chatreššar 2008 - univerzita karlovausj.ff.cuni.cz/system/files/chatressar 2008.pdf · there was...

Chatreššar 2008

Chatreššar 2008

© The editors and contributors

Executive editors:Jan BičovskýPavel Čech

Published by the Charles University in Prague, Faculty of Arts. Printed in the Czech Republic by BCS, s.r.o., Poříčany

The Chatreššar is edited by the Institute of Comparative Linguistics, Charles University, Prague.The members of the Editorial Board are: Jost Gippert (Frankfurt a.M.), Hans-Christian Luschützky (Vienna), Petr Vavroušek (Prague), Martin Worthington (Cambridge), Andrzej Zaborski (Cracow), and Petr Zemánek (Prague).

ISSN: 1803-005XISBN: 978-80-7308-280-2

ContentsAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

ArticlesDer thematische Instrumental des Plurals. Zur historischen Morphologie und Phonologie des Griechischen und Altindischen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11Ján Bakyta

Initial *x- in Slavic revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23Jan Bičovský

The Earliest Comparative Linguists . . . . . . . . . . . . . . . . . . . . . . . . . . 47Peter J. Huber

Der altbabylonische Ausdruck bīt akītim im Verwaltungskontext . . . 65Lukáš Pecha

Zur Deklination der starken Substantiven im älteren Neuschwedischen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71Hana Vaňková

NotitiæDie akkadische Synonymenliste malku : šarru . . . . . . . . . . . . . . . . . . 89Ivan Hrůša

Zur Reinheit des Körpers in der Kultur Mesopotamiens . . . . . . . . . . . . 91Walther Sallaberger

An instructive scribal error on a cuneiform medical tablet . . . . . . . . . . . 97Martin Worthington

Acknowledgments

The publication of the proceedings has been financially supported by the project MSM 0021620823 “Český národní korpus a korpusy dalších jazyků” [The Czech National Corpus and Corpora of Other Languages] of the Ministry of Education of the Czech Republic.

Editorial Note

Chatreššar was first published in 1997 as a yearbook of the Institute of Comparative Linguistics of the Charles University in Prague. As such, it mainly carried writings by the Institute’s members and associates, and was published exclusively in Czech.

The first sign of Chatreššar opening to an international scholarly readership was the publication of the proceedings of the international conference Electronic Corpora of Ancient Languages, held in Prague in November 2007. This volume appeared as Chatreššar 2007 and the core of the editorial board emerged during this occasion; it was later enlarged by the addition of a number of other respected scholars.

Thus, Chatreššar has now become an international scholarly journal with an international editorial board, where all of the contributions go through a peer-review process. Its focus is on comparative and historical linguistics (especially Indo-European and Semitic) and philology of ancient languages, especially those from the Near East.

We hope that Chatreššar will become a platform for these fields of scholarship, helping to promote original research and facilitate scholarly discussion, and that it will become an effective way of reaching the readership interested in these fields.

It has been decided that the contributions published in Chatreššar will be divided into articles and short notices (Notitiæ).

We would like to thank all of those who have helped this Journal to appear in its new shape, both the authors who submitted their articles, and those on the editorial side, who have contributed their scholarly expertise or technical skills.

The Editors

Articles

Der thematische Instrumental des Plurals Zur historischen Morphologie und Phonologie des Griechischen und

Altindischen1

Ján Bakyta

The article shows that there is only little evidence which would speak for the reconstruction of the instrumental plural ending *-ōys in the Indo-European thematic declension, with Avestan -āiš being the one possible exception. The ending should be reconstructed as *-oys (which is preserved in Greek) from *-oh1is, where h1 marks the instrumental case, i is a particle, and s is the plural marker. Consequently, the alternative Vedic thematic instrumental plural ending -ebhis arose probably from *-aHi-bhi-s by conservation of the particle i and by repeating the sign of instrumental, and so did the instrumental singular ending -ena/ā from *-aHi-n-aH. In the Old Indic instrumental ending -ais, there was never a long diphthong and we may ask whether the Old Indic are not only short diphthongs which were not monophthongized to o, e for some reason. The demonstration that Greek -οις is not shortened from the alleged Indo-European *-ōys also reduces the evidence for the Osthoff’s Law, a Lautgesetz in the history of Greek vocalism.

Als Hermann Osthoff 1881: 1593 das später nach ihm genannte Lautgesetz formulierte, nach dem im Urgriechischen Langdiphthonge vor Konso-nanten oder Langvokal vor Liquida / Nasal und Konsonant gekürzt wurde, nannte er als Beispiel zweiter Stelle den „Instrum. Plur. der o-Stämme -οις (ἵπποις) aus *-ωις = sanskr. -āis“. Dass die griech. Endung -οις auf *-ωις zurückgeht, hat Osthoff (1879: 52-64, besonders 58) schon zwei Jahre zuvor behauptet, ohne damals diesen lautgesetzlichen Vorgang dem Leser näher beleuchtet zu haben. Den dort angekündigten speziellen Aufsatz über den Instrumental Pl. der thematischen Deklination hat er, soweit ich sehe, nie geschrieben.

Die Gedanken von Bopp und Pott, die Osthoff 1879: 56 als seine Vorgänger in Sachen Gleichsetzung des griech. -οις mit dem altindisch. -ais bezeich-net hat, brauchen wir hier nicht zu verfolgen. Osthoff hielt es für möglich,

1 Dieser Artikel entstand im Jahr 2008 mit Unterstützung der Grantagentur der Karlsuniversität Prag: Grant-Nr. 8662/2008.

Ján Bakyta12

dass die Dativ-Lokativ-Instrumental-Endung -οις z.T. auch aus der loka-tivischen Endung -οισι verkürzt ist, wahrscheinlicher aber schien ihm, dass im Gegenteil die letztgenannte nur eine auf dem instrumentalen -οις < *-ωις beruhende Neubildung, eine der „ephemere[n] auf den markt gebrachte[n] sprachliche[n] kunstproducte der sängerzunft“ (1879: 62) sei, da man sonst im -οισι mit Restitution von -σ- > /-oysi/ rechnen müsste (1879: 59).

Diese Ansichten hat kurz vor seinem Tod Johannes Schmidt entschieden abgelehnt: die alte idg. thematische Endung des Instrumentals Pl. sei im Griechischen überhaupt nicht mehr erhalten (ergibt sich aus Schmidt 1905: 3), das als Instrumental sowie Dativ und Lokativ verwendete -οις sei aus dem lokativischen -οισι verkürzt, wofür allerdings nicht so sehr oder über-haupt nicht die Verallgemeinerung der elidierten Endung im antevoka-lischen Sandhi (-οισ’ V-), wie früher behauptet, als vielmehr die Übertra-gung der Gestalt des proklitischen und deswegen apokopierten Artikels auf die darauf folgenden Nomina (also τοισι θεοῖσι > τοις θεοῖσι >> τοις θεοῖς, schon seit dem Altertum jedoch akzentuiert τοῖς θεοῖς) verantwortlich sei.2

Seit der Erschließung des Mykenischen ist aber klar, dass mindestens ein griech. Dialekt im 13. Jahrhundert v. Chr. im Plural der thematischen Dek-lination noch die Endungen des Lokativ-Dativs /-oy(h)i/ und des Instrumentals /-oys/ oder /-ōys/ nebeneinander besaß.3 Es ist also offensichtlich, dass -οισι keine Neubildung und bestimmt kein „kunst-product der sängerzunft“ sein kann. Andererseits kann man nicht mit J. Schmidt -οις als alte Instrumentalendung völlig aberkennen. Die Schmidt’sche Apokopetheorie vermag uns jedoch zu erleuchten, wie und wann die Elimination der Lokativ-Dativ-Endung -οισι ihrer Form nach begonnen hat: das konnte frühestens dann geschehen, als das Demonstra-tivum ὁ, ἡ, τό als Artikel benutzt wurde, wozu sich in den homerischen Epen etwa aus dem etwa 8. Jahrhundert erste Ansätze finden.

Es ist also völlig in Ordnung, dass die neueren historischen Grammatiken des Griechischen (Rix 1976: 141f.; Sihler 1995: 263f.) für die gr. Dativ-Loka-tiv-Instrumental-Endung des Plurals -οις mit zwei Quellen rechnen, mit dem idg. Lokativ und Instrumental, nur dass sie alle nicht anzugeben ver-mögen, wie es zur Elimination der Form -οισι gekommen ist, da ihnen 2 Schmidt 1905: 4-48, insbesondere 22-26.3 Siehe z.B. Bartoněk 2003: 188-189, dazu noch 167 (zur Unwahrscheinlichkeit der Inter-pretation als /-oys/).

Der thematische Instrumental des Plurals 13

Schmidt’sche Theorie nicht mehr bekannt ist.4 Deshalb mag sich die obige Darlegung nicht ganz erübrigen, im Folgenden wird uns aber nur die Frage beschäftigen, welche Form der idg. Instrumental Pl. besaß, aus dem die gr. Endung -οις hervorging, die letztlich -οισι verdrängte und auch deren dativisch-lokativische Funktionen übernahm.

Dass diese uridg. instrumentale Urform des griech. -οις *-ōys war, wird seit Osthoff allgemein angenommen.5 Demgegenüber möchte ich im folgenden zeigen, dass im griech. -οις keine Vokal- bzw. Diphthongsverkürzung statt-gefunden hat und dass diese Endung somit ziemlich getreu die idg. Form widerspiegelt, jedenfalls nicht minder als das altindische -ais.Wenn wir das zur Verfügung stehende Material anderer idg. Sprachen übersehen, das die Rekonstruktion von *-ōys stützen soll, so stellen wir fest, dass es mit ihm eigentlich leidlich steht.

In der Sprache der Avesta finden wir die Endung des Instrumentals Pl. the-matischer Stämme -āiš (Hoffmann – Forssman 2004: 120). Das kann hier eher den Öffnungsgrad als die Quantität des betreffenden Vokals bzw. Diphthongs signalisieren (vgl. Hoffmann – Forssman 2004: 44.54). Indes-sen ist zuzugeben, dass der avestische Befund das weitaus Beste ist, was man zugunsten des idg. *-ōys anführen kann.Noch Szemerényi 1990: 197 versichert uns, dass die Endung *-ōis „sicher auch im aks. -y steckt“. In Wirklichkeit ist dies durchaus unsicher. Wir haben einen ganz klaren Kontraktionslangdiphthong in der Endung des Dativs Sg. der thematischen Stämme, /-ōy/, der im Altkirchenslavischen als /-u/ erscheint, falls die genetische Identität als gesichert gelten kann. Wenn wir also diese Endung -y mittels eines Lautwandels aus *-oys herleiten möchten, wäre das mindestens ebensogut oder ebensowenig berechtigt wie aus dem vermeintlichen *-ōys. Dagegen kann man einwenden, dass das idg. */oy/ in dieser Sprache gewöhnlich /ě/ ergab, vgl. den thematischen Loka-tiv Sg. idg. *-oy, altkirchenslavisch -ě, Perfektstamm *woyd- > věd- usw. Doch ist hier das Bild bunter: die thematische Endung des Nominativs Pl. *-oy entwickelt sich in */-ī/ und weiter ins altkirchenslavische /-i/. Wenn *-oy unter denselben lautlichen Bedingungen, d. h. im Auslaut, bald als -ě (im Lokativ Sg.), bald als -i (im Nominativ Pl.) erscheint, wäre ein dritter

4 So Sihler; Rix dagegen erwähnt die Apokope im vortonigen Artikel. Kretschmers (1909: 56f.) Kritik der Schmidt’schen Apokopetheorie scheint mir in diesem Fall ungenügend.5 Dies zu sanktionieren half natürlich besonders Osthoffs Freund Brugmann (z.B. 1904: 397f.). Siehe weiter beispielsweise Meier-Brügger 2002: 200 für die Gegenwart; hier sind allerdings Instrumental, Dativ und Ablativ Pl. versehentlich vertauscht.

