reversal of merger (Yiddish final consonants) (LONG)
Alan R. King
mccay at redestb.es
Wed Dec 2 17:46:34 UTC 1998
----------------------------Original message----------------------------
On the subject of other (possible) examples of merger reversal, I am
pleased that the problem of Yiddish final consonants was brought up (B.
Wald, Wed, 25 Nov 1998, "Re: reversal of merger"). I have long been
interested in Yiddish, the original language of my grandparents' home,
which I have studied privately off-and-on over a number of years. Though
largely self-taught, I have occasionally attended university courses on
Yiddish linguistics. I happen to have the personal satisfaction of having
"guessed" the same solution to the word-final consonant "un-merger" problem
which Kiparsky published according to Benji Wald's posting, without knowing
(until today!) that anyone else had supported what I thought was a pretty
unorthodox idea of "mine"; I might add that I was encouraged to make the
last-mentioned self-assessment by the reaction of my Yiddish linguistics
professor, to remain unnamed, who said, more or less, that he had never
heard of such a nutty idea (though not his exact words). Yet it is the
only solution which seemed, and seems, to take account of the data
adequately, though I'd like to hear any alternative suggestions. I would
therefore like to encourage that the present discussion be extended to
cover in greater depth this fascinating problem. I would also personally
appreciate it if someone would give the exact Kiparsky reference, or even
send me a copy of the article if possible, as my own (short-term) access to
libraries is very limited. Here is a summary and discussion of the facts
for the benefit of all, hoping to have my blunders corrected as usual by
more knowing parties.
--- THE BASIC FACTS ---
Modern Yiddish (which for most practical purposes corresponds to historical
Eastern Yiddish, given the virtual extinction of historical varieties of
Western Yiddish) is divided into three main dialects, all contiguous to
each other in their historical (pre-holocaust) geographical domains,
popularly known as "Lithuanian", "Polish" and "Ukrainian", and variously
otherwise designated in the scholarly literature. In "Polish" Yiddish
there is general devoicing of word-final stops and fricatives, but not in
the other varieties, in which final voiced and voiceless consonants
contrast. This devoicing is a historical innovation, so we might assume
that "Polish" Yiddish also went through the change as a contact-induced
phenomenon occurring in several languages spoken in or around the dialect's
historical territory, while "Lithuanian" and "Ukrainian" Yiddish PRESERVED
the historically older voicing contrast in word-final position. This seems
to make a lot of sense at first sight, and I can personally testify that it
is also a version that is sometimes taught in academia.
But there is an interesting problem: ALL Eastern Yiddish dialects have
VOICELESS final stops and fricatives corresponding to Old High German [OHG]
voiced (or voiceless) consonants in certain words. (Proto-Yiddish is often
treated more or less as a direct descendant of Old High German with
additional non-Germanic elements incorporated.) As Benji said, the overall
pattern seems to be that words that *universally* have devoiced final
consonants in modern Yiddish are those that are morphologically invariable,
as opposed to those entering into processes of inflection or derivation,
both involving suffixation (nouns, adjectives, verbs) in which we find the
split described above between "devoicing" and "non-devoicing" dialects.
While some historical textual documentation might conceivably contribute
evidence, this is for the most part a problem requiring diachronic
reconstruction, given that pre-twentieth-century Yiddish orthography (which
traditionally employs the so-called "Hebrew" alphabet) is generally
"conservative" (i.e. "etymological") and strongly influenced by the German
model; in consequence, final consonants are often represented as
distinctively voiced and voiceless, but it cannot be concluded that genuine
spoken pronunciations are thereby reflected. For example, the numeral 20
may be seen spelt as (in transcription) /cvancig/, even though the
pan-dialectal Eastern Yiddish form ends in /k/; this does not constitute
evidence for a Yiddish pronunciation with /g/. This orthographical
practice continued into the twentieth century and in indeed still
encountered sometimes.
