LINGUIST List 6.1747

Fri Dec 15 1995

Sum: Relative Chronology of Sound Changes

Editor for this issue: Ljuba Veselinova <lveselinemunix.emich.edu>

Directory

1. Alexis Manaster Ramer, Summary: Relative Chronology of Sound Changes

In response to my query:

 I am trying to collect cases where two sound changes A and B
 interact as follows: we find A in a whole group of dialects or
 related languages, and B only in a subset of those, yet where
 both A and B apply, B must have applied first.

I received a number of responses (see below). I also finally thought
of two particularly good examples, from Polish and Ukrainian, which I
append at the end (excerpted from a paper to appear in Linguistique
Africaine). Thanks to all who wrote in, apologies for the length of
this posting.

=================
Matthew Baerman (mbaermanviolet.berkeley.edu):

West Macedonian dialects experienced 2 innovations:
(1) Imposition of fixed antepenultimate stress (where before there
were varying degrees of freedom).
(2) Contraction of any sequence of adjacent vowels.

Both applied to the whole of the territory, though most instances of
(2) result from the elision of intervocalic consonants (usually -v-),
which is more sporadic. In most dialects (1) preceded (2), thus:
zi'ma ova 'this winter' > zi'mava po'lovina 'half' > po'loina >
po'lojna (the literary language has po'lovina but zi'mava) yielding
instances of penultimate stress where otherwise antepenultimate stress
is (nearly) universal. However, in the northwestern dialects of Polog
the order must've been reversed, since there we find instead 'zimava
and 'polojna &c.

 ========================================

David Stampe:

Various dialects (RP, Mid-Atlantic US) in this century changed [Ow] as
in _go_ via [^w] to [Ew]. These dialects all palatalize the velar.
But palatalization of velars already occurred in all dialects. Had
there been a generation which did not palatalize the velars before the
newly front vowel, that surely would have been noticed, but it was
not.

 ==========================================
John Phillips (johnccyi.ccy.yamaguchi-u.ac.jp)

The Celtic languages provide a sort of an example, though probably not
as clear-cut or as purely phonological as you would like. All the
Celtic languages have consonant mutations, take Welsh and Irish for
example. In Welsh, p-t-c-b-d-g-m change to b-d-g-f-dd- -f at the
beginning of a word in certain syntactic environments. (f is [v], dd
is the voiced dental fricative, the mutation of g was a voiced
guttural fricative in mediaeval Welsh but is now silent). In very
roughly the same positions in Irish, p-t-c-b-d-g-m change to
ph-th-ch-bh-dh-gh-mh. So the ptc are different, bdgm the same in both
languages. Historically, the change happened to any consonant between
two vowels (roughly), so borrowed Latin cathedra (with th=t) becomes
cadair in Welsh, cathair in Irish, Latin labor- becomes llafur in
Welsh, laubhair in Irish (bh=[v]). `Two chairs' are dwy gadair in
Welsh, da chathair in Irish. Obviously, this change happened after the
Latin borrowings of the Roman period.

But Welsh is a p-Celtic language, and Irish is a q-Celtic language,
cf. Welsh pedwar and Irish ceathair, `four' (Latin quatuor), and the
change from q to p in Welsh (or, rather, Welsh's predecessor) took
place before the Roman period, as evidenced by Gaulish inscriptions
and by British place and personal names in Greek and Latin documents.

So a wide-ranging change with similar and often identical effects
happened in Welsh and Irish (and all the other Celtic dialects)
several centuries after the change from q to p which is regarded as
the dividing feature of the two major dialect groups.

==============================
Lauren Sagart:

In Gan dialects of Chinese, One change (call it change a) turns th-
into h- over a certain area (map on p. 245). Then new th-'s are
recreated by a subsequent change in certain parts of that same area
AND BEYOND (ie, where th- had not been destroyed): that change (call
it change b) creates new t-'s and th-'s out of original post-alveolar
affricates tS-'s and tSh-'s (map on p. 251). It is clear that (a)
precedes (b), otherwise the th-'s created by (b) would have been
changed to h- by (a) but that does not happen. From a geographical
point of view, the area of (b) includes a smaller area where (a) has
previously taken place, the configuration you are looking for. My
interpretation is that (b) is a response to (a), that (b) arose in the
area where (a) had taken place, but that later on (b) continued
spreading without reference to its original motivation, spreading to
areas where (a) had not taken place.