Ján Bakyta14

Wandel desselben zwar ebenfalls in einer Endung, aber vor dem auslau-tenden -s denkbar.Um aber Missverständnissen vorzubeugen, sei hier bemerkt, dass ich der Möglichkeit der lautgesetzlichen Entstehung des altkirchenslavischen instrumentalen -y aus *-oys sowie *-ōys skeptisch gegenüberstehe,6 da die einzige klare Quelle des altkirchenslavischen y das idg. *ū ist. Das uns hier beschäftigende -y wird seine Existenz einer Analogie verdanken. Ich lege folgende Überlegung vor: die fragliche Endung ist, als sie noch etwa *-ěs oder *-is lautete, durch das akkusativische *-ys (> -y) ersetzt worden, dessen Entwicklung im Einzelnen leider ebenfalls unklar ist, so dass man nicht auszuschließen kann, dass dieses seinerseits von *-uns > *-ūs > *-ys der u-Stämme beeinflusst worden ist. Es sind die slavischen a-Stämme zu ver-gleichen, in denen der alte Nominativ Pl. *-ās > *-as durch den Akkusativ *-ys (> -y) verdrängt worden ist (cf. Brugmann 1904: 391), dessen Entwick-lung oder analogische Entstehung abermals unklar ist, es sei denn, es geht im Grunde genommen um expandierende *-uns der u-Stämme.Schon Schmidt 1905: 3 leugnete, dass das litauische -aĩs aus *-ōys her-zuleiten ist, weil der idg. Dativ Sg. auf *-ōy > litauisches -ui. Doch blieb auch er wegen des altindischen -ais bei der Annahme, dass hier ein Langdiph-thong vorliegt und postulierte deshalb als idg. Endung -āys.7 Während aber auch für diese These das Litauische keine Anhaltspunkte bietet, sind im litauischen -ai- die idg. Kurzdiphthonge *-ay- sowie *-oy- ziemlich regelmäßig zusammengefallen (Meier-Brügger 2002: 92f.). Im thematischen Nominalparadigma ist dies beim Nominativ Pl. *-oy > -aĩ8 zu beobachten (jedoch nicht beim Lokativ Sg. -e!). Das litauische -aĩs möchte ich deswe-gen als Beleg für die Behauptung, dass die Endung des Instrumentals Pl. *-oys lautete, entschieden beanspruchen.9

6 Versuche wie der von Pedersen (erwähnt bei Brugmann 1904: 398) *-ōys > *-ūys > *-ūws > -y sind kaum mit einem anderem Beleg zu stützen. Nebenbei sei erwähnt, dass auch Ped-ersens Lösung mit *-oys hätte beginnen können: *-oys > *-uys > *-uws > *-ūs > -y.7 Danach schwankt Schwyzer 1939: 556 Bemerkung 5 zwischen -ōys und -āys. Vgl. Hirt 1927: 62. Szemerényi 1990: 197 nimmt dies (gegenüber Brugmann 1904: 397, vgl. Bemerkung 9 unten) nicht wahr.8 Das thematische Nominativ Pl. als Argument in der Sache des Instrumental Pl. führt auch Hirt 1927: 63 an, obwohl ich seine Folgerung (im Litauischen sei der Instrumental Pl. durch das Hinzutreten eines s zum Nominativ Pl. gebildet) nicht akzeptieren kann.9 Brugmann 1904: 397f. erwähnt Schmidts Rekonstruktion von *-āys nicht, er wird aber von ihr zu der Erkenntnis geführt worden sein, dass im litauischen -aĩs der Langdiphthong ōy nicht stecken kann, wohl aber, wie ich behaupte, der Kurzdiphthong oy, da er a.a.O. die Verkürzung -ōys > -oys nicht nur im Griechischen (und Italischen, siehe weiter), sondern auch im Baltischen annahm.


Die lateinische Dativ-Ablativ-Endung Pl. -īs, die auch instrumental und lokativisch verwendet wird, kann man lautgesetzlich unmöglich aus der traditionell rekonstruierten idg. thematischen Dativ-Ablativ-Endung *-o-bhos herleiten. Im lokativischen *-oysu fand vielleicht eine Apokope statt (Sommer 1914: 350.151; Meiser 1998: 72.74.129), vielleicht wurde es aber nur von der Instrumentalform verdrängt. So würde das lateinische -īs im Wesentlichen oder sogar ausschließlich eine Fortsetzung des idg. Instrumentals darstellen. Es beruht auf -eys, dessen Vorstufe -oys ist (noch greifbar in ab oloes und , Meiser 1998: 4.134) und als uritalisch gelten darf.10 Dass sie weiter aus dem vermeintlichen idg. *-ōys verkürzt ist,11 kann man mit Sicherheit nicht ausschließen, stellt aber eine überflüs-sige Annahme dar, und zwar eine eher unwahrscheinliche angesichts der Tatsache, dass das -ōy des Dativs Sg. noch im Lateinischen des sechsten Jahrhundert erhalten ist.12

Das griech. -οις und meiner Meinung nach auch das litauische -aĩs zeu-gen also zugunsten des idg. *-oys, das avestische -āiš vielleicht zugunsten *-ōys. Die altkirchenslavische und italische Evidenz ist nicht eindeutig, lässt sich aber besonders im Fall der italischen unmittelbarer auf -oys zurück-zuführen. Die avestische Endung kann somit als eine Neuerung gelten (wohl nach dem Dativ Sg. -āi). Ob sie vielleicht schon urarischen Alters ist, hängt davon ab, wie man das altindische -ais, den bisherigen Hauptzeugen in der Sache, beurteilt.

Der noch heute (z.B. Meier-Brügger 2002: 95) allgemein akzeptierten These, dass die altindischen Kurzdiphthonge… , aus den älteren /āy/, /āw/ hervorgangen sind, kann man schon jetzt einige Fälle entgegenhalten, in denen diese Diphthonge nie lang waren: naus „Schiff“, das als *neh2-u-s zu rekonstruieren ist,13 asáu < *-séh2-u „jene“14 und mindestens teilweise

10 Sommer 1914: 350 mit Belegen. Es ist mir jedoch nicht klar, warum das oskische bisher nie als eventuelles /nowlanōys/ interpretiert worden ist, da die Endung des Dativs Sg. /-ōy/ in dieser Sprache als erscheint.11 Sommer 1914: 350 als Möglichkeit; Meiser 1998: 72.75.129.136; Brugmann 1904: 397; Sihler 1995: 263.12 Das ergibt sich aus der Kombination der Fakten, dass man die Form in der Duenos-Inschrift liest und dass man später Dativ -ō und nicht **-ŏ findet, es sei denn, dieses dativische -ō ist etwa aus dem Ablativ Sg. übertragen worden oder mit dem Ablativ Sg. identisch, was bisher natürlich niemand behaupten wollte.13 Kuryłowicz 1935: 38. Das Paradigma war amfikinetisch; in den schwachen Kasus ist im Altindischen wahrscheinlich eine Durchführung der Vollstufe der Wurzel anzuneh-men (nāvás < *neh2wés

Ján Bakyta16

auch augmentierte Verben mit synchron vokalischen Wurzelanlaut wie aicchat < *(h1)é-h2is-sḱe/o- (zum Präsens icchati).15 In all diesen Fällen fiel der intervokalische Laryngal aus und es fand eine Kontraktion statt, die Monophthongierung *ay > ē, *aw > ō aber unterblieb. Kurylowicz 1935: 38 hat das damit begründet, dass die letztgenannte Monophthongierung schon erfolgt war, als die Kontraktion begann, ebenso gut kann man aber zwei andere Erklärungen in Betracht ziehen, die er für das Unterbleiben bereits der Kontraktion in rayis < *raïs < *reh1is erwog: „lé désir de conserver la transparence morphologique de la forme“ und „le caractère dissyllabique de la forme“, also das Bemühen, die Zweisilbigkeit zu bewahren (Kurylowicz 1935: 37 Bemerkung 1).

Es ist also zu vermuten, dass die altindische Endung -ais auf dem älteren *-aïs und dieses wieder auf *-aHis < idg. *-o-His beruht und dass der Laryngal und nach Ausfall desselben der Hiatus wegen der morpholo-gischen Transparenz mindestens so lange erhalten blieben, bis die Diph-thongierung ay > ē beendet wurde.16 Zu bemerken ist noch, dass der Instru-mental Pl. des Pronomens tas im Fall der Monophthongierung *tays > **tes zu früh einsilbig geworden wäre.

15 Kurylowicz 1935: 38. Cf. Thumb – Hauschild 1959: 191f., Wackernagel 1978: 53 (beide natürlich ohne Laryngal). Ittzés 2005: 214ff. möchte neuerdings einige Fälle anders erklären (Musterbeispiel: *é-h1ews-e-t > *á-awš-a-t > *āwšat > áuṣat zu Präsens óṣati), so dass hier die Diphthonge au, ai wirklich ursprünglich lang gewesen wären. Diese These hat zweierlei zur Voraussetzung: 1. Dass sich die nach dem akzentuierten Augment tieftonige Verbal-wurzel in der Vollstufe bzw. Dehnstufe befand (diese müsste dann analogisch eingeführt worden sein, was aber gut denkbar ist). 2. Dass man eine urindogermanische Form (vgl. Ittzés 2005: 218) überliefert und zugleich in ihr mindestens den Ausfall der Laryngale, die Kontraktion und die Diphthongverkürzung ohne weiteres vorgenommen hat. In Wirkli-chkeit werden die augmentierten Formen in der Mehrheit der Fälle von neuem komponiert. Es ist natürlich fraglich, ob man sich dabei der Nullstufe oder der Vollstufe der Wurzel bedi-ent hat. Sollte hier und da der Augment mit der Vollstufe zusammengesetzt worden sein, ist ein Lautwandel ă-ē > /ăy/ bzw. ă-ō /ăw/ denkbar. Ittzés’ These ist also anreg-end, aber uns stehen keine Mittel zur Verfügung, sie für dieses und jenes vedische Wort zu verifizieren. Da aber auch Ittzés selbst nicht leugnen will, dass die Anlautdiphthonge der augmentierten von den Verben auf Vokal abgeleiteten Formen in manchen Fällen nie lang waren, ist diese Frage für uns hier im Folgenden belanglos.16 Ich bin nicht imstande, eine metrische Analyse des vedischen Textes durchzuführen. Es ist aber wahrscheinlich, dass die Zweisilbigkeit der Endung hier nicht zu spüren ist, sonst hätte man dies schon früher gemerkt. Doch halte ich diese Tatsache für keinen zwingenden Gegenbeweis der hier vorgelegten These. Jedem steht natürlich eine andere Beurteilung der Schlagkraft dieses argumentum ex silentio (die ich ja im Allgemeinen nicht verwerfe) offen.


