Hrozný and Hittite: The First Hundred Years: Proceedings of the International Conference Held At Charles University, Prague, 11-14 November 2015 9004413111, 9789004413115

Citation preview

Hrozný and Hittite: The First Hundred Years

Culture and History of the Ancient Near East Founding Editor M.H.E. Weippert Editor-in-Chief Jonathan Stökl Editors Eckart Frahm W. Randall Garr B. Halpern Theo P.J. van den Hout Leslie Anne Warden Irene J. Winter

volume 107

The titles published in this series are listed at brill.com/chan

Hrozný and Hittite The First Hundred Years Proceedings of the International Conference Held at Charles University, Prague, 11–14 November 2015

Edited by

Ronald I. Kim Jana Mynářová Peter Pavúk

leiden | boston

Library of Congress Cataloging-in-Publication Data Names: Kim, Ronald I., editor. | Mynářová, Jana, editor. | Pavúk, Peter, editor. Title: Hrozný and Hittite : the first hundred years : proceedings of the International Conference held at Charles University, Prague, 11-14 November 2015 / edited by Ronald I. Kim, Jana Mynářová, Peter Pavúk. Other titles: Culture and history of the ancient Near East ; v. 107. 1566-2055 Description: Leiden ; Boston : Brill, 2019. | Series: Culture and history of the ancient Near East, 1566-2055; volume 107 | Includes bibliographical references and index. | Summary: “This volume collects 33 papers that were presented at the international conference held at the Faculty of Arts, Charles University in November 2015 to celebrate the centenary of Bedřich Hrozný’s identification of Hittite as an Indo-European language. Contributions are grouped into three sections, “Hrozný and His Discoveries,” “Hittite and Indo-European,” and “The Hittites and Their Neighbors,” and span the full range of Hittite studies and related disciplines, from \ and Indo-European linguistics and cuneiform philology to Ancient Near Eastern archaeology, history, and religion. The authors hail from 15 countries and include leading figures as well as emerging scholars in the fields of Hittitology, Indo-European, and Ancient Near Eastern studies”-- Provided by publisher. Identifiers: LCCN 2019030332 (print) | LCCN 2019030333 (ebook) | ISBN 9789004413115 (hardback) | ISBN 9789004413122 (ebook) Subjects: LCSH: Hrozný, Bedřich, 1879-1952--Congresses. | Hittite language--Congresses. | Hittites--Congresses. Classification: LCC P945 .H76 2020 (print) | LCC P945 (ebook) | DDC 491/.998--dc23 LC record available at https://lccn.loc.gov/2019030332 LC ebook record available at https://lccn.loc.gov/2019030333

Typeface for the Latin, Greek, and Cyrillic scripts: “Brill”. See and download: brill.com/brill-typeface. ISSN 1566-2055 ISBN 978-90-04-41311-5 (hardback) ISBN 978-90-04-41312-2 (e-book) Copyright 2020 by Koninklijke Brill NV, Leiden, The Netherlands. Koninklijke Brill NV incorporates the imprints Brill, Brill Hes & De Graaf, Brill Nijhoff, Brill Rodopi, Brill Sense, Hotei Publishing, mentis Verlag, Verlag Ferdinand Schöningh and Wilhelm Fink Verlag. All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission from the publisher. Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill NV provided that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910, Danvers, MA 01923, USA. Fees are subject to change. This book is printed on acid-free paper and produced in a sustainable manner.

Contents Abbreviations  ix Introduction  1

Part 1 Hrozný and His Discoveries 1

Hrozný’s Excavations at Kültepe and the Resurrection of a Bronze Age Palace  5 Gojko Barjamovic

2

Hrozný’s Excavations, 1924–1925: Sheikh Sa’ad, Tell Erfad  32 Jan Bouzek

3

Hrozný and the Decipherment of Hieroglyphic Luwian  44 J.D. Hawkins

4

Bedřich Hrozný and the Aegean Writing Systems: An Early Decipherment Attempt  62 Artemis Karnava

5

A Fruitful Collaboration between E. Sellin and B. Hrozný during His Viennese Years: The Cuneiform Texts from Tell Taanach and Their Impact on Syro-Levantine Studies  78 Regine Pruzsinszky

Part 2 Hittite and Indo-European 6

Consonant Clusters, Defective Notation of Vowels and Syllable Structure in Caromemphite  95 Ignasi-Xavier Adiego

7

Tagging and Searching the Hittite Corpus  120 Dita Frantíková

vi

Contents

8

The Phonetics and Phonology of the Hittite Dental Stops  147 Alwin Kloekhorst

9

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi  176 Martin Joachim Kümmel

10

The Word for Wine in Anatolian, Greek, Armenian, Italic, Etruscan, Semitic and Its Indo-European Origin  195 Reiner Lipp

11

Satzanfänge im Hethitischen  230 Rosemarie Lühr

12

Hittite Historical Phonology after 100 Years (and after 20 Years)  258 H. Craig Melchert

13

duttarii̯ata/i- and Some Other Indo-European Maidens  277 Veronika Milanova

14

One Century of Heteroclitic Inflection  295 Georges-Jean Pinault

15

From Experiential Contact to Abstract Thought: Reflections on Some Hittite Outcomes of PIE *steh2- ‘to stand’ and *men- ‘to think’  317 Marianna Pozza

16

Hittite Syntax 100 Years Later: The Case of Hittite Indefinite Pronouns  335 Andrei V. Sideltsev

17

Das unerwartete in der altassyrischen Nebenüberlieferung hethitischer Wörter  354 Zsolt Simon

18

The Personal Deictic Function of Hittite kāša, kāšma and kāšat(t)a: Further Evidence from the Texts  365 Charles W. Steitler

MUNUS/f

Contents

vii

19 Lycian Erimñnuha  382 Jan Tavernier 20 The Indo-European Feminine, the Neuter, and the Diagnostic Value of the τὰ ζῷα τρέχει rule in Greek and Anatolian  396 Annette Teffeteller 21

Sidetisch – Ein Update zu Schrift und Sprache  416 Christian Zinko and Michaela Zinko

Part 3 The Hittites and Their Neighbors 22 The LÚ.MEŠ SAG and Their Rise to Prominence  435 Tayfun Bilgin 23

Virginity in Hittite Ritual  455 Billie Jean Collins

24 Venus in Furs: Sappho fr. 101 Voigt between East and West  469 Alexander Dale 25

A Problem of Meaning: Variations in Hittite Landscape as Narrated in the Sun-God’s mugawar (CTH 323)  484 Romina Della Casa

26 „Fehler“ und Fehlschreibungen in hethitischen Texten  499 Susanne Görke 27

Personennamen der hethitischen Großreichszeit als Quellen religiöser Verhältnisse  506 Manfred Hutter

28 Die Gottheit Nikarawa in Karkamiš  518 Sylvia Hutter-Braunsar 29 From Nerik to Emar  531 Patrick M. Michel

viii

Contents

30 The Last Foothold of Arzawa: The Location of Puranda and Mount Arinnanda Revisited  544 Rostislav Oreshko 31

Phrygia and the Near East  571 Maya Vassileva

32

The Disappearance of Telipinu in the Context of Indo-European Myth  583 Roger D. Woodard

33

Foreign Medical Knowledge in Ḫattuša: The Transmission and Reception of Mesopotamian Therapeutic Texts in the Hittite World  603 Valeria Zubieta Lupo

Word Index  619 Index of Modern Names  649 Index of Historical and Mythological Names  657 Index of Text Passages  661 Index of Place Names  669 Subject Index  673

Abbreviations CHD

CHT EIEC HED HEG

HHw

HW

HW 2 IEW LIV2

NIL RE RlA

Reference Works Güterbock, H.G., Hoffner, H.A. and van den Hout, Th. P.J. (eds.) (1983– ) The Hittite Dictionary of the Oriental Institute of the University of Chicago, Chicago: The Oriental Institute of the University of Chicago. Laroche, E. (1971) Catalogue des textes hittites, Paris: Klincksieck. Mallory, J.P. and Adams, D.Q. (1997) Encyclopedia of Indo-European Culture, London: Fitzroy Dearborn. Puhvel, J. (1984– ) Hittite Etymological Dictionary, Berlin, New York: Mouton de Gruyter. Tischler, J. (1983– ) Hethitisches etymologisches Glossar, mit Beiträgen von Günter Neumann [IBS 20], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Tischler, J. (2008) Hethitisches Handwörterbuch mit dem Wortschatz der Nachbarsprachen [IBS 128], Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Friedrich, J. (1952) Hethitisches Wörterbuch. Kurzgefasste kritische Samm­ lung der Deutungen hethitischer Wörter, Heidelberg: Winter (Ergänzungshefte [1, 1957; 2, 1961; 3, 1966]). Friedrich, J. and Kammenhuber, A. (1975ff.) Hethitisches Wörterbuch, 2. völlig neubearbeitete Auflage, Heidelberg: Winter. Pokorny, J. (1959) Indogermanisches etymologisches Wörterbuch, Bern: Francke (3rd ed. Tübingen, Basel: Francke, 1994). Rix, H. et al. (2001) Lexikon der indogermanischen Verben. Die Wurzeln und ihre Primärstammbildungen. Zweite, erweiterte und verbesserte Auf­ lage bearbeitet von Martin Kümmel und Helmut Rix, Wiesbaden: Reichert (2nd ed.). Wodtko, D.S., Irslinger, B. and Schneider, C. (2008) Nomina im indogermanischen Lexikon, Heidelberg: Winter. Paulys Realencyclopädie der classischen Altertumswissenschaft, Neue Bearbeitung, Stuttgart: Metzler (1893–1980). Reallexikon der Assyriologie und Vorderasiatischen Archäologie, Berlin: de Gruyter.

Series AnOr AOAT

Analecta Orientalia (Roma) Alter Orient und Altes Testament (Münster)

x AS BAR BIAA BoSt BSIELL CHANE DBH DMOA HdO HSS IBK IBS KBo KUB LIEEDS LSIE MAOG MVAeG OIMP OIP PIHANS

RGTC SANER SAOC SSGL SSLL StBoT TAVO THeth UISK WAW

Abbreviations Assyriological Studies (Chicago) British Archaeological Reports (Oxford) British Institute of Archaeology in Ankara (London) Boghazköi-Studien (Leipzig) Brill’s Studies in Indo-European Languages & Linguistics (Leiden) Culture and History of the Ancient Near East (Leiden) Dresdner Beiträge zur Hethitologie (Wiesbaden) Documenta et Monumenta Orientis Antiqui (Leiden) Handbuch der Orientalistik = Handbook of Oriental Studies (Leiden) Harvard Semitic Studies (Cambridge, MA) Innsbrucker Beiträge zur Kulturwissenschaft (Innsbruck) Innsbrucker Beiträge zur Sprachwissenschaft (Innsbruck) Keilschrifttexte aus Boghazköi (Leipzig, Berlin, 1916–2015) Keilschrifturkunden aus Boghazköi (Berlin, 1921–1990) Leiden Indo-European Etymological Dictionary Series (Leiden) Leiden Studies in Indo-European (Amsterdam) Mitteilungen der Altorientalischen Gesellschaft (Leipzig) Mitteilungen der Vorderasiatisch-Ägyptischen Gesellschaft (Leipzig) Oriental Institute Museum Publications (Chicago) Oriental Institute Publications (Chicago) Publications de l’Institut historique-archéologique néerlandais de ­Stamboul = Uitgaven van het Nederlands Historisch-Archeologisch Instituut te Istanbul Répertoire Géographique des Textes Cunéiformes (Wiesbaden) Studies in Ancient Near Eastern Records (Berlin) Studies in Ancient Oriental Civilization (Chicago) Studies in Slavic and General Linguistics (Amsterdam) Studies in Semitic Languages and Linguistics (Leiden) Studien zu den Boğazköy-Texten (Wiesbaden) Tübinger Atlas des Vorderen Orients (Wiesbaden) Texte der Hethiter (Heidelberg) Untersuchungen zur indogermanischen Sprach- und Kulturwissenschaft / Studies in Indo-European Language and Culture (Berlin) Writings from the Ancient World (Atlanta)

Journals AASOR ABSA

Annual of the American Schools of Oriental Research Annual of the British School at Athens

Abbreviations AfO AION AJA AJP AnSt AO AoF AuOr BASOR BaM BB BiOr BSL BSOAS ClAnt CRAIBL FoL GM HS HSCP IF IncLing IstMitt JAC JANEH JANER JAOS JCS JEOL JHS JNES JNSL JRAS JSOT JSS MDOG MSS N.A.B.U.

xi

Archiv für Orientforschung Annali dell’Istituto Orientale di Napoli American Journal of Archaeology American Journal of Philology Anatolian Studies Archiv Orientální Altorientalische Forschungen Aula Orientalis Bulletin of the American Schools of Oriental Research Baghdader Mitteilungen Beiträge zur Kunde der indogermanischen Sprachen (Bezzenbergers Beiträge) Bibliotheca Orientalis Bulletin de la Société de Linguistique de Paris Bulletin of the School of Oriental and African Studies Classical Antiquity Comptes Rendus des Séances de l’Académie des Inscriptions et BellesLettres Folia Linguistica Göttinger Miszellen Historische Sprachforschung Harvard Studies in Classical Philology Indogermanische Forschungen Incontri Linguistici Istanbuler Mitteilungen Journal of Ancient Civilizations Journal of Ancient Near Eastern History Journal of Ancient Near Eastern Religions Journal of the American Oriental Society Journal of Cuneiform Studies Jaarbericht Ex Oriente Lux Journal of Hellenic Studies Journal of Near Eastern Studies Journal of Northwest Semitic Languages Journal of the Royal Asiatic Society Journal for the Study of the Old Testament Journal of Semitic Studies Mitteilungen der Deutschen Orient-Gesellschaft Münchener Studien zur Sprachwissenschaft Nouvelles Assyriologiques Brèves et Utilitaires

xii OA RA RANLmor REA RHA RSO SEL SMEA SSL TIES TPhS TSBA WdO ZA ZABR ZDMG ZDPV ZOrA ZVS

Abbreviations Oriens Antiquus Revue d’Assyriologie et d’Archéologie Orientale Rendiconti dell’Accademia Nazionale dei Lincei, Classe di scienze morali, storiche e filologiche Revue des Études Anciennes Revue Hittite et Asianique Rivista degli Studi Orientali Studi Epigrafici e Linguistici Studi Micenei ed Egeo-Anatolici Studi e Saggi Linguistici Tocharian and Indo-European Studies Transactions of the Philological Society Transactions of the Society of Biblical Archaeology Die Welt des Orients Zeitschrift für Assyriologie und Verwandte Gebiete Zeitschrift für Altorientalische und Biblische Rechtsgeschichte / Journal for Ancient Near Eastern and Biblical Law Zeitschrift der Deutschen Morgenländischen Gesellschaft Zeitschrift des Deutschen Palästina-Vereins Zeitschrift für Orient-Archäologie Zeitschrift für Vergleichende Sprachforschung

Languages Alb. Albanian Arm. Armenian Av. Avestan CLuw./Luv. Cuneiform Luwian/Luvian CS Church Slavonic Engl. English Fr. French Germ. German Goth. Gothic Gr. Greek HLuw./Luv. Hieroglyphic Luwian/Luvian Hitt. Hittite It. Italian Lat. Latin Latv. Latvian

Abbreviations Lith. Lithuanian Lyc. Lycian Lyd. Lydian MH Middle Hittite MS Middle Hittite script NH Neo-Hittite (N)Pers. New Persian NS Neo-Hittite script OCS Old Church Slavonic OE Old English OH Old Hittite OHG Old High German OIr. Old Irish OLith. Old Lithuanian ON Old Norse OPers. Old Persian OPrus. Old Prussian ORus. Old Russian OSax. Old Saxon OS Old Hittite script Osc. Oscan Pal. Palaic PIE Proto-Indo-European Rus. Russian Skt. Sanskrit Sp. Spanish Toch. Tocharian Umbr. Umbrian Ved. Vedic

xiii

Introduction Ronald I. Kim, Jana Mynářová and Peter Pavúk In December 1915, a 36-year-old Czech librarian at the University of Vienna by the name of Bedřich Hrozný published an article in the Mitteilungen der Deutschen Orient-Gesellschaft titled “Die Lösung des hethitischen Problems”, in which he identified Hittite as an Indo-European language. Despite initial skepticism, Hrozný’s claim, followed by his 1917 monograph Die Sprache der Hethiter, soon won universal acceptance and launched Hittitology as a new branch of Indo-European philology and Ancient Near Eastern studies. Hrozný went on to be appointed Professor at Charles University upon the independence of Czechoslovakia in 1918 and remained there for the rest of his career, briefly serving as Rector in 1939. Yet for all the recognition he enjoyed in his lifetime, he could scarcely have imagined that the centennial of his momentous discovery would be celebrated with multiple international conferences. From November 11 to 14, 2015, scholars from 22 countries gathered in Prague for a conference entitled “Hrozný and Hittite: The First Hundred Years”, organized by the Institute of Comparative Linguistics (the successor to the Institute for Cuneiform Studies, founded by Hrozný himself), Institute of Classical Archaeology, and Czech Institute of Egyptology of the Faculty of Arts, Charles University, in collaboration with the Oriental Institute of the Czech Academy of Sciences. The participants included specialists in Hittite language, literature, history, and religion, as well as Anatolian and Indo-European historicalcomparative linguistics and Ancient Near Eastern history and archaeology. We were honored to be able to host not only leading authorities in those fields, but also numerous younger researchers and doctoral students, a sure sign that Hittite studies remains as vital as ever at the beginning of its second century. The volume is made up of 33 contributions, divided into three thematic sections. The first, “Hrozný and His Discoveries”, contains five papers which document and evaluate Hrozný’s activities in the excavation of sites in Anatolia (Kültepe) and Syria (Sheikh Saʿad, Tell Erfad), his pioneering research on the cuneiform texts from Tell Taʿanach, and his significant, if long downplayed role in the decipherment of Hieroglyphic Luwian and rather less successful attempt at deciphering the Aegean writing systems. The second section, “Hittite and Indo-European”, consists of 16 studies on all aspects of Hittite synchronic and diachronic linguistics, including phonology, morphology, syntax, semantics, pragmatics, and etymology. Several papers are devoted to other members of the Anatolian branch such as Lycian, Carian, and Sidetic, while others focus

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_002

2

Kim, Mynářová and Pavúk

on problems of Indo-European historical-comparative grammar in which Hittite evidence plays a central role. The last section, “The Hittites and Their Neighbors”, presents 12 studies on diverse aspects of Hittite philology, religion, and mythology in the context of contacts with neighboring peoples of ancient Anatolia, Mesopotamia, and the Aegean. A recurring theme in many of these articles is the unique and complex combination of Indo-European inheritance and Near Eastern influences that characterizes all aspects of Hittite civilization. We would like to express our gratitude once again to the Faculty of Arts, Charles University and the Czech Academy of Sciences for their financial and logistical support, without which we could never have imagined organizing an event on this scale. Petr Zemánek and Dita Frantíková provided invaluable assistance at all stages, the former in particular with the preparations for the conference, the latter with the editing of several papers. Student volunteers from the Institute of Comparative Linguistics and Czech Institute of Egyptology ran the registration and book displays, coordinated handouts, assisted with problems both technical and mundane, and in general ensured that everything ran as smoothly as possible. Martin Gális painstakingly compiled the first draft of the indices. Finally, we wish to thank the anonymous reviewers for their many helpful criticisms, which have greatly improved the volume, and our editors at Brill, Katelyn Chin, Erika Mandarino, and Kayla Griffin, for their efficiency and expertise in bringing the entire project to fruition. The following contributions illustrate the flourishing progress of Hittite studies a century after 1915 and attest to the enduring legacy of Hrozný’s achievement. It is our hope that they will be consulted with profit by specialists from a wide range of disciplines, and thereby further the remarkable spirit of interdisciplinary and international cooperation which Hrozný himself so embodied.

Part 1 Hrozný and His Discoveries

∵

Chapter 1

Hrozný’s Excavations at Kültepe and the Resurrection of a Bronze Age Palace Gojko Barjamovic Abstract Few scholars have been as important as Bedřich Hrozný in bringing ancient Hittite culture to light after three millennia in oblivion. His contribution to the decipherment of the Hittite language is widely celebrated. His archaeological mission to the site of Kültepe in 1925, on the other hand—partly because it was never properly published— now tends to be dismissed as second-rate work. Although there can be no doubt that Hrozný’s enthusiasm as an archaeologist did not compensate for his lack of expertise, it is upon reexamination of his data possible to extract important information from his work that can add to the results obtained by others. This article reviews the results of his soundings on the mound of Kültepe and connects them to the earlier and later excavations conducted by Ernest Chantre and Tahsin Özgüç. This leads to a proposal for a reconstructed plan of the Warshama Palace in its entirety.

Keywords Kültepe – Acemhöyük – Hrozný – Anatolia – Minoan – Middle Bronze Age – Old Assyrian – history of archaeology – early Turkish Republic – Bronze Age palaces

When Bedřich Hrozný arrived at Kültepe, he was the third scholar to initiate excavations there. His main reason for coming to Central Turkey was his hope to pin down the find spot of the so-called “Cappadocian tablets” written in an archaic dialect of the Assyrian language, which had been coming out of the Kayseri area for more than half a century.1 Through those texts, he hoped to 1 A preliminary report of the excavations appeared in Hrozný 1927 alongside a few popular accounts for the press (cf. the bibliography provided by Adamčík 2013), and a short summary was published in Hrozný 1952. For two biting critiques of his work, see T. Özgüç 1999: 74–75 and Larsen 2015: 19–22. Critical reactions among Turkish archaeologists were almost immediate,

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_003

6

Barjamovic

trace the roots of Hittite history and society, thus continuing the work initiated by his decipherment of their language a decade earlier. Ernest Chantre had previously conducted two seasons of excavations at Kültepe in 1893 and 1894, and Hugo Winckler and Hugo Grothe had spent a few days there in 1906. The source of the tablets had eluded both teams, and there was still no definite proof that they came from the site. Winckler somewhat cantankerously noted that the places where the clay tablets were being found had in fact been roughly pointed out to him, but that he found the accuracy of the information doubtful, seeing as the individuals who had the benefit of the discoveries would not be inclined to give away anything that offered them an economic advantage. The archaeological potential of the mound was also viewed with skepticism by Winckler, who left—­ apparently disillusioned—after just eight days of laying out test trenches, stating that: “Wie bereits von Chantre geschildert, ist es kaum möglich, in dem aus reinen Erdmassen bestehenden Hügel … bestimmte Schichten festzustellen. Es ist alles durcheinander gesunken.”2 Chantre had in fact achieved some success during his longer sojourn at the mound, proving not only the antiquity of the site and its probable identity as the source of the “Cappadocian tablets,” but also by publishing a volume—­ reasonably solid for its day—with a chapter devoted to his work at Kültepe that contains descriptions, plans and drawings resulting from this work.3 Upon his arrival in Kayseri in May 1893, Chantre had initially been denied even the most basic information about the location of the mound, which he had been told in Constantinople (now İstanbul) was the origin of the cuneiform tablets. He had in the end been given directions by the priors of the convent of Surb cf. e.g. the scathing comment offered by Remzi Oğuz [Arık] 1933: 23 “kıymetli lisan âlimi Çekoslovak Hrozny’nin—Şlimanın hafriyatından farksız ve Hashöyük’teki araştırmalar ­kadar arkeolojiye zararlı” (i.e. that Hrozný’s excavations were indistinguishable from those of Schliemann, and as destructive as those at Hashöyük). 2 Winckler 1906: 634. 3 Chantre 1898. See also the review by Crowfoot (1899), which shows the importance of Chantre’s finds in contemporary scholarship. Note the brief news piece in olz 3 (1900/Juli), p. 273 about villagers mining the mound at Kültepe for fertilizer and finding large stone walls made of volcanic tuff in which the individual stones were kept together by iron clamps and looked as if they formed part of a temple temenos. It also mentions the existence of a stone inscription in Greek (possibly Kt 99, purchased and later published in von der Osten et al. 1933: 89–90), a terracotta model showing a divine procession, and some cuneiform tablets (called “bricks” by the author). Presumably the local excavations took place during the years between Chantre’s visits in 1893–1894 and that of V. Waldemar Belck and C. Lehmann-Haupt in 1901. Excavations had clearly gone on at the mound for much longer, since the late Hellenistic bronze tablet coming from the site had been sold to Berlin in 1890 (cf. Robert 1963; ­Barjamovic 2015a).

Resurrecting a Bronze Age Palace at Kültepe

7

Karapet, who controlled most of the land in the valley with the exception of the village of Karahöyük, located at the foot of the mound. Having spent a long time searching for the site and in order not to waste any time in clearing new areas, Chantre began his excavations at points that had previously been dug by the villagers for fertilizer as well as treasure. With only fifteen workers, wooden shovels, and poor pickaxes at his disposal, and working out of the nearby monastery against the wishes of the local village, Chantre’s first season of work was still successful in gathering a number of objects and proving the multi-period nature of the site. Deeper excavations had to wait until the following year, when Chantre returned with tools and wheelbarrows from Europe and employed a total of almost one hundred local workmen at the site. According to his report, Chantre cleared up the vertical cuts that had been made by the locals in the eastern and southern parts of the mound. Entire “halls” 50 to 60 meters wide had been dug into the mound at numerous points, exposing cuts of several meters in height. He dedicated two squads of workers to the task of clearing those cuts, while two different groups were occupied in opening trenches at other points to find “the order of the superposition of deposits.” Unfortunately, Chantre lacked the skills to handle the highly complex stratigraphy of the mound, which causes great difficulties even to this day, and so his cuts revealed to him only “the rubble of various ages, offering just superpositions of debris in the most indescribable mess.”4 During his 1894 season, Chantre opened a total of six trenches—12 to 15 m long and 4 to 5 m wide—in the center and western parts of the mound, some of them up to 17 m deep. All he could distinguish were parts of “shapeless walls, and objects of all kinds in the same disorder as what had been discovered in the cuts made on the southern edge,” built as “mud brick walls on stone foundations.” The trenches he opened on the perimeter of the mound revealed much more substantial bastioned walls enclosing the site, some of them still visible to this day from the visitor’s parking lot on the western side of the mound. The locals referred to the point where the main path enters the mound as the “city gate,” and Chantre surmised that succeeding archaeologists, “more fortunate than I, will return some day to my excavations and discover the doors of these walls, the walls themselves, and all the buildings that must have existed for a city of this size.” That process is still ongoing a century later. Chantre was aware that the highest point of the mound, where Hrozný was later to conduct his excavation, was likely to represent the main structure of the city. He noted that “substantial walls of the citadel remains are still to be 4 Chantre 1898: 74.

8

Barjamovic

found on several other points on the mound. The largest mass is located on its central part, and forms a sort of prominence that reminds one of the ruins of a building that was to dominate the whole city. It is, according to the people of Kara-Eyuk, the location of the palace, and it is here that one must discover treasures! My excavations, which began at this prominence on its northwest side, and went down to a depth of nine meters, showed an accumulation of debris up to four meters. At this level, we began to find wall sections of two meters thick on average, which can be attributed to buildings without, however, us being able to discern the shape of the building to which they belonged.”5 What Chantre had come upon was presumably the same gigantic palatial structure that Hrozný and Tahsin Özgüç would dig after him, although the exact location of his sounding is unclear. An attempt to reconcile Chantre’s map with a satellite image of the site shown on Fig. 1.1 does not pin down the places where he worked. He may have come down into the middle of the structure, or he may have missed it altogether. Regardless, he does not seem to have damaged the building too much for Özgüç to still be able to discern its general layout. The monumental size of the mound, the palace, and the complexity of the stratigraphy proved too much for Chantre. In the end, he left little actual data to help Hrozný get a sense of the chronology and nature of the site. In fact, Chantre still knew so little about site formation that he surmised that the mound itself and the abovementioned four meters of debris, which consisted mainly of ash, charred wood and vitrified mud brick, had to have been caused by volcanic eruption. He supposed that the mound itself could have been formed as a small outlet of the nearby Erciyes Dağ that had lifted up the remains of Kültepe, but never broken into actual eruption. Needless to say, we now know that his theory had no merit, but it does still warrant mention due to the fact that it was the Stand der Forschung when Hrozný came to the site. In addition to the problems caused by his predecessor’s inability to handle the magnitude of the excavation, the period immediately following the Great War was a highly tumultuous time to work as an archaeologist. Hrozný’s native state of Czechoslovakia was just seven years old when the expedition ­departed, and the Ottoman Empire where Kültepe was located had collapsed less than a decade earlier. Only four years prior to Hrozný’s arrival, there had been a Greek army of 125,000 men standing 50 km west of the new state capital at Angora (today Ankara).6 And by 1923 both the large Armenian and Greek populations that had owned and inhabited the valley where Chantre had 5 Chantre 1898: 77. 6 Smith 1973: 232–236.

Resurrecting a Bronze Age Palace at Kültepe

9

Figure 1.1 An overlay of the plan published by Chantre (1898) on a modern satellite image. The location of the road and the village on his map only seem to fit the actual situation partially, yet some of the houses still stand, which means that neither is unlikely to have moved much. Taking into account the position of the zone marked on the plan as having been destroyed prior to his visit (the Fouilles des paysans), it does seem to render the situation correctly grosso modo. However, the scale is about 5% larger than actual measurements, and the north arrow points some 10° E of true north and 12° E of the position of magnetic north in 1894. The exact position of the two areas in which Chantre conducted his excavations near the palace thus remains unclear. According to the overlay, he would have come down inside the palace in the regions of Rooms 16, 17 and 23 (cf. Fig. 1.3). This seems to be supported by the remark of T. Özgüç (1999: 76), who states that Chantre “opened two pits on the west side of the citadel and here removed some of the rooms as he went down.” However, according to Chantre’s plan, it looks as if his excavations were carried out horizontally from the north and into the section created by the locals digging for fertilizer. Given the uncertainty of his plan, he may therefore have missed the Waršama Palace altogether, and his trenches may in fact correspond to the area visible roughly 50–100 m north and northwest of the palace where two clear precipices are visible on the satellite image on either side of a “tongue” protruding north—not unlike what is seen on his plan. Given the imprecision in scale and orientation in his plan (and the difficult conditions under which Chantre worked), either option seems possible. In contrast, his report of digging at the summit (though not marked on the plan) and finding walls “2 m thick on average” seems to suggest that he did hit the main palatial building on the site.

10

Barjamovic

worked a generation earlier had been exiled or slain.7 There was no longer any trace of the medieval monastery Surb Karapet that had been of such help to Mr. and Mrs. Chantre, and the region had seen an almost total shift in population and toponymy in less than a decade. The background of how Hrozný managed to get funding directly from the Czechoslovakian Parliament through the personal backing of Beneš and Masaryk, and to obtain the permits to work at Kültepe, has been told elsewhere.8 Suffice it to say that excavations had to be carried out under extremely difficult local conditions: poverty and great political instability were threatening se­ curity,9 there was no access to clean drinking water, and malaria was ­rampant 7 For a discussion of the local demography, cf. Der Matossian 2013: 191–193. 8 The recent work in Czech by Kopecký (2011) provides the most detailed biography on Hrozný written to date. It contains an account of the preparations for the excavation, and situates it within a broader narrative on the formative phase and identity of the post-WWI Republic of Czechoslovakia. A biography in English is found in Matouš 1949. For a narrative with focus on his work at Kültepe, cf. Larsen 2015: ch. 2. In a letter dated January 26, 1922 kept in the Yale Babylonian Collection, Aleš Hrdlička, curator of physical anthropology at the United States National Museum (now the Smithsonian Institution), wrote to the treasurer of the American Oriental Society, Prof. Albert T. Clay at Yale, interceding on behalf of his Czechoslovakian compatriot Hrozný to ask for financial support of $1000 for “one or two trips to Asia Minor … a relatively small expense, but nevertheless an expense that under the depreciation of the valute of his country is unattainable where he is.” The letter suggests that Hrozný had already begun thinking about an excavation in Turkey at this early stage (though not necessarily at Kültepe). Conversely, a postcard sent by Hrozný to Prof. Clay dated January 10, 1925 still mentions only his intention to excavate in Syria that year (“dans le printemps 1925 je prépare la continuation de nos fouilles à Tel Erbd”). Matouš (1949: 28) states that the permission to excavate at Kültepe only came through during the summer as Hrozný had already begun work in Syria. He and Petraš wrapped up their work at Tall Rifa’at, and stood ready to excavate at Kültepe by late June. 9 Adamčík (2013) offers a useful summary in Slovak of the 1924 exploratory trip that Hrozný undertook with the architect Jaroslav Cukr. The two were assaulted several times on their return from Syria, particularly while in Kayseri and trying to decide whether Chantre was correct in his identification of Kültepe as the source of the tablets, or to follow the proposal by Sidney Smith to dig at Höyük Tepe west of Kayseri. The two scholars did manage to exclude the latter possibility, but under such duress that Cukr ended up fearing for his life and refused to return with Hrozný for the following season. On June 21, 1924 Cukr had to flee an angry mob that accosted him on the main street of Kayseri, and to seek asylum in a mos­­que when a young man called him a gâvur and began firing his pistol at him. The incident was allegedly the main reason why Cukr decided to leave his place on the team to Václav Petraš. An excavation permit was obtained the following year with the aid of Hrozný’s old friend Halil Edhem Bey, Director of the Constantinople Museum, and with full backing from the Minister of Education, Hamdullah Subhi Bey, and the Director General of Antiquities, Mubarek Bey. Kemal Atatürk personally signed the permit. The attitude toward

Resurrecting a Bronze Age Palace at Kültepe

11

on the marshy plain.10 Both Hrozný and his architect Petraš contracted the disease while working at Kültepe, and it was little help that they had limited archaeological training and that the mound they had set out to tackle was huge. Hrozný’s previous experience was restricted to a field season at Tell Taanech in 1904 under the guidance of Professor Sellin, and a couple of soundings into sites in Syria and the Transjordan (Sheikh Saad and Tall Rifa’at) in 1924 and 1925. Petraš had probably never dug before. When the team consisting of Hrozný, Petraš and a German superintendent of the dig workers named Grimmek ascended Kültepe with 150 local workers on June 21, their attention was immediately drawn to the same hill rising at the center of the mound that Chantre had first identified. They correctly assumed that this central height would contain an important building in the ancient city, and they began by laying out three search trenches some way away from Chantre’s earlier dig. The workers soon hit upon large walls that confirmed the initial assumption that this had to be part of the same structure identified by Chantre. The walls, which had a thickness of 1.50 to 2.30 m, were built without mortar using large natural blocks of volcanic andesite; rough-hewn stone topped by bricks that had been vitrified in a violent conflagration. In the following weeks, the trenches were gradually extended to form a square measuring some 50 by 50 m that revealed more of what appeared to be part of the same monumental building. Hrozný saw this as the stronghold of the ruler. Its rooms, however, were mostly empty, and Hrozný failed to find the source of the Cappadocian tablets as he had hoped. Regrettably, we have only a sketchy plan and a few photographs of this important work from his brief preliminary report of the excavation in 1927 alongside a series of newspaper articles for a popular audience at home,11 and a single page in the Inscriptions cunéiformes du Kültépé, Vol. I from 1952. What little was published suggests that Hrozný kept detailed records, but we are for the time being limited to his short descriptions and what one gets from combining those with the results of later work on the site. More extensive notes and photographic records, previously said to have been lost, have allegedly appeared at the Náprstek Museum and will hopefully provide more detailed information (see Bouzek, this volume). 10 11

the foreign team in Kayseri had changed markedly by 1925, and conditions at the village of Karahöyük had become tolerable. Chantre (1898: 72) likewise refers to the village beside the mound as being “en proie à l’impaludisme le plus violent.” Cf. the bibliography in Adamčík 2013 for a useful inventory.

12

Barjamovic

So far, posterity has, and with some justification, reproached Hrozný for his work at Kültepe. In fact, the later excavator, Prof. Tahsin Özgüç of Ankara University, barely seems to contain his anger in his writings on the matter. He clearly felt that Hrozný had deprived the world of crucial data from a crucial site. However, such presentism, though fully understandable, has prevented important connections from being made between the two excavations. In addition, Özgüç’s own work on the palace was published almost exclusively in a single monograph that came out more than four decades after his excavations of the building had begun; less than 50 written pages and plates constitute the sum total of published documentation at present on one of the largest Bronze Age palaces ever excavated.12 In evaluating Hrozný’s work, one must therefore first acknowledge the obvious political and logistical challenges that he had to face, as well as the circumstance that he had no archeological type-site to help him date his layers.13 ­Secondly, posterity did nothing to frame his results in a productive or relevant way. The fact that he still did manage to distinguish between Bronze Age layers and those extensive Hellenistic remains that had severely damaged the older building is more than Chantre and Winckler accomplished. He was systematic about where to open his trenches, and both his own report and Tahsin Özgüç’s aerial photo show that he did not just dig a random round hole, as is sometimes stated in critiques of his work. He stuck to one set of trenches for the entirety of his time on the mound, producing clear evidence to reject the theory proposed by Chantre that Kültepe was situated on an active geological zone and that the settlement had been destroyed by an eruption. He further disproved Winckler’s assertion that “[e]s ist kaum möglich in dem aus reinen Erdmassen bestehenden Hügel bestimmte Schichten festzustellen. Es ist alles durcheinander gesunken.”14 Instead, he notes that “[n]ous avons, au contraire, pu constater qu’il y a dans le Kültépé des couches tout à fait distinctes.”15 Eventually, he did of course also manage to prove not only that the site was the source of the Cappadocian tablets, but also that this was the ancient city of Kaneš—no small feat at a time when close to nothing was known about Anatolian prehistory. The sin of Hrozný therefore lay not so much in the dig itself as the fact that he never published his results in a comprehensive manner. History to some 12 13 14 15

T. Özgüç 1999. This came only with the excavations at the mound of Alişar by the Oriental Institute of the University of Chicago between 1927 and 1932 under the auspices of Erich Schmidt. Winckler 1906: 634. Hrozný 1927: 3.

Resurrecting a Bronze Age Palace at Kültepe

13

extent got in his way, first with the collapse of the First Republic and the First Vienna Award, and later with the Nazi and Soviet occupation of Czechoslovakia. In 1939, he had the opportunity to emigrate, but refused to leave the country and chose instead to stay on as President of Charles University and bring it through some of its most difficult years. A heart attack in 1944 effectively put an end to his career as an active researcher.16 The remainder of this article shows how Tahsin Özgüç’s rejection of Hrozný’s work prevented connections from being made, and that one may gain valuable insights from combining data from the two scholars, particularly in reference to the palatial building on the Kültepe mound, its layout and function. Hrozný states that his workers dug down to a depth of about 8 meters in the trench on the mound, although it is difficult to make much sense of the results as they are presented in the preliminary report. Contrary to Chantre, he may initially have thought that the stone foundations were walls, removing them and the mud brick walls sitting on top of them. He published no record of any small finds that were made, and it is hard to know if he collected many.17 He also believed that the building with the large walls had to be dated to the Hittite imperial period, and that it had been burnt down around 1200 bc. He thus assumed that the Old Assyrian tablets had to come from even lower levels, and dug down deep in a futile search for them. Everything changed on July 2, when Hrozný went to the nearby village of Salur to purchase four clay tablets that had come from the site. During his trip, he accidentally learned from the coachman and excavation guard that he was looking in vain for tablets on the mound, since the texts had in the past been excavated just east of the mound in a field belonging to the heirs of a man called Hadji Mehmed. He immediately sent Petraš with a group of local councilmen to investigate, and soon the group identified a meadow measuring roughly 52 by 32 m where one could clearly discern multiple overgrown trenches. Excavations on the mound were suspended on 17 July for the harvest season, and Hrozný left the following day for Istanbul to obtain provisions and a permit to move excavations down off the mound. Negotiations over the price of the field further delayed matters, as the heirs of Hadji Mehmed demanded 160.000 czk to sell their land.18 Hrozný was forced to obtain a state valuation 16 17 18

Matouš 1949. Although his regular articles in Národní listy and elsewhere seem to suggest that he did find objects; cf. Adamčík 2013: 16 for references and see fn. 38 below. The amount corresponds to approximately 7.2 kg of gold in 1925 prices, or ca. usd $250.000 in 2020.

14

Barjamovic

under the new Turkish antiquities law that would ultimately allow him to procure the land for czk 5.500. During negotiations, it became clear that not only the family of Hadji Mehmed and their fellow villagers of Karahöyük, but also the occupants of the nearby settlements of Salur, Mancusun and Gömeç had been digging there in recent years. This had resulted in skirmishes and led to the local jandarma forbidding further activities in the field. The policemen had instead taken over the illicit digging themselves. Hrozný returned to Kültepe on August 30 in the afternoon, but since malaria broke out in the camp immediately thereafter, workers only began excavating the new site identified by Petraš on September 20. The first few days revealed only scattered tablets, and Hrozný was beginning to wonder whether he was excavating a kind of hidden storage where the residents of the ancient city had saved their records before fleeing.19 The next few days proved him wrong, however, and by September 26 he uncovered the first larger group of tablets at a depth of some 2.5 meters. Over the course of nine weeks, the team uncovered the remains of several houses, and found several additional groups of texts lying in heaps or kept in storage jars, for a total of about one thousand tablets and envelopes. Additional finds included a couple of huge storage jars dated to the Hellenistic period, and allegedly also “a fragment of the Hittite throne of the governor” that it has not been possible to trace.20 It may have been lost in the 1969 fire that destroyed the storage facilities at the Bohemian castle of Benešov nad Ploučnicí where the finds from Kültepe were kept. Importantly, however, Hrozný carefully noted down the find spots of all tablets in the field, and a preliminary plan was later published on Plate 129 of the Inscriptions cunéiformes du Kültépé, Vol. i. It is unclear whether the two strata ii and Ib of the Lower Town at Kültepe as later defined by T. Özgüç had been missed or conflated by Hrozný, but by the time that Hrozný and Petraš left Kültepe on November 21, the expedition had produced a record that was sufficiently coherent to use as the basis for later archival analysis. In 1982 this ­allowed Larsen to conduct a survey of the archive of the Assyrian merchant Imdi-ilum s. Šu-Laban based on the plan published by Hrozný that connected text and locus.21 Jan Gerrit Dercksen continued this work in 2002 by assigning individual tablets to the 12 find spots identified by Hrozný.22 And in 2014 Thomas Hertel combined the work of his predecessors with a much-improved

19 20 21 22

Adamčík 2013. See fn. 46 below. Larsen 1982. Dercksen 2001–2002.

Resurrecting a Bronze Age Palace at Kültepe

15

Figure 1.2 An overlay of the plans from Hrozný 1927 and T. Özgüç 1999. The two have been placed on top of one another and fitted as best as possible—connecting walls when that seems feasible, while filling out the entire pit left by Hrozný as visible from satellite images and ignoring his reconstructions and the odd angles shown on his plan. When compared subsequently, the scales of the two maps agree within 1% of accuracy, which may be taken as independent support for the way the two plans are fitted. A cleaned-up and straightened version of the overlay forms the basis of the reconstruction shown in Fig. 1.3.

knowledge of the archaeology of the site to include his findings in a complete plan of the Lower Town settlement.23 By combining Hrozný’s original plans and what T. Özgüç had since excavated, Hertel showed that one could connect the two and create an overlapping map of the area. 23

Hertel 2014.

16

Barjamovic

This in itself serves to vindicate Hrozný’s archaeological work to some small extent, at least in terms of precision. The evidence applied by Dercksen proves that precise records of the individual finds from the field of Hadji Mehmed were made and allowed reconstructions of the original assemblage. The work conducted by Larsen, Dercksen and Hertel on the Lower Town area thus suggests that Hrozný’s work can be trusted to some extent and that it may be of adequate precision to permit its use also to investigate the building on the citadel mound. A similar attempt to reconcile Hrozný’s plan of the palatial structure with those published by T. Özgüç is made in Fig. 1.2. To some extent, the excavation notes provided by Hrozný offer greater precision than those given by T. Özgüç for his own work in the Lower Town area between 1959 and 2005. During the latter’s first ten seasons at the site, results were regularly published, but did not normally provide the exact find spots of individual tablets in a systematic fashion.24 This renders impossible any attempt at conducting archival studies and/or combining studies of texts with the content of the houses from which they came. The data irretrievably lost at Kültepe through lack of recording and publication is overwhelming. During the first three seasons, Tahsin Özgüç was doing well in terms of both research and publication, and that there was no immediate reason why this practice could not have continued. But political changes in Turkish academia appear to have caused changes in his approach and to have brought about a quickening pace of excavation at the expense of publication. Perhaps related to this shift was the effort by Tahsin and Nimet Özgüç in Ankara and Albrecht Goetze at Yale University in 1951 to build a joint field program at Kültepe, formed as a collaboration between Ankara University and the American Schools of Oriental Research. The effort was given strong support by several influential colleagues, most notably by Prof. Benno Landsberger in Chicago, who had himself taught Tahsin and Nimet Özgüç, and the field epigrapher at Kültepe, Prof. Kemal Balkan of Ankara University. All parties appear initially to have been animated by the prospect of learning from each other and coming to grips with the overwhelming amount of data that the first three seasons of work at the site had produced. The correspondence of Goetze, now kept at Manuscripts and Archives at Yale University, shows how he, Tahsin and Nimet Özgüç, Landsberger and Balkan went about planning the joint project. Goetze spent almost two weeks at the site between Aug. 28 and Sept. 7 in 1951. Nimet Özgüç was leading the excavation while her husband was on research leave in the United Kingdom. There is an almost tangible excitement in the 24 Cf. Barjamovic 2015b for a discussion of one important exception where it is possible to assign precise find spots for a few tablets excavated in 1948 in the Assyrian “schoolhouse.”

Resurrecting a Bronze Age Palace at Kültepe

17

letters exchanged between the three from Ankara, New Haven and London after Goetze’s return around Oct. 1. These in turn translated into activity on both sides to garner financial and institutional support for the endeavor. The mood begins to change in the letters dated from December of 1951 and into 1952, as the group slowly saw its initiative clandestinely undercut by an opposing group of Turkish scholars headed by Prof. Emin Bilgiç of Ankara University. Bilgiç clearly wanted the excavation for himself, and tried his best to remove Tahsin Özgüç and his chief epigrapher Balkan from the site. At the same time, he was himself corresponding with Goetze, to be invited on a scholarship to the usa, not realizing that the latter was aware of his duplicity. Ultimately, Bilgiç failed to oust Tahsin and Nimet Özgüç, but did succeed in postponing the all-important meeting in the Turkish Historical Association that was to decide on the joint project for long enough to prevent the 1952 season from happening.25 He then went on to persuade a sufficient number of committee members to stay away from the following meeting in the Department of General Antiquities, where a vote on the project had to be passed.26 As Goetze saw it, the plan of cooperation was vetoed, because its present members “cared more for expressing nationalistic resentment of Western scholarly pre-eminence than for seizing an unusual opportunity for progress in archaeology.”27 The result of this failed collaboration on the excavations at Kültepe seems to have been profound. Presumably in fear of further attempts to dislodge him from the site, Tahsin Özgüç retreated from all further collaborations, and did not invite foreigners to work at Kültepe again. He is in fact later said to have cautioned junior Turkish colleagues from entering into foreign partnerships, while the details, and even the existence, of a projected partnership with asor were no longer spoken about. The failed collaboration between Ankara, Yale and Chicago remained essentially unknown until a digitized finding aid for Goetze’s papers turned up in a Google search during the preparation for this

25 26

27

1952 was the only year where no excavation took place between 1948 and the present. It has always been referred to as a “study season” in later conversation, but Goetze’s letters make it clear that the real reason for the break in excavation was due to the asor affair. In a letter dated March 24, 1952 to Carl Kraeling, who was the president of asor from 1949 to 1954, Goetze reports: “What you have just read is the text of la lettre fatale. We learn from it that the nationalistic clique has prevailed and that therefore the proposed cooperation with ttk [Türk Tarih Kurumu = Turkish Historical Society] is impossible … ­Nobody could have foreseen this development, and the internal twist which my visit to Turkey and our proposal gave to internal tensions among Turkish scholars is really an astonishing phenomenon. I consider the matter closed.” Online resource: Grigg 1978 Albrecht Goetze Papers, Manuscripts and Archives, Yale University Library.

18

Barjamovic

article.28 Keeping the situation in mind, however, one may perhaps explain why the material from Hrozný’s excavations, and his developing approach to archival analyses of the textual evidence, were not systematically taken into account by Tahsin Özgüç. Digging fast and producing results of a tangible kind at Kültepe had become a political endeavor, and perhaps even a precondition for keeping the site. A team led by Nimet Özgüç returned to the huge palatial building where Hrozný had worked, opening up the first soundings in 1954 and continuing excavations until 1962. After a break of a few years, a team led by Kutlu Emre resumed work from 1966 to 1982.29 The levels dug by Hrozný were defined as belonging to Mound Level 7 and found to be contemporary with Lower Town level Ib. This huge structure became known as the “Waršama Palace” after a letter found in its vicinity addressed to a local ruler of this name. We now know that the building was quite a bit older than Waršama, and that the timber used to build it was cut around 1850 bc.30 The building occupies more than one hectare and measures roughly 100 by 110 m.31 T. Özgüç estimated that it contained at least 60 rooms at ground level, with roughly 40 of those rooms dug in the northern sector of the building by the Turkish teams. But those excavations cover only about a third of the total area of the palace, and the plan published by Hrozný suggests that the Czechoslovaks dug at least another 30 rooms in the southern and central areas of the palace. In all probability, a more comprehensive estimate would be close to 100 rooms and corridors. Importantly, Tahsin Özgüç notes that only the foundations of the building appear to have survived: despite his best efforts, he was not able to identify a single door in the building. And since the walls still stood up to a height of more than 4 m, this suggests that the basement rooms were accessible from above only by ladders or stairs, either individually or in groups.32 The walls 28

29 30 31 32

An inventory of Goetze’s papers MS 648, Box 3, folders 52–54 was made by Yale librarian Susan Grigg in 1978 and digitized and put online in 2015. Scholarly collaborations resumed at Kültepe in 2006 under the new director of excavations, Prof. Fikri Kulakoğlu of Ankara University. The site is now the center of a thriving community of Turkish and international scholars, and the biannual “Kültepe International Meetings” draw colleagues from all over world, who conduct research on material related to the site. They gather there for a week to present and discuss their work with emphasis on interdisciplinary exchange and integration. In addition, there is a push to revisit the results and locations of the old excavations, and more importantly, to consult and interpret the huge legacy of notes and plans that were produced during the years of excavations under Tahsin Özgüç. Kültepe studies are entering an exciting new era. T. Özgüç 1999: xxi. Manning et al. 2016 and 2017. T. Özgüç 1999: 79. T. Özgüç 1999: 81.

Resurrecting a Bronze Age Palace at Kültepe

19

consisted of massive stone foundations up to 250 cm in height, and were topped by another 150 cm of mud-brick built using bricks of different size from the walls proper of the palace above.33 The actual walls were erected in half-timber on top of the foundations, using vertical beams, crossbeams and mud-brick, and were thickly plastered (4.5–5 cm) on the outside down to the ­foundations, though never with any trace of color. Two narrow rooms were identified by Özgüç as possible staircases based on their shape (cf. Fig. 1.3, Rooms 16 and 42), but no finds were made to support his proposition, and it is difficult to divide the rooms into functional groups, or to imagine stairwells descending into doorless rooms. The functional layout (“residential,” “storage,” etc.) shown on Fig. 1.3 is taken from Michel 2015 and based on information provided by T. Özgüç 1999 and elsewhere.34 The data provided in the primary publications is occasionally

Figure 1.3 A reconstructed plan of the Waršama Palace based on the overlay of the plans from Hrozný 1927 and T. Özgüç 1999 shown in Fig. 1.2. Room numbers follow T. Özgüç 1999 and Michel 2015 and refer to Table 1.1.

33 34

T. Özgüç 1999: 80. Michel 2015 and 2019.

20

Barjamovic

contradictory, and the precise context of the finds is often unclear. The area in the palace dug by Hrozný had later been refilled and was mostly avoided by Tahsin Özgüç (in understandable frustration). It now stands as a raised mound in the center of the building with the remainder of the structure dug around it. It seems clear, however, that the large square identified by Hrozný in his report as a paved courtyard must have marked the center of the building and the rooms of the palace essentially built around it. Although his description is vague, Hrozný seems to suggest that the courtyard was elevated compared to the surrounding walls.35 If this is correct, then it appears that the courtyard would have been directly accessed through a southern gate, and that one would have to ascend from it directly to the first floor of the building. The western entrance found by Özgüç could provide an auxiliary access to the main basements, but the southwestern part of the building was badly damaged. Tahsin Özgüç provided abundant proof that a level above the excavated walls must have constituted the main floor of the building. He both showed how roof beams, brick sections, tablets and pottery had fallen from the upper floor to be deposited outside of the building to the north, and pointed to the fact that most of the functional rooms we would expect from a palace appear to be absent in the plan, including official and private chambers and cultic installations. Özgüç also notes that Hrozný probably found the palace courtyard,36 but states that all walls had been removed in order to dig further down without records being made and plans drawn. As seen from the combined plan on Fig. 1.3, this is not entirely accurate. ­Although the plan is sketchy and cannot represent the situation entirely correctly, the overlapping plans do appear to produce identifiable patterns. For instance, the eastern and western sides of the building are said by Tahsin Özgüç to have been badly damaged by later Hellenistic and Roman activities. The plan from Hrozný’s preliminary report shows the existence of ephemeral building remains in the same area, which he dated to the Greco-Roman period based on the find of characteristic terracotta.37 One can easily see how the adjacent areas dug by Hrozný could also have been heavily damaged in 35

36 37

Hrozný 1927: 4. This was rejected by T. Özgüç (1999, 75), who seems, however, to take Hrozný to say that the entire structure, and not only the courtyard, had been built on a terrace. Whether the central courtyard was elevated vis-à-vis the surrounding rooms could presumably no longer be verified by Özgüç, since Hrozný had removed everything in his search for deeper levels. The deep fill that Hrozný dug through under the courtyard was presumably the destruction debris of the Old Palace, contemporary with the Lower Town Level ii, and brought to light by Özgüc in other areas. T. Özgüç 1999: 80. Hrozný 1927: 4.

Resurrecting a Bronze Age Palace at Kültepe

21

a­ ntiquity. A second case of an apparent coincidence between the two plans relates to the system of oversized foundations located by Özgüç in the western part of the building. These appear to have formed part of an arrangement that was also detected by Hrozný and marked by thicker walls. This not only implies a measure of accuracy in the combined plan, but also suggests that the basement foundations in this part of the structure could have been supporting a heavier weight than in the remainder of the building. This might be related to the existence of the heavily fortified side entrance to the basement found by Özgüç, or it might suggest that this part of the structure carried the main suite of reception rooms on the first floor of the building. Most (if not all) rooms in the basement had floors of hard stamped earth covered with plaster. Hrozný never mentioned these floors, and he presumably missed them. The fact that no doors were found led Tahsin Özgüç to surmise that the overall function of the rooms might reasonably mirror the general distribution of the rooms above, and that those wings of the basement that had storage and administrative purposes would refer to offices on the floor above. However, nothing definite emerges, and the interpretations offered by Özgüç and Michel, while representing the most plausible scenario based on the few surviving finds, are inevitably tentative. The rooms of the building appear to have been almost completely emptied out before they were set ablaze (strongly suggesting intent) and leave us guessing as to what their purpose was. Relatively few and only mostly minor objects, such as ceramic vessels, seals, bullae (31 total), beads and fragments of figurines were recovered during the nearly three decades that the Turkish team worked in the building (cf. Table 1.1). This also goes to explain why Hrozný mentions few objects in his preliminary ­report—he probably found very little during his two months on the mound.38 In addition, like his successors, he was mostly digging through what proved to be a heavily disturbed zone of the central courtyard and its surrounding corridors. 38

According to his summary reports in Národní listy published between February 27 and April 10, 1927, finds made in and around the paved courtyard included basalt stone, pottery, mortars, weights, spindles, and the lower half of a massive stone relief on a block depicting a pair of feet in shoes with raised tips. The latter is said to have come from near the courtyard, and might date to the late Hittite stratum identified by Tahsin Özgüç in sector lxxxvii–lxxxviii/88–92 (cf. T. Özgüç 1999: 79). However, too little is preserved to allow any strong statements on stylistic grounds. The rendering of knee, ankle and muscle (cf. Hrozný 1927: pl. ii.1) might suggest an earlier date, but finds dating to the Iron Age were also made by Hrozný in what appears to be the same area, including the blade of an iron knife, clay loom weights, and “painted pottery with geometric shapes,” cf. Adamčík 2013: 12–13. A slightly different list of finds appears in Hrozný 1927: 5.

22 Table 1.1

Room 11

Barjamovic Rooms (numbers referring to Fig. 1.3) and finds with data drawn from T. Özgüç 1999 and N. Özgüç 2001. Some additional small finds coming from the palace are listed in T. Özgüç 1999: 84–94 without information about their find spot. For the location of the epigraphic finds, see Michel 2019: 128–130.

Finds

Three bullae “tied to rims of bottles” (stoppers) (Kt t/t 9–11) TÖ 87, NÖ 181 12 8 bullae (Kt z/t 4–9, 15, 23–24 and stopper 25) TÖ 88–89, NÖ 187–189 14 1/2 bulla (Kt 73/t 27) TÖ 88, faience seal-amulet (Kt y/t 10) TÖ 91 24 Storage vessels (“hydria”), TÖ 85: “evidently arranged in regular alignment, filled with contents. These are represented by a repetitive type” (no numbers listed) 25 1 bulla (Kt 73/t 30) NÖ 189, bowls (Kt 73/t 38, 40, 52, 53, 58–59) TÖ 93–94, pitchers (Kt 73/t 41, Kt 73/t 51) TÖ 94, jar (Kt 73/k 56–57, 61) TÖ 94 26 Steatite stamp seal (Kt y/t 6) TÖ 90, bowl (Kt t/t 5) TÖ 93, pitcher (Kt 73/t 48) TÖ 94 Outside Room 28 1 bulla (Kt g/t 280) TÖ 88: “in the palace debris outside Room 28,” NÖ 182: “in the debris of the south part of the palace” 29 1 bulla (Kt u/t 1) TÖ 88, NÖ 185 33 Two pitchers (Kt 73/t 46–47) TÖ 94, red serpentine stamp seal (Kt 73/t 29) TÖ 89, gold foil fragments (Kt y/t 7) TÖ 91 34 Hollow top of gold nail (Kt v/t 39) TÖ 91 35 TÖ 85: “Complete vessels on floor” with no numbers given 42 Three bowls (Kt 82/t 6–8) TÖ 93 44 Serpentine weight (Kt z/t 1) TÖ 92, hydria (Kt 73/t 43) TÖ 94, 1 bulla (Kt s/t 64) TÖ 88: “palace debris xciv/28,” NÖ 184: “Room 44” 46 7 bullae (Kt v/t 22, 41–46) TÖ 87, NÖ 185–186 47 Head of faience figurine (Kt 73/t 22) TÖ 90–91, Gold bead (Kt f/t 411) TÖ 91, spearhead (Kt f/t 84) TÖ 91 49 Three jars (Kt v/t 144–146) TÖ 94 50 Pitcher (Kt 73/t 42) TÖ 94

Resurrecting a Bronze Age Palace at Kültepe

23

Room

Finds

51 (corridor south of Room 50)

A sardonyx bead (Kt 73/t 31) TÖ 92, bowl (Kt 73/t 54) TÖ 94, sealed bulla stopper of a jug (Kt 74/t 13) TÖ 88: “in Room 51,” NÖ 182: “in the debris of the south part of the palace,” NÖ 189: “Room 51” Stamp seal, terracotta (Kt 73/t 16) TÖ 89

Palace debris lxxxviii/89 Outside palace at Bronze stamp seal (Kt g/t 309) TÖ 90 “Palace debris in Room 11 xci/89,” N. Özgüç 1959: 46 “refuse layer of the Warshama Palace (area Y/28)” Outside palace at Envelope Kt z/t 20, presumably dating to the Lower Town Rooms 11–13 Level ii period, but cf. Michel 2015: 167 and 2019: 129 (last, unmarked entry) following TÖ 93 for its later archeological context Along the south Bowl (Kt 82/t 9) TÖ 94 wall xcvi/95 Burnt debris of Obsidian vessel fragment (Kt 73/t etüdlük) TÖ 92 north fortification Outside Room 9 The letter of Anum-Hirbi (Kt g/t 35) TÖ 93 (see below) Nine administrative records, four or five of which can be Unknown: identified (Kt g/t 36, Kt g/t 42+Kt z/t 11, Kt g/t 43–44). “Palace on the Günbattı (1987) states that they were found with Kt h/t Citadel” and 330, while Bilgiç (1964) states that the latter came from a palace debris temple. NÖ 165, 183 notes that bullae Kt h/t 102, 318, (TÖ 93, NÖ 183–184, Michel 326–27 and Kt o/t 204a came from the Palace debris, the numbers supporting Günbattı’s assertion. Kt g/t 44 is 2015: 166–167 and 2019: 128) dated to rel 202. The remaining texts studied by Donbaz (1998) came from the mound but probably not from the palace. Most date to the level ii-period. Fragments of wooden furniture, charred, straight-sided, “In the rooms, with dowel-holes, TÖ 92. Two pieces pictured in pl. 78, principally 1–2; Fig. A. 35–36. among objects on the floors” Unknown (from Bone appliqué (Kt 81/t 126) TÖ 92, bowls (Kt 73/t 39 [the number suggests it came from Room 25], Kt 81/t the palace, but 117) TÖ 94, hydria (Kt 73/t etüdlük) TÖ 94, pitcher (Kt without exact 81/t 119) TÖ 94 provenance)

24

Barjamovic

Tahsin Özgüç, followed by Michel, stated that an absence of cultic and residential rooms is unthinkable in a building of this size and importance. This in turn led him to assign the rooms marked “residential” on Fig. 1.3 to have served that purpose. However, if all rooms were windowless basements, then the offices, living quarters, and the pomp and circumstance associated with a palatial structure should be assigned to the first floor.39 The rooms allocated to administration and storage by Özgüç are arranged according to two possible staircases, but a reexamination of the published data compiled in Table 1.1 suggest a more complex distribution of finds and precludes any straightforward interpretation. It seems probable that the basements were mostly used for various types of storage. This would fit with the image we obtain from the textual record about the political and social organization of the Kaneshite state, which was governed by a ruling couple, but administered by a large cadre of civil servants in an elaborate hierarchy of titles and portfolios.40 One can easily reconcile a model of organization that involves multiple offices and associated storage space with the architectural layout of the Warshama Palace. At what appears to be southern end of the building, Hrozný detected walls of “a radically different character” from the rest.41 He tentatively ascribed this to a separate temple structure, but interestingly, the feature as marked on Fig. 1.3 seems to fit exactly into the blank on Tahsin Özgüç’s plan of the southern wall of the palace. So, assuming that Hrozný got his levels right, it may instead constitute part of a main entrance or gate structure of the palace. From the gate, one would gain access directly to the courtyard and presumably proceed by way of staircases to the main level of the building. Multilevel palatial structures are well known e.g. in the Minoan palaces, and probably also at Alalakh,42 where a so-called “piano nobile” led into reception halls located on a raised level, often above magazines. At Knossos, Phaistos, and Malia, evidence for such second-floor apartments includes fallen fragments of decorated plaster, fallen doorjambs, and even pillar bases in the first 39

40 41 42

Michel (2019: 127) talks of residential quarters and rooms reserved for ceremony in the existing plan, while Michel (2015: 165) expresses a notion closer to the present: “Dans la pièce 16 se trouvait un escalier menant à l’étage où se trouvait vraisemblablement la résidence du prince.” Cf. Erol 2007; Veenhof and Eidem 2008: 219–233; Barjamovic forthcoming for a list and discussion of various titles, and Dercksen 2004, 2005, 2008; Shi 2015 for the statecontrolled system of land tenure and organization of labor. Hrozný 1927: 4. For a possible example at Alalakh in the Hatay, cf. Woolley 1946, but note the critique in von Rüden 2014: 60–61 of Woolley’s tendency to draw Aegean comparisons to the architecture and decoration at Alalakh (when little was known from Syrian and Anatolian sites).

Resurrecting a Bronze Age Palace at Kültepe

25

floor area.43 Although Hrozný published no floor elevations from Kültepe, ­Tahsin Özgüç states that an offset between floor levels does seem to have existed; perhaps the perimeter of the structure was built around a low rise.44 The central courtyard would in turn be located at its highest point. This would explain how one could walk directly from a gate and through a courtyard into the basement of the building while effectively remaining on the same level. On the other hand, there was no propylaeum or other obvious structure marking the route of ascent, and clearly the architectural concept was different from that of Aegean contemporaries. Instead, the progression from gate to colonnaded courtyard is a standard feature in later Hittite public architecture,45 and may have had its prototype at Kültepe, as suggested by the partially preserved “halls” that surround the courtyard on at least three sides. A 16th-century parallel to the general outline of the Warshama Palace comes from the 18-ha site of Kuşaklı just 120 km northeast of Kültepe. At less than half its size, “Building C” at Kuşaklı is thought to have been a temple, but seems to share some features of the Warshama palace, including a large rectangular courtyard around which the various building wings were arranged, and doorless rooms of a southeastern wing of the building that were built on a natural slope and went below the courtyard by about one story (ca. 3.20 m). Walls were constructed on stone foundations with rising mud walls, still standing in some places to a height of 3 m, and preserving parts of the ceiling that formed the floor above the basement. The total number of rooms in Building C at Kuşaklı is estimated to have been at least 110, which probably comes close to what the palace in Kültepe had. The relative proximity of the two buildings in time and space may even suggest a genetic relationship, the later building being a halfscale version of its predecessor. The main indication of the existence of an upper floor at Kültepe is the absence of any obvious candidates for certain functions that one would assume exists in a palace, as well as the lack of doors in a building whose walls were 43 Cf. Graham 1962. Also, the flat roofs of these buildings appear to have been important ritual and public spaces, cf. Graham 1979. 44 T. Özgüç 1999: 82, and cf. Kulakoğlu 2017 for a discussion of the natural topography dictating the shape of the earlier public buildings on the mound. 45 Cf. e.g. the overview in Mielke 2011 with good examples at Büyükkale at Boğazköy (gate buildings leading into colonnaded courts) and Maşat Höyük (colonnaded courtyard). Note also the doorless basements in buildings E, F, H and K at Boğazköy (cf. Neve 1982: 92–102) and the southern wing of Building C at Kuşaklı (discussed below) that are similar to what is preserved at Kültepe. A close parallel in time, but more remote in distance, is the so-called “Burnt Palace” of Level V at Beycesultan (cf. the perspective reconstruction in Lloyd and Mellaart 1965: 30), which had no basements, but a colonnaded main courtyard and an adjacent grand staircase leading up to the first-floor apartments.

26

Barjamovic

preserved to a great height.46 Based on this observation, one may suggest that the short walls protruding on the western side of the courtyard in Hrozný’s plan mark the position of staircases leading to the upper chambers, as shown on Fig. 1.3. The surrounding halls mentioned above may in turn have supported a colonnade or peristyle on the floor above. Multistoried palatial structures are known from the excavations of the contemporary palace of Sarıkaya at Acemhöyük, which also revealed the remains of a frontal colonnade or portico in wood. Aside from postholes and carbonized remains of the timber itself, an elaborate architectural rendering found on a bathtub in the building supports the reconstruction of a colonnade. This image (Fig. 1.4) shows stacked porches on the exterior of a building with superposed slender wooden posts and large flat capitals, the lower post being taller than the striped upper one. Details include openwork railings used as parapets on the balconies, and stairs in two locations. Walls are indicated by solid black, checkerboard, or crisscross linear patterns. The image on the bathtub has been associated convincingly by Nimet Özgüç with the façade of Sarıkaya itself,47 and while one may speculate on the identity of specific details in the rendering of the architecture, the main thrust of the illustration is clear: we are looking at a building which presents an open front with two stories of balconies over a socle.48 At Sarıkaya, the portico formed part of the palace façade, as in later western examples, while it may have encircled the courtyard of the Warshama Palace in anticipation of later Hittite architecture. In both cases, the extensive use of wood in the buildings led to such high temperatures during the conflagrations that destroyed them that it completely vitrified the surrounding mud-brick and made Chantre believe that he was digging in a volcano. Interpretations such as these must necessarily remain tentative, but at least it seems possible to demonstrate that the two datasets published by Hrozný 46

47 48

An unlikely alternative would be that Hrozný obliterated all remains of the throne room during the course of his excavations. Adamčík (2013: 13) reports: “Among the last finds made at [the mound of] Kültepe … was a fragment of the side of a throne belonging to the Hittite ruler.” What object he refers to is unclear, and there is no source given for his remark. Hrozný (1927: 5) mentions no such object in a similar list. If one were nevertheless to accept the reference, and if the object in question were in fact to date to the Middle Bronze Age palace (and not some later period, perhaps paired with the relief found by Hrozný in the same building), then this could be taken as evidence that Hrozný dug away the remains of the main audience chamber of the ruler. This does not solve the problem of the missing doors in the remainder of the building. Cf. N. Özgüç 1966 and see also T. Özgüç 1998: 468 and 2003: 128–129. A detailed discussion of the piece appears in Mellink 1993 and is quoted extensively here. Wooden porticos of the same type appear in traditional Turkish architecture, cf. e.g. the examples shown in N. Özgüç 2015: 204.

Resurrecting a Bronze Age Palace at Kültepe

27

Figure 1.4 Decorated bathtub from Acemhöyük (cf. N. Özgüç 1966 and Mellink 1993) probably showing the double portico northern façade of the Sarıkaya Palace, a near-contemporary of the Warshama Palace at Kültepe.

and Tahsin Özgüç are to some extent compatible. This was clearly the case in Hertel’s study of the Lower Town,49 and seems also to apply to the much more disturbed layers excavated by them on the mound of Kültepe. A comparison between the two allows us to draw new conclusions and raise new questions. Section drawings for the palace have not yet been published, but were produced by Özgüç and his team, and are being worked on by the current excavators.50 There are also smaller areas in and around the building that were intentionally left intact by Emre for future archaeologists to work on. Finally, one might feasibly re-excavate parts of Hrozný’s old trench to see what,

49 50

Hertel 2014. I am grateful to Professors T. Özgüç and F. Kulakoğlu for sharing this information. An important goal of the ongoing excavations at Kültepe is to link Middle Bronze Age (palatial) strata to Early Bronze Age predecessors, and to obtain a better understanding of the mound stratigraphy by digging a small area where the crucial transitional layers appear to have been left undisturbed by later activities. For recent developments, cf. Ezer 2015 and Kulakoğlu 2017.

28

Barjamovic

if anything, was left behind.51 Several of these undertakings are scheduled or have already been launched by the new director of excavations at Kültepe, and there is real hope now that a number of the issues raised in the course of this discussion will become clear in the decades to come. More can probably also be done with the notebooks, photographs and plans from Hrozný’s excavations, if they do in fact survive—perhaps in a joint effort between local institutions and scholars working on Old Assyrian Kültepe. One might use the upcoming centennial of his excavations as the occasion to produce some semblance of a final report of Hrozný’s work at Kültepe in 1925. One should emphasize that it is all too easy (and sometimes even tempting) to criticize earlier colleagues with the benefit of hindsight. Hopefully, I have shown why this is not a productive approach. Unique constraints, changing research agendas, and often entirely down-to-earth priorities drove all three scholars discussed above in their work—Chantre, Hrozný and Tahsin Özgüç. Instead, a reexamination and recombination of their data with current knowledge can lead to new insights about the site of Kültepe and its main palace. Acknowledgments I am grateful to Jeffrey Blossom of the Harvard University Center for ­Geographic Analysis for producing figures 1.1–1.3. I thank Aslıhan Yener and Mogens Trolle Larsen for reading and commenting on a draft of this text. I am grateful to the Yale University Library Manuscripts and Archives for providing access to the Albrecht Goetze Papers. I thank the Yale Babylonian collection for permission to cite the correspondence between Clay, Hrdlička and Hrozný. Bibliography Adamčík, T. (2013) Československé vykopávky na Kültepe, Acheron: Acta Historica et Orientalia Neosoliensia 3/5–6, 4–16. Arık, R.O. (1944) 1942 de Türk Tarih Kurumu adına yapılan: Bitik kazısı ve Hatay tetkikleri hakkında kısa rapor, Belleten viii/30, 341–384. Barjamovic, G. (2015a) Kültepe after Kaneš, in F. Kulakoğlu and C. Michel (eds.), Proceedings of the 1st Kültepe International Meeting, Kültepe, 19–23 September 2013. Studies Dedicated to Kutlu Emre [Kültepe International Meetings 1, Subartu 35], Brepols: Turnhout, 233–242. 51

T. Özgüç (1999: 74) states that he reinvestigated the remains found by Hrozný, as far as they were preserved, but that very little remained after the futile attempt by the Czechoslovakian to find tablets in deeper levels.

Resurrecting a Bronze Age Palace at Kültepe

29

Barjamovic, G. (2015b) Contextualizing tradition: incantations, writing and domestic life in Old Assyrian Kanesh, in: P. Delnero and J. Lauinger (eds.), Texts and Contexts: Approaches to Textual Transmission in the Cuneiform World [Studies in Near Eastern Records 9], Berlin, Boston: de Gruyter, 48–86. Barjamovic, G. (forthcoming). Anatolia in the Middle Bronze Age, in: N. Moeller, D. Potts and K. Radner (eds.), The Oxford History of the Ancient Near East, Oxford: Oxford University Press. Chantre, B. (1896–1898) Le tour du monde. En Asie Mineure: Souvenirs de voyages en Cappadoce et en Cilicie, Paris: Hachette. Chantre, E. (1898) Mission en Cappadoce: 1893–1894, Paris: Ernest Leroux. Crowfoot, J.W. (1899) Review of E. Chantre 1898, The Journal of the Anthropological Institute of Great Britain and Ireland 29/3–4, 320–324. Der Matossian, B. (2013) Ottoman Armenian Kesaria/Kayseri in the nineteenth century, in: R.G. Hovannisian (ed.), Armenian Kesaria/Kayseri and Cappadocia, Costa Mesa, CA: Mazda, 187–230. Dercksen, J.G. (2001–2002) Review of K. Hecker, G. Kryszat, L. Matouš, Kappadokische Keilschrifttafeln aus den Sammlungen der Karlsuniversität Prag, Archiv für Orientforschungen 48/49, 183–196. Dercksen, J.G. (2004) Some elements of old Anatolian society in Kaniš, in: J.G. Dercksen (ed.), Assyria and Beyond: Studies Presented to Mogens Trolle Larsen [PIHANS 100], Leiden: Nederlands Intituut voor het Nabije Oosten, 137–177. Dercksen, J.G. (2005) Again on Old Assyrian tuzinnum, Nouvelles Assyriologiques Brèves et Utilitaires 2005/2, 39 (no. 38). Dercksen, J.G. (2008) Observations on land use and agriculture in Kanes, in: C. Michel (ed.), Old Assyrian Studies in Memory of Paul Garelli [PIHANS 112], Leiden: Nederlands Instituut voor het Nabije Oosten, 139–157. Erol, H. (2007) Eski Asurca metinlerde meslek adları ve unvanlarla geçen şahıs isimleri. MA thesis, University of Ankara. Ezer, S. (2015) Kültepe-Kaneš in the Early Bronze Age, in: L. Atici, F. Kulakoğlu, G. Barjamovic and A. Fairbairn (eds.), Current Research at Kültepe-Kanesh: An Interdisciplinary and Integrative Approach to Trade Networks, Internationalism, and Identity [JCS Supplemental Series 4], Atlanta: Lockwood, 5–23. Graham, J.W. (1962) The Palaces of Crete, Princeton: Princeton University Press. Graham, J.W. (1979) Further notes on Minoan palace architecture: 1. West magazines and upper halls at Knossos and Mallia, AJA 83/1, 49–63. Grigg, S. (1978) Albrecht Goetze Papers, Manuscripts and Archives, New Haven: Yale University Library. Hertel, T. (2014) The lower town at Kültepe/Kanesh: urban layout and population, in L. Atici, F. Kulakoğlu, G. Barjamovic and A. Fairbairn (eds.), Current Research at Kültepe/Kanesh: An Interdisciplinary and Integrative Approach to Trade Networks, Internationalism, and Identity [JCS Supplemental Series 4], Atlanta: Lockwood, 25–54.

