Tara McAllister Byun

Purpose: Recent research suggests that visual-acoustic biofeedback can be an effective treatment for residual speech errors, but adoption remains limited due to barriers including high cost and lack of familiarity with the technology. This case study reports results from the first participant to complete a course of visual-acoustic biofeedback using a not-for-profit iOS app, Speech Therapist's App for /r/ Treatment. Method: App-based biofeedback treatment for rhotic misarticulation was provided in weekly 30-min sessions for 20 weeks. Within-treatment progress was documented using clinician perceptual ratings and acoustic measures. Generalization gains were assessed using acoustic measures of word probes elicited during baseline, treatment, and maintenance sessions. Results: Both clinician ratings and acoustic measures indicated that the participant significantly improved her rhotic production accuracy in trials elicited during treatment sessions. However, these gains did not transfer to generalization probes. Conclusions: This study provides a proof-of-concept demonstration that app-based biofeedback is a viable alternative to costlier dedicated systems. Generalization of gains to contexts without biofeedback remains a challenge that requires further study. App-delivered biofeedback could enable clinician–research partnerships that would strengthen the evidence base while providing enhanced treatment for children with residual rhotic errors. Supplemental Material: https://doi.org/10.23641/asha.5116318

Purpose: Single-case experimental designs are widely used to study interventions for communication disorders. Traditionally, single-case experiments follow a response-guided approach, where design decisions during the study are based on participants' observed patterns of behavior. However, this approach has been criticized for its high rate of Type I error. In masked visual analysis (MVA), response-guided decisions are made by a researcher who is blinded to participants' identities and treatment assignments. MVA also makes it possible to conduct a hypothesis test assessing the significance of treatment effects. Method: This tutorial describes the principles of MVA, including both how experiments can be set up and how results can be used for hypothesis testing. We then report a case study showing how MVA was deployed in a multiple-baseline across-subjects study investigating treatment for residual errors affecting rhotics. Strengths and weaknesses of MVA are discussed. Conclusions: Given their important role in the evidence base that informs clinical decision making, it is critical for single-case experimental studies to be conducted in a way that allows researchers to draw valid inferences. As a method that can increase the rigor of single-case studies while preserving the benefits of a response-guided approach, MVA warrants expanded attention from researchers in communication disorders.

Purpose: This study documented the efficacy of visual–acoustic biofeedback intervention for residual rhotic errors, relative to a comparison condition involving traditional articulatory treatment. All participants received both treatments in a single-subject experimental design featuring alternating treatments with blocked randomization of sessions to treatment conditions.Method: Seven child and adolescent participants received 20 half-hour sessions of individual treatment over 10 weeks. Within each week, sessions were randomly assigned to feature traditional or biofeedback intervention. Perceptual accuracy of rhotic production was assessed in a blinded, randomized fashion. Each participant's response to the combined treatment package was evaluated by using effect sizes and visual inspection. Differences in the magnitude of response to traditional versus biofeedback intervention were measured with individual randomization tests.Results: Four of 7 participants demonstrated a clinically meaningful response to the combined treatment package. Three of 7 participants showed a statistically significant difference between treatment conditions. In all 3 cases, the magnitude of within-session gains associated with biofeedback exceeded the gains associated with traditional treatment.Conclusions: These results suggest that the inclusion of visual–acoustic biofeedback can enhance the efficacy of intervention for some individuals with residual rhotic errors. Further research is needed to understand which participants represent better or poorer candidates for biofeedback treatment.

Collecting independent listeners’ judgments of speech accuracy/intelligibility is an essential component of research on speech disorders. Raters can be trained clinicians, students in speech-language pathology, or naïve listeners (now commonly recruited online via crowdsourcing platforms such as Amazon Mechanical Turk/AMT). However, limited comparison data exist to guide researchers in determining which rater population to use. We describe a study (Hitchcock et al., in prep) in which 2,256 tokens of English /r/ at the word level, produced by five children receiving intervention for /r/ misarticulation, were rated in a binary fashion using the online platform Experigen. Raters were certified clinicians (n = 3), students in speech-language pathology (n = 9 unique listeners per token), or naïve listeners recruited on AMT (n = 9 unique listeners per token). Interrater reliability was higher when comparing modal ratings between clinicians and students (Cohen’s kappa=.73, CI=.7-.77) than between clinicians and naïve listeners (Cohen’s kappa=.64, CI=.6-.68). However, students offered none of the speed/efficiency advantages available through AMT. We posit that students would be more motivated to complete ratings if they received feedback on accuracy. As a future direction, we propose a hybrid crowdsourcing platform/educational game, designed to sharpen important perceptual skills while also motivating students to contribute valid ratings for research data.

