THE POSITION OF TOCHARIAN AMONG THE OTHER

INDO-EUROPEAN LANGUAGES
D ouglas Q. A dam s

U n iv e r s it y of Id a h o

The placement of Tocharian vis-a-vis the other Indo-European groups has been something of
a problem; those who have investigated the issue over the past seventy-five years have, on the
basis of very limited evidence, come to a variety of conclusions. An investigation here of the
phonological, morphological, and lexical isoglosses (and especially the latter) uniting Tocharian
with other Indo-European language families leads to the grouping of Tocharian with Meillet’s
“Northwestern” group in the first instance, with particularly close ties with Germanic. The ties
Tocharian has with both Greek and Indie outside of the Northwestern group reflect later
contacts on the part of the pre-Tocharians with the pre-Greeks and pre-Indies as the pre-
Tocharians moved progressively eastward in the late Proto-Indo-European world.

1. Since their identification as Indo-European lan­ 1970 Between Balto-Slavic and Greco-Armenian (a la
guages in the early years of this century, the exact Benveniste) with later contacts with Italo-Celtic and
relationship of Tocharian A and В with the other Indo- Slavic (with whom they enjoyed their last Indo-
European groups has been the source of considerable European contacts) and NO particular relationship
debate, particularly since Tocharian did not seem to fit with Germanic (G. S. Lane)
neatly into any of the received categories of Indo-
European dialectology. Indeed, its early recognition as Until the late thirties Tocharian was not well enough
a centum language (Tocharian A leant, В lcante ‘hun­ known to provide meaningful data about its relation­
dred*) in the “far east” of Indo-European languages ship with other Indo-European groups, so it is only
where only satem languages were expected robbed that with Holger Pedersen that any of the proposed con­
isogloss of its once primary position. However its nections takes on solidity. He based his conclusion
positive contribution to our knowledge of Indo- very largely on Tocharian’s possession of the “r-
European relationships has been more elusive, as the passive” (e.g., kaustdr ‘he is killed’). It was his opinion
following list of often conflicting hypotheses advanced that the “r-passives” in Indo-European represented an
by various investigators shows: innovation and thus Tocharian was to be grouped with
Italo-Celtic, Phrygian, and Hittite since they all shared
1913 Related to Italo-Celtic (H. Pedersen) this particular innovation. Subsequent investigators,
1914 Between Italo-Celtic on the one hand and Slavic and however, have seen the “r-passive” as an archaism and
Armenian on the other (A. Meillet) thus its presence is not diagnostic of any particularly
1917 Related most closely to Celtic (J. Charpentier) close relationship among groups that retained it.
1922 Related to Phrygian and, more distantly, to Armenian While showing differences among themselves, there
(E. Hermann) is a certain commonality in the conclusions reached by
1923 Related in the first instance to Thracian and Phrygian, Benveniste, Porzig, and Lane. All agree that there is
in the second to Armenian and Balto-Slavic, and in some connection of Tocharian with Balto-Slavic and
the third to Germanic (J. Pokorny) Greek. Beyond these groups there may be some con­
1925 Part o f a group with Italo-Celtic, Phrygian, and nection with Germanic (Porzig [but specifically denied
Hittite (H. Pedersen [this proposal was repeated in by Lane]), Italo-Celtic (Lane), or Thraco-Phrygian
1938 and 1941]) (Porzig). There is also unanimity in not seeing any
1936 Between Italo-Celtic on the one hand and Greco- special relationship with Indo-Iranian. All of these
Armenian on the other (E. Benveniste) investigators, including Pedersen, used what we might
1954 Originally most closely connected with Balto-Slavic call “traditional” methodology—grouping the various
and Germanic, more distantly with Greek and branches of Indo-European on the basis of those
Thraco-Phrygian (W. Porzig) characteristics which intuitively were diagnostic by

395
396 Journal o f the American Oriental Society 104.3 (1984)

