Page 1 of 1

What shape is he vowel space, really?

Posted: Fri Aug 26, 2016 1:23 pm
by alice
OK, I know this won't revolutionise phonetics, but it's been at the back of my mind for a while, and now it's ready to come out.

Most of you will point to something like this and reply "it's a quadrilateral; why do you need to ask?" However, there may well be grounds for treating it as triangular; see here for one argument. (There's lots other interesting stuff on that blog, btw.) Note, too, that DependencyPhonology treats vowels as mixtures of , [a], and qualities, which fits better into a triangle than a quadrilateral.

So: Has the IPA vowel diagram, in some sort of Sapir-Whorfian subterfuge, conditioned us to a suboptimal view of the physical and auditory reality? Could the IPA thus be seen as analogous to QWERTY, in an admittedly non-obvious way?

(I highly recommend praat, too; it can become addictive. I'm still not sure why I can't produce a fully back though.)

Re: What shape is he vowel space, really?

Posted: Fri Aug 26, 2016 2:22 pm
by zompist
I would normally trust William Labov on these issues, and as it happens he agrees with you— the vowel space is triangular, not quadrilateral.

However, the page you linked is limited by, in effect, looking at just two speakers. Labov has examined hundreds. His Principles of Linguistic Change contains vowel charts for many of them; not all of them are as strikingly triangular as Wells and Jones. It seems more common for the bottom of the triangle to be blunt rather than sharp.

Re: What shape is he vowel space, really?

Posted: Fri Aug 26, 2016 3:25 pm
by mèþru
Isn't this essentially the same thread as this one you posted earlier?

Re: What shape is he vowel space, really?

Posted: Fri Aug 26, 2016 3:28 pm
by Grunnen
Geoff Lindsay has made an interesting proposal for conseptualising the vowel space as fundementally a space of frequencies of the first and second formants of the acoustic speech signal. He's used that as the basis for a proposal for to change the IPA vowel chart, and he arrives at an essentially triangular layout as well. Personally I like to think his view on this subject has helped me understand vowels better. You can find his proposal on his blog: http://englishspeechservices.com/blog/the-vowel-space/

Re: What shape is he vowel space, really?

Posted: Sat Aug 27, 2016 2:06 pm
by alice
Grunnen wrote:Geoff Lindsay has made an interesting proposal for conseptualising the vowel space as fundementally a space of frequencies of the first and second formants of the acoustic speech signal. He's used that as the basis for a proposal for to change the IPA vowel chart, and he arrives at an essentially triangular layout as well. Personally I like to think his view on this subject has helped me understand vowels better. You can find his proposal on his blog: http://englishspeechservices.com/blog/the-vowel-space/
That's rather bizarre, because that page - which I actually linked to in my post! - is what prompted this thread in the first place.

And yes, this is very like my earlier thread , which I forgot I'd started :-(

Re: What shape is he vowel space, really?

Posted: Sat Aug 27, 2016 2:24 pm
by Tropylium
Related empirical question: does anyone happen to know of languages other than Danish that contrast all four of /ɛ œ ʌ ɔ/?

For four high qualities you can look at e.g. literally anything Turkic for /i y ɯ u/ (also e.g. Scandinavian for /i y ʉ u/, or Toda for /i y ɨ u/). Finding four close-mid qualities is harder, but at least a few langs (e.g. Estonian, Selkup, Yakut) have the full /e ø ɤ o/. But four open-mid qualities is a tougher call yet.

Re: What shape is he vowel space, really?

Posted: Sat Aug 27, 2016 3:44 pm
by mèþru
Considering that the Scandinavian languages exist in a dialect continuum, perhaps Norwegian or Swedish have some dialect with similar vowels
The Wikipedia article for Swedish says that the standard dialect has /ɛ œ ɵ ɔ/, but it also says the first three are actually [ɛ̝, œ̝, ɵ̞].

Re: What shape is he vowel space, really?

Posted: Sun Aug 28, 2016 4:21 pm
by vokzhen
Tropylium wrote:Related empirical question: does anyone happen to know of languages other than Danish that contrast all four of /ɛ œ ʌ ɔ/?
Maastrichan has /ɛ œ ə ɔ/ that are just above the cardinal mid-open line, though /ə/ is limited to unstressed syllables so I'm not sure it's fully contrastive. Cantonese appears to have /ɛ œ ɐ ɔ/, but according to the vowel diagram on Wikipedia /ɛ œ ɔ/ are actually between mid and mid-high, i.e. /e ɛ/ etc could be transcribed [e̝ e̞] or maybe even [e̝ e].

Re: What shape is he vowel space, really?

Posted: Mon Aug 29, 2016 11:30 am
by Rory
alice wrote:Most of you will point to something like this and reply "it's a quadrilateral; why do you need to ask?" However, there may well be grounds for treating it as triangular; see here for one argument. (There's lots other interesting stuff on that blog, btw.) Note, too, that DependencyPhonology treats vowels as mixtures of [ i], [a], and qualities, which fits better into a triangle than a quadrilateral.

The answer depends on what you mean by "the vowel space". There are at least four possible interpretations:

1) The physical space that the articulators move in (i.e. the inside of the mouth).
2) The physical acoustical signal, defined in terms of meaningful dimensions of variation (usually the first and second formants, but also F3 and a host of other spectral cues can be linguistically meaningful).
3) The perceptual space of auditory representations of the vowels.
4) The abstract phonological space in terms of features (or elements, or whatever your favourite theory has).

