full-fledged pidgin

(I’m finishing these projects out of order (wavesurfer took me longer to deal with than Praat or Audacity, so the two WaveSurfer posts have languished on the largely-completed-but-unpublished stack for a while now). Exercise 5 is below. Anyone who might be reading this and who is not, in fact, one of the professors for LSA 317: Experimental Phonology should probably move along. I’m not going to provide enough context for this to make sense, sorry. )

I can definitely see that wavesurfer would be a an incredibly useful application if it worked well. As it is, though, I’ve found it wildly frustrating (even on a mac).

question 1

The [h] seems to result in increased airflow escaping during glottal pulse. We can see the [h] just before the vowel in that area of increased oral airflow with no activity on either the EGG or the differential of the EGG. At this point, the vocal folds are separated and subglottalic pressure must be increased sufficiently to cause audible frication. I’d guess that the breathy sound on the vowel in [hi] is the result of the vocal folds and chest muscles aiming for (but not quite hitting) their articulatory targets in the shift from [h] to V and back again.

question 2

The open quotient (OQ) starts out relatively small, rises to XX, and returns to the lower range. The .5 mark indicates equilibrium ( vocal folds equally open and closed during a glottal pulse ) and occurs at the center of the modal-voiced periodic sound. Given the pattern shown in Childers & Lee’s Table III, one would expect the [h] in “has” to have been produced with vocal fry or `creaky’ voice. Indeed, the spectrogram (in Praat) also suggests this (as do my ears).

question 3

No, the [h] between .65 and .72ms is not voiced. The oscillations in the airflow and acoustic record must be the result of turbulence within the vocal tract and frication at the vocal folds. I’m imaging the vocal folds are being held close enough together to allow frication, tightly enough to prevent modal voicing. Perhaps this gesture explains the apparent creaky voice a few ms later?

question 4

Yes, the [z] in “has” is voiced: there is modal voicing with an OQ of around .5 (and actually is .5 at at least one point). The airflow increases when the vocal folds are open, decreases when they close, and the waveform shows a periodic (if low amplitude) wave. This is very different from the [h] in question 3 where the vocal folds are not being allowed to vibrate and the spectrum shows `fuzzy’ lines indicative of a range essentially random frequencies present in the sound.

question 5

(Boy, this would be much easier to do if I could listen to what I’m seeing without switching back and forth between Praat and WaveSurfer.) Looking at the VV transition between “the only” (from 1.08 to 1.18ms) I don’t see any signs of laryngealization at all. The OQ values seem to indicate modal voicing, the spectrum looks regular, the airflow looks regular, and I can’t hear anything but a smooth transition.

There’s a big drop in amplitude during the transition of “only applicable” between about 1.22 and 1.32ms, but the OQ measures are all around 0.5 again. I don’t see any evidence of laryngealization.

“Applicable answer” between 1.88 and 1.98ms has, again, modal voicing with an OQ of 0.5 and no signs of laryngealization even in the (D)EGG signal.

question 6

The closure for the [p] in “applicable” is from about 1.44 to 1.52ms It’s hard to tell exactly the timing with the uncalibrated airflow measurements, but it looks like there is a small burst of air initiated at 1.438s. This would be when the speaker’s lips close — closing the lips brings the mandible up. This reduces the volume of the vocal tract which, in turn, increases the oral air pressure and excess air is forced into the nearest area of lower pressure (e.g. the world). The closure of the lips and resulting increase of vocal tract pressure removes the pressure differential across the glottis to sustain voicing and the vocal folds return (beginning at T=1.441s) to a neutral state. The effect of all this in the acoustic signal is an interruption of the periodic sound at approximately T=1.435s, a small (probably inaudible) noise associated with the air rushing from the mouth, and then a cessation of all sound until the closure is released at T = 1.525s. The next closure of the glottis is at 1.594s for a VOT of .069s or 69ms.

[k] in “keep” looks a bit different. Whereas in [p] it was possible to see the airflow decrease prior to a measurable change at the glottis, these changes are much more synchronized for [k]. I imagine this is due to the shorter distance between the glottis and the velar closure. This leaves both (a) more air still moving forward in front of the closure and (b) a much smaller vocal tract volume. The smaller volume removes the requisite pressure differential for voicing sooner so the vocal folds return to a neutral position even as air from the phonation of the preceding vowel is still escaping the vocal tract. The closure is released at 2.561 (airflow resumes, the acoustic signal resumes at 2.563s) and the next glottal closure is at 2.621s for a VOT of 6ms.

[t] in “to” has a short closure duration ( 43ms ) between the cessation of airflow and glottal activity at 2.798s and the resumption of airflow (the release) at 2.841s. The first glottal closure after the release happens at 2.938s for a VOT of 97ms.

question 7

In a glottal stop I’d expect to see the increase in conductivity associated with a glottal closure (I just realized that this EGG plot must be “upside down” like an ERP plot — with the negative readings above the zero crossing and the positive readings below). I’d expect to see something that looks like a glottal closure that gets held since, really, that’s what a glottal stop is, but I don’t see that in either stop. Indeed, it looks to me like the glottis remains especially open during the production of these stops.

The distance of the microphone from the speaker’s mouth.

Okay, so the handout falls short of actually asking the question, but I’m assuming it’s like my mom and odd sharing of facts in a declarative mode should be interpreted as a question (or, possibly, order).

I don’t know what the weather was like when these measurements were taken, but assuming a temp of 21 C and air pressure of 1 atm (if, as my friend Markus points out, the measurements weren’t taken in the mountains on a hot as crazy day with a tornado buffeting the lab door) the mic was approximately 0.688 meters (.002 * 344) from the speaker’s glottis. Assuming an average male vocal tract length of 20.2 cm this means that the mic was 48.6cm from the speaker’s mouth (at least when T = 1.425s and the speaker’s vocal folds closed to create the sound that the mic sampled at T = 1.427s). 48.6cm seems 10 or 15cm further than I normally place a mic. Maybe I’m doing it wrong?