Diese Laryngal kann nicht nur nach der Analogie von naus < *neh2us, sondern auch aufgrund struktureller Erwägungen erschlossen werden.17

Die Diphthonglänge in dem bisher rekonstruierten idg. *-ōys vermochte man aus dem angeblichen **-o-Vys herzuleiten, nicht aber zu erklären.18 Höchstens bemerkte man, dass diese Endung wie eine Pluralisierung (durch Hinzutreten von s) von -ōy < *-o-Vy, also von Dativ Sg., aussieht,19 was vielleicht nicht mit Sicherheit auszuschließen ist, was aber aufgrund des semantischen Unterschieds zwischen Dativ und Instrumental recht son-derbar wäre.

Das Richtige aber kann man einem Blick auf den thematischen Instrumen-tal Sg. entnehmen. Hier ist zunächst ein Langvokal, *ō oder *ē, zu spüren, den man weiter als *-o-o bzw. *-e-e oder aber als *-o-h1 bzw. *-e-h1 analy-sieren kann.20 Die laryngalistische Indogermanistik wird sich zweifellos für die zweite Möglichkeit entscheiden.

Dann können wir die Urform des griech. -οις und des altindisch. -aih auf *-oh1is präzisieren und als Themavokal o + das Instrumentalzeichen h1 + die Partikel i + das Pluralzeichen s21 analysieren.

17 Erst nachdem ich zur Rekonstruktion *-oh1is gelangt war, kam mir zur Kenntnis, dass Beekes 1995: 195 in der Endung des Instrumentals Pl. der thematischen Deklination *h1 rekonstruiert hat. Beekes hat aber diesen Laryngal nicht zu begründen versucht, er hat ihn auch im Ablativ, Dativ Sg. und im Nominativ Pl. angesetzt, also überall da, wo es zu einer Kontraktion des Themavokals mit vokalischen Anlaut der Endung gekommen ist, was meiner Meinung nach unbeweisbar und mindestens im Dativ Sg. unberechtigt ist, und schließlich und vor allem, seine Rekonstruktion von *-oh1eis läuft wieder einmal auf ein Langdiphthong hinaus. Was übrigens seine Angaben „-ōd < -oh1ed < -o-ed“ usw. Bedeu-ten sollen, ist mir rätselhaft, es sei denn, der Laryngal hätte den Hiatus abgelöst, was ein sonderbarer Gedanke wäre, oder es handelt sich um ein Versehen (statt des zu erwartenden *-ōd < *-o-ed < *-oh1ed usw.).18 Thumb – Hauschild 1959: 36: „über die Herkunft des idg. Suffixes ist keine Sicher-heit zu erlangen“; Szemerényi 1990: 197: „ganz isolierte Endung“; Beekes 1995: 192: „This ending has no parallel in the consonant-stems. Perhaps it has been taken over from the pronouns.“; Sihler 1995: 263: „whose internal structure is enigmatic“.19 Debrunner – Wackernagel 1975: 66f. (mit Literatur).20 Siehe vor allem Szemerényi 1990: 195, der allerdings den Laryngal nur sehr vorsichtig anführt. Dass es sich um den ersten Laryngal handelt, ergibt sich daraus, dass im *-e-h1 die Vokalqualität erhalten blieb. Theoretisch ist vielleicht *-o-oh1 und *-e-eh1 möglich und dann würde hinter dem altindisch. Instrumental Pl. -ais doch ein Langdiphthong stecken. Eine solche Annahme jedoch erübrigt sich völlig.21 Ob dieses s ursprünglich ein Pluralzeichen war, ist fraglich. Ein Blick in Hirt 1927: 140-149 (besonders 147f.) ist lehrreich. Es wird jedoch recht allgemein angenommen, dass dies (vor allem im Akkusativ Pl. *-ns) so aufzufassen ist, siehe Meier-Brügger 2002: 164, Szemerényi 1990: 173, Thumb – Hauschild 1959: 16 usw.

Ján Bakyta18

Damit ist unsere Aufgabe, die Rekonstruktion des thematischen Instru-mentals Pl., im Wesentlichen erfüllt. Doch verdient noch die Partikel *i unsere Aufmerksamkeit. Sie ist offenbar auch in den athematischen Endun-gen des Instrumentals Pl. -bhis (Altindisch), -mìs (Litauisch), -mi (Altkirch-enslavisch), -φι (Griechisch) sowie Sg. -mь (Altkirchenslavisch) enthalten.22 Ich aber halte für durchaus wahrscheinlich, dass die vedische thematische Instrumentalendung -ebhis, die dem -ais konkurriert und ein „Prakrit-ismus“, eine Neuerung darstellt (Debrunner – Wackernagel 1975: 105ff.), nicht nur die Richtigkeit unserer Behauptung beweist, dass in -ais kein Langdiphthong vorliegt, sondern auch wahrscheinlich macht, dass die Partikel i einst mindestens fakultativ auch zum Instrumental Sg. hin-zutreten konnte: *taHi-bhi-s oder *taï-bhi-s / *tay-bhi-s wurde anstatt *taHis / *taïs / *tays (zum Pronominalstamm ta-) neu gebildet, um die mit dem Laryngal zu verblassen drohende Instrumentalität klarer zum Ausdruck zu bringen und/oder die Zweisilbigkeit zu bewahren.23 Danach bestand aber kein Grund, die Monophthongierung *taybhis > tebhis nicht durchzuführen. Aus den Pronominalformen, die in Rigveda ausschließlich auf -ebhis enden (cf. Debrunner – Wackernagel 1975: 502), begann diese Endung in die Nominalflexion durchzudringen, um während der nach-rigvedischen Zeit in der vedischen und sanskritischen schriftlichen Über-lieferung immer mehr zurückzutreten (nicht jedoch in den Vorstufen der Prakritsprachen). Die gängige Auffassung vermag das e in -ebhis mit dem vermeintlichen Langdiphthong ai in -ais nicht zu verknüpfen und seine Herkunft bleibt deshalb im Unklaren.24

Im Instrumental Sg. -ena (vedisch und wahrscheinlich ursprünglicher -enā) kann das e einfach aus dem Instrumental Pl. -ebhis bezogen wer-den, es ist aber möglich, dass auch dies eine schon ältere Neubildung um der Transparenz willen ist. Dann ist an eine Urform *taHi-n-aH / *taï-n-ā / *tay-n-ā (-n- wahrscheinlich aus den n-Stämmen, wie in agni-n-ā, śatru-ṇ-ā, cf. Thumb – Hauschild 1959: 54, Debrunner – Wackernagel 1975: 148) zu denken, in der an den alten Instrumental *ta-(H)-i mit der

22 Altkirchenslavisches -mь < *-mi und -mi < *-mī(s), das vielleicht auf *-mih1(s) beruht, so dass hier Instrumental zweifach bezeichnet wurde, mit *m und *h1. Die Agglutination in den Nominalendungen war also offensichtlich ziemlich ungeregelt.23 Es ist möglich, wenn auch nicht unabweislich, wegen des altpersichen kamnaibiš martiyaibiš und des avestischen āfrivanaēibiš (dies jedoch „recht unsicher“, Debrunner – Wackernagel 1975: 107) an eine urarische Neuerung zu denken. Dann wäre die Erhaltung des Laryngal in der Zeit der Neubildung gesichert und als die Ursache der Neubildung wäre eine drohende Einsilbigkeit ausgeschlossen.24 Thumb – Hauschild 1959: 141: „Es ist zweifelhaft, ob das e aus dem Nom. Plur. m. stammt.“


noch erhaltenen Partikel i der Verbindungskonsonant n und abermals das Instrumentalzeichen *ā oder *aH (< *idg. eh1) angehängt wurde. Und da das e in den thematischen Lokativen Sg. und Pl. völlig berechtigt ist (-e < *-o-i, -eṣu < *-o-i-s-u / *-o-i-s-su), so müssten wir weiterhin lediglich im Fall des Dativ-Ablativ Pl. -ebhyas (und des Genitiv Pl. -eṣām in der Pronomi-nalflexion) zur Annahme der Übertragung des e greifen, die wahrscheinlich vom konsonantischen Anlaut der Endung (-bhyas, -ṣām) gefördert wurde, der ähnlich in devena, deveṣu und devebhis zu beobachten war.25

Es verbleiben uns die lautlichen Folgerungen. Für das Altindische gilt aufzuklären, inwieweit es richtig ist zu behaupten, dass die altindischen , einst Langdiphthonge waren. Wir haben gezeigt, dass in einem wichtigen Fall, in dem überliefert ist und wo deshalb einstimmig ein ehemaliges */āy/ rekonstruiert wurde, dem nicht so ist. Ich bin geneigt anzunehmen, dass die wenigen idg. Kontraktionslangdiphthonge schon vor Anfang der vedischen Überlieferung gekürzt wurden und dass sich über die altindischen , sagen lässt, dass sie nicht monophthongiert wurden, keineswegs aber, dass sie ursprüngliche Langdiphthonge widerspiegeln. Die Beweise, die Wackernagel 1978: 41 für das Letztgenannte anführt, sind teilweise überholt und teilweise nicht zwingend.26 Wir werden auf diese Problematik vielleicht ein anderes Mal zurückkommen, es wäre aber nüt-zlich, dass sich auch wirkliche Sanskritkenner damit beschäftigen.

Für das Altgriechische ist ein isolierter und deshalb wichtiger Fall der Wirksamkeit des Osthoff’schen Gesetzes erledigt, mit der es meiner Mei-nung nach auch sonst schlecht steht. Letzten Endes zeigen uns das griech. -οις und das altindisch. -ais, die beide auf dem idg. *-oh1is beruhen, dass auf den Vokalismus der griechischen Dialekte mitunter mehr Verlass ist als auf den Vokalismus des Altindischen auch dort, wo man ihn bisher nicht angenommen hat.