The following set of examples is given first of all in Standard Yiddish
[SY], which is phonologically closest to "Lithuanian" Yiddish and reflects
the type of dialect with a final voicing contrast; afterwards we will look
at the corresponding set of forms in "Polish" Yiddish, with universal final
devoicing. For SY lexical forms I will throughout take Uriel Weinreich's
now-classic dictionary as the source. I will organise the sample forms into:
(1) morphologically variable lexemes ending in a consonant that was VOICED
in Proto-Eastern-Yiddish [PEY], or at least in a preceding stage such as
OHG (except for final /z/, for which see the later discussion), with no
suffix;
(2) variable lexemes ending in a consonant that was VOICELESS in PEY or
earlier;
(3) the same voiced-final variable lexemes as in 1 with an inflectional or
derivational SUFFIX;
(4) the same voiceless-final variable lexemes as in 2 with a SUFFIX; and
(5) INVARIABLE words which (except for /s/, see below) ended in a voiced
consonant in PEY or before (now everywhere voiceless).
For variety, my examples for (1-4) include some nouns and some verbs
(adjectives are discussed later). Among the suffixes I have chosen to
include plural morphemes and diminutive suffixes on nouns, plus the
occasional non-diminutive derivational form, and the infinitive morpheme
and occasionally a derivative suffix with verbs. The examples in (5) are
preverbs or adverbs, representing words not taking suffixes; some other
invariable word types will be introduced into the discussion later. I will
use lexemes ending in /g/, /b/, /k/ and /p/, and also /z/ and /s/ so as to
include a fricatives pair, but /d/ and /t/, /v/ and /f/ follow the same
pattern except that there is a variable pan-dialectal tendency to devoice
final /v/ (which doesn't occur in German-origin words), e.g. SY /'ganef/ <
Hebrew /gan'na:v/ "thief", cf. plural /ga'novim/ < Hebrew /gan'na:vi:m/.
Yiddish stops assimilate in voicing to a following consonant even across
word boundaries, so the following examples should be understood to
represent the citation or pre-pause forms. /y/ is [j]; /N/ is the velar
nasal; /c/ is a sibilant affricate; /S/ a palato-alveolar fricative.
Nasals and /l/ may be syllabic in Yiddish.
STANDARD YIDDISH
(1a) /veg/ "way, path"
(1b) /vayb/ "woman, wife"
(1c) (ix) /zog/ "(I) say"
(1d) (ix) /hob/ "(I) have"
(1e) /hoyz/ "house"
(2a) /ek/ "end, corner"
(2b) /kop/ "head"
(2c) (ikh) /kuk/ "(I) look"
(2d) (ikh) /khap/ "(I) catch, grab"
(2e) /gas/ "street"
(3a) /'vegN/ "plural of VEG"
(3b) /'vaybl/ "diminutive of VAYB"
(3c) /'zogN/ "infinitive of ZOG"; /'zoger/ "announcer (agent noun from ZOG)"
(3d) /'hobm/ "infinitive of HOB"; /'lib-hober/ "lover, amateur (/lib/
"love" + agent noun from HOB)"
(3e) /'hayzer/ "plural of HOYZ"; /'hayzl/ "diminutive of HOYZ"
(4a) /'ekN/ "plural of EK"
(4b) /'kepele/ "endearing form of KOP"
(4c) /'kukN/ "infinitive of KUK"; /'on-kuker/ "sight-seer" (from /on/
"preverb" + agent noun from KUK)"
(4d) /'khapm/ "infinitive of KHAP"
(4e) /'gasn/ "plural of GAS"; /'gesl/ "diminutive of GAS"
(5a) /a'vek/ "away" (cf. /veg/, (1a))
(5b) /a'rop/ "down" (cf. Modern High German "herab", with "b" in spelling -
but /p/ in pronunciation)
(5c) /a'roys/ "out"
Now the corresponding "Polish" forms:
"POLISH" YIDDISH
(1a') /veyk/
(1b') /va:p/
(1c') /zuk/
(1d') /(h)op/
(1e') /ho:s/
(2a') /ek/
(2b') /kop/
(2c') /kik/
(2d') /khap/
(2e') /gas/
(3a) /'veygN/
(3b') /'va:bl/
(3c') /'zugN/, /'zuger/
(3d') /'hubm/, /'li:p-huber/
(3e') /'ha:zer/, /'ha:zl/
(4a') /'ekN/
(4b') /'kepele/
(4c') /'kikN/, /'u:n-kiker/
(4d') /'khapm/
(4e') /'gasn/, /'gesl/
(5a') /a'vek/
(5b') /a'rup/
(5c') /a'ro:s/
--- FURTHER FACTS ---
(i) Semitic words
Although Yiddish is classified as a Germanic language it possesses a
significant lexical component of Semitic (Hebrew and/or Aramaic) origin.
