Take-home lab exercise 4 on "Multidimensional Scaling with Mandarin fricatives "

Ling 600.01 Phonetic Theory

This take-home exercise uses the perceptual similarity data from the experiment that you did for the in-class lab in the first class meeting of Week 6 and the final report will be due at 5:00 on the day before the second class meeting of Week 8 (or at the beginning of that class if you don't care for feedback). The "final report" for in-class lab #6 (on "Judging perceptual distance") should be the Methods section for this exercise.

Part 1 -- The data.

The stimuli for this experiment that we did in class were ten tokens each of the syllables /fa, sa, ɕa, ʂa, xa/ (spelled in pinyin as "fa", "sa", "sha", "xia", and "ha") said in tone 1 (high level tone) by Fangfang Li. You rated the similarity among pairs of these fricative tokens on a five-point Lickert scale from "same" to "very different". [The Praat experiment script (and the stimuli) are available in the sibilantsMDS.zip file in the MDSresults directory under our class web site, in case you want to run this experiment with another person, such as your language consultant.]

The following two figures show spectra taken from a 20 ms slice at a point 50 ms back from the CV boundary. [In these two figures, the different syllables are written in pinyin in the upper left corner, whereas in the scatterplots that follow them, the different initial fricatives will be identified by "f" for the labiodental fricative in the syllable that is written in pinyin as "fa", "s" for the dental sibilant in the syllable that is written in pinyin as "sa", "c" for the aleolopalatal sibilant in the syllable that is written in pinyin as "xia", "r" for the retroflex sibilant in the syllable that is written in pinyin as "sha", and "x" for the velar fricative in the syllable that is written in pinyin as "ha".] The first figures show the energy in dB against Bark (calculated with the Praat BarkFilter command) and the second is phones against Bark (calculated with the Praat Excitation command).

This next figure shows the frequency of the highest amplitude point (the "peak") in each of the dB versus Bark spectra plotted in the first graph, as a function of the second formant just after the onset of the vowel.

This next figure shows the loudness (in sones) summed over all of the phone values in each of the phones versus Bark spectra plotted in the second graph, as a function of the third formant just after the onset of the vowel.

This next figure shows the results of applying the R multidimensional scaling command to the group averages.

These last two figure shows the analogous results for Yating Shih and Tingfen Lin, two bilingual speakers of Taiwanese (a Southern Min dialect spoken by 80% of the population of Taiwan) and Guoyu (the Taiwan national standard variety of Mandarin Chinese) who took this course in 2007 and 2008, respectively. (Yating is from the south and is a fairly balanced bilingual, whereas Tingfen is from the north and is Guoyu dominant.)

Other graphs for individual members of this year's class are available in the MDSresults directory.

Part 2 -- The report.

Write a brief report (no more than one single-spaced page, exclusive of figures) about this experiment, using the IMRaD ("Introduction, Methods, Results, and Discussion") format. The Introduction could be a single very short paragraph stating the relationship (as you understand it) between the kind of production-based vowel charts that you turned in for Field Project Report 3 and the vowel features picked out in the MDS plots of Terbeek's data that Johnson shows in Figure 6.11 of Acoustic and Auditory Phonetics, and saying that we are exploring the analogous relationship for fricative features and the (pscyho)acoustic parameters that cue them. The Methods section should be a description of the experiment that we did in class. (This section is the "final report" for that lab.) The Results section should include (an appropriate subset of) the above five figures and one or two more individual plots from the MDS, that you choose in order to make some point about effects of first language experience. The prose for this part should describe briefly the relevant patterns that you see in the figures (which you want to to pick up on in your Discussion). The Discussion section should interpret the patterns in terms of the features you chose. Also include some description of the point you are trying to make with the individual plot(s) that you chose.

p.s. Here is a page from Stevens (1989) that might help you understand the relationship between the spectral peak frequency taken from the frivative itself and the third formant frequency measured at the onset of the following vowel. [pdf file of page from Stevens (1989)]