While recent research suggests that visual biofeedback can facilitate speech production training in clinical populations and second language (L2) learners, individual learners' responsiveness to biofeedback is highly variable. This study investigated the hypothesis that the type of biofeedback provided, visual-acoustic versus ultrasound, could interact with individuals' acuity in auditory and somatosensory domains. Specifically, it was hypothesized that learners with lower acuity in a sensory domain would show greater learning in response to biofeedback targeting that domain. Production variability and phonological awareness were also investigated as predictors. Sixty female native speakers of English received 30 min of training, randomly assigned to feature visual-acoustic or ultrasound biofeedback, for each of two Mandarin vowels. On average, participants showed a moderate magnitude of improvement (decrease in Euclidean distance from a native-speaker target) across both vowels and biofeedback conditions. The hypothesis of an interaction between sensory acuity and biofeedback type was not supported, but phonological awareness and production variability were predictive of learning gains, consistent with previous research. Specifically, high phonological awareness and low production variability post-training were associated with better outcomes, although these effects were mediated by vowel target. This line of research could have implications for personalized learning in both L2 pedagogy and clinical practice.
Purpose
Research in communication sciences and disorders frequently involves the collection of clusters of observations, such as a series of scores for each individual receiving treatment over the course of an intervention study. However, little discipline-specific guidance is currently available on the subject of building and interpreting multilevel models. This article offers a tutorial on multilevel models, using notation from the R statistical software, and discusses their implications for research in communication sciences and disorders.
Method
This tutorial introduces multilevel models and contrasts them with other methods to analyze repeated measures data, such as repeated measures analysis of variance or standard linear regression. It also provides guidance on interpreting the components of a multilevel model and selecting the best-fitting model. Finally, these models are illustrated through an analysis of real data from a study of speech production training in second-language speakers of English.
Conclusions
As a flexible method that can increase the rigor of modeling for clustered data, multilevel modeling represents an important tool for research in communication disorders. Given their increasingly prominent role in the analysis of experimental data in communication sciences, it is important for researchers to be familiar with the basics of building and interpreting these models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.