virtue of their being shared innovations. Of the three (though the degree of resemblance between Armenian
later investigators, only Porzig made explicit his and Balto-Slavic that the table suggests seems too high
reasons for choosing one isogloss rather than another. and I think the same is true of the degree of resemblance
suggested to exist between Armenian and Tocharian)
2. In the thirties Kroeber and Chretien (1937 [see also on the one hand and like Germanic and Italic (and to a
Kroeber, I960]) experimented with a more statistical lesser extent like Baltic) on the other. The distance of
and perhaps less subjective technique in which they the relationship with Indo-Iranian is confirmed. (In­
took a list of variables gleaned from Meillet’s Les deed, only the [negative] index of resemblance with
Dialectes indo-europeens and established their presence Iranian is statistically significant at the .05 level of
or absence in each of nine Indo-European groups confidence. The other indices, while not significant at
(Albanian, Tocharian, and initially Hittite were ex­ that level, are suggestive.)
cluded for lack of data) and then calculated, by means It is also instructive to look at the Tocharian cor­
of formulas, the degree of resemblance of one group relations from the point of view of the other Indo-
with another. Their method does not attempt to dis­ European groups. We find for instance that, aside from
tinguish common retentions from common innovations Baltic, Tocharian ranks as the group most nearly
but it does purport to show the overall phonological similar to Germanic. Similarly for both Italic and
and morphological closeness of any pair of Indo- Celtic, Tocharian ranks in second place after Celtic
European groups (no lexical data was involved). Since and Italic respectively. These data (admittedly statis­
their results agreed very closely with classifications tically questionable) indicate a closeness of relation­
based on more traditional means, it seems reasonable ship with Meillet’s “Northwestern Group” that has not
to assume that their methodology really does measure been commented on before.1
what it is supposed to.
1 have supplied the data for Tocharian and calcu­
lated the indices on the basis of Kroeber and Chr6tien’s 1 Considering the general paucity of our knowledge of
formula Q2 (Q2 = (ad - be) / (ad + be) where a is the agricultural vocabulary in Tocharian, it is perhaps significant
number of shared positives, b the number of positives that a certain number of those mostly agricultural words
in one language matching negatives in the other, c the Meillet singled out as characterizing his Northwest Group are
number of negatives in one language matching positives found in one or the other or both of the Tocharian languages:
in the other, and d the number of shared negatives). *bedh- ‘dig’ in Tocharian A pate ‘plow’ (**bhodho-) and pata-
The results appear below (and should be compared to ‘to plow’ (*bhodh-eA-);
their Table IV [1937: 99]). *A(e)w(e)sa- ‘oats’ in Tocharian В ysare ‘wheat (vel sim .)'[A
wsar] (the Klosterrechnungen demonstrate conclusively
T able I that the Tocharian В word ysare does not mean
KROEBER-CHRETIEN, 74 ELEMENTS, FORMULA Q2 ‘grain/corn’ in general [which is satre]; since Kucha
(Tocharian Added) was known to Chinese travelers [cf. Levi, 1913: 353] for
its wheat and millet [the latter probably yap], I take
Ce La Gr Ar Ir Sk SI Ba Ge To ysare to be ‘wheat,’ the other named grain in our texts
Ce 1.00 .94 -.37 -.67 -.41 -.42 -.26 -.42 -.17 .12 from a PIE *Awesa-ro- [cf. Latin avena]);
La .94 1.00 .13 -.66 -.81 -.82 -.54 -.47 .15 .24 *krAsn- ‘hornet’ in Tocharian В krohkse ‘bee’ from a PIE
Gr -.37 .13 1.00 .35 .30 .39 -.33 -.23 -.25 .34 *kr Asnukuken- (?);
Ar -.67 -.66 .35 1.00 .29 .07 .42 .38 -.27 .30 *keA w (d)- ‘strike’ in Tocharian В kau- [A ко-] ‘to kill’ and
Ir -.41 -.81 .30 .29 1.00 .96 .42 .15 -.31 -.46 kaut- [A kot-] ‘to strike;’
Sk -.42 -.82 .39 .07 .96 1.00 .26 .17 -.75 -.24 *wel- ‘dominate’in Tocharian В walo [A wiil] ‘king’from PIE
SI -.26 -.54 -.33 .42 .42 .26 1.00 .96 -.15 .13 *wel-a-nt- or В wlaw- ‘to behave oneself.’
Ba -.42 -.47 -.23 .38 .15 .17 .96 1.00 .43 .17 The Tocharian word for ‘seed,’ AB sarm, is however not
Ge -.17 .15 -.25 -.27 -.31 -.75 -.15 .43 1.00 .27 related to English seed or Latin semen. It is rather a deverba-
To .12 .24 .34 .30 -.46 -.24 .13 .17 .27 1.00 tive noun to В sary- ‘to plant’just as maim ‘measure’ and saim
‘refuge’ are to mai- ‘to measure’ and sai- ‘to support’ respec­
As this table suggests, the statistical methodology tively. In these three verbs we have a small series of deverba-
used by Kroeber and Chr£tien tends to confirm the tives in *-mn to *-yejo- presents whose root shows an
traditional conclusions of Benveniste, Porzig, and Lane. б-grade, i.e., sor-ye-mn, m od-ye-mn, and *sdd-ye-mn. Thus
We find Tocharian most like Greek and Armenian sary- is to be analyzed as sar-y- and not sa-r-y- or the like.
 A dams: The Position o f Tocharian Among the Other Indo-European Languages 397

