286Tyelpetema and phonetics vs. phonology in Quenya
- Nov 14 11:58 PMHello,
One long-standing problem in Quenya phonology has been the anomal
behaviour of the _tyelpetéma_ consonants with regard to syllable weight,
which is in turn connected with the permission/prohibition of long
vowels in non-final syllables and stress. That is, since Quenya does not
allow extra-heavy non-final syllables and at the same time a heavy
penult bears the stress, we are facing a problem of how to interpret the
(both well-attested) heavy weight of a syllable where a short vowel is
followed by a _tyelpetéma_ consonant (as in _hiruválye_, the acute
accent indicates stress) and the permitted long vowel in _máryat_ (this
time it is length that is indicated by the acute accent).
It was suggested in _An Introduction to Elvish_ that the phenomenon is
somehow connected with etymology and/or morphemic boundaries. This does
not seem too likely, since, for instance, the morpheme boundary is also
present in _hir+uva+lye_ and nevertheless the penult is heavy. One
possible solution would be noting that the word _má_ 'hand' was
originally _ma3_, with the velar voiced spirant which could be still
detected in later Quenya, which would mean that the long _á_ was a
biphonemic combination. I wouldn't bet my money on this theory, but it
is a possible solution. One flaw with it is that the primitive _3_ was a
thing quite different from Quenya's later disappearing _3_ < _g_, since
it had disappeared apparently much earlier than the latter even
appeared, since we have signs of its disappearance in the primitive
stems already (Cf. MÁ-, MA3-; TÁ-, TA3-), and much earlier than the
formation of the Quenya proper stress system and phonotactics anyway.
Besides, such an explanation fails to account for the _hiruvalye_ case.
Since the _a_ is short, we expect the penult to be closed. Thus, _ly_ is
bimoraic - whether a geminate or [l] + a glide (to which we will return
anon) - but if it is, it wouldn't be permitted initially, since Quenya
does not tolerate initial clusters. True, _lyenna_ - the only example of
initial _ly_ - is somewhat doubtful, but the other _tyelpetéma_
consonant are perfectly allowed: _nyelle_, _tyelpe_ (I know it's not
quite pure Quenya, but it's an example all right), save _ry_, which
can't be there for historical reasons, and not because of an abhorrence
of initial _tyelpetéma_ liquids (there wouldn't be anyway a lot of
places for _ly_ to go if _tyelpetéma_ liquids were avoided, since _ry_
would also be impermissible and initial _d_ is forbidden). It would be
strange if part of the series was allowed initially and part not (_ry_
is a special case apparently).
In the above discussion I have carefully avoided using the words
'palatal' or 'palatalized' with reference to _tyelpetéma_ consonants,
since we should first determine whether they are the former or the
latter. I suggest they are palatalized.
One obvious reason is that we would then have problems with interpreting
_r_. A *palatal* _r_ (as opposed to palatalized) is simply nonexistent
in the world's languages, and even if it is, the nearest _I_ can come to
a palatal trill or flap is a retroflex approximant (true, I have
inherent problems with trilling non-palatalized apical _r_ (not with
palatalized though), as Ivan can testify, but I would still much wonder
if anyone could do it), which is in phonetic effect far removed from the
other _tyelpetéma_ consonants, and from Tolkien's pronunciation of
In fact, other evidence can be both for and against: thus, the
pronunciation of Quenya _ty_ as _ch_ [tS] speaks for both
interpretations - since [t_j] is an assibilated consonant in itself, and
can be assibilated quite strongly into [tS], for instance in certain
Irish Gaelic dialects. _hy_ is somewhat a problem, since it apparently
_is_ palatal. However, velar consonants when palatalized are turned into
palatals - at least in Russian - since they articulated with the dorsum,
and thus the lifting of the dorsum towards the soft palate (which *is*
palatalization) changes their place of articulation, while the
articulation of, say dentals, is made with the apex, and the dorsum
movements are only additional articulations. However, this supposition,
in my opinion, allows to present a better theory for what we are
My suggestion is the following. The stress in _hiruvalye_ is to be
explained by the fact that Quenya stress, not being phonological, is
determined by purely phonetic environs, unlike the (phonemically
I would suggest that _ly_, being palatalized, was articulated with a
very audible [j] off-glide, which had at some time become an additional
mora. So _ly_ is bimoraic, the first mora being the rhyme of the
preceding syllable and the second the onset of the next one.
*Phonologically* this _ly_ is single, as it *is* allowed initially.
A situation when phonological entities are pronounced as two segments on
the phonetic level is not unheard-of in real-world languages. This, in
Archi the affricates behave in a strange way. The Archi consonantal
system is built on a fortes:lenes opposition, and at some point in time
the lenis affricates disappeared. Thus, on the synchronic level, only
the former fortes affricates remain. But since they are not in
opposition to any other row of consonants by the fortes:lenes ratio
alone, we have to decide they are unmarked. Now in initial position
those fortes affricates were of course pronounced as simple ones, since
geminates (not biphonemic clusters of identical consonants, as Russian
_v+ves+ti_) are quite unlikely initially. Word-medially, however, these
are pronounced as geminates - synchronically for *no* apparent reason!
(Add to it the fact that Archi also does not allow extra-heavy syllables
and forces vowel shortening in closed syllables, and we have a
troublingly familiar situation!).
Moreover, such discrepancies can even violate phonological constraints -
as in the Polish dialectal pronunciation of _bial/y_ 'white', where the
first [b_j] can be pronounced as a strong [j] (standard Polish has a
very weak glide there, more like a formant going back-forth for a
negligibly short time), and in Polish initial _bj_ is of course
impermissible (it has nowhere to come from anyway, since ProtoSl. _bj_ >
_bl_ very very long ago). If we agree that purely phonetic phenomena can
violate phonological constraints in Quenya as well, we find a
justification for the attested pronunciation of the monophonemic /k_w/
and /N_w/ as clustered [kw] and [nw].
Such a supposition (additional articulations becoming full-fledged
morae) could also explain away the stress in _cirya:quen_ (XI:372),
since we know that /k_w/ was pronounced as [kw], i.e. bimoraically.
Now with regard to length, there is a different situation. Stress,
phonologically irrelevant in Quenya, can be determined by the purely
phonetic environs. Length, being phonologically important, should be
judged on the phonological level. Now if _ry_ is a single consonant (it
apparently is *phonologically*), then the syllable division in _máryat_
is _má+ryat_. There is the question of unattested syllable- (=word-)
initial _ry_, but the tendency for open syllables must be much stronger
than the tendency for maximum onsets. The only example of VC-V syllable
division I can think of is in the syllabic languages of E and SE Asia,
as in Chinese (Putonghua) _Tian-an-men_. But Chinese works in a quite
different way from Quenya - Quenya is not a syllabic language.
Thus, my suggestion is the following: a CV[Cy]V sequence is phonetically
(for purposes of stress) divided into syllables as CV[C+y]V and
phonemically (relevant to length) as CV+[Cy]V.
I am eager to hear your comments.
Pavel Iosad pavel_iosad@...
Is mall a mharcaicheas am fear a bheachdaicheas
- Next post in topic >>