It is clear that we're only interested in how to characterize the space in terms of meaningful variation in vowel identity. So, while the physical space (1) is in reality a three-dimensional polygon, it suffices to describe articulatory movements in terms of tongue raising and fronting, ignoring the third dimension. Even so, this space is clearly not an exact triangle or a trapezoid, it is a round oblong roughly approximating a triangle or trapezoid, the exact details of which will vary from person to person. Arguing over whether the space (1) is a triangle or a trapezoid is like arguing whether the human head is the shape of a soccer ball or a rugby ball - neither answer is correct, but the truth has elements of both, and the reality varies from person to person.

Similar constraints apply to the space (2), where individual variation in production will alter the shape of the acoustic space. Again, approximately triangular or trapezoidal is a fine characterization. I'm basing this on my own experience in analysing data from hundreds of vowel spaces from years of running experiments.

(3) is an interesting space because we don't really have a good way to measure auditory perception of a multidimensional percept like vowel quality. (Unidimensional characteristics like pitch or loudness are a little easier.) Incidentally, while the IPA's chart is presented as being motivated by an articulatory basis, if you trace the history of it it's really actually an auditory motivation. When it was first put together, the ideas about what the tongue was doing were really just educated guesses (as it was impossible to see inside the skull during articulation during the early 1900s). I recall there being some discussion of this in Peter Ladefoged's introductory textbook Vowels and Consonants.

Finally, (4) is a different question entirely. Under some views, the question is meaningless as the "vowel space" is particular to a language. Thus, in English, which has an /æ/~/a/ contrast, the vowel space is trapezoidal, whereas Spanish has a triangular vowel space. If, however, you believe in universal phonological features, then I guess that you should say the vowel space is trapezoidal, and if you're a dependency/government phonologist, then it's triangular. Personally, I don't think the idea of universal (i.e. innate) features (/elements) is worth the time of day, but I accept that others do.

So: Has the IPA vowel diagram, in some sort of Sapir-Whorfian subterfuge, conditioned us to a suboptimal view of the physical and auditory reality? Could the IPA thus be seen as analogous to QWERTY, in an admittedly non-obvious way?

I'd like to encourage you to continue to be sceptical about the relationship between IPA and physical reality!

(I highly recommend praat, too; it can become addictive. I'm still not sure why I can't produce a fully back though.)

Are you American? Many Americans (and others) have fronted /u/ and fine the "cardinal" quite hard to do. Try starting with [o] and raising it.

Re: What shape is he vowel space, really?

Posted: Mon Aug 29, 2016 12:21 pm
by mèþru
He's actually Scottish, from what I understand from his website: https://web.archive.org/web/20080511160 ... about.html

Re: What shape is he vowel space, really?

Posted: Mon Aug 29, 2016 1:03 pm
by alice
mèþru wrote:He's actually Scottish, from what I understand from his website: https://web.archive.org/web/20080511160 ... about.html
Rory shouldn't need to ask, anyway; he's met me twice, both times at ZBB meets!

Re: What shape is he vowel space, really?

Posted: Mon Aug 29, 2016 2:41 pm
by opipik
Tropylium wrote:Related empirical question: does anyone happen to know of languages other than Danish that contrast all four of /ɛ œ ʌ ɔ/?
Arammba has /ɛ œ ə̆ ɜ̆ ʌ ɔ/.

Re: What shape is he vowel space, really?

Posted: Tue Aug 30, 2016 8:57 am
by Rory
alice wrote:
mèþru wrote:He's actually Scottish, from what I understand from his website: https://web.archive.org/web/20080511160 ... about.html
Rory shouldn't need to ask, anyway; he's met me twice, both times at ZBB meets!
Hi Geoff! That's embarrassing, I didn't realise you'd changed your username again. Sorry! In any case, Scottish English /u/ is also pretty far fronted.

Re: What shape is he vowel space, really?

Posted: Tue Aug 30, 2016 10:03 am
by finlay
alice wrote:
mèþru wrote:He's actually Scottish, from what I understand from his website: https://web.archive.org/web/20080511160 ... about.html
Rory shouldn't need to ask, anyway; he's met me twice, both times at ZBB meets!
geoff? where tf did you go all these years?

actually, a better question is how long have you been back and when did you last change your username because i didn't recognize you....

Re: What shape is he vowel space, really?

Posted: Tue Aug 30, 2016 2:57 pm
by alice
I reckon I've been "back" with a changed username for a few months.

Anyway, the problem I now have with , and back vowels generally, is that I can't seem to get a decent second formant - it jumps around between about 600 Hz and 1500 Hz. Whereas other vowels it's nice and clean. What am I suffering from?

Re: What shape is he vowel space, really?

Posted: Mon Oct 24, 2016 9:23 pm
by Nachtuil
The triangular vowel space is quite appealing to me as someone who struggles to keep the central vowels apart. I am inclined to be swayed by the apparent method behind the placement too. Very fascinating to read. Two things struck me, how at least one language has a vowel harmony link between /a/ and the mid central and how I also can't keep bar i and unrounded u apart so it is appealing they are essentially the same thing.