You are on page 1of 8

Review

Reviewed Work(s): Proto-Indo-European Phonology by Winfred P. Lehmann


Review by: Henry M. Hoenigswald
Source: Language , Oct. - Dec., 1954, Vol. 30, No. 4 (Oct. - Dec., 1954), pp. 468-474
Published by: Linguistic Society of America

Stable URL: https://www.jstor.org/stable/410470

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

is collaborating with JSTOR to digitize, preserve and extend access to


Language

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
468 LANGUAGE, VOLUME 30, NUMBER 4

Proto-Indo-European phonology. By WINFRED P. LEHMANN.


University of Texas Press and Linguistic Society of Amer
Reviewed by HENRY M. HOENIGSWALD, University o
New data from Hittite and a methodology refined throug
guistic theory have suggested this 'reformulation of PIE pho
an introduction the author sketches for us the classical 19th-cen
tion of Indo-European. He stresses the crucial role of Germa
stating further that Proto-Germanic 'must be considered on
servative IE dialects', the investigation of which 'may theref
information for a solution of some of the remaining phonol
PIE'. Granting that Germanic is more archaic in certain res
monly recognized, it has certainly innovated in other respe
weightier? less important?), and its alleged archaicism shou
used to favor Germanic evidence in reconstructing a given f
to report that Lehmann, in the body of his study, has not
danger.' As a trained and productive Germanic scholar he ha
and legitimate use of his erudition. IE studies have always ne
ences to offset the traditional Greco-Sanskrit bias.
The classical reconstruction has among its weaknesses those that are due to
the insufficiency of data (for purposes of the comparative method, that is),
though others are no doubt due to flaws in method. When 19th-century scholars
went wrong they were often merely inconsistent, i.e. they failed to come up to
their own standards. Had they been more consistent, they would have been, by
and large, good 'structuralists', to use Lehmann's own word. True, their methodo-
logical uncertainty would sometimes prevent them from finding the best possible
formulation: where they believed and protested that they were phoneticians,
they were in fact phonemicists at a time and at a stage when the term did not
yet exist, and the discipline on the synchronic level could not yet have existed.
There was therefore superfluous quibbling about mere notations (e.g. a suggestive
morphophonemic notation as against the more concrete phonetic). Far worse
was the hostility which kept de Saussure's work from being really accepted. But
there was hardly the thorough misunderstanding of language implied in Leh-
mann's too-harsh words. Brugmann was not a phonetician when, according to
Lehmann, he should have been a phonemicist.2 He was the latter, no matter what
he called it himself. He may list the IE diphthongs like ei as phonemes along with
the 'units' e i y p bh etc., and this may indeed have been unfortunate; but he of
all linguists should not be seriously accused of having failed to understand the

1 Weighing evidence pro and con is notoriously difficult. Counting features one by one
results in illusory exactitude; while in any less explicit attempt at evaluation our deep-
seated interest in archaisms will play us tricks: we are apt to forget mere innovations. Thus
the Tennessee mountain folk are popularly supposed to speak renaissance English.
2 Indo-Europeanists, working mostly with alphabetically written literary records, have
had phonemic analyses of a sort made for them. It is an interesting fact that the one neo-
grammarian most seriously concerned with methodology was August Leskien, one of the
very few scholars familiar with contemporary and not necessarily literary material (from
Baltic and South Slavic).

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
REVIEWS 469

systemic relations betwe


into an occasional unnece
important.
On p. 1 v is misprinted for fi. Proto-Celtic p is not a happy example to use in describing
the nature of the Celtic stop systems (3), in view of the loss of IE p in all of Celtic, and
its reappearance in only a part of the group later. For Car-6s (3 fn. 5) read Ca-r~s; for t (4)
read t.