A current need in the field of speech pathology is the development of reliable and efficient techniques for the purpose of evaluating changes in speech production over the course of treatment. The industry standard for scoring speech is time consuming and expensive, as it involves aggregating perceptual ratings across expert listeners. As techniques for automated measurement of speech improve, acoustic measures have the potential to play an expanded role in the clinical management of speech disorders. The current study asks which of several acoustic measures of children’s productions of English /r/ corresponds most closely with ratings given by trained listeners. This study fits a series of ordinal mixed effects regression models to a large sample of children’s /r/ productions that had previously been rated by three trained listeners (speech-language pathologists). Controlling for age, sex, and allophonic contextual differences, the acoustic measure that accounted for the most variance in speech rating was F3-F2 distance, normalized relative to a sample of age- and gender-matched speakers. Therefore, this acoustic measure is recommended for use in future automated scoring of children’s production of rhotic targets.

It is known that some adult listeners have more sharply defined perceptual categories than others, and listeners with more precise auditory targets are also more precise in their production of contrasts. There is additionally evidence that children who have not yet mastered production of a contrast show diminished performance on perceptual measures of the same contrast. To date, however, few studies have investigated developmental perception-production relations using the fine-grained measures typical of adult studies. Existing evidence suggests that perception and production can be closely connected in development, but this relationship may break down as perception and articulation mature at different rates. This study evaluated perception and production of the English /r-w/ contrast in 40 typically-developing children aged 9–14. Perceptual sensitivity was measured with a logistic function fitted over responses in a forced-choice identification task using two synthetic 10-step continua from rake to wake. Participants also produced rhotic and non-rhotic words. Across participants, there was a significant correlation between perceptual acuity and rhoticity in production, although this effect was only observed for one of two continua tested. These results provide preliminary evidence compatible with the hypothesis that children with a more refined auditory target for a sound also produce that sound more accurately.

Recent evidence suggests that the incorporation of visual biofeedback technologies may enhance response to treatment in individuals with residual speech errors. However, there is a need for controlled research systematically comparing biofeedback versus non-biofeedback intervention approaches. This study implemented a single-subject experimental design with a crossover component to investigate the relative efficacy of visual-acoustic biofeedback and traditional articulatory treatment for residual rhotic errors. Eleven child/adolescent participants received ten sessions of visual-acoustic biofeedback and ten sessions of traditional treatment, with the order of biofeedback and traditional phases counterbalanced across participants. Probe measures eliciting untreated rhotic words were administered in at least 3 sessions prior to the start of treatment (baseline), between the two treatment phases (midpoint), and after treatment ended (maintenance), as well as before and after each treatment session. Perceptual accuracy of rhotic production was assessed by outside listeners in a blinded, randomized fashion. Results were analyzed using a combination of visual inspection of treatment trajectories, individual effect sizes, and logistic mixed-effects regression. Effect sizes and visual inspection revealed that participants could be divided into categories of strong responders (n=4), mixed/moderate responders (n=3), and non-responders (n=4). Individual results did not reveal a reliable pattern of stronger performance in biofeedback versus traditional blocks, or vice versa. Moreover, biofeedback versus traditional treatment was not a significant predictor of accuracy in the logistic mixed-effects model examining all within-treatment word probes. However, the interaction between treatment condition and treatment order was significant: biofeedback was more effective than traditional treatment in the first phase of treatment, and traditional treatment was more effective than biofeedback in the second phase. This is consistent with existing theory and data suggesting that detailed knowledge of performance feedback is most effective in the early stages of motor learning. Further research is needed to confirm that an initial phase of biofeedback has a facilitative effect, and to determine the optimal duration of biofeedback treatment. In addition, there is a strong need for correlational studies to examine which individuals with residual speech errors are most likely to respond to treatment.

Any account aiming to provide a comprehensive picture of children's acquisition of speech must consider both the development of the phonological grammar and the maturation of the structures and motor skills used to implement the sounds of a language. Much previous literature has been marked by a tendency to draw sharp demarcations between motor and grammatical influences, or to assert that all of child speech can be reduced to one or the other. This paper argues that it is neither necessary nor desirable to segregate speech‐motor development from grammatical development when modeling speech acquisition, because they are fundamentally intertwined. The paper focuses on bringing together two literatures that have evolved largely independently. The first explores how speech‐motor patterns practiced during babbling come to be disproportionately represented in the lexicon in children's earliest stages of meaningful speech. The second posits that abstract elements of phonology – segments, features, and constraints – can be understood to emerge from generalizations over stored memory traces at a more holistic level. We argue that an emergentist model of phonological learning can be enhanced by incorporating the insight that memory traces of strings that have been heard and produced are encoded more robustly than strings that have only been heard.