Deriving individualised /r/ targets from the acoustics of children’s non-rhotic vowels

Abstract

In visual-acoustic biofeedback for rhotic errors, learners are guided to match the third formant (F3) location to a visual target on a real-time acoustic spectrum. As the acoustic properties of correct English /r/ differ across speakers, this study aimed to improve target selection by investigating the validity of individualised targets derived from children’s non-rhotic vowels. A previously proposed prediction formula was adjusted using data from a child normative sample and tested in two groups of children. Study 1 found that predicted values were unexpectedly higher than actual F3 values in children whose /r/ errors had been remediated. To understand this discrepancy, Study 2 applied the formula to typically developing children and found that predicted values were also higher than actual F3 values, suggesting that different normative data might better represent the current samples. An updated formula is proposed, which can be used to generate individualised targets within acoustic biofeedback applications.