Lehmann then restates 'Brugmann's' reconstruction (Ch. 2). The restatement


is different 'primarily in not including schwa, diphthongs, aspirated voiceless
stops, or palatal stops.' There is also an , more or less equivalent to the old schwa
secundum. Lehmann seems to reckon also with a distinctive accent for PIE,
described as changing from phonemic stress to (non-phonemic?) stress to pho-
nemic pitch to 'free pitch accent, like that of Vedic' (8), an obscurity which I
shall take up later on. Lehmann seems to find that the location of these accents
in the morphemes of the lexicon remained constant; only their phonetic character,
and the degree to which they are tied up with other features (vowel contrasts, for
instance), underwent variation in passing from older to more recent stages of
Proto-Indo-European. There is a long effort to disclaim the existence of diph-
thongs as phonemes per se, and a r6sum6 of Edgerton's semivowel ('resonant')
theory. The structure of the root is then described, much along Benveniste's
lines.3 There is, finally, a section on IE meter (i.e. Homeric and to some extent
Indic), in which consonants, semivowels in 'diphthongs', and vowel length (intro-
duced here on p. 20 as X, without comment) all function somewhat alike, sug-
gesting a phonological reconstruction in which they come out as members of the
same classification. The rather more specific argument used by Kurylowicz, and
also by Maurer, is not discussed.
In 8 fn. 1, read v for ?. Gk. bios is not from gwiwos (13). For procas 'wooer' (16) read pro-
cus. Metricians will take exception to the statement (20) that the 'accented part' of the
hexameter foot consists of a long syllable. P. Maas, for instance, simply identifies certain
locations in the line as 'longa'.

The next chapter, on the laryngeal theory, is a useful account of work in this
complex field. Included is Kurylowicz' splendid, half-forgotten explanation of
the exceptions to Brugmann's (and Pedersen's) law: 'PIE o before a non-syllabic
semivowel in an open syllable yields Indo-Iranian a.' It is Skt. 1st sg. cakdra
vs. 3rd sg. cakdra because in the 1st-person form the stem syllable was not
a The well-known limitations existing between root-initial and root-final consonants
(only in Benveniste's sense?) are not quite fully stated here; but see 17, with a reference
to Meillet, Introd. 157. Not only are (a) identical consonants and (b) a voiceless and a voiced
aspirated stop incompatible, but so are (c) two voiced unaspirated stops even if they are
not identical, and, it seems, (d) any two stops in the same place of articulation. Following
Z. S. Harris's observation that this seems to give the voiced unaspirated stops a central
position in the system, one could perhaps formulate the situation like this: the zero grade
of a root contains two different consonants (listed: b d g s y w r I m n and the laryngeals)
and further, for b d g, one of these two: B(reath) and U(nvoicing). Either one of the latter
may occur before the first, between the first and the second, or after the second consonant,
and will affect the contiguous b d g. Thus Ubd = ped, bUd = pet, bdU - bet, bdB = bedh,
Ubm or bUm = pem. U and B may or may not be identical with two of the laryngeals.

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
470 LANGUAGE, VOLUME 30, NUMBER 4

originally open: cakdra is from *kwekwolXe (Hitt. -1hi), cakdra is fr


(Hitt. -i). Similarly, jandyati from *gonXeyeti (hence not tjdndy
presuppose tgoneyeti), genX being a set base. This incidentally p
of laryngeals, in this environment, into Indo-Iranian times. The
no other explanation which would fit the facts better.

For ldos (26) read la6s. The article from Lg. 17 (quoted 33) is by W. M
becomes e etc. in Attic (and other dialects) but not in 'Greek' (33), a fact w
tain importance for the dialect chronology to be reconstructed within IE