In both cases where an early PTch *a and *u would

3. Whether ‘traditional’ or ‘statistical’ all of these give different reflexes, the evidence strongly suggests
investigations have been limited for our purposes by that the vowel was *u. Thus where a PTch *k precedes
their having been applied to Tocharian ‘after the we have: AB kul- ‘to fade away (of energy, etc.)’ from
fact’—using data that may or may not be particularly PIE *g(w)lH-o- [: Sanskrit glayati, glati ‘he is languid,
pertinent to the placement of Tocharian.*2 The rest of fades away’), A kuli [B kliye] ‘woman’ from PIE
this contribution will look at the problem of the *k(w)lHiEen- [: Mod. Irish caile ‘country woman,
connections Tocharian may or may not have with maiden1], A kulmants- ‘reed; from PIE *klHm- [: Greek
other Indo-European groups from a distinctly To­ kalamos ‘id.’], AB kulyp- ‘to desire, wish’ from PIE
charian point of view. To further this investigation we *k{w)lp-yejo- [Avestan (aibi-)xrapa- ‘to form a desire
will discuss some new data, both ‘traditional’ and (for), etc.’], A kunmask- [B kanmask-] ‘to come; from
‘statistical,’ dealing with phonology, morphology, and PIE *gwm-nA-ske/o-, AB kur- ‘to grow old, fragile’
lexicon. As we will see the new data will sharpen our from PIE *grH-o- [: Greek geras ‘gift of honor’ from
picture of Tocharian’s relationship with other well- ♦‘privilege of age’], AB kursar ‘mile’from PIE *krsru-
attested Indo-European groups. [: Latin currus ‘wagon’].
One can match this list with one of more or less
3.1. THE FATE OF THE PIE SYLLABIC RESONANTS {* R) equal size where neither Tocharian A or В shows any
It is difficult on casual inspection to decide whether trace of *w, e.g., AB kdlp- ‘to obtain’from PIE *k{w)lp-,
PIE *R gave Proto-Tocharian (PTch) *aR directly or AB kdn- ‘to happen’ from PIE *gignH-e/o, etc. Most
whether the *d here is the reflex of one of the several of these are verbs where one may readily suppose that
short vowels that regularly gave late PTch *a, e.g. PIE an original *w disappeared by analogy with other
*w, *ay or even *i or *e, though there is no trace of the forms of the verb where different ablaut grades existed,
palatalization of the preceding consonant which should e.g., AB kal- ‘to endure’ [: Tocharian В kel- in the
have been produced if the vowel had been one of the preterite], but a few morphologically isolated words
latter. The origin of the *a is of some importance since also exist, e.g., A kdnt [B kante] ‘hundred’ from PIE
the various Indo-European groups are divided by the *kmtom. The presence or absence of a PTch *w after
“color” of the vowel which is generated by the PIE an initial *k is not predictable, either when comparing
syllabic resonants: *uR in Germanic, *i/uR in Baltic Tocharian as a whole with other Indo-European groups
and Slavic, *uR (but *eN) in Italic, *Ri in Albanian, or when comparing Tocharian A with Tocharian В
*Ri in Celtic, and *aR in Greek, Armenian, and (note that Tocharian В never shows a *w when A does
Indo-Iranian. not but that the reverse is often true). Nevertheless, in
In Proto-Tocharian an early *aR or *uR would have these cases a PTch *u rather than *a after an initial *k
fallen together as *aR everywhere except under two seems sure.
sets of circumstances: (1) after a word-initial *k where As for the second case where an early PTch *a and
*u sometimes (but not always) shows up as late PTch
*u would not yield the same reflex in later Tocharian,
*wa rather than a or (2) when immediately preceding
namely in combination with a resonant before a vowel
another vowel—in such cases an early PTch *aR
(*a/uRV), there is a nice morphologically isolated
would give later *aR (cf. *krAsniyom *krasniyan -*
example in Tocharian В sprane ‘the two sides’ from
*kraniya(n) —■ •krahi ‘[nape of the] neck’ where PTch a
PIE *sprHon- [: Old English spura ‘spur*]. If the early
from PIE *A also came to stand in an open syllable).
PTch had been **sparan-, the Tocharian В outcome
In this second kind of case, the reflex of a PIE *R
would have been **sparane rather than the actual
could appear directly before a vowel in early Proto-
sprane which presupposes an early PTch *spuran-.
Tocharian only in the case where an intervening laryn­
The Proto-Tocharian treatment of PIE syllabic reso­
geal had been lost, i.e. *RHV —• *aRVor *uRV.
nants was then exactly the same as that found in Proto-
Germanic: all *R's become *uR. This isogloss uniting
Tocharian and Germanic is a highly significant one
Van Windekens’ (1976: 418) connection of sar-y- with Latin since in all likelihood it represents a shared innovation
serere ‘to sow ’ (from PIE *‘to sow in a row’) is attractive, (partly shared also with Italic for *r and */) and is thus
though the traditional derivation of the Latin word from a
PIE *si-sH -e/o- remains more likely.
2 In large part because of the poorly understood nature of relative placement of Tocharian, only Pedersen and Lane
the Tocharian data. Of all of those offering opinions on the could be described as Tocharianists.
 398 Journal o f the American Oriental Society 104.3 (1984)

