Download presentation
Presentation is loading. Please wait.
Published byAlberta Little Modified over 9 years ago
1
The New Normal: Goodness Judgments of Non-Invariant Speech Julia Drouin, Speech, Language and Hearing Sciences & Psychology, julia.drouin@uconn.edu Dr. Emily Myers, Department of Speech, Language, and Hearing Sciences Dr. Rachel Theodore, Department of Speech, Language, and Hearing Sciences Introduction In a dynamic world where relating to others is an important aspect of life, being able to speak and understand people is essential for effective communication. However, it is often the case that we have variation in our speech signal. Examples of Speech Signal Variations: Boston accent ‘r’ dropping ( “car” “cah”) Southern accent vowel breaking (“cat” “cayut”) Speaker with a lisp (“sing” “thing”) Speaking rate (people tend to talk faster on the East Coast) With all of this variation in the speech signal we encounter, how is it that we are able to keep a stable representation of the acoustic speech signal with each new talker we encounter? Kraljic & Samuel (2005) suggested two techniques of how this occurs. Two Theories of Learning New Talkers: 1. One idea is that a listener holds multiple representations of every speaker they encounter. 2. The other idea of “retuning” or “perceptual learning” suggests a change in phonemic boundaries after exposure to an ambiguous phoneme. Speech sound categories, like other cognitive/perceptual categories, have a graded internal structure, with some members of the category considered more representative of the category than others. We asked if perceptual learning caused a shift in internal category structure, or if the perceptual adjustment was limited to the boundary region. Speech Sound Category Structure: Hypothesis: Perceptual learning of ambiguous token causes a shift in goodness criterion and demonstrates a shift in internal speech sound category structure. Methods All participants (n=17) were undergraduate students at the University of Connecticut enrolled in PSYC 1100 or PSYC 1103. All participants were between the ages of 18-21 and were fluent English speakers. The paradigm (replication of well-established training paradigm (Kraljic & Samuel, 2005) consisted of 3 experimental tasks in which the participants were asked to listen to different tokens and make decisions about them. 1. Lexical Decision Task: Participants heard 100 words and 100 non-words in random order and were asked to use corresponding computer keys to indicate if what they heard was a word or a non-word. 40 of these tokens were “critical words” and contained an ambiguous sound word medially 2. Goodness Judgment Task: Participants heard a continuum of sounds and were asked to focus on the middle sound and rate how good of an example it was of an “s” sound. Participants in both the /s/ and /sh/ training groups received the same instructions for the task. Participants heard 6 points on a continuum, played randomly over 10 trials each. “rehearsal” “rehearshal” 3. Phoneme Identification Task: Participants listened to the same stimuli used in the Goodness Judgment Task, however were asked if what they heard sounded more like “asi” or “ashi.” Graphs: Figure 1. Accuracy during training was analyzed with a break- down in each token category. Figure 2. mean /s/ goodness ratings for both /s/-biasing and / ʃ /- biasing groups as a function of continuum point. Figure 3. Mean /s/ identification responses for both the /s/-biasing and / ʃ /- biasing groups as a function of percent /s/ continuum point. Results: Goodness Judgment Task: Two- way mixed ANOVA: No main effect of group, therefore similar ratings for each continuum point between /s/ and / ʃ / biasing groups: [F (1,15)=.950, p=.345]. Significant main effect of continuum point, [F (5,75)= 2.53, p=.036] Phoneme Identification Task: Two-way mixed ANOVA: No main effect between training groups [F (5, 75)= 1.298, p=.274] Significant main effect of training group: [F (1,15)= 7.331, p=.016] Significant main effect of continuum point: [F (5, 75)= 54.974, p <.01] Speaker 1Speaker 2Speaker 3 Speech Sound Category before new talker Speech Sound Category after new talker Poor /s/ Excellent /s/ Poor /s/ “ashi” “asi” “rehear?al”
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.