Effect of stress and rate on carticulation: an analysis of the variability of F2-onsets
MetadataShow full item record
Human speech production is characterized by a ‘blending’ of consonants and vowels such that there is a lack of one-to –one correspondence between the acoustic speech signal and the perceived segmental phoneme. The overlapping articulation responsible for this lack of invariance is known as coarticulation. The ubiquitous context induced variability presents a theoretical challenge to researchers who seek to understand how the brain establishes categorical equivalent classes for sounds that exhibit variation due to neighboring contexts. The test case for such coarticulatory blending is the production of stop consonant + vowel sequences. The locus equation metric (LE) derived by Lindblom (1963) and investigated by Sussman and colleagues has sought an auditory solution to this non-invariance problem. LEs plot the dynamic changes in the onset of the F2 transition, relative to its offset in the vowel nucleus of a given stop consonant across a wide array of vowel contexts. The LE scatterplots reveal, at the categorical level, a lawful orderliness to stop + vowel sequences that has eluded investigators examining individual speech tokens. This dissertation sought to assess the LE metric for stop consonants across two additional sources of suprasegmental perturbation-emphasis/stress and speech rate. In the context of these two linguistically relevant processes, the studies were designed to explore the ‘fate’ of context-induced variability that has been shown to disappear in LE scatterplots. Classic LE analyses have revealed closely clustered and linear distributions of data coordinates implicating clear coarticulatory differences across stop places of articulation (/bdg/) with vowel contextual effects no where to be seen. This research used VCV tokens of American English to (1) elicit additional variability in V.CV productions by altering speaking conditions, contrasting emphasis between V1 and V2, and by increasing speech tempo, (2) applying the traditional LE metric in order to assess their sensitivity to suprasegmental effects and (3) deriving analysis techniques capable of uncovering the absorbed variability masked by the traditional LE.