Abstract
This paper evaluates “mean responsetime”(MRT),a method used in previous studies to relate physiological evidence (recordings of electromyographic activity in the cricothyroid and sternohyoid) to acoustic evidence (fundamental frequency). Rather than averaging over tokens before correlating these signals, we calculated the bestresponse time (RT) for each token, and evaluated the pattern of variability across utterances. Furthermore, rather than correlating over whole utterances, we correlated electromyographic activity (EMG) to fundamental frequency (F0) only over intervals defined in terms of linguistically significant events in the F0 trace, identified using a linguistically motivated model of English intonation. Steep changes in the F0 tended to have better correlation coefficients than shallow ones, which we relate to the physiological model by noting the complex of components contributing to both signal types. Also, the distribution of lead times was easier to interpret when the two tones delimiting the analysis domain had some tight temporal relationship specified by the intonational phonology. Finally, lead times tended to vary as a function of what preceded the target rise or fall. In short, averaging over signals before analysis obscures patterns of variation in the data which may lead to new insights and to new directions for research.
Get full access to this article
View all access options for this article.
