LITHUANIAN QUARTERLY JOURNAL OF ARTS AND SCIENCES
Volume 33, No.2 - Summer 1987
Editor of this issue: Antanas Klimas
Copyright © 1987 LITUANUS Foundation, Inc.
WAS SLAVIC A PRUSSIAN DIALECT?
HARVEY E. MAYER
Defense Language Institute
When at the 1984 AABS (Association for the Advancement of Baltic Studies) Conference, Jules Levin said that Common Slavic was merely an Iranized Prussian dialect, I found the idea novel — novel, but incorrect. Evidently, Levin had been drifting into this position starting with his thesis, The Slavic Element in the Old Prussian Elbing Vocabulary.'1 I imagine that the more he worked with these materials, the more they looked alike to him so that eventually he found them enough alike to suggest a Slavo-Prussian protolanguage.2
But what exclusive features do Slavic and Prussian share beyond some ancient words either maintained or borrowed from language to language? The answer is none. These languages share no exclusive feature like Grimm's law which unites all Germanic languages with h versus k, th, ð, or s in cognate morphemes in other Indo-European languages. Without such phonetic evidence, mere look-alike vocabulary proves nothing.3
Common Slavic and Common Prussian both represent an unaspirated long versus short four-vowel4 satem system with s,z as reflexes of Indo-European k', g'(g'h) rather than either th, d as in Albanian and Old Persian (which is Southwest Iranian), or ð, þ as in Lithuanian. But they are not unique with this since North and East Iranian do the same. Only an absence of features splits Slavic and Prussian from North and East Iranian.
North and East Iranian like the rest of Iranian have f, th from p, t before r while Slavic and Prussian do not. Iranian f and th in other positions are not matched by the same sounds in Slavic and Prussian either. Iranian x is completely unmatched by the same sound in Prussian, and only by accident might be matched by the same sound in some rare cognate in Slavic.5 Otherwise, Iranian x and Slavic x are usually not found in the same places. Slavic and Prussian agreement on not having features like these, however, proves no common protolanguage. It is outweighed by primordial differences. The fundamental one concerns the sequence x. Its presence in Slavic alone proves that Slavic can never have been a Prussian dialect.
The question of x involves the ruki law. All evidence points to an immediate merger in Prussian, as in Latvian (to s, z) and Lithuanian (to ð, þ), or ruki law reflexes with reflexes of Indo-European palatals k', g', g'h, which is the opposite of what happened in Slavic, Iranian, Indic, Armenian, and Albanian.6
The palatals k', g', g'h were features of only Satem Dialect Indo-European. We find their systematic assimilation only in satem dialects.7 These palatals k', g', g'h were, most likely, contemporary from the very beginning with ruki law features.
Special ruki law reflexes, however, are found in only some satem languages. Still, this shows that both satemizing, that is, the rise of palatals k', g', g'h through fronting, and the ruki law with its backing of s are parallel processes compensating for the loss of laryngeals.8 The ruki law with the hushing or Slavic velarizing of Indo-European s after i, u, r, k, then, is the assimilation of front s to back y, u, r, k.9 Iranian evidence indicates that the ruki law, though not older than the rise of palatals k', g', g'h, themselves, is older than their assimilation. There we find diversity of the reflexes of palatal k', g'(g'h) but none of the ruki law ones. North and East Iranian dialects have s, z versus Southwest Iranian th, d as reflexes of palatal k', g'(g'h). But all Iranian dialects have only ð, þ as reflexes of the ruki law, reflexes mostly kept separate from those of the Indo-European palatals k', g', g'h. This is evidence that in dialectal Indo-European the appearance of the palatals k', g', g'h from pure velars k, g, gh most likely happened at the same time as the ruki law's first backwards movement of s.10 Meanwhile, we find a tendency to merge k, g, gh of any sort with s in the satem dialects since some velars are moving forward while s is moving backward. Keeping the two processes separate was doubtless difficult.
Indeed with this in mind, we might ask whether the original ruki law reflexes were some sort of ð, þ, as in Indic, Iranian, Lithuanian, and Armenian, or were some sort of voiceless x, as in Slavic, and voiced h. Thus, either original ð, þ later became x, h in some dialects, or original x, h later became ð, þ in some dialects.
In Slavic we happen to find both voiceless ruki law possibilities, velar x and palatal ð. Voiced h and þ are not found in Slavic as ruki law reflexes. These could only have arisen before voiced obstruents. Since special ruki law sequences do not occur before obstruents in Slavic, we find s rather than x or ð and z rather than h or þ in those environments.
