I, GREEK (K)SUN AND PROTO-GREEK "*KSX-"

The early Greek preposition /particle meaning 'with; together' currently lacks a convincing etymological explication. 1 What makes the analysis of Greek 'with' complex is the fact that its initial letter shows an odd indeterminacy that runs counter to the sound correspondences as we currently know them.

INITIAL KSI

Homeric (ProQuest: ... denotes non-USASCII text omitted.) (chiefly in compounds)

Early Attic (ProQuest: ... denotes non-USASCII text omitted.)

Ionic verse (ProQuest: ... denotes non-USASCII text omitted.)

Linear B ku-su

INITIAL SIGMA

Homeric (ProQuest: ... denotes non-USASCII text omitted.)

Later Attic (post 5th-c. B.C.E.) (ProQuest: ... denotes non-USASCII text omitted.)

Ionic prose (ProQuest: ... denotes non-USASCII text omitted.)

Boeotian (ProQuest: ... denotes non-USASCII text omitted.)

Note that the distribution of ksi and sigma in this word does not follow the familiar dialect breakdown of early Greek - indeed, the fact that both shapes occur within a single dialect precludes a simple dialectological approach to the problem, since it would presumably be necessary to posit, e.g., that Attic (ProQuest: ... denotes non-USASCII text omitted.) reflects a fifth-century borrowing from an external source which entirely supplanted the indigenous shape (ProQuest: ... denotes non-USASCII text omitted.). It is difficult, however, to imagine why such a basic element of the vocabulary should have been suddenly replaced with a cognate from outside Attic.

The fact that the apparent replacement of ks- by s- in Attic correlates with a historical point in time - the fifth century B.C.E. - suggests that we find before us the results of a phonological shift from a more archaic early shape in ks- toward a truncated later shape in s-. In the s- of dialects such as Boeotian, we would have to presume that a parallel but evidently independent shift occurred in that dialect's prehistory. If the s-form results from a secondary development, we must consequently conclude that this development took place on more than one occasion across the various dialects.

However, whether we posit a single shift or multiple shifts, it is difficult to accept that the replacement of ks- by s- represents a simple phonological change since we must account for the fact that the phenomenon seems not to encompass initial ks- in general but only the ksfound specifically in this word. Unless we wish to settle for a theory of diachronic phonology that countenances changes targeting individual lexical items, we must come up with some way to give a phonology-based characterization of the Greek word for 'with' that uniquely defines this word and distinguishes it from all other words containing initial ks-. In principle, we might posit that, as a function word, (k)sun had unusual suprasegmental characteristics that might account for the fact that it has apparently been subjected to eccentric phonological developments. Such a hypothesis cannot be ruled out, but, in the absence of further parallels to this development, it likewise cannot be proven.

The above qualifications notwithstanding, in the following pages I would like to suggest that the relation between ksún and sun may indeed be viewed ultimately as a matter of conventional, lautgesetzlich phonology. The hypothesis that I would like to consider here is that the apparently eccentric reflexes of (k)sun are due to the presence of something special about the Proto-Greek Anlaut underlying ksúnlsún. For indexical purposes, let us label the onset of the Pre-Greek precursor of the word for 'with' as "*ks*-" - i.e., a sequence of elements that in some manner yet to be defined differed from the typical *ks- cluster. Using this assumption as a starting point, it is possible to posit that this "*ks*-" left differing reflexes in documented Greek by undergoing at least two contrasting developments as the historical dialects took shape:

1. In some of the Greek proto-dialects, "*ks*-" merged with simple s-.

2. In the ancestor of dialects like Attic, "*ks*-" evidently remained distinct from *ks- and only later - i.e., after the onset of written records - did the outcome of "*ks*-" merge fully with s. Prior to this merger, if this view is correct, early Attic still retained a distinction between its ks- and its reflex of "(*)ks*-" but the distinction was presumably obscured on the graphic level through the use of a single character ksi to represent both.

