Dialectal and age-related acoustic variation in vowels in spontaneous speech

Abstract

Our knowledge of the acoustic characteristics of vowels is based mainly on productions obtained under fine experimental control (e.g., [hVd] tokens produced in isolation). Analyzing spontaneous speech data has its obvious challenges because the vowels of interest occur in various segmental and prosodic contexts and are additionally affected by sudden changes in speech tempo. These variations may compromise the accuracy of interpretation of linguistic phenomena such as sound change. In this study we examine if more natural productions of vowels in two dialects (as compared to those produced in citation form and read speech) show evidence of the existence of cross-generational vowel changes—including corresponding changes in spectral dynamics. A subset of vowels from a large corpus of spontaneous conversations was analyzed. The vowels occurred in variable consonantal contexts in both mono- and polysyllabic words. The obtained patterns of vowel change were consistent with those in read and citation form speech. Although the measured spectral changes were smaller due to shorter vowel durations, the dialect-specific nature of formant dynamics was maintained. These results demonstrate that the patterns of vowel change, including variation in formant dynamics, do not diminish under the circumstances of greater contextual/prosodic variability found in spontaneous speech.