TOKHARIAN AND BALTIC VERSUS SLAVIC AND ALBANIAN

HARVEY E. MAYER
Defense Language Institute

In a previous article1 I said that only Baltic and Tokharian had lost all traces of aspiration in
native forms via glottalization. In a more recent article2 I said that Slavic and Albanian, unlike Baltic,
had, in a special way, restored aspiration. Here I shall show in more detail how, starting right after
Very Late Proto-lndo-European, the emerging dialects which were to become Baltic developed
differently from those which were to become Slavic.
In Very Late Proto-lndo-European, the excessive aspiration which had been blurring the
distinction between voiced and voiceless stops was being eliminated by glottalization. The
laryngeals were beginning to be affected by this, too. Some had been eliminated in initial position
which made the vowel a a phoneme. The resulting sound system where stops were aspirated allo but
glottalized phonemically was as follows:

Eventually most dialects eliminated aspiration more from one series of stops than from the other. Prelndic, Pre-Greek, Pre-ltalic, and Pre-Celtic removed it more from voiceless stops while Pre-lranian,
Pre-Armenian, Pre-Hittite, and Pre-Germanic removed it more from voiced stops.3 Pre-Baltic, PreSlavic, and Pre-Albanian, in the central position bordered by dialects of both types, eliminated
aspiration from both voiceless and voiced stops.
A specialist,
even agreeing with
my assessment
of
Pre-Prussian
as having
delabialized H3 and o to H2 and a, might come up with the following misleading representation of the
resulting Early Pre-Baltic, Early Pre-Slavic and Early Pre-Albanian sound system with H3and o in
parentheses to indicate their absence from Pre-Prussian.4

It misleads because it suggests far more unity than there was. The critical difference between Early
Pre-Baltic and the two variants of the other central dialect, Early Pre-Slavic and Early Pre-Albanian,
was not then noticeable in the inventory of phonemes themselves, but in the presence versus
absence of one of their combinational possibilities. Early Pre-Slavic-Early Pre-Albanian retained both
initial sk- and initial ks-. Early Pre-Baltic allowed only initial sk-. It had metathesized ks- to sk-. (And
noting sk-, not *k-, we see a very early, pre-ruki law change which we can attribute to this very early
state of Pre-Baltic.) On hearing one another talk, the speakers of these different dialects might only
have been amused by the resulting "mispronunciation" of certain words: "They say *ksoudos, we say
*skoudos (or *skaudas)!" But this difference in sequencing was going to have important results.
I attribute this difference in sequencing between Early Pre-Baltic and Early Pre-Slavic-Early PreAlbanian to a difference in degree of glottalization, that is, ultimately, of the checking of
aspiration. Early Pre-Baltic was more heavily glottalized. It was more set against aspiration.
Therefore, as one device for inhibiting a change of s to h (this change now a common development in
Indo-European languages), Early Pre-Baltic metathesized ks- to sk-. Here k, near the point of
articulation of glottalization had a similar function of inhibiting the aspiration of s to h.
We must keep the early coincidence of deaspiration of both voiceless and voiced stops in Early PreBaltic and Early Pre-Slavic-Early Pre-Albanian in perspective. Some, if not most of it arose through
the accommodation which their respective speakers were making to speakers of dialects initially
deaspirating only voiced stops and to speakers of those initially deaspirating only voiceless ones. So
none of this agreement points necessarily to a single dialect. It could have all been coincidental. Far
more critical was the fact that Early Pre-Slavic-Early Pre-Albanian was less glottalized enough to
retain the initial combination ks-. It had a means which Early Pre-Baltic lacked to restore aspiration
(i.e., from initial ks-) in a new phoneme, h, after the laryngeals had disappeared. And this is what
happened. Pre-Baltic avoided this in non-initial position through morphophonemic, often, syllabic
division:-k + -s(-).
Once the laryngeals vanished, and with them the glottalized stops, differences in phonemic inventory
became apparent as the following charts show with o, o in parentheses to show their absence from
Pre-Prussian and z in parentheses to show its allophonic status.