3. It is possible that other dialects (e.g., that of Linear B) also retained the distinction between the reflexes of "*ks*-" and *ks-, but, like early Attic, did not reveal the distinction through their graphic systems. On the other hand, one can just as readily imagine that such dialects had already lost the distinction through a merger of "*ks*-" and *ks-,

The situation in the Homeric text, which shows both ks- and s- as its reflexes of "*ks*-" is interesting. Perhaps Homeric, like early Attic, still synchronically possessed a distinct reflex of u*ksx-," which it manifested through differing allophones governed by environmental factors which are not presently apparent. Alternatively, the co-occurrence of ks- and s- in Homeric may perhaps be ascribed to contact between a dialect in which "*ks*-" emerged as ks- and another dialect in which "*ks*-" resulted in s.2

II. COMPARANDA

With this perspective in mind, I would like to adduce two pieces of comparative IndoEuropean data that suggest that there may in fact be something rather complex underlying the historical phonology of the onset of this Greek preposition. Before introducing this material, however, let me briefly reflect on the remainder of the elements of which the Greek word (k)sun is composed.

1. The underlying phonology of the sound marked by the upsilon of (k)sun is ambiguous. While the obvious source for Greek u is the Indo-European high back vowel *«, there is ample evidence that, under certain circumstances, Greek -H- was also the outcome of an earlier short *o - cf. Greek núks < IE *nokwt-, dial. Greek ónuma 'name' < IE *Hnomn (Cowgill's Law). Earlier attempts at analyzing (k)sun have taken the -H- here to be an example of this raising phenomenon, 3 and, guided by comparative data to be made clear below, we shall adopt this interpretation of the vowel of (k)sun and regard it as the reflex of an original mid-height vowel *o.

2. The original nature of the nasal of (k)sun is similarly unclear, since pre-Greek lost the distinction between *n and *m in word-final position. In what follows we shall use the character *N to represent one or another of the Indo-European nasal phonemes.

m. IRANIAN *XSAN

The first Indo-European datum is drawn from the modern Eastern Iranian language Ossetian, a relative of the Scythian and Alanic languages, which survives in the mountainous area straddling the Russian-Georgian border. The Ossetian lexicon includes an element axsan 'midpoint', which combines a close formal match to the earlier Greek form ksún with an ancillary postpositional function ('among') that is not far distant from that of the Greek preposition. In the following examples, drawn from those cited by V. Abaev in his Istorikoétimologiceskij slovar' osetinskogo slovar' (s.v. xsan), the relevant postpositional phrases are italicized. The first two examples are in the Iron dialect, on which literary Ossetian is based; the final example is in the western dialect, Digor.

"Among your sheep is a good, well-fed ram."

"They sat down among the people."

"Why am I without glory among my companions!"

Schwartz has already suggested a connection between Ossetian äxsän and Greek (ProQuest: ... denotes non-USASCII text omitted.).4 If we bear in mind that Ossetian introduced a prothetic vowel into words that originally began with a consonant cluster, and if - as we noted above - we allow for the possibility that the upsilon of ksún reflects the secondary raising of an Indo-European *o, it becomes possible to reconstruct a common forerunner of both axsan and (ProQuest: ... denotes non-USASCII text omitted.) on the order of *KsoN-. 5

The postpositional role of Ossetian axsan patently represents a secondary development, the basic function of the word being that of a noun meaning 'middle'. A postpositional phrase such as adamy xsan, in which the noun phrase stands in the genitive, is thus transparently "(at) the middle of the people." It is not difficult to imagine how a construction starting either as "*at the midpoint of a group of entities" (in which the "midpoint" was the head of a noun phrase serving in a locative function) or as "*in a group of entities, at the midpoint" (in which the locatival "midpoint" was an adverbial paralleling the "group," which itself took the form of a locatival phrase) could have developed into an expression of comitativity once the restriction that the formation refer to a plurality had been lost.6

Abaev's lexicon, the standard reference work on Ossetian etymology, makes no reference to Greek (k)sun, preferring instead to follow Vsevolod Miller's juxtaposition of axsan with the Sanskrit noun ksana- 'moment'. The adduction of ksana- is based on the presumption that a form meaning 'midpoint' could have developed from one meaning 'moment'. While it is indeed possible to envision a semantic development linking 'moment' and 'midpoint' -Abaev posits the chain '*moment' > '*interval (of time, space)' > '*center' > 'middle' - the connection is not transparent enough to be fully satisfying. Ossetian axsan does not appear in the discussion of ksana- in Mayrhofer's etymological dictionary of Sanskrit, which prefers linking ksana- to the word for 'eye' (aksi-).7

However satisfactory the semantic link between axsan and ksún may be, connecting the two brings us no closer to understanding the anomalous onset of the Greek preposition. There are two familiar phonological sources for an Iranian *xi-cluster, only one of which is compatible with a Greek ks-, This is a proto-sequence composed of an Indo-Iranian velar (< an IE labiovelar or pure velar) followed by an *s, which by all rights ought to have retained the cognate of the (labio)velar intact in Greek. The most obvious interpretation of the relation between ksún and axsan thus provides us with no insight into the problem of how sun came to lose its k-.