25 Die Partikel i halte ich für ein deiktisches Element, das als solches vor allem zu einem Pronomen gehört haben kann. Doch ist recht offensichtlich, dass es auch zur Kasusbil-dung des Nomens unabhängig von jener des Pronomens verwendet wurde. Das i im alti-nidischen nominalen -e des thematischen Lokativs Sg. (der ja früher wahrscheinlich end-ungslos gewesen war) und folglich auch des Lokativs Pl. (*-o-i-s-u / *-o-i-s-su, letzteres nach Meier-Brügger 2002: 200) wird im Grunde genommen mit dieser deiktischen Partikel identisch sein. Und dieses ist nicht vom Demonstrativpronomen, dessen Endung -smin ist, und kaum von den Personalpronomina, obwohl hier die Endung -(y)i lautet, bezeugt worden. Es ist also problematisch, das anstatt des bloßen Themavokals stehende altindische nominale -e- mit dem Verweis auf einen Einfluss der Pronominalflexion erledigen zu wol-len, umso mehr, als man auch das pronominale -e- eigentlich nicht zu motivieren vermag (wie z . B. im tebhis).26 Zu einem von ihnen vergleiche, was oben in Bemerkung 16 gesagt wurde.

Ján Bakyta20

LiteraturBartoněk, Antonín. (2003). Handbuch des mykenischen Griechisch. Hei-delberg: Universitätsverlag C. Winter.

Beekes, Robert S. P. (1995). Comparative Indo-European Linguistics: an Introduction. Amsterdam – Philadelphia: John Benjamins.Brugmann, Karl. (1904). Kurze vergleichende Grammatik der indoger-manischen Sprachen. Strassburg: Karl J. Trübner.Debrunner, Albert – Wackernagel, Jacob. (1975). Altindische Grammatik III. Nominalflexion – Zahlwort – Pronomen. Göttingen: Vandenhoeck & Ruprecht. [Unveränderter Nachdruck der 1. Auflage von 1929/1930].

Hirt, Hermann. (1927). Indogermanische Grammatik III: Das Nomen. Heidelberg: Carl Winter’s Universitätsbuchandlung.

Hoffmann, Karl – Forssman, Bernhard. (20042). Avestische Laut- und Flexionslehre. Innsbruck: Institut für Sprachen und Literaturen der Uni-versität Innsbruck.

Ittzés, Máté. (2005). „Problems of the Augment in Vedic“. Acta antiqua Academiae scentiarum Hungaricae 45/2-3, 207-223.Kretschmer, Paul (1909). „Zur Geschichte der griechischen Dialekte“. Glotta 1, 9-59.Kuryłowicz, Jerzy. (1935). Études indoeuropéennes I. Kraków: Gebeth-ner i Wolff.

Meier-Brügger, Michael. (20028). Indogermanische Sprachwissenschaft. Berlin – New York: Walter de Gruyter.

Meiser, Gerhard. (1998). Historische Laut- und Formenlehre der latein-ischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft.Osthoff, Hermann. (1879). „Kleine beiträge zur declinationslehre der indog-ermanischen Sprachen II“. In: Osthoff, Hermann – Brugman, Karl. (1879). Morphologische Untersuchungen auf dem Gebiete der indogerma-nischen Sprachen. Leipzig: S. Hirzel, 1-147.Osthoff, Hermann. (1881). Rez. „Gustav Meyer, Griechische Grammatik“. Philologische Rundschau 1, 1588-1597.Peters, Martin. (1980). Untersuchungen zur Vertretung der indogerma-nischen Laryngale im Griechischen. Wien: Verlag der Österreischischen Akademie der Wissenschaften.


Rix, Helmut. (1976). Historische Grammatik des Griechischen: Laut- und Formenlehre. Darmstadt: Wissenschaftliche Buchgesellschaft.Schmidt, Johannes. (1905). „Zur geschichte der langdiphthonge im Grie-chischen“. Zeitschrift für vergleichende Sprachforschung auf dem Gebi-ete der indogermanischen Sprachen 38, 1-52.Schwyzer, Eduard. (1939). Griechische Grammatik I. Allgemeiner Teil, Lautlehre, Wortbildung, Flexion [2]. München: C. H. Beck’sche Verlags-buchhandlung.

Sihler, Andrew L. (1995). New Comparative Grammar of Greek and Latin. New York – Oxford: Oxford University Press.Sommer, Ferdinand. (19142-3). Handbuch der lateinischen Laut- und For-menlehre. Heidelberg: Carl Winters Universitätsbuchhandlung.Szemerényi, Oswald. (19904). Einführung in die vergleichende Sprach-wissenschaft. Darmstadt: Wissenschaftliche Buchgesellschaft.Thumb, Albert – Hauschild, Richard (1958). Handbuch des Sanskrit I. Grammatik 1. Einleitung und Lautlehre. Heidelberg: Carl Winter Uni-versitätsverlag.

Thumb, Albert – Hauschild, Richard (1959). Handbuch des Sanskrit [I. Grammatik] II. Formenlehre. Carl Winter Universitätsverlag.Wackernagel, Jacob. (1978). Altindische Grammatik I. Lautlehre. Göt-tingen: Vandenhoeck & Ruprecht. [2., unveränderter Nachdruck der 1896 erschienenen ersten Auflage].

Ján Bakyta Institute of Greek and Latin Studies, Faculty of Philosophy & Arts, Charles University in Prague [email protected]

Initial *x- in Slavic revisited

Jan Bičovský

In this article, I aim to show how Slavic x- could have developed from PIE or later *sk- without the need to invoke the operation of the ruki rule. First, the whole process of ruki in Slavic must be divided into three separate and causally unrelated changes, each being motivated by a different set of factors (i.e. opti-misation of perceptory contrasts) within the synchronic phonological systems. Next, the change of *sk to *š is explained both in phonetic terms and by analogy with comparable changes in Germanic languages. This change predates the fol-lowing shift of *š to *ʂ and finally to *x. As to the reasons of Sl. *sk being often the reflex of non-Sl. IE g- or k-, the process of s-mobile is given some considera-tion. In the second part of the article, a few dozen of the x-initial lexemes in PSl. are examined in the light of this theses, adding a couple of new explana-tions to old conundrums.

The ruki ruleIt is a long-standing view of Indo-European linguistics that the voice-less velar fricative x in inherited Slavic vocabulary results from the retraction of PIE *s after *i, *u, *r, *K and *K .̫ The process was first brought to the attention of the comparative linguistics by the Danish linguist Holger Pedersen (1895) and in the Indo-European studies is commonly referred to as Pedersen’s Law, or simply the ruki rule, after the main factors of the change. I have shown elsewhere1 that this process, intricate though it appears on the surface, can be explained with some confidence if the rule is properly divided into discrete phases, of which only the first phase, the ruki-rule proper, results from automatic retraction of the PIE apical *s after all non-anterior closed (to include high vowel) oral phonemes in post-PIE.2 It follows that PIE *r was, at least in syllabic coda, retroflex or retracted

1 Bičovský 2006. But since the original is hardly available to the public, I dare include a short recapitulation of the argument.2 Since I believe laryngeals to be a glottal, velar and postvelar/uvular (possibly rounded) fricative, and therefore also closed non-anterior phonemes, the ruki rule must have applied only after these were merged and possibly debucalised to ɦ/h in syllabic coda, or assimi-lated as a glide gesture to the preceding vowel.

Jan Bičovský24

and velarised. The subsequent changes to historical š or x are not causally related.

The feature of apicality of the PIE *s is crucial for understanding the retrac-tion (or posteriorisation) on the phonetic and articulatory basis. The fact that PIE had an apical *s (as in Castilian Spanish) rather than a coronal/laminal *s (as in German, French, Italian and most Slavic languages) has already been pointed out by several scholars, first probably by André Mar-tinet in 1955. There are persuasive arguments from many IE (proto-)lan-guages for assuming that the original articulation was apical. The appear-ance of rhotacism (*s > r) in some positions in Latin (honos ~ honoris), (Northwest) Germanic (OE cēosan ‘choose” ~ (ge)coren ‘chosen”) and Indic (such as sandhi –r for -ṣ before voiced consonants, e.g. agnis+dahati > agnirdahati ‘(the) fire is burning”) points to a type of articulation involv-ing the apex of the tongue. The resulting rhotics are as a rule apical vibrants or approximants. Likewise, synchronic evidence from languages such as Spanish shows the process in real life. In certain areas (Extremadura), final apical -s is pronounced as -r before voiced consonants, e.g. es dulce ‘s/he is sweet” [erdulθe]. Next there is an argument from orthography: the spell-ing of Hittite /s/ with the set of Babylonian syllabograms for šV rather than sV is strongly reminiscent of the situation in the Arabic realms of medieval Spain, where Spanish apical /s/ was likewise transcribed with an in the Arabic script, showing that the sound was closer in perception to the Arabic /š/ than to its laminal /s/. And, last but not least, linguistic typology shows that in most languages employing a single sibilant in their phonological system, this sibilant is, as a rule, apical (take e.g. Finnish).

The ruki rule has often been invoked as further evidence for the real-ity of the satem dialect area, since both Balto-Slavic and Indo-Iranian displayed results of this process, while in Albanian and Armenian, the two notoriously difficult languages from the point of historical phonology, a few disputable examples could be adduced. Be that as it may, it appears that only the first phase of the ruki rule, *s > š can be observed in most satem languages, the other two phases, which even-tually lead to *x in Slavic, and also to /kʰ/ = [x]? in some Prakrits, such as Mādhyaniṇa (Cardona 2003: 109), are parts of very different processes. Ruki proper, that is the retraction of PIE *s to (originally purely allophonic) *š, was not motivated by the satem change of * to *ç (which would confirm the validity of the satem isogloss), nor did satemisation secure the phonological status of these allophones - it merely provoked a further retraction of *š to *ʂ, still a positional vari-

Initial *x- in Slavic revisited 25

ant of *s, and that only in some dialects, while in others some or all these sibilants merged.

The retraction from *š to *ʂ result from typologically well-documented and phonetically well-understood strategies for enhancing perceptory distinc-tion between palatal and apical sibilants (e.g. Marzena Żygis 2006)3 and have no causal relation to the said retraction. Though typological parallels for such a change are far from abundant - at least Modern Swedish and Old Spanish show the change ʂ > x, the latter was in fact adduced to already by Pedersen himself. Assuming PIE *s was apical, it follows that at some point all satem languages must have arrived at a sibilant system with both palatal and apical phonemes. From this it would also follow that the intermediate realisation between *š and *x was not a palatal *ç (which would have with all probability merged with the original *ç < PIE *) but an alveo-velar fricative [ɧ] (contemporary Swedish provoked the inclusion of this special IPA sign). The fact that the ruki rule is best attested in Slavic and Indo-Iranian, and that its effects were more regular in these two languages, is probably also due to the fact that apart from satemisation itself, these two underwent further palatalisation of velars, resulting in even more sibilant series than in the case of Baltic, and of course, Albanian and Armenian.