Before this disappearance of the laryngeals had happened, with merely allophonic aspiration removed
from all non-glottalized stops and with no regard for the difference in sequencing possibilities of
initial sk- and ks-, it might have seemed that Baltic and Slavic had come from the same IndoEuropean dialect as Meillet had put it.5
But this coincidence in allophonic distribution together with the relative geographic position of PreBaltic and Pre-Slavic-Pre-Albanian merely allows us to label this particular band of dialects as the
"Central Dialectal Continuum of Laryngeal-Glottalic Indo-European."
After this disappearance of the laryngeals had happened, it became clear that Baltic and Slavic could
not have come from the same Indo-European dialect at all, despite deceptive similarities. Something
had to have existed throughout Pre-Baltic alone which prevented the restoration of an aspirated
phoneme, h, after the disappearance of the laryngeals. Ultimately, this was glottalization whose early
excessiveness was confined to one dialectal continuum alone, the Pre-Baltic one, to which Pre-Slavic
had simply never belonged at all.
Matching the deceptive deaspirating similarities between Early Pre-Baltic and Early Pre-Slavic-Early
Pre-Albanian we find the deceptive coincidence in the loss of all trace of aspiration in only Baltic and
Tokharian. We might, in our turn, be tempted to conclude that Baltic and Tokharian came from the
same excessively glottalizing Indo-European dialect. But did they? Did the Baltic and Tokharian
apparently thorough deaspirations arise through an identical process indicating a common, exclusive
origin? They did not. Other results in the stops indicate entirely different processes leading to the
same results in deaspiration.
Tokharian stops (only p, t, k) compared with those of other Indo-European languages indicate a
relatively late completion of total deaspiration initiated by excessive glottalization. Baltic evidence
indicates the opposite, relatively early one. Strong aspiration in Pre-Tokharian must have continued
long enough to have blurred all or most voiced/voiceless distinctions in its stops before glottalization
finally eliminated aspiration. Thus Pre-Tokharian obstruents before the loss of all aspiration which in
turn resulted in the loss of all glottalization must have been as follows:

Pre-Baltic was never like that.

Pre-Baltic in its late satemized stage had the following system of obstruents. Note how it compared
with the late satemized one of Pre-Slavic-Pre-Albanian with allophones in pharentheses ( , = ruki
law reflexes and reflexes of s assimilated to , . Much, much later, in Middle Common Slavic, one of
these ruki law reflexes is x which now occurs after i, u, r, and even k, not just in initial position, as was
surely the case much earlier).6

The shifts into Proto-Baltic versus Proto-Slavic and Proto-Albanian involving the assibilation of the
palatals followed patterns set by the results of the different degrees of glottalization. Proto-Baltic's drift
was going to be different. And this was true only because it had noh and did not allow the initial
sequencing of ks-, all results of excessive glottalization. Note also that at this point (Pre-) Albanian
with reflexes an, am, n, m for syllabic nasals , , , already had been diverging from
(Pre-)Slavic with un/in, um/im, n/n, m/m.7Still, its drift in the development of Indo-European
compact obstruents k, g, h,

continued to match that of Slavic.

Indo-European , tended to become t, d in all three protolanguages. In Proto-Baltic alone t,

d tended immediately to become , especially in initial position. This conformed with the condition
of no initial ks-, a stop, any stop, followed by a fricative, set by previous excessive glottalization. In
Proto-Baltic, Proto-Slavic, and Proto-Albanian (t), (d) alternated with, probably, palatalized k',
g' (which eventually became k, g). And these in turn, alternated with old k, g. But there was one big
difference. Proto-Baltic had no h because of Pre-Baltic's excessive glottalization.
The other languages had the velar h which each, with similar drift, paired with palatalized h' from
different, individual sources which was part of the pattern of sharp/plain (palatalized/non-palatalized)
velars. Proto-Slavic's first h' probably came from k' (from Pre-Slavic, Indo-European ) in the word for
'cold' *h*old- as an expressive form. Others came later, some from s as expressive forms. The word
for 'lame', xrom-, might in originally changed form, have been either *h'rom- with
palatalized h' or *hrom- with non-palatalized h. Proto-Albanian probably built its h/h' opposition on
new velar aspirates from new initial ks- from metathesized sk- as in *khol-/*kh'wl-, *kh'il-(which
became *hhol-/*kel-, *kil-, and later *hol-/*tel-, *til-).8 Albanian evidence, h to s before front
vowels, suggests that Slavic's first palatalization of the velars might have started with x(h) to before
front vowels and spread from there as a class phenomenon in both languages to the stops with k,
g, to t, d in the early stages.
While Baltic , from (t), (d) from lndo-European

merged with , from the ruki law,9 Slavic

and Albanian compact t, d from this , became diffuse ts, dz to oppose new t, d from their
respective first palatalizations of the velars k, g before front vowels (which are now , i.e., t, in

Slavic and s, z in Albanian).10 When Albanian new t, d became strident (to avoid any sort of
merger with t from k from kh' from ks- plus front vowel), old strident ts, dz from Indo-European
,

became mellow tth, ddh. While tthsimplified to th, ddh split into d, dh which are now variant

reflexes of Indo-European .11 Slavic affricates ts, dz from Indo-European , became