The other potential source of Iranian *xs- seems even less promising as far as sun is concerned. It has long been known that there is a small but significant set of instances in which an Indo-Iranian sequence of *velar or *palatal + *-s- corresponds to a Greek sequence of a velar or labiovelar followed by a dental. While such cases led earlier Indo-Europeanists to speculate that a lost "interdental" series should be reconstructed for the ancestral language, it has since become widely accepted that such cases are better traced back to Indo-European tautosyllabic sequences of *dental + *palatal/*postpalatal - this is seen most clearly in Anatolian and Tocharian A, where as reflexes of the term meaning 'earth' we find tekan and tkam, respectively, as the cognates to Greek khthón- and Sanskrit ksám- (< IE *dhgh(e/o)m-). In principle, therefore, the -xs- of Ossetian axsan might reflect an original IE *tkw- or *tk-, although in this case we would of course expect its Greek counterpart to show as its Anlaut *kt- rather than either ks- or s-.

IV. GREEK "*KSX-" = IRANIAN *xSl THE OUTCOME OF A TRISEGMENTAL CLUSTER?

The -xs- of Ossetian axsan thus seems to go back to either *Ks- or *TK-, but neither of these will in itself account for the "vacillating" ks-/s- that we find in Greek. However, I would like to suggest a further question that has to my knowledge not yet been asked of Indo-European historical phonology. We know the outcome of *K + *s and that of *T + *K, but can we say what the outcome of a co-occurrence of both factors - i.e., of a sequence of *dental + *(labio)velar + *s - would have been? I would like to suggest that both of these forms are reconcilable in terms of comparative phonology - i.e., that the sound laws as they currently stand may be augmented in such a way that these forms do not constitute counterevidence to the principle of the exceptionlessness of phonological change - if we use a triconsonantal cluster as the starting point for our reconstruction.

To my knowledge, we have at present no good evidence enabling us to say decisively what we should expect as the regular outcome of a Proto-Indo-European word-initial sequence *TKs-, but I suggest that neither an Iranian reflex *xs- nor a Greek reflex similar to - but distinct from - ks- should be ruled out. One might in principle envision, for example, the first two elements of the proto-sequence (*TK) developing before *s as they do before a vowel (i.e., *TK-s- > Pre-Greek *KT-s-, Pre-Indo-Iranian *Ks-s-), with the following *-s- subsequently melding in one manner or another with the dental/sibilant with which it came into contact as a result of the metathesis.

Depending upon one's view of how the original Indo-European *TK- came to yield the correspondence of Greek *KT- to Indo-Iranian *K$-, it is certainly possible to imagine other developmental sequences by which a Proto-Indo-European *TKs- might have yielded both Greek "*&,$*-" and Iranian *x$-. Regardless of the details of the paths through which *TKsmight have developed, I suggest that the fact that the Greek particle (k)sun is a phonological hapax would be a natural outcome from a complex sequence of specific elements. As the criteria governing the occurrence of a given phonological outcome become more and more specific - as in this case, in which the precondition is an underlying Indo-European wordinitial sequence of a *dental, a *velar, and *s, in that order - the number of cases in which the necessary conditions are met will naturally decrease. In the case of (k)sun < *TKsoN, the number of cases in which all of these conditions were met - at least among surviving, identifiable etyma - seems to have been no more than one.

V. INDO-EUROPEAN *TKSON AND HITTITE TAKSAN

In Hittite we find the neuter noun taksan (tak-sa-an) 'middle', which is also used as a relational adverbial 'together'. Both of the meanings of taksan ('middle' and 'together') may be plausibly related to what we have seen in the semantics of the Ossetian and Greek forms mentioned above.