=====================

Theo Vennemann:

Are you by any chance thinking of the High Germanic Consonant Shift
(B) interacting with the West Germanic Syncope (A) or the West
Germanic Anaptyxis (A)?

=======================================
Greg Iverson:

how about the High German Consonant Shift? All kinds of
interpretations are possible, and have been proposed, but A here might
be the shift of PGmc theta to [d], which also occured to the North in
Dutch and Low German, by the way; then B could be the shift of PGmc
[d] to [t], which is just High German proper. People figure [d] to
[t] had to precede theta to [d] because PGmc theta doesn't come out as
[t] in German (it does in Scandinavian, but that's a different story);
could be coterminous, too, though, but then only the theta-to-[d] part
occurred in the North.

=====================================

Jussi Karlgren:

 A: in finnish, most dialects have illative case as a vowel
lengthening in the last syllable: sauna -> saunaan,
 auto -> autoon

historically, this has been saunaCan, where C most often has been
h.
 most or all dialects have had their long vowels diphtongized
("turned") as follows: ee -> ie oo -> uo o"o" -> y"o

B: savonian dialects have aa-> ua a"a" ->ia"

B before A: in savonian "saunaan" is not *"saunuan".

the vowel turning is savonian in origin and has spread westward from
savo. the consonant loss came from the west after this
happened. apparently there are western dialects where the order of
change is the opposite, so that there are such illative diphtongs to
be found.

Ref: Martti Rapola. 1923. P\"a\"apainottomiin tavuihin
kehittyneiden pitkien vokaalien k\"asittely suomen it\"amurteissa.
{\it Suomi} V:2. Helsinki: Suomen Kirjallisuuden Seura.

Also thanks to David Gohre and John Koontz.
==================================================


My own examples (with thanks to Stefan Pugh for help with the
Ukrainian examples):

 The first example comes from Polish dialects in which the series
of consonants conventionally transcribed as /s^, z^, c^/ merges with
the /s, z, c/ series, a phenomenon known as "mazurzenie" (Dejna 1973:
103-109; also map 5 at the end of the volume). Most dialects which
underwent this sound change do have /z^/ from another source, however,
namely, the change of */r'/ via /r^/ to /z^/ (Dejna 1973: 109-113).
Thus, the change of /r^/ to /z^/ occurred after mazurzenie.

 There are, however, more than a score villages, all at various
points along the boundary between mazurzenie and non-mazurzenie
dialects (for a list, see Dejna 1973: 111; also map 6 at the end of
the volume) where */r'/ is realized as /z/, and in which therefore
mazurzenie must have occurred later than the change of /r^/ to /z^/.
Indeed, it is normally assumed that the dialects that have this
ordering are ones which originally did not undergo mazurzenie at all
but adopted the mazurzenie pronunciation at some much later date and
hypercorrected by applying it even in cases where genuine mazurzenie
dialects do not.

 There can be no ambiguity about what happened because we do
possess numerous words whose underlying representations would have
been crucially changed by these sound changes, word like /z^eka/ <
*/r'eka/ 'river' or /ze/ 'that, but' with an original /z^/, where the
/z^/ does not alternate with any other sound in any related forms.
Thus we can be sure that dialects which distinguish the two kinds of
words had undergone mazurzenie before the merger of /r^/ with /z^/,
whereas the dialects that do not distinguish them underwent mazurzenie
later (or, possibly, underwent mazurzenie twice, both before and after
the change of /r^/ to /z^/).

Old Polish Mazurzenie Hyper-mazurzenie
 dialects dialects

r^eka z^eka zeka
z^e ze ze


Hence, the ordering of the two sound changes must be as follows:

 Mazurzenie Hyper-mazurzenie
 dialects dialects

 r^eka, z^e r^eka, z^e
mazurzenie r^eka, ze r^ > z^ z^eka, z^e
r^ > z^ z^eka, ze mazurzenie zeka, ze


 A rather similar situation obtains in Ukrainian, involving the
change of /o/ to /i/ in closed syllables and the palatalization of
coronal consonants before /i/. Since only some dialects palatalize
before /i/ < /o/ than have the change of /o/ to /i/ in closed
syllables, the latter change is more widespread and, if we believe
that geographical spread equal time depth, then it should be older.

