Joanne L. Miller
Neal J. Pearlmutter, David A. DeSteno
Date of Award
Doctor of Philosophy
Department or Academic Unit
College of Arts and Sciences. Department of Psychology
Psychology, Psycholinguistics, Speech perception, Talker-specificity
Phonetics (Acoustic), Speech
Talkers differ in the acoustic-phonetic information used to convey individual consonants and vowels. For many years, talker differences in phonetic properties of speech were considered as problematic noise for the perceptual system. Indeed, traditional accounts of speech perception posit that talker-specific phonetic detail is removed from the signal in the process of accessing abstract linguistic representations (e.g., Studdert-Kennedy, 1976). These accounts are challenged, however, by findings from multiple research domains indicating that talker-specific phonetic detail is retained in memory and can be used to facilitate speech processing (e.g., Goldinger, 1996, Nygaard & Pisoni, 1998). In order to develop a theoretical account of speech perception that describes how listeners accommodate talker differences in phonetic properties of speech, additional data on talker-specific phonetic detail are necessary from both the perception and production domains. This dissertation examines talker-specific phonetic detail for the case of voice-onset-time (VOT) in word-initial voiceless stops. Previous research has shown that, when holding other influences on VOT constant, talkers differ in their characteristic VOTs with some talkers having shorter VOTs than other talkers (Allen et al., 2003). Other research has shown that listeners are sensitive to such talker differences in VOT in that they can learn, for a given voiceless stop, that one talker produces short VOTs and a different talker produces longer VOTs (Allen & Miller, 2004). This dissertation consists of two projects that extend these findings. The first project examined the scope of generalization underlying listener sensitivity to talker differences in word-initial VOT. Two experiments were conducted. In both experiments, two groups of listeners were differentially exposed to characteristic VOTs for two talkers, one talker produced short VOTs and the other talker produced longer VOTs. Exposure was provided during training phases in which listeners heard both talkers produce one voiceless stop consonant, either /p/ or /k/, in the context of a word (e.g., pain or cane). In test phases, listeners were presented with a short-VOT and a long-VOT variant of the word presented during training as well as a novel word that began with a different voiceless stop than presented during training. In both cases, listeners were asked to select which of the two VOT variants was most representative of a given talker. Across the two experiments, the phonological distance between the training and novel words was manipulated, the words formed minimal pairs (pain and cane) in Experiment 1 and non-minimal pairs (pain and coal) in Experiment 2. The same pattern of results was found in both experiments. Specifically, listeners selected the VOT variant in line with exposure during training not only for the word presented during training, replicating earlier findings (Allen & Miller, 2004), but also for the novel word. Moreover, for both the minimal pair and non-minimal pair cases, the magnitude of listener sensitivity to characteristic VOTs was the same for the novel word and the training word. These findings indicate that learning a talker's characteristic VOTs does not necessitate exposure to each phonetic segment, rather, there is transfer across similar segments. In order to better inform theoretical accounts of the types of exposure listeners may require to transfer talker-specific phonetic detail across various dimensions, additional data from the production domain are necessary. To this end, the second project examined talker-specific phonetic detail in speech production. As stated above, talkers differ in VOT in word-initial stop consonants (Allen et al., 2003). Previous research also indicates that VOT is robustly affected by contextual influences, including speaking rate and place of articulation (e.g., Lisker & Abramson, 1964, Kessinger & Blumstein, 1997). This project examined whether these contextual influences on VOT are themselves talker-specific. Across two experiments, many tokens of labial /p/, alveolar /t/, and velar /k/ were elicited from talkers across a range of rates. All tokens formed syllables consisting of the voiceless stop followed by the vowel /i/ (e.g, /pi/). VOT and vowel duration (a metric of rate) were measured for each token. Hierarchical linear modeling analyses showed that: (1) VOT increased as rate slowed for all talkers, as expected, but the magnitude of the increase varied significantly across talkers, thus the effect of rate on VOT was talker-specific, (2) the talker-specific effect of rate was stable across a change in place of articulation, and (3) for all talkers VOTs were shorter for labial than for velar stops, as expected, and there was no significant variability in the magnitude of this displacement across talkers, thus the effect of place on VOT was not talker-specific. These findings provide basic information on how two contextual factors influence VOT at a talker-specific level and, in so doing, point to constraints on how listeners might accommodate such contextual variation when customizing phonetic categories for an individual talker's speech.
Rachel Marie Theodore
Theodore, Rachel Marie, "Some characteristics of talker-specific phonetic detail" (2009). Psychology Dissertations. Paper 9. http://hdl.handle.net/2047/d10017094
Click button above to open, or right-click to save.