Semitic-origin lexemes, while often fully assimilated into and indeed
differentially characteristic of Yiddish, often possess morphological
paradigms distinct from those found with the Germanic-origin part of the
lexicon. In other words, synchronically distinct paradigm classes "happen
to" correspond to diachronically distinct sources (compare the "Semitic"
/'ganef/ - /ga'novim/, mentioned above, with the other examples, all
non-Semitic). Both kinds of paradigm in modern Yiddish, the "Germanic" and
the "Semitic", apply the same rules for voice, which thus cut across both
morphological subsystems.
(ii) Final sibilants
The case of word-final /z/ in words of Germanic origin is diachronically
remarkable, since as far as I know such forms ended in /s/ in OHG and must
have developed the voiced ending subsequently in those dialects which show
it. The only obvious source for this would have been the suffixed forms in
which the sibilant, intervocalic in Proto-Yiddish, was voiced to /z/.
Compare both Modern German [haws] < OHG hu:s "house", plural [hojzer] and
"Polish" Yiddish /ho:s/, /ha:zer/ with SY /hoyz/, /hayzer/. This type
ending in /z/ in SY contrasts, I believe with diachronic regularity, with a
type in /s/, e.g. /gas/ "street" : German /'gase/, plural /gasn/. This
necessary assumption of a "reconstructed" final /z/ in non-"Polish" Yiddish
words like /hoyz/ may be very significant - see below.
(iii) Words like "montik", "tsvantsik"
The distribution of voicing in final cononants across the Yiddish lexicon
is not quite as cut-and-dried a matter as the above examples suggest, even
though they are indeed representative of the usual patterns, if only
because there are items whose analysis is open to possible debate. There
are various words with voiceless final stops coming from OHG voiced stops,
even though such words are not strictly invariable, or not quite anyway.
Consider names of days of the week, some of which are compounds the
(unstressed) second element of which is /-tik/, etymologically cognate to
SY /tog/ ("Polish" Yiddish /tuk/) "day": SY /'montik/ "Monday", /'fraytik/
"Friday" etc. (cf. German "Montag", "Freitag"). These nouns are not quite
invariable in Modern Yiddish as they have plurals, e.g. /'montikN/
"Mondays", yet /k/ is found, not /g/.
Similarly with the numerals for multiples of ten, where the element
corresponding to English "-ty" and German "-zig", with etymological /g/, is
always /k/ in Yiddish: SY /'cvancik/ "twenty", /'zibecik/ "seventy", etc.
(German "zwanzig", "siebzig"). Again, such forms can be suffixed, although
this is rather rare; a folk song, in which the singer asks for change so as
go on tipping the orchestra so they will go on playing and he may continue
to dance, begins:
Bayt zhe mir oys a finfuntsvantsiker oyf samerodne drayer.
"Please give me change in threes for a twenty-fiver."
where the compound numeral /finf-un-tsvantsik/, lit. "five-and-twenty"
takes the nominal suffix /-er/ (cf. English "fiv-er", "tenn-er").
Suffixed forms like /'montik-N/ and /'cvancik-er/ may not have been
diachronically relevant to the treatment of weekdays and multiples of ten
as invariable words for several reasons. Maybe the suffixed forms were so
rare as to be felt as peripheral; perhaps they didn't exist at the critical
time and were (re-)generated later.