 The complete facts are these: first, almost all dialects have /i/
from the vowel conventionally written */e^/ (derived from
Proto-Indo-European */oi/ etc.), fewer dialects have /i/ from /e/ in a
closed syllable, and even fewer have it before /o/ in a closed
syllable; second, all the relevant dialects palatalize coronals before
/i/ from */e^/ and /e/, but only some dialects palatalize before /i/
from /o/ (Zilyns'kyj 1979:40). Thus, palatalization before /i/ from
*/e^/ and /e/ is a more widespread rule than the change of /o/ to /i/
in closed syllables, but the latter is more widespread than
palatalization before /i/ from /o/.

 As far as relative chronology is concerned, we might be tempted
to assume that the palatalization rule is sensitive to underlying
representations. Consider some sample forms, e.g., d'il 'affairs'
(gen. pl.) < *de^l (prevocalic stem d'il-) vs. dil 'valley' < *dol
(prevocalic stem dol-). If palatalization before /i/ from /o/ takes
places after the change of /o/ to /i/, then the surface (or phonemic)
representation /dil/ will not contain enough information to tell us
whether palatalization should apply or not. We would need access to
underlying representations: in a dialect which does not palatalize
before /i/ from /o/, we would say that the rule does not apply before
an /i/ which is synchronically derived from an underlying |o|.

 Such a solution would seem to have two advantages. First, it
would conform to the principle that the more widespread rule must be
older. Second, it would allow us to have a single palatalization rule,
as opposed to accepting that palatalization before */e/ and */e^/ took
place long before the palatalization before */o/.

 However, it is rather obvious that palatalization before */e/ and
*/e^/ took place earlier than palatalization before */o/, since the
latter process is still spreading in modern times (ibid., Shevelov
1979:728), largely perhaps that is the pronunciation which was adopted
in the standard language. Zilyns'kij implies that in many areas the
differences between not palatalizing and palatalization has to do with
the age of the speaker (or rather that this was the situation at the
time of Zilyn'skij's dialectological researches in the 1930's),
whereas the change of /o/ to /i/ in closed syllables is clearly much
older, as shown by the spelling evidence, lack of variability, etc.

 Moreover, based on orthographic evidence as well as detailed
analysis of the dialects which do not have the change of */e^/, */o/
and/or /e/ to /i/, it is known that these vowel shifts went through
various intermediate stages, involving a variety of monophthongal and
diphthongal articulations (Shevelov 1979:318-334, 425-446, 596-618,
and passim). For example, */o/ in closed syllables seems to have gone
through a stage of a high front rounded vowel, IPA /y/. Even more
striking is the fate of */e/ in closed syllables: even though it ends
up as /i/ in most dialects, it was not, as we might think, simply
raised to /i/. Rather it was first retracted, rounded, and
diphthongized to something like IPA /ju/. Various intermediate stages
are attested for the other two vowels that changed to /i/ as well.
Thus it is quite likely that the palatalizations took place at times
when some or all of these vowels in question had not even yet reached
the /i/ stage, and so the palatalization processes would have been
triggered, at least in some cases, not by a following /i/ vowel but by
whatever vowel or diphthong actually occurred in the relevant position
at the relevant time, for example, at the /ju/ stage in the case of
/i/ which comes (via /ju/) from /e/.

 Even more importantly, even if we had no such direct evidence, we
would still know that such a solution would not work for Ukrainian.
The reason is that Ukrainian has a number of words with /i/ which
finds itself in an environment where no alternations are
possible. Since in fact alternations only occur in the last syllable
of a stem, examples include forms like t'itka, *titka 'aunt' < *tetka
(compare Rus. tyotya < tetya), t'ichka, *tichka 'rut, rutting time' <
*techka (compare Rus. techka), where all dialects palatalize, but
til'ky, t'ilky (depending on dialect) 'only' < *tol'ky (compare
Rus. tol'ko). In a word like til'ky, there is no way that a speaker
could synchronically determine that it is underlyingly |tol'ky| rather
than |tel'ky| or |til'ky|. (S)he would simply have no way of knowing
whether to apply the palatalization rule or not.

 As a result, we must conclude that those dialects where there is
no palatalization before */o/ did not undergo a sound change which
looked at underlying representations, but rather that there were two
palatalization processes. The palatalization before */e/ and */e^/ is
much older than the palatalization before */o/, and the latter simply
never occurred in those dialects which have forms like til'ky (as
opposed to t'il'ky).