schwa-raising, lexical diffusion &c.

[Roger Lass]

----------------------------Original message----------------------------
1. One of the contributions to this discussion raises some interesting
problems, in attempting to distinguish between 'proper' neogrammarian change
and 'secondary' change. I think this is a distinction without a difference:
any historian who works with large-scale corpora covering long time-periods
ought to be aware that neogrammarian change is not a 'kind' of change, but
simply the result of completed lexical diffusion. Attempts have been made
(notably by Labov) to distinguish the two, but they do not really succeed:
see the discussion in Joan Bybee's 2001 book on phonology and language use
for some good arguments.

Neogrammarian change is an artifact of the way historians intersect the
time-lines of languages, and the result in part of looking primarily at
institutionalised 'standard' or 'focussed' languages as the main data
sources. This is often forced on us by the textual record: but if one has
good corpora at one's disposal it seems to be the case that the normal state
for any language is to be constantly varying and constantly undergoing
change, which I suppose nobody would deny. But the standard shall we say
'Wang/Chen S-curve' of diffusing change can abort at any time, leaving
residue; it can go to completion, leaving neogrammarian Ausnahmslosigkeit;
or it can reverse, giving what Wolfgang Dressler once called 'lexical
fading', which is distinguishable from lexical diffusion only if
(contingently) you happen to know the vectoral properties of the time-line
you are intersecting. I would suggest that neogrammarian change is the same
kind of artifact as speciation in palaeontology: it is the paucity of the
record and the lack of 'intermediate stages' in so many textual traditions
that creates the illusion that 'transformative' change is even possible.
Textual historians generally have a rather different view, because their
data is of the kind that - if the traditions are rich enough - typically
show only very old changes 'complete', and most of the classical ones 'in
progress', or in mess, which is a better description.

Most important, if you can catch changes that later turn out to be
neogrammarian in progress and track them, they often turn out to be not
'events' at all, but very long term trends, which look like X > Y only if
you restrict your vision to the X and Y portions of the trajectory, or if
this is what's available, in which case faute de mieux you get neogrammarian
change. As a simple example, the (now) neogrammarian loss of postvocalic /r/
in non-rhotic dialects of English actually took over four centuries to
complete, starting very slowly indeed in the 14th century in some isolated
lexical items, but only picking up steam after 1700, and still not complete
even in 'proto-RP' in the 1870s and 80s. The same thing is true of the
lengthening of Middle English /a/ (early Modern /ae/) before voiceless
fricatives in Southern British English and its descendants (giving the
contrast in these dialects and their descendants between the vowels of 'cat'
and 'pass'): the first sporadic evidence of lengthening comes from the
1680s, but the change is still not completed even in the London 'standards'
in the 1870s, and there is still variation even in single speakers, as well
as between speakers. (On both these developments see the discussions and
data in AJ Ellis' Early English pronunciation, vol.V.)

Roger Lass