diagnostic for ascertaining the closeness of contact of Against this rather monochromatic background, the
the pre-Tocharians and pre-Germanics during some PIE *-oysi/u- seen in Greek -oi(l)n/-oiun (with per­
phase of late Proto-Indo-European.3 manent «-ephelkystikon) and Tocharian В -ais(-ah/-
The treatment of the long syllabic resonants in i)6 stands out as a startling common innovation (Winter,
Germanic and Tocharian is also identical. Both groups 1962: 126). The Tocharian form seems to represent a
show complete, uncompensated, loss of the laryngeal PIE *-oysi since *-oysu would probably have given
in most cases. For Tocharian one should compare Tch Tocharian В **-oys- (cf. soy ‘son’ from *soy us from
A kulm'dnts- ‘reed’ from PIE *klHm- with AB kulyp- PIE *suyus) but in any case it would appear to be an
‘to wish, desire’ from PIE *k(w)lp-ye/o-. Balto-Slavic old locative (with the plural ending -si/-su) added to
shows i/uR for PIE *R but T/UR for PIE *R. Italo- the *-oy that forms the neuter nominative-accusative
Celtic, on the other hand, has Ra normally, or aR(i) singular. Such a rebuilding is, so far as I know, unique
less usually, while Greek has Ra/Re/ Ro, depending on and thus highly probative of some special connection
the color of the original laryngeal, or aRa. The identity between Tocharian and Greek (pace Winter, 1962: 126).
of the Tocharian and Germanic treatments of the long
syllabic resonants is striking but probably less diag­ 33.. “s t r o n g ” a n d “ w e a k ” a d je c t iv e s
nostic than the identity of the short syllabic resonants
simply because the uncompensated loss of the laryngeal Perhaps the most striking innovation Tocharian
in this position could plausibly be an independent shares with another branch of Indo-European is the
development. possession of “strong” and “weak” (i.e., «-stem) adjec­
tives at some time in its prehistory. Both Germanic and
32.. THE GENITIVE DUAL Tocharian evidently belonged to that portion of Indo-
European, including at least Latin, Greek, and Balto-
The collective evidence of Indo-Iranian and Balto- Slavic, where «-stem nouns became productive, as
Slavic (albeit in the latter case only residually) assures “singulatives” or definites, often with affective meaning,
the reconstruction of a PIE *-ows in the genitive dual e.g., Greek Strabon to strabos ‘squinting’ or Latin Cato
and a *-ow in the locative dual. (Avestan shows an to catus ‘cunning, sharp’(see Maher, 1974: 148-49). In
innovative genitive dual [as if] from PIE *-os.) In both Germanic and Tocharian «-stem nouns were
addition, Indo-Iranian shows evidence of a genitive apparently very popular and came to replace their non-
*-oyows and locative *-oyow, originally no doubt «-stem counterparts, e.g., Gothic augo ‘eye’ or the word
restricted to pronominal use. Greek also shows traces for ‘tongue’ in both groups, Gothic tuggo and Tocharian
of a genitive in either -ой or -oiois from *-(oy)ows В kantwo (both reflecting PIE *dnghwon-).
with either distant assimilation or conflation with the More important for our current purposes, in both
competing ending -oi(i/u)n (for forms and attestations, pre-Germanic and pre-Tocharian, the singulative or
see Schwyzer, 1953: 557). While the ending itself has definitizing function of the «-stems was extended also
disappeared in Old Irish, the phonological remains of to adjectives—perhaps by way of the use of singulative
it in such cases as fer ‘of the two men’ with neutral final nouns as nominal modifiers—creating parallel declen­
consonant and “aspiration” of the following word are sions of semantically definitive adjectives (morpho­
compatible with a PIE reconstruction *-ows, thus logically «-stems) and semantically indefinite adjectives
assuring the pan-Indo-European distribution of the (those of other declensional types). Though the distinc­
genitive-locative *-ow(s) and. refuting any possible tion between “strong” and “weak” adjectives had been
inference that it is an “easternism.”4 Its antiquity is effaced in the Tocharian languages by the time of their
also suggested by its general opacity—it is not formed attestation, it had left numerous morphological traces
on any obvious analogical model.5 in the declension of both nouns and adjectives (see also
Adams, 1980). The presence of «-stem declensions of
3 Thracian may show the same development of the short the comparative of the adjective in Baltic and Greek
syllabic resonants (cf. Kati6ic, 1976: 145). (compare the thorough-going “weak” nature of the
4 No nominal duals are attested in Italic, Germanic, Ana­
tolian, Armenian, or Albanian. of *-s. Note that these forms of the dual look like the
5 One might speculate that the *-ow is the endingless corresponding cases of the singular rather than the plural.
locative o f the same stem one sees in the animate nominative- 6 The Tocharian В -tin and -i are superadded genitive
accusative, *-6w (with nominative lengthening as in *pAter). endings (ultimately from a first singular pronoun *mene and
The genitive is the same with the rather mechanical addition the *-eys o f the i-stems respectively.).
 A dams: The Position o f Tocharian Among the Other Indo-European Languages 399

