On 20 Oct 2014 20:42, "John Cowan" <cowan@mercury.ccil.org> wrote:
>
> And Rosta scripsit:
>
> > >It's settled that "ia" is /ja/.
> >
> > Does /j/ (corresponding to orthographic <i> rather than <j>) actually
> > exist? Are minimal pairs possible with /i/:/j/?
>
> I should have written [ja].
But must Lojban so specify the duration of the /i/ in /ia/? That seems unnecessarily pernickety, given the quite proper laxity of all other realization rules in Lojban. If, rather, what is settled is that /i/ in /ia/ is an onset, then I am asking why that must be so.
> And no, there is no separate /j/ phoneme,
> and I hope there will be none. An alternate analysis is possible by
> which there is /j/ and /w/ but not /h/; that is, "kai" is /kaj/ and
> "ka'i" is /kai/ with epenthetic [h]. But I see no particular merit in
> this analysis.
A merit would be getting rid of /'/ as a phoneme -- a good outcome given its anomalousness (not counting as a consonant for morphological purposes). The illicitness of /w%w/ could be accounted for by a rule that /%/ can't be adjacent to a vowel and that /w/ is both a consonant and a vowel.
>
> > These problems arise when {a'ua} is analysed as something other than
> > /a'ua/. If Lojban has no /u/:/w/ contrast -- as the impossibility of
> > minimal pairs would show -- then /ahwa/ and /awha/ are not possible
> > analyses. I think syllabification is likely an unnecessary complication,
> > but I don't see why /'/ in /a'ua/ couldn't be ambisyllabic.
>
> Again I should have written [ahwa] with square brackets. My concern
> is that this form would decay to [aWa] and then be merged with [awa].
But the contrast-preserving Lojbanist should instead say [ahua], or, more realistically, [aWua] or [axua]. Actually, the contrast-preserving Lojbanist should of course say [aTua], since it is well-established that [h] is not a reliably contrastive realization of /'/.
>
> > The simplest phonological analysis of Lojban is one in which there are
> > no clusters at all. The only phonotactic rules necessary are that a
> > C many be adjacent only to a V, a V may be adjacent only to a C or a
> > glide V of a type other than its own, and /%/ (or however we symbolize
> > the buffer vowel) may be adjacent only to a C.
>
> That would still exclude "a'ua", because there is no valid place to insert a %.
I was assuming that under these phonological rules, /'/ would be a (morphologically irregular) consonant and /u/ a vowel, with nothing counting as both vowel and consonant.
--And.
--