On 27.11.2004, at 03:22, Carl F. Hostetter wrote:
> [...] David [Salo!] relies on three unsupported premises:
> a. Lenition is to be expected in objects immediately following any
> preposition ending in a vowel [...].
> b. The grammatical (semantic) nature of the object itself cannot
> possibly have any bearing on the absence of this expected lenition [...].
> c. The presence or absence of initial consonant mutation in
> susceptible objects immediately following prepositions is solely
> conditioned by the original phonological environment (i.e., not by
> syntactic/semantic patterning or levelling)
If we look at lenition (or initial consonant mutation) as a historical
process (which, of course, it is), the following can, very briefly, be
1) It's a phonologically triggered process. (That is, it's part of the
regular 'tear and wear' of a language. Partly in reply to a discussion
on Lambengolmor with Bertrand Bellet, I might add that after looking
through extensive material and literature I could not find any
indication that word internal changes differed in their basic phonetic
processes from sentence sandhi as seen in Celtic languages. Cf.
further references to posts by Pavel Losad).
(For a short survey of the phonetics plus some interesting examples of
very vibrant mutations in Tuscan (Italian dialect) similar to those
found in Celtic go here: <http://www.ualberta.ca/~kirchner/Kirchner_final.pdf
For some general stuff on Celtic cf. McCone, Kim: _Towards a relative
chronology of ancient and medieval Celtic sound change_, Maynooth 1996
and Ball, Martin J. & Fife, James: _The Celtic Languages_, Routledge
2) Only after originally phonetically motivated sound change, (i.e. mostly
after the original phonetic context doesn't exist anymore) can
re-interpretation, levelling etc. take place.
Just one example from a different language, to illustrate the point: In
Syriac (a variety of Aramaic) the final -t, indicating 3. sg. f.
perfect becomes regularly spirantized > th. When certain other suffixes
are added, the spirantisation should phonetically disappear, but
doesn't in order to avoid confusion with other forms. Thus by secondary
analogy, a phonetic features is re-used morphologically.
3) As the basis of mutations is of phonological nature, we must work
off the primacy of phonology. Of course, once a system is established,
'grammar' can be the determining factor for application of lenition.
But all analogy etc. must be based on some original phonetic processes.
E.g. consistent lenition of adjectives after feminine nouns as in Welsh
was preceded by a _significant_ number of instances of lenition due to
regular sound change (which can be shown to be the case in Welsh).
Thus, it is certainly correct that, in principle, one should assume an
initial Sindarin stop to lenite between vowels (if Sindarin works like
real-world languages, which is reasonable to assume).
Yet, it is also certainly wrong (as Carl rightly points out) to do so
That is, secondary adjustments and other factors (namely what Carl
lists under "semantics", for which see below) come into play.
On the question of the impact of semantics on mutation Carl writes:
> [...] this premise [the semantic nature of the object cannot possibly
> have any bearing on the absence of an expected lenition] is certainly
> _not_ true of Welsh, [...] in which, for example, proper names
> (historically; the modern tendency is towards levelling) form a
> separate syntactic class from nouns in general, including with respect
> to consonant mutation; [...]
Carl makes an important point here.
The special syntax of proper names (PNs) is responsible for a special
phonetic treatment. (To explain just very briefly: In oral speech names
are most often used to address people (or Gods etc.). In Indo-European,
such addresses were in the vocative case (Zeu!, Indra!). A vocative is
special in that it forms its own sentence and has its own accent
pattern (originally stress always on the first syllable).
Here, again, it is important to note the primacy of phonology, and how
different grammatical categories influence it, just to be, in turn
influenced by it.
While names can be used to refer to persons ("Peter isn't here")
statistical surveys show that in spoken language the address function
is prevalent. Once the "critical mass" of PNs that behave in a special
way is reached, analogy may take place.
But again, grammar isn't the primary cause for mutations. Grammar may
only use such alternations as come to pass phonetically and expand or
Hence it is, I think, 4) most desirable to explain "freak mutations"
not only on the basis of grammar (which, of course, is correct in
itself) but also in terms of the historical motivation and ensuing
[Non-text portions of this message have been removed]