Prosodic prominence in English

In English, certain words are perceptually more salient than other neighboring words. The perceptual salience is signaled by acoustic cues. Prominent words are higher, longer, or louder than nonprominent words in English. Perceptual prominence is associated with meaning of a word in discourse context. Prominent words are usually new or contrastive information, while nonprominent words are given or noncontrastive information. This dissertation addresses English prominence in two separate studies. The first study investigates the prosodic prominence in relation to pitch accents, acoustic cues, and discourse meaning of a word in a public speech. The second study examines the cognitive representation of prosodic contour in a corpus of imitation. Linguists claim that the information status of a word determines the types of pitch accents in English. Prior research informs us about prominence (1) in relation to the binary given-new distinction of lexical givenness, and (2) in minimally contextualized utterances such as question-answer prompts or excerpts from a corpus. The assignment of prominence, however, can vary in relation to referential meaning as well as lexical meaning of a word in natural, more contextualized speech. This study examines the prosodic prominence as a function of pitch accents, acoustic cues, and information status in a complete public speech. Information status is considered in relation to referential, lexical givenness and alternative-based contrastive focus. The results show that accent type is probabilistically associated with information status in this speech. The accent assignment differs between referentially vs. lexically given words. Despite the weak relationship between information status and pitch accents in the speech of the speaker, non-expert listeners perceive prominence as expected: they are more likely to perceive prominence on words carrying new or contrastive information or words with high or bitonal pitch accents. Surprisingly, the listeners perceive acoustic cues differently depending on the information status or accent types of a word. Based on these results, the first study suggests that (1) the relationship between information status and accent type is not deterministic in English, (2) lexical givenness differs from referential givenness in production and perception of prominence, and (3) perceived prominence is influenced by information status, pitch accents, acoustic cues, and their interaction. The second study examines how an intonational contour is represented in the mental lexicon of English speakers. Some linguists find that speakers are able to reproduce the phonetic details of intonational features, while in other research speakers are better at reproducing intonational features than imitating phonetic details of an utterance. This study investigates the domain of intonational encoding by comparing several prosodic domains in imitated utterances. I hypothesize that the domain which best captures the similarity of intonational contour between the model speaker and imitators is the target of imitation, and that imitation can be considered as the domain of intonational encoding in cognitive representation. The results show that the f0 distance between the model speaker and imitators is best explained over an intermediate phrase. Based on these results, the second study proposes that speakers encode a time-varying f0 contour over a prosodic phrase in their mental lexicon and supports the exemplar encoding of intonational contour.LimitedAuthor requested closed access (OA after 2yrs) in Vireo ETD syste