The ruki law ð in Slavic occurs before originally front vowels e, e, i, i. 11 It coincides with ð from s of any origin.12 The traditional view is that the original Slavic ruki law reflex was x which later became ð before front vowels. This is seen as part of the Slavic first palatalization of the velars where k becomes è and g becomes þ in the same environments. For this reason, Slavists do not usually say that the reverse happened, that is, that Slavic originally had palatal ð as its ruki law reflex, kept it as ð before front vowels, and depalatalized it either to s, z before consonants or to x before back vowels. Of course, that is possible, and we shall assume that that is indeed what happened. To support this assumption, consider the following remarks about Indic, Iranian, Armenian, and Albanian.
In these languages, as in Slavic, the primary aim was to keep the reflexes of the Indo-European palatals k', g', g'h separate from those of s. Reflexes of the palatals, marked diffuse and acute, tended to be dental and palatalized. Reflexes of s tended toward opposite directions. Thus grave, non-palatalized s reflexes tended to become velars. In all the non-Slavic languages we find x and/or h, and, in Albanian gj, but not phonetically in ruki law environments. In Slavic, velars, I say, did not first appear there either, since x was sometimes used for expressive purposes and replaced both Indo-European s as in xrom- meaning 'lame' and k' as in *xold- meaning 'cold.' Like these languages, early Slavic probably had a compact, acute ð, þ as ruki law reflexes. But in Slavic the danger of s and palatal reflex merger was severe. Palatalization was characteristic of both palatal reflexes and ruki law reflexes while non-palatization alone kept non-ruki law s separate. To prevent total merger, Slavic moved its ruki law sequence before back vowels into its largely unoccupied velar slot. Indic, somewhat like Slavic, altered its ruki law sequences while Southwest Iranian and Albanian altered their palatal reflexes and thus prevented a palatal and ruki law reflex merger.
But Prussian, Latvian, and Lithuanian, unlike Indic, Iranian, Armenian, Albanian, and Slavic, seemed never to have had any restraints preventing the merging of their reflexes of the Indo-European palatals k', g', g'h with those of Indo-European s from the ruki law. And Latvian and Prussian did not seem to have had any lasting restraints against merging reflexes of palatals k', g', g'h with those of Indo-European s in any other sets of environments. The critical difference, then, between Latvian, Prussian and Slavic here is that Slavic a/ways managed to keep some of its ruki law reflexes separate from its reflexes of the Indo-European palatals k', g', g'h. Latvian and Prussian, I say, never did this, but merged these reflexes immediately.13 There- fore, Slavic cannot qualify as a Prussian dialect. Unlike Indians, Iranians, Armenians, Albanians, and Slavs, but like Lithuanians, the Latvians and Prussians identified the reflexes of the ruki law with those of the Indo-European palatals k', g', g'h and developed special features which emphasize this.
These special features, unique to Lithuanian, Latvian, and Prussian, which underscores the equation of ruki law reflexes with reflexes of Indo-European palatals k', g', g'h, are what I call "fortifying" -k- and -g-14 These began to be used, essentially, for morphological purposes. Later, depending on the language, their role became more and more expanded. They are noticeably abundant in Latvian.
Prussian, Latvian, and Lithuanian "fortifying" -k- and -g-are found primarily before an expected ruki law reflex, then before the reflexes of Indo-European palatals k', g', g'h, and finally before the reflexes of Indo-European s in otherwise analogous environments, principally when they appear before a consonant, in most cases, t or d.15 As we shall see, the primary function of this "fortifying" -k- or -g- is morphological, and as such, represents a feature borrowed from language to language like words and other morphemes. Only those languages with an early merger of ruki law s and the reflexes of Indo-European palatals, k', g', g'h, that is, Prussian, Latvian, and Lithuanian, needed this special morpheme marker, "fortifying" -k- or -g-, and Prussian and Latvian more so than Lithuanian which at least kept reflex merger ð, þ separate from Indo-European s in other contexts. In Lithuanian, this was first done to avoid confusion caused primarily by morphological elimination of the ruki law feature, as, for example, suksime 'we shall turn' rather than *sùkðime, with s, not ð, to mark the future.