Next Lehmann examines a number of special problems from G


schiirfung ('PGmc -w- was lengthened after any short vowel wh
laryngeal followed -w-; PGmc -j- was lengthened after i when reflex
followed -I-, and after a when reflex of a laryngeal preceded or
PGmc. g and (k)k corresponding to PIE /w/ (tentatively explain
agreement with Austin, by a difference between laryngeals, X1
geals after r, 1, m, n (rX > ru as in OHG hiruz, while ,X follow
sonants > a); the OHG r-preterits (like anasteroz from -stozan,
Xw and Xy, and preservation of the laryngeal throughout Proto
OHG, where it merges with r.' Lehmann also draws attention t
servation that Alemannic amar and enjr without j-as contra
'ferment'-seem to correspond to cases in which Greek has rough bre
than z-: amar : htmeros, enjr : hds, but jesan : z65). The results
in Chapter 8, where an early stage of Germanic is credited with
b x b d53 p t k s z h r I m n wj X X . Finally, Lehmann turns to ~2,
high vowel of Germanic (OE 8, Goth. j, OHG ea, ie, etc.) and trie
as a Gmc. development based on cases of eXy, i.e. full-grade but
grade i.
Intervocalic w was not lost 'in Greek', see above (36). Lat. difrutum 'm
u may still stand in the same relation to Thrac. br'itos (provided we trust
the tradition of the gloss, for which there is not the slightest necessity)
short-vowel zero grades of heavy bases to long-vowel zero grades (Wack
1.92 ff.). One could see in bhrut the sequence bhr,wXt-, in bhriit thesequen
implied in Lg. 28.291 f. This, to be sure, presupposes a somewhat differen
that propounded by Lehmann. How can OE cn be equated with Russ. sos
ing the initial stop?

As has just been hinted, laryngeals in sequence with semivow


have been involved also in the twofold representation of so-calle
Greek, viz. sometimes as z-, sometimes as h-; at least, all other
far attempted have come to grief.' Sapir and Sturtevant though
voiceless laryngeal plus y accounts for the cases with h-, while y- pu

4 This is less implausible phonetically now that W. G. Moulton, JEGP 51.


evidence for uvular r in German long before French influence played
Mendels, id. 52.559 f.
6 zugdn from tod-yugom, to-dyugom is implausible not only (75) becau
many other cases besides the nom.-acc., but also because the article is a
In early Greek tod was a demonstrative pronoun, and the phrase tod yug6d
than to zugdn was later in Attic.

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
REVIEWS 471

accounts for z-. They w


reasons. For the second,
(i.e. voiced laryngeal plu
for either y or ' in zeia
origin in PIE clusters, dy,
y' (79). This is an extrem
happened to straight ini
well be that it did not occ
Since /y w r 1 m n/ hav
initial occurrence would
theoretically possible in
not /iw.. according to E

Skt. i.tdcome
already (fulltograde
set upyaj-) or IElaryngeal
an initial *n- 'un-'before
(full the
grade ne-?), many
semivowel-if onlyscholars
on the have
general principle that initial syllabics did not originally occur--i and n no more
than e. (Actually there is often positive evidence of one kind or another for the
presence of laryngeals in such cases.7)
The rough breathing should not be called an 'aspirate' (74). In Homer z makes fully as
much position as does s before stop. The exceptions are three proper names with an iambic
sequence in the first two syllables so that another short syllable is required before the word;
in other words Lehmann's case is even better than he makes it (76). That Skt. yas- 'fer-
ment' had an initial laryngeal is probable from yeqati, which can only be a reduplicated
form, *ya-is-a-ti (cf. avocat < *a-va-uc-a-t; 78). For did-tei5 (78) read something like did-
te35.