30

Barjamovic

Hrozný, F. (1927) Rapport préliminaire sur les fouilles tchécoslovaques du Kultépé, Syria 8, 1–12. Hrozný, F. (1952) Inscriptions cunéiformes de Kültepe 1, Praha: Státní Pedagogické Nakladatelství. Kopecký, R. (2011) Bedřich Hrozný: život a dílo, Brandýs and Labem: Dar Ibn Rushd. Kulakoğlu, F. (2017) Early Bronze Age monumental structures at Kültepe, in: F. Kulakoğlu and G. Barjamovic (eds.), Proceedings of the 2nd Kültepe International Meeting, Kültepe, 26–30 July 2015. Studies Dedicated to Klaas Veenhof [Kültepe International Meetings 2, Subartu 39], Brepols: Turnhout, 217–226. Larsen, M.T. (1982) Your money or your life! A portrait of an Assyrian businessman, in: M.A. Dandamayev, I. Gerschevitch, H. Klengel, G. Komoróczy, M.T. Larsen and J.N. Postgate (eds.), Societies and Languages of the Ancient Near East. Studies in Honor of I.M. Diakonoff, Warminster: Aris and Phillips, 214–245. Larsen, M.T. (2015) Ancient Kanesh: A Merchant Colony in Bronze Age Anatolia, Cambridge: Cambridge University Press. Lloyd, S. and Mellaart, J. (1965) Beycesultan, Volume ii: Middle Bronze Age Architecture and Pottery [BIAA Occasional Monograph Series 8], London: British Institute at Ankara. Manning, S.W., Griggs, C.B., Lorentzen, B., Barjamovic, G., Ramsey, C.B., Kromer, B. and Wild, E.M. (2016) Integrated tree ring-radiocarbon high-resolution timeframe for First Babylonian Dynasty and Old Assyrian history, PLoS ONE 11/7, e0157144. DOI:10.1371/journal.pone.0157144 (accessed 10.10.2017). Manning, S.W., Barjamovic, G. and Lorentzen, B. (2017) The course of 14C dating does not run smooth: Tree-rings, radiocarbon, and potential impacts of a calibration curve wiggle on dating Mesopotamian chronology, Journal of Ancient Egyptian Interconnections 13, 70–81. Matouš, L. (1949) Bedřich Hrozný. The Life and Work of a Czech Oriental Scholar, Prague: Orbis. Mellink, M.J. (1993) Aspects of minor and major arts in Kanesh and Acemhöyük, in: M.J. Mellink, E. Porada and T. Özgüç (eds.), Aspects of Art and Iconography: Anatolia and its Neighbors. Studies in Honor of Nimet Özgüç, Ankara: Türk Tarih Kurumu Basımevi, 423–433 and pl. 74. Michel, C. (2015) L’organisation du palais de Kaneš d’après la documentation textuelle, in: C. Michel (ed.), Séminaire d’Histoire et d’Archéologie des Mondes Orientaux (SHAMO), 2012–2013. De la maison à la ville dans l’Orient ancien: bâtiments publics et lieux de pouvoir, Nanterre: CNRS, 161–174. Michel, C. (2019) Palaces at Kaneš during the Old Assyrian period, in: D. Wicke (ed.), Der Palast im antiken und islamischen Orient: 9. Internationales Colloquium der Deutschen Orient-Gesellschaft 30. März – 1. April 2016, Frankfurt am Main [CDOG 9], Wiesbaden: Harrassowitz, 121–138.

Resurrecting a Bronze Age Palace at Kültepe

31

Mielke, D.P. (2011) Hittite cities: looking for a concept, in: H. Genz and D.P. Mielke (eds.), Insights into Hittite History and Archaeology [Colloquia Antiqua 2], Leuven: Peeters, 153–194. Neve, P. (1982) Büyükkale. Die Bauwerke. Grabungen 1954–1966 [Boğazköy-Ḫattuša xii], Berlin: Gebr. Mann. Osten, H.H. von der, Bittel, K. and McEwan, C.W. (1933) Katalog. Çanak çömlekler ve kaplar. Ankara müzesine Kayseri cıvarında kâin Kültepeden getirelen yeni eserler – Oeuvres nouvelles au musée d’Ankara provenant de Kultépé, prés Kayseri – New objects brought from Kültepe, near Kayseri, to the Ankara Museum, Türk tarih, arkeologya ve etnografya dergisi 1, 64–94 (French résumé 181–182; English résumé 190). Özgüç, N. (1959) Seals from Kültepe, Anatolia (Anadolu) 4, 43–53. Özgüç, N. (1966) Excavations at Acemhöyük, Anatolia (Anadolu) 10, 29–52. Özgüç, N. (2001) Notes on the bronze vehicle from the Sarıkaya Palace at Acemhöyük, in: W.H. van Soldt et al. (eds.), Veenhof Anniversary Volume: Studies Presented to Klaas R. Veenhof on the Occasion of his Sixty-Fifth Birthday [PIHANS 89], Leiden: Nederlands Instituut voor het Nabije Oosten, 361–366. Özgüç, N. (2015) Acemhöyük – Burušhaddum I. Cylinder Seals and Bullae with Cylinder Seal Impressions [Türk Tarih Kurumu Yayınları V/7], Ankara: Türk Tarih Kurumu. Özgüç, T. (1999) The Palaces and Temples of Kültepe-Kaniš/Neša [Türk Tarih Kurumu Yayınları V/46], Ankara: Türk Tarih Kurumu. Özgüç, T. (2003) Kültepe-Kaniš/Neša: The Earliest International Trade Center and the Oldest Capital City of the Hittites, Tokyo: Middle Eastern Culture Center in Japan. Robert, L. (1963) Noms indigènes dans I’Asie Mineure gréco-romaine I [Bibl. arch. et hist. de l’Institut français d’archéologie d’Istanbul 13], Paris: Maisonneuve. von Rüden, C. (2014) Beyond the East-West dichotomy in Syrian and Levantine wall paintings, in: B.A. Brown and M.H. Feldman (eds.), Critical Approaches to Ancient Near Eastern Art. Boston, Berlin: De Gruyter, 55–78. Shi, X. (2015) Village life in Ancient Anatolia: The case of Talwahšušara, in: F. Kulakoğlu and C. Michel (eds.), Proceedings of the 1st Kültepe International Meeting, Kültepe 19–23 September, 2013. Studies Dedicated to Kutlu Emre [Kültepe International Meetings 1, Subartu 35], Brepols: Turnhout, 147–154. Smith, M.L. (1973) Ionian Vision: Greece in Asia Minor 1919–1922, London: Hurst and Company. Veenhof, K.R. and Eidem, J. (2008) Mesopotamia: The Old Assyrian Period [Orbis Biblicus et Orientalis 160/5], Fribourg: Academic Press. Winckler, H. (1906) Die im Sommer 1906 in Kleinasien ausgeführten Ausgrabungen, Orientalistische Litteratur-Zeitung 9/12, 621–634. Woolley, L. (1946) Syria as the gateway between East and West, The Geographical Journal 107/5–6, 179–190.

Chapter 2

Hrozný’s Excavations, 1924–1925: Sheikh Sa’ad, Tell Erfad Jan Bouzek Abstract This paper contains a report on B. Hrozný’s excavations in Syria and reviews the reports on the surveys of the sites in the 1990s and 2008, as well as publications of finds kept in the Czech collections. Of particular importance are Bronze Age and Early Iron Age terracottas, Hellenistic and Roman pottery, lamps and glass from Tell Erfad, and a Roman sculpture, inscriptions and part of a Neo-Hittite orthostat from Sheik Sa’ad.

Keywords Bedřich Hrozný – Syria – Sheikh Sa’ad – Tell Erfad – Bronze Age – Early Iron Age – Neo-Hittite – Roman sculpture – inscriptions

Following his groundbreaking identification of the Hittite language, Bedřich Hrozný wanted to excavate his favourite Hittites, but he had to wait for the permission of the Turkish government, even though he already had the money from the Czechoslovak government. The then newly founded Czechoslovakia had very good relations with France, and as he was by then a renowned scholar, he was offered by the French mandate government the chance to excavate in Syria. From the sites offered to him, he chose two: Sheikh Sa’ad with a ruined 4th century temple and a visible terrace of earlier date in the Hauran area (likely because of a reported earlier find of a “Hittite” statue of a lion), and a tell situated northwest of Aleppo, Tell Erfad (or Rifad). 1

Sheikh Sa’ad

At Sheikh Sa’ad, the excavations started on April 4, 1924 and continued until May 17, but several of Hrozný’s sketches of uncovered ruins in his diary date

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_004

Hrozný’s Excavations, 1924–1925

33

from the first days of June, so he must have remained there for two or three weeks after the end of the excavation. We have not yet discovered the plans made by his architect (perhaps they are lost), only the rough sketches by Hrozný himself. The main excavations were conducted below the then standing mosque, around the now largely collapsed sanctuary: at the southern face of the structure, and along the so-called Lion Wall. The lion was considered by Hrozný as being Neo-Hittite, as is also the case for the associated orthostat (Bouzek et al. 1999: 12, SS 13, Plate 13). The digs descended up to 3 m below the original surface, and Hrozný led his trenches mainly along the walls, in order to clean the ruins according to the then usual strategy (dégagement des ruines). Some important results from his digs are the stone sculpture, partly in the ­National Archaeological Museum in Damascus and partly in Prague, the inscriptions (nearly all in the Damascus museum), and the photographic documentation of the temple before its collapse in the 1930s. The main finds as preserved in drawings and photographs have already been published in csir, Czech Republic i (Bouzek et al. 1999; cf. also Bouzek 1990: 87–92).1 Being a philologist and epigraphist, Hrozný was much more interested in inscriptions and other written documents than in anything else, but he described all he could see, and his notebooks contain sketches of situations and of important finds, richly documented in his photographs. Except for several brief preliminary reports in Czech (Hrozný 1924) and one in English (Hrozný 1926), Hrozný himself did not publish any more substantial report during his lifetime, but his photographs and diaries were preserved and most of the objects acquired by him as his share (the usual practice at that time) have already been published in several places (Hrozný 1924). In 2008, a Czech team from the Faculty of Arts of Charles University conducted a brief survey at Sheikh Sa’ad. At the site, we took photographs of the ruins and their immediate surroundings, notably of the remaining visible part of the terrace (the Wall of the Lion) and the walls uncovered by Prof. Hrozný, as far as they are still visible (Figs. 2.3, 2.4, 2.5). We tried to compare the present situation with that of the 1920s as reproduced in Hrozný’s photographs (Figs. 2.1–2.2) and sketch drawings. Beside Jiří Musil and the present author, the third member of the expedition was Pavel Beran, a student of Arabic language at the same faculty, who acted as translator, and the fourth was Marie Dufková from the National Museum of Prague, where the photographic documentation of Hrozný’s excavations at Sheikh Sa’ad is now stored (cf. Bouzek et al. 2008). 1 The objects and statutes are kept in the collections of the National Gallery in Prague and are currently exhibited in the Polabské Museum – Museum of Bedřich Hrozný in his birthplace of Lysá nad Labem; others are in the National Archaeological Museum in Damascus.

34

Bouzek

Figure 2.1 Photo of the mosque at Sheikh Sa’ad 1924. Photo by bedřich hrozný (National Museum – Náprstek Museum of Asian, African and American Cultures, Inv. No. N 87)

Figure 2.2 Photo of the mosque at Sheikh Sa’ad 1924, with the ruins of the “Sanctuary”. Photo by bedřich hrozný (National Museum – Náprstek Museum of Asian, African and American Cultures, Inv. No. N 180)

Hrozný’s Excavations, 1924–1925

35

Figure 2.3 Sheikh Sa’ad 2008, showing the decay of the “Sanctuary” building. Sideview of Fig. 2.4. Photo by jan bouzek

Figure 2.4 Sheikh Sa’ad 2008, showing the decay of the “Sanctuary” building. The same view angle as in Fig. 2.2 by Hrozný. Photo by jan bouzek

36

Bouzek

Figure 2.5 Sheikh Sa’ad 2008, showing the decay of the “Sanctuary” building. Detail of the remains in Fig. 2.4, originally hidden at the back of the structure in Fig. 2.2 by Hrozný. Photo by jan bouzek

The site of Sheikh Sa’ad is now surrounded by an enlarged village, with several houses situated close to the ruins. A large part of the nave of the Early Christian temple has collapsed (according to the villagers, as early as the 1930s), but we were able to compare the present state with the old photographs taken by Hrozný, which gives some clues to enable us to interpret what he was still able to see. The underground spaces, one with (now empty) loculi and the other with vaults (in one case decorated with floral ornaments), make it clear that this building was a modest parallel to the temple of Kanawat (cf. esp. Amer et al. 1982). The architectonic members decorated with floral and similar reliefs in the Hrozný collection (Bouzek et al. 1999: Plates 15 A–C, 16–18 and very probably also Plate 14) belonged either to the decoration of the temple, or to monuments contemporary with its construction and use (late 4th–6th century ad; cf. esp. Dentzer-Feydy 1985–1986, Dentzer-Feydy 1989). The statues of Victory (Bouzek et al. 1999: Plates 1–6, 24: 3)2 may well date from the 3rd century, the “Golden Age” of the Hauran part of the province of 2 The last item is exhibited in the Damascus museum.

Hrozný’s Excavations, 1924–1925

37

Arabia, especially under the emperors Alexander Severus (222–235) and Philippus Arabus (244–249). The funeral busts (Bouzek et al. 1999: Plates 9–11) date from roughly the same period, as do other pre-Christian statues of roughly the same style: the torso of Eros (Bouzek et al. 1999: Plate 7 A–C) and the fragments of female figures in high relief (Bouzek et al. 1999: 11, SS 8, Plates 8 A–B and 12, SS 12, 12 A–B), though the exact chronology of Hauran sculpture is not yet known in the same detail as that of Palmyra (cf. esp. Boleli 1985– 1986: 311–372; Abel 1956–1957: 1–13; Dentzer and Dentzer-Feydy 1991). The base with the remains of a foot and letters (Bouzek et al. 1999: Plate 19 A–B) is roughly of the same date, while other inscriptions sketched by Hrozný at Sheikh Sa’ad and kept in the National Archaeological Museum in Damascus date from the Hellenistic, Roman and later periods. They are written in the Greek, Aramaic, Phoenician (?) and Hebrew languages (Bouzek et al. 1999: 17–19, Figures 3–7).3 A mighty structure below the Roman temple construction was termed by Hrozný the “Terrace au Lion”, and he may well be right in connecting this structure with the formerly discovered “(neo-)Hittite” lion since its lower part, reached by Hrozný’s digs, was indeed of earlier date. Since the discovery of the new Neo-Hittite finds on the citadel of Aleppo (Khayyata and Kolhmayer 1998: 69–95, esp. 89–95, Plates 21–24), the known monuments of Neo-Hittite art have come closer to the art of southern Syria, and the fragmentary relief in the National Gallery of Prague (Bouzek et al. 1999: 12, SS 13, Plate 13) may thus well have been originally part of a Neo-Hittite orthostat. The new dating of later Neo-Hittite sculpture allows one to accept a continuous development from the Bronze to the Iron Age (Orthmann 2002: 153–157 with further bibliography).4 2

Tell Erfad

The excavations at Tell Erfad were different. No visible architecture was on the surface, and here Hrozný laid his trenches to find out what lay underneath. He tried to excavate some sections of the city wall, too, but he postponed this task, and in the end never really started. Six trenches were thus opened on the

3 The Greek inscriptions have been treated by Radislav Hošek, the Aramaic, Phoenician and Hebrew by Stanislav Segert and Yona Sabar. 4 On Sheikh Sa’ad, see now also Charvát 2015 and Meynersen 2015.

38

Bouzek

southern slope of the tell. On October 17 he started with trench South ii, and on November 20 with trench South vi. The planned trenches South iii, iv and v were finally not opened at all. On the western slope, he excavated during both seasons in the sounding “West” directed E–W. On the northwest slope, Hrozný opened Trench North i on October 21, and North ii on October 24. The peak number of workmen was reached on October 31; on this day he started a trench towards the east from the top of the hill. Later Hrozný wanted to follow the city wall, but his digs were interrupted and he did not have enough time to realize the enlargement of his project. On the northern slope, he opened two trenches as late as April 21, 1925. The finds from Tell Erfad from Bedřich Hrozný’s excavations in 1924 and 1925 (Hrozný 1926) were obtained by him from the then Syrian authorities as his share of what he excavated on the tell; the more important objects were given to the National Museum of Damascus (Nováková 1971: 15–18). Hrozný’s collection was first kept in the Oriental Seminar of Charles University, but later was mostly ceded to the Oriental Department of the National Gallery. His notebooks remained in the Charles University Institute for Ancient Near Eastern Studies, while most of the negatives went to the Náprstek Museum, also in Prague. A large part of the collection was destroyed during a fire in 1969 in the castle of Benešov nad Ploučnicí, where it was temporarily kept, and many objects bear the traces of this secondary burning. A smaller part, which was lent to the Bedřich Hrozný Museum in Lysá nad Labem, survived much better. Today, the most important finds from Hrozný’s digs at Tell Erfad are the preserved small finds. It is not easy to reconstruct precisely the course of the excavations, but they lasted several weeks in both seasons, and, as the later British excavations have shown, the earliest levels were probably not reached in most parts of the digs (Seton-Williams 1961; Seton-Williams 1967).5 Most of the material kept in Prague dates from the Assyrian, Graeco-Persian, Hellenistic and Early Roman periods, but earlier finds are also represented. Nea Nováková prepared in the late 1960s the publication of the terracottas from Tell Erfad (Nováková 1971), of which a brief preliminary report by J. Krušina-Černý appeared in 1958 (Krušina-Černý 1958). In the 1980s, after finishing the publication of Antonín Salač’s excavations in Kyme and Samothrace, we formed a group to prepare the Tell Erfad materials kept in the Oriental

5 Seton-Williams distinguished the Hellenistic and Roman Level i (4th c. bc–1st c. ad), the Persian Level i A (6th–5th c. bc), the Assyrian/Aramean Level ii (9th–7th c. bc), the late second millennium Level iii, and three earlier Levels iv–vi.

Hrozný’s Excavations, 1924–1925

39

Department of the National Gallery, but some members of the team had other projects or a different life after 1989, so the project was much delayed. Our ­excavations in Beirut in 1996, however, gave us a deeper insight into the problems of Lebanese and Syrian Hellenistic and Early Roman materials, and the project was renewed. The present author was able to prepare the section dealing with “Graeco-Persian”, Hellenistic and Early Roman pottery (Fig. 2.6) during several stays in Athens and in the library of the French Institute of Beirut, and he would like to express his gratitude for their hospitality both to the French School at Athens and the French Institute at Beirut. The classification of the three Rhodian amphora handles is by the late Virginia Grace (Fig. 2.7). Ladislav Boháč kindly contributed to this project with his study of Hellenistic and Early Roman lamps (Fig. 2.8) and glass from the collection (Boháč et al. 1997; Boháč and Bouzek 1999).6

Figure 2.6 Tell Erfad. Imported Hellenistic pottery: 1) Megarian Bowl fragment; 2) white painted “West Slope” fragment (after Boháč et al. 1997: Fig. 4)

6 On Tell Erfad, see also Nováková 2015.

40

Bouzek

Figure 2.7 Tell Erfad. Imported stamped amphora handles (after Boháč et al. 1997: Fig. 2)

3 Conclusion As a philologist and epigraphist, both in Sheikh Sa’ad and in Tell Erfad Hrozný was primarily interested in inscriptions and other written documents. Nevertheless, both excavations yielded important results; furthermore, he described all he could see, and his notebooks also contain valuable sketches of the sites and the important finds. Except for several brief preliminary reports in Czech and one in English, Hrozný himself did not publish any more substantial­

Hrozný’s Excavations, 1924–1925

41

Figure 2.8 Tell Erfad. Hellenistic and Early Roman lamps (after Boháč et al. 1997: Figs. 9, 11)

reports during his lifetime, but his photographs and diaries were preserved, and most of the objects acquired by him have now been published (cf. above and the bibliography). Bibliography By Bedřich Hrozný

Hrozný, B. (1924) První výzkumná výprava československá do Orientu (The first Czechoslovak research expedition to the Orient), Národní Listy 6.4.1924. Hrozný, B. (1924) Ve starozákonní zemi Basan (In the Old Testament land Basan), Národní Listy 24.8.1924. Hrozný, B. (1924) V Kaisarii (In Caesarea), Národní Listy 24.8.1924.

42

Bouzek

Hrozný, B. (1924) Z mých potulek po Sýrii (From my trips in Syria), Národní Listy 25.12.1924. Hrozný, B. (1926) The first Czechoslovak excavations in the Near East, Central European Observer 4/29, 511–512. Hrozný, B. (1926) The first Czechoslovak excavations in the Near East, Central European Observer 4/30, 527–529.

By others

Abel, N. (1956–1957) La statuaire hauranienne, une branche provinciale de l’art romain tardif, Annales de la Société Royale d’Archéologie de Bruxelles 49, 1–13. Amer, G., Biscop, J.L. and Dentzer-Feydy, J. (1982) L’ensemble basilical de Qanavat (­Syrie de Sud), Syria 59, 257–275. Boháč, L., Bouzek, J. and Grace, V. (1997) Hellenistic and Roman finds from Bedřich Hrozný’s excavations at Tell Erfad, Syria, Parts i–ii, Eirene 33, 122–157. Boháč, L. and Bouzek, J. (1999) Hellenistic and Roman finds from Bedřich Hrozný’s excavations at Tell Erfad, Syria, Part iii and Addenda, Eirene 35, 47–60. Boleli, G. (1985–1986) La sculpture à ronde bosse du caractère indigène en Syrie du Sud, in: J.-M. Dentzer (ed.), 311–372. Bouzek, J. (1990) La sculpture romaine des fouilles de Bedřich Hrozný à Sheikh Sa’ad, Études et Travaux (Varsovie) 15, 87–92. Bouzek, J., Dufková, M., Hošek, R., Ondřejová, I. and Segert S. (eds.) (1999) Corpus signorum imperii Romani, Czech Republic i: Roman sculpture from Syria and Asia Minor in Czech collections, Prague: Karolinum. Bouzek, J., Hošek, R. and Segert, S. (1999) Sheikh Sa’ad, in: J. Bouzek et al. (eds.), 7–19. Bouzek, J., Dufková, M. and Musil, J. (2008) Hrozný’s excavations, 1924 and visit of the site in 2008: Sheikh Sa’ad. A preliminary report on the project of publication of Hrozný’s digs, Studia Hercynia 12, 98–100. Charvát, P. (2015) The archaeological investigation by Bedřich Hrozný at Sheikh Sa’ad in Syria, in: Š. Velhartická (ed.), 150–160. Dentzer, J.-M. (ed.) (1985–1986) Hauran i, Recherches archéologiques sur la Syrie de Sud à l´époque hellénistique et romaine 1–2 [Bibliothèque archéologique et historique 124], Paris: Geuthner. Dentzer, J.-M. and Dentzer-Feydy, J. (1991) Le Djebel al-’Arab. Histoire et patrimoine du Musée de Suweidā‘ [Guides archéologiques de l’Institut Français d’Archéologie du Proche-Orient 1], Paris: ERC. Dentzer-Feydy, J. (1985–1986) Décor architectonique et développement du Hauran du ier s. av. J.-C. au viie s. après J.-C., in: J.-M. Dentzer (ed.), 261–310. Dentzer-Feydy, J. (1989) Le décor architectural en Syrie aux époques hellénistique et romaine. in: J.-M. Dentzer and W. Orthmann (eds.), Archéologie et histoire de la Syrie

Hrozný’s Excavations, 1924–1925

43

ii: La Syrie de l’époque achéménide à l’avènement de l’Islam [Schriften zur Vorderasiatischen Archäologie 1], Wiesbaden: Harrassowitz, 456–476. Khayyata, W. and Kolhmayer, W. (1998) Die Zitadelle von Aleppo. Vorläufiger Bericht über die Untersuchungen 1996 und 1997, Damaszener Mitteilungen 10, 69–95. Krušina-Černý, L.J. (1958) Notes on the terracotta figurines from Tell Erfad, in: J. Frel (ed.), Epitymbion Roman Haken, Prague: Sumptibus Societatis Archaeologicae Bohemoslovenicae, 109–112. Meynersen, F. (2015) “Im archäologischen Zukunftsland.” Bedřich Hrozný und die späthellenistisch-kaiserzeitlichen Bildwerke aus Sheikh Sa’ad in Südsyrien, in: Š. Velhartická (ed.), 161–180. Nováková, N. (1971) Le terres-cuites de Tell Erfad, Vol. i, Prague, 15–18. Nováková, N. (2015) The excavations of Bedřich Hrozný at Tell Erfad, in: Š. Velhartická (ed.), 181–198. Orthmann, W. (2002) Die Bildkunst im Übergang von der Großreichszeit zur späthethitischen Periode, in: E.A. Braun-Holzinger and H. Matthäus (eds.), Die nahöstlichen Kulturen und Griechenland an der Wende vom 2. zum 1. Jahrtausend v. Chr.: Kontinuität und Wandel von Strukturen und Mechanismen kultureller Interaktion. Kolloquium des Sonderforschungsbereiches 295 “Kulturelle und sprachliche Kontakte” der ­Johannes-Gutenberg-Universität Mainz, 11.–12. Dezember 1998, Möhnesee: Bibliopolis, 153–157. Seton-Williams, M.V. (1961) Preliminary report on the excavations at Tell Rifa’at, Iraq 23, 68–87. Seton-Williams, M.V. (1967) The excavations at Tell Rifa’at 1964: Second preliminary report, Iraq 29, 16–33. Velhartická, Š. (ed.) (2015) Bedřich Hrozný a 100 let chetitologie – Bedřich Hrozný and 100 Years of Hittitology, Praha: Národní Galerie v Praze.

Chapter 3

Hrozný and the Decipherment of Hieroglyphic Luwian J.D. Hawkins Abstract In contrast to his universally acclaimed identification of Hittite in 1915, Bedřich ­Hrozný’s efforts to decipher the Hieroglyphic Luwian script are widely considered to have had only limited success. However, a reappraisal of the history of decipherment demonstrates that Hrozný and his contemporaries Meriggi, Gelb, Forrer, and Bossert, at first independently but increasingly building upon each other’s results, had reached a broad consensus on the phonetic readings of most signs already by 1935. His work lay most notably in smaller details, such as the correct evaluation of single signs, as also acknowledged by his fellow decipherers. Together with his careful preparation of editions of Hieroglyphic texts, for which he travelled to Anatolia and Syria in 1934, these pioneering contributions thus establish Hrozný as one of the important figures in the early period of Hieroglyphic Luwian studies.

Keywords Bedřich Hrozný – Hieroglyphic Luwian – Hittite – cuneiform – history of research – Emirgazi

1 Prologue It was a great pleasure on this occasion and at this location to present this paper reassessing one aspect of B. Hrozný’s contribution to Hittite studies, namely his work on the decipherment of the “Hittite Hieroglyphs” (as the script was then known; now “Anatolian Hieroglyphs”). My wish is to correct an unfair judgement on his achievement in this field, which appears in the standard reference work, the Reallexikon der Assyriologie und vorderasiatischen Archäologie (iv/6–7, 1975), in the brief entry “Hrozný” by J. Klíma. He wrote:

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_005

Hrozný and the Decipherment of Hieroglyphic Luwian

45

Die weiteren Bemühungen Hrozný’s [sc. after “Die Lösung des hethitischen Problems”] um Entzifferung der hethitischen Hieroglyphen und der kretischen Schrift haben zwar zur Zusammenstellung des entsprechenden inschriftlichen Materials beigetragen, blieben aber sonst ohne Erfolg. An examination of Hrozný’s contribution – at least insofar as it relates to the Hittite Hieroglyphs – shows this judgement to be wide of the mark. Here we may also quote the assessment of J. Friedrich (1957: 97): Meriggi, Gelb, Forrer and Bossert were the ones who brought the whole undertaking on to more solid soil about 1930 … [Friedrich goes on to cite the single main contribution of each] It is difficult to appraise the independent contribution of Hrozný, already known to the reader as the successful first investigator of cuneiform Hittite, who also began to work on hieroglyphic Hittite in 1932. One may ask why this appraisal was difficult? It may be thus useful here to summarise briefly the situation as regards the Hieroglyphic inscriptions from their discovery and early attempts at decipherment in the period 1870–1920. For more details, see Hawkins 2000: 6–9. In the middle part of the 19th century, rock reliefs from across Anatolia were beginning to be reported to European scholars. These sculptures were often accompanied by inscriptions in a Hieroglyphic script, usually much eroded. From 1870 onwards, dressed stone monuments showing the same script were beginning to be collected in the Imperial Ottoman Museum in Constantinople, and to be published in academic journals and monographs in Europe. They came from such centres as Hama (central Syria), the Tyanitis and Maraş (southern Turkey) and the site identified as Karkamiš (on the Euphrates). By 1876 A. Sayce was able to designate these as “Hittite”, combining biblical references to a people, the ḥtym, with the toponyms found in then recently deciphered Egyptian and Assyrian records (Egyptian ḫt’, then transcribed “Kheta”, Assyrian Hatti). In 1900 Messerschmidt published his Corpus Inscriptionum Hettiticarum (with supplements 1902 and 1906), which collected over 30 stone inscriptions and a number of inscribed seals (mostly surviving as impressions on clay). Meanwhile the progress of discoveries led inexorably to the central Anatolian site at the village of Boğazköy, already noted for its massive ruins and its rock reliefs, and known from c. 1890 also as a source of Cuneiform tablets.

46

Hawkins

­German excavations here, 1906–1912, garnered among other discoveries a massive haul of clay tablets, which turned out to be the royal archives of the capital city of Hatti, written in Babylonian Cuneiform in the Akkadian language and the then unknown language of the kingdom. After the premature death of the excavator, the philologist H. Winckler, the Deutsche Orient-Gesellschaft, faced with responsibility of publishing the tablets, commissioned several scholars to undertake the task, among them Hrozný, then lecturer and professor at the University of Vienna. He worked on the tablets in the Ottoman Museum from April 1914 until the outbreak of the First World War forced him to return to his post in Vienna. He had copied a number of tablets (fig. 3.1), which were later published as Keilschrifttexte aus Boghazköi v and vi (1921), and had permission to edit the tablets on which he had been working. His interest concentrated on the main language of Hatti (Hittite to us), readable (in Cuneiform) but otherwise unknown, and in 1915 he was able to publish in preliminary form his demonstration of its Indo-European relationship and the resulting readings of text passages (mdog 56, 17–50). This was followed by a detailed treatment in his Die Sprache der Hethiter (BoSt 1–2 [1916–1917]). The reading, publication and academic edition of the Hittite texts made massive progress during the 1920s: publication in the series Keilschrifttexte aus Boghazköi i–vi (1916–1923) and Keilschrifturkunden aus Boghazköi i–xxv (1921–1930), as well as editions in the series BoSt 3–10 (1919–1924) and MVAeG (1925–1933). By this time the vast access of Hittitological material opened the door to a renewed effort on the decipherment of the Hieroglyphs, where lack of progress had apparently led to a dead end. It is true that Sayce’s recognition of the Hieroglyphic equivalents of the Cuneiform logograms “king” and “country” on the digraphic “Tarkondemos” seal had permitted the identification of the toponyms Hamath, Tuwana, Gurgum (modern Maraş) and Karkamiš on the Hieroglyphic inscriptions from these centres. This could have led (as it later did) to the progressive allocation of syllabic values to the signs writing the toponyms, but while several correct readings were proposed, they were accompanied by too many erroneous ones which stalled the whole process. But now that Boğazköy and the reading of Hittite had provided so much background information, the decipherment could be restarted on a much-expanded basis. 2

Decipherment: the First Successful Phase

During the 1930s five scholars working initially independently but gradually combining to debate, correct and build on each other’s efforts resumed the task of decipherment: they were P. Meriggi (1930), I.J. Gelb (1931), E. Forrer

Hrozný and the Decipherment of Hieroglyphic Luwian

47

Figure 3.1 Hrozný’s copy of the Hittite Laws, §§xxxiv–xxxix (KBo vi 4 iv 1–35): reproduced from Bossert, Altanatolien (Abb. 727)

48

Hawkins

(1932), Th. Bossert (1932) and B. Hrozný (1933) – the dates in brackets refer to the earliest publication of each on the subject (for their photographs cf. Hawkins 2000, chli I/3 frontispiece). It will be seen that Hrozný was the latest entrant of the study, and indeed he acknowledges the contributions of all the others in his introductions (1933: 7–8), so he had their earliest proposals before him and drew on them. His aim as he sets it out was “de réviser les lectures jusqu’à présent proposées pour les principaux signes de cette écriture, d’eliminer les lectures fausses, de proposer des lectures nouvelles…” In other words, he intended to offer a detailed critique of the progress so far, which indeed he gives in his treatment of the principal phonetic signs (1933: ch. iv, 99–117), where he tabulates the sign forms, his own readings and cross-­ references to his discussion earlier in the text of the readings of others. It is presumably for this reason that J. Friedrich found it “difficult to appraise the independent contribution of Hrozný” (see above). He could not be credited, as the others were by Friedrich, with single major discoveries: his work lay in the smaller details, the correct evaluation of single signs, where he was indeed often right, as later acknowledged by his fellow decipherers. For example, Hrozný identified in the texts the words for “horn”, “dog” and “horse”, each written with the crucial sign L.448, to which he allocated on philological grounds the value ku, to give the words kurni, kuwani- and akuwi-. He was in principle right in his identifications (Hrozný 1934: 129 n. 3, 149 n. 2; 1937: 358), though the exact character of the consonant has remained uncertain: Hrozný proposed ku; Meriggi cu, Gelb su (1937: 19–20), Laroche sù? (1960: 231), supported by Mittelberger (1962: 278) and Hawkins (2000: 35–36); and Melchert zú (1987) on phonetic grounds (PIE *k̂ > Luwian z), which is now supported by some Cuneiform-Hieroglyphic correspondences and is generally accepted. It is clear that by 1935 the five decipherers, acting almost as an unofficial team, had reached a broad degree of consensus on the phonetic readings largely correct or nearly so, but with sufficient erroneous readings of important signs, especially i, ῑ / a, ā (actually zi, za / i, ia), to obscure the character of the underlying language (see fig. 3.2). At the same time, lack of evidence on the vocabulary rendered attempts at translations a matter of guess-work and generally premature. The decipherers made individual visits to Turkey to inspect for themselves the inscriptions, mainly in museums. In Hrozný’s case, he travelled in July–­ November 1934 from Istanbul and Ankara by Kayseri all the way to Aleppo. On this trip, in addition to photographs and copies of already published texts, he produced editiones principes of the inscriptions ERKILET 1 and 2, BOYBEYPINARI 1 and 2, and “CARCHEMISH i, ii, iii” (now KARKAMIŠ A24a2+3, ANKARA 1 and KARKAMIŠ A20a1). Here too he made the first usable texts of

Hrozný and the Decipherment of Hieroglyphic Luwian

Figure 3.2 Table of the decipherers’ syllabic values, 1930–1935: black = correct; green = almost correct; red = incorrect

49

50

Hawkins

the EMIRGAZI altars A, C and D. His work from this trip was published in 1937. Hrozný’s work on the Hieroglyphs was mostly presented initially as lectures, often to the Académie des Inscriptions et Belles Lettres in Paris and read there by René Dussaud, and then published in the journal Archiv Orientální. These were subsequently collected and republished in his Les inscriptions hittites hieroglyphiques (ihh, I [1933], ii [1934] and iii [1937], including the material of his 1934 Anatolian journey). This last made available good new photographs taken by him, and he also collated all the inscriptions which he saw and made his own hand copies. He did not have a naturally neat drawing hand and his copies may appear somewhat rough, yet they are generally quite accurate: we may compare these with his Cuneiform copies, of which the same may be said (see fig. 3.1). We may also compare Gelb’s more sophisticated technique in producing copies for his Hittite Hieroglyphic Monuments (1939: vii–viii), in which he drew the inscriptions in Indian ink on to photographs, which were then bleached out. Further, Hrozný in a remarkable display of the printer’s art had a Hieroglyphic type created for each inscription edited by him so that they could be presented in a linear arrangement. This was done by drawings from photographs done by Karel Dyrynk, technical director of the Czech Státní Tiskárna [State Printing Works of Securities], which were then carefully controlled by Hrozný himself. The accuracy of the representation of each sign is such that these type-set texts are still useful for reference (see below, fig. 3.5). As an example of Hrozný’s work, I present his treatment of the EMIRGAZI altar A text (EMIRGAZI 1, §§1–26 in my current numeration). These four pieces of a duplicate text offered a formidable problem to the would-be reader on account of the somewhat eroded state of the relief signs. Hrozný was the first to publish relatively clear photographs of the different sides of each monument. In support of my judgement of his work I give (1) my own photographs of altar A (fig. 3.3); (2) his hand copy of the altar A text set against my own, where I have marked in red the necessary corrections to his texts and highlighted the corresponding areas on my text (fig. 3.4); (3) a passage of his typeset composite text (§§7–12 in my numeration), also with necessary corrections in red and the occasional non-existent signs marked x (fig. 3.5). It will be seen that the bits needing correction both in his hand-copy and in the type-set text are all in areas which he has shaded to indicate damage and lack of definition. While his copy was clearly drawn free-hand, my own was traced on acetate wrapped around the stone and with the aid of a photographic lamp. We may, I think, agree in judging his EMIRGAZI treatment as a good first attempt on a

Hrozný and the Decipherment of Hieroglyphic Luwian

51

Figure 3.3 EMIRGAZI altar A: photographs D. Hawkins

very difficult text. Of his attempted translation, however, one must admit that, given the degree of reading and understanding Hieroglyhic then current, it was altogether premature. That Hrozný on his 1934 itinerary did not confine himself to museum work but ventured out into the somewhat primitive conditions of the Anatolian countryside is shown by his work on the rock inscriptions, TOPADA (ihh iii: 370–371, pl. liii) and SUVASA, illustrated by his own, very atmospheric photographs of the local villages, villagers and transport (buggies and carts), and we see Hrozný himself on a ladder working on the TOPADA inscription (fig. 3.6). He published good photographs of a cast in Beirut Museum, but also his usual type-set linear arrangement (the sign types as usual rendered with a high degree of accuracy, see ihh iii: 370–371). All the contents of Hrozný’s ihh discussed above presumably fall under Klíma’s heading “Zusammenstellung des entsprechenden inschriftlichen Materials”.

52

Hawkins

Figure 3.4 EMIRGAZI altar A: above, Hrozný’s copy (ihh iii: pl. lxxvii); below, Hawkins’ copy. Corrections to Hrozný’s text in red, corresponding passages in Hawkins highlighted.

Above, I have characterized Hrozný’s translation of EMIRGAZI as “premature”, and the same would apply to most of his ihh translations. However, if we look at his efforts on some of the more straightforward inscriptions, especially those of Karkamiš, we may take a more positive view. By way of example I offer the beginning of his translations of KARKAMIŠ A2(+3), §§1–9, from ihh ii (1934), pp. 204–205. It is presented in his usual style: text in type-set linear arrangement (very faithful to the original), transliteration and translation (fig. 3.7). On the transliteration and translation, I have marked in red the emendations required by our present state of knowledge. Although these may appear substantial, they mostly reflect the progress made by those notable advances, the discovery of the KARATEPE bilingual, the “New Readings”, and the identification of the negative (see below). The plateau of decipherment

Hrozný and the Decipherment of Hieroglyphic Luwian

53

Figure 3.5 Extract from Hrozný’s text of emirgazi 1, §§7–12 (ihh iii: 408–409) with corrections in red

reached by Hrozný at this date, largely shared with his fellow decipherers, can by no means be described as “ohne Erfolg”. We may compare the remarks of Meriggi (1934: 15–17) on the differences and convergences of the various

54

Hawkins

Figure 3.6 Hrozný at Topada (ihh iii: pl. LI (lower))

a­ ttempts. Already by 1934, the decipherers were taking note of each other’s published proposals in a manner which was developing into team-work, and so it continued through to their latest publications in this phase of the decipherment (Meriggi 1937; Hrozný 1937; Gelb 1942). 3 Epilogue It is perhaps appropriate to conclude this assessment of Hrozný’s contribution to the decipherment of Anatolian Hieroglyphs with a brief summary of progress since his time up to the present, which has been largely dependent on new discoveries bringing new evidence. 3.1 The KARATEPE Bilingual Discovered in 1946 by Th. Bossert and H. Çambel, and published in instalments 1948–1956 by the former (incomplete at the time of his death in 1960), this great Hieroglyphic-Phoenician bilingual of 75 clauses both confirmed much of the decipherment of the 1930s, and provided a massive increase of knowledge of the vocabulary by its translations (cf. Melchert 2003: pls. ii b–c). By the time of the initial publication, Hrozný’s health was already declining, leading to his death in 1952. His obituaries note that he maintained his academic interests

Figure 3.7 Hrozný’s edition of KARKAMIŠ A2, §§ 1–9 (ihh ii: 204–205) with emendations in red

Hrozný and the Decipherment of Hieroglyphic Luwian

55

56

Hawkins

Figure 3.8 Meskene digraphic seal impressions (top, cylinder; lower, two signets; bottom right, stamp): after Beyer 2001: pl. 1, Tablette 7

almost to the end, so he was doubtless aware of this momentous breakthrough, but as far as I am aware he left no record of his reaction to it. 3.2 The Digraphic Seals of Ras Shamra and Meskene The French excavations at Ras Shamra in the seasons 1953–1954 recovered the Archive Sud in courtyard V of the Palais Royal, containing state documents, letters and treaties, which illustrate the relationship of the kings of Ugarit with their overlords, the Hittite Great Kings of Hattusa and their viceroys, the kings of Karkamiš. These tablets bear many validating seal impressions of the kings and their officers, and these are often digraphic, giving the seal-holder’s name written in Cuneiform equivalents of Hieroglyphic signs. These confirmed many readings of the 1930s, corrected others and contributed new values. See Laroche 1956; Hawkins 2013. For illustrations see for example Nougayrol 1956: pls. xxiii and lxxx; Schaeffer 1956: figs. 36, 37, 79. Further French excavations at Meskene 1972–1976 recovered some 1.000 tablets (fig. 3.8) from the archives of the rulers of Emar, with a much larger ­number of digraphic seal impressions with many more Cuneiform-Hieroglyphic equivalents, greatly extending the repertoire of Hieroglyphic values. See Laroche 1981, 1982, 1983; Beyer 2001; Hawkins 2017.

Hrozný and the Decipherment of Hieroglyphic Luwian

57

3.3 The “New Readings” Curiously, neither KARATEPE nor the Ras Shamra digraphic seals provided unambiguous evidence on the most crucial Hieroglyphic signs, the two pairs    and    (L.376, 377 and L.209, 210) read since the 1930s as i, ī and a, ā. But new evidence appeared in 1969 with the publication of pithoi from the U ­ rartian site of Altıntepe by T. Özgüç (1969). On these the usual measures elsewhere written in Cuneiform were written in (Anatolian) Hieroglyphs, and the sign was found to have a Cuneiform correspondence si, thus to represent not a ­vowel but consonant+vowel (see fig. 3.9). This suggested that the signs formerly read as i, ī and a, ā should be re-evaluated as zi, za and i, ia (Hawkins, Morpurgo Davies and Neumann 1974). Much supporting evidence could be adduced; but most crucially the effect of these “new ­readings” was to bring the language of the Hieroglyphs into almost complete congruence with Luwian written in the Cuneiform script (fig. 3.10). Henceforth it was possible to refer to Hieroglyphic Luwian alongside Cuneiform Luwian, and this is now accepted by all serious scholars. Corroborating evidence was almost immediately forthcoming from a Meskene digraph (Laroche 1976, the name Amzahi; cf. fig. 3.11). 3.4 The Negatives Almost as an appendix to the “new readings”, new evidence became available showing that the sign L. 332 previously mistaken for a form of the relative ­pronoun L.329, had the value na and represented the factual negative “not” (Hawkins 1975). In fact, the KARATEPE bilingual should have established this, but by a cruel mischance both attestations of the sign occurred in (negative) relative clauses (§§ xxi, xxvi), so the proper equivalence Hieroglyphic NEGATIVE = Phoenician bl “not” was missed. This recognition of the factual negative of course reverses the sense of every clause in which it appears.

Figure 3.9 Altıntepe pithoi and their capacity measures: from Hawkins apud Bordreuil et al. 2013: 85

58

Hawkins

Figure 3.10 The “New Readings”

3.5 Corpus of Hieroglyphic Luwian Inscriptions (chli) In the fifty years after the discovery of KARATEPE, new Hieroglyphic inscriptions continued to be found and published on a regular basis, and over 30 major inscriptions and many minor ones and fragments became available. My chli volume I (published in the year 2000) included all the Iron Age inscriptions (1000–700 bc) then known as well as another group now recognized as “transitional” (1200–1000 bc). chli volume ii by H. Çambel under my editorship, published in 1999, gave the final publication of the KARATEPE bilingual. The publication and (re)edition of the Empire inscriptions, some 40 pieces, has had to wait for chli volume iii (part 1), which I am currently completing and hope to send to the printer soon. This also includes new Transitional and Iron Age inscriptions which have appeared since 2000, over 30 pieces, some of major length and importance (part 2). Further, I give a complete Syllabary (part 3), discussing each sign and the evidence for its reading. An Addenda and Corrigenda to chli i and ii takes into account new developments and

Hrozný and the Decipherment of Hieroglyphic Luwian

59

Figure 3.11 Some syllabic values of Hieroglyphs established by Meskene digraphs: taken from a transparency used in a lecture by Laroche 1976 and given to the author (after Hawkins 2013)

perceptions since 2000, and notes small and not so small errors in the transliterations and translations (part 4). It is very gratifying to see the interest in Luwian awakened by chli and the number of young scholars keen to take on this research into the future. It is clear that things will not stand still in Luwian studies, but new discoveries, new evidence and interpretations will continue apace. So, at this point it is good to remember our intellectual forebears, including Hrozný, who in the 1930s set us on the path which has led to where we are today.

60

Hawkins

4 Postscript At a Colloquium to commemorate the 100th anniversary of the decipherment of the Hittite language held in Istanbul Archaeological Museum 14–15 ­November 2015, Š. Velhartická read a paper “Bedřich Hrozný, the decipherer of the Hittite language”. This was published in M. Doğan-Alpaslan, A. Schachner and M. Alpaslan (ed.), The Discovery of an Anatolian Empire (Istanbul 2017). In it she reported recent discoveries of much archival material of Hrozný’s work and correspondence. At another conference, Beyond all Boundaries: Anatolia in the 1st Millennium B.C., held at Ascona, Switzerland, 17–22 June 1918, Velhartická reported the discovery of further material and correspondence, particularly between Hrozný and the other decipherers of the Anatolian hieroglyphs. It is clear that this material has the potential to illustrate in much fuller detail Hrozný’s own contribution to the decipherment. Bibliography Beyer, D. (2001) EMAR iv. Les sceaux [Orbis Biblicus et Orientalis 20], Göttingen: Vandenhoeck & Ruprecht. Bossert, H.T. (1932) Šantaš und Kupapa [MAOG vi/3], Leipzig: Harrassowitz. Forrer, E.O. (1932) Die hethitische Bilderschrift [SAOC 3], Chicago: University of Chicago Press. Friedrich, J. (1957) Extinct Languages, London: Peter Owen. Gelb, I.J. (1931, 1935, 1942) Hittite Hieroglyphs i, ii, iii [SAOC 2, 14, 21], Chicago: University of Chicago Press. Gelb, I.J. (1939) Hittite Hieroglyphic Monuments [OIP 45], Chicago: The University of Chicago Press. Hawkins, J.D. (1975) The negatives in Hieroglyphic Luwian, AnSt 25, 119–156. Hawkins, J.D. (2000) Corpus of Hieroglyphic Luwian Inscriptions, 3 vols. [UISK 8.1–8.3], Berlin, New York: de Gruyter. Hawkins, J.D. (2013) The digraphic seals of Ugarit. Emmanuel Laroche and the decipherment of Hieroglyphic Luwian, in: P. Bordreuil et al. (eds.), Les écritures mises au jour sur le site antique d’Ougarit, Paris: Académie des Inscriptions et Belles-Lettres. Hawkins, J.D. (2017) Laroche and the seals of Meskene-Emar, in: A. Mouton (ed.), Hittitology Today. Studies on Hittite and Neo-Hittite Anatolia in Honor of Emmanuel ­Laroche’s 100th Birthday. Istanbul, 21–22 novembre, 2014, Istanbul: Institut Français d’Études Anatoliennes Georges Dumézil, 247–266. Hawkins, J.D., Morpurgo Davies, A. and Neumann, G. (1974) Hittite Hieroglyphs and L­ uwian: New Evidence for the Connection [Nachrichten der Akademie der Wissenschaften

Hrozný and the Decipherment of Hieroglyphic Luwian

61

in Göttingen, 1. Philologisch-historische Klasse, 1973/6], Göttingen: Vandenhoeck & Ruprecht. Hrozný, F. (1915) Die Lösung des hethitischen Problems. Ein vorläufiger Bericht, MDOG 56, 17–50. Hrozný, F. (1916, 1917) Die Sprache der Hethiter [BoSt 1–2], Leipzig: Hinrichs. Hrozný, B. (1933, 1934, 1937) Les inscriptions hittites hiéroglyphiques. Essai de déchiffrement, Praha: Orientální Ústav. Laroche, E. (1956) Documents hiéroglyphiques hittites provenant du palais d’Ugarit, in: C.F.-A. Schaeffer (ed.) 1956, 97–160. Laroche, E. (1960) Les hiéroglyphes hittites, I: L’écriture, Paris: Éditions du Centre National de la Recherche Scientifique. Laroche, E. (1976) Lecture given at the xxiii R.A.I., Birmingham, 1976. Laroche, E. (1981) Les hiéroglyphes de Meskene-Emar et le style “syro-hittite”, Akkadica 22, 5–14. Laroche, E. (1982) Documents hittites et hourrites, in D. Beyer (ed.), Meskéné-Emar: dix ans de travaux, 1972–1982, Paris: Éditions Recherche sur les Civilisations. Laroche, E. (1983) Les hiéroglyphes hittites de Meskéné-Emar: un emprunt d’écriture, CRAIBL 127/1, 12–23. Melchert, H.C. (1987) PIE velars in Anatolian, in: C. Watkins (ed.), Studies in Memory of Warren Cowgill, Berlin, New York: de Gruyter, 182–204. Meriggi, P. (1930) Die hethitische Hieroglyphenschrift, ZA 5, 165–212. Meriggi, P. (1934) Die längsten Bauinschriften in hethitischen Hieroglyphen nebst Glossar zu sämtlichen Texten [MVAeG 39/1], Leipzig: Hinrichs. Meriggi, P. (1937) Listes des hiéroglyphes hittites, RHA iv/27, 69–114; iv/29, 157–200. Messerschmidt, L. (1900, 1902, 1906) Corpus Inscriptionum Hettiticarum [MAOG 4/5, 3/7, 11], Berlin: Peiser. Mittelberger, H. (1962) Review of Laroche 1960, Die Sprache 8, 276–286. Nougayrol, J. (1956) Le palais royal d’Ugarit, iv: Textes accadiens des archives sud [Mission de Ras Shamra IX], Paris: Imprimerie Nationale. Özgüç, T. (1969) Altıntepe ii. Mezarlar, depo binası ve fildişi eserler [Türk Tarih Kurumu Yayınlarından 5], Ankara: Türk Tarih Kurumu Basımevi. Sayce, A.H. (1876) The Hamathite inscriptions: lecture delivered 2 May 1876; published TSBA 5 (1877), 22–32. Schaeffer. C.F.-A. (1956) Ugaritica iii. Sceaux et cylindres hittites [Mission de Ras Shamra viii], Paris: Geuthner.

Chapter 4

Bedřich Hrozný and the Aegean Writing Systems: An Early Decipherment Attempt Artemis Karnava Abstract Before Bedřich Hrozný concluded his scientific career in the 1940s, he became ­interested, among other scripts, in the 2nd millennium bc Aegean writing systems, most notably Linear A and Linear B. There is archival evidence nowadays that his ­interest was lively for a number of years, even after his acute health problems in the mid-1940s. His interest culminated in a proposed decipherment of Linear B, which was harshly criticized by scholars such as Helmuth Bossert (1944), Alice Kober (1946), Spyr�idon Marinatos (1947) and Emmett Bennett (1950). The reviews concentrated on both dubious recognitions of phonetic values of signs, as well as the historical arguments that were used by Hrozný to support his suggestion and his lumping together of Linear A and B as representing a single Indo-European language. This paper aims to examine the phonetic transcriptions and the historical evaluations proposed by Hrozný at the time, in view of what we now know of the Aegean writing systems in particular and the Mycenaean civilization in general. His decipherment proposal was one of the first to be presented to the public and built upon the extremely limited Linear B material published at the time. His study of this material was based on the assumption that the Minoan culture was in close contact and affinity with Near Eastern cultures, a belief then widespread. An additional clue that led him astray was the fact that simple, ‘linear’ signs, such as the ones used in Aegean writing, were very similar to signs attested in the Egyptian writing systems, in the Anatolian hieroglyphs, the Indus script and even the Phoenician alphabet. This paper thus aims to insert Hrozný’s decipherment attempt into the intellectual circumstances of his era and evaluate its position in the history of the decipherment of the Linear B writing system.

Keywords Hrozný – Aegean writing systems – decipherment – Linear B

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_006

Bedřich Hrozný and the Aegean Writing Systems

63

1 Introduction In a conference aiming to celebrate the 100th jubilee of a major scholarly achievement, it might seem unbecoming to bring into the discussion a facet of that scholar’s work that went amiss. Yet it was the same scholar who contributed to a scientific breakthrough in one field but then offered a false solution in another. So, there is even perhaps a cautionary tale to be told when investigating the dynamics of success and failure at work in one and the same person. Additionally, from the epistemological and historical point of view, achieving a great discovery or reaching a dead end are processes that by no means function in a historical vacuum; the circumstances are significant and worth taking into consideration, in all fairness to the distinguished scholar. It is also probable that by examining the particulars as well as the broader framework, we ourselves will benefit as scholars and realize how much we are influenced by our own worldview and the thought patterns that are current in our times. To begin our story, before Bedřich Hrozný concluded his career in the seventh decade of his life, he demonstrated a profound interest in the 2nd millennium bc Aegean writing systems. At that time, the late 1930s and early 1940s, the writing systems employed in a great part of the area that would be identified in the subsequent millennium with Classical Greece were poorly known and had defied numerous decipherment attempts. Among them, Linear B was the most prominent, and its documents were the most abundantly preserved. Hrozný’s Linear B decipherment proposal was among the first to be presented to the public and was based on the extremely limited Linear B material published at the time. Documents in Linear A and Linear B, as well as a third writing system that escaped Hrozný’s attention, the Cretan Hieroglyphic, were discovered in systematic excavations for the first time in 1900; it was then that Arthur Evans began excavating the Minoan palace of Knossos in Crete (Evans 1899–1900). Evans can be credited with the fact that he realized almost immediately after his initial discovery that he was dealing with three different writing systems and managed to present the Cretan Hieroglyphs in a monograph relatively quickly (Evans 1909); Hrozný on the other hand preferred to ignore the fact that Linear A was a different script from Linear B and proceeded to examine them together. Evans would spend the remainder of his career and life publishing the excavations of the palace in four volumes (Evans 1921–1936), where some more material and some analysis of Linear A and B appeared in the first (1921) and fourth (1935) volumes. In the end, it was his collaborator John Myres who published the Linear B tablets after his death (Evans 1952). Consequently, in the 1930s and 1940s very few Linear B texts were at the ­disposal

64

Karnava

of scholars or the public, and this delay in publication is thought to be one of the reasons why the actual decipherment took more than 50 years after Linear B texts were first discovered. The other major cache of Linear B tablets from Pylos in the Peloponnese was discovered in 1939 (Kourouniotis and Blegen 1939), but they quickly became unavailable for study due to the outbreak of World War ii. 2

Hrozný and the Aegean Scripts

Hrozný (1879–1952), of Czech origin and a native of Bohemia, a part of the Austro-Hungarian Empire at the time, did his undergraduate studies in Berlin and received his doctoral degree in Vienna. After working for years at the Vienna University Library, he became professor at the same university in 1905, only to move in 1918 to the Charles University of Prague as a professor as soon as Czechoslovakia was declared an independent state; he subsequently became a rector at the same university (Matouš 1949; David 1953). It seems that most of the work Hrozný carried out on the Aegean scripts took place before his acute health problems in the mid-1940s. He published his investigations as a decipherment attempt in two lengthy articles. It is not clear when exactly he started working on the decipherment, or what prompted him specifically to engage in such an endeavour, but the first results were already incorporated in a monograph he published in 1940, where he dealt primarily with the prehistory of Anatolia (Hrozný 1940). He subsequently mentions that it was in the course of this investigation that he realized how much influence Western Anatolian cultures of the 3rd and 2nd millennium bc had exerted on Crete, the birthplace of the Aegean writing systems (Hrozný 1949: 1). The first article appeared in a journal of which he himself was the editor, the Archivum Orientale Pragense (Hrozný 1943a). The article, written in German, took up one third of the issue. The presentation starts rather abruptly with a listing of inscriptions presumably deciphered by him, with little explanation of the method used or the process followed. After a short introduction, where he explains that he rejects the previous attempts to decipher Linear B1 which 1 A number of attempts had already taken place before Hrozný’s effort, the most notable being that of A. Persson in 1930, who attributed some phonetic values on the basis of signs similar to the Cypriot syllabary; of F.G. Gordon in 1931, who opted for the Basque language; of F. ­Melian Stawell also in 1931, who opted for Greek (but unfortunately she also interpreted the Phaistos Disc, the Cretan Hieroglyphic and Linear A as Greek). For all of these, see Chadwick 1958: 26–31.

Bedřich Hrozný and the Aegean Writing Systems

65

were primarily based on homomorphisms of Aegean signs with signs of the Cypriot syllabary,2 he moves on to suggest similarities between the Cretan scripts and the Hittite Hieroglyphs—or Luwian Hieroglyphs, as they are known today. He further posited the provenance of this latter script from the ProtoIndus script (Parpola 1996). This is a theory he had already developed earlier in his monograph Über die älteste Völkerwanderung und das Problem der protoindischen Zivilisation (Hrozný 1939) and further presented in the second edition of his Die älteste Geschichte Vorderasiens und Indiens (Hrozný 1943b). He had also detected similarities with cuneiform and Egyptian hieroglyphs, but most strikingly he thought there was even a connection with the Phoenician alphabet. In the article he presented his decipherment of 42 inscriptions, among them a single Knossos tablet. His work was mostly based on inscriptions painted on the inscribed so-called Mycenaean stirrup jars, vases that were destined for the transport and storing of liquid products.3 From what we understand today (but this was also understood back then), the inscribed stirrup jars are the worst specimens of Linear B to start with, because they contain so many unique sign forms and spelling mistakes that they are commonly held to be the work of potters who were in all probability illiterate (Bennett 1986; Hallager 1987). Hrozný continued his efforts after the war with a second article in the same journal, which in the meantime had dropped its Latin title and was renamed in Czech as Archiv Orientální (Hrozný 1946a). By now he remained the sole editor of the journal. He presented inscriptions nos. 43–179, mostly tablets from Knossos, and the occasional sealing.4 This time the language of the article had switched to French. Both of his articles were later joined to form a book translated into French by Madeleine David (Hrozný 1949), and it is this book that reached international audiences and received scathing reviews. The book additionally contained the texts of some tablets from Knossos and Pylos that had been made available to him in the meantime, as well as a Ugaritic inscription. 2

The Cypriot syllabary is a writing system that was employed in Cyprus in the 1st millennium bc primarily for the recording of the Greek language. It was deciphered in 1874 and a number of its signs showed close similarities to Linear B signs; what was not understood at the time was that there was in fact some indirect connection between Linear B and the Cypriot syllabary, but it was wrong to assume that signs that shared the same form between the two scripts also shared by definition the same phonetic values (Chadwick 1958: 30–34). 3 See Zurbach 2016, which contains all the previous bibliography. 4 The Minoan and Mycenaean administrations made use of clay tablets that, unlike in the cuneiform world, were never stamped. Clay sealings are, however, independently attested: these were mostly stamped, but sometimes also bore incised inscriptions (Krzyszkowska 2005: passim).

66

Karnava

In the same volume of the journal, Hrozný presented the decipherment of an inscription in the Byblos script (Hrozný 1946b).5 The Byblos inscriptions are composed in a writing system referred to in modern scholarship as ‘pseudoHieroglyphic’, known from a few inscriptions on metal and stone from Byblos on the Syro-Palestinian coast. The inscriptions date to the first half of the 2nd millennium bc, but very little is known and understood about this system even today, since it is so poorly attested (Daniels 1996). Hrozný apparently felt confident that he could propose a solution to this undeciphered script as well, and so he did. In trying to understand why and how a scholar of Oriental texts and cultures would venture into such territory, a number of possible reasons emerge. His interest appears, to a certain degree, explicable. For starters, attempting to solve a problem of potentially significant historical implications does not seem something out of character for Hrozný. From the biographical account that was written in his honour during his lifetime by his student Lubor Matouš (Matouš 1949), Hrozný emerges as a personality of vivid and varied interests: besides philology, history and religion were fundamental to his research. Additionally, it seems that his successful contribution to the decipherment of Hittite Hieroglyphs (see Hawkins, this volume) had made him bold enough to proceed further and test his intellect on a more challenging set of evidence, one of the last undeciphered writing systems of the Eastern Mediterranean. Nowhere in his presentation of the decipherment does he express any worry or concern of what possible negative criticism might inflict on his reputation or scholarly standing. In fact, he appears so convinced of his results that he hardly makes any reference to the method of decipherment he had used, as any scholar who presents the results of a painstaking process would do. He does mention at some point that although he had to deal with a problem of a different nature than the Hittite decipherment, he proceeded nonetheless by use of the same ­method that he had employed before (Hrozný 1949: 2). In the instance of the Aegean scripts, both the writing as well as the language were unknown, as ­opposed to Hittite, where only the language was unknown and the phonetic values of script signs were already known. He omitted, however, any mention of ­particulars regarding his method, except for calling it ‘combinatory’ and ‘comparative’.6 5 This Byblos inscription was already included in his second edition of Die älteste Geschichte Vorderasiens und Indiens (1943b). 6 ‘Dans notre étude des inscriptions crétoises—écrites en une écriture inconnue ainsi qu’en une langue inconnue—nous avons eu recours à la même méthode, essentiellement

Bedřich Hrozný and the Aegean Writing Systems

67

But as a scholar of his time, he also kept in mind the big picture and was very much preoccupied with matters of racial attributions, affiliations and origins. Already at the end of the 19th century, the main explanation for how cultural achievements and innovations spread in the world involved population movements and even invasions. This was the premise of diffusionism, a theory that repeatedly came in competition with, but at times complemented, evolutionary historical explanations (Trigger 1989: 148–155). The grand competing theories at the beginning of the 20th century in European scholarship are best ­expressed, on one hand, through the work of the Swedish archaeologist Oscar Montelius, who became the most prominent advocate of diffusionism by suggesting that all innovations in Europe had come from the Orient (the ex oriente lux theory, Montelius 1899). Montelius’ theories were not very popular in German academia, where more chauvinistic and nationalistic approaches were current. The theories of Gustaf Kossinna, who advocated ferociously in favor of the purity and superiority of the Germanic peoples on the basis of archaeological, material and cultural remains, were in this respect more extreme and constituted the opposite end of the spectrum (Trigger 1989: 163–167). Although Hrozný was German-trained and belonged to the German-speaking academic environment, his choice of sides, while never expressed verbatim, is not hard to imagine. He was a member of a linguistic and ethnic minority, on top of being an Orientalist: it is more than likely that he subscribed to diffusionist, Montelian ideas and rejected the Germanocentric ‘Aryan’ approaches, which ­reserved no respectful place for the Slavic minority he was part of. Furthermore, Hrozný in his books actively engaged in supporting the IndoEuropean hypothesis (Mallory 1989; Anthony 2007), in that he placed the Cretans, the inventors and users of Linear A and B, from the ethnic as well as the linguistic point of view together with the Hittites as a branch of the Indo-­ European language family (esp. Hrozný 1943b: passim). By placing Hittite ­hieroglyphs in direct comparison with the Proto-Indus script, he claimed a common ethnic and linguistic background for all: Hittites, Indians and Cretans. His theory was based on the ‘decipherments’ and the consequent readings by him of otherwise undeciphered inscriptions from the above cultural circles, but his arguments were rather circular. He suggested that because all these peoples were presumably affiliated, it was this affiliation that explained superficial or even circumstantial similarities in their writing systems; in their turn, these similarities verified, in his opinion, the relevant readings of the inscriptions, which verified the ethnic affiliation in the first place (see also further below). c­ombinatoire et comparative, que pour notre déchiffrement des inscriptions hittites-­ cunéiformes, «hittites»-hiéroglyphiques, et proto-indiennes’ (Hrozný 1949: 2).

68 3

Karnava

Michael Ventris, Alice Kober and John Chadwick: the Decipherers

In the meantime, a young British architect by the name of Michael Ventris was actively working even before the war on the decipherment of Linear B (Ventris 1940).7 As with all success stories, Ventris’s successful decipherment has been frequently told, and many details about the steps leading to it are well known (Chadwick 1958; Pope 1975: 159–179). Hrozný was one of the twelve scholars to whom Ventris addressed his famous “Mid-century Report” in 1949, a questionnaire aiming to discuss the complex decipherment problems of the Aegean scripts (Ventris 1988: 29–132). Hrozný did not reply to the questionnaire but was always on the list of recipients of Ventris’s ‘Work Notes’, a sort of progress report that he would send out to scholars from 1951 onwards (Ventris 1988: 133– 333). These ‘Work Notes’ eventually culminated in the successful decipherment (Ventris and Chadwick 1953; Ventris and Chadwick 1973). Before Ventris, however, another scholar, the American Alice Kober, is known to have been active in the Linear B decipherment effort. Kober published an article that proved to be groundbreaking, one that laid the foundations for Ventris’s successful decipherment (Kober 1948). Kober’s contribution lies in the fact that she identified Linear B as recording an inflectional language, and, through her famous ‘triplets’, sets of three words with the same stem but different endings, she correctly figured out the gender endings for male and female names in Linear B. It is believed that Kober’s untimely death in 1950 at the relatively young age of 43 prevented her from contributing further to the decipherment.8 In Kober’s archive, which is now kept at the Program for Aegean Scripts in the Classics Department at the University of Texas at Austin, among her correspondence there are two postcards from Hrozný dating to 1947 and 1948.9 In the first postcard of 1947 (Fig. 4.1) Hrozný sends Kober the second article of his decipherment attempt, which had been published, as previously mentioned, in 1946. One can imagine that in the postwar usa it was not easy to acquire a copy of a journal published in Prague, so a direct request to the author would solve this problem. Hrozný also invites Kober to publish her work in his journal, if she so wishes. Finally, he appears certain that the scholarly community will eventually recognize his decipherment of the Cretan script, 7 Ventris was just 18 years old when he published his first account on the subject and reportedly hid his age from the aja editors. 8 A biography of Kober has recently appeared, where her contribution to the decipherment is illuminated (Fox 2013). 9 Kober’s archive has been digitized and is available from the University of Texas Libraries online repository.

Bedřich Hrozný and the Aegean Writing Systems

Figure 4.1 A postcard from Hrozný to Kober dated 22.9.1947 (courtesy of Tom Palaima, Program in Aegean Scripts and Prehistory and the Digital Repositories of the University of Texas at Austin Libraries)

69

70

Karnava

like his decipherment of the Hittite as well as the Proto-Indus script. It is thus evident that Hrozný draws on the weight of his recognized work on Hittite to vouch for his work on the Minoan and the Proto-Indus scripts. In his second postcard, presumably written sometime in 1948 (Fig. 4.2), Hrozný thanks Kober for sending him the photograph of a tablet from Pylos, but he complains that the photo was so bad that the writing was practically illegible. Still, he feels that his readings of the tablet confirm some of his suggestions, but he asks for a better photo. Before these cards were exchanged, it is not certain whether Hrozný had read a review Kober had published (Kober 1946) of his book (Hrozný 1943b) as well as the first part of his decipherment (Hrozný 1943a). The review is sharp and most clearly negative, because it focuses on Hrozný’s weakest point: his ideological convictions. As mentioned previously, he believed that the ­ancient populations that inhabited the area from the Eastern Mediterranean and A ­ natolia to the Indus Valley were all Indo-European and so were their ­languages. In the unsuccessful effort to support his theory, Hrozný chose to omit from his account certain areas altogether, such as Asia Minor and the ­Syro-Palestinian coast, and he makes no mention of Egypt. On the basis of ostensible and highly subjective similarities between Hittite Hieroglyphic signs and what is known today as the Indus or Proto-Harappan script, he deduced that all the l­ anguages behind them, hence all the populations that spoke them, were Indo-European. Incidentally, Hrozný’s decipherment premise comes into stark contrast with Ventris’s method. It is a well-known fact that Ventris started out his work on the decipherment of Linear B in the belief that the language of the inscriptions was non-Indo-European; instead, he favored Etruscan, and because of its affiliation with Lemnian, i.e. the pre-Greek language of Lemnos in the Aegean, he proposed what is known as the “Pelasgian solution”, that Linear B recorded the language of pre-Greek populations. This theory represented current wisdom at the time until the decipherment of Linear B. But, as opposed to Hrozný, who started out with the purpose of documenting a theory rather than let the evidence guide him to it, Ventris started out with a wrong theory, which he was flexible enough to change when the evidence became overwhelming in favour of the opposing theory. Kober was Hrozný’s most fearsome critic. She in fact delivered the most brutal of all Hrozný’s reviews while Hrozný was still alive, as she herself admits to a friend and colleague in a private letter found in her archive.10 In this letter, 10

‘I read your letter first, then looked the issue over in order to get my breath and discovered it contained my review of Hrozny. Since I never have gotten over the thrill that comes

Bedřich Hrozný and the Aegean Writing Systems

Figure 4.2 A postcard from Hrozný to Kober dated to 1948 (courtesy of Tom Palaima, Program in Aegean Scripts and Prehistory and the Digital Repositories of the University of Texas at Austin Libraries)

71

72

Karnava

she appears impressed by the fact that his outrageous theories are received in scholarship less aggressively than they deserved. John Chadwick, the philologist who collaborated with Ventris, reserved an equally brutal criticism for Hrozný’s fruitless attempt in the aftermath of the successful decipherment by Ventris and himself. Chadwick was 38 years old, i.e. relatively young, when he became the historian and narrator of his own and Ventris’s unique triumph (Chadwick 1958: 27–28). He therefore spared no mercy for an elderly and respected scholar’s defeat, and besides stressing the obvious faults in Hrozný’s contribution, or rather his absence of method, he even added an ad hominem attack on the by then late scholar.11 4

More Criticisms of Hrozný’s Decipherment

Further negative reviews of Hrozný’s Die älteste Geschichte Vorderasiens und Indiens include one by the Belgian scholar Charles Delvoye (Delvoye 1949). Delvoye lists inaccuracies and omissions committed by Hrozný with regard to the archaeological evidence he used in his book. Hrozný had a keen interest in archaeology and archaeological remains and had always maintained that theories should be formed and tested on the basis of holistic analysis of data, and linguists and philologists should also take into account archaeological evidence and vice versa. But in order to make his data fit his theory, he appears to have made selective use of archaeological data as well. Delvoye delivers an impressive review of Hrozný’s book in 20 pages, in which he tries to set the record straight on a variety of research problems, such as the Neolithic pottery of Thessaly which Hrozný compares to pottery from Jericho, thus positing a northern derivation of Syro-Palestinian populations. He also criticizes Hrozný’s ‘decipherment’, for which he thinks that Hrozný got so carried away because of his vast knowledge of Mesopotamian cultures and languages, that the Cretans were barely left enough space to demonstrate their disparity with their eastern neighbours.

11

from seeing my words in print, I read it. Considering some of the reviews on the same subject that I’ve read, that one is ferocious. Everybody seems to handle Hrozny with kid gloves. I suppose it’s because nobody thinks a man with Hrozny’s reputation could possibly be as stupid as he seems’ (letter excerpt, written by Kober in 1947, found in Fox 2013: 92). ‘The arbitrariness of Hrozný’s work is so patent that no one has taken it seriously. It is a sad story which recurs too often in the world of scholarship: an old and respected figure produces in his dotage work unworthy of his maturity, and his friends and pupils have not the courage to tell him so’ (Chadwick 1958: 28).

Bedřich Hrozný and the Aegean Writing Systems

73

Hrozný’s decipherment of the Linear B system was harshly criticized at the time by further scholars, such as the German Helmuth Bossert (Bossert 1944: 247–249, 272–273), the Greek archaeologist Spyridon Marinatos (Marinatos 1947), the British archaeologist John Myres who had, among other tasks, worked alongside Evans on the Knossos Linear B tablets (Myres 1946, 1948), and the American scholar Emmett Bennett, who was also instrumental to the correct decipherment (Bennett 1950). Bennett in his usual humorous and wellintentioned manner additionally offered some words to live by for all would-be decipherers.12 The reviews concentrated on both dubious recognitions of phonetic values of signs, as well as the historical arguments that were used by Hrozný to support his suggestion. Especially criticized was his lumping together of both Linear A and Linear B under an interpretation as a single Indo-European language. As a result of these reviews, but also because of the sweeping effect of Ventris and Chadwick’s decipherment shortly afterwards, all previous decipherment efforts were dismissed and laid to rest by the scholarly community. 5

A Concluding Appreciation of Hrozný’s Decipherment Attempt

In a list of reasons why Hrozný’s effort was unsuccessful, the first one must be the paucity of material at the aspiring decipherer’s disposal. There were simply not enough texts published at the time for anyone to venture into attempts at decipherment, let alone to try and verify them afterwards. This paucity of material is the main reason why the other 2nd millennium bc Aegean writing systems, namely the Cretan Hieroglyphs and Linear A,13 remain undeciphered to this day. The most recent overall count of Bronze Age Aegean texts amounts to 6,000 documents with 70,000 signs for Linear B, 1,500 documents with 8,000 signs for Linear A and 350 documents with 3,000 signs for the Cretan Hieroglyphs;14 the total is 7,850 documents in all these scripts combined with 81,000 signs attested, which in today’s perspective would fill a little less than 22 A4 pages in the application Word.15 At the time of Ventris and Chadwick’s 12 13 14 15

‘The problem which ought to be of most interest to those who will undertake similar studies of their own is why it is that this decipherment has satisfied its author, and has not satisfied any others’ (Bennett 1950: 81). It is also likely that the Phaistos Disc represents a unique specimen of yet another 2nd millennium bc Aegean script (Karnava 2014). The numbers are from Olivier 2008: 173–174, based on a 2004 count. These totals have not changed dramatically since then. On the basis of 3,750 characters with spaces per page.

74

Karnava

d­ ecipherment there were even fewer Linear B documents to work with, and Ventris seems to have struggled with some 30,000 signs at his disposal (Olivier 2008: 182). A combination of further reasons for Hrozný’s fruitless effort was his preconceived historical perceptions together with a series of methodological errors. Since he himself is extremely laconic on the matter of his work methods and procedures, it is not clear what initiated the vicious circle: did his theories about Völkerwanderungen and technological diffusion lead him to perform selective and arbitrary use of his material, or did his methodological errors lead him to believe that his theories were grounded on credible evidence? Hrozný’s study was based on the assumption that the Minoan and Mycenaean cultures of the Aegean region had close affinities with further Eastern cultures, a belief which was stretched to the extent that casual similarities in traits of material culture were taken to be substantial and meaningful affiliations. On the other hand, a fundamental methodological error was his reliance on presumed sign matches between different writing systems that were evidently totally unrelated; sign matches could, in theory, point to some relation or even affiliation between distant scripts, but not when these matches involve simple, ‘linear’ script signs. Relatively simple signs, such as a cross (‘X’) or a rounded sign (‘O’), can be and have been invented independently and used in totally unrelated writing systems. A further fundamental deficiency of his study was the absence of proper text editions, even for the few texts that were available. A thorough and detailed study and edition of a writing system under study is imperative, so that a researcher does not confuse signs, the phonetic value of which changes with a single stroke (e.g. ‘O’ and ‘Q’; ‘P’ and ‘R’) or because of divergent reading directions (e.g. ‘M’ and ‘W’; ‘b’ and ‘d’; ‘p’ and ‘q’). Furthermore, although Hrozný is hailed as a vital contributor to the decipherment of Hittite, what is meant by ‘decipherer’ in the case of reading the Hittite language behind its cuneiform version is someone who stepped in to interpret the language after paleographers had managed to attribute phonetic values to the signs of the writing system. It seems that there, Hrozný’s extensive knowledge of Indo-European and Semitic languages was the best equipment with which to handle problems of a linguistic nature. But in the case of Linear B, what was needed at that stage was someone who would be able to attribute phonetic values to signs of an unknown writing system, in order for the reading process to commence; the eventual recognition of the language ­involved would be a further step in the process, one that was also a desideratum.16 16 See Barber 1974: 3–33 for the fundamentals of archaeological decipherments.