Note the appearance of an etymologically unjustified -k-on -g- (-g- before voiced consonants) in the following words:
Latvian rakstit versus Lithuanian raðýti 'to write'
Latvian ligads versus Lithuanian lìzdas 'nest'
Lithuanian raikðtë, raikðtis 'tie' versus raiðtis; rìðti 'to tie'
Lithuanian þvaigþdë versus Common Slavic*gvĕzda 'star'
Prussian kleksto 'broom' versus Lithuanian klastýti 'to sweep grain off
Prussian lagzde and Latvian lagzda versus Lithuanian lazdà 'stick'16
Examples of "fortifying" -k-, -g- in ruki law environments are:
Latvian pirksts versus Lithuanian pirðtas 'finger'
Latvian auksts versus some Latvian austs 'cold' and Lithuanian áuðta 'it is getting cold.'
Latvian ligzds versus Lithuanian lìzdas 'nest'
Lithuanian raikðtë, raikðtis 'tie' versus raiðtis 'tie', rìðti 'to tie'
Lithuanian ðerkðnas 'hoar frost' versus ðirðnýti 'to cover with hoar frost'
Prussian zigzdo 'sand' and Lithuanian þiegþdros 'gravel' versus þieþdros 'gravel'
The origin of "fortifying" -k- or -g-, traceable to the first feature of Lithuanian, Latvian, and Prussian, that is, the falling together of the reflexes of the ruki law and the Indo-European palatals k', g', g'h, developed as follows. First, in those dialects of Indo-European destined to become Prussian, Latvian, and Lithuanian, special ruki law reflexes arose, reflexes of s and its voiced counterpart, z, which were said farther back in the mouth, that is, were compact rather than diffuse. Then, as the palatals k', g', g'h were assibilating, they immediately, or almost immediately, merged with the ruki law reflexes, that is assibilating k' merged with ruki law ð and assibilating g' and g'h, which, most likely, had already become g' earlier in these Indo-European dialects, merged with ruki law þ. In Lithuanian these merger reflexes remained ð, þ. In Prussian and Latvian, where, at first, they might have been transitionally s', z', they became s, z. Finally, whenever speakers of Prussian, Latvian, and Lithuanian wanted to prevent contamination or confusion of specific morphemes, most likely, those occurring in minimal pairs, or had other reasons for adjusting the phonetic shapes of given morphemes, they marked chosen ones either with k or g (that is, discontinuous back or compact), ð or þ (that is, continuous back or compact), s or z (that is, front or diffuse), or sometimes with compromise clusters kð or gþ or ks or gz, depending ultimately on the language. These clusters contain "fortifying" -k- or -g-.
Thus, from an Indo-European fourfold system of oppositions involving k/k'/ð/s and g/g'/þ/z Pre-Lithuanian became Lithuanian by merely merging the acutes and arriving at the threefold system of oppositions k/ð/s and g/þ/z. Pre-Prussian and Pre-Latvian went much farther in becoming Prussian and Latvian. They moved both k', g' and ð, þ forward from compact to diffuse so that both acutes, voiceless k', ð, and voiced g', þ, merged, respectively, with s and z. Their resulting systems were by half simplified over their old ones to twofold ones of merely k/s, g/z with k, g being compact, grave, and discontinuous against diffuse, acute, and continuous s, z since here alternations involved only s/k, z/g. For morphological purposes, mostly, as a first move, they immediately extended the k, g from k, g ruki law environments to the remaining ruki law environments where the new combinations looked very much like olds ones in -uks-/-ugz-, -iks-/-igz-, -rks-/-rgz- with etymological, non-"fortifying" -k-, -g-. Thus, we find "fortifying" morphological -k-, -g-, in places, changing -rs-/-rz-, -us-/-uz-, -is-/-iz- to -r-k-s-/-r-g-z-, u-k-s-/-u-g-z-, -i-k-s-/-i-g-z-. Only later, after morphological levelings affecting mostly -uð-, -ið- so that now we find mostly -us-, -is-, did first -u-k-ð-, -i-k-ð- with "fortifying" -k- appear in Lithuanian, probably arising first in the West under Curonian influence where this "fortifying" -k-, -g- was common, after the westward movement of Lithuanians to their present homeland.17 Since the change from s, þ (or s', z') to s, z was much more radical and complete in Prussian and Latvian than in Lithuanian, we find more cases of these "fortifying" velars in Latvian, and would, if there were more texts, in Prussian, too. Also, if Curonian influence on Lithuanian could be discounted, which I believe was a primary source of these velars in Lithuanian, I think that Prussian might then show more cases than Lithuanian of "fortifying" -k- and -g-. These velars are particularly noticeable before the reflexes of the palatals which inherited this feature from two sources: 1. Their early phonetic merger with the ruki law reflexes, 2. Their own fluctuating reflexes of k or s, g or z, which in turn helped the development of -k-, -g- before ruki law reflexes.
