reversal of merger (general remarks)

[Alan R. King]

----------------------------Original message----------------------------
The general drift of my argument when I casually introduced this issue with
specific reference to the English /mate - meat - meet/ problem was that, of
the various attempts at explanation of the "merger reversal", my hunch
(merely a hunch) is that the ultimate decision is going to be in favour of
a "sociolinguistic" explanation, i.e. one which crucially involves
relations or interactions between competing, synchronically overlapping
varieties (registers, styles...) and, at some stage, either the
*substitution* of (some element of) one phonological subsystem for another
within some variety (no doubt a prestige variety, given that it has come to
our attention at all!), or alternatively the *displacement* in some range
of functions (again including prestige functions, for the same reason) of
one variety by another within the larger sociolinguistic picture.  In other
words, I suspect that the problem will be found to exemplify the need for
historical linguistic explanation to encompass a consideration of
"diachronic sociolinguistics", if such a thing can exist, and I don't see
why on earth it shouldn't, except that data may often be lacking!  :-)  I'm
not trying to be defeatist, only faithful to linguistic reality past and
present.

If I am right that merger reversal is a "sociolinguistic" process, similar
processes are no doubt in progress all over the place right now, but
without the historical perspective (particularly the "benefits" of
distance) maybe they "look" different.  First of all, from close up,
general trends may tend to come first to our attention as individual
variation.  When I start to think I can easily come up with numerous
examples of present-day processes involving substitution of specific
dialect features (rather than out-and-out dialect substitution) in
connection with various register alternations (often associated with
various degrees of diglossia and so on), prescriptivistically motivated
restructurings, "rectifications" endorsed by educational establishments
often in conjunction with literacy training, etc. in which the sum effect
is some sort of "restoration" of "lost" distinctions, in other words, in
some sense, "merger reversal".

In English, millions of native speakers learn (or try to learn) in school
to restore /h/ as part of their phonology, and to differentiate between
word-final /n/ and /N/ (i.e. -n versus -ng); in my time at least, there was
still pressure (largely futile?) from school teachers to also "unmerge"
/w-/ and /hw-/; for speakers of Cockney, there is similar unmerging of /f/
and /th/, and of /v/ and /dh/; and I'm sure the list could go on and on.
Some Spanish (or Castilian) speakers of European varieties, where the
standard or prestige norm is to keep the phonemes /s/ and /th/ distinct
although these have merged in some dialects, probably similarly go through
the "educational" process of "reversing" the merger.

(The same is not true of the long lost /b/ versus /v/ distinction, which is
scarcely ever realized in pronunciation (and then artificially and highly
pedantically, at least if we're talking about native Castilian speakers).
But just suppose the school establishment got it into their heads to
"correct" this pronunciation "defect" on a large scale (stranger things
*have* happened); I don't think it implausible that in that unlikely event,
a future generation of Spanish speakers might actually start "unmerging"
the two pseudo-phonemes in normal speech.  Interestingly, the result would
be anti-historical, since the present-day Spanish spelling, which keeps
both letters, is badly unetymological in the way it has (re-)distributed
them across the lexicon.)

Obviously I'm not saying the social mechanisms for the *spread* of these
phonological "restorations" were the same in the past as they tend to be
now, just that the "restoration" phenomenon itself is presumably an ancient
sociolinguistic kind of process, many modern variants of which may be
observable all around us.

On the subject of other (possible) examples of merger reversal, I am
particularly happy that the problem of Yiddish final consonants has been
brought up, as it has long interested me.  I have posted a long
contribution on the subject separately.

The Michelena Basque example, which I'm glad to see Larry has now presented
explicitly for the benefit of this discussion, is in my opinion another
interesting and challenging issue which has not been very widely debated
(understandably, since expertise in Basque is scarce and most Basque
experts themselves/ourselves seem to have been preoccupied with more
pressing problems), and which could perhaps both profit from fresh
theoretical investigation and contribute to the body of existing theory.
Any new suggestions?

Alan