Thus it can be shown that neither the retraction to *ʂ nor to *x were moti-vated by the phonemes directly involved in he first phase of ruki. The crucial point for the reconstruction of PSl. *x- is precisely that the second phase, *š > *ʂ, does not depend on the factors (r, u, k, i) of the first one, nor does the third phase depend on those of the second one. And, most impor-tantly, neither of the latter two phases must rely for their input only on the results of the ruki rule proper, but would naturally apply to any seg-ment that meets the criteria for the shift regardless of its origin. The ruki rule, by definition, is a progressive assimilation. This would seem to imply that for any Slavic reflex of this rule, for any original *x, there must have existed a segment to its ‘left” that triggered the change. There is however a group of words in Slavic, well attested and mostly agreed to be of at least Balto-Slavic antiquity, which cannot meet the criteria – words with initial x-, such as *xoldъ ‘cold’, *xorbrъ ‘brave, valiant’, *xudъ ‘poor, weak’, *xormъ ‘house, temple, sacred precinct’ and a few dozen more.

The difficulty in reconstructing the original structure of these words is obvious. Apart from the verb *xoditi ‘to walk’, whose *x- may be (an may not but usually is) explained away as the result of sandhi assimilation with 3 It is, of course, possible that other phenonema played their parts in this development, but the one described by Żygis seems to be able to explain the process by itself.

Jan Bičovský26

the numerous preverbs ending in ruki factors, such as *proti, *per, *h2eu, etc., there is no good explanation for an original *s as the input for ruki changing to *x in this position, and I believe there are good reasons to refute this hypothesis for *sod > *xod as well (see bellow). Also, this may presuppose for the PBSl. language to have undergone univerbisation of these preverbs with verbs already at a very early stage in its development when ruki retraction was still applicable, and that is dubious.4 Marko Snoj 2003: 210 proposes the same explanation for *xorna ‘fodder, care” and *xor-niti, connecting this word to the PIE root *serh2 ‘take care’.

Initial x- in SlavicThe question why the factors of initial ruki, once they affected *s, would disappear5 in initial position while elsewhere they were preserved and developed in the manner expected, is still a mystery. Initial clusters *is, *us, *rs (likely *Hrs in PIE) would have yielded Slavic *jьx *vъx *ъrx, and therefore no initial *x- at all6. There only remains one possible explana-tion of the initial ruki *x, in PIE groups *Ks and *K̫ s, possibly surfacing universally as *ks regardless of the articulation of the velar (and its features such as voicing and aspiration). This would have changed to *kš and later either to *kx or more likely to *šš, then becoming *x, depending on the relative chronology of the changes in PSl. syllabic structure. Indeed, for a small group of words, etymologies with initial PIE *ks could be found, especially in Indo-Iranian and Greek, and there is nothing to disqualify the existence of such initial clusters in PIE, rare though they must have been, nor the possibility that they survived in Slavic.

4 In Baltic, reflexive pronouns are appended between the preverb and verb root, which is clearly a phenomenon comparable to e.g. Old Irish and which shows that some independ-ence of the preverb is to be expected for earlier stages. However, no traces of this can be found in Slavic, so the era of univerbisation is not likely to predate the split of Baltic and Slavic, while the first phase of ruki only operated much earlier than that. 5 Of course, in a strictly mechanistic manner, one could propose a *HsV, with a ‘disap-pearing laryngeal” to save the day, and include laryngeals in the ruki set for Slavic, with a possible dissimilatory change of *[xs] to *[ks] and [kʃ]. The few instances of possible PIE *HV > PSl. *kV, such as *koza ‘goat” and *kostь ‘bone” for *h3Vǵeh2 and *h3ostis are them-selves still a matter of debate. Such an explanation could not be ruled out on purely histor-ical-phonological grounds, but would require the reconstruction of a number of PIE roots unattested elsewhere in IE languages, and thus throw the IE status of these roots, along with the laryngeal, overboard. 6 Terence R. Carlton clearly did not perceive this problem - ‘However, the number of Indo-European roots with structures of the type k or r + s + vowel is much too meagre to explain all the instances of initial x in Slavic.” (Carlton 1991:96)


It would be a mistake of course to revert this logic completely and deny that any initial *x- could have been caused by the ruki-rule. It is tempting, for example, to link Skr. kṣudá ‘hungry’ to Slavic *xudъ ‘poor’, although I believe that there is at least one other possibility, and there are a couple more words that could find an explanation here, but by and large, most instances of Slavic x- do not find any such parallel, no matter how hard one twists and stretches their semantics and morphology.

Several possibilities for the origin of Slavic *x- have been proposed. Some relied on processes other than regular phonological change, such as affective change of whichever original segment appeared to match etymological cor-respondences best – a strategy followed by e.g. Machek: 1971, but the older the etymon seems, the more difficult it is to envisage a good argument from semantics for affective change. Moreover, providing affective change as an explanation where no other linguistic explanation can be found is always an easy target for criticism and it would have seemed more prudent to admit failure and leave the question unanswered for the time being. There have even been serious attempts to understand *x- as a result of a prefix or a morpheme of its own, the zero-grade *ks- of a root *kes- ‘to cut’ and the like, but needless to say – neither were they found very persuasive.

Recently, a difficult quasi-PIE affricate *ks has been proposed by Bańkowski 2000 to account for some instances of initial *x- (even some of the best etymologies we seem to have, such as *xorbrъ = PGerm. *skarpaz), while doing away with others by assuming an irregular change of initial *us- to *ъx- (rather than the expected *vъx- with v-prothesis) and then aphaeresis to *x-. The prefix *us, otherwise unattested in Slavic (there is only the s-less variant *vy- from *ūd-) goes back to PIE *uds, and I am in doubts as to whether the outcome of a PIE group [ts] would qualify for ruki retraction at all, since it is more likely that it would have been simplified to coronal s long after ruki ceased operating. Many scholars attempted to find a non-ruki source for Slavic *x in any consonant cluster that would impose the vague impression that it could have somehow developed into a velar fricative, such as *H-, *sgʰ- etc. These developments often contradict the overall tendencies in the development of PBSl. or PSl. consonantism, either requiring that laryngeals survived well into the PSl. period or denying any unity of development for Baltic and Slavic. This strategy, though yielding no persuasive results, at least acknowledges the fact that the ruki-rule by itself cannot account for every *x- found in the Proto-Slavic lexicon.

Jan Bičovský28

And of course, a combination of all these strategies would have been expected to arise sooner or later, picking and choosing affective change for this word and one of the phonological changes listed for another one and so on – Shevelov 1964: 136 gathered an early selection of these. ‘Thus the sources of the initial x- in Sl are:

1. IE k -ʿ affective2. IE ks- not necessarily affective3. IE s- in roots that may have pre-fixes ending in r, u, k, i

not necessarily affective, often with semantic differentiation

4. IE s- in prefixless words affective5. IE k- affective6. IE sk- affective7. x, h in loan-words not necessarily affective’

There have also been attempts to explain initial x- in Slavic as a result of an Iranian adstratum and borrowing. Attempts have been made to explain even the very phonetic process of the ruki-rule as being of Iranian ori-gin (Sussex 2006: 24 among others). Since it is universal in Indo-Iranian, the ruki rule must have been at least of Proto-Indo-Iranian antiquity (the unity of these dialects is variously dated between 2000 and 2500 BCE), and failed to apply at some stage to the newly developed Iranian *s < *, in the same manner as it did not apply to PSL *s < PIE *. At such a deep time level, rule borrowing among closely related languages is possible, but so is an independent (and similar) development of a common heritage. Also, since Iranian š does not became x, one should look for etymologi-cal relatives with š- as well as x- to examine all the possible sources. Ini-tial x- in the few Iranian languages known from the 1st millennium BCE results from older *sV- *s-, *sj-, which pertain to a much younger stage of development.

Etymological correspondencesEver since etymological correspondences were first taken into account, it appeared that Slavic *x- mostly seems to reflect sk-, k- or g- in other IE languages, with a greater number of cognates in Baltic and Germanic – something to be expected given the large percentage of shared vocabu-lary among these three branches. There are a number of almost certain correspondences, some pointing to a single velar, such as *xoldъ – Eng. cold – lat. gelidus etc., others to *sK-, such as *xorbrъ – MnE. sharp etc.


The sk- reflexes are especially abundant and many examples can be found for a sk- and K variation of the same root, possibly due to the so called s-mobile (see below), effective long after the split of PIE, and possibly well into the PSl. era.

The sk- group has already been employed as a ruki-like explanation of x-. It is, albeit superficially, plausible that these initial consonant groups were either as a rule, or probably only occasionally or in some dialect, metath-esised to *ks very early, and this metathesis fed the ruki rule while still operating. However, such metatheses are rare cross-linguistically and sus-pect in a language that generally tended to simplify and regularise conso-nant clusters along the sonority principle.

The assumption that the ruki rule in Slavic has three phases has some rather promising implications for the reconstruction of initial *x-. In fact, what we are looking for is not necessarily the source of Proto-Slavic *x – and if plausible explanation has been found during the last 120 years, it may also be due to there being none. What we are looking for is simply any phoneme that would finally turn up as x- in Slavic – and that is not a ruki *š, but simply any *š there was to undergo the following shift to *ʂ and eventually, to *x.

The fact that Sl. x- does not regularly correspond to a single reflex in Bal-tic, Germanic, or anywhere else, is still a grave problem, even if we have to solve the comparatively easier question of a non-ruki š. If we remain faithful that there is a good and regular explanation (and intuition warns us that there may be none), it is necessary to assume that although the extra-Slavic reflexes are not uniform, Proto-Slavic had only one original sound or sound-cluster and that the reflexes in other languages somehow devel-oped that sound in either one way or the other in each and every case. It is fairly obvious that finding an explanation for such changes would be as difficult, or probably even more difficult, than finding them in Slavic alone. The five most common reflexes of *x- outside Slavic are *sk- *k/* *g/* and it would seem that to account for Slavic *š- we would need to postulate at least five rules to deal with this piecemeal. Also, we would be immediately faced with the obvious problem – we know quite well what the outcomes of these phonemes are in Slavic. Or do we?