(iv) Adjectives in "-ik"
But this won't explain the fact that the frequent ending /-ik/ on
adjectives, corresponding to /-ig/ in OHG and German, is always voiceless.
Adjectives are inflected, and the usual pattern for assignment of voice to
lexeme-final consonants applies; but compare what happens when the suffix
is /-ik/ < /-ig/:
STANDARD YIDDISH
(6a) /klug/ "clever" (cf. German "klug")
(6b) /'kluge/, /'kluger/, /'klugN/ "inflected forms of KLUG"
(7a) /Stark/ "strong" (cf. German "stark")
(7b) /'Starke/, /'Starker/, /'StarkN/ "inflected forms of SHTARK"
(8a) /'milx-ik/ "dairy" (cf. German "-ig")
(8b) /'milxike/, /'milxiker/, /'milxikN/ "inflected forms of MILXIK"
"POLISH" YIDDISH
(6a') /kli:k/
(6b') /'kli:ge/, /'kli:ger/, /'kli:gN/
(7a') /Stark/
(7b') /'Starke/, /'Starker/, /'StarkN/
(8a') /'mil(e)x-ik/ "dairy"
(8b') /'mil(e)xike/, /'mil(e)xiker/, /'mil(e)xikN/
--- DISCUSSION ---
The two approaches to accounting diachronically for the distribution of
voicing in word-final stops and fricatives in Yiddish will be identified as
Theory 1 [T1] and Theory 2 [T2]. T1, the account which seems to be widely
assumed or implied and sometimes taught, is at first sight the simpler of
the two; it simply states that
THEORY 1
(i) OHG and Semitic final voiced stops and fricatives were inherited via
PEY and continued into most modern Eastern Yiddish dialects.
(ii) In "Polish" Yiddish word-final stops and fricatives are systematically
devoiced.
This theory does not yet address the problem of "exceptions" such as (5a)
"avek", (5b) "arop", "montik" (and "montikn"), "tsvantsik" (and
"tsvantsiker"), (8a) "milkhik" (and (8b) "milkhike" etc.), whose final
consonants are voiceless in all modern Eastern Yiddish dialects although
this is not predicted by T1. T2 is more complex in appearance, implying an
effective merger reversal in dialects other than "Polish":
THEORY 2
(i) Inherited original voiced stops and fricatives in word-final position
were at some stage (Stage A) devoiced in *all* Eastern Yiddish dialects.
(NOTE: Does that mean we should attribute the devoiced final consonants to
PEY itself? I don't know.)
(ii) At a later stage (Stage B) in all dialects except "Polish" Yiddish,
word-final stops and fricatives were (re-)voiced WHEN AN UNDERLYING
"VOICED" FEATURE WAS SYNCHRONICALLY RETRIEVABLE. In the "Polish" dialect
the voicing of final stops and fricatives underwent no change after Stage A
(i.e. they are all voiceless).
Retrievability of the "underlying 'voiced' feature" depended principally on
the co-existence of suffixed forms in which the suffix "shielded" the
consonant from devoicing at Stage A, cf. (3a) "vegn", (3b) "vaybl", (3c)
"zogn", "zoger", (3d) "hobn" etc. The "exceptions" itemized under T1 are
thus explained by the synchronic IRRETRIEVABILITY of any "underlying
'voiced' feature" at Stage B.
The retrievability principle invoked by T2 seems to make sense, whereas it
is difficult to think of an equally plausible principle that would motivate
the DEVOICING, in generally "non-devoicing dialects", in these same words
within the framework of T1.
For brevity's sake I referred to an "underlying feature" in the formulation
of T2. I imagine this formulation will satisfy those who like to employ
the concept of abstract underlying phonological forms, but I don't believe
the concept is strictly necessary to provide a coherent account of the
hypothesized events. An alternative is to refer merely to
morpho-phonological alternation.
It is well known that vast inflectional simplification has occurred in
Yiddish, although the basic inflectional categories are maintained. Not
only have the categories been reduced; the morphology of the surviving
paradigms has furthermore been much simplified, by greatly generalizing
some morphological markers and eliminating much original allomorphy.