[m]m=kan A-BU O\JMV-RU=ya LUMU-DU OAM=SU=ya nasma §E§ NIN=ya balluw[anzi] n=us taksan kuwabì anda tittanumi n=us [kissan an]iyami

If a father and child, husband and wife, or brother and sister are fighting, when I bring them together I handle them [thus] . . . (KBo 2.3 i 2-4)

The close match in semantics notwithstanding, the quantity of the second vowel of takSan stands in the way of interpreting this word as a cognate to the Greek and Ossetian terms. The *o that evidently underlies the upsilon of (k)sun would presumably have yielded a Hittite long -a-, for which one would expect a piene spelling (*tak-$a-a-an). In morphological terms, Hittite taksan seems to be a derivative in -an constructed from the verbal root *taks- 'join' (cf. maskan 'gift', lagan 'inclination', etc.).8 At the same time, the semantic overlap between the two senses of taksan ( 'together' and 'midpoint') and the meanings of the Greek and Ossetian forms adduced above is striking - indeed, the meaning 'midpoint' agrees more neatly with axsan than it does with taksan's etymological root (taks-), since the latter reflects most typically 'joining' in the sense of 'producing a structure' rather than 'forming an assemblage'.

The similarity between the verbal root tafcS- and the sequence posited above as the hypothetical source for the Anlaut of *TKsoN raises the possibility that takSan and *TKsoN might simply be two parallel derivatives from the same verbal root - i.e., *TKsoN would most likely have originated from the same verbal root as a neuter zero-grade thematic noun structurally analogous to *yug-o-m.9 This interpretation is gainsaid, however, by the -x- of Ossetian axsan. The -k- of the verbal root taks- seems clearly to go back to an Indo-European palatal, the reflex of which is lost in Iranian before *-s- (cf. Avestan tasa- 'cut, make', taSan- 'creator' (compare Sanskrit taksa-, taksan-) from Indo-Iranian "*tas$-" presupposing Indo-European *teks-.10

A hypothetical avenue by which these discrepancies might be resolved is by positing that an inherited Pre-Hittite *taksan- '*middle, (in the) midst' - i.e., an Anatolian cognate to Ossetian axsan and (k)sun < *TKsoN - became secondarily subsumed under the verbal root taks-. This could have come about either through a confusion of two etymologically distinct but superficially similar terms - viz., a *taksan inherited from Indo-European *TKsoN became contaminated through contact with a Hittite deverbal neologism *taksan- '*a joining' - or simply by a folk reanalysis taking *taksan to be derived via the stem-suffix -anfrom tak$-. In either case, an interpretation along these lines, while speculative, would afford a fairly simple way to account for the semantics of taksan, which otherwise are not easily accommodated with those of its putative root source taks-.

VI. POSTSCRIPT: AN ALTERNATIVE INTERPRETATION OF SANSKRIT KSAIUA

As was noted above, the standard etymological lexicon of Ossetian adduces Sanskrit ksana- 'moment' in its discussion of axsan. To be sure, it is not impossible to imagine a semantic connection between 'moment' and 'midpoint', but the connection requires enough of a leap that adopting this view demands a measure of faith. The alternative explanation for ksana-, reflected in the adduction of the term for 'eye' (Skt. aksi- < *Hekw-s-i), is similarly conceivable but entails a comparable semantic jump.

I would like to suggest that, as an alternative to Ossetian axsan and Sanskrit aksi-, a further form may be adduced that leads to an etymology for ksana- that is no less phonologically viable but more semantically transparent. It was noted above that Sanskrit ksmay be the outcome of an earlier tautosyllabic sequence of *dental + *(labio)velar/*palatal. The documented Celtic outcome of such a sequence is a simple dental - cf. Irish art 'bear' (< *Hrtko-, cf. Sanskrit fksa-, Greek (ProQuest: ... denotes non-USASCII text omitted.), Hittite hartaga-), Irish du (gen. don) 'place' (< *dhghom- 'earth', cf. Sanskrit ksám-, Greek (ProQuest: ... denotes non-USASCII text omitted.), Hittite tekan). l suggest that we add to the limited list of identified Celtic reflexes of this phenomenon the Irish a-stem noun tan 'time', perhaps most familiar as a marker of temporal subordination.

When (lit. 'the time that') she bears a child, after that she is joyful. [Ml. 129c8]

The juxtaposition of Sankrit ksana- and Irish tan points straightforwardly (if perhaps unobviously) to a reconstructable etymon *tKno-, *tKna-, This, it is suggested, is a more semantically transparent explanation of tan than the conventional etymology, which relates tan to the verbal root meaning 'stretch'.11 There is thus no reason to assume any connection between ksana- and any of the various forms that are the topic of the preceding paragraphs.

