07-12-03 14:23, alex wrote:
> If you allow Piotr, I should have a question.
> -how is one able to distinguish between changes which happened before
> Roman period and changes which happened in the time of Roman period ? It
> seems to me a very important aspect since IMHO, the only strictus way to
> control the status of some words in a certain period of time, are the
> Greek and Roman words recorded in that time.
The relative chronology of two sound changes can be determined if they
operate in overlapping environments and there is some kind of
interaction between them, that is if one of the "feeds" the other by
providing input to it or "bleeds" the other by destroying the context in
which the other would have applied. If therefore change A feeds or
bleeds change B, we conclude that A operated before B. If an earlier
application of change A ought to have fed or bled change B but it failed
to do so, we conclude that B operated before A. For example, the Old
English palatalisation of velars before front vowels must have operated
earlier than palatal umlaut, for otherwise velars would have been
palatalised also before "new" front vowels (from umlauted back vowels)
and we would have "ching", "cheep", "chitchen", "yeese" and "yiddy"
instead of <king>, <keep>, <kitchen>, <geese> and <giddy>. Using this
kind of reasoning we can establish more complex relative chronologies
involving many different changes.
Furthermore, if a change affects a loanword, this means that the change
was still a productive process at the time the word was borrowed, or
that it operated after the borrowing. This allows us to impose
_absolute_ constraints on the chronology of some changes. For example,
if a phonetic process regularly affects Latin loanwords in Albanian (as
well as inherited words, early loans from Greek, etc.), it must have
been productive at the time of borrowing or still more recently.
Examples include the change of au > a, the vocalisation and loss of
intervocalic voiced stops, such changes as s > sh and j- > (*z^-) > gj-,
and Tosk rhotacism. If, however, a process operates regularly in
inherited words but never in loans from Latin (e.g. the merger of *a and
*o, or the change of *sk > h), its operation must have predated Albanian
contacts with Balkan Latin.
Piotr