Candidacy Project Defense: Allison Johnson
Title: "RoBURSTness: Quantifying Robustness of the /t/-/k/ Contrast in Children and Adults"
Abstract: Children develop speech gradually throughout childhood, and they continue to refine articulatory movements and phonological knowledge even after they have produced a sound or sound contrast. In research and clinical practice, speech is most often characterized using coarse-grained transcription methods; however, broad phonemic classifications or simple judgments of correct/incorrect fail to capture variations in productions. Measurable sub-phonemic differences in children’s speech, such as covert contrasts, undifferentiated lingual gestures, or intermediate productions, are difficult to classify using only transcription, but they have been quantified using fine-grained acoustic or perceptual measures. These subtle differences have important clinical implications in diagnosing speech sound disorders, developing focused treatment approaches, and determining prognosis for therapy.
Many studies in phonetics have searched for acoustic correlates to characterize speech sounds or phonetic dimensions. The present study directly compared two gradient-scale measures—an acoustic measure (Centroid) and a psychoacoustic measure (Peak ERB)—in their utility for quantifying robustness of the /t/—/k/ contrast in adults (n = 21) and 2-3-year-old children (n = 163). For adults, both measures were highly successful in differentiating /t/ and /k/, but Centroid was significantly better than Peak ERB for classifying productions in back-vowel contexts. Similar results were found for the children. There was also an interaction effect of target consonant in front-vowel contexts: Centroid better differentiated /t/ and /k/ when the target consonant was /t/, but Peak ERB better differentiated /t/ and /k/ when the target consonant was /k/.
Overall, children showed a less robust contrast compared to adults, even for productions transcribed as correct. These findings provide further evidence that children acquire speech gradually and continue to refine productions even after sounds are considered acquired. For children’s productions that were transcribed as correct, the wide range in robustness of contrast highlights the importance of supplementing transcription with fine-grained measurements to capture developmentally relevant sub-phonemic differences in children’s speech. Limitations of the study and future directions are discussed.