Bedřich Hrozný and the Aegean Writing Systems

75

The attribution of proven phonetic values to script signs remains the crucial and elementary problem faced by prospective decipherers of Linear A to this day. Since sign form countertypes alone between two different writing systems cannot guarantee phonetic correspondence, it is a fact that whoever proposes to read or interpret Linear A texts on the basis of sign similarities with Linear B is moving on thin ice.17 It is imperative that phonetic values be demonstrably attributed to script signs, before one or more contributors can be credited with the decipherment of an ancient writing system. Until then, the Cretan Hieroglyphic and Linear A scripts defy all attempts at decipherment, and in order for someone to suggest sign readings beyond reasonable doubt, evidence other than circumstantial clues will be required. Bibliography Anthony, D.W. (2007) The Horse, the Wheel and Language. How Bronze-Age Riders from the Eurasian Steppes Shaped the Modern World, Princeton: Princeton University Press. Barber, E.J.W. (1974) Archaeological Decipherment. A Handbook, Princeton: Princeton University Press. Bennett, E.L. (1950) Review of Hrozný 1949, AJA 54/1, 81–82. Bennett, E.L. (1986) The inscribed stirrup jar and pinacology, in: Φίλια Έπη εις Γεώργιον Ε. Μυλωνάν δια τα 60 έτη του ανασκαφικού του έργου, Vol. i, Αθήνα: Αρχαιολογική Εταιρεία, 136–143. Bossert, H.T. (1944) Ein hethitisches Königssiegel. Neue Beiträge zur Geschichte und Entzifferung der hethitischen Hieroglyphenschrift [Istanbuler Forschungen 17], Berlin: Zweigstelle Istanbul des Archäologischen Instituts des Deutschen Reiches. Chadwick, J. (1958) The Decipherment of Linear B, Cambridge: Cambridge University Press. Daniels, P.T. (1996) Pseudo-hieroglyphs of Byblos, in: P.T. Daniels and W. Bright (eds.), 29–30. Daniels, P.T. and Bright, W. (eds.) (1996) The World’s Writing Systems, New York, Oxford: Oxford University Press. David, M. (1953) La vie et l’œuvre de Bedřich Hrozný, RA 47, 96–100.

17

A method of establishing the approximate phonetic values of some ten signs of the Linear A script can be found in Olivier 1975, with some additions to his catalogue by use of the same principle in Karnava 2016.

76

Karnava

Delvoye, C. (1949) Archéologie orientale et archéologie préhellénique [À propos de l’histoire de l’Asie antérieure, de l’Inde et de la Crète de M. Bedrich Hrozny], Revue belge de philologie et d’histoire 27, 929–948. Evans, A.J. (1899–1900) Knossos. Summary report of the excavations in 1900. i. The palace, ABSA 6, 3–70. Evans, A.J. (1909) Scripta Minoa. The Written Documents of Minoan Crete, with Special Reference to the Archives of Knossos, i: The Hieroglyphic and Primitive Linear Classes, with an Account of the Discovery of the Pre-Phoenician Scripts, their Place in Minoan Story and their Mediterranean Relations, Oxford: Oxford University Press. Evans, A.J. (1921–1936) The Palace of Minos at Knossos, London: British Museum Press. Evans, A.J. (1952) Scripta Minoa. The Written Documents of Minoan Crete, with Special Reference to the Archives on Knossos, ed. J.L. Myres, Oxford: Clarendon Press. Fox, M. (2013) The Riddle of the Labyrinth. The Quest to Crack an Ancient Code, New York: Echo. Hallager, E. (1987) The inscribed stirrup jars. Implications for Late Minoan iii B Crete, AJA 91, 171–190. Hrozný, B. (1939) Über die älteste Völkerwanderung und das Problem der proto-indischen Zivilisation. Ein Versuch, die proto-indischen Inschriften von Mohendscho-Daro zu entziffern [Monografie Archivu Orientálního 7], Prag: Orientalisches Institut. Hrozný, B. (1940) Die älteste Geschichte Vorderasiens, Prag: Melantrich. Hrozný, B. (1943a) Kretas und Vorgriechenlands Inschriften, Geschichte und Kultur, Archivum Orientale Pragense 14, 1–118. Hrozný, B. (1943b) Die älteste Geschichte Vorderasiens und Indiens, Prag: Melantrich (2nd ed.) = Histoire de l’Asie antérieure, de l’Inde et de la Crète, Paris: Pavot, 1947 (traduction française par Madeleine David). Hrozný, B. (1946a) Les inscriptions crétoises, AO 15, 158–302. Hrozný, B. (1946b) Die hieroglyphische Stele von Byblos, AO 15, 138–139. Hrozný, B. (1949) Les inscriptions crétoises. Essai de déchiffrement, Praha: Orientální Ústav (traduction française par Madeleine David). Karnava, A. (2014) Phaistos Disc, in: G.K. Giannakis (ed.), Encyclopedia of Ancient Greek Language and Linguistics, Vol. 3, Leiden, Boston: Brill, 62–63. Karnava, A. (2016) On sacred vocabulary and religious dedications: the Minoan ‘libation formula’, in: E. Alram-Stern, F. Blakolmer, D. Deger-Jalkotzy, R. Laffineur and J. Weilhartner (eds.), Metaphysis. Ritual, Myth and Symbolism in the Aegean Bronze Age. Proceedings of the 15th International Aegean Conference, Vienna, Institute for Oriental and European Archaeology, Aegean and Anatolia Department, Austrian Academy of Sciences and Institute of Classical Archaeology, University of Vienna, ­22–25 April 2014 [Aegaeum 39], Leuven, Liège: Peeters, 345–355, pl. cvi–cviii. Kober, A.E. (1946) Review of Hrozný 1943a, 1943b, AJA 50, 493–495. Kober, A.E. (1948) The Minoan scripts: fact and theory, AJA 52, 82–103.

Bedřich Hrozný and the Aegean Writing Systems

77

Kourouniotis, K. and Blegen, C. (1939) Excavations at Pylos, 1939, AJA 43, 557–576. Krzyszkowska, O. (2005) Aegean Seals. An Introduction, London: Institute of Classical Studies. Mallory, J.P. (1989) In Search of the Indo-Europeans. Language, Archaeology and Myth, London: Thames and Hudson. Marinatos, S. (1947) H ανάγνωσις της κρητομυκηναϊκής γραφής υπό του Β. Hrozny, Κρητικά Χρονικά 1, 387–390. Matouš, L. (1949) Bedřich Hrozný. The Life and Work of a Czech Oriental Scholar, Prague: Orbis. Montelius, O. (1899) Der Orient und Europa. Einfluss der orientalischen Cultur auf Europa bis zur Mitte des letzten Jahrtausends v. Chr., Stockholm: Königliche Akademie der schönen Wissenschaften, Geschichte und Altertumskunde. Myres, J.L. (1946) Professor Hrozny’s table of the Minoan signary, JHS 66, 129. Myres, J.L. (1948) The Minoan script, according to professor Bedřich Hrozný, AJA 52, 104–106. Olivier, J.-P. (1975) ‘Lire’ le linéaire A?, in: J. Bingen, G. Cambier and G. Nachtergael (eds.), Le monde grec: pensée, littérature, histoire, documents. Hommages à Claire Préaux [Université Libre de Bruxelles, Faculté de Philosophie et Lettres 62], Bruxelles: Université Libre de Bruxelles, 441–449. Olivier, J.-P. (2008) Les écritures crétoises, in: R. Τreuil, P. Darcque, J.-Cl. Poursat and G. Touchais, Les civilisations égéennes du Néolithique et de l’Âge du Bronze, Paris: Presses Universitaires de France (2nd revised ed.), 169–183. Parpola, A. (1996) The Indus script, in: P.T. Daniels and W. Bright (eds.), 165–171. Pope, M. (1975) The Story of Decipherment. From Egyptian Hieroglyphs to Maya Script, London: Thames and Hudson. Trigger, B.G. (1989) A History of Archaeological Thought, Cambridge: Cambridge University Press. Ventris, M. (1940) Introducing the Minoan language, AJA 44, 494–500. Ventris, M. (1988) Work Notes on Minoan Language Research and Other Unedited Papers, ed. A. Sacconi [Incunabula Graeca 40], Rome: Edizioni dell’Ateneo. Ventris, M. and Chadwick, J. (1953) Evidence for Greek dialect in the Mycenaean archives, JHS 73, 84–103. Ventris, M. and Chadwick, J. (1973) Documents in Mycenaean Greek, Cambridge: Cambridge University Press (2nd ed.). Zurbach, J. (2016) Vasi con iscrizioni in lineare B, in: M. del Freo and M. Perna (eds.), Manuale di epigrafia micenea. Introduzione allo studio dei testi in lineare B, Padova: Libreriauniversitaria.it, 613–622.

Chapter 5

A Fruitful Collaboration between E. Sellin and B. Hrozný during His Viennese Years: The Cuneiform Texts from Tell Taanach and Their Impact on Syro-Levantine Studies Regine Pruzsinszky Abstract At the beginning of the 20th century, excavations at Tell Taanach, situated close to Megiddo in the Jezre’el Valley, were initiated by the theologian and Biblical archaeologist Ernst Sellin, professor in Vienna at that time. During the campaigns between 1902 and 1904, the largest then known Canaanite corpus of cuneiform texts was discovered. Sellin supported the then almost unknown Assyriologist Bedřich Hrozný in appointing him as the epigraphist of the excavation, and Hrozný subsequently published the important text finds, including lists of persons and epistolary texts, dating to the mid- to late 15th century bce. This paper provides an overview of the most important characteristics of these texts and their chronological, historical and linguistic background. Furthermore, the impact of the texts in subsequent and more current research within Assyriology is discussed, with special focus on Hrozný’s achievements within the study of the Syro-Levantine region in the Bronze Age.

Keywords Late Bronze Age – Canaan – Amarna – Peripheral Akkadian – West Semitic – Hurrian – onomastics

When I first came across the cuneiform texts from Tell Taanach, I was still a student in Vienna and did not fully grasp their importance for the overall cuneiform culture. In 2006, just before the 100-year anniversary of Ernst Sellin’s publication of the excavation results of Tell Taanach, I was asked to discuss the onomastic evidence of the cuneiform texts from that site.1 I was struck by the 1 Pruzsinszky 2006: 99–114. © koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_007

A Fruitful Collaboration between E. Sellin and B. Hrozný

79

wealth of studies that dealt with various aspects of the relatively small text corpus.2 Today, another ten years later, I am going to discuss the Taanach tablets with regard to the impact of Hrozný’s edition, which took place before the anniversary celebrated by this conference on “Hrozný and Hittite: The First Hundred Years,” since it was the very beginning of the 20th century when the Austrian excavations directed by Ernst Sellin took place in Tell Taanach. On board the team was 25-year-old Bedřich (Friedrich) Hrozný, who had just completed his dissertation on South Arabian graffiti in Vienna a few years earlier in 1901.3 1

Friendship and Cooperation at Tell Taanach

From the very beginning of his professorship in Old Testament studies and Biblical archaeology in Vienna, Ernst Sellin sought for a possibility to excavate in northern Palestine. His aim was to integrate archaeological methods into the field of Old Testament studies. And thus, after strenuous journeys in the late 1890s and negotiations with the Ottoman authorities for a license, the pioneer in Biblical archaeology chose a not too large-sized tell of ca. 6 ha for a start, situated in an area where no excavations had previously taken place: Tell Taanach, situated only 8 km southeast of Megiddo at the southern end of the Jezreel Valley.4 The site’s importance was certainly due to its strategic location overlooking the Via Maris leading from Egypt to Anatolia and Mesopotamia. The excavations at Tell Taanach between 1902 and 1904 marked the beginning of a fruitful cooperation between Biblical and Ancient Near Eastern studies and were the first Austrian-German excavations in this region, to be later followed by the larger ones of Šechem and Jericho.5 Although Ernst Sellin was not an archaeologist and later was critically accused of nonprofessional work, he certainly proved to be a most capable project manager in obtaining funding for his excavation from the Imperial Ministry of Culture, the Academy of Sciences, the Deutscher Palästina-Verein, and the Austrian Lloyd company. Sellin was characterized as a “flexible and innovative intellect coupled with the effort to connect biblical scholarship and archaeology, which was a completely new endeavor.”6 Ernst Sellin’s daughter remembered: “Seine Wiener Jahre waren

2 3 4 5 6

For a summary, see Kreuzer 2008. Segert 1999: 630–631. Morandi Bonacossi 2012: 387–388 and Horowitz 2012: 387. See Palmer 2012 for a detailed study of Ernst Sellin’s life. http://www.tau.ac.il/~archpubs/megiddo/revelations4.html (accessed 15.05.2017).

80

Pruzsinszky

die erfülltesten seines Lebens. In dieser weltoffenen Stadt fanden er und seine Frau gleichgesinnte Geister, Freundschaften unter den Kollegen….”7 Among those contacts dear to him personally and professionally was his former student in Hebrew language Bedřich Hrozný, an aspiring Orientalist who had studied with the Semitist David Heinrich Müller and was at the beginning of his impressive career.8 Hrozný, who had learned a number of ancient languages as a tool to understand the cultural and historical background of ancient texts, was working at the university library to earn a living. Sellin supported this young scholar by appointing him as the epigraphist of his excavations. He reported stressing his hopes to discover some more cuneiform texts:9 Ich unternahm diese Expedition in der ausgesprochenen Hoffnung, ­ eitere keilinschriftliche Täfelchen zu finden. Da ich nun selbst nicht w imstande bin, dieselben zu entziffern, und andererseits das türkische Antiquitätengesetz einen auch nur zeitweiligen Transport außer Landes nicht gestattet, so erschien es dringend geboten, einen des Babylonischen Kundigen diesmal mitzunehmen, der die Entzifferung eventuell gleich an Ort und Stelle vornehmen konnte. Natürlich konnte ich keinen Geeigneteren dazu finden als Dr. F. Hrozný, der sich bereits durch die treffliche Edition der im Jahre 1903 von mir ausgegrabenen Texte verdient gemacht hatte. Bedřich Hrozný had studied Akkadian in Vienna, mostly as an autodidact. ­Between 1901 and 1902 he intensified his cuneiform studies with Friedrich De­ litzsch and Hugo Winckler in Berlin, and from 1905 onwards he taught cuneiform at the University of Vienna; among his students were Viktor Christian and Harry Torczyner.10 Only after the dissolution of the Austro-Hungarian Empire in 1918 was Hrozný appointed Professor of Cuneiform Studies and the History of the Ancient Near East at the Charles University in Prague. In his obituary, Ernst Weidner (1952–1953: 395) summarized: “Diese lange Zeit seiner Tätigkeit als Bibliothekar und Privatdozent wurde nur unterbrochen durch eine ­Orientreise mit E. Sellin im Jahre 1904.” Siegfried Kreuzer (2015: 200) concluded: “…man darf wohl annehmen, dass die ersten grabungsarchäologischen 7 8

9 10

Palmer 2012: 15. On Hrozný’s professional life, see e.g. Segert 1999: 629–648. Hrozný is certainly one of few scholars to have been honored with a multi-volume Festschrift, published in Archiv Orientální 17 and 18 (1949–1950). On Hrozný’s formative years in Vienna between 1897 and 1919, see now Eichner 2015: 82–125, especially pp. 95–100 on Hrozný’s collaboration with Sellin. Weidner 1952–1953: 395; Sellin 1906: 2. Hunger 2004: 20.

A Fruitful Collaboration between E. Sellin and B. Hrozný

81

­ rfahrungen in Taanach Hrozný zu seinen späteren, eigenen Grabungen in E Kültepe/Kanisch zumindest mit motivierten und ermutigten.” 2

The Mini Archive from Tell Taanach

During the 1903 excavation season, the first four cuneiform tablets, two of them in a clay chest, were discovered in a large complex of buildings then called “Burg.” These tablets were the first cuneiform tablets found in Palestine, which gave evidence of the wide use of cuneiform script beyond Mesopotamia. This discovery of cuneiform texts was the key reason for Sellin to conduct another season in 1904 to find more texts that might have been overlooked. Sellin took Hrozný along to have him read and copy the texts at the site. Hrozný, who had previously worked on tablets in the British Museum, “applied his knowledge from the Hebrew Bible by comparing Akkadian and Hebrew words and by referring to biblical texts,” as Stanislav Segert (1999: 631) stated in his sketch of Hrozný’s life. And indeed, Sellin found more tablets: in total, the excavations by Sellin resulted in 12 texts, among them letters and lists of persons.11 Hrozný received another grant from the academy to collate all the Taanach tablets stored in the Filistin collection of the archaeological museum in Istanbul12 to be published in Sellin’s excavation report. Only in the 1960s were the excavations resumed by the American archaeologist Paul Lapp, who discovered two more texts, published by Hillers and Glock.13 The mini-archive from Taanach14 predates the well-known Amarna letters by about two generations and provides information on the political and economic situation in the Egyptian province of Canaan in the late 15th century during the Late Bronze i period (ca. 1450–1200 bce).15 1 2 3 4 11 12 13 14 15

Letter: Ehli-Teššup to Talwašur Letter: Ahiami to Talwašur Administrative tablet: list of people called to service Administrative tablet: personal names Hrozný 1904 and 1906. For an overview, see now Horowitz et al., 2018: 132–155 and Berlejung 2006: 230–234. For those texts stored in the Filistin collection (Fi.) of the Archaeological Museum in ­Istanbul, see Horowitz and Oshima 2004. Hillers 1964; Glock 1971 and 1983. Horowitz et al. 2018: 14–15. See also van der Toorn 2000 for an overview of cuneiform texts discovered in the Levant, to which the newer book by Horowitz et al. (2006 and 2018) can be added.

82

Pruzsinszky

5 (4a)16 6 7 8 9 10 (8a) 11 12 13 14 15 16 17

Administrative tablet (fragment): personal names Letter: Amenhatpa to Talwašur Letter: Amenhatpa to Talwašur Administrative tablet: personal names Letter (fragment) Letter (fragment) Letter (fragment) Letter (fragment) Letter (fragment) Administrative tablet (fragment): personal names Inscribed cylinder seal of Ātanah-ilī with Egypt. Symbols Administrative tablet: personal names Alphabetic cuneiform text

Total: 9 letters, 6 administrative texts, 1 seal inscription, 1 alphabetic text Taanach was at the time a city-state within the Egyptian province of Canaan and the place from which Thutmose iii marched against Megiddo in 1468.17 The city was subordinated to Megiddo and stood in contact with the city ruler Aḫiami of Tel Rehob, situated close to Beth-Shean. It was governed by a local ruler and mariannu named Talwašur,18 who seemingly did not fully obey Amanḫatpa, the Egyptian governor in Gaza, the main Egyptian provincial administrative center. The letters deal with the supply of chariots, weaponry and persons, and it becomes obvious that Talwašur did not show up before Amanḫatpa to fulfill the provision of the requested equipment/arming. The primary purpose of the lists of persons is less clear: we might be dealing with soldiers, workmen or alike. But the names of different linguistic traits attest to a multiethnic/multilingual society. Another short and unintelligible alphabetic text contains alphabetic cuneiform signs similar to those of Ugarit, probably in a local or untrained form, and postdates the letters and lists by about 200 years.19 It is a fine example of the

16 17 18 19

4a and 8a is the numbering system according to B. Hrozný. On Canaan in the 2nd millennium bc, see Na’aman 2005 with several references to Tell Taanach. On the disputed reading of this name, see below. Sanders in Horowitz et al. 2018: 165–167.

A Fruitful Collaboration between E. Sellin and B. Hrozný

83

spread of the Ugaritic alphabetic script to a relatively small scribal center in Palestine. Summarizing, the archive provides glimpses on the economic and social history and the exchange of people (such as the “brothers,” asīru-men, ḫanakumen and prisoners) and goods (mostly raw materials and weapons) between the suppliers in the Jezreel valley and the Egyptian recipients. It is obvious that “Taanach … served as a protected center of specialists supporting the population in the region.” Although we are dealing with a relatively small archive, its very existence means “that the site served government interests.”20 Considering the number of tablets found in Palestine, the number of tablets found in Taanach is only outnumbered by the total covering the entire second millennium of the current excavations of Ḫazor (amounting to 19 tablets so far).21 The chart by Horowitz et al. (2018: 15) demonstrates that Tell Taanach still provides the highest number of Late Bronze Age tablets found so far in Canaan. The reason that relatively few tablets have been discovered in the Western Periphery is because the well-known sites, such as Aphek, Gezer, Ḫazor, Megiddo and also Taanach, continued to be inhabited after their destruction at the end of the Late Bronze Age. Resettlements certainly led to ­removal of the archives, which leaves us incidental text finds in mostly secondary contexts.22 Hrozný’s edition of the cuneiform texts a few weeks after the campaign in 1904 in the first “Endbericht,” and the “Nachlese” submitted in 1905 and published in 1906, remained the basis of discussions for many years. Despite Hrozný’s “cuneiformist eyes,”23 later collations of the tablets by Glock and Gordon, Schaffer and Rainey resulted in an improved understanding of the context of these sources.24 But due to the deteriorating state of preservation and the loss of a number of tablets, the photographs and handcopies by Hrozný remained indispensable for subsequent studies, including the reedition by Horowitz, ­Oshima and Sanders in 2006 and 2018 incorporating the unpublished collations by Glock, Gordon and Schaffer.25

20 21 22 23 24 25

Glock 1983: 66. See table 3 in Horowitz et al. 2018: 15, 63. Cohen 2008: 84–85, referring specifically to the text finds of Ḫazor. Glock 1983: 59 n. 11. For bibliographical details and the information drawn from unpublished notebooks, see Horowitz et al. 2018: 132–135. For further published collations see Rainey 1999: collation of letters 1–2 and 5–6 in 1971. See also Horowitz et al. 2004.

84 3

Pruzsinszky

Use of Cuneiform in Canaan

The Taanach letters are the earliest attestations of Canaano-Akkadian, a development of a Western dialect within a distinct scribal culture in Canaan.26 The texts demonstrate that the Canaanite scribes applied a Canaanite inflection to the Old Babylonian stems (hybrid forms especially in the system of verbal morphosyntax) sometime between the end of the Middle Bronze Age and the beginning of the Late Bronze Age. Anson F. Rainey, the author of a monumental linguistic analysis of the mixed dialect used by scribes from Canaan,27 stressed that the letters share the same Old Babylonian traits and lack of Middle Babylonian traits as do the Amarna letters from Canaan. Thus, they belong to the “Canaanite” sphere of West Semitized grammar, in sharp contrast to the peripheral Akkadian of the city-states in the north. Already in the 1970s, Rainey had devoted a separate study based on thorough collations to the verbal usage in the Taanach texts, in which he pointed out the pre-Hebrew West Semitic forms and syntax of the Babylonian texts.28 Although Hrozný’s translations indicate that he identified the singular verbal forms fairly well, his comments are rather laconic about their linguistic nature. Based on his background in Semitics, he utilized Hebrew or Arabic etymologies for the understanding of lexical evidence.29 He also made references to the Amarna evidence, then only accessible through Hugo Winckler’s publication of 1896 in Keilinschriftliche Bibliothek.30 Other comparative material was simply nonexistent at the time. The Letters 5 and 6 sent from Gaza show clear Canaanisms (suffix conjugation) and a ductus that is almost identical with those Amarna texts from nearby Megiddo.31 Albert Glock (1971: 23–26 and 1983: 59) provided a first paleographic study and even speculated whether the Megiddo Gilgamesh fragment (dating to the 14th century bce)32 could have been written by the same scribe who wrote the Amanḫatpa letters and at least one list of persons at ­Taanach (but see below). According to him, it is highly probable that “in Megiddo [there] existed a cultural tradition that influenced or even trained” 26 27 28 29 30 31 32

On very interesting remarks concerning Canaano-Akkadian, “the product of the coexistence and interaction of two Semitic languages,” see Vita 2015b: 388–396 with further relevant literature. Rainey 1996. Rainey 1977: 33–64. Furthermore, he applied his knowledge of the Hebrew Bible by referring to biblical texts throughout his accompanying commentary. The standard edition by Knudtzon was published only in 1907. Glock 1983: 59; Vita 2015a: 67–69. Goren et al. 2009: 763–766, Anthonioz 2015: 47.

A Fruitful Collaboration between E. Sellin and B. Hrozný

85

the scribes in Taanach.33 At least later on, during the Amarna period, the scribes of Biridiya, ruler of Megiddo were lent to Iašdata, ruler of Taanach.34 The ductus of letters 1 and 235 is similar to that of those Amarna letters stemming from Gath in southern Canaan, Beth-Shean, Piḫilu and Ḫazor.36 Although the comparisons lead to rather diffuse results and the petrography of the tablets has not yet been analyzed, they all do stem from within the Canaanite boundaries. Only with a few more contemporary tablets might we be able to trace down scribes who may have been occasionally exchanged with other city-states, as shown by Wilfred van Soldt (2013: 26) and Juan-Pablo Vita (2015a) for the Caananite Amarna letters. According to the present view, cuneiform (as litterae francae)37 obviously spread and established itself in Canaan via Ḫazor, an important earlier scribal center in the Old Babylonian period. Questions arose as to how schools of scribes might have developed in this region and if, as Moran (1992: xxi–xxii) suggested, this was a real spoken pidgin and how the practitioners came to speak it: the more so since the personal and geographic names show no reflection of such a pidgin or “interlanguage”—they are purely West Semitic. Those who spoke this tongue obviously kept close to an Old Babylonian dialect, in opposition to the trend to adopt the Middle Babylonian dialects at the expense of the classical tongue. As Moran (1992: xxii) put it: “The language can only be described as an entirely new code, only vaguely intelligible—if at all—to the West Semite because of the lexicon, and to the Babylonian because of the grammar.”38 According to van Soldt (1998: 596), this interlanguage or koine “was most probably not a spoken pidgin, but rather a code developed during more than a century of local school tradition.”39 The scribes obviously translated the words of the Canaanite-speaking ruler into the koine Akkadian in use in Syria and Palestine. This mixed language then became the institutionalized diplomatic 33 34 35 36 37 38 39

Glock 1971: 23. Vita 2015a: 67–68. They present a mixture of standard Old Babylonian style and usage of Canaanized features, such as the West Semitic indicative-energic modal ending or yaqtul preterits and jussives and yaqtula volitives: Rainey 1996: ii 31, passim. Glock 1983: 59–60; Vita 2015a: 69. von Dassow 2004: 670. They are thus best described as Akkadian texts with West Semitic verbal system and word order/syntax reflecting the West Semitic mother tongue of scribes and officials: translated into Canaanite or Egyptian, cf. Ugarit. See also Rainey 1996: ii, 32. von Dassow (2004, 648 and 660) speaks of a “mixed CanaanoAkkadographic code.” For a recent study on this complex issue of Canaano-Akkadian, including relevant research, see also Mandell 2015.

86

Pruzsinszky

language of the Late Bronze Age, which according to Izre’el’s linguistic studies on the Canaanite dialects of the Amarna Age40 is not uniform, but may vary according to geography or the scribal tradition and education of different cities.41 Which background and training the local scribes had and the place of writing in the predominantly oral culture of the time, as well as the identification of single scribes based on palaeographic, linguistic and petrographic analyses, are fascinating topics of ongoing research. Among the sites in Palestine, there surely existed several centers of scribal education. In Megiddo, for instance, a fragment of a tablet containing the Epic of Gilgameš and liver omen texts have been found (see above). Recent petrographic analyses by Yuval Goren, however, reveal that the Gilgameš tablet was probably produced in Gezer, where it was copied from another, original (?), text.42 Gezer seems to have been an important scribal center, from which scribes departed for other Canaanite centers, as Vita (2015a) and Goren et al. (2009) observed on the basis of later Amarna tablets. According to petrographic analyses, the school texts found in Aphek (younger) and Ḫazor (older) were obviously produced locally, while the lexical text of Ashkelon originates from the Lebanese littoral and the alphabetical text from Beth-Shemesh from the southern coastal plain of Israel. Since almost all of the Canaanite scholarly texts were composed within the local Canaanite boundaries and were clearly transported from one place to another, some sort of scholarly exchange must have existed among the Canaanite city-states. Developments in Mesopotamia were transmitted to the scribes of Canaan through intermediaries in the west, like the scribal community of Emar and especially Ugarit, where schooling ­material with West Semitic, Hittite and Hurrian translations has been found. Carole Roche-Hawley (2015: 63–65) has emphasized the role of oral instruction for the more practical needs of the local repertoire, in contrast to the basic school texts that were found in Ugarit, which mostly reflect the classical cuneiform tradition of Mesopotamia.43

40

41 42 43

See Vita 2010: 864–865 for a comprehensive and short summary of various positions. ­ mphasis in research is given to the scribes of these times, who were exchanged among E the Canaanite rulers. Thus, palaeographic studies have become indispensible within the field of Peripheral Akkadian (PA) in the past few years. Cohen (2009) studied the schools in the Western Periphery, focusing on other more affluent sites, such as Emar, Ugarit and Ḫattuša. Goren et al. 2009: 763–773. It is also in Ugarit that the local language was represented in an alphabetic form of the cuneiform script, also referred to as the “heyday of cuneiform script in Canaan” (Horowitz et al., 2018: 18).

A Fruitful Collaboration between E. Sellin and B. Hrozný

4

87

Onomastics and Their Evidence for Cultural Contacts in the Late Bronze Age

With regard to the name lists, several articles have been devoted to the onomastic evidence consisting of about 80 different personal names, the largest number of Late Bronze Age names from a single site in Palestine.44 As early as the late 1920s, Arnold Gustavs published two lengthy articles on the evidence of the personal names for the polyethnic character of northern Palestine.45 Several readings have been improved over the years, and the analyses by Rainey (1977, 1998), Glock (1983), Zadok (1996) and Hess (2003) clearly demonstrate the advances in understanding of the attested languages since the early 20th century. Ran Zadok’s study of 1996, including the unpublished collations by Glock and Gordon, is certainly the most exhaustive, with numerous comparisons demonstrating his profound knowledge of onomastics of the various attested linguistic groups. Zadok’s aim was to draw an ethno-linguistic map of Southern Canaan in the second millennium in which the evidence of Tell ­Taanach is treated carefully alongside that of the Ramesside period. He compiled all the evidence by including his own collations and meticulously corrected Hrozný’s semantic and etymological proposals, which were mostly based on Biblical parallels and readings. For instance, the local ruler’s name written ­syllabically RI-PI-SUR is now to be read Talwašur, as indicated by Egyptian evidence. It may be analyzed as Hurrian Tulwišar, presumably pronounced as Talwašor in Taanach.46 The name of the sender of letter 1 was originally read as Guli-Addu by Hrozný; according to Glock’s collation,47 it may very well represent the Hurrian or hybrid name Eḫli-Teššup or Eḫli-Addu, frequently found syllabically in texts from the Levant. Eḫli-Teššup was an overlord, probably ­residing in Ḫazor. A change in the local population can be postulated taking the evidence of personal names into account: we notice a gradual increase of non-Semitic names in the corpus of Late Bronze Age texts in Canaan.48 In the Ḫazor texts, the personal names are Semitic in character. In the Amarna corpus, however, about one third is non-Semitic. In the lists of persons from Taanach, we notice an even greater frequency of non-Semitic onomastics: about 60% are West 44 45 46 47 48

See Pruzsinszky 2006: 101–117 with relevant literature. Gustavs 1927: 1–18. Pruzsinszky 2006: 107; Berlejung 2006: 230 reads Talwišar. B. Kaldhol (https://www .academia.edu/31789773/The_god_A%C5%A1%C5%A1ur_in_Hurrian_Personal_Names) reads Talwašur and translates this name “Assur is great”. Glock 1983: 59 n. 11. Zadok 1996; Hess 2003: 49–50.

88

Pruzsinszky

­Semitic, while 20% are taken up by Indo-Aryan and Hurrian names each.49 “Thus the significant presence of northern names (mainly Hurrian and some Indo-Aryan)…mixes with the dominant West Semitic presence.”50 This also matches the evidence from the relevant Amarna letters from this region. “The presence of northern names is significant in the regions around the northern Jordan valley, the Jezreel Valley, and its westward reaches inland sites such as Aphek to the south. ... Coastal cities from Byblos to Ashkelon, except for Acco, were not subject to this northern influence and thus this cultural influence of names came through the Beqa and Jordan valleys,” as Richard Hess (2003: 50) concluded in his onomastic survey and which might also be explained by the “Syro-African rift migration” hypothesis of Nadav Na’aman.51 To understand and reconstruct the past, one also needs to look at the material culture and try to study the distribution of people in the landscape, that is, settlement patterns in antiquity. This may be facilitated by the ongoing Jezreel valley regional project,52 a long-term, multidisciplinary survey and excavation project investigating the history of human activity in the Jezreel Valley. A similar holistic approach goes for the understanding of ancient text corpuses. We can only achieve a better understanding by considering their overall context in the chronological, political, historical and linguistic-cultural environment to reconstruct the dynamics of the cultural history of the Late Bronze Age in the Levant. This in turn can only be done by considering the entire text corpus known from this region as it was set forward in the important publication ­Cuneiform in Canaan (Horowitz et al., 2006 and 2018), which will certainly be ­enriched by W. van Soldt’s forthcoming book The Transfer of Knowledge in a Cuneiform Culture.53 Bibliography Anthonioz, St. (2015) Miroir d’une transmission du Gilgameš akkadien en périphérie (deuxième partie), in: C. Roche-Hawley (ed.), 47–58. Berlejung, A. (2006) Briefe aus dem Archiv von Taanach, Texte aus der Umwelt des Alten Testaments N.F. 1/3, 230–235. 49 50 51 52 53

Berlejung 2006: 230. Hess 2003: 50. Na’aman 2012: 8, 11–12, 330, passim. Note especially the leaders’ names in this region, which are either Hurrian or Indo-Aryan. http://www.jezreelvalleyregionalproject.com (accessed 15.05.2017). See van Soldt 2013: 19–31.

A Fruitful Collaboration between E. Sellin and B. Hrozný

89

Cohen, Y. (2008) Review of Horowitz, Oshima and Sanders (2006), BASOR 349, 83–86. Cohen, Y. (2009) The Scribes and Scholars of the City of Emar in the Late Bronze Age [HSS 59], Winona Lake, IN: Eisenbrauns. Eichner, H. (2015) Friedrich Hroznýs Studien- und Arbeitsjahre in Wien (1897–1919), in: Š. Velhartická (ed.), Bedřich Hrozný a 100 let chetitologie, Prague: National Gallery, 82–125. Glock, A.E. (1971) A New Ta‘annek tablet, BASOR 204, 17–30. Glock, A.E. (1983) Texts and archaeology at Tell Ta‘annek, Berytus 31, 57–66. Goren, Y. et al. (2009) A provenance study of the Gilgamesh fragment from Megiddo, Archaeometry 51, 763–773. Gustavs, A. (1927) Die Personennamen in den Tontafeln von Tell Ta‘annek, ZDPV 50, 1–18. Gustavs, A. (1928) Die Personennamen in den Tontafeln von Tell Ta‘annek, ZDPV 51, 169–218. Hess, R.S. (2003) Preliminary perspectives on Late Bronze Age culture from the personal names in Palestinian cuneiform texts, Dutch Studies in Near Eastern Languages and Literatures 5/1–2, 35–57. Hillers, D.R. (1964) An Alphabetic cuneiform tablet from Taanach (TT433), BASOR 173, 45–50. Horowitz, W. (2012) Tall Taʻannak. A. Philologisch, in: RlA 13–5/6, 387. Horowitz, W. and Oshima, T. (2004) Cuneiform tablets from Canaan in the İstanbul Arkeoloji Müzeleri, Colloquium Anatolicum 3, 31–39. Horowitz, W., Oshima, T. and Kreuzer, S. (2004) Die Keilschrifttexte von Tell Ta’annek/ Taanach, in: S. Kreuzer (ed.), 85–97. Horowitz, W., Oshima, T. and Sanders, S. (2006) Cuneiform in Canaan, Cuneiform Sources from the Land of Israel in Ancient Times, Jerusalem: Israel Exploration Society; Hebrew University of Jerusalem. Horowitz, W., Oshima, T. and Sanders, S. (2018) Cuneiform in Canaan, The Next Generation, University Park, PA: Eisenbrauns (2nd ed.). Hunger, H. (2004) Geschichte der Altorientalistik in Wien, in: F. Schipper (ed.), Zwischen Euphrat und Tigris: Österreichische Forschungen zum Alten Orient [Wiener Offene Orientalistik 3], Wien: LIT-Verlag, 19–22. Hrozný, F. (1904) Die Keilschrifttexte von Ta’annek, in: E. Sellin, Tell Ta’annek [Abhandlungen der Kaiserlichen Akademie der Wissenschaften Denkschrift 50/4], Wien: In Kommission bei C. Gerold’s Sohn, 113–122. Hrozný, F. (1906) Die neuen Keilschrifttexte von Ta’annek, in: E. Sellin, Eine Nachlese auf dem Tell Ta’annek in Palästina [Abhandlungen der Kaiserlichen Akademie der Wissenschaften Denkschrift 52/3], Wien: In Kommission bei A. Hölder, 35–41.

90

Pruzsinszky

Kreuzer, S. (ed.) (2006) 100 Jahre Forschungen zur Archäologie, zur Geschichte, zu den Fundobjekten und zu den Keilschrifttexten [Wiener Alttestamentliche Studien 5], Wien, Frankfurt: Peter Lang. Kreuzer, S. (2008) Taanach, https://www.bibelwissenschaft.de/wibilex/das-bibellexi kon/lexikon/sachwort/anzeigen/details/taanach/ch/e42ab6b1a53a6da7a89 b3e0fee96862d/ (accessed 15.05.2017). Kreuzer, S. (2015) Geschichte, Sprache und Text: Studien zum Alten Testament und seiner Umwelt, Berlin, Boston: de Gruyter. Mandell, A.H. (2015) Scribalism and Diplomacy at the Crossroads of Cuneiform Culture. The Sociolinguistics of Canaano-Akkadian, unpublished Ph.D. dissertation, University of California. Moran, W.L., The Amarna Letters, Baltimore: Johns Hopkins University Press 1992. Morandi Bonacossi, D. (2012) Tall Taʻannak. B. Archäologisch, in: RlA 13–5/6, 387–388. Na’aman, N. (2005) Canaan in the Second Millennium B.C.E., Collected Essays, Vol. 2, Winona Lake, IN: Eisenbrauns. Palmer, E. (2012) Ernst Sellin – Alttestamentler und Archäologe [Beiträge zur Erfor­ schung des Alten Testaments und des Antiken Judentums 58], Frankfurt am Main: Peter Lang. Pruzsinszky, R. (2006) Das Onomastikon der Texte aus Tell Taanach, in: S. Kreuzer (ed.), 99–114. Rainey, A.F. (1977) Verbal usages in the Taanach texts, IOS 7, 33–64. Rainey, A.F. (1996) Canaanite in the Amarna Tablets: A Linguistic Analysis of the Mixed Dialect Used by Scribes from Canaan [HdO i/25], Leiden: Brill. Rainey, A.F. (1999) Taanach letters, Eretz Israel 26, *153–162. Roche-Hawley, C. (2015) Mesopotamian cuneiform at Ugarit. Learning versus using, in: C. Roche-Hawley (ed.), 59–66. Roche-Hawley, C. (ed.) (2015) Devins et lettrés dans l’orbite de Babylone. Travaux réalisés dans le cadre du projet ANR Mespériph 2007–2011 [Orient & Mediterranée 16], Paris: Éditions de Boccard. Segert, St. (1999) Bedřich Hrozný and the Bible, ArOr 67/4, 629–648. van Soldt, W. (1998) Review of Rainey (1996), JAOS 118, 595–597. van Soldt, W. (2013) The extent of literacy in Syria and Palestine during the second millennium B.C.E., in: Ll. Feliu et al. (eds.), Time and History in the Ancient Near East, Proceedings of the 56th Rencontre Assyriologique Internationale at Barcelona 26–30 July 2010, Winona Lake, IN: Eisenbrauns, 19–31. van der Toorn, K. (2000) Cuneiform documents from Syria-Palestine. Texts, scribes, and schools, ZDPV 116–117, 97–113. Vita, J.-P. (2010) Scribes and dialects in Late Bronze Age Canaan, in: L. Kogan (ed.), Proceedings of the 53rd Rencontre Assyriologique Internationale, Vol 1. Language in the Ancient Near East [Babel & Bible 4/2], Winona Lake, IN: Eisenbrauns, 863–894.

A Fruitful Collaboration between E. Sellin and B. Hrozný

91

Vita, J.-P. (2015a) Canaanite Scribes in the Amarna Letters [AOAT 406], Münster: Ugarit-Verlag. Vita, J.-P. (2015b) Language contact between Akkadian and Northwest Semitic languages in Syria-Palestine in the Late Bronze Age, in: A.M. Butts (ed.), Semitic Languages in Contact [SSLL 82], 375–404. von Dassow, E. (2004) Canaanite in cuneiform, JAOS 124/4, 641–674. Weidner, E.F. (1952–1953), Dem Gedächtnis der Toten. Friedrich Hrozný, AfO 16, 395–396. Zadok, R. (1996) A prosopography and ethno-linguistic characterization of Southern Canaan in the Second Millennium BCE, Michmanim 9, 97–145.

Part 2 Hittite and Indo-European

∵

Chapter 6

Consonant Clusters, Defective Notation of Vowels and Syllable Structure in Caromemphite Ignasi-Xavier Adiego Abstract One of the typically chaotic situations produced in Carian both by the singularity of the alphabet and by the scarcity of the documentation available is the so-called “defective notation of vowels”: the fact that Carian writing tends to omit the notation of vocalic segments, and does so without any clear pattern and for no apparent reason. In this paper I will try to bring some order into the chaos. I will attempt to show that, in spite of appearances, the Carian words attested in the Memphis-Saqqâra subcorpus offer a very regular and not particularly complex syllable structure, and that the defective notation of vowels can be explained if one interprets these vowels as excrescent vowels, i.e. unstressed and very short vowels of fluctuating timbre that are inserted without affecting the syllable structure of the word. The situation is not so clear in the rest of the Carian documentation, but I will present a first approach in order to establish whether a similar pattern of syllable structures and spelling procedures is at work in other Carian subcorpora.

Keywords Carian – Anatolian languages – Luwic languages – syllable structure – consonant clusters

1 Introduction One of the most challenging tasks in linguistic analysis – and, if successful, one of the most rewarding – is the search for order hidden inside an apparent ­chaos of forms. In the case of Carian, one of the typically chaotic situations produced both by the singularity of the alphabet and by the scarcity of the documentation available is what I called (in Adiego 2007: 239–242) the “defective notation of vowels”: the fact that Carian writing tends to omit the notation

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_008

96

Adiego

of vocalic segments, and does so without any clear pattern and for no apparent reason. This omission is obvious from even the briefest analysis of Carian words and sequences (i), but it can also be verified (ii) by comparison with Greek adaptations of Carian and (iii) thanks to some internal alternations ­detectable in the Carian corpus (Adiego 2007: 254–255). (i) pntmunś, idmns, tbridbδś, pnlδśwl, ýnsmsos, emsγlpn, idrayridsemδbq, mtKelŋ, trqδimrqrdstazomδ[, smδýbrs, ?/obsmsmñðñ/?… (ii) dquq – Ιδαγυγος kbjomś – Κεβιωμος ksolbś – (place name) Κασωλαβα msnordś – (place name) Μασανωραδα ntro – cf. Νετερ-βιμος, Lyc. Natrbbijẽmi plqo – Πελ(λ)εκως qlaλiś – Κολαλδις, Κυλαλδις qtblemś – Κυτβελημις, Κοτβελημος qýblsiś – (ethnic) Κυβλισσ//ος// śuγλiś – (place name) Σουαγγελα tñuś – Τοννους

(iii) βrsi – iβrsiś – iβarsiś dwśoλś – iduśolś kbidn – kbdynš pikrmś – pikarmś pnuśoλ – punwśoλś (Πονυσωλλος) psmšk(wneit) – psmaśk – pismašk / pismaśk šr(quq) – šar(kbiom) (Greek Σαρ-) trqδ- – trquδe cf. Lycian Trqqñtttbazi – ttubazi

This “defective notation” is also a major obstacle to the study of phonotactics and syllable structure in Carian. In this study, I will try to bring some order into the chaos. I stress that the study is preliminary, and as such, its scope is quite narrow: I have limited myself to a Carian subcorpus comprising the inscriptions (mostly funerary stelae) from Saqqâra, a necropolis of Memphis (hence the denomination “Caromemphite” used here). The reason for my choice is obvious to anyone who knows Carian inscriptions: this subcorpus is the most extensive and consistent of all the Carian documentation, presenting a highly standardized version of the writing, without significant variation. This feature was decisive for the decipherment of Carian alphabet and is likely to be very useful for a study such as the present one. It is not clear that the results obtained from this subcorpus can be easily extrapolated to the rest of the Carian documentation, but at least I hope to show that Caromemphite presents clear patterns of syllable structure that can be recognized if some basic and straightforward principles are accepted.

Consonant Clusters, Defective Notation of Vowels

2

97

Methodological Assumptions

The first methodological assumption is the differentiation between V and C. V represents the syllable peak or nucleus, and C the remaining segments. In the first phase of my analysis, only a, e, i, o, u, j, w, y, ý will be parsed as possible syllable nuclei; nasals and liquids will be excluded. In a later phase, I will introduce the analysis of some of these nasal and liquid segments as syllabic, and assign them the label V. The reason for this questionable procedure is that there are no graphic differences between syllabic and nonsyllabic liquids and nasals. Therefore, it is advisable initially to consider all the liquid and nasal segments as purely consonants, and then carry out a different analysis at a later stage. I recognize that the use of V and C to represent the syllable nucleus vs. the nonnuclear segments may seem rather confusing, since one might tend to simply associate V with “vowel” and C with “consonant”, but I think that it is a very straightforward way of indicating the different analyses proposed here. Henceforth, then, V = syllable nucleus or peak, C = other segments. A serious methodological problem in parsing syllables is how to analyze adjacent vowels. Here, I take only one vowel as the syllabic nucleus (V), namely the more open one, when a sequence of two adjacent vowels is found. The preceding and following vowels are thus classified as C. When they are preceding, they can be considered prenuclear glides; when they are following, they can be taken as the second element of a rising diphthong. The only examples of three adjacent vowels in our subcorpus are the following four words: kojoλ

paraeym

yjas[iś]

yiasi

In kojoλ and yjas[iś], where the intermediate vocalic element is j, it is obvious that this j functions as a consonant onset, and therefore the syllabic parsings ko.joλ and y.ja.s[iś] are assumed. This analysis can be extended to yiasi, in which case i can be considered an alternative spelling for j. More complicated is the case of paraeym. Although a triphthong cannot be ruled out, I propose to parse this form as a trisyllabic word pa.ra.eym, in which the hiatus between a and e coincides with the boundary between the two elements of the compound (clearly, we are dealing with a compound name with the well-known stem para- as first element, cf. Adiego 2007: 393). In fact, this latter form introduces another important problem: the possibility of hiatus between two adjacent vowels. Cases like mdayn or arliomś are in

98

Adiego

principle highly ambiguous: they admit /mda.yn/ or /mdayn/, /ar.li.omś/ or / ar.liomś/ as parsings. Fortunately, such examples are not particularly abundant. In general, a nonhiatus solution seems more plausible, particularly when the adjacent vowel is i, because this alternates with j (most probably a glide sound), although, regrettably, no examples of this alternation are documented within the Caromemphite corpus: note, for instance, kbjomś in our corpus vs. šarkbiom in Sais. However, the situation is not always so simple. For instance, in a name like paraiβreλ we might think of a diphthong /ai/ or alternatively consider that the compound boundary between para and iβreλ points to a hiatus parsing. This latter possibility would be backed up by the analysis of paraeym (pa.ra.eym) proposed above. In any case, the presence vs. absence of hiatus does not greatly affect our analysis, where, as we will soon see, the sonority hierarchy is of major importance. Adjacent vowel sequences, if interpreted as belonging to the same syllable, never violate the sonority hierarchy. It should also be noted that, with few exceptions, adjacent vowels are combinations of a high vowel and a low vowel (or vice versa), according to the following classification:1 high vowels: {u, i, w, j, y, ý} low vowels: {a, e, o} Therefore we find [low+high] (ai, ay, au, aw, ey, ow, etc.) and [high+low] (ja, io, ýa, ye, etc) sequences. Examples of [low+low] sequences are very few in number. I have found only two in our subcorpus: aorś and the compound paraeym mentioned above. Note that aorś is an Egyptian name (Adiego 2007: 353, following Vittmann), and that for paraeym a hiatus analysis pa.ra.eym seems preferable, as noted above. I think that this preference for avoiding two low vowels in contact favors the interpretation as diphthongs and points to a tendency to avoid hiatus.2 The examples of [high+high] combinations of adjacent vowels are somewhat more numerous, but their distribution is interesting. In principle, there are twenty theoretically possible combinations:3 1 Obviously, this classification assumes that y, ý represent a [y] sound, as I have defended in Adiego (2007). 2 The scarce presence of sequences of low vowels – ea, eo, etc. – is a general feature throughout the Carian corpus. 3 I leave aside combinations of the same timbre, as *yy, *yý, *uw, *uu, etc. In fact, there are no examples of these combinations in the Caromemphite corpus.

Consonant Clusters, Defective Notation of Vowels

99

iu iw iy iý ui uj uy uý ju jw jy jý yi yu yj yw ýi ýu ýj ýw But in reality, our subcorpus only offers examples of three types: iu, yi, and ýi.4 iu: piub[a]ziś yi: yiśbiksś ýi: šýinś Note that the example of iu poses problems of reading; what is more, it appears in an Egyptian name. If the value [y] of y, ý is assumed, the sequences yi, ýi point to a fronting change u > y triggered by the following sound, the front vowel i, as suggested summarily in Adiego (2007: 257). An analysis of the situation in the complete Carian corpus is beyond the scope of the present work, but a detailed study of all combinations of high vowels gives strong support to this fronting change. In the present work, sequences of [high+high] vowels will be taken as forming a diphthong. The problem is to ascertain which of the two high vowels functions as syllable peak. In the case of iu, this is not particularly relevant, but in ýi, yi it is more important because the way these sounds function may help us to identify the exact nature and origin of y and ý. I tend to think that here i functions as syllable peak, because it is i that affects the articulation of y, not the other way around, but this is a rather impressionistic interpretation. The opposite parsing cannot be ruled out, nor even the possibility of the coexistence of both parsings (/yj/, /ɥi/): compare for instance Spanish, where /uj/ and /wi/ coexist: muy /muj/ vs. fui /fwi/. In a purely conventional mode, I analyse yi, ýi as a rising diphthong in which i functions as syllable peak. To sum up, I assume that all sequences of two vowels, of which one is low and the other is high, or both are high, preceded and followed by consonants or by word boundaries, may be treated as diphthongs or sequences of vowels plus glides. Therefore, I include them in pre- or postnuclear C-slots. Finally, a further methodological problem is raised by the exact status of j, w. As we will see soon, the Caromemphite subcorpus clearly suggests that j is a glide. The word Kj is the only example in which j must be analysed as a syllable peak, and this word is a highly aberrant spelling documented only in one 4 A sequence yj appears only in yjas[iś], analysed above as heterosyllabic (/y.ja.siś/).

100

Adiego

­inscription. In all of the other numerous instances of the word in Caromemphite, we find the expected spelling Ki. Less clear is the situation of w (see the detailed analysis in Adiego 2007). Insofar as we find it in contexts where an interpretation as syllable peak is apparently inevitable, I will treat it as V in these contexts. However, below we will observe some important restrictions on its distribution compared to the other vocalic signs. 3

Theoretical Assumptions

3.1 Sonority Hierarchy and Sonority Sequencing The concepts of the sonority hierarchy and sonority sequencing are crucial in the syllable structure of the world’s languages. The sonority hierarchy establishes a ranking from the most sonorous to the least sonorous sounds. There are various proposals for the sonority hierarchy, both universal and languagespecific, but the discrepancies do not affect the sequence of the major classes of sounds. Therefore, the following hierarchy can be taken as universal: low vowels > high vowels > liquids > nasals > fricatives > affricates > stops. The Sonority Sequencing Generalization establishes that, in a given syllable, sonority gradually increases from the beginning to the syllable peak or nucleus, where the most sonorous segment of the syllable appears, and decreases (also gradually) from this nucleus to the end, always following the sonority hierarchy. Otherwise defined, in Blevins’s words: Sonority Sequencing Generalization (SSG): Between any member of a syllable and the syllable peak, a sonority rise or plateau must occur. blevins 1995

Certainly, violations of this sonority sequencing can occur, the most typical being the behavior of s, which in many languages (English or Latin, to cite only two examples) can occupy positions on a word boundary (English stops, Latin stare, princeps), but the SSG is generally observed in the world’s languages. 3.2 Syllable Typology Another important theoretical issue in this study is the typology of syllable structure and the classes of segments which function as syllable nuclei. Languages show variation regarding these two parameters. As for syllable structures, we find a spectrum ranging from languages that admit only very simple

Consonant Clusters, Defective Notation of Vowels

101

structures (CV) to languages with very complex consonant clusters, both in the beginning (known as the Onset of the syllable) and in the ending (the Coda). It will be interesting to establish the number of consonant segments that are permitted in onset and coda position in Caromemphite. There is also considerable variation among the world’s languages regarding which classes of segments function as syllable nuclei, but significantly, it generally obeys the sonority hierarchy. For instance, whereas in Sanskrit only vowels and liquids can be syllable nuclei, in reconstructed PIE this possibility extends to the following segment class in the sonority hierarchy, i.e., nasals. An important discussion in phonology is whether any class of segments can play the role of a syllabic peak. In descriptions and analyses of different languages, we find that fricatives and even stops can take on a syllabic function. This question is closely connected with the discussion of the existence of languages with “syllable-less” words. Perhaps one of the best known examples of this is the Salishan language Bella Coola, in which there are words that lack vocalic and sonorant segments.5 Scholars have debated whether in these words fricatives and stops are syllable nuclei, or whether these words have no syllable structure at all in Bella Coola. As this study aims to show, Caromemphite is not, despite appearances, one of these particularly idiosyncratic languages, but offers rather more frequently attested, “normal” syllabic patterns. 4

Analysis of the Subcorpus

After these methodological and theoretical prolegomena, I will analyse the words contained in the Caromemphite subcorpus. I exclude truncated forms, admitting only those whose completion is highly probable and also some examples in which the missing letters do not notably affect the analysis: for instance, in [-]owtś only the initial C segment is missing, and so the complete word can be easily analysed. 4.1 “Vowel-less” Words The Caromemphite subcorpus offers only two apparent examples of “vowelless words”, sb and mHmś. But note the following particularities: (1) sb “and” is a proclitic conjunction, as the interpunction clearly shows in the sole example we have of this word in our subcorpus: paraeym : sbpolo.

5 Note for instance: łxwtłcxw “you spat on me”, xscć “I am now fat”, xłp̓ x̣ʷłtłpłłskʷc̓ “he had had in his possession a bunchberry plant”.

102

Adiego

Therefore, I will take sbpolo as one word in which sb together with p makes up the onset of the first syllable. (2) mHmś contains two instances of m. As we will see in §4.3, this sound can function as a syllable nucleus in Caromemphite. This word poses the additional problem of our ignorance of the exact value of H. Also isolated are the examples of words where the only vocalic sign is j or w. I have already mentioned the unique case of j in Kj. As for w, we have only the words w (in E.Me 13) and pnlδśwl. In all three of these cases, I accept a value of j, w as syllable nuclei. Beyond these five cases, all Caromemphite words contain at least one of the following vocalic sounds: a, e, i, o, u. Significantly, the subcorpus does not contain unambiguous examples of the use of y, ý as the only syllabic sounds of a word (note only the case of šýinś, already mentioned above). 4.2 Words with One Vocalic Syllabic Nucleus As stated above, in my initial approach, I analyze only the vocalic segments a, e, i, o, u, y, ý, j, w as syllable nuclei (V). As noted above, there are no examples of y, ý as sole vocalic segments in a word, so in this section dedicated to (apparently) monosyllabic words, y and ý always appear preceding or following the syllable nucleus and are therefore classified as C. Table 6.1 shows all the words in the corpus that contain one vowel. I note violations of SSG in the onset by X, and violations of SSG in the coda by Y. I exclude from these violations cases in which the sound is a sibilant (s, ś, š) ­occupying the edgemost position in the onset or in the coda. Table 6.1 Words containing one vowel

C

C

C

C

C

V

K K s u ś

w e o i i j a e e

C

C

C

C

SSG violations

n r d

ś m

n

ś

Y

n

103

Consonant Clusters, Defective Notation of Vowels

C

p

C

p n

C

p m p t k p š š m p t n l

C

C

V

C

k m z k l m m p š k t p l r t q b s d r s K b t δ

t β m q w q ý [-] p b n m w k d d λ ý s r j q s t t j m τ q n s r m ś

e e e u e a a o i o o u a o a a e i o i a o i u e o a a u o i i u w

d m ŕ q t r s w k s ś ś r r y ý q n l e b ś ś s m ś/š t q r m d n l

C

C

ś ś ś ś m δ t r

ś ś ś m

ś ś n n ś ś b l r

ś k ś ś d t b ś

C

SSG violations

Y ś

Y

X X X ś ś m

ś ś δ

Y X X

X X ś

Y X X

104

Adiego

Note the following particularities: (a) Nucleus particularities: – As mentioned above, examples of w, j as nuclei (V) are extremely rare, and there are no examples of y, ý as nuclei (except perhaps the ambiguous šýinś). (b) Onset particularities: – The number of C in the onset oscillates from 1 to 5, but 5 and 4 are exceptional: pnlδśwl is the sole example of 5 C, and pntmunś is the sole example of 4 C. Both cases contain liquids and nasals among the prevocalic consonants (on this, see the following paragraph). – Violations of SSG in the onset: we find nine violations of SSG (in 43 words), which seems a fairly high number. But there is one most remarkable point: in all nine of these violations, liquid, nasal and/ or δ segments are involved: mrsiś, lkorś, mdayn, mdaýn, pnlδśwl, etc.6 (c) Coda particularities: – The number of coda segments oscillates between 1 and 4; for 4 C we have only three examples: idmns, pikrmś, tbridbδś. – There are only three clear examples of SSG violation (idmns, pikrmś, pjabrm), and two doubtful ones (tbridbδś, ýasδś). The doubts in these latter cases are posed by the exact status of δ, a question to which we will turn below (§4.4). Not surprisingly, these certain and doubtful examples of SSG violation include the three cases in which the coda has four segments. Another significant point is that these words coincide in having a liquid, nasal or δ involved in the violation. 4.3 Words with One Proper Vowel and One Syllabic Sonorant Leaving aside the cases of sibilants, the conclusion is clear: all the apparent examples of coda and onset violations of SSG involve the presence of a liquid, nasal or δ (on which see §4.4 below). The most logical solution, then, is to accept the existence of syllabic nasals and liquids or, at least, the presence of unwritten supporting vowels in the articulation of nasals and liquids in these positions. In favor of this analysis is the alternation of pikrmś with pikarmś, or of šr- in šrquq with šar- in šaruśoλ, šarwljatś. By admitting syllabic nasals and liquids, the syllable structure of these words becomes simpler and more in line with the SSG: 6 A difficult form to interpret is šdτatś, due to the doubts about the precise value of τ. In any case, the violation of SSG by š is a typical case of sibilant eccentricity.

105

Consonant Clusters, Defective Notation of Vowels Table 6.2 Words containing one vowel and one syllabic sonorant

C

p

C

V

C

C

V

C

C

C

d

p p š p j

i m m m l i r n a

m d d n q k q m b

n a a o o r u u r

s ý y r

n n d

ś

s

t

m q n m

ś ś ś

4.4 “Double” Consonants (δ, β, γ, λ, z) Contra Adiego (2007) and Kloekhorst (2008) I am now convinced that δ represents /n.d/, where /n/ may or may not be a syllabic nasal. The same can be said of β = /m.b/ (note iβrsi = /im.br.si/, βrsi = /m̥ .br.si/) and of γ = /n.g/. I think that it is possible that Carian had biphonemic heterosyllabic graphemes. Kloekhorst (2008) argues against the existence of biphonemic graphemes on the grounds that the Carian writing system had individual graphemes for the sounds involved in biphonemic sequences. However, this argument is not compelling: the Greek alphabet also had biphonemic graphemes such as or representing biphonemic sequences for which individual graphemes were also available (+, +). Moreover, it is probable that /b/ in = /m.b/ was not exactly identical to the sound represented by (a fricative /β/?), and similarly in the case of vs. . Perhaps the same biphonemic explanation can be extended to λ, which may represent /l.d/ (or /l.l/?), with the possibility that the first /l/ represents a ­syllabic liquid. Finally, a similar analysis may be suggested for , for which a  value [s.t] seems probable. Therefore, we may propose the following equivalences: δ β γ z λ

= /n.d/, /n̥ .d/ = /m.b/, /m̥ .b/ = /n.g/, /n̥ .g/ = /s.t/ = /l.d/, /l̥.d/ or /l.l/, /l̥.l/

106

Adiego

This interpretation allows us to give a straightforward analysis of words like tbridbδś, ýasδś where the number of alleged coda consonants is unusually high and/or violates SSG. If analysed as /tbridbn̥ dś/, /ýasn̥ dś/, the resulting forms become typical examples of disyllabic words with more typical consonant clusters and no violations of the SSG:7 Table 6.3 Analysis of tbridbδś and ýasδś

C

C

C

V

C

C

V

C

C

t

b

ý r

a i

d

s b

n+ n+

+d +d

ś ś

This leads us to reconsider the analysis of the words βemś, pλeqś, and zmuś in Table 6.1 in which “double consonants” are also involved: Table 6.4a Analysis of βemś and pλeqś

C

V

C

V

C

C

p

m+ l+

+b +d

e e

m q

ś ś

Table 6.4b Analysis of zmuś

C

C

C

V

C

s+

+t

m

u

ś

If the hypothesis of the double consonantal value of , and similar letters is assumed, I think that the analysis of the two words in Table 6.4a, in which the sonorant parts are parsed as syllable nuclei, is preferable. An onset / mb/ in βemś would violate the SSG. However, in pλeq a /pl:/ onset (with a sort of 7 In this table, I represent as n+d. Similarly, in subsequent tables, is represented as n+g, and as m+b, as l+d and as s+t.

Consonant Clusters, Defective Notation of Vowels

107

long or geminate l) cannot be entirely ruled out. As for Table 6.4b, if = /st/, the resulting onset /stm/ violates the SSG only in the position of the sibilant – a violation which, as we have seen, is acceptable in Carian. 4.5 Words with One Proper Vowel and Two Syllabic Sonorants An exceptional form is pnlδśwl, in which we must assume the existence of two syllabic sonorants: n and δ = /n̥ .d/ (in accordance with §4.4). Under this analysis, the syllable structure of the words become much simpler: Table 6.5 Analysis of pnlδśwl

C

V

C

V

C

C

V

C

p

n

l

n+

+d

ś

w

l

Syllable Structure: a First Conclusion 4.6 Considering Table 6.1 and the revisions of Tables 6.2 to 6.5 together, we find that the onset in the beginning of a word consists of a maximum of three consonants. In fact, there are few examples of this maximal onset; moreover, some of the examples are Egyptian names like psmaš/śk, and perhaps also pKsimtś and pttuś. The SSG in these initial-word onsets is generally fulfilled. The same  can be said of codas in word-final position: the maximum number of ­consonants is three, but this number is only obtained if the last segment is a sibilant ś. The situation in the interior of words is similar: we find a maximum of two consonantal segments. We are not sure where the boundaries of syllables fell, but it is clear that at least in Caromemphite, the onset and/or the coda can have a maximum of two consonants in word-internal position. 4.7 Words with Two Vowels Although we have seen that we must accept the existence of syllabic liquids and nasals, we continue to use the same methodological procedure to analyse Caromemphite words. In this section, we take into consideration words containing two plain vowels. As we will shortly discover, some (though not many) two-vowel words also contain syllabic liquids and nasals and therefore must be parsed as trisyllabic words.8 8 The distribution of internal C between syllables presented here is purely speculative. For the present discussion, the only important point is the number of C in internal clusters.

108

Adiego

Table 6.6 Words containg two vowels

C

C

C

C

C

V

p s s τ K l l n š p p s q q q k d m

u u w u u w i a a a w a u w a a a i i o e i a a o ý a a i i o a a ý i w w

C

q k k p r k r r r r r d d

r u r y

C

C

s s

s s

K l m

K k k

m r p r s b l n+ +d b

C

V

p p p s m m r l m i m m l ś l i t y u l m r m n b s i r r r n s i s s ś d

a e e i u u o o e e? u o e o i o a e o o w a o o a i a i a e e i o i i o o

C

C

w s n ś ś n ś l+ š m y s n

+d ś ś ś ś ś ś

ś l u s w ś s q ś ś ś ś [?] ś ś l d n ś

C

X ś

109

Consonant Clusters, Defective Notation of Vowels

C

p

C

s

C

C

C

V

p t t t m

m n s š τ p š t p p y t t r p d š r n t p s q l [q?] l k l p n t d q d t n k r t b ŕ K š k

w i a e a a a e i a i a i u w o i a a o u u a u a a a w

C

C

C

r r r r ś

l+ l+

r

k k

C

V

C

C

C

s q n n n m n ý k p b š q b l r r +d +d r ś ś ŕ p β z t n

a a u u a a a e a e i u o i i i o i i u o o o i i i a e

t u q r i ś i z r y k b

ś ś ś t ś ś ś ś m m s t

ś ś ś ś

ś ś s ś ś s l+ l+ l+ u

X

+d +d +d ś

ś [ś] [r] ś i t

X X

Of these 68 words, only four present a clear violation of the SSG in the onset, and all four cases again involve nasals and liquids: qýblsiś, tŕKata[r]ś, ntokris, and psmškwneit. Analysed as trisyllabic words with a syllabic sonorant, all four words offer a straightforward syllable structure:9 9 Against this parsing of psmškwneit one might argue that a form pismašk is attested in the rest of the Carian corpus. But (1) we are dealing with a borrowed name, so the adaptation may present alternations; (2) in this case, unlike the rest, a compound name is present.

110

Adiego

Table 6.7 Words containing two vowels and one syllabic sonorant

C

p

C

V

C

V

C

C

V

C

C

q t s

n ý ŕ m

t b K k

o l a w

k

r s t n

i i a e

s ś [r] i

ś t

š

Similar conclusions may be drawn from Tables 6.6 and 6.7 as from Tables 6.1– 6.4: maximal clusters of three members in the onset at the beginning of words and in the coda at the end of words, and fulfilment of the SSG. Perhaps the most striking new feature is given by the form ursKleś, the only example in which a sequence of four interior consonants must be assumed. But it is very interesting that the sound s appears in a central position in these clusters. Both a parsing /ur.sKleś/ and a parsing /urs.Kleś/ are possible, and clearly support the idea that in Caromemphite, as in many languages, s can occupy an exceptional position in the coda or in the onset. We therefore accept that in Caromemphite an internal cluster of four consonants may appear if at least one of them is a sibilant. 4.8 Words with Three Vowels The words that contain three plain vowels (or diphthongs) are represented to the right in Table 6.8.10 No cases of SSG violation are attested in this list. Consonant clusters are limited to a maximum of three segments both in initial and in final position, and even so there are very few such clusters. The only example of a 4 C cluster in internal position is paraiβreλś, if a parsing as /pa.raim. breλś/ is accepted. But as mentioned above, in this case a hiatus between the two compound elements para- and iβreλś should be assumed. Alternatively, we may attribute the exceptional size of the cluster to the presence of the diphthong ai. In all cases, the resulting parsings respect the SSG and produce onsets and codas of a typologically normal size.

10

Note moreover that the Greek adaptation of the name (Ψαμμήτιχος) points to an articulation in which p and s are in the onset. For the first word, ynemordś, I follow Schürr’s correct reading of the form (contra Adiego 2007, †ynemoriś).

111

Consonant Clusters, Defective Notation of Vowels Table 6.8 Words containing three vowels

C

C

k

C

V

C

C

š p š p p u

y y w y i a a a u a a a a

n i n j t r d [r] [b] r n r r r r

V

C

e a u a u y i u w w a a i

C

C

C

V

C

C

C

d

ś

i ś

m+ +b r b

o i i [i] o y a o o a e e a

r

l

m s t s r b k ś ś j

ś [ś] w r r l+ l+ t y l+ r

ś ś m ś +d +d ś ś m +d ś

4.9 Special Cases In accordance with §4.5, śuγλiq, śuγλiś must be analysed as containing respectively /n+g/ and /l+d/ or /l+l/. The problem is that we have no way of establishing whether /λ/ here represents /l̥.d/~/l̥.l/ or rather /l.l/. In the latter case, the onset would be formed by a cluster /gll/ (with a consonantal long, i.e. geminate /l/?), as shown in Table 6.9a. I think that the alternative analysis (with /l̥/ and three syllables per word) is to be preferred (Table 6.9b). Table 6.9a Analysis of śuγλiq, śuγliś

C

V

C

C

ś ś

u u

n+ +g n+ +g

C

C

V

l+ l+

+l i +l i

C q ś

Table 6.9b Alternative analysis of śuγλiq, śuγliś

C

V

C

C

V

ś š

u u

n+ n+

+g l+ +g l+

C

V

+d i +d i

C q ś

112 5

Adiego

Preliminary Conclusions about Syllable Structure in Caromemphite

(1) From the data examined, we can establish the following pattern for Caromemphite syllables: (C1)(C2)(C3)V(C4)(C5)(C6), - where C6 always = ś, and - where the SSG is fulfilled with the exception of sibilants. This pattern shows the possible onsets at the beginning of the word and the possible codas at the end. Word-internally, there is a maximum of four segments between syllable peaks. This means that, ruling out impossible parsings like VCCCC.V and V.CCCCV, in word-internal position as well there cannot be more than three C per onset and/or per coda. For instance, in ursKleś all possible parsings – except for the two impossible ones mentioned above – follow this restriction: …VCC.CCV… (…urs.kle…), …VC.CCCV… (…ur.skle…), …VCCC.CV (…ursk.le…). A similar situation is visible in the other example of a 4 C internal cluster, paraiβreλś. Here the possible parsings are …VC.CCCV… (…ai.mbre…), …VCC.CCV… (…aim.bre…) and …VCCC.CV… (…imb.re…). (2) There are no properly “vowelless” words. The only example, sb, is in fact a proclitic word that can be interpreted as a part of the onset of a complex: /sbpo.lo/. (3) Therefore, an independent word must have at least one vowel. Curiously, in monosyllabic words, this vowel is practically always one of the following: a, e, o, i, u. w only appears as a solitary vowel in the isolated word w, and j in the aberrant form k̑j. In the case of y, ý, we only have the very ambiguous example šýinś, in which y may or may not be the syllable nucleus (cf. supra §2). In monosyllabic words, no syllabic nasals or liquids appear either. (4) Point (3) may be extended to polysyllabic words, where we systematically find a, e, o, i, u at least in one of the syllables. Examples of polysyllabic words without one of these vowels are very scarce: mHmś, psHým[-]ś (note that one letter is missing, so the example may not be properly ­usable), pnlδśwl. (5) Sibilants: as in many languages, Carian sibilants do not conform to the SSG: they can occupy the edgemost position both in the onset and in the coda. (6) Syllabic nasals and liquids – or at least nasals and liquids accompanied by unwritten supporting vowels – seem to exist in Caromemphite; when we assume their existence, the syllable parsing becomes clear and consistent.

Consonant Clusters, Defective Notation of Vowels

113

(7) If we assume (5) and (6), the onset and coda respect the SSG. It must be said, however, that some “heavy” clusters of stops must be accepted. (8) Certainly, some of these “heavy” clusters appear in names borrowed from Egyptian (ptnupi-, ttbazi-, pttu-, pKsimtś). Are we dealing with structures atypical for Carian which the writing is attempting to reflect, or were these structures acceptable in Carian and incorporated into the language without difficulty? (9) Our analysis has raised the possibility that δ, β, γ, λ are properly to be analysed as heterosyllabic biphonemic graphemes. According to this view, they would represent /nd/, /mb/, /ng/, /ld/-/ll/, in which the first element could function as syllabic nucleus in certain contexts. Similarly, z may also be a biphonemic grapheme representing /st/. 5.1 Proper Vowels Conclusions (2), (3) and (4) merit further reflection. They point to a crucial differentiation between the vowels a, e, i, o, u and the rest of the sonorant ­segments. I will call this inventory of segments “proper vowels”. In the case of liquids and nasals this comes as no surprise, as we are dealing with basically consonantal sounds which in some languages can also function as syllabic nuclei. More important, in my opinion, are the consequences of conclusions ­(2)– (4) for the exact status of j, w, y, ý in Caromemphite, a very controversial question. It is clear that these signs represent sounds whose appearance must have been bound to particular conditions, because this is the only way of explaining their distribution. Apart from their use in contact with proper vowels as glides or members of diphthongs, with the exception of j11 they also seem to occupy syllabic positions. Let us briefly discuss some aspects of w and y, ý. 5.2 The Sign v Interestingly, most examples of the sign w appear in nonfinal syllables. The sole exceptions are the words semwś and pnlδśwl. Among the examples of apparent use of w as syllabic nucleus in nonfinal syllables we find the curious case, already noted in Adiego (2007), of the genitives wśoλś and compounds, in clear contrast to the use of u for the corresponding nominative. In our subcorpus, this contrast is present in the pair pnuśoλ and punwśoλś – a particularly striking case, because it also suggests an alternation between pn- and pun- at the beginning of the word. In Adiego (2007: 242) a first explanation was suggested: a displacement of  accent and use of w to represent /wV/. A shift of accent would be fully 11

As repeatedly mentioned, the sole example is the exceptional form k̑j.

114

Adiego

­consistent with the results of our analysis of syllable structure in Caromemphite: the fact that, except in extremely rare cases, w cannot appear alone in monosyllabic words and must be accompanied by a proper vowel in polysyllabic words points to a close connection between this sign and unstressed ­position. It seems therefore that w represents either directly or indirectly an unstressed vowel. In Adiego (2007: 242) I suggested an indirect solution (w = /wV/), but now I would not wish to rule out a direct solution (w = unstressed u). One way to reconcile the two possibilities, at least diachronically, would be the following: could represent a graphical reminiscence of /wV/, although would not represent a longer /w/, but simply (unstressed, also short?) /u/. There is a similar situation in Etruscan, in which v appears in certain contexts where it seems to represent /u/ because the vowel that followed it is missing: the articulated plural genitive -σvla vs. plural nominative -σva can be explained if behind the spelling -σvla one sees the previous form *-σvala (< -σva + genitive ending -la).12 It is probable that in Etruscan σvla sounded more like /šu.la/ than /šwla/, which contains a very strange cluster. Similarly, we can imagine that in punwśoλś perhaps we have simply /pú.nŭ.śoλś/ as the result of a previous /pún. wə.śóλś/, by admitting a reduction process wə > /ŭ/ following the process wá > ú (or perhaps ū́?) in the nominative. This hypothesis could explain (1) the ­existence of a specific grapheme for /w/ in Caro-Egyptian, (2) the somewhat chaotic use of this grapheme in archaic Carian and (3) its disappearance in ­later Carian alphabets: if the process of wə > u was generalized, the ­replacement of w by simple u would have contributed to the loss of this letter. 5.3 The Signs W , w Even more complicated is the case of , . First, the relationship between the two letters remains a mystery to me. No clear distributive principle can be established prima facie, and the low number of attestations of both letters does not allow a thorough study. We must therefore consider them jointly, as if they represented the same thing. The fact that they do not appear as the only vowels in Caromemphite words seems to point to a similar analysis as that provided for w, that is, that they are unstressed vowels. However, this cannot be clearly affirmed, for a variety of reasons. First of all, in the form šýinś, although I have conventionally assumed a parsing as ý = C, i = V, the fact that both ý and i represent high vowels makes 12

See Adiego 2006: 34.

Consonant Clusters, Defective Notation of Vowels

115

it impossible to establish which vowel functioned as syllable nucleus, as I noted above. Secondly, in some inscriptions that present an alphabet very close to the Caromemphite alphabet, examples of ý as the sole vowel in a monosyllabic word are attested: týn, ýbt. In this highly provisional study I have left out these other materials, but it is impossible not to mention them. Finally, I have already referred to the fact that the number of examples is too limited to draw reliable conclusions. In any case, the tendency of y, ý to appear in immediate contact with other vowels is clear: out of 18 examples of y, ý in Caromemphite, 13 are attested preceded or followed by another vocalic (or semivocalic) sign. I cannot add much more about y/ý. 5.4 A Question of Stress The fact that we have many examples of disyllabic words in which one syllable nucleus is occupied by one of the proper vowels, whereas a syllabic liquid or nasal appears in the other nucleus, points clearly to the existence of a stressed vs. an unstressed syllable. The study of Caromemphite stress remains beyond the scope of the present study, but it is a very interesting aspect which should be studied in the future. 6

The Problem of the “Defective Notation of Vowels”

The present study has allowed us to establish that Caromemphite had a rather regular and not excessively complex syllable structure, in any case not very different to that of (for instance) Latin. This conclusion is reached by simply assuming the existence of syllabic liquids and nasals. Certainly, there are some examples of stop (or stop+fricative) clusters (such as ttbazi[ś], sbpolo), but even in such cases, the SSG is not greatly violated and the restriction on the number of C segments allowed in the onset and coda is respected. Neither “vowelless” words nor extraordinary clusters are present. However, the establishment of a regular and unexceptional syllable structure in Caromemphite raises an intriguing question: how should we interpret the so-called defective notation of vowels? Clearly, we cannot claim that this phenomenon existed in Caromemphite on the phonetic level; if it had, we would have found a large number of unanalysable consonant clusters. In other words, although the supposed vowels do not appear, the extant sequences of segments constitute permissible syllable structures. This is an unexpected finding in a writing system in which vowels are omitted either completely (as for instance in Phoenician) or partially (as supposedly in Carian).