Within T2, the proposed Stage B changes in dialects other than "Polish"
Yiddish can be seen simply as one further, wide-ranging morphological
adjustment with a secondary phonological consequence. The phonological
consequence is the (re-)introduction of a voicing contrast in word-final
stops and fricatives. It just happens that the resulting post-Stage-B
voicing distribution corresponds largely, though not entirely, to that
reconstructed pre-Stage-A distribution.
The motivation of the morphological adjustment in question is the
elimination of voicing alternations that had arisen in the inflectional and
derivational paradigms as a consequence of the Stage-A word-final
devoicing. The voicing alternations are maintained in "Polish" Yiddish,
which now has a more complex (and conservative) morphology than the other
dialects in this one respect. Compare the "simplified" paradigms in
Standard Yiddish (and non-"Polish" dialects) to that in "Polish" Yiddish,
here given first since, according to T2, the latter represents the
pan-Eastern-Yiddish situation before Stage B (from here on I will employ
standard transcriptions):
"POLISH" YIDDISH
* Singular Plural
veyk >>> veygn
ek ekn
mu:ntik mu:ntikn
* First-person Infinitive
zuk >>> zugn
kik kikn
* Predicative Masculine nominative singular
kli:k >>> kli:ger
shtark shtarker
milkhik milkhiker
* Base Diminutive
va:p >>> va:bl
kep kepele
* Base Derivative
zuk >>> zuger
kik u:n u:nkiker
tsvantsik tsvantsiker
I have marked with >>> the suffixed pradigm forms which introduce
complexity due to alternation between voiceless and voiced consonants.
STANDARD YIDDISH
* Singular Plural
>>> veg vegn
ek ekn
montik montikn
* First-person Infinitive
>>> zog zogn
kuk kukn
* Predicative Masculine nominative singular
>>> klug kluger
shtark shtarker
milkhik milkhiker
* Base Diminutive
>>> vayb vaybl
kep kepele
* Base Derivative
>>> zog zoger
kuk on onkuker
tsvantsik tsvantsiker
Here I have marked with >>> base forms which, according to T2, have been
remodelled on the suffixed forms, thereby eliminating voicing alternations.
The advantage, in terms of paradigm simplification, of the Stage B change
is evident. In principle either the base form or the suffixed forms could
have been generalized, e.g. either veg : vegn or *vek : *vekn would have
served the purpose, but the second route would have wasted a phonemic
distinction whereas the route taken, generalization of the suffixed forms,
maximizes that distinction, while also, incidentally, regularizing the
phonotactics (by eliminating the ban on word-final voiced stops and
fricative).
Some other points argue in favour of T2:
(i) De-voicing in non-"Polish" dialects
While in "Polish" there is systematic and exceptionless devoicing of final
stops and fricatives, there is also "sporadic" devoicing in the other
dialects, in forms such as those mentioned: "arop", "tsvantsik", "milkhik"
etc. This is perhaps not decisive evidence for a *shared* devoicing of
final consonants, but it is certainly at least plausible that a common
development might be involved. The fact that, to a varying extent, the
different items that devoice in non-"Polish" dialects are, or may be,
precisely those for which the reverse route (reconstruction of the
"original" voiced consonant) becomes obscured *after devoicing has
occurred*, raises difficult questions if the T1 position is adhered to.
Furthermore, there is no obvious motivation for the devoicing of (some)
final consonants in these dialects under the T1 hypothesis: the
phonological system remains unaffected by sporadic devoicing, and there is
no advantage with respect to the complexity of inflectional or derivational
paradigms, which are largely unaffected by the development.
(ii) Final /s/ and /z/
If my understanding of the situation with final /s/ and /z/ is correct,
then this may constitute one of the strongest arguments in support of T2.