Yes – we do know for certain that initial * and * yielded late PSl. *s and *z respectively, at least they did if no resonants followed. The situation is slightly more complex with *r, *l and * whose outcomes are not uniform, sometimes we even get two reflexes of the same PIE root, such

Jan Bičovský30

as *eit ‘to shine’ giving both *svět- and *květ- in Slavic, or *kloniti and *sloniti, both ultimately from PIE *el, enlarged to *len, also found in e.g. Eng. lean, Gr. κλίνω etc. Especially in Baltic, such examples abound (Kortlandt 1978). In both cases we probably deal with reflexes of two dialect areas, the distribution of which had long been lost – they might have, for all we know, merged, or one might have disappeared leaving only slight lexical traces – the existence of both original *kR and *kʷR ensured that the shift of *R to *kR did not result in a combination unnatural to the language.7 We also know for certain that until the first palatalisation, the neutral PBSl. velars remained unchanged. So it is possible to rule out *- or *- as direct sources for Slavic *x-. It is likely that the ancestors of Balts and Slavs lived in the same region, and since their languages display such strong affinities, they must have been in constant contact since IE times. Therefore, they originally formed a dialect continuum, in which there must have been intermediate dialect stages between the core of the Slavic homeland and the core of the Baltic region. Forms of one language could have been borrowed freely and frequently into the other, quite often no phonological adjustment would have been necessary, since even after the split their phonologies were very close, and so were their grammar and lexi-con. Later, population movement would have occurred from time to time resulting in temporary bilinguism and later in amalgamation of the two dialects (compare the results of Norse-Anglo-Saxon contacts in medieval English community). The existence of Slavic doublets (or even triplets) of initial x- K- and sk- is to be expected under such circumstances.8 But there still remains one fact that we do not know for certain: what are indeed the regular reflexes of PIE *sk- groups in Early-Proto-Slavic? The data is not very impressive - examples are surprisingly scarce in compari-son with, on the one hand, Slavic sp- and st- and, on the other, with sk- in Baltic and Germanic, the two closest IE branches. Mike Southern inter-

7 It should be noted that in the course of its development from PIE to Late-Proto-Slavic, the language traversed several millenia and it seems debatable that for all this long time it remained a single language with little dialectal differentiation. As is the case with any other language, several stages of dialectal disintegration and later congruence would have taken place, leaving traces both in the grammar and the lexicon. We should not shirk from acknowledging that some of the variation is rather due to dialectal congruence then to anything else – variation should be expected in any language and it is uniformity which should raise suspicion. 8 As Jost Gippert reminded me, the same phenomenon exists in Germanic, e.g., Germ. schmelzen and Engl. melt, which, as he adds, ‘should be regarded as part of the s-mobile problem, which need not be explained on the basis of interdialectal borrowing or the like but which could yield doublets off-hand even within one given language.”


preted the overall scarcity9 of *sC- reflexes of s-mobile in Slavic as result-ing from the loss of s-, i.e. to a reversal of the s-mobile process (though for him it is still part of the process, working both ways on the initial con-sonant (groups)). In fact, there are very few good examples of Sl. sk- or šč- that would unequivocally point to an original PIE *sk-, s-mobile or not, thus attested in some other IE branch. It is true that the material is not easy to interpret in the first place: many such groups in PSl. resulted from earlier combinations of the preverb *sъ (< PIE *(k)sun) + *k- and were quite often written with –ъ- even in CSL, so that e.g. *skokъ ‘leap’ appears either as skokъ or sъkokъ in CSL. The fact that this almost never happens with original st- or sp- is suspect. A likely explanation would be that the lexical integrity of these roots never allowed multiple analyses, whereas *sk- roots were, for some reason, more ambiguous in this respect. Both the relative scarcity of sk- and their greater ambiguity support the conclusion that most original *sk- groups were at some point lost in that dialect of PSl which later became dominant. There is nothing to disqualify the possibility that several *sk- groups were either reintroduced from some minor dialect, or formed anew. A perfect analogy for such a development is found in (Old) English. Numerous etymological doublets such as shirt-skirt display such a phenomenon – the sh- variants are (Southern) English, the sk- variants northern and/or Old Norse by origin.

Also, most sk- examples in Slavic do point to s-mobile variation, many even to such variation within Slavic, and often the s-less variant seems to be an original verbal root in Slavic. To list but a few – *skora ‘hide (n.)’ < *(s)er ‘cut’ (cf. kora ‘bark, rind’ and Cz. choroš ‘polyporus, a growth on the bark of a tree’); *skopьcь ‘eunuch, ram’ < *(s)kep ‘cut’ (cf. *kopati ‘dig); *skala ‘rock’ < *(s)kel ‘cut’ (cf. *klati ‘hit, hew’). Most such forms result from morphological derivation that could have taken place almost at any point in time during the development of PSl. I therefore assume that:

(1) PSl. *sk- became *šš universaly in one dialect. (2) Where PSl. *x- corresponds to e.g. Germanic k- or h-, initial *s was either lost in Germanic, or appended earlier in Slavic.

Here (1) requires a solid argument for such a development, and it seems that it can be provided both from phonetics and from linguistic typology. It is (2) which is more difficult to account for. Let me first address the Slavic part of the problem. It is an established fact that in PIE, only initial clus-ters of s+unvoiced consonant were admissible. Since variants in roots with 9 Slavic s-mobile is not as scarce as Southern claims: there are many examples of both s+ variants in Slavic and variation within this group, though mostly not *sk-/*K- variants.

Jan Bičovský32

Media-/sTenuis- are sometimes found, there seems to be an agreement that the sibilant in this position caused devoicing (and possibly deaspira-tion) of the following media. It has also been argued by many that in the vicinity of *s PIE palatovelars were retracted, so that the group resulted in *sk10. In *s groups the assimilation apparently worked either way, pos-sibly according to rules of syllabification or other phenomena. Therefore, whatever the initial velar in any other IE language is, Slavic would most probably have a *sk-. Accepting now tentatively that the scarcity of sk- in Slavic is not a coinci-dence but a trace of some phonetic change, the argument follows naturally: the anlaut consonant cluster *sk-, in reality an apical plus a dorsal, was mostly assimilated to *šš (via *šx). It is certainly easier to envisage such a change than any change that would lead to *x- directly. How would such a change proceed? Here we are fortunate enough to have a perfect example of such a development in a well documented IE language, whose dialects and historical stages are well studied and we can be quite confident that our knowledge is very accurate, viz. German. That Old (High) German *s was an apical is shown by numerous loan words in Slavic languages – such as Cz. škola, růže ‘school” ‘rose’ = Germ. Schule, Rose etc. The reflexes of earlier sC and s were really very close to (Slavic) šC and š/ž, and after the original *s in most environments merged with the new s < *ts < *t and became a laminal, residual sC clusters were further retracted to šC, com-pare Eng. swim Germ. Schwimmen etc.11 But šk continued to assimilate – and the intermediate stages are still to be observed in the pronunciation of Dutch – [sx]; in Dutch, apical s is still the default pronunciation of the sibilant. It is worth noting that no such change occurred in sp- and st- groups. Before the regular development to German š , a geminate šš seems to be the most plausible outcome of further assimilation (compare the development of šč in Russian, or the sci/sce in Italian etc. for the ini-tial geminate sibilant).

In English, examples such as shower, shore, shorn, shrink from OE scūr scora scoren scrincan show that palatalisation here was not the main fac-tor in the change of sk- through šš- (note that in contrast to original VsV groups, in VšV groups no lengthening of vowels in medial ‘open’ syllables takes place – the sequence was still a geminate) to š- and it was rather a pro-gressive assimilation in the consonant cluster as was the case in German.

10 For an exhaustive treatment of the *s+- problem see Lubotsky (2001).11 Almost the same process can be followed in the evolution of Portuguese (Camara 1972).


Although tempting, I believe that a comparison of Slavic *sk > *x to late Prakrit sk > kh- is not fruitful (see details in Cardona 2003: 109). In the latter case, the development proceeded along these lines: first s- lost its oral stricture and became a mere voiceless glottal spirant (reminiscent of contemporary Latin American Spanish VsC > VhC), next, this spirant was reanalysed as a preaspiration of the following stop, and finally, it was shifted to post-aspiration, which is a more natural ordering of the two articulatory gestures. For one reason, this could not have happened in Slavic. In Prakrits, all the initial SC clusters, regardless of the articulatory position of the stop, underwent the same development, which resulted in ph and th. Nothing comparable happened in PSl if we are to judge from its latest stages. Of course, if *x was already in the system, while *tʰ *pʰ were not, this putative *kʰ could have merged with *x, while aspiration of the other two groups was simply lost. But if, as is more likely, it was still *š at that time, no such merger is imaginable.

Following degemination, which would only be the logical next step in such an initial cluster (compare such developments in e.g. Italian, German, English), in Slavic, this later *š merged with the ruki *š, which probably also resulted in a phonologisation of the latter, since prior to this merger, the situation resembled that of Vedic, where, loanwords exluded, the ret-roflex ṣ was still only an allophone of s. The second problem, namely what happened to the k- and g- reflexes in other PIE languages, is probably the harder. Although in general, historical linguists do not have difficulties in linking roots with and without initial s- together, the problem of this so called s-mobile is still an enigma. In my opinion, this s- is indeed, as Edgerton 1958 proposed, the result of a san-dhi metanalysis in PIE and likewise in later languages. Edgerton further speculates about the possibility of e.g. nom.sg. o-stem –s being ambigu-ous as to whether it, in sandhi position before a consonant, represented a single or a double s, the second s being part of the next word. His argu-mentation uses the example of PIE *h1es-si ‘thou art” > *h1esi as a proof of degemination of *–ss- in PIE. Needless to say, this example stands alone and moreover, comes from a word that would surely score among the most frequently used in the language, so such a degemination is rather due to frequency than anything else. And, last but not least, in the case of s-mobile we speak of external sandhi. Mayrhofer goes on to argue that any possible group of *–s s- would have been treated as –s- anyway, so the speaker would not have been able to decide whether any *–s- in san-dhi position represents */–s-/ or */–ss-/ and would often select the latter. I

Jan Bičovský34

find it difficult to believe that such a metanalysis would have taken place very often. By far the most common type of metanalysis on such a large scale are combinations of phrases such as article+pronoun/noun (cf. Eng-lish apron, adder etc., parallels can be found in Italian, French and other languages), preposition+pronoun (cf. Slavic, where the final segment of prepositions ending in *–n (*sъn ‘with”, *vъn ‘in”etc.) whose final segment shifted to the onset of the pronoun *j-, thus creating a secondary subset of 3prs. pronouns) or preverb+verb (cf. Cz. bahnit se ‘to concieve (of a ewe)”), original from ob jagnitъ se,̨ in the same way also Cz. bořiti ‘demolish’ (< ob-oriti). A sequence of PIE Subject+Verb hardly compares to these exam-ples: all such structures are parts of the same syntactic phrase, even of the same prosodic sequence. Thus, if indeed s-mobile results from some such process of metanalysis some other source must be found. Since neither PIE nor its daughter languages in the earlier phases did have articles and since most roots with s-mobile are clearly verbal anyway, the candidates for such a source are to be found only among the primary adverbs/particles. Inci-dentally (or maybe not), many PIE adverbs do exhibit an *–s alternating with Ø, notably *eʰ-s, *sub-s, *h2ed-s, *ud-s etc. This *–s, apparently not a part of the root but some kind of desinence, might have played a distinct role in the PIE language, most likely signalling a free-standing adverb as opposed to a (future) preposition or preverb.12 In later stages, this role was mostly lost and the two forms merged, with either of the two becoming the default representation. If at some stage there existed both univerbisation such as *eʰ-bʰereti ‘brings out > pronounces etc.’, cf. OIr. epert ‘s/he said’ and juxtaposed *eʰs (...) bʰereti with probably the same meaning ‘brings out’, it is only a matter of time before loose particles cease to function as such (as it happened in Classical Greek, Latin, Classical Old Irish, Slavic and Baltic and most later Indo-Iranian languages) and common juxtapositions with *eʰs bereti would be treated as by-forms or equivalents of univerbised forms, with no clear role of the *–s- element. Metanalysis, both phonologi-cal (leading to .sC rather than s.C, thus naturally favouring open syllables) and semantic (the s+form would become the ‘marked form’ in the pair of, say, *sker and *ker, leading to a reinterpretation of their semantic relation ‘cut” and ‘cut abruptly, cut badly etc.”) was to be expected.

