Understanding VOT

Question

I am a third year bachelor student of Linguistics. It would be nice if I don't get mean comments, because I genuinely do not understand what I am about to ask.

I have to write a paper on phonetic categorisation in perception. Everything that we will write in this paper has to be based on William F. Ganong's (1980) paper "Phonetic Categorization in Auditory Word Perception".

So, what we had to do is run a similar experiment, for a very short time (1 week), and choose a different stimuli and language than that presented in the paper. I chose Slovenian, and my stimuli was on the /b/-/p/ voicing distinction. Therefore, the only thing my stimuli differ in is voicing.

I recorded the stimuli with my own voice.

I was told by my professor that the only thing I need to do to create a /b/-/p/ continuum is change the VOT of those phonemes. So, I thought I did that, but now I am getting really confused about VOT.

Oh btw, we were supposed to modify our phonemes in steps, so that at a certain point a /b/ would sound like a /p/ by changing the "VOT".

But the thing is...Slovenian doesn't even have aspiration, and therefore it doesn't have VOT!!!! So what I was actually modifying my stimuli by cutting away the voicing bit of /b/ and adding it to /p/ in several steps until they created a continuum of /b/ to /p/.

To sum it up: -until now I thought that what I was modifying was VOT

-I have not actually modifying VOT, but cutting out the voicing bits from /b/ and adding it to the /p/ and vice versa.

My questions then are:

-is voicing and VOT the same thing?

-what was I modifying? Is there any study that looks at a simple voicing distinction?

-am I doomed?

Answer 1

The quick answer is that voicing and VOT are not same thing. Voicing is a physical phenomenon ("the vocal folds vibrate"), and VOT is a measurement ("time from release of a consonant to voicing"). You might use VOT as a means of detecting that a consonant is voiced, in case the consonant is voiced throughout its closure (the VOT value will be negative).

You probably cannot reproduce the Ganong experiment with another language, which used stimuli created with Haskins Laboratories FOVE program. But you can do something similar, and perhaps manipulation of natural speech tokens can be subjected to suitable manipulations: or you can use a different synthesis method (presumably using Praat). I can't find anything on Slovenian VOT, but from this article we know what the expected range of VOT in Serbian is (as well as Serbian English): mean VOT for Serbian /k/ is 56 msc (Serbian English /k/ = 71 msc: comparable to Lisker & Abramson 80 msc). Reporting about voiced consonants in English is more problematic since sometimes stops are voiced throughout and sometimes they are not but have a short positive voice onset time. But even if you average them together, the result should be that you get a small possibly negative VOT for a voiced consonant.

Doom is a harsh evaluation, but you might be existence-challenged. Ganong's experiment relied on the fact that there is a point of (small) positive VOT where English speakers perceive stimuli as voiced. You may be able to distinguish Slovenian /p/ and /b/ on that basis: or perhaps Slovenian speakers do not distinguish /p/ and /b/ based on short vs long VOT lag. Either result is fine (which is probably the point of doing this), as long as your interest is "what are the facts?" and not "how do I support theory X?".

The most challenging problem as far as I can see is creating the stimuli. You will spend too much time on the project if you try to create "task/dask, tash/dash" type pairs; but it's not clear to what extent your experiment is supposed to test lexical vs. phonetic processing. You might create non-word monosyllables [pa, ba], [pi, bi] using Praat, but that assumes you can do synthesis with Praat. Simple cut and past editing is very problematic (i.e. don't). The thing that does not make sense to me is the presumption that you could do this in a week.

Dutch is an example of a language with negative VOT for voiceless stops, likewise Spanish and Hungarian. You could read the article by van den Berg, "Perception of voicing in Dutch two-obstruent sequences: Covariation of voicing cues" (Lingua, 1989) to see a languages that is more like Slovenian. I also assume you have a supply of Slovenian-speaking subjects.