116

Adiego

But it is also clear that in Caromemphite we find examples of this “defective notation of vowels” if we make comparisons with Carian names attested in Greek sources. Let us go back to one of the examples mentioned at the beginning of this paper. The name qlaλi- is attested in Greek sources as Κολαλδις, Κυλαλδις. It is a typical example of a name in which the appearance of a vowel between /k/ and /l/ in the Greek version cannot be easily attributed to its adaptation to Greek, because the cluster /kl/ exists in this language. At this point, I think that it is very interesting to introduce the differentiation between excrescent and epenthetic vowels, established by Levin (1987) and adopted by, among others, Bagemihl in his fundamental article on Bella Coola (Bagemihl 1991). Although the framework of Levin’s article is the Generative Phonology of the 1980s, the concepts can be easily applied to our discussion. Two distinctive traits of excrescent vs. epenthetic vowels from Bagemihl’s summary are relevant here (Bagemihl 1991: 600): Excrescent vowel: (a) “[its] quality is variable, generally determined by phonetic coarticulation effects (either language-particular or universal) and frequently tends toward schwa”; (b) “its insertion is triggered not by unsyllabified consonants but rather by the need to mediate a transition between adjacent articulations requiring some degree of constriction in the oral tract”. Epenthetic vowel: (a) it generally has “a nonfluctuating quality”; (b) its “insertion is conditioned by unsyllabified consonants”. Hall (2011: 1584–1585) exemplifies excrescent vowels by means of the short vowels that occur in consonant clusters in Piro, an Arawakan language spoken in the Western Amazon. She mentions four other traits of excrescent vowels in Piro which are highly relevant to our discussion: – the presence of the excrescent vowel varies with the “syllabification” of a consonant, as in /whene/ ‘child’, phonetically realized as [whene], [wəhene], [wohene], [wïhene], [wuhene]; – excrescent vowels cannot bear any kind of stress; – excrescent vowels are of much shorter duration than lexical vowels; – in Piro orthography, the excrescent vowels are not written. One aspect which has a strong bearing on our discussion is the fact that o/u in Κολαλδις/Κυλαλδις corresponds closely to these traits of excrescent vowels: the back and/or round quality of the inserted vowel seems to be determined by the

Consonant Clusters, Defective Notation of Vowels

117

articulation of the preceding consonant, q, a uvular or labiovelar sound that triggers similar insertions of back round vowels in other words. The oscillation between o and u also points to the fluctuating nature of the sound and also perhaps to a very short duration. Finally, Carian orthography does not reflect this sound, and consequently it was hardly ever a stressed vowel. This o/u sound could be inserted in order to mediate the transition between q and l. But the most interesting point is that this insertion was not determined by structural reasons. In other words, it was – or could have been – inserted in an already well-constructed syllable structure. We can therefore imagine that qlaλiś, a syllabically well-formed word, was phonetically articulated with the (variable) insertion of a short excrescent vowel whose quality was determined by the preceding stop: /qlal.diś/ [qolal.diś] In fact, supporting vowels accompanying syllabic liquids and nasals can also be interpreted as excrescent vowels, as they do not affect the syllable structure. pλeqś šrquqś plqo pikrmś ksolbś kbjomś pnuśoλ śuγλiś qýblsiś ntokris (Egyptian)

/pl̥.deqś/ /šr̥.quqś/ /pl̥.qo/ /pi.k̥rmś/ /ksolbś/ /kbjomś/ /pnu.śold/ /śun.gl̥.liś/ /qý.bl̥.siś/ /n.to.kris/

[pel̥.deqś] [šar̥.quqś] [pel̥e.qo] [pi.karmś] [kasolabś] [ke?bjomś] [ponu.śold] [śun.gel̥.liś] [qý.bl̥.siś] [ni?.to.kris]

cf. Πελδηκος cf. Σαρcf. Πελ(λ)εκως cf. pikarmś cf. Κασωλαβα cf. Κεβιωμος cf. Πονυσωλλος cf. Σουαγγελα cf. Κυβλισσ//ος// cf. Νίτωκρις

The following objection to this analysis might be raised: in some cases, we know or can imagine that the excrescent vowel corresponds to an originally existing vowel. For instance, in sb=polo, we can parse /sbpo.lo/, with a complex cluster, and might postulate one or two excrescent vowels for the complex onset created by the univerbation of the proclitic conjunction and the word: [s(e)b(e)po.lo]. However, is this not the original situation, as shown by Milyan sebe, and partially also Lycian se? I think that we can assume that excrescent vowels in cases like this are either the result of a reduction of an original vowel to a schwa-like sound that does not count in the syllable structure, or the result of an insertion after reduction.

118

Adiego

In fact, there are cases in which it is clear that the excrescent vowel is not etymological: in šar-/šr-, we can imagine a šr̥ from *sr(i). This form was articulated alternatively with an excrescent vowel /a/ as šar-. This vowel is perhaps a case of the tendency to phonologize the excrescent vowels coming from r̥ and make them epenthetic, which would explain the frequency of this vowel even in Carian writing. Also in the case of kbjomś- Κεβιωμος, the kb-cluster is p­ ossibly original if it is directly derived from a Luwic /hwiyamma/i-/ “running, runner, attendant” per Melchert (2013, 25), but this is a very hypothetical etymology. Another interesting possibility is that Carian may have had alternant forms, in which the proper vowels alternate with absence of vowels, which eventually gave rise to excrescent vowels. The example of pn- vs. pun- thus perhaps reflects a wider phenomenon in Caromemphite and in Carian as a whole. The proper vowels of some forms might also determine the timbre of the excrescent vowels inserted in a consonant cluster in other forms of the same paradigm: punuśoλ vs. punwśoλś. This imitation of timbre can be explained either as analogical or as a result of the reduction process that led to the (practically total) loss of the unstressed/short vowels. 7 Conclusions In spite of appearances, Caromemphite words offer a very regular and not particularly complex syllable structure, with a maximum of three consonants in the onset and the coda, and a degree of fulfillment of the Sonority Sequencing Generalization (SSG) similar to other languages in which only sibilants behave in a particular manner. This analysis requires only that we accept the existence of syllabic nasals and liquids in Caromemphite. The above analysis creates a problem: if Caromemphite, as it is spelled, offers words with well-formed syllable structures, how can we explain the socalled defective notation of vowels? A possible solution is to interpret these vowels as excrescent vowels in the sense of Levin, Bagemihl, Hall and others, i.e. unstressed and very short vowels of fluctuating timbre that are inserted without affecting the syllable structure of the word. These vowels may be either the result of a reduction in unstressed positions of actual vowels – which would explain certain correspondences with other Luwic languages – or a mere insertion of a transitional sound in inherited consonant clusters. These sounds were not usually written in Caromemphite, but they do appear in Greek adaptations of Carian names. Further research is needed in order to establish whether a similar pattern of syllable structures and spelling procedures is at work in the rest of the Carian documentation.

Consonant Clusters, Defective Notation of Vowels

119

References Adiego, I.-X. (2006) “Observaciones sobre el plural en etrusco”, in G. del Olmo, L. Feliu and A. Millet (eds.), Šapal tibnin mû illakû, Studies Presented to Joaquín Sanmartín on the Occasion of His 65th Birthday, Sabadell: \, 1–13, reproduced in: A. Ancillotti and A. Calderini (eds.) L’umbro e le altre lingue dell’Italia mediana antica, Perugia: Jama, 2009, 29–41. Adiego, I.J. (2007) The Carian Language [HdO I/86], Leiden, Boston: Brill. Bagemihl, B. (1991) Syllable structure in Bella Coola, Linguistic Inquiry 22/4, 589–646. Blevins, J. (1995) The syllable in phonological theory, in J.A. Goldsmith (ed.), The Handbook of Phonological Theory, Cambridge, MA, Oxford: Blackwell, 206–244. Hall, N. (2011) Vocalic epenthesis, in M. Van Oostendorp et al. (eds.), The Blackwell Companion to Phonology, iii, Cambridge, MA, Oxford: Blackwell, 1576–1596. Kloekhorst, A. (2008) Studies in Lycian and Carian phonology and morphology, Kadmos 47, 117–146. Levin, J. (1987) Between epenthetic and excrescent vowels (or what happens after redundancy rules), in M. Crowhurst (ed.), Proceedings of the West Coast Conference on Formal Linguistics 6, Stanford: Stanford Linguistics Association, 187–201. Melchert, H.C. (2013) Naming practices in second- and first-millennium Western Anatolia, in R. Parker (ed.), Personal Names in Ancient Anatolia, Oxford: Oxford University Press, 31–49.

Chapter 7

Tagging and Searching the Hittite Corpus Dita Frantíková Abstract This article discusses the issues accompanying the creation of an electronic search tool for the Hittite corpus. The research is based upon the project “Electronic Corpus of Hittite”,1 which entails creating a working frame for inclusion of texts, tagging and searching. The descriptive part of the working frame is not the core of the presented text; instead, after a general introduction, the paper focuses on selected problems of the electronic treatment of ancient cuneiform texts. Part 1 exposes the limits of digitizing the Hittite texts. Part 2 focuses on tokenizing and tagging, and Part 3 on the area of electronic search. Of particular interest is the choice of search algorithms: standard search methods are based on comparing letters, and the syllabic script has to be tested for accuracy of such implementation. As full-text search is an insufficient tool for Hittite, word approximation is necessary. Special issues of the data structure (such as ­encoding of clitic chains) are also discussed. As the author is not aware of any fundamental description to date of encoding cuneiform languages, the presented article covers a large area of study, which leads to restrictions in detail but lays the foundation for further digital analysis.

Keywords Hittite – smart search – corpus – tokenizing clitics – digitizing cuneiform

1 Introduction 1.1 From Cuneiform Tablet to Smart Search Texts preserved to us through the centuries and millennia, written on perishable material which may undergo damage, captured in graphemic systems familiar to (at best) a group of local scribes, require special attention. Ever since 1 The work on digitizing the Hittite corpus is in progress at the Institute of Comparative ­Linguistics, Charles University, Prague.

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_009

Tagging and Searching the Hittite Corpus

121

the Brown corpus (Kučera and Francis: 1967), it has proved obvious that linguistic work cannot proceed without digitally elaborated texts. The computerized corpus of a language is meant to serve as an instrument which provides comparable accurate data and can be a starting point for linguistic inquiry. It is not meant to be a goal of the effort, but rather a shared platform of scientific research. The issue of providing a digital tool for the study of an ancient language is a complex task. The overall work covers questions of digitizing, annotation and analysis. Each of these fields may profit from the achievements of current corpus linguistics, but certainly cannot apply its methods without making adjustments. In this article, I tackle some of the problems encountered while processing Hittite for computer use. Besides briefly commenting on general issues of digitizing and annotating as well as searching, I focus on the problems of digitizing enclitic chains (§2.1) and lemma ambiguity (§3), and introduce the problem of smart search in Hittite (§5) with discussion of the suitability of common string matching algorithms (§5.2). 1.2 The Three Milestones: Digitizing, Annotating and Searching Text digitizing is a vital step in the development of Hittite studies. Recently, attempts to create online text databases have come into focus. While such a database is not presently available, there are several research centres that are attempting to provide one. Until this day, readers of Hittite texts must depend on printed editions such as Studien zu den Boğazköy-Texten, edited by the Akademie der Wissenschaften und der Literatur Mainz, and Dresdner Beiträge zur Hethitologie, which from 2002 until today has issued transliterations of the Keilschrifttexte aus Boghazköi (KBo) and Keilschrifturkunden aus Boghazköi (KUB); both of these editions provide transliterated texts. (The autographs of the cuneiform tablets have been published in several collections and are identified by a unique number.) The free online sources, such as Textzeugnisse der Hethiter, provide only the tip of the iceberg of the sheer volume of Hittite texts, spanning a wide range of genres preserved on several tens of thousands of tablets.2 To search the corpus for a specific problem, the researcher must depend on their knowledge of texts (hardly detailed for the number of tablets) and their own digitized texts, stored typically as a Word editor file, which they may have collected or copied themselves. Digitizing thus clearly emerges as a 2 Aside from the rituals, the Hittite text files made available by the titus project (Thesaurus Indogermanischer Text- und Sprachmaterialien, http://titus.uni-frankfurt.de/) are restricted to members.

122

Frantíková

­ riority for the further technology-oriented development of Hittite linguistic p studies. Token annotation is the next step, one which cannot be undertaken without many language-specific decisions. The questions of annotating are complex and require detailed investigation. The syntactic annotation of Hittite texts following the framework of Universal Dependencies is in progress at the University of Pavia (Inglese 2016) and provides useful background for annotating an ancient cuneiform language. The desired levels of tagging are manifold; their achievement is based not only on current possibilities, but also on the quality of material and possibilities of reaching conclusions on selected features of the texts. Searching Hittite texts requires a so-called smart search, where the user need not give an exact match of the searched text. For Hittite, a smart search is only superficially a matter of “going the extra mile”. In fact, without it, the digitized search tool remains only a weak support of a dictionary volume search. The originals of the texts in question are ancient tablets, mostly broken, with a high number of incomplete clauses and incomplete or only partially readable forms. At the same time, the Hittite corpus is diachronic in nature, covering four centuries of writing development. In such a case, a plain full text search cannot yield satisfactory results. It must be made possible to find incomplete forms as well as forms with variant spellings. 2 Digitizing The scribes who created the Hittite texts wrote in cuneiform script; the texts were recorded (mostly) on clay tablets. The most accurate sources of data are therefore the tablets themselves. The desired outcome of text processing (parts of which are already undertaken by different scientific institutions)3 is to have quality tablet photos alongside their transliterations and translations, and possibly autographs. The available material is extremely extensive; as stated of cuneiform artefacts by Cammarosano (Fisseler et al. 2014: 171), “in view of 500.000 fragments of clay tablets, the availability of fast, reliable and automatic methods is urgently 3 Cf. the project of the Akademie der Wissenschaften und der Literatur in Mainz, Das Corpus der hethitischen Festrituale: staatliche Verwaltung des Kultwesens im spätbronzezeitlichen Anatolien; the Annotated Corpus of Luwian Texts (by I. Yakubovich, in progress); the Old Hittite Treebank (Inglese 2016); the Leiden Corpus of Hittite Texts (by A. Kloekhorst, in progress); and others.

Tagging and Searching the Hittite Corpus

123

Figure 7.1 Tablet Kt 94/k 581b. (Kültepe 1994, excavated in the kārum (lower town), published by M. Larsen as akt 6a 9b in The Archive of the Šalim-Aššur Family, Volume 1: The First Two Generations [Kültepe Tabletleri vi-a], Ankara: Türk Tarih Kurumu Basımevi, 2010.)

­ ecessary”. Digitizing cuneiform texts has recently progressed for Akkadian, n where the digital database of cuneiform texts, both tablet images and their transliterations with translations, is soon to appear online.4 For Hittite studies, the primary goal is to digitize the multiple available transliterations. At present, tablet autographs are often the only source of the original cuneiform script. Tablet photos available online often do not provide sufficient standards of readability and high quality. Easily legible pictures, such as the one in the example below, are (so far) scarce.

4 The Old Assyrian Research Environment, part of the Online Cultural and Historical Research Environment hosted by the University of Chicago at ochre.lib.uchicago.edu.

124

Frantíková

Preparing an autograph is not simply a mechanical enterprise, but rather one that requires the author’s mental input. As some signs and above all, partly broken signs may be understood in various ways, creating an autograph will always be a scientific interpretation of the text (not to mention the possibility of mistakes). This holds all the more for transliterations; the reader inevitably relies on their accuracy. Even if produced with the best possible quality, because of the need for disambiguation on certain points (e.g. whether a Sumerogram is functioning as a logogram or a determinative), interpretation cannot and should not be avoided in every case. Thus, when digitizing the Hittite corpus, the raw data cannot be compared to modern corpora where one can rely upon the language capacity of the creators and users of the corpus. At present, it appears sensible to include available transliterations and allow for inevitable corrections at a later date. The layer of photos, autographs and automatic recognition of cuneiform signs (i.e., programmes that can rewrite signs from a photograph) mark the next logical step. A layer of transcription can then be created automatically. As this process is unidirectional (as different transliterated signs are represented in the same manner, transliteration is not recoverable from transcription), these two levels are not interchangeable. 2.1 The Marking of Clitics Clitic chains, attached to the first word of a sentence, are a frequently encountered sentential element of Hittite. The clitics are ordered in a specific ­sequence (Laroche 1958: 161, Garrett 1990, Hoffner and Melchert 2008: 410, Kloekhorst 2014). The denotation of clitics is one of the major problems in digitizing Hittite transliterations (Inglese 2016: 13).5 The handling of clitics in transliteration can take several forms (in transcription, they may easily be separated conventionally by the equal sign, e.g. nu=nnaš). One option is not to isolate clitics as such in the script at all. Thus, one would write e.g. nu-un-na-aš and the tokens would be separated by an advanced tool which, based on certain criteria, would be able to suggest the position of the word split and create individual tokens nu and (u)n-na-aš while preserving the information that they form a unit. This is not a recommended procedure as, because of the variation in analyzing the enclitics, the amount of manual work would still be very high. The other option is to mark up clitics in the text. The convention is to use either the “­Frakturhyphen” of CHD or the equal sign “=” (e.g. Melchert 2001). In case of a 5 The denotation of clitics determines proper tokenization, as well as proper treatment of the cuneiform signs.

Tagging and Searching the Hittite Corpus

125

sign which is an outcome of a phonemic merger, inserting “=” in the “middle” of a sign is the rule, i.e. nu-u=n-na-aš or ma-a-n=a-aš as opposed to nu=uš. In such cases, the unadjusted automatic split (where the equal sign marks up as a split sign) creates such tokens as e.g. ma-a-n and a-aš. However, there is no cuneiform sign n, nor is the word -aš preceded by a vowel sign a (its transcription would than falsely yield āš). These inaccuracies lead to various adjustments: for example, following the practice of some cuneiform databases one can create on the one hand an epigraphic version of the text, and on the other hand a discourse version in which linguistic information is coded, separate from the epigraphic. While the best adjustment of the algorithm is up to the decision of the analyst, the following “illegal” results must be avoided: – creating signs which are against cuneiform rules, nonexistent or at least not present (e.g. n); – preventing signs from functioning as a single unit (e.g. allowing for u=n to function as un); – preventing words from being accurately full-searched, based on the peculiarity of clitic notation (e.g. un-na-aš stands for the lexeme nnaš/naš); – creating false transcriptions (e.g. by allowing for a word split a-aš, which yields āš in transcription). A suggested treatment which returns valid results is to adjust texts so as to avoid splitting signs. As the reason for splitting signs by the equal sign is pragmatic (to allow for easier search in the text editor), this can well be avoided and rearranged as shown in the following table: With this adjustment, the forms preserve their original orthography, as long as they retain the equal sign as an integral component. When lemmatized, they thus retain the feature “clitic”, and any specifics (e.g. uš-ši for šši) can be explained as the result of their syntactic position. Table 7.1  Treatment of clitic chains for tokenizing purposes

Clitic chain

Unadjusted transcription

Adjusted chain

Adjusted tokens, transcription

n=a-at ú-et=ma=at n=a-aš nu-u=š-ši n=a-an=kán

n; āt uet; ma; at n; āš nū; šši n; ān; kan

na=at ú-et=ma=at na=aš nu=uš-ši na=an=kán

na=; =at uet=; =ma=; =at na=; =aš nu=; =ušši na=; =an=; =kan

126

Frantíková

2.2 Typographic Issues The unification of typographic style requires a set of transliteration standards. As different authors and editions use different signs mainly for the metadata, there arises a need to employ such a standardized mark-up which will enable most readers to use it without specific training. Special codes have to be used for text borders, paragraphs, denotations of broken signs, etc. Sumerograms and Akkadograms have to be made graphically distinct from Hittite words, as is the convention in transliterations. 3 Tokenizing Text tokenizing is a prerequisite to search, translation and tagging. The problem to deal with in the case of cuneiform material, beside incomplete tokens, is the absence of unified orthography, which makes it more difficult to interpret the text properly. As the basic unit of annotation is the syntactic, not orthographic word (UD: 2016), decisions have to be made in the area of clitic chains (see below). Otherwise, Hittite does not tend to aggregate forms; the negation particles are an extra unit (Hoffner and Melchert 2008: 341 with refs.), and compounds are exclusively lexical (Hoffner 1966). From the perspective of the coding recorded by a Hittite scribe, the spoken language was captured with a graphotactics partly based on phonological perception and partly on other features (tradition; scribal school). The morphological layer was not the decisive factor for the method of recording. The effort to interpret such records from a morphological perspective may thus well yield ambiguous results, as discussed below in the case of clitic chains. There are specific reasons for lemma ambiguity in a syllabically rendered language (Krebernik 1996: 234) as compared to a modern language corpus. These include form parallelism and semantic parallelism, as well as the inconsistent representation of word forms, with inconsistent transliteration on top of that to boot. On the other hand, form parallelism is not a major issue in the Hittite corpus. Although expected for a morphology-marking language (Kaalep 2012: 85), it is mainly avoided due to the insufficiency of data. Most morphological forms are preserved in a handful of (or even one) attestations, and some forms are typically missing. Parallel lexical roots are usually realized in different morphological forms. The following example shows both the variability of writing and the overlap of the verbs pai- “go, pass” and pai-/pi- “give, pay”:

127

Tagging and Searching the Hittite Corpus

Table 7.2  Selected morphological slots of Hittite verbs pai- “go, pass” and pai-/pi- “give, pay”

Morphological form

pai- “go, pass”

1SG.PRES.ACT

pé-e-eḫ-ḫi, pé-eḫ-ḫi, pé-e-eḫ-ḫé pa-it-ti, pa-iš-ti, pé-eš-ti

2SG.PRES.ACT 3SG.PRES.ACT 1PL.PRES.ACT

2SG.PRET.ACT 3PL.PRET.ACT

pai-/pi- “give, pay”

pa-i-mi, pa-a-i-mi, pa-a-mi pa-i-ši, pa-a-i-ši, pa-it-ti, pa-a-ši pa-a-i, pa-i pa-iz-zi, pa-i-iz-zi, pa-a-iz-zi pí-i-ú-e-ni, pí-ú-e-ni, pí-ya- pa-i-wa-ni, pa-a-i-wa-a-ni, u-e-ni, pí-i-ya-u-e-ni pa-a-i-wa-ni, pa-i-ú-wa-ni, pa-a-i-u-e-ni- pa-i-u-e-ni, pa-a-u-e-ni pa-it-ta, pé-eš-ta pa-it-ta pí-i-e-er, pí-i-er, pí-e-er, pa-a-er, pa-i-er, pa-er, pí-ya-er, pí-er pa-a-e-er, pa-e-er

As this example shows, the heterogeneity of the records is high, and so is the heterogeneity of morphological form: overlap is found only in some forms of the second singular of both present and past forms (marked in the table in bold). As for semantic parallelism, the disambiguation of ambiguous forms in the case of Hittite is a matter of informed decision based on textual analysis. As the most frequent words of a language are as a rule polysemous (Agirre and Edmonds 2007: 1), disambiguation must be based on the environment of a particular word form. Without doubt, this task will eventually find support in computational techniques, but at present and for the near future, it must depend on the researcher’s understanding of the context, genre and usage of a given lexeme. Furthermore, the problem of text translation is not so much the problem of unknown lexemes, but rather of multiple meanings for a given lexeme that make text interpretation so difficult. Incomplete clauses are another frequent obstacle for semantic interpretation in a text-corpus language such as Hittite. A specific problem for Hittite is the tokenizing of clitics. As shown in the previous section (§2.1), clitic chains must be properly digitized. Cuneiform orthography poses numerous challenges in this regard: e.g. the orthography of nu-uš-ši seems to imply that such a chain of signs stands for nu+ušši, but in fact it is the rendering of nu+šši. Likewise, one must take into account consonant

128

Frantíková

mergers, e.g. iš-ki-še-et representing iškiš+šet, not iški+šet; at-ta-aš-mi-iš representing attaš+šmiš, not atta+šmiš. If one strives to retain a single textual layer (not keeping an extra epigraphic and discourse layer), it proves useful to retain “=” as a part of the token (disregarded when dealing with individual signs) and keep the word boundary at the sign boundary, as e.g. iš-ki=še-et. The morphological form (in this case, nom.sg.) will be disambiguated on the tagging level. However, as it is important to retain the graphemic representation of tokens, it is advisable to add a special sign denoting a “missing” part, e.g. iš-ki-{iš}=. It is obvious that such insertion will require an interpretation which, as with any interpretation, introduces the risk of error. 3.1 Treatment of Text in a Token Editor The following example shows the level of preservation of a prototypical Hittite text. Incomplete or added tokens are given in italics: 1[ ]URU[ 2[ °-a]n LÚ URUU[r-šu-° 3[ an-d]a=pát IS.-BAT ša=a[n 4 na=an nam-ma ḫu-ul-le-et ši-°[ 5 ša=an tu-ut-tu-ut a-aš-šu=uš-še-et [da-a-aš 6 [ku]-it-ki ú-e-mi-et x[ 7 [ ]MÁŠ ŠEŠ=ŠU URUḪa-[at-tu-ša u-da-aš 8 [ÉRIN.MEŠ URUZa]-al-pa ú-et nu URU[ 9 [mA-nu-um-ḫ]é-er-wa+a-š=a pa-it [ 10 [ t]i a-x[ The example shows that in ten lines, eighteen tokens are complete; the rest of the text is to some extent unreliable. Some tokens are supplied and cannot be used for a search, while others may be reconstructed with some level of reliability (e.g. kuitki). The following image shows the way a transliterated text may be handled in a token editor.6 The editor automatically splits the text (the full text is recorded in the box on the right side) into lines (on the left). When a discrete line is unfolded, tokens are split into sub-lines. In the first column of the TableEditForm, each token is given its label, e.g. PreD (pre-determinative), PostD (postdeterminative), blank (word), brackets (supplied). The value is paired automatically and, if necessary, edited manually. The second column of the TableEditForm contains the respective token. In the third column, following the token, it is assigned a quality label: “unreliable”, “unreadable” or “ ” (i.e., 6 The software solution, including the editor, is part of the outcome of the project “Electronic Corpus of Hittite”, which will be made available by Charles University.

Tagging and Searching the Hittite Corpus

129

blank); other ­labels may be added on demand. In the fourth column of the editor, words of “blank” quality are paired with a lemma.

Figure 7.2 Transliterated text in token editor

Tokenized Corpus Usage 3.2 The tokenized text may be utilized for the following purposes: Table 7.3  Employment of tokenized corpus

Token search (Ctrl+F equivalent) Lexeme search (repeated Ctrl+F equivalent) Morphological search Textual analysis Morphological analysis

Looking up all occurrences of a specific word form (e.g., attaš) Looking up all occurrences of all forms of a lemma (e.g., attaš, attan, attuš etc.) Looking up occurrences of specific morphological feature (e.g., genitive singular; apposition) Identification and examination of the properties of textual information (frequencies, keywords etc.) Analysis of word formation and grammatical structures of words.

130

Frantíková

Table 7.3  Employment of tokenized corpus (cont.)

Byproduct: automatic dictionary Study tool

As tokens are aligned to lemmas, an automatic dictionary is produced within the process. The tokenized corpus is a perfect aid in understanding words and contexts.

The token search conducted upon digitalized and tokenized texts can be developed in multiple ways, as discussed in detail in §5 below. Not only can it search for exact matches as a regular text finder in a text editor, but it also allows for approximation and provides suggestions for further usage. 4 Annotating The primary goal of annotating a corpus (diachronic as well as synchronic) is to link a range of tags and metadata to the original text. It should help to retain linguistically relevant information from the original, e.g. information on the script and preservation quality (DIAKORP 2016). The choice of tagging layers is optional; tagging can either focus exclusively on grammatical (morphological) information or provide additional data (syntactic layer; information on text genre; information on the original text; image of the original; layer of cuneiform script). The Golden Rule of morphology states that the pair serves as a unique identifier of each word form (Hlaváčová 2009: 6). It leads to such conclusions as preferring unclassified tags to ambiguous tags or tags based on the interpretation of the annotator. While most attributes are not part of the text itself and do not interfere with text counts, some of the tags must receive their own text position, as do the word tokens, and consequently participate in the statistics. They include at least tags that indicate breaches, added information (if considered relevant)7 and clitic boundaries. The need may arise to somehow capture scribal errors or other unusual unexpected graphic features. It has proved useful to use a special attribute which shows the original writing, captured in any useful way without a need for

7 The author of the corpus has the choice of including or excluding incomplete text; either possibility has to be tagged with an internal tag.

Tagging and Searching the Hittite Corpus

131

s­ pecial restrictions, e.g. /aš, where the brackets enclose an emended piece of text and the original sign is found transcribed after the slash. As for syntactic annotation, the scheme for handling fragmentary sentences is discussed by Inglese (2016), following Zemánek (2007), Giusfredi (2015), Molina and Molin (2016) and others. He suggests a unified account of the annotation of fragmentary sentences (p. 61), within which he aims at reducing the level of speculation over possible syntactic relations of missing or broken tokens. Although the issue is studied in detail, for many features of Hittite no satisfactory theoretical solutions have been reached so far. This covers mainly syntactic annotation of structures with damaged or missing valence heads. It is difficult to retain a reasonable level of simplicity while at the same time tagging the quality and reliability of broken or missing tokens. 4.1 Analysing Hittite Morphology for Annotating Purposes Morphological analysis aimed at tagging text tokens works with isolated word forms, regardless of their context; its outcome are morphological indices. The other part of the result is a lemma, which identifies the abstract lexical unit, in the sense of unambiguous identification of its dictionary entry. In general, morphological analysis cannot avoid ambiguity; many tokens cannot be properly tagged without considering their context, semantic and syntactic. 4.2 Tag Positions The basic unit of the corpus is a document , which can be split into texts (every document contains at least one text) and further into paragraphs

and sentences . Of these, metadata is usually connected to the document and text. For a document, they are above all the title, Id number, CTH number, publication Id, text type, genre, historical period (OH/MH/NH), and type of script (NS/MS/OS). The tags assigned to the texts are imprinted to all of its tokens. Every tag identifying a token is represented by a string of indices. While it is reasonable to keep them at a manageable level, it is certainly desirable to occupy as many slots as possible, as they become the data for searches based on grammatical and other categories. For Hittite, the positions may supply the information displayed in the following table: Note on position 1, “word class”: Beside the prototypical word classes (noun, adjective, verb, pronoun, number, particle [including the negation particles], interjection, conjunction, adposition, adverb, preverb, and the class of determinatives), it is advisable to introduce three new categories: oscillating word

132

Frantíková

Table 7.4  Tag positions for the Hittite corpus

Example position in a tag

Information

Position 1 Position 2

Word class (core word class of the token) Subposition (global detailed category relevant for the whole paradigm, eg. deverbative substantive) Gender, nominal (relevant for substantives, adjectives, some numerals, some pronouns, some infinite verb forms) Number (relevant both for nominal and verbal declensions) Case (relevant for nominal declensions and also for adpositions, where it denotes their valence possibilities) Possesum gender Possesum number Person (relevant for pronouns and verbs) Tense Grade (relevant for adjectives and adverbs) Verb form (relevant for verbs and uninflected infinite verbals; includes mood and type of non-finite form) Voice (relevant for verbs and participles) Enclitic (relevant for members of the enclitic chain) Textual completeness of the token Aspect (relevant for verbs)

Position 3 Position 4 Position 5 Position 6 Position 7 Position 8 Position 9 Position 10 Position 11 Position 12 Position 13 Position 14 Position 15

form8 (e.g. adposition/preverb or substantivized adjective, in cases where it is not possible to make the distinction or where the distinction is unnecessary), foreign word (this category includes only words that are not adapted to the Hittite morphological system) and unknown words (those that cannot be classed under any other label; they may be misspelled words, ciphers etc.). ­Hittite words coded by special orthography – the logograms – are coined with

8 This distinction helps to identify cases of homonymy, and also helps to specify statistics which are influenced by different opinions of the annotators (Osolsobě et al. 2017: 225–234).

Tagging and Searching the Hittite Corpus

133

the tags appropriate to Hittite morphology (e.g., an Akkadian genitive form denoting the Hittite nominative is tagged as “nominative”). The descriptive issues of logograms must be handled in another layer of annotation. 5 Searching Since the Archie search engine of 1990, web search software has been one of the fastest developing and most discussed (Brin and Page 1998; Lawrence and Giles 1999). From the original full text search, the focus is on exploring the user’s behaviour and adjusting the results based on anticipation of the searched item. The following picture illustrates the connection between the original text and its search. The information from the original tablets is converted into a printable version of autographs and/or transliterations. Such data serves as the source of the corpus, which in turn is the source of the dictionary. When the texts of the tablets are tokenized (i.e., when word forms are separated), they may be annotated in multiple layers. The basic layers are composed of the morphological

Figure 7.3 From table to smart search

134

Frantíková

and the lemma layer. Based on the annotated data, it is possible to perform statistics and analyses of different kinds. On the web interface, the user can enter queries of different kinds and perform textual analysis. To match the expectations of the inquirer, the search must enable advanced functions other than a plain exact search. It is thus quite important to employ a so-called smart search. On the surface, a smart search provides the user with a single, easy-to-use search field into which the user enters their queries without prior understanding of the system’s operation. On a deeper level, the search combines multiple functions which foresee the user’s behaviour and contribute to obtaining the proper results. 5.1 Justification for Smart Search In working with the Hittite corpus, one encounters alternations in writing conventions within a single lexeme (see Table 7.2, the example of pai-/pi, above). Some changes are diachrony-based (the texts span approximately 400 years), but other factors are also at work: the possibility of recording a unit in several ways, the native language of the scribe and not least the existence of erroneous spellings. Attestations of a unit (word) with equal grammatical categories often differ in form, as shown above for verbs. This is also true for other categories, e.g.: palḫi- ‘wide, broad’ nom. pl. commune: pal-ḫa-a-e-eš, pal-ḫa-a-eš, pal-ḫa-e-eš, pal-ḫi-e-eš In the case of palḫi-, the reasons for the different spellings may be understood in diachronic terms. But regardless of the cause, the Hittite corpus abounds in variant spellings and therefore it is necessary to employ a smart search tool. While the easiest way is to search by lexeme and thus obtain a reference to an index, this would require a faultlessly tagged corpus for lexemes. The results of such a search would thus be only partially productive.9 The user of a digitized corpus should not be forced to enter an accurate query, but rather should receive suggestions which allow for specification of the original query not only by means of translation but also based on grammatical categories, lexeme index or their respective combinations (e.g. lemma + grammatical category).

9 While incomplete words in general must be omitted from the search for their unreliability, there are cases when e.g. only the ending is missing or part of one of the signs is broken. In such cases, the criterion of incompleteness would rule out much relevant material.

Tagging and Searching the Hittite Corpus

135

5.2 The Problem of String Matching The type of searching algorithm has to be chosen based on the character of the corpus; in our case, a diachronic corpus, written in a syllabic script, and in an extinct language, which implies that it is being searched by a nonnative speaker. To maximize the usefulness of the search, it is necessary to use an approximate string matching algorithm (i.e., string matching algorithm with errors), where the data need to be analysed for preprocessing. The first problem is the question of the character of the script. When working with a transliterated string, the search can be run based on letters (including or excluding separators) or based on syllables, which better corresponds to the original writing. In the course of its history, Hittite employed over five hundred signs; they are transliterated for the use of modern science by as few as sixteen graphemes of the Latin script (A, E, H, I, K, L, M, N, P, Š, T, U, W, Z) plus two which are found in transcription of Sumerograms (B, S). For search by a sign, every sign can be appointed a unique value (e.g., a number) and as such be used in search algorithms. The problem of such an approach, although seemingly closest to reality, is that the edit distance (i.e., the minimum number of edits required to convert one string to another) between e.g. nu and na is the same as between e.g. nu and šu (i.e., maximal). Such a search yields very vague results. Example: Edit distance,10 letter and syllable based pal-ḫa-a-e-eš x pal-ḫi-e-eš (both “wide, broad”, nom. pl.) Letter-based (including separators): 3/13 Letter-based (without separators): 2/9 Syllable-based: 2/5 As is shown in the example, the absolute edit distance is either 3 or 2, while the relative edit distance (i.e. with regards to the number of units) shows great differences (3/13, 2/9; 2/5). The letter-based transformation captures the relative closeness of the two forms, while the cuneiform sign-based transformation equates (in the sense of necessary edits needed in order to obtain pal-ḫa-a-e-eš) the word pal-ḫi-e-eš to e.g. ḫar-ša-a-e-eš “thick-bread”, nom. pl. ­Consequently, a

10

The edit operations used for string transformation as understood here are insertion, deletion and substitution. In this elementary example, they are used in their simplest version; each operation counts as one unit. In other algorithms, the operations are typically defined with a weight function, e.g. wins(x).

136

Frantíková

search based on syllables proves less useful than one based on letters. (The role of separators will be discussed below.) The types of algorithms used for comparing strings in other fields of science and business are phonetic-based algorithms, users’ behaviour-based algorithms and string-based algorithms (i.e., letter-based). Phonetic-based algorithms, such as Soundex,11 Metaphone (Philips 1990: 39) or Caverphone (­Rodichevski 2016), encode homophones with the same representation. This aids the user in receiving an accurate output even if an inexact query has been inserted. Phonetic coding is chiefly consonant-based (vowels are accounted for only if they appear in the initial position in the token). They are created for the use of English speakers (especially for American English pronunciation) and help to accommodate spelling mistakes of words used in the English-speaking environment (typically names used in the United States; they were first created in order to compare names in the U.S. census in 1880). For this reason, they are inappropriate for a diachronic corpus of a language other than English. However, it cannot be excluded that their use may yield results comparable to a search based on orthography. The fact that needs to be accounted for is that the search is being done by a speaker of the English language. Misspellings will most probably be based on a mental picture of Hittite that originates in the person’s most commonly used language. As English has become a lingua franca of today’s science, phonetic-based algorithms conforming to the user’s language may thus be a reasonable option. Algorithms used by modern web search engines base their results on users’ behaviour, on their entry and on statistical methods. For example, Google prefers results which have been found satisfactory by previous users; they are assigned higher ranking. However, a corpus of the size of the Hittite texts is too small to successfully apply such statistical methods. String matching algorithms, such as the Levenshtein distance12 (Levenshtein 1965: 845), the Knuth-Morris-Prat algorithm (Knuth et al. 1977), the Ratcliff-­Obershelp pattern recognition13 (Ratcliff 1988: 46) or the Jaro-Winkler 11 12 13

Soundex was patented by Robert C. Russell in 1918 (US Patent No 1,261,167Russell) and served originally as a tool of disambiguation of names in genealogical search. It was not connected to electronic applications until several decades later. Levenshtein definitions: (1) The smallest number of insertions, deletions, and substitutions required to change one string or tree into another; (2) A Θ(m × n) algorithm to compute the distance between strings, where m and n are the lengths of the strings (Black 2004). Ratcliff-Obershelp Similarity definition: Compute the similarity of two strings as the doubled number of matching characters divided by the total number of characters in the two strings. Matching characters are those in the longest common subsequence plus, ­re­cursively, matching characters in the unmatched region on either side of the longest common subsequence (Black 2004). The best agreement has the value 1, no agreement has 0.

Tagging and Searching the Hittite Corpus

137

­algorithm (Jaro 1976), operate on the basis of comparing textual similarity metrics.14 While Levenshtein computes the edit distance between strings, Ratcliff-Obershelp also takes into account the total number of characters in the two strings, while the Jaro-Winkler measure operates with the weighted sum of percentage of matched and transposed characters from each file and with matching initial characters, later rescaled so that the intervals and weights depend on the string quality (Black 2014). These algorithms use a scale between 0 and 1, where 0 denotes totally nonmatching strings and 1 denotes equal strings. On a sample of a Hittite text, we tested several algorithms (Jaccard Distance, Sorensen Dice Distance, Tanimoto Coefficient, etc.). The table below gives the results for (unfitting) Ratcliff-Obershelp, which indexes the distance between ma-a-an and nam-ma as equal to the distance of ma-a-(a)n (adjusted enclitic form ma-a-na) and ma-a-an: More appropriate results for cuneiform transliteration are returned by the Levenshtein and the Jaro-Winkler algorithms. These two are worth pursuing further for accuracy and behaviour on strings adjusted according to certain criteria. Table 7.5  Ratcliff-Obershelp algorithm, search pattern: ma-a-an

Ratcliff Obershelp Similarity – including separators

Ratcliff Obershelp Similaritywithout separators

------------------nam-ma | ma-a-an :0,615384615384615 ma-a-(a)n | ma-a-an :0,615384615384615 ma-a-an | ma-a-an :0,571428571428571 a-an | ma-a-an :0,545454545454545 I-NA | ma-a-an :0,545454545454545 A-NA | ma-a-an :0,545454545454545 MU-an-ti | ma-a-an :0,533333333333333

------------------ma-a-(a)n | ma-a-an :0,666666666666667 ma-a-an | ma-a-an :0,6 nam-ma | ma-a-an :0,6 ma | ma-a-an :0,571428571428571 MA | ma-a-an :0,571428571428571 MU-an-ti | ma-a-an :0,545454545454545

14

These algorithms are representatives of many other, refined searches, such as Simple Optimal String Matching Algorithm (Allauzen and Raffinot 2000) or fast approximate string matching (Baeza-Yates and Navarro 1998). Today´s trends are towards focusing on the speed of the search, e.g. for exact string matching, graphics processing units applied to string matching (Frías and Lombardo 2017). Such features are of limited interest to corpora of ancient languages, as time optimization, for example, is not the major goal of search improvement for the Hittite corpus described here.

138

Frantíková

The next table gives the results of three different types of Levenshtein (github). Because Levenshtein counts the number of necessary edits, the best match gets the value 0, while the value 3 means that three edits are required in order to transform the first string into the other: Table 7.6  Levenshtein algorithm, search pattern: ma-a-an

(Including separators)

(Without separators)

(Dictionary entry)

------------------ma-a-an | ma-a-an :0 ta-a-an | ma-a-an :1 ma-a-(a)n | ma-a-an :1 pa-a-un | ma-a-an :2 ma-aš | ma-a-an :3 a-pa-aš | ma-a-an :3 ša-an | ma-a-an :3 a-an | ma-a-an :3 pa-ra-a | ma-a-an :3 an-da-an | ma-a-an :3 ša-ra-a | ma-a-an :3 da-a-ir | ma-a-an :3 pa-a-ir | ma-a-an :3 ka-a-ni | ma-a-an :3 pé-ra-an | ma-a-an :3

-----------------ma-a-an | ma-a-an :0 ta-a-an | ma-a-an :1 ma-a-(a) | ma-a-an :1 ma-aš | ma-a-an :2 ša-an | ma-a-an :2 a-an | ma-a-an :2 š-ša-an | ma-a-an :2 pa-a-un | ma-a-an :2 a-aš | ma-a-an :3 ma | ma-a-an :3 a-ap-pa-an | ma-a-an :3 a-pa-aš | ma-a-an :3 a-at | ma-a-an :3 š-ša-an-n | ma-a-an :3 pa-ra-a | ma-a-an :3

------------------ma-a-an | ma-a-an :0 ma-a-(a)n | ma-a-an :0 ta-a-an | ma-a-an :1 m | ma-a-an :2 n | ma-a-an :2 ma | ma-a-an :2 a-an | ma-a-an :2 mu | ma-a-an :2 m-mi | ma-a-an :2 1en | ma-a-an :2 ka-a-ni | ma-a-an :2 pa-a-un | ma-a-an :2 a-aš | ma-a-an :3 d | ma-a-an :3 U | ma-a-an :3

In the first column, the separators form an integral part of the token on the entry. Under such conditions, Levenshtein returns as zero edits only the exact match ma-a-an, while the counts ma-a-(a)n15 and ta-a-an are treated as equal. The tokens indexed “3” are generally irrelevant. In the second column, the separators are omitted. The deletion of separators does not eliminate the false result in the second line. In the third column, the transliterated words are ­automatically transcribed and as such enter the Levenshtein search. At the end, they are converted back to their original form. Such a convention i­ mproves the search results so that the two forms ma-a-an and ma-a-(a)n require zero edits while ta-a-an requires one edit. 15

The form ma-a-(a)n is the result of transliteration of the enclitic: ma-a-na-aš.

139

Tagging and Searching the Hittite Corpus

In the following table, a Jaro-Winkler algorithm is used for a text with separators. In the first column, the optimized algorithm is used. In the second, the word prefix16 is given the scaling factor (i.e., weight) 0.2, while the standard scaling factor is 0.1; and in the third, the prefix is given the highest possible weight of 0.25. The alignments that get a score above 1.0, as seen in the third column, are “illegal” and therefore the prefix weight must be adjusted in order not to obtain such results.17 With higher prefix weight, the algorithm (falsely) includes more forms as “perfect matches”, not only ku-iš and ku-i-(i)š, but also ku-in, ku-it. The following table shows similar results for a search run on a text without separators: Table 7.7  Jaro-Winkler algorithm (github), search pattern: ku-iš

(Including separators)

(Including separators, prefix 2.0)

(Including separators, prefix 2.5)

------------------ku-iš | ku-iš :1 ku-i-(i)š | ku-iš :0,966666638851166 ku-iš-ki | ku-iš :0,9375

-----------------ku-iš-ki | ku-iš :1 ku-iš | ku-iš :1

------------------ku-iš-ki | ku-iš :1,03125 ku-i-(i)š | ku-iš :1

ku-i-(i)š | ku-iš :0,98888885974884 ku-in | ku-iš :0,973333358764648 ku-it | ku-iš :0,973333358764648 Ku-uš-ša-ra | ku-iš :0,888484835624695 Ka-ni-iš | ku-iš :0,813333332538605

ku-in | ku-iš :1

ku-in | ku-iš :0,920000016689301 ku-it | ku-iš :0,920000016689301 Ku-uš-ša-ra | ku-iš :0,804848492145538 Ka-ni-iš | ku-iš :0,789999961853027

16

17

ku-iš | ku-iš :1 ku-it | ku-iš :1 Ku-uš-ša-ra | ku-iš :0,930303037166595 Ka-ni-iš | ku-iš :0,824999988079071

The notion prefix is used in the algorithm with a different meaning than is usual in linguistics and is not concerned with word morphology: for higher values of the so-called prefix, the algorithm gives more favourable ratings in case the strings match from the beginning rather than from the end. These false alignments are due to the weights of matching and transposing in the algorithm.

140

Frantíková

Table 7.7  Jaro-Winkler algorithm (github), search pattern: ku-iš (cont.)

(Including separators)

(Including separators, prefix 2.0)

(Including separators, prefix 2.5)

u-uš | ku-iš :0,783333361148834 IŠKUR-aš | ku-iš :0,741666674613953 pa-iš | ku-iš :0,733333349227905 kat-ti | ku-iš :0,730000019073486 ták-ki-iš-ta | ku-iš :0,711111128330231 Ku-uš-ša-ra-aš | ku-iš :0,69523811340332 KUR | ku-iš :0,688888907432556 ku-e-da-ni-ik-ki | ku-iš :0,683333337306976

u-uš | ku-iš :0,783333361148834 kat-ti | ku-iš :0,759999990463257 IŠKUR-aš | ku-iš :0,741666674613953 pa-iš | ku-iš :0,733333349227905 ták-ki-iš-ta | ku-iš :0,711111128330231 Ku-uš-ša-ra-aš | ku-iš :0,69523811340332 KUR | ku-iš :0,688888907432556 ku-e-da-ni-ik-ki | ku-iš :0,683333337306976

u-uš | ku-iš :0,783333361148834 kat-ti | ku-iš :0,774999976158142 IŠKUR-aš | ku-iš :0,741666674613953 pa-iš | ku-iš :0,733333349227905 ták-ki-iš-ta | ku-iš :0,711111128330231 Ku-uš-ša-ra-aš | ku-iš :0,69523811340332 KUR | ku-iš :0,688888907432556 ku-e-da-ni-ik-ki | ku-iš :0,683333337306976

Table 7.8  Jaro-Winkler algorithm (github), search pattern: ku-iš

(Without separators)

(Without separators, prefix 2.0)

(Without separators, prefix 2.5)

----------------ku-i-(i)š | ku-iš :1 ku-iš | ku-iš :1 ku-iš-ki | ku-iš :0,933333337306976 ku-in | ku-iš :0,883333325386047 ku-it | ku-iš :0,883333325386047 KUR | ku-iš :0,777777791023254

--------------------ku-i-(i)š | ku-iš :1 ku-iš | ku-iš :1 ku-iš-ki | ku-iš :0,977777779102325 ku-in | ku-iš :0,933333337306976 ku-it | ku-iš :0,933333337306976 KUR | ku-iš :0,833333313465118

--------------------ku-iš-ki | ku-iš :1 ku-i-(i)š | ku-iš :1 ku-iš | ku-iš :1 ku-in | ku-iš :0,958333313465118 ku-it | ku-iš :0,958333313465118 KUR | ku-iš :0,861111104488373

141

Tagging and Searching the Hittite Corpus

(Without separators)

(Without separators, prefix 2.0)

(Without separators, prefix 2.5)

Ka-ni-iš | ku-iš :0,774999976158142 ki-iš-ha | ku-iš :0,774999976158142 Ku-uš-ša-ra | ku-iš :0,766666650772095 né-ku-uš | ku-iš :0,75 u-uš | ku-iš :0,722222208976746 ták-ki-iš-ta | ku-iš :0,694444417953491 Ku-uš-ša-ra-aš | ku-iš :0,683333337306976 ku-e-da-ni-ik-ki | ku-iš :0,674242436885834 pa-iš | ku-iš :0,666666686534882

Ku-uš-ša-ra | ku-iš :0,824999988079071 Ka-ni-iš | ku-iš :0,800000011920929 ki-iš-ha | ku-iš :0,800000011920929 né-ku-uš | ku-iš :0,75 u-uš | ku-iš :0,722222208976746 ták-ki-iš-ta | ku-iš :0,694444417953491 Ku-uš-ša-ra-aš | ku-iš :0,683333337306976 ku-e-da-ni-ik-ki | ku-iš :0,674242436885834 pa-iš | ku-iš :0,666666686534882

Ku-uš-ša-ra | ku-iš :0,854166626930237 Ka-ni-iš | ku-iš :0,8125 ki-iš-ha | ku-iš :0,8125 né-ku-uš | ku-iš :0,75 u-uš | ku-iš :0,722222208976746 ták-ki-iš-ta | ku-iš :0,694444417953491 Ku-uš-ša-ra-aš | ku-iš :0,683333337306976 ku-e-da-ni-ik-ki | ku-iš :0,674242436885834 pa-iš | ku-iš :0,666666686534882

When executed on data without separators and prefix weight up to 2.0, the algorithm does not return false results (i.e., listing ku-iš-ki as a best alignment for ku-iš) and keeps a good alignment distance from tokens belonging to other lexemes. Corresponding results would be obtained if the data were automatically transcribed using the transcription rules conventional for Hittite. Levenshtein with separators returns worse results for ku-iš (e.g. ku-i-(i)š and ku-in get the same alignment of value 1) than without separators. We conclude that both algorithms, the Levenshtein and Jaro-Winkler, yield quality results in searches of a syllabically transliterated language, where the possibility of higher prefix weight must be tested for a specific language and may prove useful in languages where morphological marking is mainly suffixed. 5.3 Logic of Smart Search The query can be executed based on different kind of entries. See the following examples of possible searched strings:

142

Frantíková

Table 7.9  Possibilities of query execution on the entry

Search string

Result

KBo iii 22 “a-as-su-us” āssu / assu asu / su a-as-su-us e-es-ta

Transcription of the text on the KBo iii 22 tablet Exact matches + suggested lexeme Best approximated form → lexeme + alternatives ditto Clauses with at least one lexeme, clauses with both lexemes first “a-as-su-us” AND e-es-ta Clauses with both exact match of the token and the lexeme good Best approximated translation in dictionary → lexeme + alternatives

Figure 7.4 Search visualization

On the interface, the user should have the opportunity to type any desirable request into an easy-to-operate entry (see Figure 7.4), as will be shown in the following example of smart search. The following picture (Figure 7.5) illustrates the process of the user’s activity and the response of the software. Once the (inaccurate) form is entered into the search line, the program provides results for the best approximated lexeme. Further, it suggests other lexemes and their respective forms. Simultaneously, it provides a list of token attestations of the best approximated lexeme.

Tagging and Searching the Hittite Corpus

143

Figure 7.5 Search and response process

6 Conclusion Digitizing Hittite in order to enable advanced search functions is unquestionably one of the most important tasks of Hittite studies today. The enormous recent progress in corpus computation worldwide aids the field of study of ancient languages by providing both the electronic tools as well as theoretical background in morphological standards and tagging. The crucial differences in digitizing Hittite compared to digitizing modern languages lie in the way the texts have been preserved to the present day. When dealing with a language without speakers, we rely on the graphemic interpretation of the language, where the script conventions and the motivation of scribal practices may blur the original underlying structure. Therefore, it is very important to try to avoid interpretations on every level and strive to interpret the texts as objectively as possible. Digitizing is just the first step to corpus usage. The effort to create accurately digitized texts is hampered by different transliteration practices and ­incompleteness of the original material. As for tokenizing, the specific issues arise from tokenizing partly broken texts and from the occurrences of clitic chains in the beginning of clauses, as is prototypical for Hittite. The chains have to be parsed in a way so that the members can be automatically assigned to the corresponding lemmata and the information on the use of the original cuneiform signs is not lost. I suggest that in order to retain one layer of corpus text, the clitics are not parsed within a cuneiform sign but rather between two signs, with adjustments to the tokens later on. On the level of tagging, the

144

Frantíková

­digital analysis adds information beyond mere transliteration, such as morphological information, connection of the token to the dictionary entry (lemma or hyperlemma), syntactic and textual information etc. Compared to corpora for modern languages, one highly specific problem in creating a Hittite digital corpus is the issue of search. The original orthography operates on the basis of syllabic and vocalic signs, while the search algorithms operate on basis of Latin-script graphemes. It is important to adjust the search in a way that yields valid results. In this article, this issue is tackled only briefly (§5.2), but it will be the subject of further research. The provisional conclusion is that Latin-script-based algorithms are a more efficient search method than cuneiform-based ones. Best suited for the Hittite text type are the edit distance algorithm (Levenshtein) and Jaro-Winkler algorithm. The search executed upon text without separators and with increased prefix weight yields more accurate results than the search with separators and with standard value of the scaling factor. Acknowledgments This contribution is part of the project Digitizing Hittite Corpus, No. 308315, completed at the Faculty of Arts, Charles University, Prague in 2015–2016 and supported by the Grant Agency of Charles University (GA UK). In addition, I wish to thank Vladimír Petkevič (Institute of the Czech National Corpus) and Ronald I. Kim (Institute of Comparative Linguistics) for their useful remarks and corrections. Bibliography Agirre, E. and Edmonds, P. (eds.) (2007) Word Sense Disambiguation: Algorithms and Applications [Text, Speech and Language Technology, 33], Dordrecht: Springer ­Science & Business Media. Allauzen, C. and Raffinot, M. (2000) Simple Optimal String Matching Algorithm. Journal of Algortihms 36/1, 102–116. Baeze-Yates, R. and Navarro, G. (1998) Fast Approximate String Matching in a Dictionary. In Proc. SPIRE’98. IEEE Computer Press, 14–22. Fisseler, D., Weichert, F., Müller, G.G.W. and Cammarosano, M. (2014) Extending philological research with methods of 3D computer graphics applied to analysis of cultural heritage. In: R. Klein and P. Santos (eds.), EUROGRAPHICS Workshop on Graphics and Cultural Heritage, The Eurographics Association. DOI: 10.2312/ gch.20141314.

Tagging and Searching the Hittite Corpus

145

Frías, R.R. and Lombardo, M.V. (2017) A Review of String Matching Algorithms and Recent Implementations using GPU. International Journal of Security and Its Applications 11/6, 69–76. Garrett, A.J. (1990) The Syntax of Anatolian Pronominal Clitics. Ph.D. Diss., Harvard University. Hoffner, H.A. (1966) Composite Nouns, Verbs, and Adjectives in Hittite. Or NS 35, 377–402. Hoffner, H.A. and Melchert, C. (2008) A Grammar of the Hittite Language. Part 1: Reference Grammar. Winona Lake, IN: Eisenbrauns. Hlaváčová, J. (2009) Formalizace systému české morfologie s ohledem na automatické zpracování českých textů, Dissertation, Charles University in Prague. Inglese, G. (2016) Towards an Old Hittite Treebank, Dissertation, IUSS Pavia. Jaro, M.A. (1976) UNIMATCH: A Record Linkage System: User’s Manual, Washington, DC: U.S. Bureau of the Census. Kaalep, H.J. et al. (2012) A trivial method of choosing the right lemma. In: A. Tavast, K. Muischnek and M. Koit: Human Language Technologies – The Baltic Perspective: Proceedings of the Fifth International Conference Baltic HLT 2012 [Frontiers in Artificial Intelligence and Applications 247], Amsterdam: IOS Press, 82–89. Kloekhorst, A. (2014). Accent in Hittite: A study in plene spelling, consonant gradation, clitics, and metrics (StBoT 56). Wiesbaden: Harrassowitz. Knuth, D.E., Morris, J.H. and Pratt, V.R. (1977) Fast pattern matching in strings. SIAM journal on computing 6, 323–350. Krebernik, M. (1996) The Linguistic Classification of Eblaite: Methods, Problems and Results. In: Cooper, G.S. and Schwartz, G.M., The Study of the Ancient Near East in the 21st century. The William Foxwell Albright Centennial Conference. Winona Lake, IN: Eisenbrauns, 233–249. Kučera, H. and Francis, W.N. (1967) Computational Analysis of Present-Day American English, Providence: Brown University Press. Laroche, E. (1958) Comparaison du louvite et du lycien. BSL 53, 159–197. Laroche, E. (1971) Catalogue des textes hittites, Paris: Klincksieck. Laroche, E. (1972) Catalogue des textes hittites, premier supplément. RHA xxx, 94–133. Lawrence, S. and Giles, C.L. (1999) Accessibility of information on the Web, Nature 400 (6740), 107–109. Levenshtein, V.I. (1965) Binary codes capable of correcting deletions, insertions, and reversals, Doklady Akademii Nauk SSSR 163/4, 845–848. Osolsobě, K., Hlaváčová, J., Petkevič, V., Svášek, M. and Šimandl, J. (2017) Nová automatická morfologická analýza češtiny, Naše řeč 100/4, 225–234. Philips, L. (1990). Hanging on the Metaphone, Computer Language 7/12, 39–44. Ratcliff, J.W. and Metzener, D. (1988) Pattern matching: The gestalt approach, Dr. Dobb’s Journal 46, 49–51.

146

Frantíková

Zemánek, P. (2007) A treebank of Ugaritic. Annotating fragmentary attested languages. In: K. De Smedt, J. Hajič and S. Kübler (eds.), Proceedings of the Sixth International Workshop on Treebanks and Linguistic Theories, Bergen: NEATL, 212–218.

Online Black, P.E. (2014) Jaro-Winkler, in: V. Pieterse and P.E. Black (eds.), Dictionary of Algorithms and Data Structures, https://www.nist.gov/dads/HTML/jaroWinkler.html (accessed 8.10.2016). Brin, S. and Page, L. (1998) The anatomy of a large-scale hypertextual Web search engine. In: Seventh International World-Wide Web Conference (WWW 1998), April 14–18, 1998, Brisbane, Australia, http://ilpubs.stanford.edu:8090/361/1/1998-8.pdf ­(accessed 1.9.2016). DIAKORP – Český národní korpus, Praha: Ústav Českého národního korpusu FF UK, http://www.korpus.cz (accessed 1.9.2016). Giusfredi, F. (2015) Phrase structure and ancient Anatolian languages. Methodology and challenges for a Luwian syntactic annotation. In: Proceedings of CLiC-it, http:// clic.humnet.unipi.it/proceedings/Proceedings-CLICit-2014.pdf (accessed 1.9.2016). Kučera, K., Řehořková, A. and Stluka, M. (2015) DIAKORP: Diachronní korpus, verze 6 z 18.12. 2015. Praha: Ústav Českého národního korpusu FF UK, http://www.korpus.cz (accessed 2.1.2016). Melchert, H.C. (2001) Cuneiform Luvian Corpus online, http://www.linguistics.ucla .edu/people/Melchert/CLUVIAN.pdf (accessed 11.10.2016). Molina, M. and Molin, A. (2016) In a lacuna: building a syntactically annotated corpus for a dead cuneiform language (on the basis of Hittite), in: Proceedings of the International Conference “Dialogue 2016”, http://www.dialog-21.ru/en/digest/2016/on line/ (accessed 1.9.2016). Müller, G.G.W. and Wilhelm, G. (2002–2013) Textzeugnisse der Hethiter, Hethitologie Portal Mainz, http://www.hethport.uni-wuerzburg.de/HPM/txthetlink.php (accessed 1.1.2016). Rodichevski, A. (2016), Approximate string-matching algorithms, part 1, http://www .morfoedro.it/doc.php?n=222&lang=en (accessed 1.9.2016).

Chapter 8

The Phonetics and Phonology of the Hittite Dental Stops Alwin Kloekhorst Abstract This article provides a comprehensive analysis of the phonetics and phonology of the Hittite dental stops, based especially on a detailed treatment of the usage of the cuneiform signs TA and DA in all positions in the word, and in all chronological stages of Hittite.

Keywords Hittite – phonology – phonetics – cuneiform script – Indo-European linguistics

1 Introduction The cuneiform syllabary that was taken over by the Hittites from their North Syrian neighbours possesses in its CV series separate signs to distinguish voiceless from voiced stops, e.g. TA vs. DA, KA vs. GA, KI vs. GI, etc. Since in Hittite the members of such sign pairs are sometimes used interchangeably (e.g., the word for ‘they eat’ is spelled a-ta-an-zi as well as a-da-an-zi, the word for ‘he opens’ is spelled ki-nu-uz-zi as well as gi-nu-uz-zi, etc.), it is generally stated in the Hittitological literature that in spelling the choice between the signs for the voiceless stop and the signs for the voiced stop is random, and that the use of a specific sign in a given word has no bearing whatsoever on the phonology of the stop it denotes (e.g. Melchert 1994: 13–14; Kimball 1999: 89–90; Hoffner and Melchert 2008: 16; Patri 2009: 89), a view that I, too, adhered to in my etymological dictionary of Hittite (Kloekhorst 2008: 21). In a series of recent articles (Kloekhorst 2010a, 2013, 2016) I have retracted this view, however, arguing that in some periods of Hittite the signs for the voiceless stops (TA, KA, KI, etc.) in some positions in the word do represent

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_010

148

Kloekhorst

phonologically different sounds from those represented by the signs for the voiced stops (DA, GA, GI, etc.). In the present paper I will provide a follow-up to these articles, presenting all additional evidence regarding the pair TA vs. DA that I have gathered over the last years, which results in a detailed analysis of the phonetics and phonology of the Hittite dental stops in all positions in the word throughout the entire Hittite period. 2

Dental Stops in Intervocalic Position: The OH Situation

In Kloekhorst 2013, an article that dealt with the phonetic difference between the signs TA and DA in Old Hittite, I argued that in Old Hittite we have to distinguish three dental stops in intervocalic position, namely: 1. A geminate spelled stop that is always written with the sign TA, (-)Vt-ta(-), and that etymologically corresponds to PIE *t. It was argued that this consonant phonetically represents a voiceless long stop [tː], which in this article will be called fortis. 2. A geminate spelled stop that is written both with the sign TA and with the sign DA, (-)Vt-ta(-) and (-)Vd-da(-), and that etymologically corresponds to the PIE cluster *TH. It was argued that this consonant phonetically represents a voiceless long postglottalized stop [tːʔ], which in this article will be called ejective. 3. A single spelled stop that is written both with the sign TA and the sign DA, °V-ta(-) and °V-da(-), and that etymologically corresponds to PIE *d and *dh. It was argued that this consonant phonetically represents a voiced short stop [d], which in this article will be called lenis. Although the lenis stop is voiced in this position, it was argued that its voice is only allophonic and that the basic distinction between the fortis and the lenis stop is length instead of voice.1 We can therefore set up the following three dental phonemes for Old Hittite:

1 This was argued on the basis of the presence in Hittite of clusters consisting of a lenis and a fortis stop, like the one in e-ku-ut-ta ‘he drank’ /ʔékwtːa/. If the basic distinction between the lenis and the fortis stops were voice, we would expect to find voice assimilation in such clusters, yielding either two fortis stops, **e-ek-ku-ut-ta, or two lenis stops, **e-ku-ta (cf. Kloekhorst 2008: 2; 2016: 1–2). Since this did not happen, the distinction between the two kinds of stops apparently was not voice. This argument is corroborated by the fact that, as we will see below, after obstruents the lenis dental stop is realized as a short voiceless stop [t], whereas the fortis dental stop is in that position realized as a long voiceless stop [tː]. The only

The Phonetics and Phonology of the Hittite Dental Stops

149

fortis /tː/ ejective /tːʔ/ lenis /t/ In the article mentioned, only the Old Hittite situation was investigated, but not the situation in Middle and New Hittite. I will therefore do so here. 3

Dental Stops in Intervocalic Position: the MH and NH Situation

There can be no doubt that of the three dental phonemes that have to be distinguished for Old Hittite, at least the fortis and lenis stops must have remained distinct phonemes also after the OH period: in MH and NH texts, too, they are consistently distinguished in spelling, namely by geminate vs. single spelling, respectively (= Sturtevant’s Law). The status of the ejective stop in Middle and New Hittite is less clear, however. The postulation of an intervocalic ejective stop /tːʔ/ in Old Hittite was based on the existence of four words that in OS texts show geminate spelling with the sign DA, (-)Vd-da(-), which correlates with the etymological presence of a cluster of a dental stop + laryngeal in their reconstructed preform: paddaḫḫi ‘I dig’ < *bhodhh2-, paddar / paddan- ‘basket’ < *péth2-r / *p(e)th2-én-, piddāi ‘he flees’ < *pth1/2-ói-ei, and uddār ‘words’ < *uth2-ṓr. They thus contrast with the Old Hittite words that show consistent geminate spelling with the sign TA, which always correlates with the etymological presence of a *t in their preform. Since in Akkadian a spelling (-)Vd-da(-) can also be read as (-)Vṭ-ṭa(-), i.e. as containing a geminate emphatic stop, which phonetically must have been a long postglottalized stop [tːʔ],2 it was argued that in Hittite, too, the spelling (-)Vd-da(-) in these four words represents the presence of a long ejective stop /tːʔ/, which can then be regarded as the regular outcome of an intervocalic cluster *-TH-.3 The first step required to determine to what extent the ejective stop is still a separate phoneme in MH and NH times is to investigate the spelling of these four words in MS and NS texts: how often do they show the sign TA or the sign DA?

­distinction between the two is length, which therefore can be regarded to have been the basic distinction between the two stops. 2 Kouwenberg 2003: 81–82. 3 Kloekhorst 2013: 127–131.

150

Kloekhorst

padd(a)-i ‘to dig’:4 In OS texts, we only find the 1sg.pres.act. form pád-daaḫ-ḫi, but in MS and NS texts also other relevant forms are attested. In MS texts, we find no forms spelled with TA, but one form spelled with DA (3pl.pres.act. pád-da-a-an-zi (1x)). The ratio of forms spelled with the sign TA to forms spelled with the sign DA is thus 0 : 1 = 0%. In NS texts, we find no forms spelled with TA, but 31 forms spelled with DA (1sg.pres.act. pádda-aḫ-ḫi (5x), 3sg.pres.act. pád-da-a-i (8x), pád-da-i (5x), 3pl.pres.act. pád-da-an-zi (3x), pád-da-a-an-zi (5x), 1sg.pret.act. pád-da-aḫ-ḫu-un (1x), 3sg.pret.act. pád-d[a-…] (1x), 3sg.pres.mid. pád-da-a-ri (1x), part. pád-daan-t° (2x)). The ratio of forms spelled with the sign TA to forms spelled with the sign DA is thus 0 : 31 = 0%. If we combine the MH and NH numbers, we arrive at 0 : 32 = 0%. paddar / paddan- ‘basket’: In OS texts, we find five attestations spelled with DA (nom.-acc.sg. pád-da-r° (1x), dat.-loc.sg. pád-da-ni (1x), pád-daa-ni (1x), [p]ád-da-ni-i (1x), pád-da[-ni] (1x)), and four attestations with TA (nom.-acc.sg. pát-ta-ar (2x), instr. pát-ta-ni-it (2x)).5 In MS texts, we find no forms spelled with TA, but five forms spelled with DA (dat.-loc.sg. pád-da-ni (5x)),6 yielding a ratio of 0 : 5 = 0%. In NS texts, we find three forms spelled with TA (dat.-loc.sg. pát-ta-a-ni (1x), instr. pát-ta-ni-it (2x)) and 27 spelled with DA (dat.-loc.sg. pád-da-ni-i (10x), pád-da-a-ni (3x), pád-da-ni (11x), instr. pád-da-ni-it (1x), abl.(?) pád-da-n[a-az] (1x), uncl. pád-da-na-aš (1x)),7 yielding a ratio of 3 : 27 = 10%. If we combine the MH and NH numbers, we arrive at 3 : 32 = 8,6%. piddai-i ‘to flee’:8 In OS texts, we find two attestations of this verb spelled with DA (3sg.pres.act. píd-da-a-i (2x)), and none spelled with TA. In MS texts, we find no forms spelled with TA, and seven forms spelled with DA (3sg.pres.act. píd-da-a-i (2x), píd-da-i (1x), 3sg.pret.act. [p]íd-da-iš (1x), píd-da-a-it (1x), 2sg.imp.act. píd-da-a-i (2x)), yielding a ratio of 0 : 7 = 0%. In NS texts, we find six forms spelled with TA (3sg.pres.act. pí-it-t[a-i] (1x), pít-ta-a-iz-zi (2x), 3pl.pret.act. pít-ta-a-er (2x), imperf. pít-ta-iš-k° (1x)) and thirty forms with DA (3sg.pres.act. píd-da-a-i (1x), píd-da-a-iz-zi (3x), 1pl.pres.act. píd-da-a-u-e-ni (1x), 3pl.pres.act. píd-da-a-an-zi (2x), 4 All numbers are based on the attestations of this verb as gathered in CHD P: 235–236. 5 Cf. Kloekhorst 2013: 127. 6 See Kloekhorst 2014: 358 for attestations. 7 See CHD P: 241 for attestations (note that the ‘passim’ mentioned for KUB 27.67 refers to ii 19, iii 13, 18, 24; and that GIpád-da-a-ni as cited for KUB 9.6 i 3 is in fact GIpát-ta-a-ni). 8 Numbers based on the attestations as gathered in CHD P: 352–353.

The Phonetics and Phonology of the Hittite Dental Stops

151

píd-da-an-zi (1x), 2sg.pret.act. píd-da-it-ti (1x), 3sg.pret.act. píd-da-a-iš (6x), píd-da-a-it (2x), pí[d-d]a-it (1x), 3pl.pret.act. píd-da-a-er (1x), 2sg.imp.act. píd-da-i (1x), píd-da-a-i (2x), 2pl.imp.act. píd-da-at-ten (1x), píd-da-a-at-tén (1x), imperf. píd-da-a-eš-k° (1x), píd-da-a-iš-k° (1x), píd-da-iš-k° (2x), píd-da-eš-k° (2x)), yielding a ratio of 6 : 30 = 16,7%. If we ­combine the MH and NH numbers, we arrive at 6 : 37 = 13,9%. uddar / uddan- ‘word’: The only relevant form of this word attested in OS texts is the nom.-acc.pl. form, which was attested once as ut-ta-a-ar, but once as ud-d[a?-]a?-ar as well.9 In MS and NS texts, we also find other forms of this word that are spelled either with TA or with DA. In MS texts, this word is attested four times with the sign TA (gen.sg. ut-ta-na-a-aš (1x), nom.-acc.pl. ut-ta-a-ar (2x), erg.pl. ut-ta-na-a-an-te-eš (1x)), and 58 times with the sign DA (gen.sg. ud-da-na-a-aš (2x), ud-da-na-aš (1x), dat.loc.sg. ud-da-ni-i (18x), ud-da-ni (1x), abl. ud-da-na-a-az (1x), ud-da-na-az (4x), ud-da-na-za (1x), nom.-acc.pl. ud-da-a-ar (29x), erg.sg. ud-da-na-anza (1x)),10 yielding a ratio of 4 : 58 = 6,5%. In NS texts, we find seven times a spelling with TA (gen.sg. ut-ta-na-aš (2x), nom.-acc.pl. ut-ta-a-ar (5x)) and 135 times with DA (gen.sg. ud-da-na-aš (2x), dat.-loc.sg. ud-da-ni-i (20x), ud-da-ni (5x), ud-da-a-ni-i (3x), erg.sg. ud-da-na-an-za (4x), abl. udda-na-az (4x), nom.-acc.pl. ud-da-a-ar (89x), ud-da-ar (4x), erg.pl. ­ud-da-na-a-an-te-eš (1x), [u]d-da-na-an-te-eš (1x)),11 yielding a ratio of 7 : 135 = 4,9%. If we combine the MH and NH numbers, we arrive at 11 : 193 = 5,4%. We see that in MS and NS texts in all four words the number of forms spelled with the sign DA is much larger than the number of forms spelled with TA. The next step is to answer this same question for the words that in OS texts are consistently spelled with the sign TA, and that therefore must contain a (non-ejective) fortis stop /tː/: how are these spelled in MS and NS texts? Since it would be too time-consuming to investigate all relevant words, I have selected a few representative examples that are attested often enough to give statistically relevant numbers.

9 Cf. Kloekhorst 2013: 129. 10 For attestations, cf. Kloekhorst 2014: 319 n. 1234 (gen.sg.), 454 (dat.-loc.sg.), 299 (erg.sg.), 320 (abl.), 240 n. 869 (nom.-acc.pl.), 299 (erg.pl.). 11 For attestations, cf. Kloekhorst 2014: 319 n. 1235 (gen.sg.), 454 (dat.-loc.sg.), 299 (erg.sg.), 320 (abl.), 241 n. 870 (nom.-acc.pl.), 299 (erg.pl.).

152

Kloekhorst

katta ‘down’ < *ḱmto: In OS texts, this word is consistently spelled kat-ta (33x), with the sign TA, and never with the sign DA. Also in MS and NS texts it is always spelled kat-ta (ca. 700 times in my files) and never **­kad-da. The ratio of spellings with the sign TA to spellings with the sign DA is therefore 100%. kitta(ri) ‘he lies’ < *ḱéito(ri): This word is in OS texts consistently spelled ki-it-ta (23x),12 with the sign TA, and never with the sign DA. Also in MS and NS texts it is in the overwhelming majority of cases spelled ki-it-ta(ri) (ca. 200 times in my files), with the sign TA. Only three times do we find ki-id-da(-ri).13 The ratio of spellings with the sign TA to spellings with the sign DA is thus approximately 200 : 3 = 98,5%. lukkatta ‘it dawns’ < *lukoto:14 This word is in OS texts in all its four attestations spelled with the sign TA, and not with the sign DA.15 Also in MS and NS texts it is always (more than 20 times) spelled with the sign TA (lu-uk-kat-ta, lu-ug-ga-at-ta, lu-kat-ta), and never with the sign DA. The ratio of forms spelled with TA to the forms spelled with DA is thus 100%. nutta ‘and to you’ < *nu=tuo: This word is in OS texts attested once as nuut-ta,16 spelled with the sign TA. In my files of MS and NS texts, it occurs ca. 230 times as nu-ut-ta, with the sign TA, and once as nu-ud-da (KUB 33.70 iii 16 (OH/NS)), with the sign DA. The ratio of forms spelled with TA to forms spelled with DA is thus 230 : 1 = 99,6%. In all these words, the ratios of the number of forms spelled with the sign TA to the number of forms spelled with the sign DA (100%, 98,5%, 100%, and 99.6%, respectively) are totally opposite to the ratios of TA to DA in the words padd(a)-i, paddar / paddan-, piddai-i and uddar / uddan-, which were 0%, 6,7%, 13,9%, and 5,4%, respectively. This massive difference in spelling between these two groups of words proves that also in MH and NH times the ejective dental stop /tːʔ/ was still phonemically distinct from the fortis dental stop /tː/. 12 Cf. Kloekhorst 2014: 419 n. 1623 for attestations. 13 ki-id-da (KBo 3.21 ii 9 (MH/MS)), ki-id-da-ri (KUB 30.15 obv. 6, 13 (OH/LNS)). 14 Cf. CHD L-N: 75 for attestations. 15 lu-ug-ga-at-ta (StBoT 25.4 iv 21 (OS)), lu-uk-kat-ta (StBoT 25.3 ii 30, iv 7 (OS)), [(lu-uk-­katt)]a (StBoT 12+ i 31 (OS)). 16 nu-ut-ta (KUB 43.27 i 8 (OS)).