comparative in Germanic) may also be a part of this have been largely random and it is wise to check these
phenomenon, reflecting an early and prematurely cur­ intuitive choices by looking at whole groups of words
tailed stage of the extension of «-stem adjectives as so that the comparison might yield sounder results.
definitive or singulative markers beside the non­ For this purpose I have chosen two semantic fields
definitive or non-singulative adjectives of other declen­ dealing with “basic” vocabulary: (1) parts of the
sions. (human) body and (2) “biological terminology”—the
designations for animals and plants and their parts.
34. . THE MERGER OF HYSTEROKINETIC AND HOLOKINETIC In addition, I took one semantically random group,
Г- AND /-STEMS namely those verbs which began with (Tocharian) k-.
The three groups of words totaled some 365 words.
Not quite so diagnostic perhaps, since it might In each of these groups we are, of course, limited in
plausibly be thought to represent independent develop­ our comparisons of those words which are attested in
ments, is the merger of the hysterokinetic and holo- the surviving texts of the two Tocharian languages and
kinetic r- and /-stems in both Tocharian and Latin. For to those which have a convincing Indo-European
instance the paradigm of the word for ‘father’ (a etymology but in all three cases there is a sufficiently
hysterokinetic noun in Proto-Indo-European) shows large number of words to suggest a certain amount of
the same hysterokinetic nominative paired with an confidence in the results. Resemblances in all three
originally holokinetic accusative in both Latin and fields are “scored” the same way: complete morpho­
Tocharian:7 logical identity of the Tocharian word with its cognate
in some other Indo-European branch gets a “3,” root
Latin Tocharian В Late PIE identity only scores a “ 1,” and something that is in
Nom. pater pacer *pAter between in some way gets a “2.” Thus Tocharian В ante
Acc. patrem pa tar *pAtrm [A ant] ‘surface, forehead’ is an exact match for
Sanskrit anta- which results in a “3” for Indie whereas
The same pattern is seen among certain PIE /-stems in it has only root cognates in Greek, Celtic, and Ana­
both languages, e.g.: tolian, giving each of those groups one point. Because
unique resemblances, those shared by Tocharian and
Latin Tocharian В Late PIE only one other Indo-European branch, would seem to
Nom. plehes asce *-e(y)s be particularly important in determining degrees of
Acc. plebem asc *-im relationship, I noted such cases with a plus and tallied
the pluses separately (labeled “dyads” in the totals).
Judging from the paradigm of the one extent w-stem Thus Tocharian В pratsako [A pratsak] ‘chest’ matches
adjective in Tocharian B, colye [nom. sg.8] ‘wild,’ col* Greek prosopon exactly, except for the stem class, thus
[acc. sg.], colyi* [nom. pi.], colam [acc. pi.], Tocharian giving Greek a “2.” Since no other Indo-European
once had this pattern among w-stems also—an exten­ group shows the same compound (PIE *proty-dkw-),
sion of this pattern that does not appear to have ever Greek gets a “2+.” The totals for these comparisons
characterized Latin. are given in T a b l e II.

. . STATISTICAL COMPARSION OF LEXICAL RESEMBLANCE:

35 T able II
INDICES OF COMPARISON FOR LEXICAL RELATIONSHIP
The number and kind of lexical resemblances joining BETWEEN TOCHARIAN AND OTHER INDO-EUROPEAN GROUPS
two or more members of a particular linguistic family
are of course important to determining the degree of Ge Ba SI Ir In Ar Gk A1 La Ce An
interrelationship between the two branches. Various Total points
investigators of Tocharian have been quick to note for three groups 96 65 48 34 63 2 73 9 35 34 11
individual isoglosses joining it with this or that other of words
branch of Indo-European. So far such investigations Number of
“Dyads” 1 6 8 4 0 4 0 8 0 2 2 2

7 The evidence is ambiguous but the same pattern in the The rank order for closeness of overall lexical resem­
r-stems might be found in Germanic also. blance is: (1) Germanic, (2) Greek, (3) Baltic, (4) Indie,
8 Found once, at K23 b3. (5) Slavic, (6) Latin, Celtic, Iranian, (9) Albanian,
400 Journal o f the American Oriental Society 104.3 (1984)