Finally, there are the voiceless aspirates (Chapter 11), regarded since de Saus-
sure's day as clusters of voiceless stop plus a voiceless laryngeal, perhaps a
specific one. As Lehmann points out, kh (unlike k, g, gh) is not subject to pal-
atalization in Indo-Iranian; hence it was still a cluster, the laryngeal keeping the
velar stop from contact with y in Skt. khyati 'he sees' and a few other cases.
While the conclusion is not absolutely cogent, it does simplify the picture far
more efficiently than do the author's somewhat parallel conclusions with regard
to diphthongs. Together with the earlier discovery of the survival of laryngeals
after resonants, it has great force. The clusters did in fact become 'unit phonemes'
to all intents and purposes in both Indic and Iranian, as witness their alphabetic
treatment. One might regard this latter step as minor. The important fact is that
in Indic (more clearly here than in Iranian) those particular clusters survived as
a special entity of some kind, while all the other IE dialects let them fall together
with one or another of the existing unit phonemes.
The last four chapters (Laryngeals in PIE; The PIE Phonemic System; The
6 However, initial y- occurring in this position seems to have resulted in Gk. z- also, as
Sapir and Sturtevant thought. If Xyug were set up, the root, with the infixed present seen
in Skt. yundkti etc., would become a monstrosity. See Benveniste, Origines 1.159 ff.; A.
Martinet in Word 9.287.
7 Lehmann himself, Sturtevant (CGHL2, Vol. 1), and others are not consistent in writing
laryngeals in forms with so-called initial vowel. A master's thesis by J. A. Reif (University
of Pennsylvania, unpublished) makes it probable that semivowels were very largely pre-
ceded by an initial laryngeal.

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
472 LANGUAGE, VOLUME 30, NUMBER 4

Allophones of the Laryngeals; Development of the PIE P


present Lehmann's final view. Once again there is much criti
mann's and Hirt's notations), restatement, and summary
doublet Gk. (a)thdnatos : thnto's is blamed by Lehmann (in
others) on accent: when the accent did not fall on the next s
between resonant and laryngeal; when it did, this e was lost (89).
effect of a laryngeal on preceding e a o and on preceeding
allophones of /y/ and /w/) in closed syllables had already ta
PIE, or eX etc. before consonant or pause had already merged
grade e (though this was apparently not so in the case of eXy
of schwa (or of the three 'Greek' schwas), Lehmann reckons w
,y (95). The voiced laryngeal y changed 'the timbre of a conti
it coalesced with it', so that the result was 5, somewhat un
laryngeals x and h with their effect on contiguous vowels un
This is a most important difference between Lehmann's and
tions on 'normal-grade o'. Vowel e occurs only accented, onl
By way of completing the picture, two familiar cruces, not her
the laryngeals, are briefly examined. One is b and its family
for the correspondence between Gk. Celt. Tokh. Hitt. t wit
other languages (Gk. drktos, Skt. fkaa 'bear')-a correspondenc
a velar stop. Perhaps kfb does not 'contrast MINIMALLY [emp
succession of velar and dental obstruent phonemes' (100); but
small comfort in view of Skt. uktd, where the contrast, thou
unmistakable. If it were not, no more would be needed than t
rkga and posit a conditioned sound change from t to sibilant
As it is, Benveniste's effort to set up an additional guttural se
phophonemic advantages so far as the build of the root is con
be taken seriously.
The other perennial question has to do precisely with t
series. The fact that no historical IE language has more than two
forbid reconstructing three; otherwise the Semitic sibilants could
reconstructed without a knowledge of South Arabic, the only
to preserve all the original contrasts. Still, there are indicatio
the palatals and the 'pure velars' (i.e. both the nonlabialized s
time allophones determined by the segments of sound followi
is convinced of this, and he supports his judgment with a convin
structural argument (101-2). The fact remains that it is not e
all the etymologies in which k' and k, so-called, find themse
sets of correspondences. The labiovelars are kept in the sys
Lehmann, although others have been equally prone to see in
groups consisting of a velar stop and the nonsyllabic alloph
allophones of the four Sturtevantian laryngeals are tentativel
by means of a good deal of ingenious reasoning and arguing f
the sound changes conditioned by them, as follows: [x] for /x
for /h/, [B] or [b] for /2/ (108).
This late PIE as arrived at by the comparative method is f

