Over the last decade, neurofeedback has become increasingly popular as a potential treatment option for various neurodevelopmental and neurological disorders. As a completely noninvasive procedure, electroencephalographic (EEG) neurofeedback uses auditory or visual methods to provide an individual with information derived from scalp-electrode recordings of his or her own brain activity. The feedback information may relate to the spatial distribution of activation across the brain or the energy distribution across specific frequency bands. To date, no studies have been undertaken to investigate the potential use of EEG neurofeedback in the clinical management of stuttering. Stuttering is a high-prevalence (1% of the population) speech disorder that often develops into a chronic, life-long problem with severe social implications. The ultimate causes are unknown, but recent studies have revealed structural and functional neural deficits that impact sensorimotor functioning. The present project therefore aims to test the feasibility of developing EEG neurofeedback procedures that may reduce the severity of stuttering in adult participants. Initial testing with 19 data channels and BCI2000 software focuses on reliably discriminating between rest and speech or nonspeech motor activity and designing conceptually relevant feedback displays that represent the spatial or energy distribution of the recorded brain signals.  If successful, this research could have important clinical implications in the treatment of stuttering.

Previous studies have shown that neurologically healthy subjects learn to generate adaptive motor commands when either sensory feedback (e.g., visual or auditory information) or the movements themselves (e.g., arm movements for reaching or jaw movements for speech) are experimentally perturbed - a phenomenon known as sensorimotor adaptation. In parallel, a different line of research has shown that for both nonspeech and speech movements the central nervous system (CNS) differentiates between self-generated and externally-caused sensory input by comparing actual and predicted inputs - a detected mismatch causes the input to be attributed increasingly to external sources. Here, we investigate these two phenomena in combination by means of two experiments that examine whether sensorimotor adaptation in speech production is diminished, or even abolished, when an auditory feedback perturbation (shifted formant frequencies) is either delivered with increasing delays or made increasingly large. Based on our own pilot data and recent findings from other labs, we hypothesize that increasing either the delay or the extent of the formant perturbation will cause the CNS to increasingly interpret the perceived error as an external manipulation rather than a production error that requires adaptive corrections in future trials. For each experiment, acoustic analyses are used to determine adjustments in subjects’ produced formant frequencies from a baseline phase (unaltered feedback) to the perturbation phase (formant-shifted feedback) across conditions varying in delay and amount of shift in the auditory feedback signal.

Despite a surge in research efforts over the past two decades, scientific understanding of the mechanisms underlying motor learning remains limited. Although many procedures have been designed to assess an individual’s ability to update motor commands in an altered environment, few attempts have been made to design procedures that assess, and possibly improve, an individual’s ability to learn a completely novel mapping of motor commands and sensory consequences. This is unfortunate because the latter approach has potential clinical applications: various brain disorders cause sensorimotor deficits, and learning alternative movement strategies may aid recovery. Whereas the above studies focused on hand or arm movements, our own research focus is on the speech motor system. The present project aims to design protocols to test and quantify the ability of the speech motor system to learn neural representations of a novel link between motor commands and sensory results. We make use of 3D electromagnetic motion capture technology that tracks 3 position coordinates and 2 angles for up to 9 sensors on the tongue, lips, and jaw; thus for a maximum of 45 dimensions. These 45 dimensions are mapped onto 2 dimensions of movement for a cursor displayed on a computer monitor. As subjects are instructed to move the cursor, their performance (i.e., learning of the novel mapping) is quantified. We will also test individuals from a clinical population (such as those with motor speech disorders or stuttering) and analyze their performance over time to design the most optimal techniques and procedures for quantifying the speech sensorimotor system’s ability to learn the complex relationship between a high-dimensional movement space and a 2D visual space.

Despite major advancements in our understanding of the neural systems involved in persistent developmental stuttering (PDS), the causal mechanisms underlying acquired neurogenic stuttering (ANS) remain poorly understood. ANS typically occurs after stroke, and its primary symptoms resemble those of PDS (i.e., within-word dysfluencies). However, individuals with ANS often show no adaptation effect and no fluency-enhancement with altered auditory feedback. To date, little is known about differences and similarities between ANS and PDS in terms of the loci of stuttering—that is, the locations of stuttering moments within spoken utterances. Brown’s research on PDS has shown that stuttering is more likely to occur on (1) longer words, (2) words that occur early in the sentence, (3) content words, and (4) words that start with a consonant. In this study, we therefore compared ANS and PDS with regard to the distribution of stuttering moments as quantified by Brown’s word weights (for each stuttered or fluent word, 0-4 points are assigned based on the aforementioned four factors). As expected, adults with PDS showed greater word weights for stuttered versus fluent words. Individuals with ANS showed a highly similar pattern with greater word weights for stuttered versus fluent words. Calculating word weights for each factor separately (word length, sentence position, grammatical class, initial consonant) revealed that the ANS group also did not differ from the PDS group in the relative influence of each individual factor. Thus, although there are several known differences between acquired neurogenic and developmental stuttering, the loci of stuttering appear to be influenced by similar factors in both disorders. Of those factors, word length and grammatical class had the strongest influence on the occurrence of dysfluencies in both groups. These findings can be used to generate new hypotheses regarding the neural basis of fluent and dysfluent speech production.