[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
Sat Dec 13, 2014 6:43 pm (PST) . Posted by: "And Rosta" andjamin
>
> I meant what's the limit on the number and type of phonetic contrasts
> and hence on the size of the phonological segment inventory -- for
> a lg spoken only nonnatively. The Speedtalky IAL should be close
> to but within that limit. Maybe this is what you propose to address
> with your Crimbo lucubrations? I look forward to seeing the results
> in the new year...
I'm still trying to assess an easily calculated way to measure robustness
- Laver did a 0-100 scale for English phonemes, with acoustically similar
sounds, such as |p| and |k| about 15 units apart (which I think was the
smallest interval), but i don't know how he calculated the distances.
One paper I perused (I think it was by Clements) seemed to indicate that
phonemic contrasts are more robust than phonetic ones, so perhaps a
combination of distinctive feature matrix and acoustic data (formant
frequencies, perhaps) could yield a robustness metric. But so far I
haven't found any examples of this approach in the literature.
I'm still working through the consonant inventories of the 100 most widely
spoken languages, though this list tends to skew heavily toward Indo-
European languages, with Standard / Mandarin Chinese at the top of the
list and only a few non-IE languages in the top 50.
My best guess is that an optimal vowel system would be between 5 and
8 vowels, with 5 = {i, e, a, o, u} and 8 = {i, y, e, ø, ɯ, u, a, o}, while the
optimal phoneme inventory will probably be between 25 and 50 consonants.
More to follow...
Bfowol