Thus, in Latvian we find raksts versus Lithuanian raðtas 'writing'. The special purpose this served in identifying in basic minimal pairs the morpheme meaning 'to write' and its more acute need in Latvian than in Lithuanian can easily be seen. Latvian needs "fortifying" -k- to keep raksts 'writing' distinct from rasts 'found.' Lithuanian raðtas 'writing' needs no -k- to be different from ràstas 'found.'
The choice of -k-, -g- as "fortifying" or morphophonemic marking consonants was logical primarily because as such they themselves provided the primary, optimal environment for the assimilation of front (that is, diffuse) s to back (that is, compact) ð (or s'). Also, the fluctuation in reflexes itself of the palatals from velar k, g to sibilant ð, þ (or s', z') and s, z played a part in this. Thus, the falling together of ruki law reflexes with the reflexes of the assibilated palatals k', g'(g'h) to set Prussian, Latvian, and Lithuanian apart from the rest of Indo-European was a final logical step toward completion of the tendency to merge acute compact consonantal sequences to which the presence of "fortifying" -k-, -g- with their compactness attests. Though "fortifying" -k-, -g- were added later in different proportions and various times in Prussian, Latvian, and Lithuanian, still their presence tells us that the ruki law and the velars were connected from the beginning, partly, no doubt, by the role of the velars themselves as participants in, that is, environmental contributors to the ruki law. Once the distinction between ruki law and assibilated palatal reflexes became essentially blurred, the function of velars as environmental contributors became expanded not only to original ruki law situations, but also to areas where no ruki law had existed. Thus "fortifying" -k-, -g- were placed before assibilated palatal reflexes to provide a neo-ruki law environment. This happened principally in Latvian and Prussian when no s' (or ð), z' (or þ) existed to provide a contrast to s, z for morphological purposes.
All these developments show that possible initial or early similarities between Slavic and Prussian are insignificant against their initial or early differences which are reflected by the following features showing the Slavic could not have been a Prussian dialect that was later "Iranized" the way Lithuanian, most likely, really had been.
Prussian has "fortifying" morphophonemic or morphological marking -k-, -g-. Slavic does not. Slavic with even its opposite trend of systematic reduction of consonant clusters has no trace whatever of these Prussian features. Prussian had a reason for developing such features which Slavic did not which was the falling together of reflexes of the ruki law and the assibilated palatals k', g'(g'h) which led to the merger of reflexes of the ruki law, of assibilated palatals k', g'(g'h), and of s(z) in non-ruki law contexts all into s, z, which, incidentally, we also find in Latvian. This we do not find in Slavic. In fact, the presence of "fortifying" -k-, -g- is evidence that compact ruki law reflexes had once existed before consonants, even obstruents, in Pre-Prussian and Pre-Latvian. We have no evidence of this at all in Slavic where the development of compact ruki law features was inhibited or their maintenance discouraged wherever a consonant followed. This was a purely early Slavic limitation showing a considerable early difference in the development of ruki law features, between Slavic and Prussian.
Levin at the Montreal conference said I was exaggerating the importance of the ruki law and that the loss of its special reflexes was merely a matter of area where it was characteristic, regardless of origin, of any local dialect, be it Polish or even Yiddish, to lose ð and þ. This, I say, cannot apply to early Prussian, or even late Prussian and late Latvian, which under different, new conditions recreated ð and þ, that is from s + j and z + j, and, in the case of Latvian, also from t + j and d + j. Otherwise, what was it about this particular "area" that caused ð and þ to become s and z in Polish and Yiddish dialects? I believe it had something to do with one of its possible intrusive dialects, and I do not mean a Finnish, or Estonian, or any other Finnic dialect.
We must not forget that the area in Poland whose dialect shows the most confusion of ð/s, þ/z is in and around Torun, the area of Warmia. I find it significant that Torun, a northern Polish city, is due south of Old Prussian territory. Since Poland, for many decades, was a haven for those escaping German persecution in the late Middle Ages, Torun was a logical place for Prussians avoiding Germans to go. The resulting bilingualism or multilingualism provided the right conditions for simplification of sibilant systems of dialects in the area, that is, Polish and Yiddish ones, all influenced by the confusion caused by Prussians whose language had far fewer cases of ð and þ, especially in cognate morphemes and who, therefore, had shibboleth problems with Polish or Judeo-German.