OHG (from which the Germanic component of Yiddish is descended) did not
have /z/ in word-final position. Eastern Yiddish word-final /s/ or /z/ can
have the following sources (@ = schwa; "sz" is for "ess-zett" in German,
now pronounced like /s/ (and sometimes spelt "ss") but historically distinct):
(a) a final VOICELESS sibilant in OHG, compare e.g. German "Haus", SY
/hoyz/ "house"; German "Nusz", SY /nus/ "nut"; German "grosz", SY /groys/
"big"; German "heraus", SY /aroys/ "out";
(b) a VOICELESS sibilant followed by a word-final vowel in OHG (Yiddish
lost the final vowel), e.g. German "Gasse", SY /gas/ "street";
(c) a VOICED sibilant followed by a word-final vowel in OHG, e.g. German
"boese", SY /beyz/ "wicked";
(d) a non-Germanic-origin VOICELESS final sibilant (including /s/ from
lenited Hebrew /t/): Hebrew /ko:s/ > SY /kos/ "cup"; Hebrew /b@'ri:th/ >
Ashkenaz Hebrew, SY /bris/ "covenant, circumcision";
(e) a non-Germanic-origin VOICED final sibilant: Hebrew /b at ro:'gez/ > SY
/broyges/ "angry"; Hebrew /k at ru:z/ > SY /kruz/ "decree".
An ultimately important point here is that in Germanic items which
originally ended in a simple /*s/ in OHG, like /hu:s/, the sibilant is
voiced when it became intervocalic through suffixation, cf. German "Haus"
/haws/, plural "Häuser" /hojzer/. In other Germanic items ("Nusz",
"Gasse", "boese") and all non-Germanic words there was originally no such
voicing alternation.
Yet in non-"Polish" Yiddish, "house" is /hoyz/ (plural /hayzer/, diminutive
/hayzl/), and similarly /moyz/ "mouse" (diminutive /mayzl/), and so on.
We also find /z/ in words with original final /z/ in about the same
conditions in which these dialects keep (or restore) a final voiced stop,
thus e.g. "beyz" (nominative masculine singular "beyzer"), "kruz" (plural
"kruzim"), and /s/ elsewhere, i.e. not only when the sibilant was
originally voiceless, e.g. "nus" (diminutive "nisl"), "gas" (plural "gasn",
diminutive "gesl"), "kos" (plural "koyses"), but also where the sibilant
was originally voiced but suffixed forms are lacking, e.g. "aroys",
"broyges". (However, the verbs /muz/ "must", infinitive /muzn/, and /loz/
"let, leave", infinitive "lozn", seem to be exceptions; etymologically I
believe we should expect /s/ here (cf. German "mueszen", "laszen"). Or does
OHG intervocalic /sz/ regularly give Yiddish /z/??)
Thus /s/ and /z/ by and large follow the pattern for other consonants that
enter into these kinds of modifications, BUT the form "hoyz" can ONLY be
explained as an analogical form reshaped after suffixed forms such as
"hayzer", "hayzl". If, as T1 claims, voicing of final consonants in most
dialects is merely a continuation of the original situation, then these
dialects ought to have "*hoys", not "hoyz". ("Polish" Yiddish, of course,
has /ho:s/, but /ha:zer/ etc.)
But once we admit that Stage-B-type analogical voicing occured in the
non-"Polish" dialects in the case of final /s/ (where there is NO
"merger-reversal" since the original form was voiceless), there is no
longer much point in resisting the idea that the same process took place,
not only in words like "hoyz", but also in those like "veg", "vayb", "zog"
and "hob" (where there IS, on that supposition, a "merger-reversal" since
the original forms indeed had final voiced consonants).
I end with a tabulation of some of the sibilant-final examples, unsuffixed
and suffixed, in Standard Yiddish, by way of summary of the present point.
* Singular Plural
hoyz hayzer
gas gasn
kos koyses
bris brisn
kruz kruzim
* First-person Infinitive
muz muzn "must" (cf. German "muss", "muessen")
es esn "eat" (cf. German "esse", "essen")
* Predicative Masculine nominative singular
beyz beyzer "wicked" (cf. German "boese")
groys groyser "big" (cf. German "gross")
* Base Diminutive
hoyz hayzl
gas gesl
nus nisl
More information about the Histling
mailing list