Re: Location of phonemic boundaries (Richard Pastore )


Subject: Re: Location of phonemic boundaries
From:    Richard Pastore  <pastore@xxxxxxxx>
Date:    Tue, 12 Aug 2008 17:50:36 -0400
List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>

Dear List: Richard Warren suggests using the Gating Paradigm to define the perceptual beginning and end of phonemes. This paradigm is used quite widely. However, the validity of Gating Paradigm results depends upon the implicit assumption that later or earlier portions of the stimulus are perceptually irrelevant to the needed recognition process. Specifically, this is an assumption that there is minimal Backward Recognition Masking (BRM) and Forward Recognition Contrast (FRC), phenomena largely studied in the 1970s. We have a manuscript coming out in the next issue of Perception & Psychophysics that, at least for relative pitch judgment of tones, challenges this assumption. We measured the duration threshold for judgment of tones separated by a Musical Fifth. When asked to judge relative pitch for tones in isolation, the target tones need to be about 8 ms in duration. When followed by one or more contextual tones, the target tones need to be at least 25 ms, and sometimes as long as 45 ms., in duration. The relevance to the Gating Paradigm is discussed in the manuscript. We currently are preparing a manuscript describing a completed follow-up study where the task was to judge the direction of (pitch) change for linear frequency glides, measuring thresholds in terms of size of glide frequency change. Compared to threshold for isolated glides, there are very significant (and very large) amounts of BRM when the glide is followed by one or more steady-state tones, and there can be significant FRC when the glide is preceded by such a tone. If these findings with simple stimuli generalize in any way to the processes needed for phoneme identification, then Gating Paradigm results may not be accurate indicators of the perceptual boundaries of phonemes. Dick Pastore Richard Pastore Distinguished Service Professor Department of Psychology Binghamton University Binghamton, NY 13902-6000 ----- Original Message ----- From: "Richard Warren" <rmwarren@xxxxxxxx> To: <AUDITORY@xxxxxxxx> Sent: Tuesday, August 12, 2008 12:42 PM Subject: Location of phonemic boundaries Dear List, Jim Miller pointed out on 8 August that “It is well established that the acoustic information used by a listener to identify a consonant or a vowel is overlapping and distributed acoustically across a considerable span of time.” He indicated that although some have attempted to identify the acoustic locations of consonants and vowels in running speech, they have for the most part failed since coarticulation extends well into adjacent phonemes. But if the question is changed from “acoustic” boundaries to “perceptual” boundaries, the task becomes rather easy. When a sentence is abruptly terminated, the last speech sound is easily perceived. By using an arbitrary starting point before the beginning of a recorded sentence, and moving the time of the cutoff through the sentence, it is easy to map the perceptual beginning and end of each phoneme within a few milliseconds. We have been using this procedure for several decades.


This message came from the mail archive
http://www.auditory.org/postings/2008/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University