Clearly, Levin's remark that Slavic as a Prussian dialect was "Iranized," that is, phonetically influenced by Iranian, is untrue concerning ruki law reflexes. A look at Iranian shows that its influence could only have been retentive, not interpretive, innovative, restorative or reinstative, as far as these sequences were concerned.
Slavic, whether or not a Prussian dialect, would have had to have come in contact very early with Iranian, had it needed its help to develop x as a special reflex of the ruki law. If Iranian contacts with this "Slavo-Prussian" had occurred after the Pre-Prussian merger of the reflexes of the ruki law with those of assibilated palatals k', g'(g'h) and Indo-European s in other contexts, it would have been too late for the Slavs to have interpreted Iranian ruki law ð, þ as anything but s, z. They would also not have had any inspiration to innovate with x since Iranian had x and h for Indo-European s mostly in non-ruki law contexts. And they would have had, of course, nothing to retain. If Iranian contacts with this "Slavo-Prussian" had occurred, still in Pre-Prussian, after the merger of the reflexes of the ruki law with those of assibilated palatals k', g'(g'h), but before the merger of all these reflexes with those of Indo-European s in all other contexts, assuming this intermediary stage existed, then why did only the Slavic ruki law reflex become x? The reflex of assibilated palatal k' would also have become x. And the Slavic reflex of assibilated palatals g'(g'h) should have become h (but did not). In Lithuanian which, most likely, did come under Iranian influence so that it probably was inspired to retain or maintain its compact ruki law features, ð, i, it also kept ð, þ from the assibilated palatals k', g'(g'h). These same ð, þ from both sources have retained hush pronunciations which set them apart from the hissed pronunciations of s, z from Indo-European s in other contexts, but, in themselves, do not indicate their differences in origin, which is what we should expect, but do not indicate their differences in origin, which is what we should expect, but do not get in Slavic. Iranian influence, apparently, affected only the pronunciation of Lithuanian ð, þ, not their distribution. Note these limited results in a possibly truly "Iranized" West Satem language.18
The most likely time for Iranian influence on Slavic to have been effective even for retaining a special pronunciation of a ruki law feature or sequence was before its pronunciation had changed from ð followed by a back vowel to x. This assumes that ð followed by a front vowel never changed to x in Slavic. This was the time before Slavic o following a j (yod) changed to e, an Early Common Slavic change preceding the Middle Common Slavic one of ai (from earlier oi) to e (yat') which caused the second palatalization of the velars.19 In those early days Slavic, like Iranian, had a hush pronunciation for the ruki law sequence and other pronunciations for sibilants of other origins. Once Slavic had changed ruki law ð to x, one would expect Iranian influence to have weakened since Iranian regularly used hushes for ruki law features and had x or h for Indo-European s in other contexts which would have bewildered Slavs.
The most unlikely thing about all this supposed early Iranian influence on this supposed "Slavo-Prussian" prototype was its persistence in guiding only Slavic to keep a ruki law reflex distinct, first as ð separate from any reflex of Indo-European palatals k', g', g'h and then as x separate from non-ruki law s. Thus, once x became a new ruki law reflex, it could serve only as such and not be used in place of any other kind of s heard in matching places as x in Iranian cognates. Now if "Slavo-Prussian" and Iranian contacts were influential only before ruki law ð became x in Slavic, no longer "Slavo-Prussian," then are we to believe that at this very early historical period these so-called "Prussians" were so isolated from the rest that only their dialect could have been affected? Would not the rest of the Prussians or Pre-Prussians have thought it useful to keep a special ruki law reflex once they heard the "Slavo-Prussians" insistently use it? They never seemed to have. Thus, with only the physical presence of ruki law ð common to both languages, it is unlikely that only one dialectal variant of Prussian, "Slavo-Prussian," would have been so susceptible to Iranian influence that it would show an interruption of its basic "Prussian" merging drift of ruki law and Indo-European palatal k, g, gh reflexes. Slavic, therefore, must be seen correctly as not a Prussian dialect, but as a non-Prussian one which shows non-Prussian non-merging drift of ruki law features from the very beginning with reasons of its own for keeping a special ruki law reflex.
