Citerar Benct Philip Jonsson <[log in to unmask]>:

> Andreas Johansson skrev:
> > Citerar Benct Philip Jonsson <[log in to unmask]>:
> >
> >
> >>I still don't think length is phonemic in Swedish
> >
> >
> > And I still find analyzing vocalic length as subphonemic
> > perverse.
> Hey, 'perverse' is a value judgment...

If native intuition isn't an argument, why don't we consider /h/ and /N/ the
same phoneme?

Well, I suppose one can take two, at least, approaches to phonemic analysis. One
might try and come up with the most economical account of surface realizations,
native intuition be damned, and one may try and model what actually goes on the
speakers' mind. In the later case it would seem relevant if the speakers feel an
analysis is perverse.

> > (The monthly instalment of "Andreas's reasons for phonemic
> > vowel length in Swedish": if vocalic length is subphonemic,
> > how am I supposed to account for the fact that _kvart_
> > [kvat`] and _fart_ [fA:t`] don't rhyme?)
> They don't rime because they _kvart_ is /kvartt/ and _fart_
> is /fart/.  More precisely:
> 1) An rC cluster (other than /rr/) doesn't suffice to
> make a preceeding vowel short.
> 2) There are rC_1C_1 (i.e. /r/ + geminate) clusters even
> though the spelling system fails to distinguish them, and
> those do cause shortening.  In fact r + geminate is more
> frequent than r + single consonant.  It is well known to
> Finnish speakers that Swedish speakers mispronounce words
> like _Turku_ as _**Turkku_.  There are BTW rrC clusters
> as well, though only in loanwords.

(Apparently not to the Finn who told me Swedes can't pronounce geminates. Well,
it's a pretty odd claim to start with.)

> I readily admit that rC against rCC is most frequent with
> r + coronal, and that's probably no accident, and that most
> words with r + coronal geminate are loanwords, and that the
> phenomenon of r + coronal being realized as postalveolars
> probably has something to do with it, the [R] dialects
> mostly having short vowels before all kinds of clusters.

I can't seem to think of any word with long vowel + r + non-coronal consonant.

> Still the r+geminate analysis is more economical than the
> vowel length analysis, since even if you claim that your
> lect has no (surface) geminates vowel length is predictable
> from consonantal structure in 90 per cent of all cases,
> provided that one takes into account that:
> 1) Vowels can be long only in  syllables with primary or
> secondary stress, and
> 2) Morpheme boundaries matter in length assignment, in that
> a morpheme boundary between two consonants in a cluster
> usually -- i.e. in most lects -- prevents shortening. That's
> why you get [ku:kt] from _kok#t_ and [E:gde] from _äg#de_.
> Notably geminates in many lects shorten preceding vowels
> even if a boundary intervenes, thus [got:]/[gOt:] from
> _got#t_.

Incidentally, my 'lect has no word [ku:kt] - the past participle of _koka_ is
[kUkt]. This would seem to be an irregularity, however: cf _smekt_ [sme:kt].

Well, your explanation seems to work, altho it still *feels* wrong.

> > But to connect to what you said about the phantasmal nature of standard
> Swedish;
> > we're arguing from 'lects that can't be reduced to a common phonology.
> So you are saying that we speak different languages?
> Interresting!

Only if Bohuslän recently acquired an army and a navy. ;)

You seem to be assuming that if two speech-forms have differing phonologies,
they should be considered different languages. I see no particular reason to
adopt that viewpoint, and it would certainly imply a use of the word "language"
quite different from that of ordinary speech.

> >>I must point out that there is nothing freakish about my
> >>pronunciation: it is a quite normal West Coast
> >>pronunciation, i.e. I hear this kind of pronunciation around
> >>me every day, though of course most people are not aware of
> >>the different allophones in their own speech.
> >
> >
> > With such an open goal left before me, how can I fail to remark that West
> Coast
> > pronunciations in general are freakish? :p
> The point was that I'm not a *lone* freak.  Besides
> non-distinction of long /e/ and /E/ sounds freakish
> to me, so I guess everybody is someone else's freak!


> > To a first approximation, written |au| is [au] in stressed position and [a]
> in
> > unstressed position in my speech.
> Well, I think that the most common words with 'unstressed |au|
> -- _chaufför_ and _restaurang_ -- are simply 'misspelled': they
> might as well be spelled with _å_ just like _fåtölj_; at least
> in the case of _chaufför_ I think it is simply a case of the
> word being adopted at a later time, and by people who harbored
> secret aversions against Leopold's scheme for respelling of
> French loanwords.  Both words definitely have /o/ for me.

Equally definitely, neither has /o/ for me. They're /xa'f2:r/ (or /xa'f2r/ if
you prefer) and /rEst8'raN/ for me. The former is presumably a spelling
pronunciation, the later is just weird. Needless to say, I'd be less than
thrilled to see them respelt as _chåfför_ and _restårang_!

But the most common word with unstressed |au| in my speak is definitely not
_chafför_, and probably not _restaurang_; it's probably _automat_ /atu'ma:t/
(/atu'mat/) with derivatives.