I’ve made some improved measurements for nine Scottish English vowels and made them public on my academic homepage as pdfs.
There are two charts:
- One shows measurements taken from minimal pairs (here) – my friendly Scot, a female native speaker of Scottish English from the Central Belt, read aloud several sets of words which differed only in the vowel (eg heat, hit, hate, het, etc; coot, coat, cot, cut, etc). When the phonetic context is controlled like this, it makes the formant values much clearer to interpret
- The other shows measurements taken from the same speaker reading aloud a short story which was composed in such a way that it contains at least one example of most of the relevant vowels (shown here, alongside the minimal pair values for comparison)
I feel a lot happier with these charts, compared to the original one here cobbled together from single repetitions of words where the phonetic context was uncontrolled, and the subsequent one here which still suffers from lack of control although it made use of a greater number of tokens.
Won’t you get a record of how the person thinks they speak (or ought to speak), if they have to specifically differentiate sounds like that, rather than have them recorded in natural speech?
LikeLike
There is some risk of that, although it’s minimised when the speaker is used to being recorded and feels no need to alter the way they speak when they’re being recorded.
There is an ongoing debate about how comparable “lab speech” is with natural speech, but there are disadvantages and limitations with both. Natural speech is more difficult to record, especially if you’re determined to only record the most natural and unaffected speech, and what you gain in terms of naturalness you lose in terms of replicability and experimental control.
Since my original aim was to find a more representative chart for contemporary Scottish English than the published one/s, it was important to get values that can be as broadly generalisable as possible. I think that the the figures plotted in those charts can be confidently expected to be very similar to what I’d get if I recorded any other speaker with roughly the same characteristics as this speaker (gender, age, education/SES, regional background, etc).
Maybe I’ll consult with my speaker and see if I can get consent to make the audio files available and put your mind at rest :)
LikeLike
In a recent Language Log post Mark Liberman has given an argument for using ‘artificially designed’ material for speakers to read out. It’s to answer a different & more complex question, measuring cross-linguistic final lengthening, but i think you can generalise many of the points –
http://itre.cis.upenn.edu/~myl/languagelog/archives/005127.html
LikeLike
Incidentally, it turns out that there’s also a vowel chart available in an article on the acquisition of Scottish English by Scobbie et al in the 2006 QMUC Working Papers. There’s a reference there to Abercrombie 1979, but i’m not sure if the chart itself comes from there or if the ref just covers what’s in the text:
Acquisition of Scottish English Phonology: an overview
Click to access scobbie_et_al%202006%20WP7.pdf
LikeLike