(10) Anatolian, (11) Armenian.9 Independently of this Indo-European. But when taken all together they point
investigation, van Windekens (1976: 614-619) has also very strongly to a relatively close relationship that
tallied “concordances exclusives ou especiales” between must have existed between pre-Germanic and pre-
Tocharian vocabulary and the vocabularies of other Tocharian dialects of late Proto-Indo-European. Next
Indo-European groups. His ranked order is (1) Ger­ in importance are those resemblances which tie Tocha­
manic, (2) Greek, (3) Indie, (4) Baltic and Iranian, rian with Greek; lesser ties existed between Tocharian
(6) Latin, (7) Slavic, (8) Celtic, (9) Anatolian, (10) Ar­ on the one hand and Baltic, Latin, and Indie on the
menian, and (11) Albanian. Excepting the last three, other. Such a position amidst other Indo-European
which in any case are to be grouped together in some groups is similar to that posited by earlier investigators
fashion at the bottom of the list, the only serious though not quite the same—particularly the closeness
difference between the two lists is the relative positions of Tocharian with Germanic has not been emphasized
of Baltic and Iranian. The two lists reinforce one before. Nor have investigators ever been struck by any
another to a remarkable degree. particular resemblance with Indie, though it should be
noted that this resemblance does not extend to Iranian.
.
4 No individual bit of phonological, morphological, It is possible, on the basis of these relationships with
or lexical data discussed here would be at all decisive as Germanic, Greek, etc., to “place” Tocharian geographi­
to Tocharian’s relationship with the other branches of cally in the late Proto-Indo-European world in some
manner, say, between Germanic (on the “north”?) and
9 One might suppose, however, that these results have been Greek (on the “south”?). It is also possible, and
skewed by the differing rates of vocabulary loss that the perhaps more realistic, to assume that the resemblances
various Indo-European groups have undergone. It is obvious, Tocharian shares with Greek, and perhaps also with
for instance, that both Armenian and Albanian have replaced Indie, are (when they are not simply common reten­
a far higher proportion of their inherited word-stock than, tions) the result of later contacts in the immediately
say, Germanic or Greek. We can 1 think correct for such post Indo-European period of the pre-Tocharians with
skewing, at least to a certain extent. Some thirty years ago the pre-Greeks (and pre-Indies). These contacts would
Britton (reported in Ross [1950]) calculated the number of have occurred as the pre-Tocharians moved from a
Proto-Indo-European roots, as recorded in Walde and Pokorny northwestern location near the pre-Germanics (and in
(1927-32), retained in some fashion in each Indo-European the neighborhood of the pre-Latins and pre-Balts) to
group. (Unfortunately he grouped Baltic together with Slavic, an increasingly southern and, ultimately, eastern posi­
Celtic with Latin, and Indie with Iranian; there were no figures tion vis-a-vis the rest of Indo-European.
calculated for Anatolian or Tocharian.) The percentages of The supposition that Tocharian contact with Greek
retention are: Celtic-Italic 63.7, Greek 62.6, Armenian 23.8, is later than the contact with Germanic receives some
lndo-lranian 54.6, Balto-Slavic 65.2, Germanic 67.5, and support from the fact that the pre-Greek with which
Albanian 15.6. From these figures we can calculate more pre-Tocharian speakers came into contact with seems
nearly the original degree of lexical closeness of the non- already to have some specifically Greek features. For
Tocharian groups with Tocharian, the degree of relationship instance Tocharian В pratsako ‘chest’ can be matched,
obtaining before the varying degrees of vocabulary loss except for the difference in declensional class, sound by
obscured the picture. Thus for instance the total score (regular sound with Greek prosopon ‘face’. Both appear to
scores plus “dyads”) for Germanic is 112 which when divided reflect a PIE *proty-dkw-o- but the actual PIE form
by 67.5 (Germanic’s overall retention rate) is 165.9 which would appear to have been *proti-Okw-o- with a zero
would have been the score presumably if Germanic had lost grade (cf. Skt. pratlka- or, with a different prefix,
none of its Indo-European lexicon. The scores thus corrected Latin antiquus). Only in Greek [: Greek enope ‘face,
are: (1) Germanic 165.9, (2) Greek 129.4, (3) Indie 122.7, countenance’ metopon forehead’],10 and in this one
(4) Baltic 112.0, (5) Slavic 79.8, (6) Iranian 62.3, (7) Latin Tocharian form, do we see a generalization of the
58.1, (8) Albanian 57.7, (9) Celtic 56.5, (10) Armenian 8.4. lengthened grade in this formation. The obvious infer­
Albanian moves up a bit but otherwise the relative standings ence then is that pre-Tocharian speakers borrowed the
are much the same. already transformed word as a whole from pre-Greek
No matter what corrections we may make, we will probably speakers. (It is also perhaps pertinent to note that only
never raise these figures to the level of statistical validity. The in this word has the Indo-European prefix *proti- left
data doesn’t lend itself to that kind of precision. But the any trace in Tocharian.)
numbers are suggestive and, when coupled with other data
leaning in the same direction, their suggestiveness becomes a
presumption of correctness. 10 Cf. Hamp (1973:84).
 A dams: The Position o f Tocharian Among the Other Indo-European Languages 401

