Abstract
This paper presents some methodological implications for the segmentation of final V+/l/ and V+/r/ sequences in American English stressed monosyllables. Between the two segments in the sequences a transitional schwa-like element characterized by variable durational and spectral values as a function of the preceding vowel, the following consonant and speaking rate can be identified. Given the dynamic nature of this element, problems related to boundary placement often arise. A segmentation method based solely on spectrographic observation and auditory corroboration has proven in previous studies to be too subjective to be reliable. A more objective method based on first derivative curve extraction, which provides us with first derivative formant traces that show peaks of formant change given by velocity maxima and minima, is more suitable for our purposes. However, the decision as to whether to choose F1, F2 or F3 traces as reference points for boundary placement, together with the presence of too many peaks in some cases and of not enough peaks in other cases, poses problems to the segmentation procedure and makes it necessary for the subjective method of segmentation based on spectrographic observation and auditory corroboration to come into play. Both methods thus complement each other.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call