1. A draft of this paper was presented at the 2007 annual meeting of the American Oriental Society held in San Antonio, Texas, in a session held jointly with the North American Conference on Afro-Asiatic Linguistics. I am grateful for the comments provided by those in attendance.

2. I leave unaddressed here the question of the Cypriote element DV- (cf. Hesychius (ProQuest: ... denotes non-USASCII text omitted.) [glossed as (ProQuest: ... denotes non-USASCII text omitted.) as well as epigraphic (ProQuest: ... denotes non-USASCII text omitted.)). While this element has been regarded as a dialectal counterpart to αv, Brugmann and Hirt compared it with Attic - see the literature cited in Standerwick, Etymological Studies in the Greek Dialect-Inscriptions (1932), 42. Perhaps one should not rule out the possibility that in pre-Cypriote the reflex of our hypothetical "*ks*-" either merged with the reflex of IE *s- or simply disappeared.

3. Andrew Sihler, New Comparative Grammar of Greek and Latin (New York: Oxford Univ. Press, 1995), 406.

4. See Martin Schwartz, "Scatology and Eschatology in Zoroaster: On the Paronomasia of Yasna 48. 10 and on Indo-European H2eg- 'to make taboo' and the Reciprocity Verbs *kwsen(w) and *megh" in Papers in Honour of Professor Mary Boyce, Acta Ironica, ser. II, XI (1985), 495. Schwartz traced xsan back ultimately to a verbal root *kwsen- 'give one thing for another, exchange, requite' (Hittite kussan- 'requital, payment', Avestan xsnu- 'requite satisfactorily, give hospitality'), seeing further derivatives reflected in Greek *ksenwos 'guest-friend, guest, host' and Irish (ar) son '(in) exchange'. I do not believe that he addressed the curious sigma of (ProQuest: ... denotes non-USASCII text omitted.). I am grateful to Hya Yakubovich for drawing my attention to this reference.

5. Here and in what follows we shall use the symbol "*K" to refer to an undetermined member of the set of Indo-Iranian velar phonemes, the precise identity of which cannot be determined in the absence of further evidence. Indo-Iranian "*K" would thus represent the reflex of an Indo-European labiovelar or pure velar, belonging most probably to either the voiceless or voiced unaspirate series (hence, IE *k, *kw, *g, or *gw). It is unlikely - but not impossible - that the "*K" of *KsaN might also reflect a member of the IE voiced aspirate series, since Bartholomae's Law leads us to expect that such an element would have retained its voicing before a following *s (i.e., IndoIranian *gzh < IE *gh-s or *gwh-s).

The historical underpinnings of the final nasal of Ossetian äxsän cannot be resolved beyond doubt. While the Ossetian reflexes of IE *m and *n largely remain distinct, in a few etyma a word-final Ossetian -n seems to reflex an earlier *m - cf. dan '(I) am', which evidently goes back to *ahmi < *esmi, the origin of the Ossetian d- here remaining unclear.

6. Regarding the second of the syntactic hypotheses - in which the ancestor of the preposition stood parallel to the locative phrase rather subsuming it into a prepositional phrase - note the instances in which Greek sun served as a adverbial 'together, at once, besides, also' rather than as a preposition - cf. Od. 10.42 (ProQuest: ... denotes non-USASCII text omitted.) and similar passages cited by Liddell and Scott, s.v. (ProQuest: ... denotes non-USASCII text omitted.) C.

7. As support for connecting 'moment' and 'eye', Mayrhofer cites German Augenblick. See below, however, for an alternative analysis for Sanskrit ksana-.

8. A. Kloekhorst, Etymological Dictionary of the Hittite Lexicon (2008), 814-15.

9. For the outcome of the cluster seen in taks- < *TKs-, cf. takn-as (ultimately < *dh$hm-os), the genitive of tekan 'earth'.

10. In principle, it might be possible to posit that, while *f was generally lost in the early Iranian cluster *-s's-, it survived in one manner or another as *x in the triconsonantal sequence we have posited above (äxsän < *tísam < *tksom), but this approach seems contrived.

11. Cf. Joseph Vendiyes, Lexique étymologique de l'irlandais ancien (1978), T-25-26.

Author affiliation:

DAVID TESTEN

RESTON, VA