It appears likely that the Greek-Tocharian contact groups, with secondary ties with Baltic and Latin. The
occurred at a time when both groups were in contact Tocharians,13 however, separated themselves very early
with at least one non-Indo-European language. In this from the rest of this group. They moved south and/or
context we should note Winter’s (1970) comparison of east and came into contact with another group of
Tocharian A natak ‘lord’, nasi ‘lady’ with Greek wanaks Proto-Indo-European speakers, the Greeks, perhaps in
(stem wanakt- from *wanatk-) ‘king’ and wanassa Moldavia or thereabouts, perhaps in the first half of
‘queen’ (from *wanakya-). The un-Indo-European the third millenium—certainly before the Greeks
alternation of *-k- and *-tk- suggests an external entered the Balkans proper. Subsequently the Tochari­
source for these words and, given the meaning of the ans must have continued to drift eastward across the
words, a source of some considerable prestige.11 north Pontic steppes and then Central Asian steppes,
To place the Tocharian contacts with Germanic, perhaps in this latter location briefly associating with
Greek, and other late Proto-Indo-European groups in some group of pre-Indic speakers,14 ultimately to
time and space is of course very speculative. But it appear in history two thousand years later in Chinese
might not be too out of line to suggest that that variety Turkestan.15
of late Proto-Indo-European that was to become In the absence of archeological data, the exact route
Proto-Tocharian first took shape and became in some of the Tocharians to their final home must remain
measure distinguishable from other varieties of Proto- somewhat speculative. The speculation is not however
Indo-European in the course of the last half of the “idle,” since it will be testable as our knowledge of
fourth millenium b.c . amidst the other members of Tocharian linguistics and Eurasian archeology becomes
Meillet’s Northwestern Group (to which we should more complete. The “wanderings” of the Tocharians,
also add Albanian, whose closest ties are to Balto- whatever the exact details, are significant because they
Slavic [Huld, 1979: 258, and Hamp, personal com­ are important to our understanding of the interrela­
munication]). tionships of the various branches of Indo-European
Geographically the Northwestern Group is surely to and to our knowledge of the peopling of an at times
be located north of the Carpathians, somewhere be­ significant, if always out of the way, corner of Asia.
tween the Elbe and the Dniester. It is tempting to
Pontic steppes was curtailed. A group settling in a particular
equate this Northwestern Group, defined linguistically,
area found its contacts, linguistic and otherwise, restricted.
with the culture known to archeologists as the Globular
Linguistic idiosyncracies were less likely to be leveled out and
Amphora culture which covered East Germany, Poland,
thus there was more divergence in speech. Conversely the
Galicia, and Volhynia during the last half of the fourth
steppe-dwellers lived in an environment where mobility was
millenium (Gimbutas, 1977: 303; see map, Gimbutas,
easy for a mounted people. Even though the steppes were
1980: 293).12
large, constant interaction tended to efface incipient linguistic
Within the Northwestern Group, Tocharian was
differences.
closest to Germanic of the surviving Indo-European
13 Here as a shorthand for “speakers of pre-Tocharian”
likewise for “Greek,’’etc.
11 It might be from this same non-Indo-European source 14 One should compare Thomas’ supposition (1982: 81) that
that come Tocharian В riye ‘city’ and Thracian bria ‘id.’ the pre-Tocharians may have formed one group of the
12 Bosch-Gimpera (1961) places the Tocharians aboriginally Andronovo cultural complex, a complex which also included
in Volhynia at 1,500 b . c . or so. The time is surely too late but the Indo-Iranians.
the placement may not be so far off. 15 Gimbutas (1963) in her review of Bosch-Gimpera (cf.
One should note that of the well-attested Indo-European footnote 12) rejects his placement of the Tocharians in
groups, seven (Celtic, Italic, Germanic, Tocharian, Baltic, Volhynia because there is no linguistic or archeological
Slavic, and Albanian) belong to this enlarged Northwestern evidence of a west to east movement across the Pontic steppes
Group. Four (Greek, Armenian, Iranian, and Indie) belong in either the bronze or the iron ages. 1 think I have provided
together in another group, one we might denominate “South­ here serious linguistic evidence for the western origin of the
eastern.” Anatolian stands apart from both. It is striking that Tocharians. I don’t think the absence of an archeological
so many Indo-European groups derive from one small por­ record is terribly damaging. Could we trace archeologically
tion, and a “colonial” one at that, of the total Indo-European the historic migrations of the Slavs in Greece or, before them,
area (which stretched at least from the Elbe to the Volga, and the Vandals or Visigoths? In any case the movement of the
probably beyond). Tocharians need not have been a mass migration, at least at
The reasons I think are largely ecological. As some of the first, but that of a relatively small group of the warrior elite
steppe-dwelling Indo-Europeans moved into the forested that grew in strength only as they got east of the Volga,
zone north of the Carpathians, the easy nomadism of the “Tocharianizing” their recruits in the process.
402 Journal o f the American Oriental Society 104.3 (1984)