The Phonetics and Phonology of the Hittite Dental Stops

153

This finding has some interesting consequences for several other words and morphemes. apadda(n) ‘there, thither’: This word, which is unattested in OS texts, is in MS and NS texts spelled as follows:17 eleven times we find a spelling with the sign TA (a-pa-at-ta (3x), a-pa-a-at-ta (6x), a-pát-ta-an (2x)), and 92 times a spelling with the sign DA (a-pád-da (53x), a-pád-da-an (38x), apa-da-an (1x)).18 The ratio of spellings with TA to spellings with DA is thus 11 : 92 = 10,7%, which matches the ratios of padd(a)-i, etc. I therefore conclude that this word must have contained an ejective stop as well: /ʔapatːʔa(n)/. Melchert (2008: 369–370) reconstructed this form as end�ing in *-éd-h2o, in which *-h2o would be the preform of the allative ending that is attached to the stem *h1obhéd- as visible in dat.-loc.sg. apedani, etc. According to Melchert, the short a of the medial syllable of apadda reflects an earlier *e that has been colored to a because it stood before a cluster *-dh2-. This idea is now confirmed by the spellings with DA, which point to the presence of the ejective stop /tːʔ/ and forms an independent argument in favor of a reconstruction with a cluster *-TH-. It should be noted, however, that since I rather reconstruct the all.sg. ending as *-o, and not as *-h2o,19 I cannot accept all details of Melchert’s etymology. To my mind, we should rather interpret apadda as reflecting *h1obhé­ dhh2e, a form that consists of the pronominal oblique stem *h1obhé- (cf. gen.sg. apel ‘of his’, etc.) to which the locatival element *-dhh2e is attached that is known from Gr. ἔνθα ‘there’ and Skt. ihá ‘here’, and which may also be present in Hitt. anda ‘into’ < *h1n-dhh2e.20 natta ‘not’: This word is in OS, MS and NS texts consistently spelled naat-ta (more than 150 times in my files), with the sign TA, and never **naad-da.21 We should therefore analyze it as /natːa/, with a fortis, and not an ejective stop. The etymology of this word is not fully clear. It is obviously related to *ne ‘not’ as attested in many IE languages, but details regarding its latter part are unclear. Melchert (2008: 372) proposes to reconstruct *né-th2oh1, a form that structurally would be the same as Skt. táthā ‘thus’ 17 Counts based on the attestations listed in HW2 A: 168–170. 18 Twice we find the spelling a-pát-tén, but its interpretation is unclear. 19 Cf. Kloekhorst 2008: 161. 20 Cf. Kloekhorst 2013: 139 for this etymology of anda. 21 Cf. CHD L-N: 409.

154

Kloekhorst

and káthā ‘how’, and in which the e would be colored to a because of the following *-TH- cluster (just as in apadda < *h1obhedhh2e). Although especially this latter argument is attractive (it is otherwise difficult to account for the Hitt. a in natta vis-à-vis the *e found in the *ne as reflected in all other IE languages, cf. Melchert 2008: 371), I would rather expect that a preform *néth2oh1 should have yielded Hitt. **/nátːʔa/, spelled **na-adda. Melchert’s etymology can therefore hardly be ­correct. I am unable, however, to offer an alternative one. -tta (2sg.pret.act. ending of the ḫi-conj.): This ending is generally connected with the 2sg. perfect ending as found in Sanskrit (-tha) and Greek (-θα) and that is reconstructed as *-th2e. On the basis of the foregoing, we would expect this ending in postvocalic position to have yielded Hitt. /-tːʔa/, which then should be spelled …-Vd-da. Yet this is not the case. Although it is unattested in OS texts, in MS and NS texts this ending is in postvocalic position always spelled …-Vt-ta (ḫal-za-it-ta (OH/NS) ‘you screamed’, na-it-ta (OH/MS) ‘you turned’, pa-it-ta (OH/MS) ‘you gave’, da-a-at-ta (MH/MS) ‘you took’, etc.), which rather points to a phonological shape /-tːa/, with a fortis, non-ejective /tː/. As we will see below, there are indications that although PIE *-TH- did in Hittite develop into an ejective stop /tːʔ/ in postvocalic position (as well as after n and in word-initial position), it yielded a non-ejective fortis stop /tː/ when preceded by an obstruent, *r, or *l. This means that the 2sg. pret.act. ending of the ḫi-conjugation would originally have had two allomorphs, namely postvocalic and postnasal /-tːʔa/ vs. /-tːa/ in other positions. It seems quite possible to me that in such a situation one of the variants ousted the other, and the spelling …-Vt-ta clearly indicates that in this case it is the ending /-tːa/ that has become the productive one. Support in favor of this theory may come from the spelling of the corresponding 2sg.pres.act. ending. Whereas the normal spelling of this ending in postvocalic position is …-Vt-ti, we do find in OS texts two forms with the spelling …-Vd-di: pé-e-da-ad-d[i] (KUB 33.59 ii 2 (OS)) and ú-daad-di (KUB 33.59 iii 3 (OS)). Although not all details regarding the phonetic difference between the signs TI and DI have been clarified, it seems attractive to assume that these two forms spell the original postvocalic ending /-tːʔi/ (< *-th2e+i), which later on was replaced by the postconsonantal variant /-tːi/, spelled …-Vt-ti. We may conclude that in intervocalic position the phonemic three-way distinction between fortis /tː/, ejective /tːʔ/ and lenis /t/ (phonetically realized as

The Phonetics and Phonology of the Hittite Dental Stops

155

[tː], [tːʔ], and [d], respectively) was retained as such throughout the history of Hittite.22 4

Dental Stops after n: the OH Situation

In Kloekhorst 2013, I also treated the spelling in Old Hittite texts of dental stops after n. It turned out that for this position we have to distinguish three different stops as well, namely: 1. A stop that is consistently spelled with the sign TA, °n-ta(-), and that therefore was interpreted as a voiceless stop [t]. Etymologically, the cluster [nt] corresponds to PIE *nd. 2. A stop that is consistently spelled with the sign DA, °n-da(-), and that therefore was interpreted as a postglottalized stop [tʔ]. Etymologically, the cluster [ntʔ] corresponds to PIE *nTH. 3. A stop that is spelled both with the sign TA and with the sign DA, °n-ta(-) and °n-da(-), and that was interpreted as a voiced stop [d]. Etymologically, the cluster [nd] corresponds to PIE *nt and *ndh.

22

There are two words that do not fit the pattern seen thus far because they show a considerable number of spellings both with (-)Vt-ta(-) and with (-)Vd-da(-). The verb ḫatt-a(ri), ḫazzii̯e/a-zi ‘to make a hole, to pierce, to prick’ shows, beside spellings with the sign TA (ḫa-at-ta(-…), 54 attestations in my files), also quite a few spellings with the sign DA (3sg. pres.mid. ḫa-ad-da(-ri), 3pl.pres.act. ḫa-ad-da-an-zi, ptc. ḫa-ad-da-an-t°, etc., 18 attestations in my files). Its ratio of TA vs. DA spellings, namely 54 : 18 = 75%, is too low to classify this verb as belonging to the words containing the phoneme /tː/, but also much too high to belong to the group containing /tːʔ/. Interestingly, the DA-spellings all come from NS texts, at which period the active stem of this verb, which originally was ḫazzii̯e/a-zi, has been reshaped after the tarn(a)-class to inflect ḫatta-i / ḫatt- (e.g. 3sg.pres.mid. ḫa-atta(-a)-i, ḫa-ad-da(-a)-i). It therefore seems quite possible to me that the original /tː/ of the verbal root /hatː-/ has been replaced by /tːʔ/ by analogy with the verb padda-i / padd- ‘to dig’ /patːʔ(a)-/, which shows the same tarna-class inflection and is furthermore semantically close (cf. also the derivatives ḫatteššar ‘hole, pit’ and patteššar ‘pit, hole in the ground’). The second example, the noun atta- (c.) ‘father’, is spelled both at-ta(-) (ca. 70% of its attestations) and ad-da(-) (ca. 30% of its attestations) and thus does not fit the numbers belonging to the phonemes /tː/ and /tːʔ/ either. Since this noun is clearly a word originating from childrens’ language, we may be allowed to assume that it contains a unique sound, namely a long voiced stop /dː/. As Ron Kim reminds me, this may be compared to the fact that Goth. atta ‘father’ shows a geminate -tt- that otherwise is very rare in Gothic: it is only found at morpheme boundaries, e.g. at-tiuhan ‘pull towards, bring’, and in the noun skatts ‘money’, which is of unknown origin. Likewise HLuw. tati- ‘father’, which is the only word in this language consistently spelled with word-initial , which probably indicates the presence of a unique sound (Xander Vertegaal, p.c.).

156

Kloekhorst

I did not discuss the question, however, how these postnasal stops correlate with the intervocalic stops. In other words: what is the phonemic status of these stops that are found after n? It seems obvious to me that the three postnasal stops can be equated with the three intervocalic stops in the following way: 1. The postnasal voiceless stop [t] is to be equated with the intervocalic fortis stop /tː/. 2. The postnasal postglottalized stop [tʔ] is to be equated with the intervocalic ejective stop /tːʔ/. 3. The postnasal voiced stop [d] is to be equated with the intervocalic lenis stop /t/. In other words, the phonemic length contrast between, on the one hand, the fortis and ejective stops and, on the other, the lenis stop is in postnasal position phonetically realized as a voice contrast: 1. /ntː/ is realized as [nt]. 2. /ntːʔ/ is realized as [ntʔ]. 3. /nt/ is realized as [nd]. An interesting outcome of this equation is the fact that although in intervocalic position PIE *d and *dh merge (into the lenis stop /t/) and remain distinct from PIE *t (which yielded the fortis stop /tː/), after *n a merger took place between PIE *t and *dh (which yielded lenis /nt/ = [nd]), whereas PIE *d remained distinct (as fortis /ntː/ = [nt]). This is in my opinion directly linked to the fact that the pre-Proto-Anatolian correspondents of PIE *t, *d and *dh were */tː/, */ʔt/ and */t/, respectively.23 In intervocalic position the preglottalic feature of */ʔt/ was reinterpreted as a separate phoneme, */ʔ/ (which was subsequently lost with compensatory lengthening of the preceding vowel), causing a merger of */ʔt/ and */t/ into Hittite /t/, which contrasted with /tː/. However, after *n the length of */tː/ was lost, causing it to merge with */t/ into */nt/, which through voice assimilation yielded [nd]. Moreover, */ʔt/ remained distinct because when also in this position its preglottalic feature was reinterpreted as a separate phoneme, */nʔt/ > */nʔt/, this glottal stop blocked any voice assimilation, causing */nʔt/ to develop into [nt]. In the case of *nTH, we have to assume that, just as in intervocalic position, the laryngeal caused a preceding stop to lengthen,24 which therefore was not subject to voice assimilation either. We can set up the following chronology in order to explain all the facts: 23 Cf. Kloekhorst 2012: 258–259; 2014: 230–235, 405–414, 574–583; and 2016 for the reconstruction of preglottalized voiceless short stops (/ʔp/, /ʔt/, etc.) as the pre-Proto-Anatolian correspondents of the PIE mediae (*b, *d, etc.). 24 Cf. Kloekhorst 2013: 130–131.

157

The Phonetics and Phonology of the Hittite Dental Stops Table 1

Development of PIE clusters of *n + dental (+ laryngeal)

(1) PIE *nt PIE *ndh PIE *nd PIE *ntH PIE *ndhH PIE *ndH

~ ~ ~ ~ ~ ~

pre-PAnat. */ntː/ pre-PAnat. */nt/ pre-PAnat. */nʔt/ pre-PAnat. */ntːʔ/ pre-PAnat. */ntʔ/ pre-PAnat. */nʔtʔ/

> > > > > >

(2) */nt/ */nt/ */nʔt/ */ntʔ/ */ntʔ/ */nʔtʔ/

> > > > > >

(3) */nt/ > */nt/ > */nʔt/ > */ntːʔ/ > */ntːʔ/ > */nʔtːʔ/ >

[nd] [nd] [nt] [ntʔ] [ntʔ] [ntʔ]

= = = = = =

/nt/ /nt/ /ntː/ /ntːʔ/ /ntːʔ/ /ntːʔ/

1.

Loss of consonantal length after *n, causing the merger of */tː/ and */t/ into */t/. 2. Lengthening of a short consonant by a following laryngeal, causing the shift of */tʔ/ to */tːʔ/, and reinterpretation of the preglottalic feature of */ʔt/ as */ʔ/. 3. Voice assimilation of */nt/ to [nd], but not of */ntː/. The presence of a */ʔ/ between */n/ and */t/ blocks the assimilation. Subsequent loss of interconsonantal */ʔ/, and a reinterpretation of the cluster */tːʔ/ as a postglottalized stop /tːʔ/. In Kloekhorst 2013, I only treated the Old Hittite situation regarding dental stops in postnasal position, not that of Middle and New Hittite, which I will do here. 5

Dental Stops after n: the MH and NH Situation

First, I will treat the fate of the ejective stop. We have seen above that in OS texts the ejective stop /tːʔ/, which after n is phonetically realized as [tʔ], was indicated in spelling by the consistent use of the sign DA, e.g. in an-da(-an) ‘into; inside’. If we investigate the spelling of this word in MS and NS texts, we find more than 2400 attestations spelled an-da(-an), with the sign DA, and only two attestations with the sign TA, namely an-ta? (KBo 20.10 i 4 (OH/OS or MS)) and an-tạ-an (KUB 20.76 iv 8 (OH/NS)) (note that in both cases the sign TA is either broken or questionable). It is therefore justified to say that also in the post-OH period anda(n) is consistently spelled with the sign DA. To my mind, this indicates that it has retained its ejective stop as a phonemic entity: /əntːʔa(n)/.

158

Kloekhorst

The presence of a phonemic fortis stop, /tː/, which after n is phonetically realized as [t], was based on the consistent spelling in OS texts of the form ­ši-pa-an-ta-an-zi ‘they libate’ with the sign TA, which was supported by the consistent spelling of its corresponding 3sg. form ši-pa(-a)-an-ti / iš-pa(-a)-anti ‘he libates’ with the sign TI, and its derivative iš-pa-an-tu-uz-zi- ‘libation ­vessel’ with the sign TU. These words were therefore interpreted as [sipːəntánt͡si] = /sipːəntːántsːi/, [sipːā́nti] = /sipːā́ntːi/, [ɨspā́nti] = /ɨspːā́ntːi/, and [ɨspəntut͡si-] = /ɨspːəntːutsːi-/, respectively. They thus contrast with words that in OS texts are spelled both with the sign TA and with the sign DA (or with TI as well as DI, or TU as well as DU), which rather points to the presence of a lenis stop, /t/, which after n was realized as a voiced stop, [d], e.g. e-ša-an-ta, e-ša-an-da ‘they sit down’ [ʔḗsənda] = /ʔḗsənta/, a-ša-an-tu, a-ša-an-du ‘they must be’ [əsándu] = /əsántu/. If we now look at the spelling of the form for ‘they libate’ in MS and NS texts, we find that there it is spelled both with TA and with DA, however: ši(-ip)-pa-an-ta-an-zi25 as well as ši(-ip)-pa-an-da-an-zi.26 Likewise the form for ‘libation vessel’, which in MS and NS texts is spelled iš-pa-an-tu-uz-zi-27 as well as iš-pa-an-du-uz-zi-.28 They thus are in the post-OH period spelled the same way as words that contain a lenis /t/. We must therefore assume that their OH /tː/, which was realized as a voiceless [t], has in the post-OH period changed to /t/, which was realized as [d], probably due to voice assimilation: OH [sipːəntánt͡si] = /sipːəntːántsːi/ > MH/NS [sipːəndánt͡si] = /sipːəntántsːi/ and OH [ɨspəntut͡si-] = /ɨspːəntːutsːi-/ > MH/NH [ɨspəndut͡si-] = /ɨspːəntutsːi-/.29 This does not mean, however, that in the post-OH period after n the contrast between fortis and lenis stops has been given up. Consider the word kuen­ ta ‘he killed’. Etymologically, this form is generally reconstructed as *gwhén-to, in which the ending is the 3sg.mid. ending *-to,30 which has replaced the original 3sg.pret.act. ending *-t because the latter was regularly lost in postconsonantal position. On the basis of what we have seen above, we would expect the sequence *-nt-, through a pre-PAnat. */-ntː-/, to have developed into OH [-nd-] = /-nt-/, with a lenis stop /t/. Yet, if we look at the spelling of the word 25 26 27 28 29 30

Attested 9 times in my files of MS and NS texts. Attested ca. 60 times in my files of MS and NS texts. Attested ca. 45 times in my files of MS and NS texts. Attested ca. 30 times in my files of MS and NS texts. Note that the 3sg.pres.act. form of ‘to libate’ is also in MS and NS texts consistently spelled with the sign TI: ši-pa-an-ti, ši-ip-pa-an-ti, BAL-an-ti, never *-an-di. The rationale behind this fact is not yet clear to me, and needs further investigation. Kloekhorst 2008: 800–801.

The Phonetics and Phonology of the Hittite Dental Stops

159

kuenta in MS and NS texts,31 we find that it is always (32 times) spelled with the sign TA (26x ku-en-ta, 5x ku-e-en-ta, 1x ku-in-ta), and never with the sign DA. This spelling thus rather points to the presence of a voiceless, i.e. fortis stop, [t] = /tː/: [kwénta] = /kwéntːa/. Does this mean that our view of the development of the cluster *-nt- is incorrect? To my mind, this is not the case. As we will see below, after all other consonants other than *n, PIE *t regularly develops into a fortis stop. This means that in all verbs in which the stem ends in a consonant other than *n, the verbal ending *-to regularly developed into /-t:a/ with a fortis /tː/, e.g. *h1égwh-to > Hitt. e-ku-ut-ta ‘he drank’ /ʔékwtːa/. Only after an *n would we expect the ending to have developed into /-ta/, with a lenis /t/ (which after n was realized as a voiced stop, [-da]). It seems unproblematic to me to assume that this allomorphy between /-tːa/ and /-ta/ was levelled out in favor of the fortis ­variant. Thus, when after the OH period the sequence /ntː/ = [nt] underwent voice assimilation to /nt/ = [nd], the fortis character of the dental stop of the ending -tta was again restored, resulting in ku(-e)-en-ta [kwénta] = /kwéntːa/, not **[kwénda] = **/kwénta/. 6

Dental Stops after Obstruents

Although it was stated in Kloekhorst 2013: 131 that in Old Hittite texts after the consonants ḫ, k, p, and š only the sign TA is found, and never DA, no conclusion was attached to this fact. I will therefore treat this fact in more detail here. The absence of DA after obstruents is not limited to OS texts; also in MS and NS texts we virtually only find the sign TA following ḫ, k, p, and š.32 To my mind, this virtual complete absence of spellings with the sign DA after obstruents indicates that the dental stops that occur in this position were phonetically neither ejective nor voiced. The absence of ejectives in this position is interesting, since there are certainly Hittite words that in their preform contain a cluster of obstruent + *TH, cf. e.g. ḫaštai- ‘bone’ < *h2/3ésth1oi-,33 or the 2sg.pret.act. ending -tta as treated 31 32

It is unattested in OS texts. I have counted in my files 244x °ḫ-ta(-) vs. 3x °ḫ-da(-); 190x °k-ta(-) vs. 2x °k-da(-); 144x °p-ta(-) vs. 7x °p-da(-) (4 of which occur in a single text, namely KBo 18.54); and 3013x °šta(-) vs. 17x °š-da(-). We see that the number of spellings with the sign DA is negligeable when compared to the number of spellings with the sign TA. Moreover, the spellings with DA do not seem to occur in any systematic pattern. 33 Although PIE *t in Hittite normally undergoes assibilation to z when followed by *i, this is not the case in the oblique cases of ‘bone’ (e.g. gen.sg. ḫaštii̯aš), which have retained

160

Kloekhorst

above. This means that in such sequences either the laryngeal was lost without causing glottalization (*-CTHV- > */-CtːʔV(-)/ > Hitt. /-CtːV(-)/), or the laryngeal at first did cause glottalization (*-CTHV- > */-CtːʔV(-)/ > pre-Hitt. */-CtːʔV(-)/), after which the glottalization was lost, yielding Hitt. /-CtːV(-)/. The latter scenario effectively entails that after obstruents, original ejectives have in pre-Hittite times merged with the fortis stops. The absence of voiced stops could at first sight be interpreted as a sign that in this position original fortis and lenis stops have merged into a single stop, which is realized as voiceless. Yet there are indications that we have to distinguish two types of stops in this position. The first type of stop is found in words that show spelling alternations like the one between e-uk-ta and e-ku-ut-ta ‘he drank’, and between li-in-ik-ta and li-in-kat-ta ‘he swore’. In these words, postconsonantal spelling with the sign TA, °C-ta(-), alternates with geminate spelling in graphic intervocalic position, (-)Vt-ta(-). This clearly shows that the dental stop in these words was a long voiceless stop, [ʔékwtːa] and [línktːa], and we may therefore interpret it as a fortis stop: /ʔékwtːa/ and /línktːa/. Since in these cases the dental stop etymologically goes back to PIE *t (*h1égwhto, *h1lénǵhto), it shows that in such clusters fortis stops were retained as such. The second type of stop is found in the word a-ku-ta-al-l° ‘container of water’,34 where we find a dental stop that is spelled single. Since it is likely that the preceding -ku- represents a labiovelar (because aku- can then be derived from the verbal stem *h1gwh- ‘to drink’), we can assume that, although the t is in graphic intervocalic position, it is in fact postconsonantal. The spelling with the sign TA indicates that it is voiceless, whereas the single spelling indicates that it is short. We can therefore assume that the word phonetically must have been [əkwtalː-], with a voiceless short [t], which undoubtedly must be interpreted as the lenis stop /t/: /əkwtalː-/. These words, in which the dental stops are in graphic intervocalic position but in fact stand in postconsonantal position, show that both fortis and lenis stops can occur in this position, and that the former is realized as a voiceless long stop [tː], but the latter as a voiceless short stop [t]. This means that spellings of the structure °C-ta(-) can in principle denote both /°Ctːa(-)/, with a

34

their *t. This can only be explained by the presence of the laryngeal between *t and *i, which then blocked the assibilation. Since the assibilation in *ti is a specifically Hittite development, the laryngeal must have been still present at that moment, and can have been lost only later on. In that sense, the laryngeal did leave an indirect trace in this word, albeit not glottalization. Attested twice: instr. a-ku-ta-al-li-it (KUB 9.20, 5), a-ku-ta!-al-li-it (KUB 2.13 i 8 (text: -ga-)).

The Phonetics and Phonology of the Hittite Dental Stops

161

­fortis stop, and /°Cta(-)/, with a lenis stop, and that one can only decide between the two on the basis of alternative spellings where the dental stop occurs in graphic intervocalic position, or on the basis of etymological considerations. 7

Dental Stops after r and l

Words in which dental stops follow the resonants r and l (note that in genuine Hittite words we never find a dental stop following m) were also left out of consideration in Kloekhorst 2013. I will therefore treat these here. First, I will look at the spelling of dental stops after r. In OS texts, the 3sg. pres.mid. form for ‘he stands’ is consistently spelled ar-ta(-ri) (7 times) with the sign TA, and never with the sign DA. This indicates the presence of a voiceless stop: [ərta(ri)]. Also in MS and NS texts, this word is consistently spelled with TA (ca. 160 times in my files), and never with DA, showing that also in Middle and New Hittite times it contained a voiceless stop, [ərta(ri)]. Etymologically, this word is generally assumed to reflect a preform *h3r-to(-), which would mean that here the PIE sequence *-rt- yielded Hitt. [-rt-]. This differs from the outcome of PIE *t after *n, where it underwent voice assimilation to OH [d]. However, since the *t in *h3r-to(-) is part of an ending, it cannot be excluded that an analogy to verbal stems ending in an obstruent has taken place (cf. the case of kuenta ‘he killed’ as treated above). In order to investigate the regular outcome of PIE *-rt- in Hittite, it is better to treat words in which analogical influence can be excluded. A possible candidate is the verb ḫuu̯ art-i / ḫurt- ‘to curse’ and its derivative ḫurtai- / ḫurti- ‘curse’, which on the basis of an etymological connection with OPrus. wertemmai ‘we swear’ may be reconstructed as *h2uort- / *h2urt- and *h2urt-oi- / *h2urt-i-, respectively.35 Unfortunately, both words are unattested in OS texts, but in MS and NS texts, they both occur spelled with the sign TA as well as with DA: e.g. 1sg.pret.act. ḫur-ta-aḫ-ḫu-un and ḫur-da-aḫ-ḫu-un; acc.pl. ḫur-ta-a-uš and ḫur-da-a-uš.36 These spellings point to the presence of a voiced stop [d], which would mean that in these words PIE *-rt- has undergone voice assimilation to [-rd-]. Since the dental stop is part of the root, it cannot have been influenced analogically, which would mean that we should regard this 35 36

Cf. Puhvel HED 3: 436, Kloekhorst 2008: 373, LIV 2: 292. Note that Sturtevant’s connection (1930: 128) with Lat. verbum ‘word’, Lith. var̃das ‘name’, OPrus. wirds ‘word’, Goth. waurds ‘word’ would point to a root *h2uerdh-, with a *dh. Cf. Puhvel HED 3: 433–434 and Kloekhorst 2008: 372–373 for attestations.

162

Kloekhorst

development as the phonologically regular one, whereas the presence of [t] in [ərta(ri)] must then be due to restoration of the ending. It should be noted, however, that because of the absence of OS attestations of ḫuu̯ art-i / ḫurt- and ḫurtai- / ḫurti-, we cannot be certain whether the voice assimilation of *-rt- to [-rd-] had already taken place before Old Hittite, or is instead a post-OH development. Moreover, since the etymology of ḫuu̯ art-i / ḫurt- and ḫurtai- / ḫurti- is not fully secure (a reconstruction with root-final *dh has been proposed as well, cf. footnote 35), these conclusions must remain tentative in any case. The outcome of the PIE sequence *-rd- is clearer, since it is present in the oblique stem of the word for ‘heart’, *ḱrd-, a word that is well attested in Hittite. In OS texts, the oblique cases of ‘heart’ are spelled kar-ta[-…] (KBo 25.107, 4 (OS)), kar-ta-az=(š)-mi-it (StBoT 25.7 iv 6 (OS)), kar-ti-i=š-mi (KBo 22.2 obv. 13 (OS)), kar-di-i=š-ši (KBo 25.102 ii 6 (OS)) and kar-di-i=š-mi (StBoT 25.3 i 12 (OS)). Although no attestations with DA are found, the alternation between TI and DI37 points to the presence of a voiced stop: [gərd-].38 This is supported by the attestations from MS and NS texts, where the oblique cases of ‘heart’ are spelled kar-ta(-) as well as kar-da(-), and kar-ti(-) as well as kar-di(-),39 pointing to the presence of a voiced dental stop as well: [gərd-]. The development of PIE *-rd- to OH [-rd-], with a voiced stop, differs, however, from the development of PIE *-nd-, which, through pre-PAnat. *[-nʔt-], yielded OH [-nt-], with a voiceless stop, which phonemically was fortis, /-ntː-/. We therefore must assume that in the PIE cluster *-rd-, which for pre-Proto-Anatolian can be assumed to have been *[-rʔt-], first the preglottalic feature of the dental stop was lost, yielding pre-Hitt. *[-rt-], after which the cluster underwent voice assimilation to OH [-rd-]. Since this latter cluster contrasts with the cluster [-rt-] as found in arta(ri), we should interpret [-rd-] phonologically as /-rt-/, with the lenis stop /t/, and [-rt-] as /-rtː-/, with the fortis stop /tː/. The outcome of the PIE cluster *-rTH- may be visible in the verb šarta-i / šart- ‘to wipe, to rub’, if this really reflects a root *serdhh2/3-.40 In OS texts, this verb is attested three times, namely in 3sg.pres.act. šar-ta-i (KBo 17.18 ii 16 (OS), KBo 17.43 i 14 (OS), KUB 36.110 rev. 20 (OS)). Although the numbers are low, the absence of spellings with the sign DA seems to indicate that the dental stop of this word was not ejective, but rather plain voiceless: [sartai]. After r, the cluster *-TH- therefore seems to behave in the same way as after obstruents, i.e. it 37

If we are allowed to assume that in Old Hittite times, just as after n, so also after r the pair TI vs. DI shows the same distribution in spelling as TA vs. DA. 38 For the assumption that the initial consonant was phonetically a voiced stop [g-], see Kloekhorst 2014: 426 n. 1666. 39 Cf. Puhvel HED: 190–191 for attestations. 40 Cf. Kloekhorst 2008: 737–738 for this reconstruction.

The Phonetics and Phonology of the Hittite Dental Stops

163

loses its laryngeal without causing glottalization. However, since *-rdhH- seems to have yielded OH [-rt-], and not [-rd-], we may assume that the laryngeal did lengthen the preceding stop before it was totally lost. So the PIE cluster *-rdhH-, which in pre-Proto-Anatolian terms can be written as *[-rtʔ-], first yielded pre-Hitt. [-rtːʔ-] (through lengthening of the *[t] because of the following laryngeal), after which its outcome in Old Hittite was [-rt-]. Although the verb šarta-i / šart- is not well attested in younger texts,41 we do find a 3sg.pres. act. form [šar-]da-a-iz-zi (Bo 4869 ii 3 (Neu 1980: 103) (undat.)), which may indicate that the post-OH form of this verb was [sard°], with a voiced stop. The development of OH [-rt-] to MH/NS [-rd-] would then be identical to the development of OH [-nt-] (the outcome of PIE *-nd- = pre-PAnat. *[-nʔt-]) to MH/NH [-nd-]. For the position after l, the material is likewise scanty. There are, as far as I know, no good examples for the development of the PIE clusters *-lt- and *-ld-. The cluster *-ldh- is attested in the verb mālt-i / malt- ‘to recite’, which is generally derived from the PIE root *meldh-.42 Its 3sg.pres.act. form is in OS texts attested 10 times as ma(-a)-al-di, with the sign DI, and once as [ma-]ạ-al-ti, with the sign TI.43 Although the relative number of attestations with the sign DI is remarkably high, I assume that this form must be interpreted as [mā́ldi], with a voiced stop [d].44 This would mean that the PIE cluster *-ldh-, which should correspond to pre-PAnat. */-lt-/, through voice assimilation yielded OH [-ld-]. Also in MS and NS texts, we find the spelling ma(-a)-al-di next to ma(-a)al-ti, but also 1sg.pres.act. ma-al-da-aḫ-ḫi besides ma-al-ta-aḫ-ḫi,45 clearly pointing to the presence of a [d]. The PIE cluster *-lTH- may be visible in the word for ‘shoulder’, paltan-, which can be reconstructed as *pélth2-n, *plth2-én-.46 In OS texts, it is attested

41 Cf. CHD Š: 290–291. 42 Cf. Kloekhorst 2008: 550–551. 43 Cf. Kloekhorst 2014: 268 for attestations. 44 Again, assuming that in Old Hittite times, just as after n, so also after l the pair TI vs. DI shows the same distribution in spelling as TA vs. DA. 45 Cf. CHD L-N: 132 for attestations. 46 Although this word is usually cited as an a-stem paltana- (thus CHD P: 79–80; Kloekhorst 2008: 622; Puhvel HED 8: 76–79; Tischler HEG P: 401–402), Giorgieri (1992: 72–74) has convincingly argued that the OH instr. form paltant (KBo 30.30 rev. 5 (OS) [pa]l-ta-an-t=aat=kán ~ KUB 58.111 rev. 13 (OH/NS) pal-[t]a-an[-t=a-at=kán]) shows that the noun for ‘shoulder’ originally was an n-stem, and not an a-stem. According to Giorgieri, n-stem forms are also found in the two nom.sg. forms attested in KUB 43.53 i 7, 24. The former of these (i 7) is cited in CHD (P: 80) as “pa[l]-t[a-n]a-aš-ša-pa” (following Neu (1980: 26), who reads the form as “pạl-t[a-n]a-aš-ša-pa”), i.e. as paltanašš=a=pa, but Giorgieri rather

164

Kloekhorst

only once, namely in the instr. form [pa]l-ta-an-t=a-at=kán (KBo 30.30 rev. 5 (OS)). The spelling with the sign TA instead of the sign DA seems to indicate that this word did not contain an ejective stop. We therefore may assume that after *l the sequence *-TH- behaves the same as after *r and obstruents, namely that it loses its laryngeal without causing glottalization. Although the one attestation with the sign TA is not enough to prove whether the stop was voiceless or voiced, I assume that, just as in šartai = [sartai], the OS form paltant represents [pəltan-], with a voiceless [t]. In MS and NS texts, we find attestations spelled with the sign DA as well (cf. CHD P: 79 for attestations), showing that the OH sequence [-lt-] in younger times has undergone voice assimilation to [-ld-]. All in all, we can conclude that after *r and *l the outcome of the dental stops in Old Hittite seems to be the same as after obstruents, namely that etymological *t yields the fortis stop /tː/, etymological *d and *dh merge into the lenis stop /t/, and clusters with a laryngeal, *-TH-, after having undergone lengthening of the dental stop, lose their laryngeal without causing glottalization and thus merge with the fortis stop /tː/. The only difference is that after r and l the distinction between the fortis and lenis stops is phonetically realized as a difference in voice, namely [t] vs. [d]. Moreover, after the Old Hittite p ­ eriod reads it as pạl-t[a-aš-]ši-ša-pa. The latter (i 24) is cited in CHD (P: 80) as “[pal-ta-n]a-ašši-ša!(text -ta)-aš-ta” (whereas Neu (1980: 26) reads “[pal-t]a--aš-ši!-ta-aš-ta”), i.e. paltanaš=šiš=ašta, but Giorgieri is clearly right in reading [pal-t]a-aš-ši-ta-aš-ta. Because of the form pạl-t[a-aš-]ši-ša-pa in i 7, which Giorgieri analyses as paltaš=šiš=apa, he assumes that the form [pal-t]a-aš-ši-ta-aš-ta must be analyzed as ‘paltaš=šiš!=ašta’. The form paltaš would then be the nom.sg. form of a common-gender n-stem noun paltan-, just as the nom.sg. form of the common-gender n-stem noun ḫāran- ‘eagle’ is ḫāraš. I fully agree with Giorgieri’s analyses, except for one detail: I would personally rather interpret the form [pal-t]a-aš-ši-ta-aš-ta attested in line i 24 as representing palta(n)=ššit=ašta, and read the form from line i 7 as pạl-t[a-aš-]ši-tạ-pa = palta(n)=ššit=apa, i.e. as containing a nom.sg. form paltan from a neuter n-stem paltan-. In this way, the several Sumerographic acc.sg. forms uzuZAG(.LU)-an (cf. CHD P: 79 for attestations) may then be seen as representing paltan, and not paltanan. Moreover, the interpretation of this noun as neuter would also better fit the acc.pl. form pal-ta-na (KBo 8.91 obv. 15 (MS)), which in CHD (P: 79) is unconvincingly read as pal-ta-na[-aš]. Furthermore, it explains the suffixal accentuation in dat.-loc.sg. paltani /pəltáni/ and dat.-loc.pl. paltānaš /pəltánas/ (cf. Kloekhorst 2014: 456), which can now be explained by reconstructing a PIE proterodynamic neuter n-stem *pélth2-n, *plth2-én- (cf. e.g. Kloekhorst 2008: 622 for the root etymology). The specific a-stem forms that point to a common gender noun paltana- are only found in New Hittite texts (nom.sg.c. pal-ta-na-aš (KBo 1.42 ii 13, iv 14 (fr.) (NH/NS)), acc.sg. UZUpalta-na-a[n] in Bo 3640 iii? 9 (NS), acc.pl.c. pal-ta-nu-uš (KBo 1.42 ii 32 (NH/NS)), and, as Giorgieri (1992: 73) stated, can easily have been the result of a NH thematizisation of an original n-stem paltan-.

The Phonetics and Phonology of the Hittite Dental Stops

165

the fortis stops become lenis stops, which phonetically can be explained as a case of voice assimilation. 8

Dental Stops before Consonants

When standing before another consonant, dental stops are usually only written with signs of the shape Vt, which in Akkadian can be read Vd and Vṭ as well. Such spellings (e.g. ḫa-at-k° ‘to close’) therefore do not say anything about the phonetic realization of these stops. Occasionally, we find an alteration in the spelling of a dental stop before a consonant. For instance, the verb ‘to install’ is spelled ti-it-nu-, but more often ti-it-ta-nu-. This implies that the dental stop was long and voiceless: [titːnu-]. Another case is the verb ‘to cause to dry up’, which is spelled ḫa-at-nu-, but also once ḫa-da-nu-. This seems to imply that in this verb the dental stop was short and voiced: [hadnu-]. A third case is the verb ‘to confiscate’, which is usually spelled ap-pa-at-ri°, but also once ap-pata-ri°. This implies that the stop was short, but it cannot be decided whether it was voiced or voiceless: [əpːatrie/a-] or [əpːadrie/a-]. Note that in all cases the dental stop precedes a resonant. We may therefore assume that only here, a distinction between fortis /tː/ (realized as a long voiceless stop [tː]) and lenis /t/ (realized as a short voiced stop [d]) was made. There is no evidence for the ejective /tːʔ/ in this position. Since tit(ta)nu-zi etymologically probably reflects *dhi-dhh1-neu-, we may assume that before consonants original ejectives eventually merged with the fortis stops. Although before resonants a distinction between /tː/ and /t/ was made, we may assume that before stops this distinction was neutralized. Since e.g. ḫatk-i is never spelled **ḫa-at-ta-k°, **ḫa-ta-k° or **ḫa-da-k°, we may assume that the phonetic realization of the dental stop in this position was short and voiceless: [t]. 9

Dental Stops in Word-initial Position: the OH Situation

In Kloekhorst 2010a: 202–207 and Kloekhorst 2016, I treated the spelling of dental stops in word-initial position in Old Hittite texts, and argued (1) that consistent spelling with the sign TA denotes the presence of a plain voiceless stop [t], which corresponds to PIE *t, *d, and *dh; (2) that consistent spelling with the sign DA rather points to the presence of a postglottalized stop [tʔ], the outcome of PIE *TH-; and (3) that alternation in spelling between the signs TA and DA represents the presence of a voiced stop [d], which only occurs in

166

Kloekhorst

loanwords. It was argued that we should equate these three stops with the ­intervocalic ones in the following way: 1. The word-initial voiceless stop [t] is to be equated with the fortis stop /tː/. 2. The word-initial postglottalized stop [tʔ] is to be equated with the ejective stop /tːʔ/. 3. The word-initial voiced stop [d] is to be equated with the lenis stop /t/. As we have seen, the voiceless stop [t], which we now can identify as the fortis stop /tː/, derives from PIE *t, *d, and *dh, which means that these apparently have merged at some point in the prehistory of Hittite. This merger can be dated on the basis of the following argumentation. Dental stops followed by the vowel *i are subject to assibilation in Hittite. This assibilation does not occur in Luwian (cf. CLuw. tiu̯ at- ‘sun-god’ vs. Hitt. šī̆u̯att- ‘day’ < *diéu̯ ot-) and therefore cannot have been Proto-Anatolian, but must have been specifically Hittite. Since the outcome of word-initial *ti̯-, which yields Hitt. z- [ts-],47 is different from the outcome of *di̯-, which yields Hitt. š- [s-],48 we see that at this moment in time the fortis and the lenis stop were still phonemically distinct.49 The merger of word-initial PIE *t, *d, and *dh into a single stop [t-] = /tː-/ must therefore have been specifically Hittite as well. Apart from giving evidence for the relative dating of the merger of the initial fortis and lenis stops, it was argued in Kloekhorst 2016 that the assibilation also provides a crucial argument for determining the exact phonetic difference between the fortis and lenis dental stops at that moment. As we have seen, the outcome of PIE *ti̯- is Hitt. z-, i.e. [ts-], whereas the outcome of PIE *di̯- is š-, i.e. [s-]. The difference in outcome between the two would be inexplicable if the two clusters differed phonetically from each other in voice: we would then expect either an outcome [ts-] vs. [dz-], or [s-] vs. [z-], but not [ts-] vs. [s-]. Instead, the difference in outcome between *ti̯- and *di̯- can only be explained by assuming that at that time the phonetic difference between the fortis and the lenis dental stops was one of consonantal length: *ti̯- = *[tːj-] > *[tːj-] > [ts-], whereas *di̯- = *[tj-] > *[tj-] > [s-].50 In other words, at the time of ­assibilation, i.e. in post-Proto-Anatolian, pre-Hittite times, the fortis dental stop (­corresponding 47 48 49

E.g. *tieh2- > Hitt. zaḫḫ- ‘battle’, cf. Kloekhorst 2008: 1019–1020. E.g. *diēu̯ - > Hitt. šīu̯ - ‘god’, cf. Kloekhorst 2008: 763–764. Although we do not have good examples for the outcome of PIE *dhi̯- in Hittite, it seems safe to assume that it would have yielded the same result as *di̯- since in almost all other contexts the PIE stops *d and *dh have merged in Hittite as a lenis stop. 50 Cf. Kloekhorst 2008: 92 for this analysis, although at that time I had not yet realized its consequences.

The Phonetics and Phonology of the Hittite Dental Stops

167

to PIE *t) was a long voiceless stop *[tː-], and the lenis stop (corresponding to PIE *d and *dh) a short voiceless stop *[t-].51 Later on, in Old Hittite, the two merged into a short voiceless stop [t-], which is consistently spelled with the sign TA. Although phonetically the fortis and the lenis stop have merged by loss of length of the fortis stop, as is typologically common,52 it is synchronically best to phonologically interpret the outcome of this merger as the fortis stop /tː/, since a new word-initial dental stop [d-] has entered the language through loanwords (spelled both with TA and with DA), which w ­ ithin the overall phonological system of Old Hittite is best interpreted as corresponding to the lenis stop /t/ that in some other environments is realized as [d] as well. 10

Dental Stops in Word-initial Position: the MH/NH situation

Words that in OS texts are consistently spelled with the sign DA (dā-i / d- ‘to take’ and dai-i / ti- ‘to put’) show also in MS and NS texts consistent spelling with DA. However, words that in OS texts are exclusively spelled with the sign TA are spelled in MS and NS texts with the sign DA as well. Consider for instance the word tamai- / tame- ‘other’, whose ratio of spellings with the sign TA to spellings with the sign DA is in OS texts 7 : 0 = 100%, in MH/MS texts 22 : 3 = 88%, and in NH/NS texts 55 : 44 = 56%. Likewise the noun tagān(ze/ipa-) ‘earth’, whose ratio of TA to DA is in OS texts 6 : 0 = 100%, in MS texts 17 : 1 = 94%, and in NS texts 13 : 62 = 17%.53 In both cases we see that, although in OS texts these words are exclusively spelled with the sign TA, in MS and especially in NS texts we encounter many spellings with the sign DA as well. In Kloekhorst 2010a: 209, I tried to explain this phenomenon by stating that “the sign DA is taking over the place of TA, eventually on its way to ousting it completely”, and that “[t]his probably indicates that on a phonetic level, the opposition between word-initial /ta-/ and /tʔa-/, which was still present in OH times, is disappearing from MH times onwards”. In the meantime I have changed my mind, however. I now regard the fact that dā-i and dai-i keep on being consistently spelled with the sign DA, also in MS and NS texts, as an indication that their initial ejective stop was retained as such in post-OH times. So /tːʔā-/ = [tʔā-] ‘to take’ and /tːʔai-/ = [tʔai-] ‘to put’ remain unaltered throughout the history of Hittite. Moreover, in the case of words that in OS texts are exclusively 51

As stated in Kloekhorst 2016, although phonemic consonantal length in word-initial position is cross-linguistically rare, it certainly is attested, for instance in the Thurgovian ­dialect of Swiss German, in Pattani Malay, in Leti, etc. 52 Cf. Kümmel 2007: 135. 53 Cf. Kloekhorst 2010: 208.

168

Kloekhorst

spelled with the sign TA but in MS and NS texts start being spelled with the sign DA as well, I now regard this fact as an indication that their initial voiceless stop [t-] after the Old Hittite period starts to undergo voicing to [d-]. Since this development of OH voiceless [t-] to MH/NH [d-] in phonological terms can be described as a development of fortis /tː-/ to lenis /t-/, we can say that this is a case of a post-OH word-initial lenition. All in all, we can set up the following chronology of sound laws to account for the dental stops in initial position. Table 2

PIE

Development of PIE dental stops in initial position

pre-PAnat. (1) PAnat. (2) pre-Hitt. (3) OH

*t-

~

*[tː-]

*d-

~

*[ʔt-]

*dh-

~

*[t-]

>

}

*[tː-]

>

*[tː-]

*[t-]

>

*[t-]

}

[t-] = /tː-/

influx from loanwords: [d-] = /t-/

(4) MH/NH

}

[d-] = /t-/

*tiV-

~

*[tːiV-]

>

*[tːiV-]

>

*[tsV-]

>

[tsV-] = /tsːV-/

=

[tsV-] = /tsːV-/

*diV-

~

*[ʔtiV-]

*dhiV-

~

*[tiV-]

}

*[tiV-]

>

*[sV-]

>

[sV-] = /sV-/

=

[sV-] = /sV-/

*tH-

~

*[tːʔ-]

>

*[tːʔ-]

*dH-

~

*[ʔtʔ-]

>

[tʔ-] = /tːʔ-/

=

[tʔ-] = /tːʔ-/

~

*[tʔ-]

}

*[tːʔ-]

*dhH-

}

*[tʔ-]

(1) Loss of the preglottalic feature of *[ʔt-], causing it to merge with *[t-]. (2) Assibilation of *[tː-] and *[t-] because of a following *i; lengthening of *[t-] to *[tː-] because of a following */ʔ/. (3) Phonetic loss of length in word-initial position, causing *[tː-] and *[t-] to merge as [t-], and [tːʔ-] to shorten to [tʔ-]. Influx of [d-] from loanwords. (4) Merger of OH [t-] and [d-] into MH/NH [d-]. 11

Dental Stops in Word-final Postvocalic Position

In word-final postvocalic position, there is only one way that dental stops are spelled, namely with the signs at, e/it, and ut. It is therefore usually assumed

The Phonetics and Phonology of the Hittite Dental Stops

169

that Hittite knew only one type of dental stop in this position. Since in ­Akkadian these signs are ambiguous with regard to the dental stop they contain (beside Vt, they can be read Vd and Vṭ as well) we cannot on the basis of these signs say anything about the phonetic rendering of the word-final dental stops. We therefore have to look for other evidence. On the basis of the form pa-i-ta-aš ‘he went’, which consists of the 3sg.pret. form pait ‘went’ to which the enclitic pronoun =aš ‘(s)he’ is added, and in which the word-final dental stop of pait is intervocalically spelled as a singleton, i.e. as a lenis stop, Melchert (1994: 85) states that in word-final position “[v]oiced stops have been generalized” (note that according to Melchert lenis stops were distinctively voiced).54 Yet as I have argued in Kloekhorst 2008: 24 and Kloekhorst 2016, the form paitaš cannot be used as evidence, since the single spelling of t in this form is grammatically relevant. More telling is the case of the gen.sg. of the word šeppitt- ‘grain’. In OS texts, this form is spelled še-ep-pí-da-aš, pointing to a phonetic form [sepːidas], which can phonologically be interpreted as /sépːitas/, with a stem-final lenis /t/ = [d]. It is generally assumed that this lenis /t/ derives from PIE *t through the second Anatolian lenition rule, which states that original fortis stops are lenited when standing between two unaccented vowels in posttonic position,55 so PIE *sépitos > OH [sépːidas] /sépːitas/. Already in Old Hittite, the form še-eppí-da-aš is replaced by še-ep-pí-it-ta-aš, however, with geminate spelling of the -tt-, pointing to the presence of a long voiceless stop [tː], which can phonologically be interpreted as fortis /tː/. It is commonly thought that this means that the original stem-final fortis consonant of šeppitt- < *sépit- has been restored throughout the paradigm. The question is, however, what the exact model was for this restoration. As I have argued in Kloekhorst 2016, all oblique cases of the word šeppitt- (including the nom.-acc.pl. form) contained an ending starting in a vowel, *sép-it-V°, which means that in all these forms the stem-final *t regularly would have undergone lenition to /t/ = [d]: *sép-it-V° > Hitt. /sépːitV°/ = [sépːidV°]. These forms therefore cannot have been the source on the basis of which the fortis /tː/ was generalized. This means that we are only left with the nom.-acc.sg.n. form šeppit as the possible source for restoration of the stemfinal fortis /tː/. As a consequence, we must assume that this word represents /sépːitː/ = [sépːitː], containing a word-final postvocalic fortis /tː/. Since in the labiovelar series there is evidence for a d­ istinction between word-final lenis and fortis stops, namely in tak-ku /takwː/ ‘if’ < *tokwe vs. e-ku /ʔḗkw/ ‘drink!’ < *h1égwh, it was argued in Kloekhorst 2016 that it is likely that this distinction 54

A view that has been followed by many scholars, e.g. Vanséveren 2006: 40; Hoffner and Melchert 2008: 36; Rieken 2011: 40; van den Hout 2011: 65. 55 Cf. Eichner 1973: 10086; Morpurgo Davies 1982/83: 262; Kloekhorst 2014: 559–564.

170

Kloekhorst

was made in the dental series as well, and that še-ep-pí-it /sépːitː/ < *sépit probably contrasted with e.g. e-et /ʔḗt/ ‘eat!’ < *h1éd.56 As far as I am aware, there is, besides fortis /tː/ and lenis /t/, in word-final, postvocalic position no trace of a third phoneme that can be identified with the ejective phoneme /tːʔ/. 12

Dental Stops in Word-final Postconsonantal Position

In word-final postconsonantal position the presence of dental stops is rare, since in pre-Hittite a sound law *-CT# > *-C# has taken place (e.g. nom.-acc. sg.n. appan ‘taken’ < *h1pónt). The few cases of word-final postconsonantal dental stops that we do find must therefore all be the result of restoration. One such case is found in the OH adverb mānḫanda, māḫḫanda ‘just as’ (the latter of which is the regular outcome of the former within Old Hittite). This adverb is spelled in OS texts with the sign DA (ma-a-an-ḫa-an-da, ma-aaḫ-ḫa-an-da) as well as TA (ma-a-aḫ-ḫa-an-ta),57 which points to the presence of a [d]. Since this word develops in the post-OH period to māḫḫan, I have in Kloekhorst 2010b argued that the dental stop in mānḫanda, māḫḫanda was word-final, [mānhand] > [māhːand], which in the post-OH period regularly was lost, yielding MH māḫḫan [māhːan]. The original form mānḫanda [mānhand] was explained as a univerbation of the adverb [mān] and a form [hand], which was argued to originally have been the nom.-acc.sg. form of the noun ḫant- ‘forehead’ < PIE *h2ent-. The word-final dental stop of [hand] must then have been restored on the basis of the other forms of the paradigm of this word, which contained a [d] as well (cf. the OS spelling of the dat.-loc.sg. form ḫa-an-ti58 besides ḫa-an-di,59 pointing to [handi]), which is the regular outcome of PIE *t after an *n. Since the [d] of [hand] is taken over from the forms of the paradigm in which it stood in word-internal position, the word mānḫanda, māḫḫanda cannot be used as an argument for the development of word-final dental stops per se. Another group of words where we find a word-final dental stop is formed by archaic instrumentals of r-stems and r/n-stems. Especially the word denoting ‘by hand’ is telling, since it is spelled both ki-iš-šar-ta and ki-iš-šar-at, indicating 56

As stated in Kloekhorst 2016, although crosslinguistically it is rare to find a contrast in consonantal length in word-final position, there are certainly languages that have such contrasts, like e.g. Tashlhiyt Berber, Moroccan Arabic, the Wixli dialect of the Lak language, and Tabasaran. 57 Cf. Kloekhorst 2010b: 2185 for attestations. 58 ḫa-an-ti (KBo 6.2 ii 8 (OS), KBo 17.30 iii 6 (OS)). 59 ḫa-an-di (IBoT 1.121 rev. 17 (OS), KBo 25.37 rev. 8 (OS), KBo 25.38, 7 (OS)).

The Phonetics and Phonology of the Hittite Dental Stops

171

that the dental stop in such instrumentals is really word-final. The dental stops of these forms must have been restored on the basis of instrumentals of nouns with a stem ending in a vowel (e.g. ganut ‘by knee’). Etymologically, this dental stop is often reconstructed as *-d, but this need not be correct.60 As I have explained in Kloekhorst 2008: 799, the Anatolian evidence rather points to an original *-t.61 On the basis of the treatment of word-final dentals in postvocalic position as given above, we would now expect that after vowels this ending would have yielded Hitt. /-tː/. In OS texts, we find that the instrumental of r/n-stems is spelled as follows: ša-kán-da (KBo 22.2 obv. 2 (OS)) ‘with grease’ and ú-i-ta-an-ta (StBoT 25.56 i 5 (OS)) ‘with water’. The fact that the ending is spelled both with the sign TA and with the sign DA indicates that it consists of a voiced stop, [-d]. This contrasts with its postvocalic shape, which is /-tː/. Apparently, also in word-final position an original */tː/ when standing after n was shortened and subject to voice assimilation, just as it was in word-internal position.62 In MS and NS texts, we do find instrumentals in -anda and -anta as well, but these are usually found in younger copies of OH compositions. In MH compositions, the instrumentals of r/n-stems rather end in -enit, a renewed form that undoubtedly was created because the OH instrumental ending in [-and] regularly lost its word-final stop (just like OH māḫḫanda [māhːand] lost its stop, yielding MH māḫḫan [māhːan]). In the case of the instrumental of the word for ‘hand’, we find in OS texts three attestations spelled with the sign TA, ki-iš-šar-ta,63 but none with the sign DA. Although numbers are low, this could mean that in this word the ending consists of a voiceless stop, [-t]. If this is indeed the case, it would mean that, unlike after n, after r no voice assimilation has taken place. The difference in the outcome of the dental stop /-tː/ after r and n need not surprise too much: 60

Although Sanskrit did not know an opposition between word-final t and d, its ablative forms that correspond to the Hittite instrumental are often cited as ending in -d, e.g. mád ‘from me’, tvád ‘from you’, etc. This is undoubtedly done on the basis of the presence of a d in the OLat. thematic abl.sg. ending -ōd (> Class. Lat. -ō). However, since word-final *-t regularly yielded OLat. -d (e.g. 3sg.opt. *h1siéh1t > OLat. sied ‘he be’; Weiss 2009: 155), we cannot on the basis of Sanskrit and Latin decide whether we should reconstruct the ending with a *t or a *d. 61 The argument runs as follows. Since within Hittite the ablative in -z, which can only ­reflect pre-Hittite *-ti and not *-d(h)i, can be seen as a derivative of the instrumental in -(e/i)t (addition of the locative particle / ending *-i), it strongly suggests that the latter ending goes back to *-(e)t with a *t. 62 Note that these developments must have taken place after the ending of the instrumental was restored in these r/n-stems, which may be information that can be used when setting up a relative chronology of the linguistic prehistory of Hittite. 63 Cf. Kloekhorst 2014: 422 for attestations.

172

Kloekhorst

as we have seen above, in word-internal position r and n also have a different effect on the dental stops following them. Just like in the instrumentals in -anda, -anta, the OH instrumental form kiššarta [kɨsːárt] is replaced in the post-OH period, namely by kiš(ša)ret [kɨsːrétː], probably because the postconsonantal word-final stop of the former form was regularly lost.64 If it is indeed correct that in Old Hittite after an r the instrumental ending is [-t], whereas after an n it is [-d], we may assume two different phonemes in this position, and I would equate the voiceless stop [t] with the fortis phoneme /tː/, and the voiced stop [d] with the lenis phoneme /t/. As far as I am aware, there is in word-final, postconsonantal position no trace of a third phoneme that could be identified with the ejective phoneme /tːʔ/. 13

Dental Stops: Conclusion

We can conclude that Hittite knew three phonemically distinct dental stops: a fortis one, /tː/; an ejective one, /tːʔ/; and a lenis one, /t/. In Old Hittite, the distinction between fortis and lenis was present in all positions in the word; it was not until after the OH period that in some specific environments the fortis and lenis series merged. The ejective is only distinctively present in word-initial, intervocalic and postnasal position: in all other environments it had probably merged with the fortis series already in pre-Hittite times. The phonetic realization of the three phonemes differs by environment, as indicated in the table below. Table 3

Phonetic realizations of Hittite /tː/, /tːʔ/, /t/

Phonological Phonetic realizations per environment value #TV- -VTV- -nTV- -r/lTV- -CTV- -tR- -tC- -RT# -VT# fortis /tː/ ejective /tːʔ/ lenis /t/

[t] [tʔ] [d]

[tː] [tːʔ] [d]

[t] [tʔ] [d]

[t] – [d]

[tː] – [t]

[tː] – – – [d] [t]

[t] – [d]

[tː] – [t]

Although in word-initial position and after resonants the synchronic phonetic distinction between the fortis and the lenis stops is one of voice ([t] vs. [d], 64 Cf. Kloekhorst 2014: 422–423 for a detailed analysis of the inner-Hittite diachronic development of the instrumental form of ‘hand’.

173

The Phonetics and Phonology of the Hittite Dental Stops

respectively), it was argued above that for word-initial position this distinction originally was one of consonantal length and not of voice. I therefore assume that this originally was the case in the position after resonants as well, just as it synchronically still is after obstruents, so *[tː] vs. *[t]. In intervocalic position, the phonetic distinction between the fortis and the lenis stops is one of both length and voice, namely [tː] vs. [d], respectively. Because the distinction is marked in a twofold, and therefore redundant, way, and because after obstruents the distinction is one of length only (as it originally was in word-initial position), it seems obvious to me that the voiced character of the lenis stops in intervocalic position is allophonic. I therefore regard it as justified to set up for all positions in the word an underlying phonemic difference for the fortis vs. lenis stops that consists of consonantal length only: /tː/ vs. /t/. The ejective stop is realized as a postglottalized long stop [tːʔ] in intervocalic position, but as a postglottalized short stop [tʔ] in word-initial and postnasal position (in the other positions it does not seem to occur). Since the fortis stop, which underlyingly is long [tː], is in word-initial and postnasal position ­realized as a short stop [t], we may assume the same for the ejective stop. This indicates that the length of the intervocalic variant is original, and that the underlying phoneme should be set up as a postglottalized long stop /tːʔ/. One could argue, however, that in this way it is redundantly marked vis-à-vis the fortis and the lenis stops (/tː/ and /t/, respectively), and that it would suffice to set up the ejective stop as postglottalized only, /tʔ/. However, since in intervocalic position consonantal length is relevant as far as whether the preceding vowel stands in an open or closed syllable, I prefer to keep the long character of the ejective stop in my phonemic representation, and therefore write /tːʔ/. An overview of the development of PIE *t, *d, *dh and *TH in Hittite can be given as follows (note that the outcomes are given in their phonetic, not phonological shape): Table 4

Development of PIE dental stops and *TH in Hittite

#TVPIE

prePAnat.

*TH

*[tːʔ]

*t

*[tː]

*dh

*[t]

*d

*[ʔt]

-VTV-

OH M/NH O/M/NH [tʔ]

-nTVOH

[tːʔ]

[t]

[d]

[d]

-CTV-

M/NH OH M/NH O/M/NH [tʔ]

[tː]

-r/lTV-

[t]

[d]

[tː]

[d] [t]

[d]

[d]

[t]

174

Kloekhorst

Table 4.

Development of PIE dental stops and *TH in Hittite (cont.)

PIE

pre-PAnat.

*TH

*[tːʔ]

*t

*[tː]

*dh

*[t]

*d

*[ʔt]

-tR-

-tC-

-VnT#

-VrT#

-VT#

O/M/NH

O/M/NH

OH

OH

O/M/NH

?

?

?

[d]a

[t]a

[tː]

?

?

[t]

[tː] [t] [d]

a When restored analogically.

14 Outlook The phonetic and phonological interpretations of the Hittite dental stops as presented in this article are for the most part based on an analysis of the distributions in usage of the signs TA and DA. The distributions of the signs TE, TI and TU on the one hand, and the signs DE/I and DU on the other, have only been taken into account in the analysis of the spelling of dental stops following resonants (n, r, and l), and in the spelling of initial stops in Old Hittite (cf. Kloekhorst 2010a: 209–211). A full analysis of their usage in other positions in the word remains an important task for the future. References Eichner, H. (1973) Die Etymologie von heth. mehur, MSS 31, 53–107. Giorgieri, M. (1992) Un rituale di scongiuro antico ittita per Labarna-Ḫattušili, SMEA 29, 47–98. Hoffner Jr., H.A. and Melchert, H.C. (2008) A Grammar of the Hittite Language, Winona Lake, IN: Eisenbrauns. Kimball, S.E. (1999) Hittite Historical Phonology [IBS 95], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Kloekhorst, A. (2008) Etymological Dictionary of the Hittite Inherited Lexicon [LIEEDS 5], Leiden, Boston: Brill. Kloekhorst, A. (2010a) Initial stops in Hittite (with an excursus on the spelling of stops in Alalaḫ Akkadian), Zeitschrift für Assyriologie 100, 197–241. Kloekhorst, A. (2010b) Hittite mān, maḫḫan, māḫḫan, māḫḫanda and mānḫanda, in: R. Kim et al. (eds.), Ex Anatolia Lux. Anatolian and Indo-European Studies in Honor

The Phonetics and Phonology of the Hittite Dental Stops

175

of H. Craig Melchert on the Occasion of his Sixty-Fifth Birthday, Ann Arbor, New York: Beech Stave, 217–226. Kloekhorst, A. (2012) The phonological interpretation of plene and non-plene spelled e in Hittite, in: B. Nielsen Whitehead et al. (eds.), The Sound of Indo-European. Phonetics, Phonemics, and Morphophonemics, Copenhagen: Museum Tusculanum, 243–261. Kloekhorst, A. (2013) The signs TA and DA in Old Hittite: evidence for a phonetic difference, AoF 40, 125–141. Kloekhorst, A. (2014) Accent in Hittite: A Study in Plene Spelling, Consonant Gradation, Clitics, and Metrics [StBoT 56], Wiesbaden: Harrassowitz. Kloekhorst, A. (2016) The Anatolian stop system and the Indo-Hittite hypothesis, IF 121, 213–248. Kouwenberg, N.J.C. (2003) Evidence for post-glottalized consonants in Assyrian, JCS 55, 75–86. Kümmel, M.J. (2007) Konsonantenwandel. Bausteine zu einer Typologie des Lautwandels und ihre Konsequenzen für die vergleichende Rekonstruktion, Wiesbaden: Reichert. Melchert, H.C. (1994) Anatolian Historical Phonology [LSIE 3], Amsterdam, Atlanta: Rodopi. Melchert, H.C. (2008) Problems in Hittite pronominal inflection, in: A. Lubotsky et al. (eds.), Evidence and Counter-Evidence. Essays in Honour of Frederik Kortlandt, Vol. 1: Balto-Slavic and Indo-European Linguistics [SSGL 32], Amsterdam, Atlanta: Rodopi, 367–375. Morpurgo Davies, A. (1982/83) Dentals, rhotacism and verbal endings in the Luwian languages, ZVS 96, 245–270. Neu, E. (1980) Althethitische Ritualtexte in Umschrift [StBoT 25], Wiesbaden: Harrassowitz. Patri, S. (2009) La perception des consonnes hittites dans les langues étrangères au xiiie siècle, ZA 99, 87–127. Rieken, E. (2011) Einführung in die hethitische Sprache und Schrift, Münster: Ugarit-Verlag. van den Hout, Th. P.J. (2011) The Elements of Hittite, Cambridge: Cambridge University Press. Vanséveren, S. (2006) Nisili. Manuel de langue hittite, Volume i, Leuven: Peeters. Weiss, M. (2009) Outline of the Historical and Comparative Grammar of Latin, Ann ­Arbor, New York: Beech Stave.

Chapter 9

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi Martin Joachim Kümmel Abstract Many Hittite verbs in -ja- show a 3rd singular present in -ia-Iz-zi and similar forms. These have been explained as resulting from Luwian influence. However, the phenomenon also occurs in cases where such an influence is improbable, and is extremely widespread. The distribution of forms written with -ia-I° points to an innovation for older -i-e-I°, parallel to slightly earlier variant writings with -ia-a° in the same forms (while other forms always had only -ia-a°). It is argued that this variation is best explained by phonetic similarity of /ja/ and /je/ resulting in the possibility of using -iafor /je/ and thus allowing scribes to write -ia-I° in the sense of -je-e°. Thus, there was a graphical innovation but no real linguistic change.

Keywords Luwian influence – verb inflection – writing – phonetics – graphic innovation

1 Einführung1 Oettinger (1979, 27, 382) konstatiert, dass sich „bei manchen i̯e-Stämmen“ sekundär nur in den dritten Personen des Singulars ein Stammauslaut °(i)i̯aezeige, „andere Singularpersonen sind kaum, Pluralpersonen nicht betroffen“, d.h. die übrigen Formen würden gewissermaßen von den i̯e-Stämmen sup­ pliert. Ibid. S. 382 wird diese Tendenz beim Prototyp, nach dem sich dann andere gerichtet hätten, auf keilschriftluwische Vorbilder zurückgeführt, z. B. kluw. šapiyai-mma- → heth. Imperativ ša-pí-ia-i. 1

Mitgemeint sind auch alle anderen Formen mit der gleichen Graphie, also auch 3. Sg. Prät. -ia-It; 3. Sg. Med. -ia-It-ta(-). Für Diskussion zu danken habe ich H. Craig Melchert und Alwin Kloekhorst. Die syllabische Transliteration folgt der Praxis des CHD, ebenso die Abkürzungen für Text- und Schriftklassen sowie Belegangaben. Die interpretierende Umschrift verzichtet auf unnötige Diakritika, also s = š, h = ḫ, j = i̯, w = u̯ .

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_011

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

177

Beispiele: tekussiezzi OS, tekkussie[zzi] MH/MS ~ tekkussijazzi OH/NS aniēzzi OS ~ anijazzi NS pessiēzzi OS ~ pessijazzi MH/MS wassiezzi MH/MS+OH/NS ~ wassijazzi OH/NS arsiēzzi OS ~ arsijazzi OH/NS tarkummiēzzi NS kardimijaz-zi, Med. kardimijattari MH/MS

tekkussijaizzi MH/MS anijaēzzi NS pessijaizzi OH/NS wassijaizzi NS ārsijaizzi OH/NS tarkummijaizzi NS kartimmijaizzi, Med. kardimijaittat MH/MS

Zwar haben einige dieser Verben wirklich luwische Entsprechungen, doch die betreffenden Verben gehören im Luwischen teilweise gerade zu den gewöhnlichen -i/ja-Stämmen und zeigen keine Formen auf -ai-, vgl. 3. Sg. Prät. kluw. tar-kum-mi-i-[ta]; hluw. LEPUS+ra/i-ta /t(a)barrita/, LEPUS+ra/i-ia-ta /t(a)barrijata/. Letztere erscheinen zudem gerade nicht in der 3. Singular, in der luwisch nur -it(t)i, -it(t)a oder -at(t)i, -at(t)a bezeugt ist. Melchert (2005: 454f.) hat sich denn auch gegen Oettingers Erklärung gewandt und rechnet statt mit Entlehnung der ganzen Flexion aus dem Luwischen nur mit einem direkten Einfluss luwischer Formen der 3. Sg. auf -jai, die dann im Hethitischen zu -jai-zzi umgebildet wurden; von dort aus habe sich die Stammvariante -jai- sporadisch auch auf andere Formen ausbreiten können, die im Hethitischen sonst -ae-/ai- haben können, vgl. z. B. einige Belege der 2. Sg. Imperativ mit -ia-i: kluw. hanhanijai, paddalijai; hluw. taparrijai* (ungesichert). Es gibt jedoch kaum direkte Entsprechungen von luw. -jai = heth. -ia-I(z-zi); gefunden habe ich nur heth. ta-par-ri-ia-Iz-zi ≃ hluw. *taparijai (?); die Klasse ist im Luwischen offenbar auch nicht sehr häufig. Von Bedeutung ist auch der Beleg ạ-ni-ia-e-ez-zi (KUB 41.15 obv. 13, NS), der zeigt, dass man nicht ohne weiteres /-aittsi/ lesen kann, wie Melcherts Erklärung das voraussetzt. Wichtiger noch ist, dass sich auch viele Fälle von unerwartetem ‹ia-I°›, ‹ia-e°› finden, für die sich kein Vorbild mit -ai- finden lässt, z. B. die folgenden Formen: 3. Pl. Prät. pí-ia-er ‘gaben’ (KUB 31.68:43, NH) neben älterem pí-i-(e-)er OS, pí-e-er MH/MS zum athematischen hi-Verb pai-/pi- ‘geben’; keine 3. Sg. *pí-ia-iz-zi als Ausgangspunkt Vgl. aber 1. Pl. pijawen und luw. /pijai/: hluw. pi-ia-i

178

Kümmel

da-ia-er ‘stahlen’ MH/MS neben da-i-e-er MH/MS (erst spät 3. Sg. Präs. [ta-a-]ia-Iz-zi) i-ia-er ‘machten’ (KUB 34.90:7, NS) neben älterem i-e-er OS+ (← 3. Sg. Präs. i-ia-Iz-zi NS?) ti-i-ia-er ‘traten’ (KBo 16.35:6, NH/spätjh.) neben ti-(i-)e-er, ti-i-er NH (← 3. Sg. Präs. ti-ia-Iz-zi NH?) a-ni-ia-er OH/NS neben a-ni-i-er MH/MS (← 3. Sg. Präs. a-ni-ia-e-ez-zi NS?). 3. Pl. Präs. i-ia-en-zi (KBo 15.22+ i9, NS) ‘sie machen’ (kaum nach vereinzeltem aheth. aruwaenzi) Älter 3. Pl. i-(e-)en-zi (OH/MS) gut belegt neben i-ia-an-zi OS, ia-an-zi OH/MS, i-an-zi MH/MS. Im Folgenden wird die Distribution von Schreibungen mit -ia-I° genauer untersucht, zu den Einzelheiten der Belegdaten siehe die Tabellen im Anhang. 2

Untersuchung der Distribution

2.1 Chronologische Distribution Eine Schreibung mit -ia-I° taucht nur einmal auf einem althethitischen Original auf, dann etwas öfter in mittelhethitischen Originalen, sie wird aber erst auf junghethitischen Tafeln häufiger. 2.2 Intraparadigmatische Distribution Die betreffenden Formen sind in der älteren Sprache meist mit -i-(e-)I°-, seltener mit -ia-a°- geschrieben, auch später fast nie mit ‹-ia-a-I°› (nur ši-i-ia-a-Iz-zi, ši-ia-a-It NS),2 und auch andere Formen dieser Verben geben (wie schon bei Oettinger bemerkt) keinen eindeutigen Hinweis auf ein Paradigma des āe-/āTyps. Bei Verben ohne ältere Belege sind zwar manchmal neben der S­ chreibung mit -ia-I° keine anderen Schreibungen der gleichen Formen belegt, z. B. bei tuhusija(i)- ‘abwarten’, hantija(i)- ‘unterstützen (?)’, huntarija(i)- ‘furzen’, daraus kann aber mangels älterer Belege nichts geschlossen werden. Schreibungen mit -ia-I° erscheinen demnach ausschließlich in Formen, die in der älteren

2 Plene wird nach -ia- aber auch sonst alt vermieden, da die Ligatur I.A -ia- wie -i-a- gebraucht wurde.

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

179

Sprache vorwiegend mit -i-(e-)I° geschrieben sind, niemals in Formen, die schon früher überwiegend nur -ia-a° zeigen.3 2.3 Interparadigmatische Distribution Unabhängig von der Morphologie sind solche Schreibungen bei beinahe allen häufigeren Verbformen mit altem -ie- belegt. Bei Nominalformen scheint es keine Schreibungen mit -ia-I° zu geben, wobei hier aber nur Abstrakta auf °iessar und der Nom. Pl. der i- und ja-Stämme auf °i-e-eš einschlägig wären. Für die letzteren ist ab und zu °i-ia-aš belegt. 3 Folgerungen Die Schreibung mit -i-(e-)I° selbst gehört dagegen überwiegend der älteren Sprache an und beginnt in dem Zeitraum langsam zu verschwinden, in dem die Schreibungen mit -ia-I° auftauchen. Es liegt also der Verdacht nahe, dass -ia-I° die ältere Schreibung -i-(e-)I° ersetzt. Das bedeutet, dass genau die Formen, die eine ältere Lautung -je- aufwiesen, dem Wandel unterliegen. Da eine Diphthongierung e > ai nur hinter j wohl eher unwahrscheinlich ist, sollte der Wandel eher graphisch als lautlich begründet sein. Könnte es sich also bei -iaI° einfach um eine jüngere Schreibweise für /je/ handeln, das Zeichen -ia- also auch als -je- zu lesen sein? Eine Lesung -je-/-ji- für das Zeichen I.A ist im Akkadischen belegt (altbabylonisch und mittel/neuassyrisch, selten; s. von Soden and Röllig 1967: 19) und auch schon früher für das Hethitische erwogen worden (vgl. Kronasser 1966: 74f.; Zinko 1981), wurde jedoch meist nicht akzeptiert (vgl. Melchert 1994: 24f.; Kimball 1999: 173–175 mit Lit.), vor allem aufgrund der in älterer Sprache konsequenten Schreibung von -aC-Zeichen hinter -ia-. 3.1 Schwanken -ia- ~ -i-e- beim Verbum Nun werden aber auch die Schreibungen (°i)-ia-C°, -Ci-a-, °i-aC-, (°i)-ia-aCschon alt- und mittelhethitisch häufiger an Stellen verwendet, wo man ­lautliches /je/ erwarten würde, und zwar im Wechsel mit (C)i-(e-)-C° bzw. (C) i-(e-)IC-, vgl. 3. Sg. Präsens i-(e-)Iz-zi OS ~ ia-az-zi MS ‘macht’ (KUB 36.108 obv. 12). Auch der umgekehrte Wechsel ist belegt, z. B. 1. Sg. Präsens i-ia-mi OS ‘mache’ ~ i-e-mi OS (StBoT 8 ii 22); pé-eš-ši-ia-mi OS ~ pé-eš-ši-e-mi OS ‘werfe weg’; 3 Vgl. Hoffner and Melchert 1998: 202 „The e-grade forms with /-ye-/ (-Ci-(i-)e-) are more frequent in the older language and gradually disappear from use during NH.“

180

Kümmel

a-ni-e-mi OS ~ a-ni-ia-mi NS ‘führe durch’; ḫu-la-a-li-ia-mi OS ~ ḫu-la-a-li-e-mi OS ‘umwickle’; 3. Pl. Präsens ti-(ia-)an-zi OS ~ ti-(i-)en-zi OS ‘stellen sich hin’. Häufiger ist einer der beiden Schreibungen etwas früher belegt, aber die andere ebenfalls schon mittelhethitisch, z. B. 2. Sg. OS i-e-ši ~ MS i-ia-ši ‘machst’; 3. Sg. OS ú-e-mi-Iz-zi ~ MS ú-e-mi-e-Iz-zi ~ ú-emi-ia-az-zi ‘findet’; 3. Pl. OS (i-)iaan-zi ~ MS i-(ia-)an-zi ~ i-(e-)en-zi ‘machen’. Die herkömmliche Erklärung nimmt an, dass hier analogischer Ausgleich zwischen den beiden Varianten des Themavokals stattgefunden hat. Auffällig wäre dann freilich, dass der analogische Ausgleich im gleichen Text und der gleichen Periode in beide Richtungen abgelaufen sein müsste und zudem weitgehend auf die je-Verben beschränkt blieb. Bei den ske-Verben ist ein solches synchrones Schwanken nämlich sehr selten und taucht nur bei der 1. 2. Plural und im Medium auf. Auch bei ae- und ue-Verben findet es sich nur vereinzelt: niemals 3. Sg. *-a-az-zi neben -a-(e-)ez-zi; nur einmal 3. Sg. ú-wa-az-zi NS ‘kommt’ neben ú-(e-)ez-zi OS+, und in der 3. Plural selten -en-zi: a-ru-wa-(a-) an-zi OS ~ a-ru-wa-en-zi OS ‘verbeugen sich’; ú-wa-an-zi, ú-an-zi OS ~ ú-en-zi OS ‘kommen’; kap-pu-u-en-zi OS neben späterem kap-pu-wa-an-zi NS, kap-puu-an-zi OH/NS ‘zählen’. Das wirft die folgenden Fragen auf: Ist es wirklich wahrscheinlich, dass im Althethitischen sowohl /°jettsi/ als auch /°jattsi/ mögliche Ausgänge der 3. Sg. Präsens waren, und dass es für die je-Klasse jahrhundertelang unklar war, welchen Vokal die Ausgänge eigentlich jeweils hatten? Wäre es nicht vorteilhafter, wenigstens mit einer sehr großen phonetischen Ähnlichkeit rechnen zu können, die ungenaue Schreibungen erlaubte? 3.2 Schwanken -ia- ~ -i-e- sonst Unterstützend kommt hinzu, dass auch außerhalb der thematischen Verben in einigen Fällen für ursprüngliches /ja/ Schreibungen mit -i-(ia-)e- erscheinen, die nicht analogisch erklärbar sind, wie z. B. vgl. 3. Pl. pí-ia-er (s. oben); 3. Sg. Med. [z]é-i-e-ri (KBo 18.201 rev. 8, NS) für älteres zé-(e-)a-ri, zé-(e-)ia-ri MH/MS, in dessen Paradigma keine Formen mit altem -je- vorkamen; ta-i-Iz-zi-la- ‘Diebstahl’ MH/NS neben ta-ia-az-zi-il, ta-ia-zi-la- OS, da-ia-zi-la- OH/MS; iš-ḫi-iani-uš ‘Haare’ neben iš-ḫi-e-ni-uš (beide OH/NS) von ishi(a)- ‘binden’. U ­ mgekehrt steht auch -i-ia- für etymologisches -i-e-, z. B. im Nom. Pl. der i-Stämme °i-ia-aš für (meist älteres) °i-e-eš;4 3. Pl. Prät. ú-e-mi-ia-ar, ša-pa-ši-ia-ar, ḫa-a-ni-ia-r° (vgl. Neu 1989; Melchert 1995); šu-up-pí-ia-aš-n° ‘Reinheit’ n ­ eben šu-up-pí-eš-n° < *suppi-esn-. 4 Vgl. Melchert 1995; Oettinger 2003; Yakubovich 2009: 355.