Since Slavic contact with Iranian had to have been established before the reflexes of the ruki law and of the assibilated palatals k', g'(g'h) had merged in Prussian, that is, most likely, when Slavic was still a dialect of Indo-European, why should we call Slavic a dialect of Prussian? What special phonological or phonetic features or even feature did Slavic then, or even later, share with Prussian to be considered a Prussian dialect? The answer is, once again, none. Whatever special ancient lexical correspondences there are or may have been between Slavic and Prussian can be explained, if not as an accident, then as having resulted from contacts caused by adjacent homelands starting from very, very early times which allowed Prussians and Slavs to borrow words continually from one another, or, more exactly, from one another's languages.
It is my view that the Slavs were the most westerly speakers of a satem language or dialect type which maintained at least one distinct, special reflex of the ruki law.20 Their position was unique since the Lithuanians to the East and the Latvians, Curonians, and Prussians to the North spoke language or dialect types which no longer did. Though it might be tempting to accept a "Slave-Prussian" protolanguage, that is, a somewhat abbreviated version of the old theoretical "Balto-Slavic" one with its more obvious problems stemming from some irreconcilable differences between Common Slavic and a Common Baltic reconstructed from Lithuanian and Latvian as well as Prussian data, to do so would be erroneous.
Many of what seem like "Slavo-Prussian" concordances are late and incomplete and usually unsystematic and sporadic and are, therefore, misleading. They are, for that reason, no more and may even be less valid as indications of a protolanguage than the alternations k/c, g/dz in Slavic and Latvian. Thus, the sporadic cases of ja to je in Prussian manuscripts are no more valid as proof of a special protolanguage for Prussian and Slavic with its regular jo to je than the sporadic parallel cases of Slavic-like ju to ji in Iranian dialects are as proof of a special Slavo-lranian protolanguage when one considers that these sporadic Prussian and Iranian phenomena hardly match the consistent systematic fronting of back vowels following j (yod) in Common Slavic and its immediate successors.
Jules Levin will have to adjust his thinking to the fact that the difference in reflexes of the ruki law in Slavic and Prussian raises serious objections to a theoretical "Slavo-Prussian" protolanguage. Though by doing away with embarassing Lithuanian and Latvian data like vowel systems that do not seem to match a Common Slavic one,21 Levin may think that matching Slavic and Baltic features will go more smoothly if the Baltic ones are only Prussian, he will still be in error. Slavic, in all dialects, has kept in some positions a special ruki law feature, x, one distinct from those of assibilated palatals k', g'(g'h) and Indo-European s in other contexts. Prussian, like Latvian and Lithuanian, did this nowhere. Therefore, Slavic never was a dialect of Prussian. Still, without Levin's views to combat, I might never have discovered much of what I mention in this paper.
1 Jules F. Levin, The Slavic Element in the Old Prussian Elbing Vocabulary. (Berkeley: University of California Press, 1974.)
2 I, not Levin, coined the term "Slavo-Prussian protolanguage."
3 I make this point in my article, "Two Linguistic Myths: Blato-Slavic and Common Baltic," Lituanus, 27 (1981), 64-65.
4 For the sake of argument we allow ourselves here to interpret all evidence to mean that from Indo-European Slavic and Prussian emerged with long versus short i/i, u/u, e/e, a/a. These long versus short a/a equaled, respectively, long versus short a/a and long versus short o/o in other Indo-European languages.
5 For Example, the same root ending in -k followed by different suffixes, -s- in Slavic, -r- in Iranian, would both have x, but from different sources: *rek-s- would (and does) become rex- in Slavic while *rek-r-would become *raxr- in Iranian.
6 Harvey E. Mayer, "Baltic Membership in the West Satem Subgroup," Journal of Baltic Studies, 11 (1980), 357.
7 I prefer to view k', g', g'h, transitional stage from Proto-Indo-European k, g, gh, as having originated in dialects of later Common Indo-European rather than much later in the daughter languages since the common tendency to develop these palatals, despite later differences in reflexes, was so wide-spread in Indo-European. Also, no daughter language has kept old k', g', g'h.
8 Mayer, "Baltic Membership . . .," 357.
9 This assimilation essentially involved consonants. Thus, i was an allophone of y while u was an allophone of w with gravity, that it, peripheries of the oral cavity, as the common denominator of u and w. So diffuse s assimilated to compact i/y, u/w, r (velar), k.