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
REVIEWS 473

internal reconstruction
logically unique relics, and
treated by Pedersen, Da
tains the following pictu
tem are associated with
nemicization of pitch ac
Both periods of distinctiv
distinctive accent. In the
it could stand on any sy
timbre or vowel length. In
the changes of accent an
ments, notably those in
here to give an idea of th
way of those who wish
far-reaching conclusions
quoted passage, would n
a terminology vaguely r
called phonemic only wh
the varying location of
to consider this pitch acce
last chapter. Qualitative
differences in pitch accen
but ? > o. When the two
phone associated with th
the laryngeals were lost-a
tendency (to which Leh
more surviving larynge
Before this sound change
course also a-colored allo
from /e/ when standing
its syllabic status' (111).
such a shrewd judge as E
wrong. In an early stage
ones. The phonetic chara
of them, makes it probab
difference between /e/ an
arrives, as have other st
ture of a language with
and reduced these two en
,as determined by the seq
prefers, of e), in other w
to the French schwa as
Lehmann does not openly
can posit an oldest, pre-
vowels'), laryngeals, and
been neglected (if we ha

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms
474 LANGUAGE, VOLUME 30, NUMBER 4

would seem to be the fact that throughout all the purported


centual system, the location of the various peaks, variously de
ity, stress, (higher) pitch-the last two both 'phonemic' (or '
otherwise?-is constant, inasmuch as it is that which is reco
Greek, Vedic, Verner's Germanic, and Slavic, and inasmuch as
with the ablaut data. Here one suspects Lehmann of being a p
he ought to be as much of a phonemicist as Brugmann was!
Hirt's symbol i is curiously misprinted (86 and passim) as a dotless i.
have 'ou' for IE u and a (86, table)? Is Lehmann alluding to Beotian o
OCS i for L (ibid). gelasma etc. are not compounds (93). Read est6resa f
molein for molein (94). Read pibati for pibati (107; also 84, 118). 'Ove
novating area the back vowels /a/ and /o/ coincide' (114): at 101 fn
pointed out, however, that the 'Gmc. development ... may be dated aft
Grimm's Law' and has no direct connection with what went on in In
and Slavic.

Aside from its several handsome contributions to special problems, the present
work must be rated as a courageous effort in a difficult field, in which it ta
verve and an independent attitude to achieve anything. It would be unjust
say that in such a field any effort at all must be welcome: Lehmann's is by
odds one of the most serious efforts that have been made. Nevertheless, stim
lating as his presentation is in its broad outlines, some of us will be even mo
grateful to him for his seemingly smaller results, and also for showing us, i
plicitly, where there is detailed work still to be done-for instance, in connect
with the relation between syllabicity, conditioned sound change, and word
boundary; see above on word-initial 'y-'. In his attempt to give us an overa
picture we see Lehmann wrestling with a task that may well be beyond th
Indo-Europeanist's ability at the present time; we see him draw precarious c
clusions from scanty and ambiguous evidence, and make claims for the speci
efficacy of a structural method of reconstruction which, in the nature of things
as they have developed in modern linguistics, cannot be so radically differen
from traditional work as he wants it to be. He will earn our special thanks b
telling us in the future more about certain matters which in this volume ha
suffered from over-condensation or incomplete treatment. We need all the d
cussion we can get.

Buddhist Hybrid Sanskrit grammar and dictionary. By FRANKLIN EDGERTO


(William Dwight Whitney linguistic series.) Vol. 1: Grammar, pp. xxx, 23
Vol. 2: Dictionary, pp. [ix], 627. New Haven: Yale University Press, 1953.
Buddhist Hybrid Sanskrit reader. Edited with notes by FRANKLIN EDGERTO
(William Dwight Whitney linguistic series.) Pp. ix, 76. New Haven: Y
University Press, 1953.
Reviewed by M. B. EMENEAU, University of California, Berkeley
To review this Grammar and dictionary and the Reader of Buddhist Hybri
Sanskrit (BHS, as the author abbreviates it) is a privilege and a pleasure. We
may at last congratulate the author on having brought to a successful-or rath

This content downloaded from


180.176.178.34 on Sun, 23 Jan 2022 14:02:41 UTC
All use subject to https://about.jstor.org/terms

You might also like