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

181

Das kann wohl nur so verstanden werden, dass der Vokal in der Folge /ja/ wenigstens phonetisch einem /e/ so ähnlich war, dass umgekehrte Schreibungen möglich waren. Ein echter Zusammenfall mit /je/ muss dafür nicht vorliegen, es könnte genügen, wenn z.B. /ja/ = [jæ~jɛ] gegenüber /je/ = [jɛ~je] und /e/ = [ɛ] vorlag, so dass die Realisierung von /a/ nach /j/ mit einer (häufigen) Realisierung von /e/ identisch sein konnte. Ein vollständiger Zusammenfall wie im Litauischen (mit ‹Cia› = ‹Ce› /Cʲa/, /Cʲā/ = [Cʲɛ], [Cʲæ:]) oder im frühen Slavischen (mit *Cja > *Cje) fand also wohl nicht statt.5 Melchert (1984: 161; vgl. Kimball 1999: 174f.) argumentiert allerdings, dass eine analogische Erklärung für -ia- anstelle von -i-e- deswegen wahrscheinlicher sei, weil diese Schreibungen sich graduell über die Paradigmen ausbreiten und die jeweilige Häufigkeit von -a- und -e- also morphologisch bedingt sei. Dies spräche gegen einen Lautwandel mit völligem Zusammenfall der Lautungen. 3.3 Zum Wechsel -e- ~ -a- in der thematischen Flexion In der Tat zeigen die Formen der 3. Singular viel häufiger -e- als alle übrigen Formen, und dies zeigt, dass es wenigstens im älteren Hethitischen eine lautliche Differenz gegeben haben muss, die derjenigen bei anderen thematischen Verbklassen entsprach. Auch diese zeigen den Vokal -e- vor allem in der 3. Sg., während sonst -a- stark dominiert (mit Ausnahme der 2. Sg. Imperativ Aktiv auf -i). Nur die ske-Verben verhalten sich anders, insofern als althethitisch der Vokal -a- überhaupt nur vor nz, nt und hh erscheint, also dort, wo altes *e lautgesetzlich als hethitisch a erscheint. Es sieht also so aus, als sei als verbaler Themavokal im Hethitischen und Anatolischen generell nur *e verwendet worden, das heth. betont meist é ergab, unbetont meist *e > ‹a› [ǝ?], außer in der 2. Sg. Ipv. -i und in der 3. Sg. -ezz(i), -et.6 Dieser Zustand dürfte auch bei den je-Verben gegolten haben, wahrscheinlich aber mit viel größerer phonetischer Ähnlichkeit der Varianten: /je/ [jɛ~je] ~ /ja/ [jæ~jɛ / jə?]. Hinzu kam wohl Einfluss des Luwischen, in dem es keine phonologische Differenz von /e/ vs. /a/ gab (vgl. Yakubovich 2009: 355ff.). Die Frage ist nun, wie man diese Differenzen schreiben konnte oder ob man sie schreiben musste. Für die Schreiber gab es zwei Möglichkeiten, die Lautung -je- wiederzugeben: Eine phonetisch möglichst eindeutige Schreibung mit 5 Das spricht auch gegen die Annahme Oettingers (1984, 1985), es habe einen besonderen Umlaut ja > je vor C + i/e gegeben. 6 Vgl. Yoshida 2009, 2010, 2014; Kümmel 2012: 206f. Nicht ganz klar ist wegen der wenigen Belege der Vokalismus der 2. Sg. Präteritum. Eigentlich wäre aus unbetontem *-es lautgesetzlich -is zu erwarten, belegt ist aber außer -is in °a-(a-)iš (NS) auch -es in °a-a-eš (OH/MS+) und -as in ú-wa-aš ‘du kamst’ (NS) und °a-a-es.

182

Kümmel

°i-(i-)(e-)-IC oder eine morphonologische mit dem auch in anderen Formen verwendeten (°i-)ia-(aC)-. Die letztere Schreibweise war freilich missverständlich; sie wurde erst mittelhethitisch häufiger, konnte sich aber anscheinend nicht als Norm durchsetzen. In mittelhethitischer Zeit konnte aus solchen Schreibungen eine Regel abstrahiert werden, nach der das Graphem I.A (Rüster and Neu 1989: 218) auch im phonologischen Sinne als /je/ lesbar war, und diese Lesung konnte man durch ein folgendes IC-Zeichen markieren, so dass nun -ia-IC als -je-eC zu verstehen war. Damit hatte man eine Möglichkeit, die wahrscheinlich immer noch vorhandene lautliche Differenz von -ja- und -jezu schreiben, ohne dass eine stark abweichende Graphie nötig war. 4

Alternative: Umbildungen nach der produktiven hatrae-Klasse?

Könnte es sich aber nicht doch um Umbildungen der produktiven hatrae-Klasse handeln? Zwar ist in der Tat diese Klasse in jüngerer Zeit sehr expansiv, doch eigentlich erst junghethitisch, nicht schon mittelhethitisch. In Originalen der älteren Sprache erscheinen Formen, die man zu dieser Flexionsklasse rechnen könnte, neben tatsächlich alten Verben dieses Typs und den hier besprochenen Formen mit -ia-I° nur bei zwei weiteren Verbalklassen: 1. Bei ursprünglich athematischen hi-Verben auf °ai- (zuerst 3. Sg. Präteritum °a-a-it) und bei lā- ‘lassen’, wo die 3. Sg. la-a-Iz-zi schon mittelhethitisch geläufig ist, ebenso beim Kompositum dālaizzi (MH/MS). Die ­Flexion des Kompositums setzt einen Stamm lai- voraus, so dass lāizzi zu den Verben auf °ai- zu stellen sein könnte. 2. Bei thematischen Verben auf -ue-: Schon mittelhethitisch findet sich die 3. Sg. Präteritum ša-ru-wa-It (MH/MS) zu dem aufgrund der 3. Pl. ša-ruu-e-er anzusetzenden Stamm sarue-, vgl. auch kap-pu-u-wa-iz-zi, k­ ap-pu-wa-it OH/NS vs. kap-pu-u-Iz-zi MH/MS, kap-pu-u-(e-)et OS. Bemerkenswert ist aber auch hier wie bei den Formen mit -ia-I°, dass die sonst alt- und besonders mittelhethitisch beinahe ausnahmslose Pleneschreibung von a fehlt, und bei dieser Klasse schwankt der Themavokal fast genauso stark wie bei den ie-Verben, vgl. kap-pu-u-e-ši OH/NS, kappu-u-en-zi OS ~ kap-pu-u-wa-ši, kap-pu-u-an-zi OH/NS. Daher könnte der Verdacht aufkommen, dass -wa-I° vielleicht auch als °we-e° zu lesen sein könnte.7 Immerhin sind hier jedoch — anders als bei -ia-I° — graphisch eindeutige Formen der ae-Klasse wenigstens junghethitisch klar belegt, 7 Vgl. Kronasser 1966: 71, wogegen Melchert 1994: 24f. Das Zeichen PI -wa- wird auch akk. früh als -we-/-wi- gebraucht, s. von Soden – Röllig 1967: 43.

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

183

vgl. ša-ru-wa-a-It NH, kap-pu-u-wa-a-an-zi, kap-pu-u-wa-a-it NS, šu-ú-waa-iz-zi OH/NS. Da beide Suffixe vor allem Denominative bilden, wäre eine direkte Umbildung auch funktional leichter verständlich. Erst junghethitisch sind dann auch in anderen Klassen sekundäre Übergänge belegbar, z. B. bei mehrsilbigen hi-Verben auf °a-, die ihrerseits teilweise aus älteren einsilbigen Konsonantenstämmen entstanden waren, vgl. z. B. OS lāhui ‘gießt’ → OH/NS lā̆huwā̆i → NH lāhuwāizzi. Die Produktivität der ae-Klasse war also zunächst durchaus begrenzt, und für thematische Stämme (außer vielleicht den ue-Verben) ist sie eigentlich nicht nachweisbar. Eine rein morphologische Erklärung für die Schreibungen mit -ia-IC kann also nicht weiter gestützt werden. 5 Ergebnis Es bleibt also festzuhalten: Hinweise auf große phonetische Ähnlichkeit von / ja/ und /je/ finden sich schon seit dem Althethitischen. Daraus kann sich dann eine Verwendung des Zeichens -ia- auch als Graphem für /je/ entwickelt haben, was dann Schreibungen mit -ia-IC- = [-jeC-] für älteres -i-IC-, -i-e-IC- ermöglichte, in Konkurrenz mit -ia-aC- mit konservativer Graphemkombination (⇒ aC- hier ≃ eC-). Das Phänomen tritt fast nur beim Verbum auf, und der Hauptgrund dafür dürfte eine Tendenz zu paradigmatischer (morphonologischer) Schreibung sein, da der Wechsel /je/ ~ /ja/ nur dort vorkam. Die Sonderschreibung war also anderswo nicht motiviert und unterblieb daher. Möglicherweise zu trennen sind luwische Lehnwörter wie taparrija(i)-, für die echte ja-Flexion unsicher ist (Gruppe 5 im Anhang). Abkürzungen akk. akkadisch heth. hethitisch hluw. hieroglyphenluwisch Ipv. Imperativ kluw. keilschriftluwisch luw. luwisch Präs. Präsens Prät. Präteritum spjh. spätjunghethitisch

184

Kümmel

Bibliographie Hoffner, H.A., Jr. and Melchert, H.C. (2008) A Grammar of the Hittite Language, Part 1: Reference Grammar, Part 2: Tutorial, Winona Lake, IN: Eisenbrauns. Kimball, S.E. (1999) Hittite Historical Phonology [IBS 95], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Kronasser, H. (1956) Vergleichende Laut- und Formenlehre des Hethitischen, Heidelberg: Winter. Kronasser, H. (1966) Etymologie der hethitischen Sprache, Band 1, i. Zur Schreibung und Lautung des Hethitischen, ii. Wortbildung des Hethitischen, Wiesbaden: Harrassowitz. Kümmel, M.J. (2012) The inflexion of the Hittite verb class of mema/i- and related questions, in: H.C. Melchert (ed.), The Indo-European Verb. Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles 13–15 September 2010, Wiesbaden: Reichert, 201–208. Melchert, H.C. (1984) Studies in Hittite Historical Phonology [ZVS Ergänzungshefte 32], Göttingen: Vandenhoeck & Ruprecht. Melchert, H.C. (1994) Anatolian Historical Phonology [LSIE 3], Amsterdam, Atlanta: Rodopi. Melchert, H.C. (1995) Neo-Hittite nominal inflection, in: O. Carruba, M. Giorgieri and C. Mora (eds.), Atti del ii Congresso Internazionale di Ittitologia, Pavia: Iuculano, 269–74. Melchert, H.C. (2005) The problem of Luvian influence on Hittite, in: G. Meiser and O. Hackstein (eds.), Sprachkontakt und Sprachwandel. Akten der xi. Fachtagung der Indogermanischen Gesellschaft, 17.–23. September 2000, Halle an der Saale, Wiesbaden: Reichert, 445–459. Neu, E. (1989) Zu einer hethitischen Präteritalendung -ar, HS 102, 16–20. Oettinger, N. (1979) Die Stammbildung des hethitischen Verbums, Nürnberg: Carl. Oettinger, N. (1984) Sekundärwirkungen des Umlauts beim hethitischen Nomen, ZVS 97, 44–57. Oettinger, N. (1985) Thematische Verbalklassen des Hethitischen: Umlaut und Ablaut beim Themavokal, in: B. Schlerath (ed.), Grammatische Kategorien. Akten der vii. Fachtagung der Indogermanischen Gesellschaft, Berlin, 20.–25. Februar 1983, Wiesbaden: Reichert, 296–312. Oettinger, N. (1992) Die hethitischen Verbalstämme, in: O. Carruba (ed.), Per una grammatica ittita (Towards a Hittite Grammar), Pavia: Iuculano, 213–252. Oettinger, N. (2003) Zum Ablaut von n-Stämmen im Anatolischen und der Brechung ē > ya, in: E. Tichy, D.S. Wodtko and B. Irslinger (eds.), Indogermanisches Nomen. Derivation, Flexion, und Ablaut, Bremen: Hempen, 141–52. Rüster, C. and Neu, E. (1989) Hethitisches Zeichenlexikon. Inventar und Interpretation der Keilschriftzeichen aus den Boğazköy-Texten, Wiesbaden: Harrassowitz.

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

185

von Soden, W. and Röllig, W. (1967) Das akkadische Syllabar, Roma: Pontificum Institutum Biblicum (2., völlig neubearbeitete Auflage). Yakubovich, I. (2009) Sociolinguistics of the Luvian Language [BSIELL 2], Leiden, Boston: Brill. Yoshida, K. (2009) On the origin of thematic vowels in Indo-European verbs, in: K. Yoshida and B. Vine (eds.), East and West: Papers in Indo-European Studies, Bremen: Hempen, 265–280. Yoshida, K. (2010) Observations on the prehistory of Hittite i̯e/a-verbs, in: R. Kim, N. Oettinger, E. Rieken and M. Weiss (eds.), Ex Anatolia Lux. Anatolian and Indo-­ European Studies in honor of H. Craig Melchert on the Occasion of his Sixty-Fifth Birthday, Ann Arbor, New York: Beech Stave, 385–393. Yoshida, K. (2014) The thematic vowel *e/o in Hittite verbs, in: H.C. Melchert, E. Rieken and T. Steer (eds.), Munus amicitiae. Norbert Oettinger a collegis et amicis dicatum, Ann Arbor: Beech Stave, 373–384. Zinko, C. (1981) Die Vertretung der grundsprachlichen Diphthonge im Hethitischen, Dissertation Graz.

Form

iš-ḫu-uz-zi-ia-Iz-i [iš-ḫu-u]z-zi-ia-i-e-zi iš-ḫu-zi-ia-It

ur-ki-ia-Iz-zi

-ia-IC-

iš-ḫu-u[z-zi-Id-du] iš-ḫu-zi-Id-du –

u-ur-kI-Iz-[zi]

-i-e-IC-

3. Sg. Prät. M. kar-di-mi-ia-It-ta-at –

kardimmia- ‘wütend 3. Sg. Präs. kar-tim-mi-ia-Iz-zi sein’ 3. Sg. Präs. M. ka[r-di-mi-]ia-It-ta

3. Sg. Prät. 3. Sg. Ipv.

1: OS urkia- ‘aufspüren’ 3. Sg. Präs. 2: MS ishuzzia- ‘umgürten’ 3. Sg. Präs.

Verbstamm

kar-di-mi-ia-at-ta-ri kar-tim-mi-ia-at-ta[-ri] kar-tim-mi-ia-ta-ri kar-tim-mi-ia-at-t[a] kar-tim-mi-at-ta-at

kar-di-mi-ia-az-zi

–

-ia-aC-

Schematische Markierung der Bezeugung: OS OH/MS OH/NS MH/MS=OH/NS MH/MS MH/NS NH

Anhang: Formenübersicht

–

–

–

-ia-a-(aC-) luw.

NS

186 Kümmel

wia- ‘schreien’ (← wai-) 3: OH-MH/NS ania- ‘durchführen’

tekkussia- ‘zeigen’

tarkummia‘übersetzen’

tāja- ‘stehlen’

ta-(a-)i-Iz-zi da-a-i-Iz-zi da-i-e-ez-zi, ta-a-Iz-zi [d]a-a-i-e-ez-zi, ta-a-i-e-ez-zi da-ia-at da-a-i-ia-at da-ia-er da-i-e-er tar-kum-mi-ia-Iz-zi tar-kum-mi-e-ez-zi

a-ni-Iz-zi, a-ni-e-ez-zi a-ni-i-e-et a-ni-(i-)er, a-ni-i-e-e[r]

2. Sg. Präs. 3. Sg. Präs. 3. Sg. Prät. 3. Pl. Prät.

a-ni-ia-er

ạ-ni-ia-e-ez-zi

a-ni-e-mi

tar-ku-mi-ia-i te-ek-ku-uš-ši-ia-Iz- te-ku-uš-ši-Iz-zi zi te-ek-ku-uš-ši-I[z-zi] te-ek-ku-uš-ši-i-Iz-zi te-ek-ku-uš-ši-e-et ú-i-ia-Iz-z[i] ú-i-e-ez-zi ú-i-ia-Iz-zi

[ta-a-]ia-Iz-zi

1. Sg. Präs.

3. Sg. Prät. 3. Sg. Präs.

2. Sg. Ipv. 3. Sg. Präs.

3. Pl. Prät. 3. Sg. Präs.

3. Sg. Prät.

3. Sg. Präs.

a-ni-a-mi a-ni-ia-mi a-ni-ia-ši a-ni-ia-(az-)zi, an-ni-ia-az-zi a-ni-ia-at, a-ni-at

ú-i-ia-an-[zi]

te-ek-ku-uš-ši-ia-az-zi

da-a-ia-az-zi ta-ia-az-zi da-i-ia-zi da-a-i-ia-zi, ta-a-i-ia-zi

a-an-ni-i-

–

–

tar-kum-mi-i-

–

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

187

1. Sg. Präs. 3. Sg. Präs.

harija- ‘vergraben’

ija- ‘machen’

3. Sg. Präs.

arsia- ‘fließen’ (← ars-)

i-ia-er

3. Pl. Prät.

3. Sg. Ipv.

i-ia-en-zi

i-ia-Iz-zi

i-e-er i-e-er i-e-ed-du

i-e-ši i-e-zi, i-Iz-zi, i-e-ez-zi i-e-ez-zi i-e-eš i-e-et i-e-et, e-et i-e-et? i-en-zi, i-e-en-zi i-en-zi, i-e-en-zi

ar-ši-e-ez-zi ar-ši-i-e-ez-zi, ar-ši-Iz-zi a-ar-aš-ši-Iz-zi ḫa-ri-e-mi ḫa-ri-Iz-zi, [ḫa-ri-]e-ez-zi, ḫa-a-ri-e-ez-zi ḫa-ri-e-nu-un i-e-mi

a-ar-ši-ia-Iz-zi

ḫa-ri-ia-Iz-zi ?

-i-e-IC-

-ia-IC-

3. Pl. Präs.

2. Sg. Prät. 3. Sg. Prät.

2. Sg. Präs. 3. Sg. Präs.

1. Sg. Prät. 1. Sg. Präs.

Form

Verbstamm

i-ia-ad-du, i-ad-du i-ia-ad-du

i-ia-an-zi, ia-an-zi, i-an-zi i-an-zi, i-ia-an-zi i-ia-an-zi

i-ia-mi i-ia-am-mi, i-ia-mi i-ia-ši, i-ia-ši ia-az-zi, i-ia-az-zi i-ia-zi, i-ia-zi, i-ia-az-zi i-ia-aš i-ia-at i-ia-at

ḫa-a-ri-a-mi, ḫa-a-ri-ia-mi ḫa-ri-ia-zi, ḫa-ri-ia-az-zi, ḫa-ari-ia-zi

ar-ši-ia-(az-)zi, [a]r-ši-ia-zi a-ar-aš-ši-ia-zi

-ia-aC-

(a-ia-)

–

a-ar-ši-ia-

-ia-a-(aC-) luw.

188 Kümmel

sarria- ‘trennen’ (← sarr-)

sallia- ‘schmelzen’

iskia- ‘salben’

pessia- ‘wegwerfen, aufgeben’

3. Sg. Prät. 3. Sg. Präs. M.

3. Sg. Prät. 3. Sg. Ipv. 3. Sg. Präs. M. šal-li-ia-It-ta 3. Sg. Ipv. M. 3. Sg. Präs. šar-ri-ia-Iz-zi

iš-ki-ia-Iz-zi

(iš-kI-mi) iš-kI-Iz-zi iš-kI-Iz-zi iš-kI-It iš-kI-Id-du [šal-l]i-i-e-et-ta šal-li-(e-)et-ta-ru šar-rI-Iz-zi šar-rI-e-ez-zi šar-ri-i-e-et šar-rỊ-Ịt[-ta]

pé-(e-)eš-ši-(i-e-)et peš-še-et, pé-še-et

pé-iš-ši-ia-It

3. Sg. Prät.

1. Sg. Präs. 3. Sg. Präs.

pé-eš-ši-i-(e-)ez-zi pé-iš-ši-(i-)e-ez-zi pé-eš-ši-e-zi, pé-eš-še-ez-zi pé-eš-ši-zi, peš-ši-Iz-zi

pé-eš-ši-ia-Iz-zi pé-iš-ši-ia-Iz-zi pé-eš-ši-ia-i (!)

pé-eš-ši-e-mi

3. Sg. Präs.

1. Sg. Präs.

šar-ri-ia-zi

pé-eš-ši-ia-(az-)zi pé-iš-ši-ia-az-zi pé-iš-ši-az-zi, pé-eš-še-ia-az-zi pé-eš-še-ia-zi, pé-ši-ia-az-zi pé-ši-az-zi, peš-ši-ia-zi pé-iš-ši-ia-at pé-eš-ši-at pé-eš-ši-ia-at, pé-iš-še!-ia-at pé-iš-ši-at, peš-ši-ia-at (iš-ga-a-mi) iš-ki-ia-zi

pé-eš-ši-ia-mi pé-eš-še-ia-mi, pé-iš-ši-ia-mi

–

–

?

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

189

wemija- ‘finden’

duwarnia‘zerbrechen’ (← duwarne-)

tija- ‘treten’

3. Sg. Prät. 3. Pl. Prät.

3. Sg. Präs.

3. Pl. Prät. 3. Sg. Präs.

3. Sg. Prät.

3. Pl. Präs.

ti-Iz-zi, ti-e-ez-zi ti-i-Iz-zi, ti-i-e-ez-zi ti-i-in-zi, ti-i-en-zi ti-en-zi ti-e-en-zi ti-i-e-et ti-e-et ti-i-ia-er ti-e-er, ti-i-er, ti-i-e-er du-wa-ar-ni-ia-Iz-zi du-wa-ar-nI-Iz-zi tu-wa-ar-nI-Iz-zi du-wa-ar-nI-zi du-wa-ar-nI-e-ez-zi ú-e-mi-ia-Iz-zi ú-e-mi-(ez-)zi ú-e-mi-e-ez-zi ú-e-mi-It ú-e-mi-er, ú-e-mi-(i-)e-er

ti-ia-Iz-zi

ši-(i-)ia-It ši-ia-er

3. Sg. Prät. 3. Pl. Prät.

3. Sg. Präs.

ši-i-e-ez-zi ši-e-ez-zi, ši-i-e-z[i] – –

ši-(i-)ia-Iz-zi

3. Sg. Präs.

sija- ‘schießen’

-i-e-IC-

-ia-IC-

Form

Verbstamm

ú-e-mi-ia-at ụ́-ẹ-mỊ-iạ-ạr

ú-e-mi-ia-(az-)zi

– du-wa-ar-ni-ia-az-zi, du-wa-ar-ni-ia-zi

ti-ia-at, ti-i-ia-at

ti-an-zi, ti-i-ia-an-zi

ti-ia-zi, ti-i-ia-az-zi

– –

ši-ia-az-zi

-ia-aC-

h. wa/i-mi-

?la-wa-arri-i°

ši-i-ia-a-iz- – zi ši-ia-a-It ?ši-ia-a-ue-ni ši-ia-(a-) nu-un –

-ia-a-(aC-) luw.

190 Kümmel

3. Sg. Präs. M. 3. Sg. Prät. M.

hulalia- ‘umwickeln’ 3. Sg. Präs.

ḫu-la-a-li-ia-Iz-zi

ḫu-la-li-It-ta-at

ḫu-la-a-LI-e-z[i], ḫu-la-a-li-Iz-zi ḫu-u-la-a-li-Iz-zi, ḫu-u-la-li-(e-)ez-zi

ḫu-It-ti-e-ez-zi

ḫar-ki-It-ta-ru ḫar-ni-e-ez-zi

a-ri-ia-at a-ri-(i-)e-er, a-ri-er [ḫar-]ki-e-ed-du

3. Sg. Prät. 3. Pl. Prät. 3. Sg. Ipv.

a-ri-ia-er

a-ri-e-ez-zi

3. Sg. Präs.

3. Sg. Ipv. M. ḫar-ki-ia-It-ta-ru harnia- ‘besprengen’ 3. Sg. Präs. ḫar-ni-ia-Iz-zi (← harn-) 3. Sg. Ipv. huittija- ‘ziehen’ 3. Sg. Präs. ḫu-It-ti-ia-i (!)

harkia‘verschwinden’

4: NH, NS aria- ‘Orakel befragen’

ḫu-la-li-ia-at-ta-ri

ḫar-ni-ia-ad-du ḫu-It-ti-ia-zi ḫu-u-It-ti-ia-zi ḫu-(u-)It-ti-ia-az-zi, ḫu-It-ia-az-zi ḫu-la-a-li-az-zi, ḫu-la-li-ia-azzi, [ḫu-]ul-la-li-ia-az-zi

ḫar-ni-ia-zi

a-ri-ia-zi, a-ri-ia-az-zi

–

–

–

–

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

191

tastasia- ‘tuscheln’

taistia- ‘laden’ (← taistai-)

pi- ‘geben’

3. Sg. Prät.

3. Pl. Prät. 3. Sg. Präs.

3. Sg. Prät.

1. Pl. Prät. 3. Pl. Prät.

3. Sg. Prät. 1. Pl. Präs. 3. Pl. Präs.

ta-(a-)aš-ta-ši-iaIz-zi ta-aš-ta-ši-ia-It

da-iš-ti-ia-Iz-zi

pí-ia-er

da-iš-te-i-e-er

pí-i-er, pí-i-e-er pí-er, pí-(i-)(e-)er –

pí-en-zi

pé-i-e-ši pé-i-e-ez-zi, pé-e-ez-zi pé-i-e-zi pé-i-e-et

pé-e-i-ia-I[z-zi]

2. Sg. Präs. 3. Sg. Präs.

peja- ‘senden’

pár-ši-i-e-ez-zi, pár-ši-zi pár-ši-It-ta-ri pár-ši-It-ta-ru pé-i-e-mi, pé-e-i-mi

3. Sg. Präs. pár-ši-ia-Iz-zi 3. Sg. Präs. M. 3. Sg. Ipv. M. 1. Sg. Präs.

-i-e-IC-

parsia- ‘brechen’

-ia-IC-

Form

Verbstamm

ta-aš-ta-ši-ia-zi

ta-[e]š-ti-ia-zi

pí-(i-)ia-ú-e-ni pí-(ia-)an-zi, pí-an-zi pí-(ia-)an-zi pí-ia-ú-en

–

–

pí-(i-)iah.pi-ia-/pi-i-

–

pár-ši-ia-ad-da-ru pé-i-ia-mi pé-ia-mi pé-e-ia-zi

–

-ia-a-(aC-) luw.

pár-ši-(ia-)az-zi

-ia-aC-

192 Kümmel

wa-aš-ši-ia-Iz-zi

3. Sg. Prät. 3. Sg. Präs.

3. Sg. Präs. 3. Sg. Präs.

ak-kur-ri-ia-i (!) ḫa-an-ti-ia-I[z-zi]

3. Sg. Präs. ú-i-šu-ri-ia-Iz-zi 3. Sg. Präs. M. 3. Sg. Prät. M.

wa-ag-ga-ri-ia-Iz-zi

3. Sg. Präs.

ḫa-an-ti-ia-It 3. Sg. Prät. huntaria- ‘furzen’ 3. Sg. Präs. M. ḫu-un-ta-ri-ia-It-ta lahhia- ‘marschieren’ 3. Sg. Präs. la-aḫ-ḫi-ia-Iz?-zi la-ḫi-ia-Iz-zi 3. Sg. Prät. la-aḫ-ḫi-ia-It sapia- ‘reinigen’ 2. Sg. Ipv. ša-pí-ia-i

5: nur -ia(i)akkurria- ‘?’ hantia‘unterstützen’

wassia- ‘bekleiden’ (← wasse-) wisuria- ‘ersticken’

wakkaria‘revoltieren’

wa-aš-ši-Iz-zi wa-aš-ši-e-ez-zi ú-i-šu-u-ri-Iz-zi

wa-ag-ga-ri-Iz-zi

ú-i-šu-u-ri-ia-at-ta-ri ú-i-šu-u-ri-ia-ta-ti ú-i-šu-(u-)ri-ia-at-ta-ti ú-e-šu-ri-ia-at-ta-ti ú-i-šu-ri-ia-ad-da-at

wa-ak-ri-ia-zi wa-ag-ga-ri-ia-(az-)zi wa-ak-k4a-(a-)ri-ia-zi wa-ag-ga-ri-ia-at wa-aš-ši-ia-(az-)zi

ša-pí-ia-i-

– la-aḫ-ḫi-i°

–

–

–

–

Über die hethitische 3. Sg. Präsens auf -ia-Iz-zi

193

Form

tuḫusia- ‘abwarten’

3. Sg. Präs. 3. Sg. Prät. 3. Pl. Prät. 2. Sg. Ipv. 3. Sg. Prät.

taparria- ‘herrschen’ 2. Sg. Präs.

Verbstamm

ta-pár-ri-ia-Iz-zi ta-pa-ri-ia-It da-pár-ri-ia-e-er ta-pár-ri-ia-i du-ḫu-ši-ia-It, tu-ḫu-ši-ia-I[t] tu-ḫu-uš-ši-ia-It tu-u-ḫu-ši-ia-It

ta-pár-ri-ia-ši

-ia-IC-

-i-e-IC-

-ia-aC-

ta-ḫu-šiia-a(-)

h. taparija-*

-ia-a-(aC-) luw.

194 Kümmel

Chapter 10

The Word for Wine in Anatolian, Greek, Armenian, Italic, Etruscan, Semitic and Its Indo-European Origin Reiner Lipp Abstract It can be shown that Hittite u̯ ii̯an(a)- ‘grapevine; wine’ continues a Proto-Indo-­ European nomen agentis *u̯ éi̯h1-on-/ *u̯ ih1-n-́ ‘twiner, creeping plant, grapevine’, and that this stem formation provides the derivational base for further Indo-European words for ‘wine’ such as Latin vīnum < *u̯ ih1-n-ó- and Greek οἶνος < *u̯ ói̯n-o- < *u̯ ói̯(h1)n-o-, with the corresponding affiliation adjective preserved in Armenian gini ‘wine’ < *gu̯ īníi̯o < *u̯ oi̯n-ih2ó-. The morphological structure of the respective base word and its derivatives corresponds to regular Proto-Indo-European patterns of word formation and is therefore congruous with the assumption of an Indo-European origin of the lexemes in question. Accordingly, it is likely that the etymologically isolated Semitic *u̯ ai̯nu ‘wine’ was borrowed from Mycenaean Greek and that Proto-Kartvelian *γu̯ ini̯o ‘wine’ is a loanword from Proto-Armenian or its prestage.

Keywords wine – Indo-European – Hittite – Greek – Latin – Etruscan – Semitic – etymology – internal derivation – hypostasis



Introduction

In the Neolithic period the wild grapevine was to be found in the northern Mediterranean, on the lower and middle Danube, on the north coast of the Black Sea (Pontus), in the area of the northern and southern Caucasus, on the south coast of the Caspian Sea, in the Zagros region and in Anatolia on its Black Sea and Mediterranean coasts. Since already in the Neolithic period the berries of the wild grapevine were used for fermentation to wine, the ProtoIndo-Europeans in their homeland north of the Black Sea and the Caucasus also probably knew how to produce wine from local wild vines. Therefore it is © koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_012

196

Lipp

legitimate to search for an explanation of the origin of the various designations for ‘wine’, which resemble each other in the individual Indo-European languages, on the basis of the phonology, morphology as well as the lexical and derivational patterns of Proto-Indo-European and the early stages of the ­relevant individual languages. For this purpose, the evidence from the Old ­Indo-European languages is presented in the following and supplemented by material from the adjacent languages. The first evidence for the production of wine (with admixture of resin for better preservation) is provided by a residuum of tartaric acid in a ceramic vessel found at a neolithic site from 5400–5000 bc in the northern Zagros area (Hajji Firuz Tepe, Iran). See McGovern Glusker, Exner and Voigt 1996, McGovern 2003: 64–74

On the various forms of the word for ‘wine’ and ‘grapevine’ in the IndoEuropean area and neighbouring regions, the prehistoric distribution of the wild grapevine (also in the North Pontus area of the original IndoEuropean homeland) as well as the beginning of wine-growing and its spread after the domestication of the grapevine, we are informed concisely by EIEC: 644–646 and in detail by the monographs of McGovern, Fleming and Katz 1996, McGovern 2003 with further references. A fundamental study on the origins and spread of wine, which ties together data from as many sources as possible (linguistics, archaeology, literature), has recently been provided by the as yet unpublished Ph.D. thesis of Luke Gorton (Gorton 2014) and an article based on it (Gorton 2017). Further references to historico-cultural, philological, and linguistic treatments are to be found in Aspesi 2013, Brogyanyi and Lipp 2016 and the comprehensive etymological lemma on the Hittite word for ‘wine; grapevine’ in HEG IV: 556–564. 1

The Anatolian Evidence

1.1 Hittite u̯ ii̯an(a)- c. ‘wine; grapevine’ (written mostly as the Sumerogram GEŠTIN ‘wine’ or GIŠGEŠTIN beside its simplified variant GEŠTIN ‘grapevine’ with or without phonetic complements) is to be regarded as the continuant of the Proto-Anatolian form *u̯ íi̯on- < Pre-Proto-Anatolian *u̯ íh1-on- (*/h1/ without vowel coloration, i.e. without oral constriction, therefore probably articulated as a glottal fricative [h] as in German Haus, English house).

The Word for wine

197

According to Kloekhorst (2008: 1012), the Hittite form represents an original n-stem u̯ ii̯an- matching Hieroglyphic Luwian u̯ ii̯an-i-, u̯ in-i- c. ‘grapevine’ and the base of the corresponding adjective Cuneiform Luwian u̯ in-ii̯a- (see below §1.2), which display apophony of the athematic stem (no vowel breaking, see below §7.3), and there is no clear evidence for a thematic a-stem in Anatolian. In fact, within the Hittite paradigm itself, practically all attested forms with phonetic complements or in phonetic spelling (except for the instr.sg. ú-i-i̯aan-ta in fragmentary context, see below) are ambiguous as to their stem formation, as can be seen in the following sample (for the Hittite attestations in their context, see HEG IV: 557–558 with further references): acc.sg. ú-i-i̯a-na-an = /u̯ ii̯anan/ ‘wine’ (KBo xii 126+ i 69 [MH/MS]) in purely phonetic spelling besides GEŠTIN-na-an ‘wine’ (KUB xxxvi 110 Rs. 7 [OS]; KBo xvii 11 + Rs. iv 23 [OS]; KUB lvi 50 ii 2 [NH]), GEŠTIN-an ‘wine’ (KUB xxix 1 iii 2 [OH/NS]; StBoT viii Rs. iii 14 [OH]; Otten and Souček 1969: 30–31), gen.sg. GEŠTIN-i̯a-na-aš = u̯ i5-i̯a-na-aš ‘of the wine’ (in /u̯ ii̯anas ispanduzii̯assar/ ‘vessel for the libation of the wine’ KUB lvi 50 ii 5 [NH]; GEŠTIN = u̯ i5-), GEŠTIN-aš ‘of the wine’ (KBo xxiv 34+ Vs. 23 [MH]), abl.sg. GEŠTIN-i̯a-na-az = u̯ i5-i̯a-na-az ‘with wine’ (KBo xiv 103 Rs. iv 11; GEŠTIN = u̯ i5-), instr.sg. GEŠTIN-ni-it = /u̯ ii̯anid/ ‘with wine’ (KUB X 95 Vs. iii 6 [NH]), GEŠTIN-it ‘with wine’ (vsnf12, 21 Vs. iii 8 [NH]); in fragmentary context: instr.sg. GEŠTIN-ta šu-un[-na-aḫ-ḫé ‘and with wine fill[ed i’ (KBo xvii 4 iii 16’ [OH]; cf. Otten and Souček 1969: 34– 35, 129; -ta = /-t-a/ < */-d-ha/ with enclitic -a ‘and’ < *-ha = Pal., CLuw. -ḫa ‘and, also’ < *-h2o ‘at that’); with clearly athematic stem formation, on the other hand, instr.sg. ú-i-i̯a-an-ta = /u̯ ii̯an-d/ (IBoT iii 93 i 12), which for its fragmentary context is unclear in meaning however: ‘with wine’ or ‘with grapevine’? (on instr.sg. GEŠTIN-ta and ú-i-i̯a-an-ta cf. Oettinger 2003: 143–144 n. 4; for further attestations of forms from the paradigm of the Hittite word for ‘grapevine; wine’, see below §§1.1.3–4). The Sumerogram GEŠTIN = /u̯ ii̯an-/ was also used by acrophony as a syllabic sign with the value , which in the philological classification is represented as the syllabogram u̯ i5-, e.g. pa-an-ga-u-u̯ i5 = pa-an-ga-u-i dat.-loc. of panku- ‘assembly’ and na-a-u̯ i5 = na-a-ú-i ‘not yet’. Cf. silvestri 1974: 266 with n. 1, rüster and neu 1989: 153–154 Nr. 131, hoffner and melchert 2008: 21

For the shape of the instrumental ending, cf. athematic Hitt. genu- ntr. ‘knee’ (OHitt.+ nom.acc. gēnu, gen. gēnuu̯ aš ← PIE *ǵón-u/*ǵén-u- → *ǵnéu̯ -) with instr. OHitt. genut, ganut; Hitt. keššar-/kišr-/kiššer- c. → ntr. ‘hand’ (< *ǵʰés-or-/obl. *ǵʰes-r-́ ← PIE loc. *ǵʰes-ér from root noun

198

Lipp

­*ǵhós-/*ǵhés- ‘hand’ [Neri 2013: 195–196, 200]; nom. OHitt. keššar < *ǵʰésōr and remodelled kišširaš etc., Gr. χείρ f. < obl. *ǵʰes-r-) with instr. OHitt. kiššarat, MHitt. kiššarta = /gisar-d/, MHitt. kiššarit, kišrīt; and Hitt. išḫiman-/išḫimen- c. ‘cord, rope’ (< *sh2i-món-/obl. *sh2i-mén- ← PIE *sh2éi̯-mon-/obl. *sh2i-mén-; nom. OHitt. isḫimāš < *sh2i-mṓ+s ← PIE *sh2éi̯-mō(n) vs. later remodelled nom. išḫimanaš; see below §§1.1.3, 7.4) with instr. OHitt.+ išḫimanta = /isḫiman-d/, MHitt. išḫimanit as against thematic šākuu̯ a coll.pl. ‘eyes’ with instr. šākuu̯ at, šakuit and secondarily thematic antuḫša- ‘human being, man’ with instr. antuḫšet. On the Hittite paradigm forms see hoffner and melchert 2008: 13, 77, 79–80, 101, 111–112, 115–116; on the genesis of the athematic instrumental ending /-d/, /-id/ see melchert and oettinger 2009

1.1.1 H.C. Melchert (p.c.) assumes, however, that, corresponding to Luwian /u̯ ii̯an-ī-/, /u̯ ī̆n-ī-/ ‘grapevine’ from athematic *u̯ ii̯an- with a further derivative /u̯ in-ī-/ for the designation of ‘wine’ in Hier.Luw. (continuing the affiliation adjective u̯ inii̯a- ‘pertaining to a grapevine’ as attested in Cun.Luw.; see below §1.2), in Hittite the athematic stem u̯ ii̯an- was limited to the designation of the ‘grapevine’, which was rendered by the Sumerogram GIŠGEŠTIN or simplified GEŠTIN ‘grapevine’, and that for the designation of ‘wine’, rendered by the Sumerogram GEŠTIN ‘wine’, Hittite used a descriptively thematic hypostasis u̯ ii̯an-a-*, which was based on the athematic stem and thus denoted *‘drink from the grapevine’; this hypostasis would then underlie the attested inflectional forms for ‘wine’, which are ambiguous as to their stem formation like the afore-­ mentioned forms, e.g. gen.sg. GEŠTIN-i̯a-na-aš = u̯ i5-i̯a-na-aš ‘of wine’. According to Melchert’s analysis, a hypothetical nom.sg. Hitt. */u̯ ii̯an-a-s/ ‘wine’ would be a hypostasis of the gen. /u̯ ii̯an-as/ ‘(that) of the grapevine’, which belongs to the athematic paradigm of /u̯ ii̯an-/ ‘grapevine’ (for this analysis, see Gorton 2014 with n. 91, 2017: 7–8). This explanation would account for the formation in question, because the mechanism of forming a hypostasis from the genitive was productive in Hittite, cf. u̯ aštul ntr. ‘sin’ with gen. u̯ aštulaš ‘(the one) of the sin’ = ‘the sinner’, tai̯azil ntr. ‘theft’ with gen. tai̯azilaš ‘(the one) of the theft’ = ‘the thief’ (cf. Friedrich 1974: 123, Hoffner and Melchert 2008: 256, Puhvel 2012). 1.1.2 In fact, for finding out the underlying stem class of the word for ‘wine’ and ‘grapevine’ we have only the attestation of a clearly athematic instrumental form in fragmentary context, that is ú-i-i̯a-an-ta (see above §1.1). But as its

The Word for wine

199

meaning is unclear from the context, it therefore does not tell us if the athematic stem u̯ ii̯an- also had the meaning ‘wine’ beside ‘grapevine’. For a descriptively thematic stem *u̯ ii̯an-a- ‘wine’ as base of a derivational form, one could perhaps adduce the Anatolian toponym /u̯ ii̯ana-u̯ anta-/ (synchronically understandable as *u̯ ii̯ana-u̯ ant- ‘abundant in wine’), which in the Hittite transmission appears as URUu̯ i5-i̯a-na-u̯ a-an-da KBo iv 3 1 18 (in the contract between the Hittite king Muršiliš ii and Kupanta-dKAL, king of the Luvian states Mira and Kuu̯ alii̯a), URUu̯ i5-i̯a-na-u-u̯ a-an-ta (Brandenstein 1943: 14, 53–54, Text 2 Vs. ii 1 [Text 2 = vat 6688 + Bo 2496]), URUu̯ i5-i̯a-na-u̯ a-an[-taaš] KUB xxvi 43 1 15 (in the royal decree by Tudḫalii̯aš iv and Puduḫepa, beside the variant given below). In the classical epoch, this toponym has a correspondence in the name of the Lycian city of Οἰνόανδα (ntr.pl.) = Oenianda in the upper Xanthos valley (Strabo xiii 631; Ptolemy, Geog. v 3,6; Pliny, Nat. Hist. v 101) and Oeniandos, later called Epiphania, in southeast Cilicia (Pliny, Nat. Hist. v 93). As in the Greek version there is an adaption to the thematic noun οἶνος ‘wine’, one cannot decide on this evidence if the Anatolian base word had a final stem vowel or not. But the use of the syllabogram instead of before or shows that there is a real a-vowel after the preceding /n/ and that therefore the following /u̯ / is intervocalic, a fact which can be expressed orthographically by the use of an additional vowel syllabogram, thus besides to render a sequence /-nau̯ a-/ (for this orthographic usage, cf. Kimball 1999: 105). On the other hand, there are indeed Hittite versions of this toponym without a final a-vowel of the derivational base, i.e. /u̯ ii̯an-u̯ anta-/ attested in u̯ i5-i̯anu-an-ta-aš KBo ii 7 1 18, and /u̯ in-u̯ anta-/ = u̯ i5-nu-u̯ a-an-ta-aš KUB xxvi 50 1 4 (in the royal decree by Tudḫalii̯aš iv and Puduḫepa, beside the variant given above), ú-i-nu-an-da KBo xvii 103 1 16, 17. These forms continue the possessive formations *u̯ ii̯an-u̯ ant- and *u̯ in-u̯ ant- ‘abundant in grapevine/ vine-stocks’ based on athematic *u̯ ii̯an-/*u̯ īn- ‘grapevine, vine-stock’ as reflected by Luwian (Hieroglyphic Luwian u̯ ii̯an-i-, u̯ in-i- c. ‘grapevine, vine-stock’, Cuneiform Luwian u̯ in-ii̯a- adj. ‘pertaining to a grapevine’; see below §1.2). The variant /u̯ ii̯ana-u̯ anta-/ might thus be a remodelling of these primary forms under the influence of an existing thematic noun *u̯ ii̯an-a- ‘wine’, whereas Οἰνόανδα, the epichoric form of classical times, might be a Greek adaption of /u̯ in-u̯ anta-/ (derived from athematic *u̯ īn- ‘grapevine’) with u̯ being rendered by Gr. in the absence of /u̯ / in Greek. As for Oeniandos, this has been supposed to be based on Hitt.-Luw. u̯ inii̯ant- as a toponym with a different formation meaning originally *‘wine, must, grape juice’ (cf. Neumann 1962: 208, 211; for the formation, see below §1.1.5), but it has to be considered that also Οἰνόανδα shows the variant Oenianda in Latin rendering with the vowel -i- perhaps as in

200

Lipp

­compounds. (For the analysis of the variants, see Silvestri 1974: 272–273, 274; for the attestations id. 269–271 [Cuneiform], 272–273 [classical] on the basis of Laroche 1961: 61; for the Cuneiform attestations, cf. del Monte and Tischler 1978: 482–483, HEG IV: 560 with further refs.; for the attestation and localization of the respective Anatolian toponyms in classical times, cf. Zgusta 1984: 431–432 with further refs.; and for the identification of Hieroglyphic Luwian (REGIO)VITIS ‘vine-stock country’ with the toponym /u̯ ii̯an(a)u̯ anta-/, see Poetto 1993: 48–49, 80 with n. 193 and further refs., Hawkins 1995: 29, 54–55, 78.) 1.1.3 On the other hand, one also has to consider the possibility of the occurrence of a secondary thematic stem in -a-, which is neither conditioned by derivation nor hypostasis, but rather due to late paradigmatic remodelling of an athematic stem. For Old Hittite, one can expect a primary athematic nom.sg. */u̯ ii̯as/ or */u̯ ii̯ās/ (< Pre-Proto-Anatolian *u̯ íh1-ō+s or *u̯ ih1-ṓ+s ← PIE nom.sg. *u̯ éi̯h1-ō(n) beside acc.sg. *u̯ ei̯h1-ón- with accent shift according to the *ku̯ etu̯ óres rule, and obl. *u̯ ih1-n-́; for the reconstruction of the stem formation and its apophony, see below §§6.1, 7.1, 7.4, 9.1). Within the inner history of Hittite, this form could have been replaced by a descriptively thematic nom.sg. */u̯ ii̯an-a-s/ backformed in analogy to acc.sg. GEŠTIN-na-an = /u̯ ii̯an-an/ etc.; cf. the n-stem išḫiman- c. ‘cord, rope’ with nom.sg. OHitt. išḫimāš (< *sh2i-mṓ+s = OSax. sīmo, OE sīma ‘rope’ < Pre-Proto-Germanic *sih2-mō̃ ← PIE *sh2éi̯-mō(n)/obl. *sh2imén-) and its younger remodelling into nom.sg. Hitt. išḫimanaš in analogy to acc.sg. išḫimanan (< *sh2i-món- ← PIE *sh2ei̯-món- with accent shift by the “ku̯ etu̯ óres rule”, see below §7.4; for the paradigm, cf. Harðarson 1987: 122, Hoffner and Melchert 2008: 111). And there is indeed an Old Hittite form GEŠTIN-aš in the meaning ‘grapevine, vine-stock’, which can either be interpreted as nom.sg. */u̯ ii̯as/ or */u̯ ii̯ās/ (from athematic /u̯ ii̯an-/) or */u̯ ii̯an-a-s/ (backformation from the athematic acc.sg. /u̯ ii̯an-an/ etc.) or as gen.sg. */u̯ ii̯an-as/ (from athematic /u̯ ii̯an-/): n=ašta anda ḫalkiaš dGÌR-aš G[EŠ]TIN-aš kitta ‘and in it (i.e. in the fleece) lies/is put the field god of the grain and (of) the vine-stock’ (KUB xvii 10 iv 30 [OH/MS]; cf. HEG IV: 557–558). Thus, even if a Hitt. nom.sg. of the form */u̯ ii̯an-a-s/ in the meaning ‘wine’ were unambigously attested, we still could not tell if this formation is merely a secondary inner-Hittite remodelling within the paradigm of an athematic stem u̯ ii̯an- ‘wine ← grapevine’ showing a metonymic semantic shift (cf. Akkad. karānu ‘grape, grapevine; wine’; see below §11), or if it continues a Pre-Hittite hypostasis *u̯ ii̯an-a-s ‘wine’ = ‘drink from the grapevine’, which arose by reanalysis of the athematic gen.sg. /u̯ ii̯an-as/ ‘(that) of the grapevine’ (see above

The Word for wine

201

§1.1.1). In this case, we are therefore confronted with a heuristic dilemma and can only say that in Hittite the existence of a distinct hypostasized a-stem *u̯ ii̯an-a- meaning ‘wine’ beside an n-stem u̯ ii̯an- meaning ‘grapevine’ is possible and inferable with some plausibility, but not provable [so also L. Gorton per litteras]. 1.1.4 Nom.sg. Hitt. GEŠTIN-iš = /u̯ ii̯an-i-s/ ‘grapevine, vine-stock’ appears as expression of a generic singular in the following passage of a New Hittite version of an Old Hittite text (KBo vi 26 i 18–19; HEG IV: 558): takku GEŠTIN-iš GIŠ ḪAŠḪUR [GIŠḪAŠḪUR.KUR.RA našma GIŠŠ]ENNUR u̯ arāni ‘if the grapevine/vine-stock, the apple tree, the apricot tree or medlar tree burns down’. Due to the fact that the Sumerograms in this passage do not show plural markers and the verb is in the singular (u̯ arāni ‘burns down’), it is likely that in the series of plants mentioned each is enumerated in the generic singular; therefore it is less likely that GEŠTIN-iš is a form of the nom.pl. with the late New Hittite ending -iš (for this ending, cf. Hoffner and Melchert 2008: 71 with refs.). This i-stem formation meaning ‘grapevine, vine-stock’ is probably a late Hittite borrowing or remodelling on the basis of Luwian /u̯ ii̯an-ī-s/ ‘grapevine, vinestock’, as attested by Hieroglyphic Luwian (‘VITIS’)wa/i-ia-ni-sa, with normal Luw. ī-mutation in the genus commune (see below §1.2). A clearly hybrid Hitt. i-stem formation like Hitt. GEŠTIN-iš, which was triggered by the productive Luw. ī-mutation of Luwians speaking Hittite, can also be seen in another NHitt. version of an OHitt. text, that is nom. sg. antuu̯ aḫḫiš ‘human being’ as a remodelling of a clearly archaic form antuu̯ aḫḫaš stemming from the OHitt. original (< nom.sg. PIE *en-dhu̯ éh2ōs from s-stem *en-dhu̯ éh2-os-/*en-dhuh2-s- ‘having breath in oneself’). Already in the spoken language of the Middle Hittite epoch, this archaic form was replaced with antuḫšaš by generalization of antuḫša- based on the oblique stem as in OHitt. gen.sg. anduḫšaš (< *en-dhuh2-s-ós). See rieken 1999: 191–192 with n. 884, kimball 1999: 184, hoffner and melchert 2008: 79–80 with n. 1

1.1.5 Luwian origin also has to be considered for the derivative Hitt. (d)u̯ inii̯ant- ‘(the genius of the) grape juice, must’ < *u̯ inii̯-ant- ‘consisting of grape juice’ with acc.sg. dú-i-ni-i̯a-an-ta-an (KUB lv 65 iv 16), ú-i-ni-i̯a-an-da-an (KUB xxv 37 iii 17. 19 [NH]; KUB xxv 37 iv 6 [NH]) = /u̯ inii̯ant-an/ because of the formation of the base word corresponding to Hieroglyphic Luwian /u̯ inī-/ < *u̯ in-ii̯a- ‘wine,

202

Lipp

must, grape juice’ = ‘drink made from grapevine’, derived from u̯ ii̯an(-i)-, u̯ in(-i)‘grapevine’ (cf. Starke 1990: 381 n. 1378, Oettinger 2003: 143 with refs., HEG IV: 563–564 with text attestations; see below §1.2). 1.2 Hieroglyphic Luwian u̯ ii̯an-i-, u̯ in-i- c. ‘grapevine, vine-stock’ < Proto-Anatolian *u̯ íi̯on-/loc. *u̯ ii̯én(-i) (é > Luw. a)/obl. *u̯ ī̆n- < Pre-Proto-Anatolian *u̯ íh1-on-/ loc. *u̯ ih1-én(-i)/obl. *u̯ ih1-n-: e.g. nom.sg. (‘vitis’)wa/i-ia-ni-sa = /u̯ ii̯an-ī-s/ (SULTANHAN §7), wa/i-ia-ni-sá (SULTANHAN §15), wa/i-ia-ni-sa (SULTANHAN §23); acc.sg. (VITIS)wa/i-ni-na = /u̯ ī̆n-ī-n/ (KÖRKÜN §11), ?dat.-loc.sg. wa/i-ia-ni-[x]-i = /u̯ ii̯an-i/? (KULULU 1 §8). In the nominative and accusative, this original n-stem shows the Luwian enhancement of the stem by /-ī-/, which is regular in the genus commune and is called ī-mutation. Due to the effect of the widespread, but not pervasive Luwian contraction ii̯a > i (= /-ī-/ presumably), which becomes more frequent in the 1st millenium bc (cf. Yakubovich 2015: 10 §5.2.1), some instances of Luw. u̯ in- as in the acc. /u̯ ī̆n-ī-n/ might go back to the original strong stem u̯ ii̯an- as inferable for the acc. */u̯ ii̯an-ī-n/ besides the attested nom. /u̯ ii̯an-ī-s/, but in other cases Luw. u̯ in- can continue the weak stem variant *u̯ ī̆n- < Pre-Proto-Anatolian *u̯ ih1-n-. A derivative of this n-stem is attested in the form of the instr.sg. Hier.Luw. wa/i-ní-ti ‘with (measures of) wine’ (KARKAMIŠ A30h §3; for the semantic interpretation see Hawkins 2000: 178, 468) = /u̯ inīdi/ < *u̯ inii̯adi with the Luwian contraction ii̯a > i. Thus in the case of this Hieroglyphic Luwian form, we are dealing with a stem /u̯ inī-/ < *u̯ in-ii̯a- (< *u̯ īn-ii̯ó-; for the respective contraction also in Luw. -ii̯astems, cf. Yakubovich 2015: 13 §6.2) meaning ‘wine, must, grape juice’ = ‘drink made from grapevine’ as a designation of affiliation derived from u̯ ii̯an(-i)-, u̯ in(-i)- ‘grapevine’ (for this morphological and functional analysis based on a suggestion by C. Melchert, see Gorton 2014 with nn. 89, 90; 2017: 6f.; see above §1.1.1), like Cuneiform Luwian u̯ in-ii̯a- adj. ‘pertaining to a grapevine’: coll.pl. ú-i-ni-i̯a = /u̯ inii̯a/ (KBo xiii 263,2; quantity of -ī̆- in /u̯ ī̆n-/ depending on the position of the Anatolian accent). (For the attestations, cf. Starke 1990: 381 n. 1378, Melchert 1993: 269, Kloekhorst 2008: 1012, ACLT s.v. Hieroglyphic wiyani(ya)-, HEG IV: 558–559.) 1.2.1 The usual designation for ‘wine’ in Luwian has a different etymological origin: CLuw. ma-ad-du = /máddu/, HLuw. (VITIS)ma-tú-sà, ma-tu-sà = /máddu-sa/ ntr. ‘wine’ (Luwian gemination by fortition of a lenis after inherited *é in open syllable by Lex Čop) < Proto-Anatolian *médu ‘sweet grape-juice’ < PIE *médhu- ‘honey, mead’ = Ved. mádhu- ‘sweet drink, honey’, Gr. μέθυ ‘intoxicating

The Word for wine

203

drink, wine’, OHG metu ‘mead’, OCS medъ ‘honey’ etc. (cf. Starke 1990: 191 n. 624, HEG II: 165–166, Melchert 1993: 144–145, ACLT s.v. Hieroglyphic mattu-, s.v. Cuneiform mattu-). 2

The Greek Evidence

Greek οἶνος m. ‘wine’ (Hom.Il.+), gen.sg. Cret. ϝοινο (Gortyn), Cypr. wo-i-no = /u̯ oi̯nō/, and already Mycenaean wo-no = /u̯ oi̯nos/, Myc. ox name wo-no-qo-so = /u̯ oinoku̯ -orsos/ *‘having wine-coloured buttocks’ (for the Mycenaean evidence, see DMic. ii 442–445; cf. Hom. οἶνοπ- ‘wine-coloured, dark red’ < *u̯ oi̯noku̯ - ‘wine-coloured, having the appearance of wine’) < Proto-Gr. *u̯ ói̯no- < PIE *u̯ ói̯(h1)-no- ‘wine’ (on the formation, see below §§10.2.1–2), from which the collective Gk. οἴνη f. ‘grapevine, vine-stock’ (Hesiod+) < *u̯ oi̯nā (in the functional meaning *‘wine plant’) is derived (cf. ἐλαίᾱ, Ιon. -η ‘olive-tree, olive’ : ἔλαιον ‘olive oil’). Beside these forms, there were semantically corresponding continuants of *médh-u- ‘honey, mead’ in Myc. me-tu-wo ne-wo gen. ‘in the season of the new wine’, Gr. Hom.Il.+ μέθυ ‘intoxicating drink, wine’, which in everyday language soon fell out of use and survived only as an archaic word in later poetry (for the Myc. evidence, see DMic. I 446; for the lexeme, see above §1.2.1 and below §5.6.2). Context-sensitive laryngeal loss by Saussure’s Law: oRH.C > oR.C. Cf. *tórh1-mo- ‘hole’ > *tór-mo- > Gr. τόρμος ‘hole’ against *térh1-tro- > Gr. τέρετρον ‘driller’; *sólh2-u̯ o- ‘whole, entire, undamaged’ > *sól-u̯ o- > Gr. ὅλος, Ved. sárva- against *sl ̥h2-eu̯ -ó- > Proto-Italic *salau̯ o- > Lat. salvus, Osc. σαλαϝσ, *sl ̥h2-i- ‘prosperity’ > CLuw. šalḫi-tti- (for further specification of Saussure’s law, see below §3). As etymologically related formations, note also the n-stem Gr. υἱήν, ἢ υἱόν· τὴν ἄμπελον (Hesychius) = *ϝιήν ‘grapevine, vine-stock’ < Proto-Gr. *u̯ ii̯ḗn < PIE *u̯ ih1ḗn(ontheformation,seebelow§8),andtheu-stemMyc.we-je-we=/u̯ ei̯-eu̯ -es/nom. pl. f. ‘(vine) shoots, sprouts’ (DMic. ii 418) from *u̯ éi̯h1-u-/*u̯ ih1-éu̯ - > *u̯ éi̯u-/ *u̯ ii̯éu̯ - (for the analysis as feminine u-stem with apophonically conditioned suffix allomorph *-éu̯ - and not as masculine eu̯ -stem, cf. Perpillou 2012: 111) beside *u̯ éi̯-tu-/*u̯ i-téu̯ - ‘(willow) rod’ (with secondary root apophony) > Hom.+ ἴτυς [ῐ] ‘rim of a wheel/shield’ (Aeolic *ϝίτυς, acc. to Terentianus Maurus 658: vitym gens Aeoli; cf. Lat. vītis ‘tendril, grapevine’), and derived Hom.+ ἰτέη [ῐ], Att. ἰτέα [ῑ] ‘willow; wicker shield’ (γιτέα [= *ϝιτέα]· ἰτέα Hesychius), Εἰτέα Attic deme name (for the reshaped apophony, cf. Neri 2017: 324–325; see below §6.1).

204 3

Lipp

The Armenian Evidence

Armenian gini ‘wine’ (gen. ginwoy = [ginu̯ o] with -u̯ o < *-i̯ói̯o < *-ii̯óhi̯o undergoing pretonic syncope of unaccented *i due to the prehistoric penultima ­accent) < (Pre-)Proto-Armenian *gu̯ īníi̯o < *u̯ ēnii̯on < *u̯ oi̯n-ii̯onm ntr. < *u̯ oi̯nih2ó- as adjective of affiliation on the basis of the PIE noun *u̯ ói̯(h1)-n-o- > *u̯ ói̯n-o-, which is continued by Gr. (ϝ)οἶνος ‘wine’ (on the development and paradigm of Arm. gini, cf. Schmitt 1981: 43, 52, 68, 69, 71, 95, Matzinger 2005: 20, 66–67 with n. 304, and Olsen 1999: 439–440, who explains the Arm. vacillation between wo- and ea-stem on the basis of an old neuter). For derivational reasons, the Arm. form is less likely developed from (*u̯ ēnii̯o- oR.C or .HRo > Ro with tautosyllabic reduction according to Neri (2003: 32 n. 69). Cf. *h2u̯ ors-ó-/-éh2- ‘drop; dribble, rain’ > PIE *u̯ ors-ó- resp. *u̯ ors-éh2- > Hitt. u̯ arša- ‘fog, rain’, Luw. u̯ arša- ‘drop’, causative *h2u̯ ors-éi̯e/o- ‘to drop, let rain’ > PIE *u̯ ors-éi̯e/o- > Gr. οὐρέω ‘urinate’ vs. PIE *h2u̯ érs-h2/-éh2- ‘drop’ > Gr. *au̯ ers-a-/-ā- > Ion.Att. ἕρση, Hom. also ἐέρση, Sappho ἐέρσᾱ, Pi. ἔερσᾰ, Papyrus ἀέρση, Cret. ἄερσᾰ (Hsch.) ‘dew(drop)’. For laryngeal loss also in a sequence with post-resonantic thematic vowel -o- per Neri (2003: 285 n. 949; 302 n. 1025), cf. *peth1- ‘to fall’ in Gr.Hom. πεπτεώς ‘ducked, crouching’ < perf.ptc. *pé-pth1-u̯ os- vs. the verbal noun *póth1-mo- ‘fall(ing)’ = *pót. h1mo- > PIE *pót.mo- > Gr. πότμος ‘lot, destiny’. Cf. lipp 2009b: 417 n. 141; 454, 456, 446 n. 222 and 340–341 n. 66 with further refs. to Saussure’s Law

The sound shape of (Pre-)Proto-Armenian *gu̯ īníi̯o (gen. *gu̯ īni̯ói̯o) is to be presupposed for the borrowing into Proto-Kartvelian (= South Caucasian) with *γu̯ ini̯o ‘wine’ > Georgian γvino, Old Georgian nom. γwnoy, γwinoy = /γu̯ ino-i̯/, Mingrelian gvini (with different adaptions of final *-i̯o [Sergio Neri, p.c.]). (For alternative interpretations and specifications, cf. Gippert 1994: 117–121; on the origin of vine cultivation in the sub-Caucasian area of Georgia and the Urartean-Armenian highland, cf. Greppin 2008.) 4

The Albanian Evidence

Albanian Tosk. vérë, Geg. vênë ‘wine’ < *ven- < Proto-Albanian *u̯ oi̯nā ← PIE *u̯ ói̯(h1)-n-o- (see Demiraj 1997: 414).

The Word for wine

5

205

The Italic Evidence

5.1 Latin vīnum ntr. ‘wine’ (since Livius Andronicus, Ennius, Plautus, Cato), Faliscan acc.sg. UI[NO]M [?] = /vīnom/ (ca. 600 bc; but interpretation doubtful, see below §5.2), UINO = /vīnom/ (4th c. bc), Umbrian acc.sg. VINU = /vīnom/, abl. sg. VINU = /vīnu/ (TI 3rd c. bc), uinu (TI 2nd c. bc) ‘wine’, Volscian abl.sg. uinu (ca. 300 bc) ‘wine’. 5.1.1 Attestations from the neighbouring non-Italic languages of Ancient Italy: Etruscan VINUN (Tarquinia, end of 6th c. bc), VINỤ[M:] (Populonia, second half of 5th c. bc), VINUM (Liber Linteus and Ager Saenensis, 2nd c. bc) ‘wine’ (ET I p. 107; ii pp. 1–8, 67, 136, 167; Agostiniani 1998: 1–2; for the question whether Italic *u̯ īnom was borrowed into Etruscan or vice versa, see below §§5.6, 5.6.1–2); (Celtic) Lepontic VINOM (Ornavasso, 2nd c. bc; interpretation uncertain: ‘wine’ according to Lejeune 1987: 502–509, followed by Solinas 1995: 375, Uhlich 2007: 407, or ‘white, shining’ < *u̯ indo- = Gaulish vindo- according to Tibiletti Bruno 1981: 163–164, cf. Agostiniani 1998: 2 n. 10; if Lep. VINOM means ‘wine’, then it might be borrowed from Etruscan or an Italic language). 5.2 First attestations in Italic: Faliscan LF 1 = LDIA: 525–526 (ca. 600 bc): UI[NO]M:P[ORE]KADEUIOS ‘vinum porrigat Euius (= Bacchus)’ (‘Evius shall procure the wine’), LF 5 = LDIA: 529–530 (4th c. bc) i: FOIED.UINO.PIPAFO. CRA.CAREFO ii: FOIED.UINO.PAFO.CRA.CAREFO ‘hodie vinum bibam, cras carebo’ (‘today I will drink wine, tomorrow I will be without’). According to the new edition of the Faliscan inscriptions by Bakkum, the traditional restitution of the reading UI[NO]M in LF 1 = EF 1 is doubtful (Bakkum 2009: 199, 393–406); assured is thus only UINO in LF 5 = MF 59, 60 (Bakkum 2009: 199, 434–435). First attestations in Etruscan (ET ii pp. 67, 136): Ta 0.6 (vas, end of 6th c. BC) VINUN ‘vinum’; Po 4.3 (cippus, second half of 5th c. BC) FASLE:VINỤ[M:] M[…]AŚ ‘vas / libatio vinum…’. 5.3 Probable cognates of Lat. vīnum, Umbr. VINU etc. are to be found in the personal names Pre-Samnite Ps 3 nom. VINUXS = /u̯ īnọ̄χs/, patronymic dat. VINICIIU = [u̯ īnikii̯ọ̄] (Campania, 500–450 bc; the Pre-Samnite inscriptions from Campania, Lucania and Bruttium convey a dialect form close to that of South Picene and date back to the period before the invasion of the area by the Samnites beginning in 440 bc; see ST pp. 5–6), Oscan Po 3 gentile name nom.

206

Lipp

VIÍNIKIÍS = [u̯ īniki̯ẹs] with dissimilation from *-i̯is < *-ii̯īs < *-ii̯-ii̯o-s as derivative of the patronymic (cf. Meiser 1986: 62–66), adapted as Lat. Vīnucius, Vīnicius, Etr. gen. Vinucenas (for the Sabellic attestations cf. ST pp. 71, 104; for the corresponding forms in Latin and Etruscan cf. Schulze 1966: 110 n. 3; 380) and also perhaps as Faliscan uịnụ[cio] (Bakkum 2009: 276, 556). These forms can be analysed as continuants or derivatives in *-ii̯ó- (< *-ih2ó-) of the compound *u̯ īn-oku̯ -, nom.sg. *u̯ īn-ōku̯ -s ‘wine-coloured, having the appearance of wine’ (cf. Goth. wein-nas m. a-stem ‘having a nose (red) of wine’; Casaretto 2004: 58), thus *u̯ īnok-ii̯o- > Pre-Samnite VINICIIU with delabialisation by analogy to VINUXS = /u̯ īnọ̄χs/ < *u̯ īnōks. For the apophony of the second compound member, cf. etymologically related Gr. oἰνώψ Soph., οἶνοπ- Hom. ‘winecolored, dark red’ < *u̯ oi̯n-ōku̯ -s, *u̯ oi̯n-oku̯ -, also used as personal name in Hom.Od. 21,144 Οἶνοπος υἱός ‘son of Oinops’ (see above §2; formation like Lat. atrōx, -ōcis ‘terrible in appearance, dreadful’, but without generalisation of the lengthened grade in Pre-Samnite, pace Meiser 1986: 172 n. 3). The stem displays the usual analogical lengthened grade in the nom.sg. of pie monosyllabic root nouns and shows the effect of the preconsonantal Italic delabialisation of *ku̯ as in the development of nom.sg. PIE *u̯ ōku̯ -s ‘voice’ (Ved. vāḱ , Av. vāxš vs. Gr. acc. ὄπ-α) > Italic *u̯ ōks > Lat. vōx, which led to an analogical delabialization in the derived denominative *u̯ oku̯ -ā-i̯e/o- ‘to call’ > Italic *u̯ oku̯ āe/o- → *u̯ okāe/o- > Lat. vocō, -āre, Umbr. subocau ‘invoco’ (for the delabialisation in Sabellic, cf. Buck 1928: 95). In contrast to the Etruscan gentile name Vinacna (Tarquinii, 6th c. bc), which is derived from the Sabellic stem in the shape *u̯ īnok-, the Pre-Samnite form VINICIIU < *u̯ īnok-ii̯o- with medial -i- beside VINUXS < *u̯ īnōk-s either shows the effect of a vowel weakening in medial syllables or might have developed a context-sensitive anaptyctic vowel -i- after the beginning of Sabellic vowel syncope in open medial syllables like the later Oscan form VIÍNIKIÍS, but the variants with medial -u- such as Lat. Vīnucius, Falisc. uịnụ[cio], Etr. Vinucenas rather speak in favour of the first solution, i.e. for vowel weakening (for cases of vowel weakening in Sabellic instead of syncope, cf. Buck 1928: 55–57, Meiser 1986: 203, 267–272, Lipp 2009b: 131–132; for vowel anaptyxis in Oscan, cf. Buck 1928: 50–53). On the other hand, there has been the attempt by Kölligan (2013: 120–121) to derive the aforementioned Sabellic names from a compound form *u̯ ī-noku̯ -o‘who harms/kills by force’, connecting it with Lat. vīs f., abl. vī ‘force’ and Lat. noxa ‘crime, damage’, nex ‘death’ on the base of an alleged root *neku̯ - ‘to disappear; perish’. But in view of the etymologically corresponding forms Lat. noceō, -ēre ‘to harm’ = Ved. nāśáyati ‘makes disappear, destroys’ etc., one has to posit the PIE root as *neḱ- ‘to disappear, get lost’ (e.g. Ved. subj.aor. naśanti ‘will get lost’; LIV2: 451–452), which would entail an Italic compound form

The Word for wine

207

*u̯ ī-nok-o- ‘having/provoking disappearance/harm by force’, without there being however other verbal governing compounds in *-nok-o- (as against the attested formations Lat. nocuus ‘harmful’, innocuus ‘harmless, innocent’, noxius ‘harmful’) or fixed parallel phrasal collocations with Lat. vī ‘by force’. 5.4 The forms of the word for ‘wine’ in the individual Italic languages can only go back to Proto-Italic *u̯ īnom, which would plausibly continue a morphologically regular IE ablaut form *u̯ ih1-n-o- (with zero grade of the root as Hitt. u̯ ii̯an- ‘wine; grapevine’, HLuv. u̯ ii̯ani-, u̯ ini- ‘grapevine’ < *u̯ ih1-on-/*u̯ ih1-nagainst o-grade in Gr. (ϝ)οἶνος < *u̯ ói̯(h1)-n-o-). On the other hand, the following derivation is to be rejected and only presented argumenti causa: it starts from a presumed Italic *u̯ oi̯nom as continuant of PIE *u̯ ói̯(h1)-n-o- or borrowing of Gr. u̯ oi̯non*, acc.sg. of (ϝ)οἶνος m. (such a borrowing has been assumed by Szemerényi 1989: 166, Sihler 1995: 53): Italic *u̯ oi̯nom > early Lat. *u̯ einom (4th–3rd c. bc) > *u̯ ẹ̄nom (3rd c. bc) > vīnum (ca. 200 bc; *u̯ oi̯- > *u̯ ei̯- > *u̯ ẹ̄- as in PIE *u̯ ói̯ḱ-o- ‘settlement’ [= Gr. (ϝ)οἶκος ‘house’] > Italic *u̯ oi̯ko- > early Lat. *u̯ ei̯ko- > OLat. /u̯ ẹ̄ko-/ in inscr. vecos, veicus > Lat. vīcus ‘village’). Contrary to the actual evidence of the very early attested forms Falisc. /vīnom/ (4th c. bc), Umbr. /vīnom/ (3rd c. bc+), which cannot be borrowed from Latin, one would then expect as correspondences in Faliscan *u̯ oi̯nom > *u̯ ōnom (cf. Falisc. LOIFIRTA > LOFERTA = Lat. līberta), in Oscan *u̯ oi̯nom, and in Umbrian *u̯ ōnom (cf. *moi̯ni- [= Goth. ga-mains ‘common’] > Lat. com-mūnis, Osc. nom.sg.f. MÚÍNÍKÚ /moi̯nẹko/ ‘common’, Umbr. MUNEKLU /mōnẹklom/ ‘collection, offering’ < *moi̯ni-tlo-, nomen instrumenti from the denominal verb *moi̯ni-i̯e/o- ‘to make common’). 5.5 The Faliscan and Umbrian forms are therefore incompatible with a phonological derivation from older *u̯ oi̯nom (with Falisc. *u̯ oi̯nom, Falisc. Umbr. *u̯ ōnom as expected phonetic outcomes instead of actually attested Falisc. /vīnom/, Umbr. /vīnom/). In case of a development from Italic *u̯ oi̯nom, the Latin sound shape vīnum would have developed too late (ca. 200 bc) and thus cannot be the source for an early borrowing into Faliscan (UI[NO]M [?], ca. 600 bc, UINO 4th c. bc), Etruscan (VINUN, end of 6th c. bc) and Umbrian (VINU, 3rd c. bc with text tradition from 5th–4th c. bc). The Latin form vīnum phonetically matching Faliscan /vīnom/, Umbrian /vīnom/ hence cannot continue an inherited form *u̯ oi̯nom or a loan form *u̯ oi̯nom from Greek acc.sg. /u̯ oi̯non/*, which thus also cannot be the source of the other Italic forms. Therefore, the individual Italic forms presuppose *u̯ īnom as common base form (see Blümel

208

Lipp

1972: 26, Beekes 1987: 22–23, Flobert 1994: 290–291, who exclude a representation of former *oi̯ or *ei̯ in the respective Italic forms; likewise Gorton 2017: 12–13). 5.6 On the other hand, Agostiniani (1998: 5–6; similarly Gérard 2009: 19–20, 21) considers Etruscan VINUN, VINUM as a loan from Greek, being based on the acc. Gr. u̯ oi̯non* and displaying an Etruscan sound substitution /u̯ i-/ < */u̯ ui̯-/ ← Gr. /u̯ oi̯-/ similar to the process in the Etruscan name VILATAS ← Gr. Οἰλιάδᾱς with resyllabification of initial /u̯ i-/ = < */ui̯-/ ← Gr. /oi̯-/. However, in view of the incomplete match in the phonetic pattern of the compared forms, de Simone (2011: 476 with n. 17) is sceptical about Agostiniani’s loanword hypothesis, which he also rejects for palaeobotanical and chronological reasons. The Etruscan initial realization /u̯ i-/ = < */ui̯-/ ← Gr. /oi̯-/ is orthographically and phonologically distinct from the medial representation /ui̯/ = ← Gr. /oi̯/ as in Etr. TRUIA ← Gr. Τροία, Etr. ΦUINIS ← Gr. Φοῖνιξ; the diphthong /ui̯/ also appears in genuine Etruscan words like PAPUI, loc. in -i of the month name. In contrast to Etruscan VINUM with normal Etruscan substitution of the final nasal by -m as in PRUXUM ← Gr. acc. πρόχουν from πρόχους ‘jug’, Etruscan VINUN shows a rare final -n as in the vessel designation QUTUN (already attested in the 7th c. bc) beside the synchronically more frequent variant QUTUM ← Gr. κώθων. Cf. agostiniani 1998: 1–2

According to Agostiniani (1998: 6–7, 12–13 with n. 76), the Etruscan form VINUM spread together with the development of viniculture all over Italy from the end of the 8th century bc. In Agostiniani’s view, the Etruscan lexeme VINUM = /u̯ inum/ was adapted within the individual Italic languages in the shape */u̯ īnom/, the long vowel /ī/ being due to analogy of the semantically related Italic *u̯ ītis > Lat. vītis ‘grapevine’. But within such a scenario, one could also suppose that with regard to duration in accented position or to the degree of opening, the vowel phoneme /i/ of Etruscan (which lacked a phonological opposition of vowel quantity) had a closer phonetic resemblance to the Italic long vowel /ī/ than to the Italic short vowel /i/, which in contrast had a rather open articulation; cf. Lat. Catamītus ← Etruscan Catmite ← Gr. Γανυμήδης. Thus in Agostiniani’s view, Latin vīnum, Faliscan /vīnom/, Umbrian /vīnom/ < *u̯ īnom do not continue an inherited Italic form, but are loans from Etruscan VINUM, which for its part must be borrowed from Greek.