10 Mayer, "Baltic Membership . . . ," 357.
11 Ruki law ð plus originally long e becomes ða in Slavic.
12 The sequence ð in Slavic can also be traced to s plus j as in Russian nað 'our'/nas 'us' where s is original and in bróðu 'I throw'/brósit' 'to throw' where s is a Slavic reflex of Indo-European palatal k'.
13 There are two arguments supporting my contention that the reflexes of the ruki law and the Indo-European palatals k', g', g'h fell together early in Prussian, Latvian, and Lithuanian while they did not in Slavic, Iranian, Indic, Armenian, and Albanian.
I. After the dissolution of Common Indo-European, no Indo-European dialect has ever been recorded as having had and then lost entirely a special, exclusive ruki law feature, nor is any Indo-European language recorded with some dialects having a special, exclusive ruki law feature. Thus, uniform minus-special-ruki law results in the dialects of Prussian, Latvian, and Lithuanian and uniform plus-special-ruki law results in the dialects of Slavic, Iranian, Indic, and Armenian coupled with special reflexes of the Indo-European palatals k', g', g'h in Albanian support my theory of the early convergence of reflexes in Prussian, Latvian, and Lithuanian. See Mayer, "Can the Baltic Languages be Divided by the Reflexes of Indo-European k', g', g'h?" Journal of Baltic Studies, 16 (1985), 83-85.
II. In Indo-European word-initial tk'-, dg'-, dg'h-, k't-, g'd-, g'hd-, did not seem to exist. But word-initial st- was common. So was word-initial ks-. Yet, mysteriously, word-initial st- was common. So was word-initial ks-. Yet, mysteriously, word-initial ks- or kð- does not exist in Prussian, Latvian, and Lithuanian. Pre-Prussian, Pre-Latvian, and Pre-Lithuanian dialects of Indo-European the merger of palatals k', g'(g'h) with ruki law reflexes was evidently in progress. Since there was to be no compact reflex ð, þ etc. from k', g'(g'h) plus þ, d or any other stop in word-initial clusters, any same-sounding one seemed out of place there, be it even from the ruki law, since only diffuse s plus stop, like st-, seemed proper initially. One way to eliminate an initial ruki law compact reflex in this position permanently was to eliminate it environment through metathesis. Hence the replacement of Indo-European initial cluster ks-by sk- in Prussian, Latvian, and Lithuanian so that we get only items in Lithuanians like skaudù 'painful' to match Slavic xudo 'poor' and Sanskrit ksudráh 'small, few, low', ksodiyan (comparative) rather than *ksaudù, *kðaudù, or even *ðkaudù. Thus, the ruki law condition with k, most typical in word-non-initial position, was eliminated by metathesis very early in word-initial position and its reflex was also eliminated, that is, replaced by diffuse s-.
Further, fluctuations between k and ð (vaðkas 'wax' from *wosk'os/aðis 'axis' from *ak'sis) and ð and s (vaðkas from *wosk'os/skaudù from *kðoudu) were parallel and contemporary since Pre-Lithuanian. The formula was: in clusters without palatals k remained constant while s fluctuated; in clusters with palatals s became ð permanently while k' fluctuated. Thence, pilkðvas 'grayish'/skaudù from *kðoudu with ð/s (both, at one time, from k'). (the Slavic formula was somewhat intermediate between the Lithuanian and Iranian ones. Like Lithuanian, Slavic with osi from *ak'sis emerged with a prevailing k' reflex in ruki law contexts, assuming that Lithuanian ð here is the reflex of k, not ruki law s, which is also possible. But like Iranian, where ruki law reflexes predominate, it kept ruki law reflexes constant in clusters with k, that is, without palatals. Thus Slavic xud- from *kðoud- 'poor' like Avestan ðoithrem with Sanskrit ksetram 'country' from *kðoitr- versus Lithuanian skaudù 'painful.') The permanence of k in its ruki law contexts helped mark it for morphological marking or "fortifying" purposes. The choice in changing k to either k or ð was made initially in the Pre-Lithuanian dialect of Indo-European (when k' still existed) and was parallel to the assimilative change of s to ð before a palatal k'. The merger of the k' reflex with that of ruki law ð is seen most clearly in aðis. This alone places the merger's beginning in Indo-European. Thus, Lithuanian ð has three sources: Indo-European k', ruki law s (progressive assimilation including k'), s followed by Indo-European palatal k' (regressive assimilation involving k'). The merger of ruki law ð with ð from palatal k' is as old as the change of k' to ð or k as in aðis or vaðkas. Then, ð from k', k'ð (ruki law), ðk' Were all pronounced alike.
