2008/04/23

Johnson, K. (2006). Resonance in an exemplar-based lexicon: The emergence of social identity and phonology. Journal of Phonetics, 34, 485-499.

Presentation: Renee

Summary: Shelly

This study intends to deal with the sociophonetic issue of gender differences with the approach of speech perception theory. Gender difference has long been considered as a result of the difference in physical vocal tract length between males and females, because the longer vocal track will produce lower F₀ (males), and vice versa (females). However, some recent studies on cross-language comparisons have shown that the relationship between vocal tract length and gender F₀ difference are weak, and neither is the relationship between body height and gender F₀ difference. Therefore, it is very possible that gender difference is actually a result of performance, rather than solely physical differences. If gender difference is a matter of performance, then one could reasonably infer that there is a stereotype for how each gender should sound in our mental cognition so that we can perform according to that stereotype. Such stereotype, as proposed by this study, should be able to be reflected on listeners’ speech perception, and the present study aims to examine this phenomenon by using the exemplar-based model. The exemplar model they use here is the one proposed by Johnson (1997), who considered that listeners do not recognize the gender of the voice by normalizing the input to match the stereotype, but by matching the input with the exemplars in mind, and such matching would be strengthened after the resonance back and forth between the stereotype and the exemplar, which would lead to the eventual recognition. The reason for them to use the resonance model is because it could better explain the different performance and construction of gender among different languages and cultures, while the normalization model could not explain since it predicts the universal vocal track normalization algorithm. 

They first reviewed a study by Strand (2000), who examined how accurate the listeners would be in word recognition if they listened to speakers with stereotypical/nonstereotypical male/female voices, and the results showed that, when listening to stereotypical voices, listeners would more correctly identify the word, as compared to when listening to nonstereotypical voices. When given a nonstereotypical female voice, listeners would misidentify the gender of the speaker. Then, they use the stimuli of Strand’s experiment to feed the resonance exemplar-based model, and found that the results predicted by the model could well match the results in Strand’s study. Therefore they concluded that the social phenomenon of gender differences in speech could be explained quite well with the exemplar-based model.