The Word for wine

209

5.6.1 Against this scenario, one might object that *u̯ īno- is already attested from 500 bc (i.e. within the oldest text layer) as a compound member in ancient Italic names like Pre-Samnite VINUXS (< *u̯ īn-ōk(u̯ )-s ‘wine-coloured, having the appearance of wine’) and its derivative Pre-Samnite VINICIIU, Osc. VIÍNIKIÍS (< *u̯ īn-ok(u̯)-ii̯o-) → Lat. Vīnicius etc. (see above §5.3). This means that in the Italic area the use of *u̯ īno- as an appellative for ‘wine’ must be older by several centuries. In favour of the borrowing of the word for ‘wine’ from Italic into Etruscan, and not vice versa, one can also point to FUFLUNS, the Etruscan name of the wine god Bacchus, which was probably borrowed from Italic, as, contrary to unaccented word-internal vowels, in this form the vowel of the second syllable is never syncopated, therefore probably reflecting a long vowel of Sabellic origin (because of the phonetic length of accented vowels in native Etruscan words, the last word-internal long vowel of a form borrowed from a foreign language was evidently interpreted by Etruscan speakers as being accented and therefore did not undergo syncope, cf. Rix 2008: 147). The following derivations have been proposed: Etruscan FUFLUNS ← (dissimilation) *fluflū́ns ← Italic *flōβlōno- < *bʰlōdʰlōno- < transposite *bʰleh3-dʰlo-h3no- ‘master of the garden’ from *bʰléh3-dʰlo- ‘place of blossom, garden’ related to Lat. flōs m. ‘blossom’ < *bʰléh3-os- (analysis by Meiser 1986: 9, 215–216); or rather, also based on *bʰleh3- ‘to blossom’, the following better founded alternative (based on Rix 1998 with modification): noun *bʰe-bʰl(h3)-o- ‘what is blooming, sprouting’ = ‘vegetation; ivy’ (reduplicated form of the type *ku̯ e-ku̯ l(h1)-o- ‘what is ­turning’ = ‘wheel’) → transposite *bʰe-bʰl(h3)-o-h3no- ‘master of the vegetation / of ivy’ > Italic *feβlōno- > Sabellic *foβlọ̄n(o)s (labial coloration e > o) > Palaeo-Umbrian *foβlọ̄ns (with long vowel ọ̄) → Etruscan FUFLUNS = [fuflū́ns]. For the Common Italic coloration e > o between two tautosyllabic labials as in Sabellic *pompe ‘five’ < *ku̯ énku̯ e, cf. the comparable reduplicated form Lat. populus ‘people’ ← ‘army’, Umbr. poplo ‘army’ < *peplo- < *pe-pl(h1)-o- ‘group of lance throwers’ = ‘group brandishing lances’ from *pelh1- ‘to swing, brandish’ (etymology per Rix 1995: 82; cf. Gr. πόλεμος ‘war, fight’ < *pól(h1)-mo- ‘the brandishing of lances’). For the borrowing process of a morphonologically corresponding formation into Etruscan, cf. the theonym Etruscan NEΘUNS ← Palaeo-Umbrian *nehtūns < *neftūns (with long vowel ū) = Lat. Neptūnus < Italic *neptūno- < transposite *nebʰ-tu-h3no- ‘master of the wet element / of the moisture’. 5.6.2 In contrast to vīnum, attested in Latin since Livius Andronicus, the archaic expression tēmētum (Old Latin and poetic) was a general term for an intoxicating

210

Lipp

drink, which was also used to refer to strong unmixed wine (also called merum ‘pure, unmixed wine’). Such a usage as a term for ‘wine’ is referred to by Paulus ex Festo 501, 6 (L): temetum vinum, unde temulentia et temulentus. It does not represent the old generic term for ‘wine’, however, as is also suggested by the related adjectives tēmulentus ‘drunk’, abstēmius ‘sober, abstemious’ ← *abs tēmō ‘without intoxication’ from *tēm-o- ‘intoxication, intoxicating substance’, which continues a vr̥ddhi derivative *tēmH-ó- ‘characterized by fainting, lead́ yati ing to faint’ from *temH- ‘to faint, become dark’ (related to Vedic tām ‘faints’ < *tḿ̥H-i̯e/o-; see de Vaan 2008: 609 and cf. LEW ii 657, DELL 679–680, IEW 1063, LIV2: 624). That tēmētum could still be associated with abstēmius in ­classical times is obvious not only from Quintilian, Institutio Oratoria I 7–9 (‘abstemius’, quia ex abstinentia temeti composita vox est), but also from the testimony of Gellius, Noctes Atticae X 23, 1, according to which tēmētum is the designation for ‘wine’ in the ancient language: Qui de victu atque cultu populi Romani scripserunt, mulieres Romae atque in Latio aetatem abstemias egisse, hoc est vino semper, quod ‘temetum’ prisca lingua appellabatur, abstinuisse dicunt. Those who have written about the diet and the way of life of the ­Roman people, say that the women in Rome and Latium lived an abstemious life, that is, they abstained permanently from wine, which in the ancient language was called ‘temetum. Contrary to Agostiniani’s reasoning (1998: 9), however, this need not imply that tēmētum is the original designation for ‘wine’ in Latin and vīnum a term only secondarily introduced from outside, because in the passage given the word tēmētum was rather adduced for its antiquity and its apparent relationship to abstēmius, but not because it was a designation used exclusively in former times instead of the later current expression vīnum; and it has to be stresssed that Lat. vīnum is already attested as early as Livius Andronicus, Ennius, Plautus, and Cato, i.e. in the earliest layer of literary texts, and appears as a base in old derivatives like vīnea f. ‘vineyard, rows of vine’, attested since the xii Tabulae, Plautus, and Cato. Besides, the transmission of lexical antiquities certainly does not go back to a time horizon beyond the 6th c. bc; in other words, it cannot reflect an epoch in which the lexeme vīnum was not yet in existence in the Italic languages. This reasoning is even consistent with Agostiniani’s own analysis, according to which this lexeme spread from Etruscan as a loanword within Italic, including Latino-Faliscan, from the end of the 8th c. bc (see above §5.6). The situation might in some respect be comparable to Greek,

The Word for wine

211

where both οἶνος ‘wine’ and μέθυ ‘intoxicating drink, wine’ were inherited and already attested in Mycenaean and in both the Iliad and the Odyssey, but where μέθυ had fallen out of use by the classical epoch (see above §2). 5.7 Not acceptable is the derivation by Gamkrelidze and Ivanov (1995: 557) of Lat. vīnum from a supposedly inherited Italic form *u̯ ei̯no- < IE *u̯ ei̯(h1)-no-, because late Lat. ī (ca. 200 bc) < *ei̯ cannot be responsible for the attestations with = /ī/ occurring in Ancient Italy since the middle of the first millennium bc like Falisc. UINO etc. (see above §§5.1–2). Even if we assume an early borrowing from Latin at the stage of an older phonetic shape *u̯ ei̯nom or start from *u̯ ei̯nom as the Common Italic form underlying the attestations in the individual Italic languages, in Faliscan, Umbrian and Volscian this form would not lead to the attested form /vīnom/, but to *u̯ ēnom (cf. Falisc. hec = Lat. hīc, OLat. heic < loc. *gʰei̯-ke, Umbr. eetu /ētu/ = Lat. ītō < imptv. *ei̯-tōd; see Beekes 1987: 22 with reference to Blümel 1972: 33–34 for the representation of diphthongs in the Italic languages). 5.8 As a result of the historically identifiable spread of viticulture to the North, Lat. vīnum was borrowed in turn into Celtic as the source of OIr. fín, Welsh gwin and into Germanic as the source of Goth. wein, OE wīn, OHG wīn, which as a borrowing underlies OCS vino and Lith. vỹnas (IEW 1121). On the basis of compounds such as Goth. weina-gards, OE wīn-geard, OHG wīn-garto ‘vineyard’ and OCS vinjaga ‘grape’, vinogradъ ‘vineyard’, Gamkrelidze and Ivanov (1995: 557–558 with nn. 59, 60) suppose, however, that these are inherited forms from IE with their first compound member being genetically identical with Lat. vīnum, but not borrowed, and thus going back to IE *u̯ ei̯no- or alternatively *u̯ īno-. However, these Germanic and Slavonic compound forms are certainly not of Proto-Germanic or Proto-­Slavonic origin, but merely Common Germanic and Common Slavonic formations respectively and therefore young enough to contain word forms of the individual language branches, which are based on borrowings from Lat. vīnum. 6

Reconstruction

6.1 The evidence of the forms for ‘wine’ and ‘grapevine’ in the individual IE languages allows (in modification of the basic analysis by Beekes 1987: 24–25) the

212

Lipp

reconstruction of an amphikinetic nomen agentis (genus commune = masc./ fem.) of the type *spéḱ-on- ‘scout, spy, lookout’ (> YAv. spasan-, OHG speho): PIE *u̯ éi̯h1-on- (nom.sg. *u̯ éi̯h1-ō(n))/obl. *u̯ ih1-n-́/loc. *u̯ ih1-én ‘twiner, creeping plant, tendril, grapevine’ from the verbal root *u̯ ie̯ h1- ‘to wind around, to twine around’. (For PIE origin of the athematic n-stem meaning ‘grapevine’, see also Gorton 2014 with nn. 74–92; 2017: 2–8, 17–19.) The base form of the strong stem **u̯ ié̯ h1-on- → *u̯ éi̯h1-on- shows Schwebeablaut before a vowel-initial suffix. For prevocalic Schwebeablaut of a root morpheme containing a resonant in the onset and ending in a laryngeal, cf. inj.aor. *bʰ(u)u̯ eh2-s/t > OLat. subj. (nē) fuās, fuat ‘you shall / (s)he shall (not) be’ against subj.aor. *bʰéu̯ h2-e-ti > pres. Ved. bhávati ‘(s)he will be, becomes’; *pleh1- ‘fill (oneself)’ in Gr. πίμπλημι ‘I fill’, πλήθω ‘I fill myself, become full’, Lat. plēnus ‘full’ against *pólh1u/*pélh1-u- ‘fullness’ ntr. > Gr. πολύ ntr. → πολύς ‘much’, Goth. filu ntr. ‘much’. Cf. lipp 2009b: 150–151 n. 62 with further refs.

A lexicalisation with similar semantics is to be seen in the verbal noun *u̯ ié̯ h1ti-/*u̯ ih1-téi̯- ‘the twining (around)’ → ‘twining twig, tendril, flexible branch’ = Lat. vītis ‘tendril, grapevine’ (by poetic metonymy ‘wine’), Lith. vytìs ‘willow rod’. In proportional analogy to the zero grade *u̯ ih1-, a phonotactically renewed full grade *u̯ ei̯h1- corresponding to the normal root type CeR(H)- was introduced in YAv. vaēiti- ‘willow rod’ < *u̯ éi̯h1-ti- etc. (on the paradigmatic conditioning of the reshaped apophony, cf. Neri 2017: 324–325 with further refs.); perhaps the related form *u̯ éi̯h1-on- ‘twiner, grapevine’ (with prevocalic Schwebeablaut) also exerted some analogical influence. 6.2 It is possible that the type of nomina agentis like *spéḱ-on-, *u̯ éi̯h1-on- results from the hypostasis of locatives with postposition *-en, thus *spéḱ-on- ‘scout, lookout’ = ‘being at looking out’ from loc. *spéḱ-en or *speḱ-én ‘at looking out’ (from a root noun *spéḱ-/*speḱ-́ ‘looking out’); hence *u̯ éi̯h1-on- ‘twiner, creeping plant’ = ‘being at twining, creeping around’ from loc. *u̯ ih1-én ‘at twining, creeping around’ (from a root noun *u̯ ié̯ h1-/*u̯ ih1-́ ‘the twining, creeping around’; see below §9.2). Under this perspective, the phonotactically regularised full grade of the root in *u̯ éi̯h1-on- would be due to its derivation from the zero-grade-containing form of the loc. *u̯ ih1-én (the locative also providing the basis for the nominal stem *u̯ ih1-én-, nom.sg. *u̯ ih1-ḗn ‘(stock) being at the grapevine’ > Proto-Greek *u̯ ii̯ḗn > Gk. υἱήν Hesychius = *ϝιήν ‘vine-stock’; see below §8). For the type of derivation, cf. loc. *dhǵhm-én ‘on the earth’ (> Ved. jmán) → *dhǵhém-on-/*dhǵhm-n- ‘human being’ = ‘earthling, being on the earth’

The Word for wine

213

(> Lat. homō, -inis etc.; cf. Nussbaum 1986: 187–190, 289–291, Lipp 2009b: 115– 132 with further refs., Lipp forthcoming: §9.1.1). A similar derivation has been proposed by Neri (2017: 324 n. 181), who takes Hitt. u̯ ii̯an- ‘wine’ as the substantivised continuant of the adjectival hypostasis *u̯ ih1-on- ‘belonging to the grapevine’ from the locative *u̯ ih1-én ‘at the grapevine’ (belonging to a root noun *u̯ ói̯h1-/*u̯ ih1-́ ‘twining plant’ = ‘willow; ­grapevine’). Apart from the fact that the root had the primary shape *u̯ ie̯ h1forming a verifiable root noun *u̯ ié̯ h1-/*u̯ ih1-́ ‘the wrapping’ (see below §§9.1, 9.2), one has to keep in mind, however, that the primary meaning of Hitt. u̯ ii̯anand Luw. u̯ ii̯an-ī-/u̯ in-ī- (with ī-mutation in nom. + acc.) < *u̯ ih1-on- was ‘grapevine’, the meaning ‘wine’ in the case of Hitt. u̯ ii̯an(-a)- being secondary due to metonymy or hypostasis of the genitive (‘that of the grapevine’) or conditioned by derivation in the case of Luw. u̯ in-ī- (obl.) ‘wine’ < u̯ in-ii̯a- ‘pertaining to a grapevine’ (see above §§1.1, 1.1.1, 1.2 and below §7.1). 7

Morphological Representation of the Word for ‘Grapevine’ in Anatolian

7.1 PIE *u̯ éi̯h1-on-/loc. *u̯ ih1-én/obl. *u̯ ih1-n-́ led by generalisation of the zero grade of the root to Pre-Proto-Anatolian *u̯ íh1-on-/*u̯ ih1-én(-i)/*u̯ ih1-n-́ > Proto-­ Anatolian *u̯ íi̯on-/*u̯ ii̯én(-i)/*u̯ ī̆n- ‘grapevine’ > Hittite u̯ ii̯an- ‘grapevine’ and ‘wine’, the latter either by metonymy as *‘product from grapevine’ or by hypostasis from gen. u̯ ii̯an-aš ‘(that) of the grapevine’ displaying a thematic stem u̯ ii̯an-a-* in the meaning ‘wine’ (the meaning ‘wine’ in Hittite then being restricted to the derivative; but the existence of a separate derivative u̯ ii̯an-a-* ‘wine’ cannot be ascertained due to the ambiguity of the paradigmatic forms attested, see above §§1.1, 1.1.1, 1.1.3). Luwian u̯ ii̯an-ī-/u̯ in-ī- ‘grapevine’ either displays Luw. ī-mutation in nom. + acc. of the genus commune on the basis of the strong stem (partially with Luw. contraction ii̯a > i), or alternatively continues a hypostasis in the meaning ‘vine-stock’ = ‘(stock) being at the grapevine, (plant) consisting of grapevine’ from the locative *u̯ ii̯éni ‘at/in the grapevine’ > Luw. u̯ ii̯ani (with é > Luw. a). 7.2 For the Luwian formation, one can compare the ī-mutation in an athematic stem like nom.acc. PIE *dʰéǵʰ-ōm ‘earth’ (→ Post-PIE *dʰǵʰṓm > Gr. χθών etc.) > Proto-Anatolian *dégom (genus commune; Hitt. = /dḗgan/ ntr. secondary) → Pre-Proto-Luwian *deγom-ī́+s/m (with animate-marker *-ī́ in genus

214

Lipp

commune as relic of *-ih2/*-i̯éh2- corresponding to the type Ved. dev-ī́/gen. dev-yā́s f. ‘goddess’) > nom. *dei̯om-ī́-s, acc. *dei̯om-ī́-m (intervocalic loss of g > γ with insertion of i̯ as hiatus filler after front vowel), whence by analogy dat.loc. *dei̯om-í > Proto-Luwian nom. *dii̯amm-ī́-s, acc. *dii̯amm-ī́-n, dat.-loc. *dii̯amm-í (raising of pretonic e to i; fortis realisation resp. gemination of the nasal before following accent) > CLuw. nom. tii̯ammiš, acc. tii̯ammin, dat.-loc. tii̯ammi = /dii̯ammī́s/, /dii̯ammī́n/, /dii̯ammí/. But for this Luwian word for ‘earth’, there is also the alternative analysis as a locative hypostasis (Kimball 1994a: 78, 84–85 with n. 29), thus loc. PIE *dʰǵʰém(-i) > Proto-Anatolian *dəgém(i) > 1. loc. Hitt. tagān = /dagāń / ‘on the earth/ground’, 2. locative PreProto-Luwian *dəγém(i) > *dəi̯ém(mi) (palatalization of velar or rather intervocalic loss of g > γ with insertion of i̯ as hiatus filler before front vowel; nasal gemination according to Lex Čop) > Proto-Luwian *dii̯ám(mi) ‘on the earth/ ground’ → hypostasis CLuw. tii̯ammi- = /dii̯ámmi-/ c. ‘earth, land, ground’ (= *‘being located on the earth’?). (See Lipp 2009b: 56–60 with further refs.). 7.3 For the Anatolian forms, Oettinger (2003: 143–144) offers however the following hypothesis based on the assumption that their sound shape was conditioned by Anatolian vowel breaking: PIE *u̯ ói̯h1no- ‘wine’ (= Greek οἶνος m.) > Proto-Anatolian *u̯ ēno- > Hitt. u̯ ii̯ana- (breaking); *u̯ ói̯h1no- ‘wine’ → *u̯ ói̯h1ni‘grapevine’ > Proto-Anatolian *u̯ ēni- > Luw. *u̯ īni- > HLuw. u̯ ii̯ani- (breaking), by restitution also u̯ īni-. But in this scenario it remains unclear under which conditions the breaking of ē and ī (as against their preservation in other lexemes and paradigms) and the partial restitution of inherited ī take place (no conditioning by accent, chronological or dialectal distribution being recognizable; for the material, see above §§1.1, 1.2). On the other hand, vowel breaking must indeed be admitted as an irregular phenomenon, e.g. CLuw. mii̯asa- ntr. ‘flesh’ (nom.acc. mi-i̯a-ša-an-za = /mii̯asantsa/) beside misa- ntr. ‘id.’ = /mīsa-/ (nom.acc. mi-i-ša-an-za, mi-ša-an-za = /mīsantsa/) < *mēs-a- < PIE *mḗs ntr. < **mḗm-s ~ PIE *mēms-ó- (Ved. mā́s- ~ māṁs-á-, Goth. mimz ntr. etc.; interpretation and etymological analysis of CLuw. mi(i̯a)sa- according to Poetto 1995: 30, 34–35; for the attestations, cf. also Melchert 1993: 149, ACLT s.v. CLuw. mi(ya)sa- ntr. ‘flesh’). But one has to keep in mind that the Hittite form of the word for ‘grapevine’ is represented by the athematic n-stem u̯ ii̯an-, and the Hittite word for ‘wine’ is either identical with it due to metonymic extension of the denotation or represents a thematic hypostasis u̯ ii̯an-a-* based on u̯ ii̯an(the underlying stem formation of the attested paradigmatic forms being ambiguous, see above §§1.1, 1.1.1, 1.1.3). For the explanation of the Anatolian forms, one therefore has to start from an inherited athematic base form *u̯ ii̯an- in the

The Word for wine

215

meaning ‘grapevine’ (not ‘wine’), and this form cannot be derived from a thematic form *u̯ ói̯h1no- corresponding to Gr. οἶνος ‘wine’ and containing preconsonantal *oi̯ in the root morpheme, which would provide the starting point for the alleged Anatolian phonological development consisting of the monophthongization of the diphthong to *ē (> *ī) and subsequent vowel breaking to *ii̯a (and †u̯ oi̯(h1)-n- as an independent athematic stem of the genus commune is neither morphologically nor apophonically admissible). But in this respect H.C. Melchert (p.c.; similarly already Gérard 2009: 20) has pointed out as a decisive argument that before coronal continuants, the inherited diphthong *oi̯ was not monophthongized in Anatolian, but according to Kimball was retained as ai̯ in Hittite (as exemplified for the position before n or tautosyllabic s, just as also inherited *ou̯ and *au̯ were not monophthongized before resonants; Kimball 1994b: 14–22, 13–14, Kimball 1999: 216, 225–226; followed by Melchert 1994: 148 with the assumption of a broader context of preservation of the respective diphthongs before coronal continuants). Therefore, the presumed *oi̯ cannot provide the monophthongization product *ē as input for the vowel breaking to *ii̯a. For the precoronal retention of ai̯ < *oi̯, cf. Hitt. LÚ kaina- ‘relative by marriage’ < PIE *ḱói̯-no- (cf. Ved. śéva- ‘friendly, dear’, OHG hīwa, Latv. siẽva ‘wife’ < *ḱéi̯-u̯ o-/u̯ eh2-, OIr. coím ‘dear’ < *ḱói̯-mo-); Hitt. šuu̯ aiš ‘bird’ < *su̯ oi̯-s (with shortening by Osthoff’s law) < PIE *su̯ ṓi̯-s, nom. of *(s)h2u̯ ói̯-/*(s)h2u̯ éi̯- (Ved. váy-, nom. vés, Lat. avis). 7.4 Hitt. /u̯ ii̯an-/ c. ‘wine; grape-vine’ < Proto-Anatolian *u̯ íi̯on- < *u̯ íh1-on- with gen.sg. Hitt. GEŠTIN-i̯a-na-aš = u̯ i5-i̯a-na-aš < *u̯ íi̯on-os, instr.sg. Hitt. ú-i-i̯a-anta < *u̯ íi̯on-d̥ shows generalization of the stem formant *-on- after the strong stem of the amphikinetic paradigm Proto-Anatolian *u̯ íi̯on-/*u̯ ī̆n-́ < *u̯ íh1-on-/ *u̯ ih1-n-́. A comparable paradigmatic structure is offered by Hitt. nom.sg. ḫāraš, acc.sg. ḫāranan → gen.sg. ḫāranaš ‘eagle’ (for the Hittite paradigm, cf. Hoffner and Melchert 2008: 111–112) with apophonic levelling within the original amphikinetic paradigm PIE *h3ér-on- : obliquus *h3r̥-n-́ (nom.sg. PIE *h3ér-ō(n) = *[h3ór-o˜¯] > Proto-Anatolian *h3óro+s > Hitt. ḫāraš). The original ­apophony of the stem formant is reflected by OHG aro ‘eagle’ < Proto-­Germanic *ar-o˜¯ < PIE *h3ér-ō(n) : OHG arn < *arn-i-, ON ǫrn < *arn-u- ‘id.’ < Proto-­Germanic *ar-n- ← *ur-n- < PIE *h3r̥-n-́. (For the apophonic structure of the paradigm of *h3ér-on‘eagle’, see Harðarson 1987: 118–122 and also Melchert 1983: 8–9, 10, Zucha 1988: 170–171, 172–173.) The original PIE paradigm underlying Hitt. /u̯ ii̯an-/ c. ‘wine; grapevine’ thus was neither protero- nor hysterokinetic. This also becomes clear from the diverging paradigm structure in the following two examples.

216

Lipp

(1) Hitt. išḫiman-/išḫimen- c. ‘cord, rope’ with nom.sg. išḫimāš, acc.sg. išḫimanan → išḫimenan (on the analogy of the oblique stem as in unattested gen.sg. išḫimenaš*; for the Hittite paradigm, seeō Hoffner and Melchert 2008: 111–112) on the basis of a proterokinetic paradigm *sh2éi̯mon- : obl. *sh2i-mén- with nom.sg. *sh2éi̯-mō(n) = *[sh2ái̯-mō ]̃ → ­Proto-Anatolian *sh2i-mṓ+s > Hitt. išḫimāš (= OSax. sīmo, OE sīma ‘rope’ < Proto-Germanic *sīmo˜¯ < *sih2-mo˜¯ with laryngeal metathesis and probably analogical suffix accent after acc.sg. *sih2-món-m̥ < *sh2i-món-m̥ ). For the stem formation, cf. Gr. ἱμονιά ‘well-rope’ < *simon-i̯ā, due to a remodelling in the acc.sg. *sh2i-món-m̥ ← PIE *sh2ei̯-món-m̥ = *[sh2ai̯-món-m̥ ] < **sh2éi̯-mon-m̥ under the influence of the oblique stem PIE *sh2i-mén-, likewise displaying suffix accent. On the PIE accent shift in trisyllabic words with the vowel structure é-o-V → e-ó-V as in the case of the PIE numeral *ku̯ etu̯ óres m. ‘four’ < **ku̯ étu̯ ores, the so-called “ku̯ etu̯ óres rule” postulated by Klingenschmitt, cf. Lipp 2009b: 96 with n. 143 and further refs. For the apo­ phonic structure of the paradigm, see Harðarson 1987: 118–122, partially different Melchert 1983: 10, Zucha 1988: 170–171 with n. 28; 172–173.

(2) Hitt. ištanzan- c. ‘soul’ with nom.sg. ZI-an-za = ištanza*, gen.sg. ištanzanaš based on a hysterokinetic paradigm *pst-én- : obl. *pst-n-́ with nom.sg. *pst-ḗn > Proto-Anatolian *stḗn+s > Hitt. ištanza*. For the etymology, cf. Ved. stána- m., YAv. fštāna- m., Arm. stin ‘female breast’, Gr. στήνιον· στῆθος (Hesychius) ‘breast’. For the apophonic structure, see Harðarson 1987: 118–122 on the basis of the etymological analysis by Eichner 1973: 98 n. 78, which however has been disputed by HED i/ii 471, Kloekhorst 2008: 415, cf. EWAia ii 752.

8 The Inherited Greek Word for ‘Grapevine’ By hypostasis of the locative *u̯ ih1-én ‘at/in the grapevine/tendril’ in the prehistory of Greek there was formed a new nominal stem *u̯ ih1-én-, nom.sg. *u̯ ih1-ḗn ‘(stock) being at the grapevine, (plant) consisting of grapevine/tendril’ > ProtoGr. *u̯ ii̯ḗn > Gr. υἱήν, ἢ υἱόν· τὴν ἄμπελον (Hesychius), to be interpreted as *ϝιήν ‘grapevine, vine-stock’. For the internal derivation, cf. Ved. RV+ pári-jman- adj. ‘being all around, ubiquitous’ ← RV pári-jman adv. ‘all around’ < *pári ǰmán loc. ‘all around on earth’ (RV jmán loc. ‘on earth’). A comparable constellation of internal derivation based on a locative (without derivational suffix) can be seen in the s-stem *h2ép-os/

The Word for wine

217

*h2ep-és- ntr. ‘work, opus’ with loc. *h2ep-és(-i) ‘at work’ > Ved. ápas/apás‘work’ with loc. apási and in the adjectival hypostasis (derived from the locative) *h2ep-és- ‘being at work’ (nom.sg. *h2ep-ḗs), continued in the Ved. adjective apás- ‘working, active’. For the vocalism of the derivative, cf. Gr. ψευδής ‘lying, false’ as derivative of the s-stem ψεῦδος ntr. ‘lie, falsehood’ with dative ψεύδει < loc. *-és-i. See lipp 2009b: 120 with further refs., following j. schindler

9

Etymological Connections of the PIE Word for ‘Grapevine’

9.1 PIE *u̯ éi̯h1-on- ‘twiner, creeping plant, grapevine’ is a nomen agentis based on the verbal root *u̯ ie̯ h1- ‘to wind around, to twine around’ (LIV 2: 695), which is represented e.g. in the thematic aor. Ved. ávyat ‘has enwrapped’ ← root aor. 3 pl. *u̯ iá̯ nt, remodelled on the stem of root aor.sg. *u̯ ia¯̯´ - < PIE *u̯ ié̯ h1- from 3.pl. *u̯ ii̯ánt < PIE *u̯ ih1-ént, further in pres. Ved. vyáyati ‘enwraps’ ← *u̯ ii̯ái̯a- (after aor. vya- ← *u̯ iā̯ �-) = Lat. vieō, -ēre ‘to bind, twist’ < iterative *u̯ ih1-éi̯e/o- (noncoloration of the stem suffix proves *h1 and excludes *h2 as final root consonant: †u̯ ih2-éi̯e- > †u̯ ih2ai̯e- > Lat. †viāre), and in Lith. vejù (výti) ‘to wind’ < *u̯ éi̯h1-e/o- ← **u̯ ié̯ h1-e/o- (Schwebeablaut before vowel-initial suffix, see above §6.1). From Ved. vyáyati ‘enwraps’ ← PIE *u̯ ih1-éi̯e/o- one has to distinguish etymologically the phonologically and semantically similar verb in pres. Ved. váyati ‘weaves’ < PIE *Hu̯ -éi̯e/o-, perf. Ved. ūvur ‘they have woven’ (< *Hu-Hu̯ -ŕ̥s ← *He-Hu̯ -ŕ̥s), verbal adj. vy-ùta- ‘woven’ from *Heu̯ -, further Lith. áudžiu, áusti ‘to weave’ < *Heu̯ -d- (LIV 2: 224). 9.2 In Avestan, the related root noun viiā- is still attested occasionally in the basic (i.e. not derivationally conditioned) lexical meaning ‘wrapping, envelopment’. The acc.sg. OAv. viiąm is measured as bisyllabic (either /u̯ ia̯ 'am/ as reflex of PIE *u̯ ié̯ h1-m̥ or /u̯ ii̯ām/ as lexicalised Lindeman variant from PIE *u̯ (i)i̯ḗm < **u̯ ié̯ h1-m, if Stang’s law applied not only to **°ah2m > *°ām, but also to **°eh1m > *°ēm; on occasional reflexes of Sievers and Lindeman variants in Old Avestan, see Schindler 1977: 58–59, cf. Lipp 2009b: 357 n. 18): Yasna 48,7 b-c yōi ā vaŋhə̄uš manaŋhō dīdraγžō.duiiē // aṣ̌ā viiąm ‘you, who wish to tighten the envelopment of good thinking in accordance with the right order’. Further forms of the loc.pl. occur in Yašt 13,11 f-g viiāhuua uruuat̰.caēm astica gaonaca ‘in the wombs (= envelopments) I put together bones and hairs’; Yašt 8,9 b–d satauuaēsō…viiāhuua yat̰ jasaiti ‘when (the rain-bringing star) Satavaēsa goes

218

Lipp

to his hiding places (= shells, envelopments by the clouds)’ (cf. Humbach 1959: 78, Schindler 1972: 70, Kellens 1984: 56, 340, Kellens and Pirart 1988: 170; 1990: 308; 1991: 224, Humbach 1991: 201–202). In PIE this particular root noun was probably a nomen actionis with the meaning ‘wrapping, envelopment’, which displayed mobile accent and therefore the stem allomorphs *u̯ ié̯ h1-/*u̯ ih1-́ (i.e. not the acrostatic pattern †u̯ ió̯ h1-/†u̯ ié̯ h1- of a nomen agentis or resultativum), and therefore the Hesychius gloss Gr. γίς· ἱμάς ‘belt’ taken as *ϝίς = */u̯ ī́s/ < PreProto-Gr. *u̯ íh1-s ‘the wrapping’ (← PIE *u̯ ié̯ h1-s) can be connected as a formation with generalised zero grade of the root. 10

Morphological Representation of the Words for ‘Wine’ in Italic and Greek

10.1 Lat. vīnum, Falisc. /vīnom/, Umbr. /vīnom/ ‘wine’ etc. < Proto-Italic *u̯ īno- < PIE *u̯ ih1no- is derivationally analysable as *u̯ ih1-n-ó-, i.e. an original adjectival possessive derivation based on the oblique (weak) stem of PIE *u̯ éi̯h1-on-/obl. *u̯ ih1-n-́ ‘grapevine, tendril’ (stem variant as in Luw. /u̯ in-ī-/ ‘wine’ < u̯ in-ii̯a- < *u̯ īn-ii̯ó- ‘pertaining to a grapevine’; see above §1.2), which has the derivational meaning ‘having/containing grapevine, consisting of grapevine’ → neuter ‘Rebensaft: grape juice, must, wine’ (in Italic lexicalised as a noun). For possessive adjectives derived from the oblique stem, cf. Ved. tamasá- ‘dark-coloured’ < *temH-es-ó- ‘having darkness, consisting of darkness’ belonging to Ved. támas‘darkness’ ntr. < PIE *témH-os/*temH-és- ← *témH-os/*tm̥ H-és-, and Gr. βιός m. ‘bow’ < *gu̯ih2-ó- ‘having a tendon/string, provided with a tendon’ to Ved. jya¯´ - f. < *gu̯i̯éh2-/*gu̯ih2-́ ‘tendon’. For Italic possessive derivatives based on the oblique stem, cf. Lat. salvus ‘sound, unharmed’ = Osc. σαλαϝσ, SALAVS, Umbr. saluom, saluuom (< Proto-Sabellic *salau̯ o-) < Proto-Italic *salau̯ o- = OP fra-haravam adv. ‘entirely, altogether’ < Indo-Iranian *sər-au̯ -á- < *sr̥ H-au̯ -á- < PIE *[sl ̥h2au̯ ó-] = *sl ̥h2-eu̯ -ó- (accent like Gr. κενεός ‘empty’ < *ken-eu̯ ó-). These forms are etymologically related to CLuw. šalḫitti- ‘growth’ < *sl ̥h2-i-ti- to Hitt. šalli-, CLuw. šalḫa/i- ‘grown, adult’ ← adj. *sélh2-i-/*sl ̥h2-éi̯-. PIE *sl ̥h2-eu̯ -ó- in the meaning ‘having health/welfare’ is a possessive formation based on the oblique stem of the neuter noun *sólh2-u/obl. *sl ̥h2-éu̯ - ‘health, welfare’ (← *sólh2-u/obl. *sélh2-u-, originally belonging to the acrostatic type *ǵón-u/obl. *ǵén-u- → *ǵn-éu̯ - ‘knee’). This noun is continued 1. by the Italic noun *sal-u-/*sal-ou̯ - ← *sl ̥h2-u-/*sl ̥h2-au̯ - (← PIE *sólh2-u/*sl ̥h2-éu̯ -)

The Word for wine

219

as base of the secondary Lat. salūs, -ūtis ‘health, welfare’ < *salou̯ -ti-; and 2. by the Ved. adverb pra-salaví ‘entirely, altogether’ < Proto-Indo-Iranian loc. *sar-áu̯ -i belonging to *sár-u/*sar-áu̯ - < *sálH-u/*salH-áu̯ - < *sólh2u/*solh2-áu̯ - (← PIE *sólh2-u/*sl ̥h2-éu̯ -; see Lipp 2009b: 339–341 and nn. 64, 66 for refs.). 10.2 10.2.1 Gr. οἶνος m. ‘wine’, Myc. wo-no = /u̯ oi̯no-/ etc. < Proto-Gr. *u̯ ói̯no- < PIE *u̯ ói̯(h1)no- (laryngeal loss according to Lex Saussure, see above §2) is analysable in its derivation as *u̯ ói̯h1-n-o- ‘Rebensaft: grape juice, must, wine’ ← *u̯ oi̯h1n-ó- ‘having/containing grapevine, consisting of grapevine’, i.e. as an early substantivisation of an archaic adjectival possessive formation, characterised by o-grade of the root and initial accent. The adjectival possessive formation *u̯ oi̯h1-n-ó- is probably derived from the strong stem of PIE *u̯ éi̯h1-on-/obl. *u̯ ih1-n-́ ‘tendril, grapevine’ (by paradigmatic levelling Pre-Proto-Anatolian *u̯ ih1-on- > Hitt. u̯ ii̯an(a)- ‘grapevine; wine’); morphologically and functionally comparable is e.g. Goth. hals, Lat. collum, OLat. collus ‘neck’ < *kólso- (k < ku̯ by delabialisation before rounded vowel o) < PIE *ku̯ ólso- ‘joint for a turning motion’ ← *ku̯ olsó- (Lex Saussure: oR.C < oRH.C, see above §§2, 3) < *ku̯ olh1-s-ó‘having a turning motion, performing a turning motion’ as derivative of the verbal noun *ku̯ élh1-os/-és- ntr. ‘turning (motion)’ > Gr. τέλος ‘end’ = *‘turning point’. The archaic formation of the underlying possessive adjectives *u̯ oi̯h1n-ó- and *ku̯ olh1-s-ó- shows a morphonologically conditioned o-grade of the root morpheme, which was unaccented in the derivative, but accented in the derivational bases *u̯ éi̯h1-on- c. and *ku̯ élh1-os ntr. respectively. Such derivations might have provided a prototype for substantivisations with the vr̥ddhi vowel *-o- (for such analyses, see below §10.2.2). From these formations, one must however distinguish younger and more frequent possessive formations with e-grade of the root like possibly *ku̯ elh1-s-ó- ‘having a turning point, being at the turning point’ (> Welsh pell ‘far’, on which see Neri’s different analysis in the next paragraph); cf. the derivation from the base *ḱér-u- ‘horn, point’ (> Ved. śáru- f. RV ‘arrow, spear’) in the case of PIE *ḱer-u̯ -ó- ‘having horns, horned’ or *ḱér-u̯ -o- ‘animal having horns, horned animal’ > Lat. cervus ‘stag, deer’ beside a probably older and therefore already Pre-PIE formation *kor-u̯ -ó- ‘having horns, horned’ → PIE *kór-u̯ -o- ‘horned animal, cattle’ (early lexicalised form without later satem palatalisation like in the base form *ḱér-u- < **kér-u-) → Balto-Slavic *kṓru̯ -ā- f. ‘cow’ (vr̥ddhi derivation meaning *‘the female belonging to the horned animal’) > RussCS krava, Lith. kárvė (< *-ii̯ā-) ‘id.’ (on the formation type, see Lipp 2009a: 14 and n. 21; 60–61 and n. 161 with further refs.,

220

Lipp

Lipp forthcoming: §9.3.2; Lat. corvus ‘crow, raven’ could be related as *‘animal with horny beak’ from *kór-u̯ -o- ‘horned animal’, but in general an onomatopoetic origin is assumed for this form). For substantivisation with initial accent, cf. the derivations PIE *u̯ l ̥ku̯ ó‘dangerous’ (Ved. a-vr̥ká- ‘harmless’ = *‘undangerous’ and Hitt. u̯ alku(u̯ )a- ntr. ‘monstrosity’, if not from *u̯ élku̯ o- ‘the dangerous thing’) → PIE *u̯ ĺk�̥ u̯ o- ‘dangerous animal’ (Luw. u̯ alu̯ a-/u̯ alu̯ i- ‘lion’, Ved. vŕ̥ka-, Gr. λύκος etc. ‘wolf’), Ved. kr̥ ṣṇá- ‘black’ → kŕ̥ṣṇa- ‘black antelope’, *anná- ‘eaten’ → ánna- ‘food’, Gr. λευκός ‘white’ → λεῦκος ‘white fish’, λεύκη ‘abele, white poplar’, δολιχός ‘long’ → δόλιχος ‘long run’ (cf. Lipp 2009b: 144–145 with refs.). For the substantivisation of a possessive formation, cf. (as derivative of *rót-o- ‘the running, the rolling motion/device’ = OHG rad ‘wheel’ etc. → *rót-h2/*rot-éh2- ‘roller, wheel’ = Lat. rota ‘wheel’) the possessive adj. *roth2-ó- ‘having wheels’ → substantive *róth2-o- ‘chariot’ = Ved. rátha-, Av. raθa- ‘chariot’ (see Lipp 2009b: 387 n. 87; 453 with refs.). 10.2.2 A different derivation for Gr. οἶνος m. ‘wine’ < *u̯ ói̯(h1)no- is proposed by Neri (2017: 324 n. 181), who regards it as a barytone substantivisation with o-grade on the basis of a vṛddhi derivative *u̯ ei̯h1-n-ó- ‘belonging to the grapevine’, i.e. as a formation like the aforementioned (but in 10.2.1. differently analysed) form Goth. hals ‘neck’ < Germ. *h(w)alsa- < *ku̯ ól(h1)-s-o- ← *ku̯ elh1-s-ó- ‘having a turning motion/point’ > Welsh pell ‘far’ (= *‘having a turning point, being at the turning point’), and Lith. žãstas (2) ‘upper arm’ < *ǵhós-to- ← *ǵhes-tó- ‘belonging to the hand’. In Neri’s analysis, *u̯ ei̯h1-n-ó- is a thematic hypostasis of a locative *u̯ ih1-én ‘at the grapevine’ (→ *u̯ ih1-on- ‘belonging to the grapevine’ > Hitt. u̯ ii̯an- ‘wine’) from a root noun *u̯ ói̯h1-/*u̯ ih1-́ ‘twining plant’ = ‘willow; grapevine’ (the existence of which cannot be ascertained, however; see above §§6.2, 9.2). According to Gippert (1994: 119 n. 16), on the other hand, Gr. οἶνος m. ‘wine’ < *u̯ ói̯no- < *u̯ ói̯(H)no- continues a substantivisation with secondary o-grade vṛddhi in the meaning ‘grape’ = ‘the fruit belonging to the twining plant/grapevine’, which is derived from the verbal adjective *u̯ iH-nó- ‘twined’, whose neuter form meaning ‘twining plant, grapevine’ is continued by Lat. vīnum ‘wine’. The Armenian form gini, gen. ginwoy ‘wine’ < *u̯ ēnii̯o- < *u̯ oi̯(H)nio- is assumed to be an -io- derivative from *u̯ ói̯(H)no- ‘grape’ with the meaning ‘wine’ as designation for the product made from grapes. The Anatolian forms like Hitt. u̯ ii̯an(a)- ‘grapevine; wine’ would however be based on different derivations from the respective verbal root.

The Word for wine

11

221

The Semitic Evidence

The earliest attestations of Semitic *u̯ ai̯nu ‘wine’ are represented by the Old Cannaanite forms i̯e-nu, i̯e-ni (14th c. bc) from Akkadian diplomatic texts in the Egyptian archive of Tell el-Amarna. In the textual tradition of the respective Semitic languages there are the following attestations: Ugaritic yn = /yēnu/ ‘wine’ (13th c. bc; cf. Aistleitner 1967: 130–131 Nr. 1183, del Olmo Lete and Sanmartín 2003: 968–971, for the retrieval of the forms, see Zemánek 1995: esp. 121–122), Phoenician yn = /yēn/, Old Aramaic yyn, Hebrew yyn, vocalised yayin (Samaritan Ostraca yn = /yēn/), Ammonite yn = /yēn/ ‘wine’ (each of these forms showing the Northwest Semitic change of initial *u̯ - > i̯-); Old Sabaean wyn, yyn ‘vineyard’ (with vacillation between u̯ - and i̯-); Ethiopic (Geʻez) wayən ‘wine, grapevine, vineyard’; further only occasionally attested Arabian wayn ‘black grapes’. (On the Semitic forms and their possible interdependence with IE and non-IE forms, cf. Brown 1969: 147–148, Koehler and Baumgartner 1974: 391, Cohen 1996: 534, Lipiński 1997: 561, Zamora 2000: 269, 270, 274, Leslau 2006: 623, Aspesi 2013, Gorton 2014 with nn. 28–38, 105–111; assumption of a borrowing from Hittite by Rabin 1963: 138–139, unconvincing attempt at an etymological connection of the lexeme within Semitic by van Selms 1974, which is ­rejected by Aspesi 2013: 6 with n. 22, Gorton 2014 with n. 37). The lexeme in question is represented only in West Semitic, not in East Semitic, as the traditionally connected form Akkad. inu appears only in Young Babylonian lexicographical lists and as such is not old and also of unknown meaning (CAD 7: 152 s.v. inu C, 157 s.v. īnu 1.(e)l, Rabin 1963: 138 n. 8 versus AHw I 383). Since ancient times, the normal expression in East Semitic for ‘wine’ has been Akkad. karānu (also ‘grape, grapevine’), often rendered by the Sumerian logogram GEŠTIN. And because of the phonotactic structure of its potential root element (*u̯ in̯ ), *u̯ ai̯nu is also without an inner-Semitic etymological connection. But also as a borrowing it would not be derivable from the Anatolian forms Hitt. u̯ ii̯an(a)- ‘wine; grapevine’, HLuv. u̯ ii̯ani-, u̯ ini- ‘grapevine’ (< *u̯ ih1on-/*u̯ ih1-n-; see above §§1.1, 1.1.1, 1.2, 7.1). The following hypothesis, however, might offer a solution: The lexeme *u̯ ai̯nu could have been borrowed into Northwest Semitic (Ugaritic + ­Aramaic + Canaanite [= Phoenician, Hebrew, Moabitic, Ammonite]) around 1500 bc by way of the Levant trade from Mycenaean, where in the textual evidence of Linear B the word is attested as wo-no = /u̯ oi̯no-/ ‘wine’ (> later Greek οἶνος; see above §2). As according to the evidence from Amorite names in Akkadian and Egyptian sources and from Old Syrian and Old Canaanite glosses in the correspondence from Tell el-Amarna (cf. Moscati 1964: 46) the Northwest Semitic initial

222

Lipp

change *u̯ - > i̯- had already taken place by the first quarter of the second millennium bc, one can suppose that the lexeme *u̯ ai̯nu as a loan form, which was taken over from Mycenaean only in the middle of the second millennium bc, would first have preserved its initial *u̯ - in Northwest Semitic and as such would soon have spread from the Northwest Semitic area further within the Semitic group to the south (into North and South Arabian, Ethiopic). As by the earlier Northwest Semitic sound change *u̯ - > i̯- all former instances of initial *u̯ - had been removed, it is easily conceivable that after some time a secondary sound substitution of new isolated cases of initial *u̯ - would have occurred as in the loanword *u̯ ai̯nu ‘wine’, which therefore within Northwest Semitic changed secondarily for structural reasons into *i̯ai̯nu. Thus only in specific phonetic constellations or still later acquired lexemes was initial u̯ preserved in Northwest Semitic, as for instance in the case of the proclitic conjunction *u̯ a- ‘and’ > Ugarit.Phoen. w-, Hebr.Aram. wə- corresponding to Arab. wa- and Akkad. u < *u̯ a- (cf. Moscati 1964: 46, 121, Sivan 2001: 25–26). 12

Conclusion

To sum up: In historical linguistic analysis, the set of forms of the word for ‘wine’ represented in the individual Indo-European languages can be derived from a formation based on the PIE root *u̯ ie̯ h1- ‘to wind around, to twine around’, i.e. a nominal nasal stem PIE *u̯ éi̯h1-on-/obl. *u̯ ih1-n-́/loc. *u̯ ih1-én ‘twiner, creeping plant, tendril, grapevine’ (with Schwebeablaut of the full grade of the root morpheme before the vowel-initial suffix). This athematic stem is continued on the one hand by Hittite u̯ ii̯an- c. ‘wine; grapevine’ (whether the meaning ‘wine’ was restricted to a thematic hypostasis u̯ ii̯an-a-* cannot be decided in view of the ambiguity of the paradigmatic forms attested) and Hieroglyphic Luwian u̯ ii̯ani-, u̯ ini- c. ‘grapevine’ (with generalized zero grade of the root). On the other hand, it provides the derivational base for the possessive formations 1. *u̯ oi̯(h1)-n-ó- ‘consisting of grapevine’ (from the strong stem *u̯ éi̯h1-on-) → noun *u̯ ói̯n-o- ‘juice from grapevine, grape juice, Rebensaft’ > Greek (ϝ)οἶνος m. ‘wine’, Armenian gini ‘wine’ (< *u̯ oi̯n-ii̯o-m ntr. < *u̯ oi̯n-ih2ó- ← *u̯ ói̯n-o-) and 2. *u̯ ih1-n-ó- ‘consisting of grapevine’ (from the weak stem *u̯ ih1n-́) → Proto-Italic *u̯ īnom ntr. ‘juice from grapevine, grape juice, Rebensaft’ > Latin vīnum, Faliscan /vīnom/, Umbrian /vīnom/ ‘wine’; under this assumption the Etruscan form VINUN, VINUM continues a loan adopted from an Italic language no later than the 7th century bc. A debatable alternative is offered by Agostiniani’s analysis, according to which the Italic forms are borrowed from the Etruscan language, which in turn would have taken over this lexeme from

The Word for wine

223

Greek, i.e. Common Italic /u̯ īnom/ ← Etruscan VINUN, VINUM ← accusative Greek u̯ oi̯non*. The Semitic forms Ugaritic yn = /yēnu/, Phoenician yn = /yēn/, Hebrew yayin ‘wine’, etc. < *u̯ ai̯nu may be based on a borrowing in the mid-2nd millennium bc from Mycenaean Greek (Linear B wo-no = /u̯ oi̯no-/), whereas the source of the Caucasian forms Georgian γvino, Mingrelian gvini ‘wine’ is probably a loan acquired during the 1st millennium bc from Proto-Armenian or its prestage (Proto-Kartvelian *γu̯ ini̯o ← [Pre-]Proto-Armenian *gu̯ īníi̯o, gen. *gu̯ īni̯ói̯o < *u̯ oi̯n-ii̯o- = Armenian gini, gen. ginwoy). Acknowledgements This paper presented at the Hrozný memorial conference in Prague in November 2015 is based on a joint talk given with Bela Brogyanyi at the iii Convegno irdau in Perugia and Gubbio in September 2011 (Brogyanyi and Lipp 2016). In the present version, special attention has been paid to the Anatolian and Italic evidence. For valuable advice on the formation and attestation of the words for ‘grapevine’ and ‘wine’ in Anatolian I would like to thank Craig Melchert and Massimo Poetto, who have also generously provided me with important bibliographical references. Further references I owe to Marianna Pozza. I am also indebted to Ilya Yakubovich for making accessible precious archival material, and to Luke Gorton for putting his unpublished PhD thesis at my disposal (Gorton 2014). For advice on Semitic issues, I would like to thank my department colleagues Viktor Bielický and Pavel Čech. I would also like to acknowledge the utmost care taken by the volume editor Ronald Kim in processing my ­typographically complex text and in improving the style of the paper. For substantial suggestions and support in providing research literature, I am grateful to my friends Bela Brogyanyi, Martin Gális and Sergio Neri. The research for this paper has been supported by the Czech Science Foundation in the form of ­projects GA 14-10673S and GA 14-12987S and by Charles University in Prague within the project ‘Progres 4: Language in the shiftings of time, space, and culture’. Bibliography ACLT = I. Yakubovich (principal investigator), Annotated Corpus of the Luwian Language, http://web-corpora.net/LuwianCorpus/search/ (accessed 14.05.2016). Agostiniani, L. (1998) Sull’origine del nome del vino in etrusco e nelle altre lingue dell’Italia antica, in: L. Agostiniani et al. (eds.), do-ra-qe pe-re. Studi in memoria di

224

Lipp

Adriana Quattordio Moreschini, Roma: Istituti editoriali e poligrafici internazionali, 1–13. AHw I = Wolfram von Soden (1965) Akkadisches Handwörterbuch, Band I: A–L, Wiesbaden: Harrassowitz. Aistleitner, J. (1967) Wörterbuch der ugaritischen Sprache, hg. von Otto Eißfeldt, Berlin: Akademie-Verlag (Dritte, durchgesehene und ergänzte Auflage). Aspesi, F. (2013 [2012–2013]) Sacro vino, Ἀλεξάνδρεια/Alessandria 6–7 (= ΦΙΛΟΙΝ – ­Scritti in onore di Mario Enrietti e Renato Gendre), 3–17. Bakkum, G.C.L.M. (2009) The Latin Dialect of the Ager Faliscus: 150 Years of Scholarship, 2 parts, Amsterdam: Amsterdam University Press. Beekes, R.S.P. (1987) On Indo-European ‘wine’, MSS 48, 21–26. Blümel, W. (1972) Untersuchungen zu Lautsystem und Morphologie des vorklassischen Lateins, München: Kitzinger. von Brandenstein, C.-G. (1943) Hethitische Götter nach Bildbeschreibungen in Keilschrifttexten [MVAeG 46], Leipzig: Hinrichs. Brogyanyi, B. and Lipp, R. (2016). Wein im Anatolischen, Griechischen, Italischen und Indogermanischen, in: A. Ancilotti, A. Calderini and R. Massarelli (eds.), Forme e strutture della religione nell’Italia mediana antica – Forms and Structures of Religion in Ancient Central Italy. III Convegno internazionale dell’Istituto di Ricerche e Documentazione sugli Antichi Umbri, 21–25 settembre 2011, Roma: «L’Erma» di Bretschneider, 65–77. Brown, J.P. (1969) The Mediterranean vocabulary of the vine, Vetus Testamentum 19, 146–170. Buck, C.D. (1928) A Grammar of Oscan and Umbrian, with a Collection of Inscriptions and a Glossary, reprinted with additions and corrections, Boston: Athenaeum Press (reprinted 1979, Hildesheim: Olms). CAD 7 = I.J. Gelb, B. Landsberger and A.L. Oppenheim (eds.) (1960) The Assyrian ­Dictionary of the Oriental Institute of the University of Chicago, Volume 7: I and J, Chicago: Oriental Institute – Glückstadt: J.J. Augustin. Casaretto, A. (2004) Nominale Wortbildung der gotischen Sprache. Die Derivation der Substantive, Heidelberg: Winter. Cohen, D. (ed.) 1996. Dictionnaire des racines sémitiques ou attestées dans les langues sémitiques. ii. fasc. 6: W – WHLP. Leuven: Peeters. Demiraj, B. (1997) Albanische Etymologien (Untersuchungen zum albanischen Erbwortschatz), Amsterdam, Atlanta: Rodopi. DELL = Ernout, A. and Meillet, A. (1967) Dictionnaire étymologique de la langue latine. Histoire des mots, Paris: Klincksieck (Quatrième édition, deuxième tirage augmenté de corrections nouvelles). DMic. = Aura Jorro, F. (1985–1993) Diccionario griego-español, anejo ii: Diccionario micénico (DMic.), volumen I – volumen ii, bajo la dirección de F.R. Adrados, Madrid: Consejo superior de investigaciones científicas.

The Word for wine

225

Eichner, H. (1973) Die Etymologie von heth. mehur, MSS 31, 53–107. ET = Rix, H. (ed.) (1991) Etruskische Texte. Editio minor, Band I: Einleitung, Konkordanz, Indices, Band ii: Texte, Tübingen: Narr [revised edition: Meiser, G. (ed.) (2014) Etruskische Texte. Editio minor, Teil 1: Einleitung, Konkordanz, Indices, Teil 2: Texte, Hamburg: Baar]. EWAia ii = Mayrhofer, M. (1996) Etymologisches Wörterbuch des Altindoarischen, ii. Band, Heidelberg: Winter. Flobert, P. (1994) Les débuts de la vigne et du vin en Italie et en Gaule d’après le vo­cab­ ulai­re, Bulletin de la Société Nationale des Antiquaires de France 1992.1, 289– 301. Friedrich, J. (1974) Hethitisches Elementarbuch, 1. Teil: Kurzgefaßte Grammatik, Heidelberg: Winter. Gamkrelidze, T.V. and Ivanov, V.V. (1995) Indo-European and the Indo-Europeans: A ­Reconstruction and Historical Analysis of a Proto-Language and a Proto-Culture, Part I: The Text (English version by J. Nichols, edited by W. Winter), Berlin, New York: Mouton de Gruyter. Gérard, R. (2009) Sur l’origine du nom hittito-louvite *wiyana- « vin », in: C. Cannuyer, A. Degrève and R. Gérard (eds.), Vin, bière et ivresse dans les civilisations orientales. Entre plaisir et interdit. René Lebrun in honorem [Acta Orientalia Belgica 22], Louvain-la-Neuve: Société Belge d’Études Orientales, 15–21. Gippert, J. (1994) Die Glottaltheorie und die Frage urindogermanisch-kaukasischer Sprachkontakte, in: J.E. Rasmussen (ed.), In honorem Holger Pedersen. Kolloquium der Indogermanischen Gesellschaft vom 25. bis 28. März 1993 in Kopenhagen, Wiesbaden: Reichert. 107–123. Gorton, L. (2014) Through the Grapevine: Tracing the Origins of Wine, Ph.D. dissertation, Ohio State University. Gorton, L. (2017) Revisiting Indo-European “wine”, Journal of Indo-European Studies 45, 1–26. Greppin, J.A.C. (2008) Did the good wines come from Armenia?, Aramazd: Armenian Journal of Near Eastern Studies 3, 47–52. Harðarson, J.A. (1987) Das uridg. Wort für „Frau“, MSS 48, 115–137. Hawkins, J.D. (1995) The Hieroglyphic Inscription of the Sacred Pool Complex at Hattusa (SÜDBURG), with an archaeological introduction by Peter Neve, Wiesbaden: Harrassowitz. Hawkins, J.D. (2000) Corpus of Hieroglyphic Luwian Inscriptions, Volume I: Inscriptions of the Iron Age, Part 1: Text. Introduction, Karatepe, Karkamiš, Tell Ahmar, Maraș, Malatya, Commagene [UISK 8.1], Berlin, New York: de Gruyter. Hoffner, H.A., Jr. and Melchert, H.C. (2008) A Grammar of the Hittite Language, Part 1: Reference grammar, Winona Lake, IN: Eisenbrauns. Humbach, H. (1959) Die Gathas des Zarathustra, Band ii: Kommentar, Heidelberg: Winter.

226

Lipp

Humbach, H. (1991) The Gāthās of Zarathushtra and the other Old Avestan texts, in collaboration with J. Elfenbein and P.O. Skjærvø, Part ii: Commentary, Heidelberg: Winter. Kellens, J. (1984) Le verbe avestique, Wiesbaden: Reichert. Kellens, J. and Pirart, E. (1988) Les textes vieil-avestiques, Volume I: Introduction, texte et traduction, Wiesbaden: Reichert. Kellens, J. and Pirart, E. (1990) Les textes vieil-avestiques, Volume ii: Répertoires grammaticaux et lexique, Wiesbaden: Reichert. Kellens, J. and Pirart, E. (1991) Les textes vieil-avestiques, Volume iii: Commentaire, Wiesbaden: Reichert. Kimball, S.E. (1994a) Loss and retention of voiced velars in Luvian: another look, IF 99, 75–85. Kimball, S.E. (1994b) The IE short diphthongs *oi, *ai, *ou and *au in Hittite, Die Sprache 36, 1–28. Kimball, S.E. (1999) Hittite Historical Phonology [IBS 95], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Kloekhorst, A. (2008) Etymological Dictionary of the Hittite Inherited Lexicon [LIEEDS 5], Leiden, Boston: Brill. Koehler, L. and Baumgartner, W. (1974) Hebräisches und aramäisches Lexikon zum Alten Testament, Lieferung ii, Leiden: Brill (Dritte Auflage neu bearbeitet von W. Baumgartner). Kölligan, D. (2013) Greek and Latin elements in Oscan personal names, Linguarum V ­ arietas 2, 119–124. Laroche, E. (1961) Études de toponymie anatolienne, RHA 19, 57–98. LDIA = Giacomelli, G. (1978) Il falisco, in: Popoli e civiltà dell’Italia antica. Volume sesto a cura di Aldo L. Prosdocimi: Lingue e dialetti, Roma: Biblioteca di storia patria, 505–542. Lejeune, M. (1987) Le vase de Latumaros (Discussions sur l’alphabet de Lugano), Latomus 46, 493–509. Leslau, W. (2006) Comparative Dictionary of Geʿez (Classical Ethiopic): Geʿez-English – English-Geʿez, with an Index of the Semitic Roots, Wiesbaden: Harrassowitz. LEW ii = Walde, A. and Hofmann, J.B. (1972) Lateinisches Etymologisches Wörterbuch, Zweiter Band, Heidelberg: Winter (Fünfte Auflage). LF [with number of inscription] = Giacomelli, G. (1963) La lingua falisca, Firenze: Olschki. Lipiński, E. (1997) Semitic Languages: Outline of a Comparative Grammar [Orientalia Lovaniensia Analecta 80], Leuven: Peeters. Lipp, R. (2009a) Die indogermanischen und einzelsprachlichen Palatale im Indoiranischen, Band I: Neurekonstruktion, Nuristan-Sprachen, Genese der indoarischen Retroflexe, Indoarisch von Mitanni, Heidelberg: Winter.

The Word for wine

227

Lipp, R. (2009b) Die indogermanischen und einzelsprachlichen Palatale im Indoiranischen, Band ii: Thorn-Problem, indoiranische Laryngalvokalisation, Heidelberg: Winter. Lipp, R. (forthcoming) The Proto-Indo-European *-r/n- stem suppletion and the locative of heteroclitic neuters, in: R.I. Kim (ed.), Diachronic Perspectives on Suppletion, Hamburg: Baar. Martirosyan, H.K. (2010) Etymological Dictionary of the Armenian Inherited Lexicon [LIEEDS 8], Leiden, Boston: Brill. Matzinger, J. (2005) Untersuchungen zum altarmenischen Nomen. Die Flexion des Substantivs, Dettelbach: Röll. McGovern, P.E. (2003) Ancient Wine: The Search for the Origins of Viniculture, Princeton, Oxford: Princeton University Press. McGovern, P.E., Fleming, S.J. and Katz, S.H. (eds.) (1996) The Origins and Ancient History of Wine, London, New York: Routledge (digital printing 2006). McGovern, P.E., Glusker, D.L., Exner, L.J. and Voigt, M.M. (1996) Neolithic resinated wine, Nature 381, 480–481. Meiser, G. (1986) Lautgeschichte der umbrischen Sprache [IBS 51], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Melchert, H.C. (1983) A ‘new’ PIE *men suffix, Die Sprache 29, 1–26. Melchert, H.C. (1993) Cuneiform Luvian Lexicon [Lexica Anatolica 2], Chapel Hill: self-published. Melchert, H.C. (1994) Anatolian Historical Phonology [LSIE 3], Amsterdam, Atlanta: Rodopi. Melchert, H.C. and Oettinger, N. (2009) Ablativ und Instrumental im Hethitischen und Indogermanischen: ein Beitrag zur relativen Chronologie, IncLing 32, 53–73. del Monte, G.F. and Tischler, J. (1978) Die Orts- und Gewässernamen der hethitischen Texte [TAVO Beihefte B 7; RGTC 6/1], Wiesbaden: Reichert. Moscati, S. (ed.) (1964) An Introduction to the Comparative Grammar of the Semitic Languages: Phonology and Morphology, Wiesbaden: Harrassowitz. Neri, S. (2003) I sostantivi in -u del gotico. Morfologia e preistoria [IBS 108], Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Neri, S. (2013) Zum urindogermanischen Wort für ‚Hand‘, in: A.I. Cooper, J. Rau and M. Weiss (eds.), Multi Nominis Grammaticus. Studies in Classical and Indo-European Linguistics in Honor of Alan J. Nussbaum on the Occasion of his Sixty-Fifth Birthday, Ann Arbor, New York: Beech Stave, 185–205. Neri, S. (2017) Wetter: Etymologie und Lautgesetz, Università degli Studi di Perugia: Centro Stampa Morlacchi. Neumann, G. (1962) Beiträge zum Lykischen ii, Die Sprache 8, 203–212. Nussbaum, A.J. (1986) Head and Horn in Indo-European [UISK 2], Berlin, New York: de Gruyter.

228

Lipp

Oettinger, N. (2003) Zum Ablaut von n-Stämmen im Anatolischen und der Brechung ē > ya, in: E. Tichy, D.S. Wodtko and B. Irslinger (eds.), Indogermanisches Nomen: Derivation, Flexion und Ablaut. Akten der Arbeitstagung der Indogermanischen Gesellschaft, Freiburg, 19. bis 22. September 2001, Bremen: Hempen, 141–152. del Olmo Lete, G. and Sanmartín, J. (2003) A Dictionary of the Ugaritic Language in the Alphabetic Tradition, translated by W.G.E. Watson. 2 vols. [HdO I/67], Leiden, Boston: Brill. Olsen, B.A. (1999) The Noun in Biblical Armenian: Origin and Word-Formation. With Special Emphasis on the Indo-European Heritage, Berlin, New York: Mouton de Gruyter. Otten, H. and Souček, V. (1969) Ein althethitisches Ritual für das Königspaar, Wiesbaden: Harrassowitz. Perpillou, J.-L. (2012) La plante à vin, in: A. Blanc, L. Dubois and C. de Lamberterie (eds.), Πολύμητις. Mélanges en l’honneur de Françoise Bader, Leuven, Paris: Peeters, 109–114. Poetto, M. (1993) L’iscrizione luvio-geroglifica di Yalburt. Nuove acquisizioni relative alla geografia dell’Anatolia sud-occidentale, Pavia: Iuculano. Poetto, M. (1995) Luvio mi(ya)sa- nell’àmbito dell’interpretazione di KUB 35.45 ii 22–24, HS 108, 30–38. Prosdocimi, A. (1984) Le tavole iguvine. I + Album delle tavole, Firenze: Olschki. Puhvel, J. (2012) Elliptic genitives and hypostatic nouns in Hittite, in: J. Puhvel, Ultima Indoeuropaea [IBS 143], Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck, 122–126. Rabin, C. (1963) Hittite words in Hebrew, Orientalia N.S. 32, 113–139. Rieken, E. (1999) Untersuchungen zur nominalen Stammbildung des Hethitischen [StBoT 44], Wiesbaden: Harrassowitz. Rix, H. (1995) Il latino e l’etrusco, in: Atti del convegno internazionale “Nomen latinum” – Latini e Romani prima di Annibale, Roma 24–26 ottobre 1995 [Eutopia 4.2], Roma: Quasar, 73–88. Rix, H. (1998) Teonimi etruschi e teonimi italici, Annali della Fondazione per il museo “Claudio Faina” 5, 207–229. Rix, H. (2008) Etruscan, in: R.D. Woodard (ed.), The Ancient Languages of Europe, Cambridge: Cambridge University Press, 141–164. Rüster, C. and Neu, E. (1989) Hethitisches Zeichenlexikon. Inventar und Interpretation der Keilschriftzeichen aus den Boğazköy-Texten, Wiesbaden: Harrassowitz. Schindler, J. (1972) Das Wurzelnomen im Arischen und Griechischen, Ph.D. dissertation, Universität Würzburg. Schindler, J. (1977) Notizen zum Sieversschen Gesetz, Die Sprache 23, 56–65. Schmitt, R. (1981) Grammatik des Klassisch-Armenischen mit sprachvergleichenden E­ rläuterungen [IBS 32], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. (2., durchgesehene Auflage, 2007.)

The Word for wine

229

Schulze, W. (1966) Zur Geschichte lateinischer Eigennamen, Berlin, Zürich, Dublin: Weidmann (2. unveränderte Auflage). van Selms, A. (1974) The etymology of yayin, “wine”, JNSL 3, 76–84. Sihler, A.L. (1995) New Comparative Grammar of Greek and Latin, New York, Oxford: Oxford University Press. Silvestri, D. (1974) URUU̯ ii̯anau̯ anda: un problema di toponomastica anatolica, SSL 14, 266–274. de Simone, C. (2011) Il nome del “vino” nei dialetti dell’Italia antica, in: La vigna di ­Dioniso: Vite, vino e culti in Magna Grecia. Atti del quarantanovesimo convegno di studi sulla Magna Grecia, Taranto, 24–28 settembre 2009, Taranto: Istituto per la Storia e l’Archeologia della Magna Grecia, 473–479 [interventi: 545]. Sivan, D. (2001) A Grammar of the Ugaritic Language [HdO I/28], Atlanta: Society of Biblical Literature – Leiden: Brill (second impression with corrections). Solinas, P. (1995 [1994]) Il celtico in Italia, Studi Etruschi 60, 311–408. ST = Rix, H. (ed.) (2002) Sabellische Texte: Die Texte des Oskischen, Umbrischen und Südpikenischen, Heidelberg: Winter. Starke, F. (1990) Untersuchung zur Stammbildung des keilschrift-luwischen Nomens [StBoT 31], Wiesbaden: Harrassowitz. Szemerényi, O. (1989) An den Quellen des lateinischen Wortschatzes [IBS 56], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. TI = Tabulae Iguvinae; current editions: Prosdocimi 1984 and ST Um1. Tibiletti Bruno, M.G. (1981) Le iscrizioni celtiche d’Italia, in: E. Campanile (ed.), I celti d’Italia, Pisa: Giardini, 157–207. Uhlich, J. (2007) More on the linguistic classification of Lepontic, in: P.-Y. Lambert and G.-J. Pinault (eds.), Gaulois et celtique continental, Genève: Librairie Droz, 373–411. de Vaan, M. (2008) Etymological Dictionary of Latin and the Other Italic Languages [LIEEDS 7], Leiden, Boston: Brill. Yakubovich, I. (2015) The Luwian Language. In: Oxford Handbooks Online, http://www .oxfordhandbooks.com/view/10.1093/oxfordhb/9780199935345.001.0001/oxfordhb -9780199935345-e-18 (accessed 21.10.2015). Zamora, J.-Á. (2000) La vid y el vino en Ugarit, Madrid: Consejo superior de investigaciones científicas. Zemánek, P. (1995) Ugaritischer Wortformenindex [Lexicographia Orientalis 4], Hamburg: Buske. Zgusta, L. (1984) Kleinasiatische Ortsnamen. Heidelberg: Winter. Zucha, I. (1988) The Nominal Stem Types in Hittite, Ph.D. dissertation, Oxford University.

Chapter 11

Satzanfänge im Hethitischen Rosemarie Lühr Abstract Am Satzanfang können im Hethitischen metatextuelle Ausdrücke und informationsstrukturelle Einheiten auftreten: Frames, Topics, Contrastive Topics, Foci. Hinzu kommen Ausdrücke für Gegebenheiten der Sprechsituation, Anreden, Sprecher und ­Adressat, Anredefloskeln. Wackernagel-Elemente, Pronomina, die “Reflexivpartikel” -za, “Lokalpartikeln,” Partikeln der zitierten Rede, tragen darüber hinaus zur Gliederung des Satzes bei. Weshalb das Hethitische so viele Wackernagel-Elemente hat, könnte auf eine ursprünglich inkrementelle Syntax weisen.

Keywords Anredefloskel – Contrastive Topic – Focus – Frame – gesprochene Sprache – inkrementelle Syntax – Partikel der zitierten Rede – Reflexivpartikel – Topic – Wackernagel-Element

Die altindogermanischen Sprachen besitzen keine dem deutschen Vorfeld, Mittelfeld und Nachfeld entsprechende Feldgliederung. Daher ist eine Frage, wie diese Sprachen die solchen Feldern zukommenden informationsstrukturellen Aufgaben erfüllen.1 Dabei können Wackernagel-Elemente ein Gliederungssignal im Satz sein. Eine altindogermanische Sprache, die für diese ­Positionen umfangreiches Material bietet, ist das Hethitische. Das Hethitische hat unter den altindogermanischen Sprachen nicht nur die meisten Elemente 1 Krisch (2007) postuliert auch für das Altindische eine Gliederung des Satzes nach Vorfeld, Mittelfeld und Nachfeld: Das Nachfeld beginnt hinter dem finiten Verb in Endposition, das Mittelfeld besteht aus dem Scrambling-Bereich, und das Vorfeld umfasst gegebenenfalls zwei topikalisierte Konstituenten, die erste mit Wackernagel-Element. Auch ein Teil einer Konstituente, die durch einen “Regenerierungsprozess” zu einer vollständigen Phrase werden kann (Riemsdijk 1989; Krisch 1998: 374; Fanselow 2004), erscheint vor der Wackernagel-Position. Eine inhaltliche Bestimmung der Elemente am Satzanfang nimmt Speyer (2009) für das Lateinische, Germanische und Griechische vor. Er konstatiert in dieser Position eine Bevorzugung von “Rahmenbildnern,” “Listenelementen” und “Themen,” sofern sie explizit sind. © koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_013

Satzanfänge im Hethitischen

231

in der Wackernagelposition, auch ist die Anordnung in der Partikelkette grammatikalisiert. Daneben gibt es im Hethitischen aber auch Satzanfänge ohne Wackernagel-Elemente. Die Frage, ob auch bei den Sätzen ohne Wackernagel-Elemente bestimmte einigermaßen regelmäßig an der Satzspitze stehen, ist das eine, um das es im Folgenden geht. Die andere Frage betrifft die Elemente in den Wackernagel-Positionen selbst. 1

Satzanfänge ohne Wackernagel-Elemente

Unter den Satzanfängen ohne Wackernagel-Elemente gibt es metatextuelle Ausdrücke, informationsstrukturelle Einheiten und Bezeichnungen von Gegebenheiten der Sprechsituation. 1.1 Metatextuelle Ausdrücke Das Korpus, das hauptsächlich aus dem Gebet Muwatalli ii. an die Götterversammlung (CTH 381; jh.), dem Erlass des Telipinu (CTH 19), einem Gerichtsprotokoll (CTH 293; jh.) und Ritualen (CTH 443; 447) besteht, enthält nur ganz wenige Sätze ohne Wackernagel-Elemente. Ein Beispiel mit einem Textadverb ist (1). ki-iš-ša-an “folgendermaßen” steht am Anfang des Satzes und dient der metatextuellen Gliederung: (1) Muwatallis Gebet (CTH 381), Vs. i 1–2 UM-MA ta-ba-ar-na mNIR.GÁL folgendermaßen Tabarna: NOM.SG Muwatalli URU LUGAL.GAL LUGAL KUR Ḫa-at-ti Großkönig König Land Stadt Ḫatti  U M-MA ta-ba-ar-na mNIR.GÁL LUGAL.GAL LUGAL KUR URUḪa-at-ti … “Folgendermaßen spricht der Tabarna Muwatalli, der Großkönig, König des Landes der Stadt Ḫatti …”2 1.2 Informationsstrukturelle Einheiten Auch informationsstrukturelle Einheiten erscheinen am Satzanfang. Unterschieden wird nach New Information Focus und Contrastive Focus, wobei zum 2 Textadverbien kommen aber auch in der Mitte des Satzes vor, z.B. Muwatallis Gebet (CTH 381), Rs. iii 73–74: nu ḫa-an-da-an ú-uk ki-iš-ša-an me-ma-al-lu und getreu ich:nom folgendermaßen sprechen: PRS.IMP.ACT1SG “Getreu möge ich folgendermaßen sprechen.”

232

Lühr

Focus auch die Contrastive Topics gezählt werden, dann nach New Aboutness Topic oder Shifting Topic und Non-New Aboutness Topic und nach den Kategorisierungen der Centering-Theorie mit den Relationen Continue, Retain, Smooth Shift und Rough Shift. Hinzu kommen Frames. 1.2.1 Frames Frames sind propositionsbezogen. Als frame-setting-Adjunkte geben sie den Rahmen an, in dem der Satz wahr ist (Chafe 1976). (2)(a) In Deutschland bin ich weltberühmt. (H. Juhnke) (Maienborn 2001) (2)(b)  Im Falle eines Sieges wird die Mannschaft eine Extrazulage erhalten.3 Darüber hinaus wird durch den Frame ein Wechsel des Standpunkts angezeigt. Im Deutschen befinden sich solche scene-setting-Elemente, sofern sie referentiell sind und kontextuelle Bindung vorliegt, im Vorfeld; ebenso erscheinen Frames im Hethitischen am Satzanfang. Man findet lokale, temporale und modale Adverbialia mit Autosemantika, Deiktika und Temporalsätzen. Formen ohne Wackernagel-Partikeln sind wieder nur vereinzelt. Für einen lokalen Frame mit einer lokativischen Fügung vgl.: (3) Telipinu-Erlass (CTH 19), Vs. i 19–20

a-pí-e-el-l= a ŠU-I URU DIDLI.ḪI.A GAL.GAL TIM dieser: GEN.SG und Hand: LOC.SG Stadt: PL groß: PL ti-it-ti-i̯a-an-te-eš e- še-ir zuordnen: PTCP.NOM.C.PL sein: IPF.IND.ACT3PL a-pi-e-el-la ŠU-i URU DIDLI.ḪI.A GAL.GALTIM ti-it-ti-i̯a-an-te-eš e- še-ir “Und in dessen Hand wurden die großen Städte zugeordnet.”