Even up to recent times, in English as in other Germanic languages, s-mo-bile survives as a quasi-grammatical element (this was the way Mark South-ern treated the phenomenon in Germanic), yielding English squeeze from obs. quise < OE cwysan, splotch from plotch or blotch, squab from obs.

12 In this way having the same role as the –s of PIE nominative sg. and pl.


quab, swirl from whirl, scratch from cratch etc13. For Slavic, I would assume that the input for *š- (future *x-) was either PIE *sk, or a later (but still very early in the PBSl. period) *sk from s-mobile enlarged roots. If s-mobile is to blame for at least some sC roots, s-less forms of the same roots would be expected to exist in the language – and it seems that indeed they do.

Since of all the IE language branches, it is Germanic and Baltic where most examples of s-mobile are found, in Germanic even post-Grimm Law examples, and since these two branches were the closest to Slavic in most respects, it should not be surprising that in the same way as there are dif-ferences between Baltic and Germanic, with sometimes the sigmatic variant in Germanic, sometimes in Baltic, there could have been sigmatic variants in Slavic, where none survived in Germanic and Baltic.

The corpusTaking sk- as the default source for PSl. *x-, I shall now examine some of the more certain etyma, assuming that this theory brings further argu-ment for accepting or renouncing any etymology. Words commonly held as Slavic (excluding possible onomatopoeia and loans) are the following: *xorbrъ ‘brave, valiant’ *xoldъ ‘cold’ *xlǫdъ ‘rod’ *xabъ ‘weak’ *xalǫga ‘jet-som, osiers’ *xipъ ‘arrow, stick’ *xirъ ‘wide’ *xmura ‘cloud’ *xoditi ‘walk, go’ *xajati ‘care’ *xotěti ‘want’ *xlębъ ‘waterfall’ *xomǫtъ ‘hames’ *xorna ‘food’ xorniti ‘to protect, to feed’ *xromъ ‘lame’ *xudъ ‘poor’ *xvojь ‘twig’ *xujь ‘penis’ *xvostъ ‘broom, tail’ *xvorstъ ‘brushwood’ *xrošťь ‘dry twigs’ *xovati ‘look after’ xvorъ ‘sick’ xribъ ‘mountain range’ *xormъ ‘temple’ * xvatiti *xytiti ‘catch’ *xytrъ ‘quick, clever’ *xvala ‘praise’ *xala ‘rugs??’ *xybati ‘go amiss’ *xolpъ ‘man’ *xoliti ‘shear’ *xъrtъ ‘greyhound’ *xyliti ‘to lean’ *xestъ ‘six’ *xibati ‘move to and fro’ *xapati ‘grasp’ *xerъ ‘dark, grey’ *xvějati ‘sway’ *xrędnǫti ‘to be sick’ *xumъ ‘rustle’, *xelmъ ‘scum’, *šedъ ‘grey”, all in all, twenty four examples. In these reconstructions, forms with initial *sk- correspond to Proto-Balto-Slavic, those with *š- to Early-Proto-Slavic, and those with *x- to Late-Proto-Slavic.

*xorbrъ < *skorb-r-osFormally an o-stem adjective with r-enlargement; this formation is well attested in Slavic – cf. *dob-rъ ‘good’ *mok-rъ ‘wet’. *xorb- seems to have

13 Examples from Oxford English Dictionary.

Jan Bičovský36

very close relatives in both Germanic *skarpaz ‘sharp’ and Lith. skarbs ‘sharp, cruel’, and Latv. škerbs – since *xorbrъ primarily desribes the good qualities of a warrior, a concept such as ‘a worrior as sharp as a...’ seems natural. MIr. cerb ‘sharp’ shows the asigmatic variant of the root. Incidentally, if the root goes back to PIE (which need not be assumed), it would be one of the rare examples of PIE *b.

*xoldъ < *skoldos < *s-oldʰ-osThe exact semantic parallel in both Germanic *kaldaz and Lith. šaltas for Sl. *xoldъ ‘cold’ leaves little room for doubt that these words are indeed related. One would also add Lat. gelidus with the same meaning. It is therefore very annoying that it seems impossible to reconstruct a single PIE root to oblige all three terms. Judging by Lat. –idus, it would seem that the root is *KelH-d ,h to allow for the –i-, resulting no doubt from the reduction of a medial syllable. Both Slavic and Germanic agree on *d ,h whatever we may think about the onset of the root. Latin and Germanic also agree on initial *g-/*-, so the one odd form left to be explained (apart from Slavic) is the overall unvoiced Lith. šaltas. Here I believe we can still count on earlier *žaldas – at least many instances of similar devoicing have been studied in the development of Baltic, see especially Machek 1934: 7-36 – the semantic and structural parallel is too strong to exclude relatedness, however, this word (and the ones studied by Machek) may also be the result of borrowing from some so-far unidentifiable IE language. In Slavic, with the *s- added, both devoicing and depalatalising took place, resulting in *skoldos and later *šoldos.

*xlǫdъ < *sklond-os In the sense of ‘rod’ or ‘stick’, Lith. sklanda ‘stick’ is the exact parallel, the expected Slavic form would then be *šlondos. It is difficult to posit a PIE root, though. Most likely, this BSl. root is based on PIE *kel ‘cut’, but the origin of the *–ond- enlargement is obscure.

*xabъ < *skobʰ-osThis word, although agreed to be of PSl. ancestry, is by no means easily explained. The basic meaning is probably ‘weak’ or even more likely ‘slack’, and meanings of words derived from it range from Cz. ochabnout ‘slacken, weaken’, Sln. habéti ‘weaken’ to LSorb. chamny ‘poor’ (from *chabny by


assimilation) and Pol. chabanina ‘rotten meat’. Since few extra-Slavic par-allels can be found, Jiří Rejzek 2001 attempts to relate these words to Czech chobot ‘trunk’, OCS xobotъ ‘tail’ and skoba ‘hooked nail, rivet’, and out-side of Slavic, to the Lith. verb kabéti ‘to hang’. The original meaning of the root would thus be something like ‘to hang loose’. Little evidence for other explanations can be adduced: there is Lith. skabeti ‘cut’, ON skamma ‘hurt’, which seem to fit formally, but the semantic relation is probably too remote to allow any useful comparison.

*xalǫga < *skalonga < *(s)kʷāl-The original meaning of Sl. *xalǫga is difficult. OCS xalǫga has the mean-ing ‘hedge’ or ‘osier fence’, while Sln. halóga ‘kelp’ and SCr. hàluga ‘flotsam osiers’ both seem to imply some kind of organic material. Czech chaluha ‘kelp’ is a late loan from Sln. and provides no clue. Lat. squālidus ‘dirty, unkempt’ and squālus ‘dirt’ both come near semantically, but probably not near enough. On the other hand, if ‘oisers’ come as secondary, ‘flotsam’, the kind of organic material accumulated in flowing water outside the main current, would work very well for both Slavic and Latin. Slavic *kalъ ‘sediment in water’ might be connected as well. The ending *–ǫg- is rather scarce in Slavic, but if *ostrǫga ‘spur’, *bělǫga ‘sturgeon’, *pьstrǫgъ ‘trout” all derived from primary adjectives (‘sharp”, ‘white”, ‘pied”), are examples of the same derivative suffix, then perhaps *xalǫga does come from *kalъ ‘dirt, sediments in water’ as ‘a place with sediments’. I propose then a PIE root *(s)kʷeh2l or *skʷāl ‘sediment?”, to which Gr. πήλος ‘mud’ would be related. There is also the large group of cognates for PSl. *kalǫga or *kalǫža ‘a (dirty) puddle’, which exhibits both the root *kal and at least partly also the suffix –ǫg-. Alternatively, the two words might even be dialectal forms of the same etymon.

*xipъ < *skeip-osMost often, the three Skr. words, kṣípati ‘to throw’ kṣip ‘finger’, and kṣiprá ‘quick’ are cited as cognates, although it is difficult to see how all of them could be related. Initial *ks- would of course solve the problem at once assuming the ruki rule. The Skr. words themselves are not very clear, and at least for one of them it is not even sure whether it reflects original *ks- rather than *sk- (chipra vs. kṣipra). There is, however, a large group of words with almost the same meaning (‘stick’, ‘rod’) in a number of IE languages – Lat. scipio ‘rod’, Gr. σκίπων ‘rod’ and possibly σκοιπ̃ος ‘beam’,

Jan Bičovský38

Lith. skie p̃as ‘rod’ - that seem to be a more likely match. In Slavic, there are two forms of the underlying root, *skeip, the first one in *xipъ > *šipъ and the second in *cepъ from *koipos. I assume that the verb *štěpiti, if related at all, is secondary to *cěpiti, with sъ- ‘down, away from’ – that would also explain the semantic difference between *cěpiti ‘to strike’ and *štěpiti ‘to cut away, cleave’.

*xirъ < *skeir-osI suppose the best explanation to star from is the root which underlies *čirъ ‘clean’, probably a by-form going back to *skeiros ‘clear”.

*xmura < *smuraA difficult word by all means. Presumably, there are forms of the same word which show initial sm- or šm- in Russian and Czech, there is also a form reflecting earlier *mur- in Czech mourovatý ‘striped (of a cat) ‘ and a whole group of words with the sense ‘soot’ or ‘coal’ derived from PSl. *murъ. These latter seem to have a cognate in Germanic *smar- ‘to annoint’ = Eng. smear. But it would prove difficult to reconcile the two structurally, even though the semantics seem close enough.