‘Slavs.’” Journal o f Indo-European Studies

BIBLIOGRAPHY 2:143-156.
Meillet, Andr6 (1914). “Le Tokharien.” Indogermanischer
Adams, Douglas Q. (1980). “Towards a History of PIE n- Jahrbuch l:15ff.
stems in Tocharian.” Journal o f the American Oriental Meillet, Andr6 (1967). The Indo-European Dialects. (Trans­
Society 100:439-443. lated by Samuel N. Rosenberg.) University, Alabama,
Benveniste, ЁшПе (1936). “Tokharien et Indo-Europeen.” In University of Alabama Press.
Festschrift fu r Hermann H irt, Vol. II, pp. 229-240. Pedersen, Holger (1913). Vergleichende Grammatik der Kel-
Heidelberg, Winter. tischen Sprachen, Vol. II. Gottingen, Vandenhoeck und
Bosch-Gimpera, P. (1961). Les Indo-Europeens: Problemes Ruprecht.
Archeologiques. Paris, Payot. Pedersen, Holger (1925). Le Groupement des dialectes indo-
Charpentier, Jarl (1917). “Die ethnographische Stellung der europeens. Det Kgl. Danske Videnskabernes Selskab.
Tocharer.” Z D M G 71:347-388. Historisk-filologiske Meddelelser, XI.3.
Gimbutas, Marija (1963). “The Indo-Europeans: Archeologi­ Pedersen, Holger (1938). Hittitisch und die anderen indo-
cal Problems.” American Anthropologist 65:815-836. europaischen Sprachen. Det Kgl. Danske Videnskaber­
Gimbutas, Marija (1977). “The First Wave of Eurasian Steppe nes Selskab. Historisk-filologisk Meddelelser, XXV. 1.
Pastoralists into Copper Age Europe.” Journal o f Indo- Pedersen, Holger (1941). Tocharisch. Det Kgl. Danske
European Studies 5:277-338. Videnskabernes Selskab. Historisk-filologisk Meddelel­
Gimbutas, Marija (1980). “The Kurgan Wave #2 (c. 3400- ser XXVIII. 1.
3200 B.c.) into Europe and the Following Transforma­ Pokorny, Julius (1923). “Die Stellung des Tocharischen im
tion of Culture.” Journal o f Indo-European Studies Kreise der indog. Sprachen.” Berichte des Forschungs-
8:273-316. Institut fu r Osten und Orient 111:24-53.
Hamp, Eric P. (1973). “Formations indoeurop6ennes a second Porzig, Walter (1954). Die Gliederung des Indogermanischen
Element *-(Ho)kw-.” Bulletin de la Societe Unguistique Sprachgebiets. Heidelberg, Winter.
68:77-92. Ross, Alan S. C. (1950). “Philological Probability Problems.”
Hermann, E. (1922). Review of E. Sieg’s Tocharische Sprach- Journal o f the Royal Statistical Society. Series B, 12:19-
reste. K Z 50:296-314. 59.
Huld, Martin E. (1979). An Etymological Glossary o f Selected Schwyzer, Eduard (1953). Griechische Grammatik. Fourth
Albanian Items. Ann Arbor, University Microfilms Inter­ ed. Munich, С. H. Beck.
national. Thomas, Homer L. (1982). “Archeological Evidence for the
Katidic, Radoslav (1976). Ancient Languages o f the Balkans, Migrations of the Indo-Europeans.” In The Indo-
Part I. (Trends in Linguistics, State-of-the-Art Reports, Europeans in the Fourth and Third Millenia, Edgar C.
4.) The Hague, Mouton. Polome (ed.), pp. 61-86. Ann Arbor, Karoma.
Kroeber, A. L. (1960). “Statistics, Indo-European, and Tax­ van Windekens, A. J. (1976). Le Tokharien confronte avec les
onomy.” Language 36:1-21. autres langues indo-europeennes, Vol. 1: La phonetique
Kroeber, A. L., and C. D. Chr6tien (1937). “Quantitative et le vocabulaire. Louvain, Centre International de
Classification of Indo-European Languages.” Language Dialectologie Generate.
13:83-103. Walde, Alois, and Julius Pokorny (1927-1932). Vergleich-
Lane, George Sherman (1970). “Tocharian: Indo-European and endes Worterbuch der Indogermanischen Sprachen.
Non-Indo-European Relationships.” In Indo-European Berlin and Leipzig, Walter de Gruyter.
and Indo-Europeans, Cardona, Hoenigswald, and Senn Winter, Werner (1962). “Nominal and Pronominal Dual in
(edd.), pp. 73-88. Philadelphia, University of Pennsyl­ Tocharian.” Language 38:111-134.
vania Press. Winter, Werner (1970). “Some Widespread Indo-European
Levi, Sylvain (1913). “Le Tokharien B,’ Langue de Koutcha.” Titles.” In Indo-European and Indo-Europeans, Car­
Journal Asiatique 1913.2:311-380. dona, Hoeningswald, and Senn (edd.), pp. 49-54. Phila­
Maher, John P. (1974). “The Etymology of Common Slavic delphia, University of Pennsylvania Press.