The structure of variation in spontaneous American English stops

speech production;variability;stop consonants;compensation;acoustic cues

Khasanova, Alina

关键词: speech production;    variability;    stop consonants;    compensation;    acoustic cues;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/45322/Alina_Khasanova.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

One of the inherent characteristics of speech is that the same discrete units likephonemes, syllables, or words appear vastly different in the articulatory andacoustic domain.Acoustic variation can be thought of as the consequence of variousforces driving speech production, including production effort and the need tomaintain perceptual distinctiveness of contrastive elements.Variability is thennot just random noise, but on the contrary, is structured.Yet, little is knownabout the general structure of acoustic variation, with primary focus having been onits sources thus far.This work attempts to fill the gap through an investigationof the general structure in the acoustic variation of stops, to see how variation isrelated to a stop's phonological features.The primary focus of this thesis is on the relationship between acoustic cues to astop's contrastive phonological features, and whether that relationship ismaintained or changes under variation.Two patterns of cue co-variation areconsidered here: uniform co-variation, i.e. where both cues either get weaker orstronger, or compensatory cue co-variation, where one cue weakens as the otherstrengthens.The patterns of cue co-variation reflect distinct properties ofspeech, such as the continuity of articulation, on one hand, and the limitedcapacity for sequential independence between gestures that make up the stop.Withinthis framework, two major questions are investigated. The first questoin is whethertwo acoustic cues related through patterns of uniform or compensatory co-variation.Specific hypotheses tested are (a) that uniform co-variation of cues is moreprevalent than compensatory variation due to additional effort most likely requiredfor the latter; (b) that compensatory variation is likely to occur when necessitatedby communicative goals; and (c) that the burst portion of the stop is more likely tohost compensatory co-variation than the closure, because of different degrees offreedom available to the speaker.The second question is how phonological featuresof contrast are preserved under variation, as expressed in the parameters of thecues relationship.To answer these questions, linear regression models were fitted to the measurementsof acoustic duration and amplitude of the closure and burst intervals for thecanonical realizations of stop consonants drawn from the Buckeye speech corpus.Theslopes of regression lines were evaluated by simple slope analyses in order toassess their sign and magnitude: the sign carries information whether therelationship of cues is uniform or compensatory, while the magnitude reflects thedifferences in the categorical affiliation of the stop (place of articulation,voicing, and position in the word).Results indicate that phonological contrast isoften preserved under variation in either the direction or, more commonly, themagnitude of the slopes of the regression lines relating a pair of cues.Themajority of cue pairs are in uniform co-variation.Compensation is observed insubsets of the data for some of the cue pairs, most clearly in voiced stops andinvolving burst cues, i.e. when most needed perceptually or most easily achievedarticulatorily.The implications of these findings are discussed with respect tocurrent theories of speech production and, more generally, the search for invariancein speech.