14 In Prussian and Latvian, which along with Lithuanian, are the only Indo-European languages with the demonstrable total merger of the reflexes of the ruki law and the Indo-European palatals k', g', g'h, and also the only Indo-European languages with morphological marking "fortifying" -k-, -g-, the reflexes of the ruki law and the Indo-European palatals k, g, gh could have first fallen together to ð, þ as in Lithuanian, or to other sequences like s, z as in Indie, s, þ as in Polish, s', z' as in Russian, or th, dh.
15 Since "fortifying" -k-, -g- occur only in word-non-initial position, the commonest ruki law environment, I assume it developed first in ruki law contexts. Before consonants, in imitation of these processes, Lithuanian, Latvian, and maybe Prussian regularly metathesized the clusters ðk/sk, þg/zg to kð/ks, gþ/gz as in Lithuanian plaðtakà 'flat of the hand, palm' (with no k); Latvian plaskains 'flat'/Lithuanian plókðèias (from *plað-k-tias) where -k- here, not by accident, resembles a "fortifying" one. Christian S. Stang in his Vergleichende Grammatik der Baltischen Sprache (Oslo: Universitetsforlaget, 1966), 110-111, in fact, suggests this metathesis as a possible source of "fortifying" -k-, -g- in general. I believe the opposite happened, that "fortifying" -k-, -g- motivated these metatheses which occurred as a sort of analogy in sympathy with insertions of velars at first in contexts similar to raksts 'writing'/rasts 'found' (Latvian).
16 Lazda without -g- also occurs in Latvian. All Examples in these lists are from Stang, 108-110.
17 Mayer, "Baltic Membership . . . ", 359.
18 I coined the term "West Satem" in my "Baltic Membership ..." article.
19 Early Common Slavic started with only two palatal sequences, j, that is, y (an allophone of i) and ruki law reflex ð beyond possible s', z' from Indo-European k', g', (g'h) which depalatalized before the changes j+o to j + e, etc. (Note suto 'hundred', *zolto 'gold' with s from k and z from g'h). The ruki law reflex remained ð everywhere before vowels as long as all palatal sequences appeared before back vowels. Once the combinations j plus o or u of any length/and in a previous article, "Kann das Baltische als Muster für das Slavische gelten?" Zeitschrift für slavische Philologie, 39 (1976), 32-42, I demonstrate that in Early Common Slavic short a, o had fallen together to short o before long a, o had failed together to long a/ started becoming j plus e or i of any length (assimilation class one, a progressive one like the older dialectal Indo-European ruki law, with vowels assimilating to the preceding palatal sequence), it was clear that the same would happen with ruki law ð plus o and u of any length. This, along with the first palatalization of the velars (k, g plus front vowel to è, þ plus front vowel), was part of a general tendency to reduce distinctive feature oppositions in syllables and was apparently being matched simultaneously by a change with a similar aim, that of reducing consonant clusters, so that s plus j and s (or then, perhaps, still s') from Indo-European palatal k' plus j, and ruki law ð plus j were all tending to simplify to ð. To maintain a special ruki law feature at least somewhere while these changes were starting, or still in progress, the Slavs changed ruki law ð to x in the only remaining favorable, phonetically neutral environment, that is before back vowels (assimilation class two, a regressive on, more typical of Slavic, this time, however, affecting a ruki law feature rather than effecting one, with the palatal sequence assimilating to vowels) while they remained back vowels. (Once back vowels were completely fronted after j so that only front vowels remained in that environment, the same was bound to happen after ruki law ð if it were to remain a palatal sequence everywhere. The Slavs were, apparently, afraid of this change leading to the total loss of a special, distinct ruki law reflex.) The total fronting of long and short rising diphthongs jo, jo to je, je was eventually matched later in counterpart long and short falling diphthongs ai (from earlier ai, oi) and ai (from earlier ai, oi) resulting in monophthongs e and sometimes i. These e and i gave rise to the second palatalization of the velars including the new velar x from the ruki law where k, g, x plus e (from long ai. and short ai) or i (sometimes from short ai) became c, dz, ð
20 Mayer, "Baltic Membership . . .," 361.
21 Lithuanian and Latvian have inherited five-vowel systems, not four-vowel ones, from Indo-European.