fricatives s, z after the second and third palatalization of the velars k, g to t, d. This was done to
prevent merger of old and new mellow t, d. With strident, diffuse ts, dz simplified to s, z, new
second and third palatalization mellow t, d redefined as diffuse could now be opposed to old first
palatalization mellow t, d which remained compact and is thus noted as t (= ), d )= , later
simplified to ). There now was room for this with the number of diffuse affricates reduced to one
(mellow) series.12
Further changes in form and distribution of ruki law reflexes in Slavic and Albanian will not be
discussed here. As for the Baltic joint ruki law-palatal reflexes, other evidence of them is available with
insert -k-, -g- (Latvian prksts versus Lithuanian pi tas 'finger', etc.). These -k-, -g- were, I believe,
inserted to mark these reflexes and keep them separate from s, z from original non-ruke
law s (especially needed in Latvian and Prussian where they, the joint reflexes, had changed from
mellow , to strident s z). It was these -k-, -g-which had helped maintain the integrity, that is, the
original shape, more or less, of the Baltic syllable. This, I believe, was the major reason that the Baltic
languages stayed conservative versus the innovations of Slavic and Albanian. And mainly the
influence of Baltic prevented Slavic from becoming a lot more innovative.
We have seen that subtle differences have been more important than deceptive similarities. The lack
of aspiration in native forms in Tokharian and Baltic is a deception. It points to no real special common
prototype. The early agreement of Pre-Baltic, Pre-Slavic, and Pre-Albanian on deaspirated stops is
another deception. It, in its turn, points to no special common prototype. It is completely overriden by
the distribution of only initial sk- to Pre-Baltic versus both sk- and ks- to Pre-Slavic and Pre-Albanian.
This means that not only was Pre-Baltic merely more glottalized at the beginning, but also it was,
because of this different beginning, bound to continue having a development different from PreSlavic's. For its part, Pre-Slavic with Pre-Albanian-like starting conditions had a development which,
for a long time, was parallel to that of Pre-Albanian, most likely, its closest relative.
Noting the above, can anyone validly persist in believing in a so-called "Balto-Slavic"
protolanguage?

1 Mayer, Harvey E. 'Aspiration and Native Baltic Forms'. Lituanus. 34.2.5-18.1988. There I say that once aspiration disappears,
glottalization
does,
too.
2
.'Slavic,
a
Balticized
Albanian?'
1988.
To
appear
in Lituanus.
3 Iranian has f, , x from p, t, k plus laryngeal. Hittite has -pp-, -tt-, -kk- which indicate a tense, and, therefore, aspirated
pronunciation. Neither language matches these extensive phenomena in a voiced counterparts. Irish
has b from g"W versus g from g(h)W. Only aspiration could have prevented the reflex of g(h)W from merging with that of g"W. No
matching example exists with reflexes of k"W and k(h)W. Note that in most Indo-European languages reflex differences between
aspirated/non-aspirated
stops
are
now
phonemic
where
once
they
had
been
merely
allophonic.
4 See Mayer, 'Prussian, an Aboriginal 3-Language?' and 'Toward Reconstructing a Proto-Baltic Phonological System', to
appear in Lituanus. There and in 'Was Slavic a Prussian Dialect?'. Lituanus. 33.2.5-21,1987, I am reluctant to accept a special
genetic Baltic unity. With similar caution I speak here, for the sake of convenience, of a "Proto-Baltic continuum". Since the
delabialization of H3 to H2 preceded the deaspiration of stops, Pre-Prussian, from the start, seems to have differed
fundamentally
from
Pre-Lithuanian
and
Pre-Latvian.
5
Meillet,
Antoine. Les
dialectes
indoeuropens. Paris.
1908.
6 Contrary to popular opinion, I see no need to assume a simultaneous, uniform primary development of ruki law reflexes in all
positions in (Pre-)Slavic and (Pre-)Albanian. Slavic initial ruki law x- comes originally only from ks-. Slavic medial ruki
law x comes also from is, us, rs. These added sources plus any shifting syllabic boundaries (e.g., -ks- + -to -k- + -s-) might
well have combined to retard the change of medial ruki law s to x. Initial ks-, subject to none of this, most likely, went directly
to x(h) first. This was so because initial k- tended to be more heavily aspirated. Note the examples of Slavic x- versus
Lithuanian k- from Indo-European k-: xobof/kab ti, xljabat'/kleb ti, chrastiti/krems ti, xritati sja/kritti, xoteti/ket
ti given by Herbert Bruer in Slavische Sprachwissenschaft, 184. Berlin. 1961. Incidentally, in Middle Common
Slavic k before x, most
likely,
disappeared.
7
Mayer,
'Slavic...
'
8 These are transitional stages of *ksol-/*ksel-, ksil- (from *skol-/*skel-, skil-) to Modern Albanian hal 'cinder, chaff', l,
il (Gheg)
'I
open',
words
connected
with
Lithuanian sklti 'split', sklti 'split
off'.
9

Note

the

same in

Lithuanian virs 'top'

from s after r and s 'this'

from

Indo-European

10 At this point Slavic *tsirn, now Russian serna 'chamois' was borrowed into Baltic and immediately, via a metathesizing
pattern set in Pre-Baltic with initial ks- to sk-, became *stirn 'deer' which is now strna in Lithuanian and sti na in Latvian.
11
abej,
Eqrem.
Hyrje
n
historin
e
gjuhs
shqipe.
251.
Tiran.
1975.
12 Slavic s, z and Albanian th, dh/d, reflexes of IndoEuropean palatals, represent one general simplification of dental clusters.