|
|
Research Interests| Collaborators | Publications| Grants & Awards | My Research in simple words
Research Interests: I am primarily interested in the different theories of speech perception and production. I am strongly swayed by Fowler's idea of Direct Perception of speech as well as ideas of embodied perception.Outlandish as these ideas seemed to me at first sight, I feel these ideas could not only help us understand aspects of speech perception but would naturally lead us in understand speech perception in the context of speech production and vice versa. Liberman's analysis by synthesis perhaps was the first "embodied" speech perception
I am also interested in studying Language as used in communicative settings rather than under atypical conditions. However, much of my research until now has been restricted to the Phonetic level studied under atypical conditions of language use!
My current research endeavor involves differentiating gestural and auditory theories of speech perception. My experiments involve an in depth examination of the phenomenon of Compensation for Coarticulation (CfC) and its implication for various theories of speech perception. We hope to develop a unitary theory of CfC that explains both the Lexical and the classical forms of CfC. The (ambitious) goal here is to understand how sensory processes (e.g. auditory contrast), perceptual processes (gestural compensation) and prior knowledge ( lexical compensation effects) interact.
Collaborators: I have been fortunate to have the opportunity to work with and learn from the following brilliant minds.
Faculty: Carol Fowler (Advisor), Jim Magnuson (Co Advisor).
Fellow Graduate Students: Stephen Tobin, Anne Olmstead, Paula Silva, Karen Aicher and Anuenue Kukona
Undergraduate Students: Ann Kulikowski, Sara Pomerantz and Brittney Bauer
Publications
2006
Viswanathan, N., Magnuson, J.S., & Fowler, C.A. (2006, November). Compensation for coarticulation : Comparing Contrast and Gestural theories. Poster presented at the 47th Annual meeting of the Psychonomic Society.(pdf of poster)
Viswanathan, N., Magnuson, J.S., & Fowler, C.A. (2006, September). Disentangling gestural and auditory contrast accounts of compensation for coarticulation. Proceedings of the Ninth International Conference on Spoken Language Processing (Interspeech 2006 - ICSLP), Pittsburgh, PA 861-864. (pdf of paper)
Viswanathan, N., Magnuson, J.S., & Fowler, C.A. (2006, June). Compensation for coarticulation : Three theories compared. Journal of the Acoustical Society of America, 119 (5), 3241 (pdf of poster)
Grants and Awards:
National Science of Foundation (3 year award).
Compensation for Coarticulation: Implications for the Basis and Architecture of Speech Perception
(PI) : Jim Magnuson. Co-PI(s) Carol Fowler and Navin Viswanathan
Awarded the Alvin Liberman Prize, March 8th 2007 at Haskins Laboratories
My Research in simple words
We produce and understand speech almost effortlessly. This ease belies the complex process that underlies both the production and understanding speech. In the field of speech research there exists a debate that has raged on for about 50 years now. Alvin Liberman first proposed that people understand speech by understanding how it was produced in the first place. This view called the gestural viewpoint states that listeners understand speech by attending to the speaker's vocal movements. This view is in contrast to the view that people understand speech by mapping patterns of sounds into sound categories in their head (called an auditory view).
In our lab we try to determine which viewpoint can better explain listeners' behavior when they are exposed to certain special stimuli. These special stimuli are artificially created sounds such as "da" and " ga". Now these sounds can be created such that we can obtain a range of sounds that start out sounding like a da and slowly sounds more and more like a ga. This range of sounds is called a continuum, specifically a da-ga continuum in this case.
Now if we take the sound that is right in the middle of this continuum, the listener hears it 50% of the time as a da and the other 50% a ga. Prior research has shown that this proportion can vary depending on what sound occurs before. Specifically if the listener hears an "al" then the ambiguous sound is more likely to be heard as a "ga" ( the proportion shifts e.g. to 70% ga and 30% da). Similarly if the sound that occurs is an "ar" then the ambiguous sound is now likely to be heard as a "da" ( the proportion of judgments now flip in the opposite direction 70% da and 30% ga). This phenomenon was first interpreted by Virginia Mann to be an indication that the listeners were tracking speakers vocal movements. The explanation was that when a speaker says "al" his tongue is in the front of his mouth and he cannot get back to the correct place to produce a good "ga". So when the listener hears an "al" and the ambiguous sound the listeners reason that the ambiguous sound must have been a ga that could not be produced well leading to more "ga" responses. The story is similar for more "da" responses after "ar".
Lotto and Kluender challenged this interpretation. In a very intriguing set of experiments, they showed that the frequency pattern of the syllables lead to the phenomenon observed before. They suggested that this phenomenon is due to a form of contrast. To provide an analogy, if we had three pails of water: cold, lukewarm and hot. Depending on which bucket we dipped our hand before, the lukewarm water will seem hot or cold due to the contrast. The energy pattern of these sounds is such that the ambiguous (akin to lukewarm) is either called a da or ga depending on the sound that occurred before.
Our research involves setting up conditions such that only one of these explanations can hold good. One such case was observed by using a Spanish "r" ( like the Taco bell Carrrnes commercial !) as one of the sounds. Because this sound is produced in the front of the mouth, a gestural theory would expect the ambiguous sound to be heard as a "ga" more often. The frequency distribution of this sound is such that we expect the ambiguous sound occurring after this to be heard more often as a "da". This situation thus allows us to choose between theories as they make starkly different predictions. Our results showed that the ambiguous sound was more often heard as a "ga" preliminarily supporting a gestural explanation. We are in the process of running more experiments to better understand this phenomenon in general.
|
