My first thought is that it's odd to have those sort of constraints in
a lang that allows initial mf-.  Is that m vocalic?

That's also a lot of compensatory strategies to throw into the mix.  A
timeline of when each was dominant might be in order?

But based on your account of Sardinian, it does at least seem plausible.


On 5/6/08, Henrik Theiling <[log in to unmask]> wrote:
> Hi!
>
> In my current project Terkunan, I am thinking about further working
> with the r/l stuff.  I am kind of fond of what Sardinian and Calabrese
> do to r/l: Sardinian has a lot of metathesis:
>
>       VL  capra   >  Sard.  kraBa    'goat'
> etc.
>
> And both languages have frequently switched r with l, depending on
> phonological context, e.g. 'volcano' is 'vurcanu' in Calabrese.
>
> Terkunan has r/l switches, too, mainly l > r, e.g.  'ultimum' >
> 'urtime'.  Sometimes, it switches r > l if another r precedes:
> 'arborem' > 'arbul'.
>
> Now, I am thinking about doing more shifts to 'r', namely to define
> that r+consonant is not tolerated by phonology, some time after
> the l>r shifts, so a resolving strategy has to be invented.  My current
> idea is to let metathesis with the previous vowel happend when stop +
> r is the result or r becomes word-initial:
>
>     gardin     > gradin         'garden'
>     eternitat  > etrenitat      'eternity'
>     Gran Karle > Gran Kral      'Charlemagne'
>     parle      > pral           'to talk'
>     portu      > protu          'habour'
>     karkin     > krakin         'lime'
>     arbul      > rabul          'tree'
>     urtime     > rutime         'last, ultimate'
>     ornali     > ronali         'ordinary'
>
> This happens in Sardinian (and in the Sardinian dialect of Catalan).
>
> In all other cases, rC could be resolved by echoing the previous
> vowel.  This is no problem if the echoed vowel is not stressed:
>
>     mer'kat    > mere'kat        'market'
>     vur'kan    > vuru'kan        'volcano'
>
> In other cases, stress, which is currently totally regular, might
> trigger a problem:
>
>     'serve     > ?se'reve        'to serve'
>
> In these cases, we might need another resolving strategy (e.g., drop
> of 'r' or 'v' in the above case).  I do not want to introduce
> (unstressed) schwas.  Maybe rC is only forbidden in unstressed
> syllables: of the above examples, 'portu', 'parle', 'Karle' would
> remain unchanged in this case, and maybe 'eternitat', too, because it
> is a compound 'etern'+'itat'.
>
> This is a bit like Modern Dutch's 'kerruk' for 'kerk' and 'melluk' for
> 'melk' and 'errug' for 'erg' (only a schwa is inserted here, and the l
> is involved too, which might also be thinkable in Terkunan).  Words
> with lC in Terkunan include:
>
>     kalde       'warm'
>     Alman       'German'
>
> I had experimented with l > r here already, but decided that it only
> happens before stop or voiceless consonants (IIRC).
>
> Another problem is that I don't what to rename Terkunan to Trekunan,
> so I might need some exceptions, or simply define that Modern Terkunan
> does not have such a constraint in r.
>
> There are some open questions, e.g. what to do in compounds:
>
>    vir   'true'     + tat   >  ?viritat / ?virtat (currently: virtat)
>    mfril 'inferior' + tat   >  ??                 (currently: mfriltat)
>    sul   'alone'    + mentu >  ??                 (currently: sulmentu)
>
> The endings are actually -itat and -amentu, but -i-/-a- is (currently)
> dropped after vowel, r, and l.
>
>
> So does this add a nice touch to the language?  Does this look
> plausible?  I like many of the examples above, but I would like to
> hear your opinion, so what do you think?  And how do you think to
> resolve the open issues?
>
>   **Henrik
>

-- 
Sent from Gmail for mobile | mobile.google.com

Mark J. Reed <[log in to unmask]>