*xoditi < *skodʰ < s-gʰodʰI have already expressed my doubts that this is a case of sandhi change. Apart from the chronology, there is one more difficulty in equating the root with e.g. Gk. ὅδοσ ‘way’. According to Winter’s Law- now well estab-lished and seemingly as universally applicable a law as any PIE sound law usually is - the sequence *od should have given PBSl. **xōd > **xād, as it does in, e.g. *ed ‘eat’ > *ēd. Only a root such as **sodʰ would satisfy the rules of historical phonology, but alas, we do not have such a root attested elsewhere. However, it does display ablaut alternations (*xoditi, alongside the frequentative *xōdjeti, the l-participle *xьdlъ, past act. participle *šedъ), so apparently the root is quite old in Slavic. Also, it should be noted that we do not seem to have other such roots with sandhi *x- among the large group of PIE verbal roots with *sV- as likely to have been prefixed as is *sod, such as *sek ‘cut’. PIE *gʰedʰ seems to fit phonologically. The mean-ing ‘unite, come together, fit together’, appears in Skr. gadhitaḥ ‘connect-ing, held together’ and seems to have the same meaning in Germ. – cf. OE tō-gædere ‘together’. From the same root comes Germ. *gōd- ‘fitting, good’


and of course Sl. *god- whose derivatives range from ‘(right) time’, ‘year’, ‘hour’, ‘happen’ ‘feast’ to ‘fitting, opportune’. If the basic form of the whole paradigm is originally a noun *sgʰodʰos > *skodʰos ‘comming together’ or ‘assembly’, the other verbal forms are referring to ‘getting together’ with a broadening of the sense to ‘going’.

*xajati < *sko-

This root is mostly attested only in the negative compound *ne xajati ‘to let be, to allow’ – while the original meaning can still be found in some Slavic languages, as ‘care about/for’. Greek ἄσκεω ‘to strive, to attempt’ comes near, if the original meaning was something like ‘to care for’ or ‘to look after’. The semantic parallel in words such as OIr. scíth ‘weary’ or Toch. B. *skai ‘work’, both of which have been compared to the Slavic word, is less satisfactory.

*xotěti < *skoteteiThere is only one possible parallel with at least partly matching semantics, viz. Lith. ketéti ‘to make ready’, but phonetically this is not persuasive. Without etymology thus far.

*xlębъ < *sklembosThis word is mainly attested in East and South Slavic, by R. xljab, SCr. hljeb and OCS xlęb. The only related form is found in the Lith. verb sklemt̃i ‘to slide off”, but the semantic fit is almost perfect, and therefore *sklembos is warranted at least for BSl.

*xomǫtъ < *skomontosSince this word, ‘hame”, refers to an item connected to one of the trade-marks of IE culture – horsemanship, there is a natural bias towards recon-structing it as a PIE root. However, only in Germanic and Baltic do we find closely related words - Lith. kãmanos ‘leather bridles’ and Dutch haam > Eng. hame. Only the initial part of these words seems to match – all could possibly come from the PIE root *em ‘to cover” – with different deriva-tions – Germ. *χamaz, Balt. *kamanā and EPSl. *skomontos. The change in Balt. of PIE * to *k might very well be the result of original s-mobile, which we would expect in Sl. as well. Although in Balt. * often under-goes depalatalisation, this is mainly true in the vicinity of liquids, *s and

Jan Bičovský40

* – none of which would apply here. I therefore propose BSl. *skomont- as the original form. Since the BSl. and Germ. forms seem to be very close, I believe that a borrowing from an unidentified Turkic language is not pos-sible – unless it happened before the operation of the Grimm’s Law – and as far as I know there were no Turkic peoples in the proximity of Balto-Slavs at that time (500 BC is commonly taken as the time when the Germanic Consonant Shift commenced).

*xorna < *skornaFirst of all, the often invoked Avestan arəna (last probably Derksen 2008 : 205), although formaly almost a perfect match to PSl. *xorna, suffers from one weakness – its cognate in Scythian, the most likely donor of this word, was farna, and it is difficult to understand why an initial f, absent from PSl. phonemic repertoir, would be rendered as *x- rather than *p- or *-. Moreo-ver Scyth. farna means ‘plenty’ rather than ‘food’ or ‘fodder’ and comes from the PII. *parHnas ‘rule”, cognate to Skr. parṇa of the same meaning and PIE *pelh1 ‘fill”14. Sl. krъmja ‘fodder’, itself without cognates, could be connected via a PSl. root *ker, the former na-derivative *skorna (comp. *stŕeti ‘spread’ > *storna ‘side, extension, country’), the latter from *kr-ma. The semantic shift proceeded from ‘care, custody’ to ‘feeding’.

*xromъ < *skromos < *erSl. *xromъ ‘lame, crippled’, well attested in all major Slavic languages, seems to have a tempting parallel in Skr. śramáḥ of the same meaning. This formation, however, probably reflects a participle in *–omos, productive both in Indo-Iranian and Slavic, to the root *er ‘cut’ – probably with the meaning ‘injured’. For Slavic, we would expect **sromъ > **stromъ (comp. *ob-srov- ‘island, lit. flow-around” > *ostrovъ), and there seems no other conclusion but to project into PSl. a by-form *skromos, possibly with a somewhat stronger and more affective meaning. It is possible that Pol. poskromić ‘tame – to cut out’ and PGerm. *skrama ‘wound’ both reflect the original *sk- by-form. Perhaps the term applied to animals and people somehow mutilated so as not to be able to walk or breed properly. Germ. *χarm- (Eng. harm) comes very close in its consonantism and meaning, but the ra – ar change seems to disqualify it.

14 See the discussion of this etymon in Lubotsky 2002.


*xudъ < *ksoudosI have not been able to find any *sk- cognates elsewhere. Skr. kṣudra ‘hun-gry’, with the expected zero-grade in an r-derivative, seems to offer the best explanation by far and *xudъ would therefore be one of the rare instances of ruki-rule initiated initial *x- in Slavic. But there is a minor objection, namely the non-application of Winter’s law, which would have yielded a long vowel and therefore an acute. Derksen op.cit. 206 proposes Meillet’s law to overcome this obstacle. But there is also Skr. kṣudhā- ‘hunger’, with paralels in other Indo-Iranian languages (NAv. šuδō ‘hunger’, MPer. šwd, Khot. kṣū) with original *–dʰ- or *–dh2. This may be as good an explana-tion as the connection to kṣudra, but also allows us to circumvent Win-ter’s law.

*xvojъ < *skojis Systems, diss. under review (to be published in Linguistik Typology). Lith. skuja ‘pine needle’, OIr. scé ‘pine’, and W. ysbyddad ‘pine’all point to a pre-form *skojā possibly from a root *ske ‘to cover”. *xuj ‘penis’ is likely to be related to this etymon as well, in the metaphorical sense of ‘twig, stick’.

*xvostъ < *sgostosThis word has been often compared to PSl. *gvozdъ ‘branch, bush”, on account of its almost perfect phonological and semantic match. In phonol-ogy, the main obstacle, leaving *xv- aside, is the voiceless *–st- for voiced *–zd-, a situation reminiscent of other words on this list – see above for Lith. šaltas ‘cold” instead of an expected **žaldas. By operation of the s-mobile, *skozdos would have been the desired outcome of *s-gozdos, possibly a derivative of *ges or *ges of uncertain meaning. PSl. *gvozdъ seems to have cognates at least in Germanic, e.g. OHG questa ‘twig”. A pos-sible desigmatised form is (*skost- > *kost-) in Cz. koště ‘broom”, OCz. koščiščě, alongside chvoščiščě.

*x(v)orstъ < *sk()orstosThis well attested Slavic word has a good parallel in PGerm. *χurstiz < *kʷstis, from which come OE hyrst ‘copse’, MLG horst ‘bush’, and PCelt. *kʷriston later to become W. prys. The origin of Sl. –v- is obscure.

Jan Bičovský42

*xovati - *skoateiBoth sigmatic and asigmatic paralleles can be found outside Slavic. PGerm. *χauwian and *skauwian of the same meaning ‘look, watch’ display the Germanic s-mobile variation – OE scēawian (> MnE. show) and hēawian ‘look’. An asigmatic form can be found in Lat. caveo ‘look after, care’, Gr. κοέω ‘look’.

*xvorъ < *skoros < *s-gorosPSl. *xvorъ has often been compared to PGerm. *swaras, whence MnE. sore, G. schwären ‘to get inflamed’ and therefore with PIE *ser ‘ache, be inflamed’, which can also by found in Av. ara ‘wound’ (note that Scythian equivalent would have been *fara!). Although in the light of the present theory I tend towards a parallel with Lith. iš-gvèrsti ‘to get weak’ and Toch. kwär ‘to get old/sick’, because deriving Sl. *xv- from PIE *s- would require an ad hoc shift from *s to *š, I do agree that *ser remains a very persuasive explanation.

*xribъ < *skrībos Psl. *xribъ denotes either a ‘hump’ or a ‘hill’ and the two meanings are not difficult to reconcile. There are few cognates in Slavic and none, as far as I know, outside this group. PSl. *xrьbъtъ ‘back’ displays the same consonant sequence *xrb and the vowel *i, but the vocalism is puzzling. Either the root was *skrib, from which a lengthened zero grade *šrīb- yields *xribъ, or *skreib ,h definitely an s-mobile root, whence zero grade *šrib- > *xrьb-. The other Slavic cognate is the well attested *grьbьtъ ‘back’, identical but for the anlaut with *xrьbъtъ, and its cognate (probably derived from) *grъbъ ‘hump’. Rejzek 2001: 216 notes that by-forms with initial k- and even sk- can be found in Slavic languages. If the word is of PIE origin, the original root could only have been *(s)grebʰ and the ī in *šrīb- must be an innova-tion.

*xormъ < *skormosThere are two ways to approach the semantics of this word and both ultimately point to the PIE root *(s)er- ‘cut’. In most Slavic languages, descendants of *xormъ denote various kinds of buildings, from Czech chrám ‘church, cathedral’ to OR. xorómъ ‘hut’. Among others, Snoj (: 210) argues that the original meaning was ‘a piece of hide cut out’ presum-


ably to provide cover or shelter. Comparisons have been made to PGerm. *skirma or *skerma ‘shield’ as shields were commonly made of animal hides or of wood strengthened with hides, and the transition of meaning from ‘shield’ to ‘shelter’ (note that MnE. shelter comes from shield) seems natu-ral enough, and from ‘shelter’ both ‘house’ and ‘temple’ could be derived easily. On the other hand, the ‘shield’ stage may not be necessary if ‘temple’ was the original meaning. Before larger constructions were erected above sanctuaries, any piece of land would do – provided it was somehow (even only symbolically) separated from its surroundings – ‘cut out”. This kind of marking out of a sacred place is widely attested all over the Indo-Euro-pean area, but it is not an exclusive feature of this culture. Lat. templum comes from a root *tem ‘to cut’ and means originally ‘something cut out’, from the same root comes Gr. τέμενος ‘precinct’. So does castellum, from which Slavic kostelь ‘temple/castle’ comes. It is tempting to conclude that PSl. *xormъ shares this origin, which likewise matches well with the Indo-European origin of Slavs. The original meaning would then be ‘(sacred) precinct’, an m-derivation of *sker.

*xvat- *xyt- < *šāt- *šūt- < *skeh2t- *skuh2t-The large group of words derived from the root *xut or *xvot includes *x(v)otěti or *xъtěti ‘to want, wish’, *xvatati and *xyti

chatreššar 2008 - univerzita karlovausj.ff.cuni.cz/system/files/chatressar 2008.pdf · there was...

Documents