feedforward control (Max et al., 2004). Taken together, the findings from these studies provide novel insights into the sensorimotor integration impairments underlying stuttering. The sensorimotor learning studies demonstrated that sensory prediction errors may not be correctly integrated for subsequent movement planning in both CWS and AWS, and that this limitation reflects less than optimal implicit learning processes. The sensorimotor control study confirmed that AWS are indeed more dependent on online feedback for immediate within-movement corrections. This necessary but inefficient control strategy may ultimately lead to the repetitive corrections or postural fixations that are perceived as stuttering moments during speech production (Max et al., 2004; Max & Daliri, 2019).
- Speech