Für temporale Frames mit Adverbien bzw. einem Temporalsatz: (4) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), I 29–30

ka-ru-ú li-in-ku-un früher schwören: IPF.IND.ACT1SG ka-ru-ú li-in-ku-un “Ich habe früher geschworen.”

3 Der Gültigkeitsbereich der Proposition wird auch eingeschränkt in: Arbeitsmäßig geht es mir gut. (domain adjunct).

233

Satzanfänge im Hethitischen

(5) Muwatallis Gebet (CTH 381), Rs. iv 59

nam-ma ŠA IṢ-ṢI 2 GUNNI dann von Holz: GEN.SG 2 Feuerstelle PRS.IND.ACT3PL nam-ma ŠA IṢ-ṢI 2 GUNNI DÙ-an-zi “Dann machen sie 2 Feuerstellen mit Holz.”

DÙ-an-zi machen:

(6) Telipinu-Erlass (CTH 19), Vs. i 39–40

URU [ma-a-an mḪ]a-an-ti-li-iš-š= a Ta-ga-ra-ma a-a[r-aš als Ḫantili: nom und Stadt Tagarama: loc ankommen: IPF.IND.ACT3SG nu me-mi-iš-ki-]u-wa-an [da-a-iš] und sprechen: sup beginnen: IPF.IND.ACT3SG [ma-a-an mḪ]a-an-ti-li-iš-ša URUTa-ga-ra-ma a-a[r-aš nu me-mi-iš-ki-]u-wa-an [da-a-iš] “Und als Ḫantili in Tagarama ankam, begann er zu sprechen.”

Für modale Frames mit einer präpositionalen Fügung bzw. Adverb vgl.: (7) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Vs. ii 12

[QA-D]U ku-úr-da-a-li i-da-a-la-mu-uš zusammen mit kurtali-Behälter(N): INSTR.SG böse: ACC.C.PL EME ḪI.A ar-ḫa šu-uḫ-ḫa-ir Zunge: PL weg schütten: IPF.IND.ACT3PL [QA-D]U ku-úr-da-a-li i-da-a-la-mu-uš EME ḪI.A ar-ḫa šu-uḫ-ḫa-ir “Zusammen mit dem kurtali-Behälter schütteten sie die bösen Zungen weg.”

(8) Muwatallis Gebet (CTH 381), Rs. iii 73–74

nu ḫa-an-da-an ú-uk ki-iš-ša-an me-ma-al-lu und getreu ich: nom folgendermaßen sprechen: PRS.IMP.ACT1SG nu ḫa-an-da-an ú-uk ki-iš-ša-an me-ma-al-lu “Getreu möge ich folgendermaßen sprechen.”

1.2.2 Topics Gegenüber den häufigen aus enklitischen Personalpronomina bestehenden Topics in der Wackernagel-Domäne sind Topics an der ersten Satzposition selten. Belege finden sich für das deiktische Demonstrativpronomen “dieser.” In (9) und (10) erscheint die nominale Fügung mit diesem Pronomen als Shifting Topic in einer Rough-Shift-Relation, d.h.: Es ist kein Bezug auf das Subjekt des vorausgehenden Satzes gegeben, sondern das Subjekt ist neu eingeführt.

234

Lühr

(9) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Vs. i 33–34 nu ki-e i-da-a-la-u-e-eš al-wa-an-zi-in-ni-eš und dieser: NOM.C.PL böse: NOM.C.PL zauberkräftig: NOM.C.PL

eme[ḪI.A wa-ḫ]a-an-du BE-L[Í] QA-DU Zunge: PL sich wenden: PRS.IMP.ACT3PL Herr: DAT zusammen mit DAM- -ŠU DUMU MEŠ- -ŠU É- -ZU Ehefrau sein Sohn: PL sein Haus sein  nu ki-e i-da-a-la-u-e-eš al-wa-an-zi-in-ni-eš eme[ḪI.A wa-ḫ]a-an-du BE-L[Í] QA-DU DAMŠU DUMU MEŠ-ŠU É-ZU “Diese bösen zauberkräftigen Zungen sollen sich zugunsten des Herrn zusammen mit seiner Ehefrau, seinen Söhnen und seinem Haus wenden.”

Der Satz davor lautet: DUTU-uš DUTU-wa-aš DINGIR.LÚ MEŠ qa-a-ša-aš-ma-aš kar-di-mi-i̯a-at-ta-aš ḫ[ar-]ši-in pí-iḫ-ḫu-un “Oh Sonnengott, oh männliche Götter des Sonnengottes, siehe, ich habe euch das Dickbrot des Zornes gegeben.” (10) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Rs. iii 20’ [nu ki-i al-wa-an-za-tar(?) EGIR-pa fZi

und dieser: NOM.N.SG Zauberei(N): NOM.SG wieder Zi QA-DU] DUMU MEŠ- -ŠU ḫar-du zusammen mit Sohn: PL sein halten: PRS.IMP.ACT3SG [nu ki-i al-wa-an-za-tar(?) EGIR-pa fZi QA-DU] DUMU MEŠ-ŠU ḫar-du “Diese Zauberei halte wieder die Zi zusammen mit ihren Söhnen.”

Der vorausgehende Kontext ist: [ku-u-un SÍGpít-tu-u-la-an ma-a-aḫ-ḫa-an tuḫ-ša-an-n]a-aḫ-ḫu-un Ù ŠA fZi [i-da-a-lu ut-tar al-wa-an-za-tar ar-ḫa QA-TAM-MA tu]ḫ-ša-an e-eš-tu “Wie ich diese Schnur abgeschnitten habe, ebenso sei auch das böse Wort (und) die Zauberei von Zi abgeschnitten.” Ohne Demonstrativpronomen tritt die definite Nominalphrase als Subjekt in einer Smooth-Shift-Relation in (11) auf:

Satzanfänge im Hethitischen

235

(11) Telipinu-Erlass (CTH 19), Vs. i 12

[(nu URU DIDLI.ḪI.A GAL.GALTIM [(t)]i-it-ti-i̯a-an-te-eš und Stadt: PL groß: PL zuordnen: PTCP.NOM.C.PL e- šir sein: IPF.IND.ACT3PL [(nu URUDIDLI .ḪI.A GAL.GALTIM [(t)]i-it-ti-i̯a-an-te-eš e-šir “Und die großen Städte wurden (ihnen) zugeordnet.”

Smooth-Shift bedeutet einen Subjektswechsel gegenüber dem Subjekt im vorausgehenden Kontext. Die Ortsbezeichnungen werden durch “die großen Städte” aufgenommen: Voraus geht: (11)(a) Telipinu-Erlass (CTH 19), Vs. i 10  [(URUḪ)]u-u-piš-na URUTu-u-u̯ [(a-an)]u-u̯ a URUNe-na-aš-ša URULa-a-an-da URUZa-al- la-ra

[(URU)]Pár-šu-ḫa-an-ta URULu[(-u-u)]š-na nu ut-ne-e ma-ni-i̯a-aḫ-ḫi-eš-ki-ir “In Hupišna, Tuwanuwa, Nenašša, Lansa, Zallara, Paršuḫanta, Lušna verwalteten sie das Land.”

1.2.3 Contrastive Topics Des Weiteren kommt der eine Teil von zwei Contrastive Topics an der Satzspitze ohne Wackernagel-Partikel vor: (12) Muwatallis Gebet (CTH 381), Vs. i 5–6

URU 1 GIŠBANŠUR ANA dUTU TÚL-na 1 Tisch zu Sonnengöttin Stadt Arinna GIŠ Ù DINGIR MEŠ LÚ MEŠ [1 BA]NŠUR und Gott: PL Mann: PL 1 Tisch da-a-i setzen: PRS.IND.ACT3SG 1 GIŠBANŠUR A-NA dUTU URUTÚL-na Ù DINGIR MEŠ LÚ MEŠ [1 GIŠBA]NŠUR da-a-i “Einen Tisch stellt er für die Sonnengöttin der Stadt Arinna auf und für die männlichen Götter einen Tisch.”

Nach Lang (1984, 1991) sind solche Topics einer gemeinsamen Einordnungsinstanz, einem Common Integrator, hier Ritualgegenständen, zuzuordnen. Die kontrastive Lesart wird dabei durch ein folgendes kontrastiertes Element ­signalisiert, “für die Sonnengöttin der Stadt Arinna.” Es ist ein Contrastive

236

Lühr

F­ ocus; darauf erscheint als weiterer Contrastive Focus “für die männlichen Götter” und schließlich der zweite Teil des Contrastive Topic “einen Tisch.” Den New Information Focus bildet dann da-a-i “stellt auf.” Im Deutschen haben Sätze mit einem Contrastive Topic vor einem folgenden ebenfalls hervorgehobenen Wort ein spezifisches Tonmuster, weshalb ­Jacobs (1997) von einem Intonational Topic spricht. In der Tat werden Sätze mit Intonational Topics als zweigipflige (Hut-)Kontur wahrgenommen, während man in Sätzen mit neutralen Topics lediglich eine Hervorhebung auf dem Fokusexponenten hört. Bei dem steigenden Akzent der Hutkontur handelt es sich um einen Kontrastakzent, der zusammen mit dem Kontext die Interpretation von Alternativen auslöst. Wie an anderer Stelle bemerkt (Lühr 2008), dürften solche Sätze auch in altindogermanischen Sprachen artikuliert worden sein. Im Hethitischen haben weiterhin Sätze mit der kontrastiven Partikel -ma “aber” wohl eine Intonationskontur: (13) Muwatallis Gebet (CTH 381), Rs. iii 64–67 nu NINDA.GUR 4.RA iš-pa-an-du-uz-zi-

-i̯a [ku-i]n A-NA und dickes Brot Spende(N): ACC.SG und welcher: ACC.C.SG zu dU pí-ḫa-aš-ša-aš-ši EN- -I̯A pé-eš-ki-mi Wettergott zum Blitz gehörig: DAT.SG Herr mein geben: ITER.PRS.IND.ACT1SG na- -an- -ši und er: ACC.C.SG er: DAT.SG du-uš-ga-ra-u-wa-an-za pí-iš-ke-el-lu píd-du-li-i ̯a-u-wa-an-za freudig: NOM.C.SG geben: ITER.PRS.IMP.ACT1SG zögernd: NOM.C.SG -ma- -da le-e pé-eš-ki-mi aber du: DAT nicht geben: ITER.PRS.IND.ACT1SG  nu NINDA.GUR 4.RA iš-pa-an-du-uz-zi-i̯a [ku-i]n A-NA dU pí-ḫa-aš-ša-aš-ši EN-I̯A pé-eški-mi na-an-ši du-uš-ga-ra-u-wa-an-za pí-iš-ke-el-lu píd-du-li-i̯a-u-wa-an-za-ma-da le-e pé-eš-ki-mi “Das dicke Brot und die Spende, die ich dem zum Blitz gehörigen Wettergott, meinem Herrn, stets gebe, das gebe ich ihm freudig, nicht aber soll ich es dir zögernd geben.”

Gibt man den zweiten Satz der Satzreihe in der deutschen Wortstellung ­wieder – “zögernd aber soll ich es dir nicht geben” – mit “zögernd” als Contrastive Topic, ergibt sich wieder eine Hutkontur. píd-du-li-i̯a-u-wa-an-za “zögernd” steht dem vorausgehenden Adjektiv du-uš-ga-ra-u-wa-an-za “freudig” gegenüber, wobei der Common Integrator für beide Adjektive das Konzept ­“psychische Disposition” ist. Der Contrastive Focus besteht aus le-e “nicht,” der New Information Focus aus pé-eš-ki-mi.

237

Satzanfänge im Hethitischen

1.2.4 Foci Während die Position des New Information Focus dem Satzende zu ist, können Contrastive Foci an der Satzspitze erscheinen. Tabelle 11.1.a

Focus an erster position

Focus an 1. Position

Gerichtsprotokoll

Muwatalli

Ritual 443

insgesamt

1

10

10

Tabelle 11.1.b

Telipinu Ritual 447

5

4

Focus gegen Satzende zu

Focus final

Gerichtsprotokoll

Muwatalli

Ritual 443

insgesamt

23

51

81

Telipinu Ritual 447 41

41

Zum Vergleich die Anzahl der annotierten Texteinheiten und Tokens:

Muwatalli Ritual 443 Telipinu Ritual 447

Texteinheiten

Tokens

177 163 48 105

2015 1445 378 835

In folgendem Beleg ist das Temporaladverb a-pí-a “dann” mit der additiven ­Fokuspartikel ya- verknüpft: (14) Muwatallis Gebet (CTH 381), Rs. iii 44

nu a-pí-i̯a- -i̯a dU pí-ḫa-aš-ša-aš-ši-in und dann auch Wettergott zum Blitz gehörig: ACC.C.SG šar-li-iš-ki-mi preisen: ITER.PRS.IND.ACT1SG nu a-pí-i̯a-i̯a dU pí-ḫa-aš-ša-aš-ši-in šar-li-iš-ki-mi “Und auch dann preise ich (immerfort) den zum Blitz gehörigen Wettergott.”

Die Partikel setzt die kontrastiv betonte Konstituente “in eine quantifizierende Beziehung zu typgleichen Konstituenten” (Thurmair 1989: 17). “Auch dann”

238

Lühr

i­ mpliziert “immer,” wie es auch in dem iterativen Verb šar-li-iš-ki-mi “ich preise immerfort” zum Ausdruck kommt. Fungiert -ya als Fokuspartikel, verbindet es sich mit dem Contrastive Focus, wo immer dieses Wort im Satz steht. -ya ist also keine Wackernagel-Partikel (Hoffner and Melchert 2008: 401). 1.3 Sprechsituation 1.3.1 Anrede Entitäten der Sprechsituation sind Anreden. Vokative sind an der Satzspitze bezeugt: (15) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Vs. ii 43–44 [iš-ḫa-na-aš DUTU-]uš DIŠKUR-aš

Blut(N): GEN.SG Sonnengott(C): NOM.SG Wettergott(C): NOM.SG qa-a-ša- -aš-ma-aš -ta     B[E-L]U QA-DU DAM siehe euch dort Herr zusammen mit Ehefrau MEŠ -ŠU DUMU - -ŠU [ga-an-kán-za(?)] sein Sohn: PL sein abwiegen: PTCP.NOM.C.SG [iš-ḫa-na-aš DUTU-]uš DIŠKUR-aš qa-a-ša-aš-ma-aš-ta B[E-L]U QA-DU DAM-ŠU DUMU MEŠ-ŠU [ga-an-kán-za(?)] “Oh Sonnengott des Blutes, oh Wettergott, siehe, für euch dort ist der Herr zusammen mit seiner Ehefrau und seinen Söhnen abgewogen.”

1.3.2 Sprecher und Adressat Weitere Entitäten der Sprechsituation sind Sprecher und Adressat; vgl. das Pronomen der 2. Person hinter einem Vokativ. (16) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Rs. iii 50 a-iš EME-aš ga-ga-aš Mund(N): NOM.SG Zunge(C): NOM.SG Zahn(C): NOM.SG šu-me-eš az-zi-ki-te-en ihr: NOM essen: ITER.PRS.IMP.ACT2PL a-iš EME-aš ga-ga-aš šu-me-eš az-zi-ki-te-en “Oh Mund, Zunge, Zahn, ihr esst.”

1.3.3 Anredefloskeln Ferner erscheinen Anredefloskeln am Satzanfang; vgl. ka-a-ša “siehe”:

239

Satzanfänge im Hethitischen

(17) Muwatallis Gebet (CTH 381), Rs. iii 18–19 nu ka-a-ša am-mu-uk MNIR.GÁL und siehe ich: NOM Muwatalli LUGAL-uš LÚSANGA ŠA dUTU König(C): NOM.SG Priester von Sonnengöttin URU TÚL-na Ù DINGIR MEŠ ḫu-u-ma-an-da-aš Stadt Arinna und Gott: PL all: GEN.PL ne-pí-ša-aš dUTU-i ar-ku-iš-ki-mi Himmel(N): GEN.SG Sonnengott(C): DAT.SG bitten: ITER.PRS.IND.ACT1SG nu ka-a-ša am-mu-uk MNIR.GÁL LUGAL-uš LÚSANGA ŠA dUTU URUTÚL-na Ù DINGIR MEŠ ḫu-u-ma-an-da-aš ne-pí ša-aš dUTU-i ar-ku-iš-ki-mi “Siehe, ich, Muwatalli, der König und Priester der Sonnengöttin der Stadt Arinna und aller ­Götter, bitte den Sonnengott des Himmels.”

Und ḫa-an-da-an “wahrlich” mit Quotativpartikel: (18) Muwatallis Gebet (CTH 381), Rs. iii 57–58

ḫa-an-da-an- -wa DINGIRLIM šar-ku-uš wahrlich QUOT Gott: NOM.SG erhaben: NOM.C.SG UR.SAG-iš pa-ra-a ḫa-an-da-a[n-za DINGIRLU ]M Held(C): NOM.SG völlig ordnen: NOM.C.SG Gott: NOM.SG ḫa-an-da-an-wa DINGIRLIM šar-ku-uš UR.SAG-iš pa-ra-a ḫa-an-da-a[n-za DINGIRLU]M “Wahrlich, der Gott ist ein erhabener Held, ein ordnender Gott.”

2 Wackernagel-Elemente Betrachtet man nun Belege mit Wackernagel-Elementen, so ist die Anzahl der Sätze mit solchen Elementen viel höher: Tabelle 11.2

Distribution der Wackernagelelemente

614 Hauptsätze 109 Nebensätze 1. Ordnung 3 Nebensätze 2. Ordnung

523 mit Wackernagel-Elementen 79 mit Wackernagel-Elementen 2 mit Wackernagel-Elementen

Die Partikelkette hat bekanntlich folgende Slots.

240 Tabelle 11.3

Lühr Slots in der Wackernagelkette (Hoffner and Melchert 2008: 410)

1

2

part/conj quot acc/dat.pl -(m)a -wa(r) -nnaš -(y)a -šmaš nu -šmaš.dat. šu ta

3

4

5

6

nom/acc.sg -aš.nom.sg.c -an.acc.sg.c -at.nom.acc.sg.n -e.nom(/acc).pl -uš.acc.pl.c

acc/dat.sg refl part 1. -mmu -za -ašta 2. -tta/-ddu -apa 3. -šše.dat -kan -šan -an

2.1 Pronomina Bei den Pronomina handelt es sich um zugängliche Information. Bezeichnet werden Sprecher und Adressat und anaphorische Relationen mit dem Pronomen der 3. Person. Die Reihenfolge der enklitischen Formen entspricht nur im Dativ Plural der unmarkierten nominalen Abfolge von Dativ vor Akkusativ; vgl. mit Substantiven: (19) Ritual gegen unterirdische Mächte (CTH 447), A2, 27–28

nu A-NA LUGAL MUNUS.LUGAL DUMU.NITA MEŠ und für König Königin Sohn: PL MEŠ MEŠ DUMU.MUNUS DUMU.DUMU - -ŠU-NU pa-a-i Tochter: PL Enkel: PL ihr geben: PRS.IMP.ACT2SG nu A-NA LUGAL MUNUS.LUGAL DUMU.NITA MEŠ DUMU.MUNUS MEŠ DUMU.DUMU MEŠ-ŠU-NU pa-a-i “Gib dem König (und) der Königin Söhne, Töchter (und) ihre Enkel!”

Doch schließen sich Pronomina in Slot 2 und 4 gegenseitig aus. (20) aš-ma-ša-at “euch ihn” vs. -an-mu “ihn mir” Vgl. mit -an-mu “ihn mir”: (21) Muwatallis Gebet (CTH 381), Rs. iii 45–47

nu am-me-el ku-wa-pí A-WA-TE MEŠ und ich: gen wenn Wort: PL DINGIR MEŠ iš-ta-ma-aš-ša-an-zi Gott: PL hören: PRS.IND.ACT3PL nu- -mu- -kán ku-iš und ich: DAT.SG PART welcher: NOM.C.SG

241

Satzanfänge im Hethitischen

i-da-lu-uš me-mi-aš ZI-ni böse: NOM.C.SG Wort(C): NOM.SG Seele: LOC.SG an-da na- -an- -mu in und er: ACC.C.SG ich: DAT.SG DINGIR MEŠ EGIR-pa SIG5-aḫ-ḫa-an-zi Gott: PL wieder gut machen: PRS.IND.ACT3PL  nu am-me-el ku-wa-pí A-WA-TE MEŠ DINGIR MEŠ iš-ta-ma-aš-ša-an-zi nu-mu-kán ku-iš i-dalu-uš me-mi-aš ZI-ni an-da na-an-mu DINGIR MEŠ EGIR-pa SIG5-aḫ-ḫa-an-zi “Wenn die Götter meine Worte hören, machen sie das böse Wort, das in meiner Seele ist, für mich wieder gut.”

Mit Quotativpartikel und -at-mu “sie mir”: (22) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), i 37

nu- -wa-ra- -at- -mu und QUOT er: NOM.C.PL mir GAM-an e-kir bei sterben: IPF.IND.ACT3PL nu-wa-ra-at-mu GAM-an ekir “Sie starben bei mir.”

Den Grund für diese Distribution haben Agbayani and Golston (2012) aufgezeigt: Tabelle 11.4

Phonologie der Slots 2–4

Onset + coda

No onset

No coda

2 naš šmaš šmaš

3 aš an at e uš e

4 mu ta/du še

Pronomina in Slot 2 gehen denen in Slot 3 voraus, so dass der auslautende Konsonant von =naš und =šmaš einen Onset zu den mit Vokal anlautenden Pronomina =aš, =an, =at, =e, =uš und =e bildet. Dagegen stehen Pronomina in Slot

242

Lühr

3 vor denen in Slot 4, weil dadurch ein Hiat vermieden wird: 4–3 *mu=an [mu. an] mit Hiat. 2.2 “Reflexivpartikel” -za Slot 5 der Partikelkette enthält die sogenannte Reflexivpartikel -za. Nach Agbayani and Golston (2012) ist der semantische Beitrag dieser Partikel oftmals nicht klar. Wegen ihrer idiosynkratischen und nicht-kompositionellen Bedeutung nehmen sie einen syntaktischen Kopf an, der eng mit dem Hauptverb verbunden und als besondere Zwischenschicht “little vP” innerhalb der Tempusphrase u.a. für die Semantik des Subjekts ausschlaggebend ist.4 Für die hethitische Partikel -za sind jedoch die von Kemmer (1993) für das Medium aufgestellten “Situationstypen” entscheidend (vgl. auch Kemmer 1994; Klaiman 1991; Fox and Hopper 1994; Lühr 2012). Sie differenziert folgendermaßen (nach Kaufmann 2004: 8–9): I. Reflexives Medium: Ein Partizipant übernimmt zwei Partizipantenrollen; diese Rollen sind kaum unterscheidbar; z.B. Positionsveränderungsverben: sich setzen, to sit down, πέτεσθαι “fliegen”; Fortbewegungsverben: sich ­entfernen; to remove; indirektes Medium: πορίζεσθαι “sich etwas verschaffen” – der Benefizient wird hier als koreferent mit dem Agens interpretiert. II. Emotions- und Kognitionsmedium: Die Partizipantenrollen sind nicht unterscheidbar; z.B. Emotionsverben: sich erschrecken, to get frightened, irascor; emotive Sprechakte: sich beschweren, to complain, queror, ὀλοφύρεσθαι “sich beklagen”; Kognitionsverben: sich überlegen, to consider, meditor. III. Reziprokes Medium: Teilereignisse und Partizipantenrollen sind bestimmbar; z.B. natürliche reziproke Ereignisse: sich umarmen, to embrace, amplector; natürliche kollektive Ereignisse: sich versammeln, to assemble. IV. Spontane und passivische Situationstypen; z.B. spontane Ereignisse: sich bilden, sich auflösen, to evolve; passivische Ereignisse: the torch inflames. Alle Typen erscheinen nun auch im Hethitischen.5 In unserem Korpus wird z.B. eine Positionsveränderung durch EGIR -za bezeichnet: 4 vP ist auch für die Vergabe des Akkusativs ans Objekt zuständig, so dass zusätzliche Bewegungen angenommen werden müssen. 5 Vgl. dagegen den recht-reflexiven Gebrauch: warpanzi-ma-wa-šmaš ŪL “but they do not wash themselves” (KUB 16.34 i 8–9; Hoffner and Melchert 2008: 358).

243

Satzanfänge im Hethitischen

(23) Muwatallis Gebet (CTH 381), Rs. iii 41–42

ú-u[k- -ma- -z]a- -[ká]n dU ich: nom aber REFL.PART PART Wettergott pí-ḫa-aš-ša-aš-ši-in EN- -I̯A EGIR-pa zum Blitz gehörig: Herr mein zurück AṢ-BAT nu[- -mu ergreifen und ich: acc TI-]-nu-ut leben: CAUS.IPF.IND.ACT3SG ú-u[k-ma-z]a-[ká]n dU pí-ḫa-aš-ša-aš-ši-in EN-I̯A EGIR-pa AṢ-BAT nu[-mu TI-]-nu-ut “Ich aber zog mich zu dem zum Blitz gehörigen Wettergott zurück, meinem Herrn, und er ließ mich leben.”

Ein indirekt-reflexives Medium in: (24) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), I 28 nu- -wa-ra- -at- und QUOT er: ACC.N.SG da-aḫ-ḫi nehmen: PRS.IND.ACT1SG nu-wa-ra-at-za da-aḫ-ḫi “Ich nehme es.”

-za REFL.PART

Ein Kognitionsmedium in: (25) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Vs. ii 62 [nu- -za fZI-an- und REFL.PART Zi(C): ACC.SG ša-a-a]k kennen: PRS.IMP.ACT2SG [nu-za fZI-an-pát ša-a-a]k “Erkenne die Zi!”

-pát PART

Auch Transformationssituationen kann man hier anschließen. Prozesse sind Teilereignisse, die einen vorhergehenden Zustand präsupponieren und so eine Veränderung signalisieren (Engelberg 2000). So erklärt sich nicht nur die ­transformierende, sondern auch die transitivierende Funktion von -za; vgl. etwa:

244

Lühr

(26) tarḫ- “die Oberhand haben” vs. tarḫ- + -za “unterwerfen” (Hoffner and Melchert 2008: 361) (27) Muwatallis Gebet (CTH 381), Rs. iii 28–29

nu- -mu- -za A-BU-    -I̯A und ich: acc REFL.PART Vater    mein DÙ-at dU pí-ḫa-aš-ša-aš-ši-iš machen: IPF.IND.ACT3SG Wettergott zum Blitz gehörig: NOM.C.SG -ma- -mu an-na-az da-a-aš aber ich: acc Mutter(C): ABL.SG nehmen: IPF.IND.ACT3SG nu-mu-za A-BU-I̯A DÙ-at dU pí-ḫa-aš-ša-aš-ši-iš-ma-mu an-na-az d-a-aš “Mein Vater machte mich, aber der zum Blitz gehörige Wettergott nahm mich von der Mutter.”

Dem Sinn nach ein Emotionsmedium ist d]u-uš-kat-ti + -za “freut sich”: (28) Muwatallis Gebet (CTH 381), Rs. iii 62–64

ÉMEŠ DINGIR MEŠ- -i̯a-at- -ta ku-e Haus: PL Gott: PL und du: dat welcher: ACC.N.PL [i-i̯]a-mi ša-ak-la-uš- -ša machen: PRS.IND.ACT1SG Ritus(C): ACC.PL und -da ku-i-e-eš [i-i̯a-]mi du: DAT welcher: ACC.C.PL machen: PRS.IND.ACT1SG nu- -za- -kán dU pí-ḫa-aš-ša-aš-ši-iš und REFL.PART PART Wettergott zum Blitz gehörig: NOM.C.SG EN- -I̯A pa-[ra-a d]u-uš-kat-ti Herr mein hin sich freuen: PRS.IND.ACT2SG ÉMEŠ DINGIR MEŠ-i̯a-at-ta ku-e [i-i̯]a-mi ša-ak-la-uš-ša-da ku-i-e-eš [i-i̯a-]mi nu-za-kán dU pí-ḫa-aš-ša-aš-ši-iš EN-I̯A pa-[ra-ad]u-uš-kat-ti “Welche Gotteshäuser, die ich dir gebaut habe, und welche Riten, die ich dir vollzogen habe, (daran) erfreust du dich, zum Blitz gehöriger Wettergott, mein Herr.”

Die Beschreibung eines emotiven Sprechakts liegt vor in ar-ku-wa-ar [D]Ù-zi + -za “ein Gebet richten an”: (29) Muwatallis Gebet (CTH 381), Vs. i 2–4

ma-a-an UN-[ši] [me-m]i-aš wenn Mann: DAT.SG Wort(C): NOM.SG ku-iš-ki na-ak-ki-i̯a-aš-zi irgendein: NOM.C.SG drücken: PRS.IND.ACT3SG nu -za A-NA DINGIR MEŠ und REFL.PART zu Gott: PL

Satzanfänge im Hethitischen

245

ar-ku-wa-ar [D]Ù-zi Gebet(N): ACC.SG machen: PRS.IND.ACT3SG  ma-a-an UN-[ši] [me-m]i-aš ku-iš-ki na-ak-ki-i̯a-aš-zi nu-za A-NA DINGIR MEŠ ar-ku-wa-ar [D]Ù-zi “Wenn einen Mann ein Wort drückt, richtet er ein Gebet an einen Gott.”

Auch eine überaus geringe “Elaboration der Ereignisse” wird mit der Partikel -za bezeichnet, wie sie in prädikativen Strukturen mit nominalem Prädikatsnomen bzw. Partizip gegeben ist: (30) Šaušgamuwa-Vertrag (CTH 105), Vs. i 40–42

GIM-an- -ma- -za mNIR.GÁL-iš ŠEŠ als aber REFL.PART Muwatalli(C): NOM.SG Bruder(C):NOM.SG A-BI DUTU ŠI DINGIR LIM-iš Vater(C):GEN.SG Majestät(C): GEN.SG Gott(C): NOM.SG ki-ša-at werden: IPF.IND.MED3SG GIM-an-ma-za mNIR.GÁL-iš ŠEŠ A-BI DUTU ŠI DINGIRLIM-iš ki-ša-at “Als aber Muwatalli, der Vaterbruder der Majestät, Gott geworden war…”

(31) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), I 30–31 nu- -wa- -za a-pí-e-da-[n]i- -ya und QUOT REFL.PART jener: DAT.SG auch me-mi-ni iš-ḫi-ú-la-aḫ-ḫa-an-za Sache(C): DAT.SG einen Vertrag machen: PTCP.NOM.C.SG nu-wa-za a-pí-e-da-[n]i-ya me-mi-ni iš-ḫi-ú-la-aḫ-ḫa-an-za “Auch jener Sache bin ich durch Vertrag verpflichtet.”

Auch Possessivkonstruktionen weisen -za auf: (32) Muwatallis Gebet (CTH 381), Rs. iii 48

nu- -za ku-e-el wa-al-li-i̯a-tar und REFL.PART welcher: GEN.SG Preislied(N): NOM.SG nu-za ku-e-el wa-al-li-i̯a-tar “Wessen ist das Preislied?”

Durch die Partikel -za werden also im Hethitischen unterschiedliche Grade der Ereigniselaboration angezeigt.6 Das ist nicht idiosynkratisch. Dass sich dabei 6 Nach allgemeiner Auffassung ergibt sich bei der Unterscheidung der Partizipanten im Falle der Ereignistypen eine Skala, deren positives Ende durch zwei Partizipantenereignisse und

246

Lühr

Kontrollverhältnisse wie beim Aktiv ergeben können, wie im Falle des “transitivity toggle” mit tarḫ- + za “überwinden,” ist weniger maßgeblich, vielmehr kommt es darauf an, anzugeben, wie sehr ein Partizipant in ein Ereignis involviert oder davon affiziert ist. Dieser Partizipant wird in der Regel durch das Subjekt bezeichnet, aber auch durch andere Kasusrollen wie den Dativ als Benefaktiv oder den Genitiv und das Possessivpronomen als Possessivus. 2.3 “Lokalpartikeln” Zur Statistik in den Korpora vgl.: Tabelle 11.5

Distribution der Lokalpartikeln

Gerichtsprotokoll Muwatalli Ritual 443 Ritual 447 Telipinu Gesamt

-kan

-šan

-apa

-ašta

-an

4 34 24 14 4 80

– 2 29 4 – 35

– 2 – – 5 7

– – – 5 1 6

– – – – – –

In gleicher Weise wie den Gebrauch der “Reflexivpartikel” -za halten A ­ gbayani and Golston (2012) die Verwendung der sogenannten “Lokalpartikeln” für idiosynkratisch. Sie verweisen dafür auf Hoffner and Melchert: The particles -an, -apa, -ašta, -kan, and -šan belong to a single class of sentence particles which modify the action expressed by the main verb and its ‘adverbal’ adjuncts. The nature of that modification is disputed: some thinking that it marks primarily local relationships … and another … that it marks verbal aspect. These two viewpoints are not mutually exclusive. hoffner and melchert 2008: 364–365; vgl. auch tjerkstra 1999

Der Sprachgebrauch dieser Partikeln hat aber einen klaren Situationsbezug: Die Proposition wird wie bei den Frames in eine Situation eingebettet. Auch deren negatives Ende durch ein Partizipantenereignis gebildet wird. Reflexive Ereignisse sind dazwischen angesiedelt (Kaufmann 2004: 9). Vgl. Hopper and Thompson (1980: 252): Das Medium unterscheide sich vom Aktiv durch “transitivity.”

Satzanfänge im Hethitischen

247

wenn sich, bedingt durch die jeweilige Situation, Bedeutungsschattierungen ergeben, kann man bei allen “Lokalpartikeln” in der deutschen Übersetzung das Wort da verwenden, lokal, temporal oder in der Bedeutung “in dieser Situation.” Über den lokalen und demonstrativen Gebrauch dieses Wortes hinaus geht die Verwendung als Sprechhandlungspartikel, z.B. (33)(a) Ich habe da gestern jemanden (auf der Straße getroffen). (33)(b) Da haben wir’s. Die Übersetzung von -kán mit “da, in dieser Situation” passt auf jeden Fall dann, wenn in dem Satz ein lokales Adverbiale vorhanden ist: (34) Muwatallis Gebet (CTH 381), Vs. i 4–5

šu-uḫ-ḫi- -kán še-er dUTU-i Dach(C): LOC.SG SIT.PART auf Sonnengott: DAT.SG GIŠ me-na-aḫ-ḫa-an-da 2 BANŠUR gegenüber 2 Tafel AD.KID [k]a-ri-i̯a-an-da da-a-i Rohrgeflecht bedeckt: ACC.N.PL setzen: PRS.IND.ACT3SG  šu-uḫ-ḫi-kán še-er dUTU-i me-na-aḫ-ḫa-an-da 2 GIŠBANŠUR AD.KID [k]a-ri-i̯a-an-da da-a-i “Auf dem Dach da stellt er dem Sonnengott (d.h. der Sonne) gegenüber zwei Tafeln von Rohrgeflecht zugedeckt auf.”

Auch wenn eine temporale Bedeutung mitschwingt, wie in der Beschreibung von emotiven Sprechakten, ist die Wiedergabe mit “da” gegeben. Mit einem einfachen Verb vgl.: (35) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), iii 35

mGAL-dU-aš- -za- -ká[n GAL-dU(C): NOM.SG REFL.PART SIT.PART A-N]A PA-NI DINGIRLIM für Vorderseite: STAT.CONSTR Gott: GEN pa-ri-ya-an ki-i IQ-BI außerdem dieser: ACC.N.SG er sprach mGAL-dU-aš-za-ká[n A-N]A PA-NI DINGIRLIM pa-ri-ya-an ki-i IQ-BI “GAL-dU sprach da/in dieser Situation im Angesicht des Gottes außerdem das:”

Mit einem präfigierten Verb:

248

Lühr

(36) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), i 10 nu- -za- -kán li-in-ki-ya und REFL.PART SIT.PART Eid(C/N): LOC.SG an-da kiš-an pí-e-da-aš hinein folgendermaßen hinschaffen: IPF.IND.ACT3SG nu-za-kán li-in-ki-ya an-da kiš-an pí-e-da-aš “Folgendermaßen sagte er da/in dieser Situation unter Eid aus.”

Etymologisch ist für die Partikel -kán wie für die Partikel -šan eine Entwicklung von “dabei” (lat. cum < *kom; *som “zusammen mit”) > “in dieser Situation” > “da” möglich; vgl. engl. at that. Die beiden Partikeln sind miteinander austauschbar (zur Literatur vgl. Hoffner and Melchert 2008: 373, 377). Vgl. Belege mit -šan in lokal-temporalem Bezug: (37) Ritual gegen unterirdische Mächte (CTH 447), B4, 5



nu- -uš- -[-š]a-an ḫal-zi-an-zi und er: ACC.C.PL SIT.PART rufen: PRS.IND.ACT3PL nu-uš[-š]a-an ḫal-zi-an-zi “Und man ruft sie da.”7

(38) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Rs. iii 23–24 na- -at- -ša-an A-N[A DUMU MEŠ(?) und er: NOM.N.SG SIT.PART für Sohn: PL f]Zi [EGIR-pa ták-ša-an Zi wieder zufügen: NOM.N.SG e-eš-tu] sein: PRS.IMP.IND.ACT3SG na-at-ša-an A-N[A DUMU MEŠ(?) f]Zi [EGIR-pa ták-ša-an e-eš-tu] “Es sei da den Söhnen der Zi wieder zugefügt.”

Mehr lokal ist der Gebrauch in: (39) Ritual gegen unterirdische Mächte (CTH 447), A2, 29–30 ku-iš- -ša-an UDU ḪI.A-aš welcher: NOM.C.SG SIT.PART Schaf: GEN.PL ḫu-li-i̯a-aš nu- -uš Wolle(C): NOM.SG und er: ACC.C.PL

7 Görke and Melzer 2015; Popko 2003: 39.

Satzanfänge im Hethitischen

249

LUGAL-i MUNUS.LUGAL-ri König(C): DAT.SG Königin(C): DAT.SG MU KAM ḪI.A GÍD.DA pa-a-i Jahr: PL lang geben: PRS.IMP.ACT2SG  ku-iš-ša-an UDU ḪI.A-aš ḫu-li-ja-aš nu-uš LUGAL-i MUNUS.LUGAL-ri MU KAM ḪI.A GÍD.DA pa-a-i “Was dabei die Wolle der Schafe ist, gib dem König (und) der Königin diese an langen Jahren!”

Mit -kán und -šan im selben Satz, also austauschbar: (40) Ritual gegen unterirdische Mächte (CTH 447), B2, 38–40

ma-a-na- -at- -kán ták-na-aš wenn er: ACC.N.SG SIT.PART Erde(N): GEN.SG DUTU-uš šar-ra-at-ta- -ma Sonnengott(C): NOM.SG überschreiten: PRS.IND.MED2SG aber nu- -uš-ša-an  LUGAL-i MUNUS.LUGAL-i und SIT.PART König(C): DAT.SG Königin: DAT.SG Ú-UL ḫar-ap-ši nicht eng verbunden sein: PRS.IND.ACT2SG nu- -ut-ta ú-it-tu4 und dich kommen: PRS.IMP.ACT3SG ke-e-el ŠA SÍSKUR [l]i-in-ki-ja-an-za dieser: GEN.C.SG von Opfer Schwurgottheit(C): NOM.SG e-ep-du fassen: PRS.IMP.ACT3SG ma-a-na-at-kán ták-na-aš DUTU-uš šar-ra-at-ta-ma nu-uš-ša-an LUGAL-i MUNUS.LUGAL-i Ú-UL ḫar-ap-ši nu-ut-ta-ú-it-tu4 ke-e-el ŠA SÍSKUR [l]i-in-ki-ja-an-za e-ep-du “Wenn du aber, oh Sonnengöttin der Erde, es da brichst und dich da nicht für den König (und) die Königin bemühst, soll der Eidgott dieses Opfers kommen, dich ergreifen.”

Dann mit -aš-ta wieder mit lokal-temporaler Färbung: (41) Telipinu-Erlass (CTH 19), Vs. i 55 mI-la-l]i-ú-ma- -aš-ta [DUMU]MEŠ.É.GAL Ilaljuna: NOM SIT.PART Palastangestellter: PL [(pa-ra-a du-ud-da-mi-li u-i-)i̯(a-a)t aus heimlich schicken: IPF.IND.ACT3SG MEŠ mI-la-l]i-ú-ma-aš-ta [DUMU] .É.GAL [(pa-ra-a du-ud-da-mi-li u-i-)i̯(a-a)t “Ilaljuna schickte da heimlich Palastangestellte aus.”

250

Lühr

Und mit lokal-temporalem -apa: (42) Telipinu-Erlass (CTH 19), Vs. i 42

[na- -pa mMur-ši-i-li-i̯a-aš] und SIT.PART Mursili: GEN.SG e-eš-ḫar DINGIR MEŠ-iš Blut(N): ACC.SG Gott: PL ša-an-ḫ[i-(i)r] fordern: IPF.IND.ACT3PL [na-pa mMur-ši-i-li-i̯a-aš] e-eš-ḫar DINGIR MEŠ-iš ša-an-ḫ[i-(i)r] “Und da forderten die Götter das Blut des Muršili.”

Auch in Verbindung mit Imperativen kommen die situierenden Partikeln vor; vgl. mit aš-ta: (43) Ritual gegen unterirdische Mächte (CTH 447), A3, 18

D na- -aš-ta KI-aš UTU-uš und SIT.PART Erde(N): GEN.SG Sonnengott(C): NOM.SG i-i̯a-an-ni gehen: PRS.IMP.ACT2G na-aš-ta KI-aš DUTU-uš i-ja-an-ni “Geh da, oh Sonnengöttin der Erde!”

(44) Ritual gegen unterirdische Mächte (CTH 447), A2, 33–34

na- -aš-ta KI-aš DUTU-uš und SIT.PART Erde(N): GEN.SG Sonnengott(C): NOM.SG [(ki-i ut-tar aš-nu-ut)] dieser: ACC.N.SG Wort(N): ACC.SG herrichten: PRS.IMP.ACT2SG na-aš-ta KI-aš DUTU-uš [(ki-i ut-tar aš-nu-ut)] “Und da, oh Sonnengöttin der Erde, dieses Wort mache wahr.”

Wieder mit -apa in einem verneinten Wunsch: (45) Entsühnungsritual für Tutḫaliya und Nikalmati (CTH 443), Vs. ii 16 na- -ta- -pa und er: NOM.N.SG SIT.PART EGIR-pa li-e ú-iz-zi wieder nicht kommen: PRS.IND.ACT3SG na-ta-pa EGIR-pa li-e ú-iz-zi “Und sie sollen da nicht wiederkommen.”

Satzanfänge im Hethitischen

251

Überlegt man sich nun, in welchem Medium von Sprache situierende Partikeln zu erwarten sind, so ist dies die gesprochene Sprache:8 Partikeln sind ein Merkmal dieser Sprachform. Demgegenüber gilt die geschriebene Sprache als partikelarm.9 In der Tat könnten die vielfältigen Lokalpartikeln im Hethitischen einen Reflex dieser Sprachform darstellen und in die geschriebene Sprache übernommen sein. So finden sich die situierenden Partikeln in wörtlicher und zitierter Rede und wirken da wie in der heutigen gesprochenen Sprache gelegentlich als redundant. Vgl. mit Quotativpartikel -wa: Im Aussagesatz in der 1. Person: (46) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), i 26

LUGAL- -ya- -wa- -kán me-mi-y[a-nu-u]š König und QUOT SIT.PART Wort(C): ACC.PL pa-ra-a Ú-UL ku-it-ki ta-me-en-ga-nu-uš-k[i-mi] danach nicht irgendwie anheften: CAUS.ITER.PRS.IND.ACT1SG LUGAL-ya-wa-kán me-mi-y[a-nu-u]š pa-ra-a Ú-UL ku-it-ki ta-me-en-ga-nu-uš-k[i-mi] “Und des Königs Worte mache ich da in Zukunft in keiner Weise mehr unverständlich.”

(47) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), i 27–28 ANA LUGAL- -wa-ra- -atfür König QUOT er: ACC.N.SG -kán me-ir-nu-un SIT.PART verschwinden: CAUS.IPF.IND.ACT1SG A-NA LUGAL-wa-ra-at-kán me-ir-nu-un “Dem König unterschlug ich da/in dieser Situation das.”

(48) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), I 13–14

MUNUS.LUGAL- -wa- -mu ku-it EGIR-pa Königin QUOT mir welcher: ACC.N.SG wieder pí-eš-ki-it nu- -wa- -kán geben: ITER.IPF.IND.ACT3SG und QUOT SIT.PART Ú-UL ku-it-ki pí-ni-nu-nu-un nicht irgendwer: ACC.N.SG geben: CAUS.IPF.IND.ACT1SG

8 “Lokalpartikeln” sind an der Situierung des “kommunikativen Sinns” von Äußerungen (vgl. Löbner 2003) beteiligt (Brinker and Sager 2001). Damit gehen sie über die Charakterisierung der räumlichen Bedingungen eines Sachverhalts hinaus. 9 Nach Hoffner and Melchert (2008: 365) finden sich die Partikeln -an und -apa hauptsächlich in Texten, die während der althethitischen Periode verfasst wurden. Nachdem -an, -apa und auch -ašta desolet geworden waren, wurden ihre Funktionen von -šan und -kan übernommen.

252

Lühr MUNUS.LUGAL-wa-mu ku-it/ EGIR-pa pí-eš-ki-it nu-wa-kán Ú-UL ku-it-ki pí-ni-nunu-un “Was mir die Königin jeweils übergeben hat, da/in dieser Situation ließ ich nichts weggeben.”

In der 3. Person: (49) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), i 24

INIM wa-al-wa-ya-al-la-aš- -ma- -wa- -kán nam-ma Sache Verleumdung: GEN.SG aber Quot part weiter EGIR-an kat-ta pa-it danach unten gehen: IPF.IND.ACT3SG INIM wa-al-wa-ya-al-la-aš-ma-wa-kán nam-ma EGIR-an kat-ta pa-it “Die Sache der Verleumdung verschlechterte sich da danach aber wieder.”

Im Fragesatz haben “Lokalpartikeln” möglicherweise auch den Sinn von dt. denn; vgl. in direkten Fragen: (50) Telipinu-Erlass (CTH 19), Vs. i 58–61

[(ma-a-an mḪa-an-ti-i-li-iš MUNUS.LUGAL URUŠu-u)k-zi-ya als mḪantili Königin Stadt Ḫukziya MEŠ Ú DUMU - -ŠU [(EGIR-an ša-an-aḫ-ta und Sohn: PL ihr nach suchen: IPF.IND.ACT3SG ku-iš- -u̯ a-r-a- -aš- -kán wer: NOM.SG QUOT sie: ACC.C.PL SIT.PART ku-)en-ta] [(GAL DUMU MEŠ.É.GAL ḫa-lu-kán töten: IPF.IND.ACT3SG oberster Palastangestellter Botschaft(C): ACC.SG ú-da-aš bringen: IPF.IND.ACT3SG [(ma-a-an mḪa-an-ti-i-li-iš MUNUS.LUGAL URUŠu-u)k-zi-ya Ú DUMU MEŠ-ŠU [(EGIR-an Ša-an-aḫ-ta ku-iš-u̯ a-ra-aš-kán ku-)en-ta] [(GAL DUMU MEŠ.É.GAL ḫa-lu-kán ú-da-aš “Als Ḫantili nach der Königin der Stadt Šukziya und ihren Söhnen suchte (mit den Worten) “Wer tötete sie da/denn?,” brachte der oberste Palastangestellte eine Botschaft.”

(51) Telipinu-Erlass (CTH 19), Vs. i 35

[nu- (-u̯ a- -kán pa-aḫ-)ḫa-aš-ḫa und QUOT SIT.PAT schützen: PRS.IND.MED1SG [nu-(-u̯ a-kán pa-aḫ-)ḫa-aš-ḫa “Bin ich da/denn beschützt?”

Satzanfänge im Hethitischen

253

In rhetorischen Fragen: (52) Der Prozeß gegen GAL-DU und seinen Vater Ukkura (CTH 293), i 18–20

nu- -wa ku-it-ma-an I-NA KUR URUKa-ra-an-du-ni-ya-aš und QUOT während in Land Babylon pa-a-un [] gehen: IPF.IND.ACT1SG ku-it-ma-an- -wa EGIR-pa ú-wa-nu-un nusolange als QUOT wieder kommen: IPF.IND.ACT1SG und -wa- -kán EGIR-az wa-al-wa-i̯a-al-li QUOT SIT.PART Rückseite: ABL.SG Verleumdung(N): ACC.SG an-da Ú-UL ku-iš-ki pí-eš-ši-iš-ki-it hinein nicht irgendwer: NOM.C.SG werfen: ITER.IPF.IND.ACT3SG nu-wa ku-it-ma-an I-NA KUR URUKa-ra-an-du-ni-ya-aš pa-a-un [ ] ku-it-ma-an-wa EGIRpa ú-wa-nu-un nu-wa-kán EGIR-az wa-al-wa-ya-al-li an-da Ú-UL ku-iš-ki pí-eš-ši-iš-ki-it “Während ich ins Land der Stadt Babylon ging…, solange bis ich wiederkam, hat da nicht irgendwer hinterrücks Verleumdung vorgenommen?”

2.4 Partikeln der zitierten Rede Vielleicht hat auch die Partikel der zitierten Rede – sie folgt in der satzeinleitenden Partikelkette enklitischen Konjunktionen, satzverknüpften Partikeln und topikalisierten Elementen – ihren Ursprung in der gesprochenen Sprache. -wa(r) könnte in dieser Sprachform nicht nur der Einleitung der direkten Redewiedergabe gedient haben, sondern in der mündlichen Kommunikation ein Merkmal der sogenannten “animierten Rede” (Ehmer 2011) gewesen sein. Der aktuelle Sprecher führt in einer solchen Rede ein imaginiertes Sprechereignis vor, wobei die Inszenierung und Stilisierung fremder Rede eng mit den Bewertungen der Sprechenden hinsichtlich der zitierten Figuren und Äußerungen verwoben ist (Günthner 2002: 60). Solch ein Hinweis muss in der mündlichen Kommunikation möglichst nahe am Satzanfang kommen, damit der Adressat die mit der direkten Wiedergabe verbundene Kommunikationsabsicht richtig versteht. Die Übernahme in die geschriebene Sprache, vor allem in historische Traktate, Verträge, Annalen, Briefe, zeigt dann das Bestreben, in Textsorten, in denen es auf den genauen Wortlaut von Zitaten ankommt, diesen als oratio recta stets eindeutig zu kennzeichnen. Die Setzung dieser Partikel wurde so in der “­Kanzleisprache” regulär, als eine Art “stylistic standardization” (Fortson 1994, 1998).10 10

Zu einer Erklärung der sporadischen Setzung der Partikel -wa(r) vgl. Pecora 1984 (dazu aber Fortson 1998).

254 3

Lühr

Partikeln und inkrementelle Syntax

Sind aber diese Überlegungen zu der Partikelkette in der Wackernagel-Position richtig, so hätte man hier einen Fall von inkrementeller Syntax. Eine solche Syntax folgt psycholinguistischen Befunden: In der gesprochenen Sprache ­basiert sprachliche Produktion und Rezeption auf dem Phänomen der Projektabilität, das an die Zeitlichkeit der Entfaltung von Sprache im Gebrauch ­gebunden ist. Bei einer inkrementellen Sprachverarbeitung hat ein Adressat gleich nach den ersten Worten im Satz eine Ahnung davon, worum es im ­Folgenden geht, und zwar unabhängig vom Kontext. Denn die fortlaufenden Projektionen über den Verlauf einer emergenten syntaktischen Struktur ­ ­erlauben es dem Hörer, den entstehenden Redebeitrag ohne Verzögerung zu prozessieren (Auer 2007, 2009, 2015).11 Für die hethitische Partikelkette besagt das: Zitierte Rede Anschluss an Gegebenheiten der Sprechsituation und anaphorische Wiederaufnahme Grade der Ereigniselaboration in der Regel mit Bezug auf das Subjekt Situierung der Proposition

-wa (oder Verben des Sagens) Pronomina “Reflexivpartikel” -za “Lokalpartikeln”

Lässt man die Vokative und Anredeformen an der Satzspitze beiseite, so ist gegenüber den Sätzen ohne Wackernagel-Elementen, aber mit Textadverbien, Frames, Contrastive Topics und Foci eine Syntax mit Wackernagel-Elementen wohl adressatenfreundlicher. Die Wackernagel-Syntax wäre so ein unmittelbarer Reflex der gesprochenen Sprache. Das könnte der Grund sein, weshalb es im Hethitischen so viele Partikeln gibt.

Zusätzliche Abkürzungen

act Aktiv C Commune con Konjunktion 11

Dieser Aspekt wird u.a. auch in Hoppers Emergent Grammar fokussiert (Hopper 1987, 2001, 2011). Vgl. auch Hoffner 1986.

Satzanfänge im Hethitischen

255

ipf Imperfekt jh. junghethitisch part Partikel REFL.PART Reflexivpartikel SIT.PART Situationspartikel STAT.CONSTR Status constructus sup Supinum

Literatur Agbayani, B. and Golston, Ch. (2012) Clitic order in Hittite, in: St. W. Jamison, H.C. ­Melchert and B. Vine (eds.) Proceedings of the 23rd Annual UCLA Indo-European Conference, Los Angeles, October 28th and 29th, 2011, Bremen: Hempen, 1–15. Auer, P. (2007) Syntax als Prozess, in: H. Hausendorf (ed.), Gespräch als Prozess. ­Linguistische Aspekte der Zeitlichkeit verbaler Interaktion, Tübingen: Narr, 95–124. Auer, P. (2009) On-line syntax: thoughts on the temporality of spoken language, ­Language Sciences 31, 1–13. Auer, P. (2015) The temporality of language in interaction: projection and latency, in: A. Deppermann and S. Günthner (ed.), Temporality in Interaction. Amsterdam: ­Benjamins, 27–56. Brinker, K. and Sager, S.F. (2001) Linguistische Gesprächsanalyse. Eine Einführung,3 ­Berlin: Schmidt. Chafe, W.L. (1976) Givenness, contrastiveness, definiteness, subjects, topics and point of view, in: C. Li (ed.), Subject and Topic, New York: Academic Press, 25–56. Ehmer, O. (2011) Imagination und Animation. Die Herstellung mentaler Räume durch animierte Rede. Berlin, New York: de Gruyter. Engelberg, St. (2000) Verben, Ereignisse und das Lexikon, Tübingen: Niemeyer. Fanselow, G. (2004) Cyclic phonology-syntax interaction: movement to first position in German, in: S. Ishihara, M. Schmitz and A. Schwarz (eds.), Working Papers of the SFB 632, Interdisciplinary Studies on Information Structure (ISIS) 1, Potsdam: Universitätsverlag Potsdam, 1–42. Fox, B. and Hopper, P.J. (eds.) (1994) Voice. Form and Function, Amsterdam, Philadelphia: John Benjamins. Fortson, B. (1994) Prolegomena to a synchronic analysis of the Hittite quotative particle, Harvard Working Papers in Linguistics 4, 37–51. Fortson, B. (1998) A new study of Hittite -wa(r), JCS 50, 21–34. Görke, S. and Melzer, S. (eds.) (2015) hethiter.net/: CTH 447 (TX 20.07.2015, TRde 20.07.2015).

256

Lühr

Günthner, S. (2002) Stimmenvielfalt im Diskurs: Formen der Stilisierung und Ästhetisierung in der Redewiedergabe, Gesprächsforschung. Online-Zeitschrift zur verbalen Interaktion 3, 59–80. Hoffner, H.A., Jr. (1986) Studies in Hittite grammar, in: H.A. Hoffner, Jr. and G.M. Beckman (eds.), Kaniššuwar. A Tribute to Hans G. Güterbock on his Seventy-Fifth Birthday, May 27, 1983 [AS 23], Chicago: The Oriental Institute of the University of Chicago, 83–94. Hoffner, H.A., Jr. and Melchert, H.C. (2008): A Grammar of the Hittite Language, 1: Reference Grammar, Winona Lake, IN: Eisenbrauns. Hopper, P.J. (1987) Emergent grammar, Proceedings of the Annual Meeting of the Berkeley Linguistics Society 13, 139–157. Hopper, P.J. (2001) Grammatical constructions and their discourse origins: prototype or family resemblance? In: M. Pütz, S. Niemeier and R. Dirven (eds.), Applied Cognitive Linguistics. i: Theory and Language Acquisition [Cognitive Linguistics Research 19.1], Berlin, New York: de Gruyter, 109–129. Hopper, P.J. (2011): Emergent grammar and temporality in interactional linguistics, in: P. Auer and St. Pfander (ed.), Constructions: Emerging and Emergent [Linguae & Litterae 6], Berlin, Boston: de Gruyter, 22–44. Hopper, P.J. and Thompson, S.A. (1980) Transitivity in grammar and discourse, Language 56, 251–299. Jacobs, J. (1997) I-Topikalisierung, Linguistische Berichte 169, 91–133. Kaufmann, I. (2004) Medium und Reflexiv. Eine Studie zur Verbsemantik, Tübingen: Niemeyer. Kemmer, S. (1993) The Middle Voice, Amsterdam, Philadelphia: John Benjamins. Kemmer, S. (1994) Middle voice, transitivity, and the elaboration of events, in: Fox and Hopper 1994: 179–230. Klaiman, M. (1991) Grammatical Voice, Cambridge: Cambridge University Press. Krisch, Th. (1998) Zum Hyperbaton in altindogermanischen Sprachen, in: W. Meid (ed.), Sprache und Kultur der Indogermanen. Akten der X. Fachtagung der Indogermanischen Gesellschaft, Innsbruck, 22.–28. September 1996 [IBS 93], Innsbruck: ­Institut für Sprachwissenschaft der Universität Innsbruck, 351–384. Krisch, Th. (2007) Vorstellen eines Arbeitsvorhabens: Scrambling in altindogermanischen Sprachen (Vortrag gehalten auf dem Syntaxtreffen in Marburg, September 2007). Lang, E. (1984) The Semantics of Coordination, Amsterdam: John Benjamins. Lang, E. (1991) Koordinierende Konjunktionen, in: A. von Stechow and D. Wunderlich (eds.) (1991) Semantik. Ein internationales Handbuch der zeitgenössischen Forschung [HSK 6], Berlin: de Gruyter, 597–623. Löbner, S. (2003) Semantik: Eine Einführung. Berlin: de Gruyter.

Satzanfänge im Hethitischen

257

Lühr, R. (2008) Information structure in ancient Greek, in: A. Steube (ed.), Discourse Potential of Underspecified Structures [Language, Context, and Cognition 8], Berlin, New York: de Gruyter, 487–512. Lühr R. (2012) Ereignistyp und Diathesenwechsel im Indogermanischen, in: H.C. Melchert (ed.), The Indo-European Verb. Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles, 13–15 September 2010, Wiesbaden: Reichert, 213–224. Maienborn, C. (2001) On the position and interpretation of locative modifiers, Natural Language Semantics 9, 191–240. Pecora, L. (1984) La particella -wa(r)- e il discorso diretto in antico-eteo, IF 89, 104–124. Popko, M. (2003) Das hethitische Ritual CTH 447, Warszawa: Agade. Riemsdijk, H. van (1989) Movement and regeneration, in: P. Benincà (ed.), Dialect Variation and the Theory of Grammar, Dordrecht, Providence: Foris, 105–135. Speyer, A. (2009) Versuch zur Syntax im Protoindoeuropäischen, in: E. Rieken and P. Widmer (eds.), Pragmatische Kategorien. Form, Funktion und Diachronie. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 24. bis 26. September 2007 in Marburg, Wiesbaden: Reichert, 287–305. Thurmair, M. (1989) Modalpartikeln und ihre Kombinationen, Tübingen: Niemeyer. Tjerkstra, F.A. (1999) Principles of the Relation between Local Adverb, Verb and Sentence Particle in Hittite [Cuneiform Monographs 15], Groningen: Styx.

Chapter 12

Hittite Historical Phonology after 100 Years (and after 20 Years) H. Craig Melchert Abstract On the centennial of Hrozný’s identification of Hittite as an Indo-European language I review the major issues in Hittite historical phonology, in comparison not only with his sketch of 1917 but also with my own treatment of 1994. After some methodological preliminaries and an overview of major revisions required by the last two decades of scholarship, I focus on the PIE syllabic sonorants. Hrozný (1917: 187) already tentatively concluded that their regular outcome in Hittite was aR, and this has become the standard view, but he also entertained that syllabic nasals before stops could appear as simply a. The demonstration by Goedegebuure (2010) that CLuvian zanta is the cognate of Hittite katta ‘down’ has renewed the question of the development of the syllabic nasals before stops. Confirmation that in Hittite spells /o/, including in the result of *wR̥ , also casts doubt on the alleged direct change of *wR̥ > uR by resyllabification (Melchert 1994a: 126–127). I reexamine the entire question of the development of syllabic sonorants in Hittite in the light of these new findings.

Keywords laws of finals – nasals – orthography – phonology – sonorants – syllabification – vocalism

I

Introduction: Methodological Issues

A Hrozný’s Methodology That Hrozný’s identification of Hittite as an Indo-European language was a great accomplishment has never been seriously questioned. However, there has been a longstanding and widespread narrative in Indo-European studies that Hrozný relied too heavily on the etymological method, and that his analysis of Hittite was only widely accepted when confirmed through combinatory

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_014

Hittite Historical Phonology after 100 Years

259

analysis by others, notably by Ferdinand Sommer (see e.g. the excellent historical summary of Eichner 1980: 123–124 with notes and references). It is high time after one hundred years to put this canard to rest. A cursory survey of his Glossar (Hrozný 1917 [2002]: 212–246) shows that Hrozný correctly identified approximately eighty percent of the lexical morphemes; less than fifteen percent are wholly incorrect. Of the latter, only his interpretation of dā‘take’ and dai- ‘put’ as ‘give’, of ḫar(k)- ‘have, hold’ as ‘take’ (also confused with ḫark- ‘perish’ and ḫarnink- ‘destroy’), and of uwa- ‘come’ also as ‘go’ are truly due to etymological assumptions. Others are simply predictable errors of a first decipherment (Hrozný did have a particular problem in analyzing spatial relations, for a review of which see Frantíková 2015). It is true that Hrozný’s success rate was obviously due in part to alternations of syllabically spelled Hittite words with known Sumerian and Akkadian equivalents, but this fact merely confirms that his method was mostly combinatory. Likewise, his successful tentative interpretation of many words that were in 1917 hapax or near-hapax necessarily resulted from an analysis of the context, not from presumed etymological connections. In sum, Hrozný employed a blend of the combinatory and etymological methods, following what is a widespread and standard practice in the analysis of newly discovered and only partially understood corpus languages. As was to be expected, further study by Hrozný himself and by others corrected his most serious initial errors within the next decade, but he had successfully established the basic facts of Hittite grammar already by 1917 (for historical phonology see further below). B Interpretation of Orthography Much recent study of Hittite phonology, synchronic and diachronic, has unfortunately been based on the widespread pernicious false premise that all nonrandom orthographic patterns must at all costs reflect linguistically real ­contrasts: see e.g. Adiego (2007: 235–237), Eichner (1980: 143–144, note 63, and 149, note 73, and 1992: 57), Kloekhorst (2010: 204 and passim; 2012: 247, 2013: passim, et alibi), Rieken (2010a: 305–307 and 2010b: 653 and passim), and Simon (2012: 494 and 2013: 16, with note 37). The preceding premise reflects a profound misunderstanding of how orthographies developed by and for native speakers—especially by and for ­scribal elites—actually function. Nonrandom patterns (including those established as statistically significant) may be due to a wide variety of factors: ­established norms, aesthetic considerations (e.g. “initial-a final” and the spelling of word-initial /a-/ in Hieroglyphic Luvian Anatolian hieroglyphs, as discussed by Melchert 2010c and Rieken 2015: 226–227), and pure convention.

260

Melchert

These ­features can be due to the actions of a single influential individual: without knowledge of the existence of Noah Webster and no evidence for English before the eighteenth century, one might assume that certain sound changes took place ­between British and American English (-ise > -ize, -our >-or, etc.).1 Given the severe limitations on our knowledge of the Hittite scribal hierarchy, we cannot preclude that certain Hittite orthographic practices (and changes therein over time) are likewise the result of decisions made by a handful of powerful chief scribes. Some orthographic systems with a long history such as Modern English or French naturally do reveal phonological changes through ‘historical spellings’, almost always imperfectly due to earlier changes being overlaid by later ones and resulting misunderstandings.2 However, suggestive but less than exceptionless orthographic patterns in Hittite and Luvian written in cuneiform cannot be attributed to ‘historical’ spellings reflecting an earlier prehistoric stage of the languages, because there is no basis for supposing that they were written before the late sixteenth century bce (at the earliest; see now van den Hout 2009 for the claim of written Hittite only in the fifteenth century). Therefore, while it is always proper and beneficial to periodically review the status of particular patterns in the light of new evidence and arguments (see Section ii.B.1 below for some examples), the validity of claims of linguistically real contrasts depends solely on the degree of crosslinguistic plausibility, synchronic and diachronic, of the pattern of contrasts claimed. In the ­absence of such plausibility, there is no basis for the claimed distinctions, no matter how striking the nonrandom spelling pattern may be. An orthographic pattern is merely a necessary, not a sufficient basis for assuming a linguistic contrast. II

Historical Phonology

A Hrozný’s Presentation of 1917 Reflecting the true facts about his methodology presented in I.A above, Hrozný’s initial sketch of Hittite historical phonology (1917: 186–190) is predictably correct in many respects: 1 Webster’s predominant role in establishing the norms of American English spelling is succinctly described in the Wikipedia entry https://en.wikipedia.org/wiki/Noah_Webster (accessed February 12, 2016) under the section ‘Blue-backed Speller’. 2 For example, not every final ‘silent e’ in English reflects a true original final vowel lost by apocope, though most do. Modern English horse is a prominent exception.

Hittite Historical Phonology after 100 Years

261

1 Consonantism a. He identifies Hittite as a ‘centum’ language with merger of *K̂ and *K and preservation of *Kw. b. He concludes that*R̥ > aR (but also entertains some cases of *N̥ > a, based on katta ‘down’ matching Gr. κατά—see further Section ii.C.1 below). c. He recognizes that *s > z after nasal in anz- ‘us’ < *n̥ s-. d. He notes that Hittite t continues *th as well as *t (i.e., in the Pret2Sg ending -tta < *-th2e, as in ḫarta ‘you had, held’). e. He assumes unrounding of *Kw > K before u (in kuššan ‘when?’ [recte ‘when’]). f. He takes note of the change *m > n in word-final position. 2 Vocalism a. He assumes a general merger of short and long *ō̆ with *ā̆. b. But he notices the special change of *o > u before nasal in monosyllables (*tons > tuš [sic!]). c. Much of what is said about *ē̆ and *ī� ̆ is also correct. d. He correctly derives zīk ‘thou’ < *tū (contra Simon 2018 et al.). 3 Weaknesses These are mostly predictable, given the difficulties of cuneiform orthography, Hrozný’s limited command of Indo-European linguistics, and no recognition of ‘laryngeals’ in PIE: a. There is no recognition of ‘Sturtevant’s Law’ or the general distribution of voiceless and voiced stops. b. There is no realization of the correct conditioning for *t > ts /__i or of the PIE ‘double dental’ rule. c. He derives Hittite ḫ(ḫ) beside k, g from dorsal stops with not even an attempt to condition the alternate reflexes. d. There is much confusion in the vocalism due to the problem of e- and i-spellings in cuneiform. e. His rudimentary account of diphthongs is far from clear. One should, however, in all fairness acknowledge that points 1b, 2d, 3a, 3d and 3e remain subjects of controversy to this day (see below!), and point 3c was still debated into the 1960s. B Anatolian Historical Phonology Two Decades after Melchert 1994 1 Significant Revisions New facts as well as new arguments made by a number of scholars have predictably falsified many claims that I made in my historical phonology of more

262

Melchert

than twenty years ago, compelling either complete retraction or varying ­degrees of revision of analyses given there. The following selection aims to ­acknowledge those of greatest import for understanding the (pre)history of Hittite and for the reconstruction of Proto-Indo-European. The list is by no means exhaustive. I note explicitly that some newer analyses based on observance of nonrandom spelling patterns have met the high standard of linguistic plausibility demanded above in Section i.B. The convincing new arguments that versus spellings reflect a real synchronic contrast in vocalism deserve special notice. The far-reaching implications of this demonstration have undoubtedly not yet been fully recognized. a Consonantism: i. ‘Laryngeals’ (1) *h2w > Proto-Anatolian unitary *[xw] (Kloekhorst 2006a: 97–101 and 2008: 836–839): Hitt. tarḫu-/taruḫ- [sic!] ‘overcome’ = /tarxw-/ (no variant †tarḫ- exists!); PA *[xw] > Lyc. [kw], as in trqqñt- ‘Stormgod’ < *tr̥h2wn̥ t-´, cognate with Luvo-Hitt. Tarḫunt-. (2) *h3 is preserved in Hittite and Luvian medially as ḫ, at least after sonorant: Hitt. walḫ- ‘strike’ < *welh3- (with LIV2: 679 and Kloekhorst 2008: 946, contra Melchert 1994a: 50 et alibi); CLuv. tarḫ‘break, crush’ (sic!) < *t(é)rh3- ~ Gr. τρώω ‘wound’, etc. (Kloekhorst 2008: 838–839). (3) As a corollary to (1) above, *h3w > PA unitary [ɣw] (spelled in cuneiform) medially (Melchert 2011): Hitt. lāḫw- ‘pour’ < *lóh3w(source of the Core IE ‘root’ *leuh3- backformed from the metathesized preconsonantal zero-grade *luh3-C- < *lh3u-C-, whence Gr. λoέω ‘wash’ etc.). ii. ‘Lenition’ rules of Proto-Anatolian (1) Per Adiego (2001), we may and should assume a single rule of *T > D (including *h2 [x] > [ɣ]) between unaccented morae, with *V̄ equivalent to VV; thus *dhéh1ti > PA *dǽæti > *dǽædi > Lyc. tadi just like abl.-inst. *´-oti > *´-odi > CLuv. -ati, HLuv. /-adi/ ~ /-ari/, Lyc. -edi. (2) Contra Melchert (1994a: 69) this rule includes voiceless stops following *-ā́- < *-éh2-: *-éh2T- > *-áaT- > *-áaD-, as in *mnéh2ti > *mnáati > mnáadi > CLuv. m(a)nāti ‘sees’ (with Starke 1980: 47 and LIV2: 447 contra Melchert 1994a: 236); thus also most economically *-éh2-tr̥ > *-áatr̥ > *-áadr̥ > Hitt. abstract -ātar (contra Melchert 1994a: 86). iii. Alleged examples for a ‘limited Čop’s Law’ in Proto-Anatolian (i.e., *#é. C1- > *#áC1C1-) are now extremely sparse and arguable (on Hitt. ammug

Hittite Historical Phonology after 100 Years

263

‘me’ see below). Existence of such a sound change is thus unlikely (contra Melchert 1994a: 74–75 and 1994b). iv. The three-way contrast of dorsal stops in Luvo-Lycian is due to a conditioned split of palato-velars before their merger with velars, not an ­unconditioned three-way contrast preserved from Proto-Indo-European. Anatolian is thus, per already Hrozný, ‘centum’ (Melchert 2012a). b Vocalism i. Contra Melchert (1994a: 26), Kimball (1999: 79–80), Hoffner and Melchert (2008: 26), et al., cuneiform and are contrastive, with standing for /o(:)/ (or similar) versus for /u(:)/, with Held and Schmalstieg (1969: 105–109), Eichner (1980: 156), Hart (1983: 124–132), Rieken (2005) and above all Kloekhorst (2008: 35–60), who presents the best formulation thus far of the respective prehistoric sources. I insist that the contrast is also valid for Palaic and Luvian with differences only in detail (cf. Melchert 2010a: appendix). Thus: (1) /o:/ < *ow, in mu-u-ga-a-i- ‘incite’ [sic!] < virtual *mowkoye/o(Melchert 2010b, revising Kloekhorst 2008: 586), CLuv. zu-u-wa‘food’ < virtual *gyówh3-o- (Melchert 2012a: 212–213); also u- < *aw ‘away’ in u-i-ya-, originally ‘send/drive away’ [sic!] (Melchert forthcoming). (2) /o(:)/ < *u adjacent to *h2/3: Hitt. coll. pl. āššū (a-aš-šu-u) ‘goods’ < *-uh2; cf. also Hitt. N-ASgNt šu-u-ú ‘full’ /so:(w)u/ < *séwh3-u vs. Pal. šu-ú-na-at ‘filled’ < *su-néh3-t (for accent ‘retraction’ in the latter see Yates 2015: 148–155). (3) /o:/ also from *-óm(s)# > -Cu-u-un in ASg/Pl ku-u-un/ku-u-uš and a-pu-u-un/a-pu-u-uš < *k̂óm(s), *obhóm(s) (Kloekhorst 2008: 54 and 57, revising Melchert 1994a: 186–187). (4) But /u:/ < *ew with Kloekhorst (2008: 53–57), as in -nu-ú- < *-néu(wa-aḫ-nu-ú-mi ‘I turn’ and ḫu-e-nu-ú-ut ‘caused to flee’), ku-ú-ša‘daughter/son-in-law; bride’ < *ĝéuso- *‘chosen one’ (after Rieken 1999: 257), i-ú-uk ‘brace, pair’ < *yéug (Kloekhorst 2008: 423 after Rieken 1999: 61–62). (5) /u:/ also < accented *ú in an original open syllable: Hittite nouns in -ú-ul < *-úlom with syncope (Rieken 2008). (6) Hitt. ú-ug ‘I’ < *ūg(h2) with *ū ← 2Sg subject form *tū, vs. ammug ← 2Sg non-subject *tŭ (revising Melchert 1983: 161–163), contra Simon 2018 et al. ii. There is no basis for an alleged PA phoneme */ẹ̄/ < *ey distinct from */ē/ < *ē (contra Melchert 1994a: 56, after 1984: 102–103, 112–113, and 143). Late

264

Melchert

Hittite ī for ē is analogical, per Yakubovich (2010: 315–318): nīya- for nēya‘turn’ after other ḫi-verbs in -i-; Dat-LocSg kīdani etc. after N-ASgNt kī (note the absence of †apīd- < apēd-). In any case, oblique pronominal -ed- is from *-éd-, not *-é/óyd-; see the concession by Melchert 2008: 369– 370 with references. 2 Significant Retentions I continue to reject some new claims for alleged linguistic contrasts based on orthographic patterns, because I find the contrasts linguistically implausible as formulated thus far. I therefore note here explicitly that I retain some analyses of 1994: i. Consonantism a. By the time of attested Hittite, Luvian, and the other Anatolian languages, word-initial voiced stops (including the reflexes of PIE voiced ­aspirated stops) had all devoiced. However, the different treatment of *#G(h)- in Luvian from that of *#K- shows that this change is a post-Proto-Anatolian areal feature, per Melchert (1994a: 18–20). I reject the implausible claim of a partially preserved contrast in Hittite by Kloekhorst (2010). b. I retain the formulation of ‘Čop’s Law’ in Luvian as given by Čop (1970): *é.C1 > áC1.C1. Contra Kloekhorst (2006b), *ó.T does not regularly lead to Hittite ā.D: see dākki ‘matches’ < *dók̂ei (the root has no final laryngeal, per LIV2: 109) and ḫ(u)wappi ‘throws’ < *h2wópei (also with no final laryngeal; NB Vedic past participle uptá-).3 c. Word-initial *h3- is preserved as /x-/ in Hittite and Luvian (with initial devoicing of obstruents per a.), except /__r. For Hitt. arai- ‘rise’ < *h3róy-ei see Oettinger 2004: 402; see also Hitt. ar- ‘stand (up)’ < *h3ér-tor, *h3r-óntor (after LIV2: 299); phonologically regular but synchronically aberrant *ḫārta, *aranta was leveled to attested ārta, aranta. The loss in both cases may be attributed either to the ‘Saussure-Hirt effect’ (so Oettinger) or a more general Hittite loss of word-initial *h3- before *r. d. There is no evidence for word-initial *h1 preserved as [Ɂ] in Hittite and Luvian, contra Kloekhorst (2004, 2006a: 77–81 and 2008: 205

3 Kloekhorst (2014: 571–574) has made cogent arguments that the match in the stem between Hieroglyphic Luvian Dat-LocPl á-pa-ta-za and Lycian ebette means that ‘Čop’s Law’ is not exclusively Luvian, in which case we must actually define the change as Luvo-Lycian *°ĕ́C1V > *ĕ́C1.C1V.

Hittite Historical Phonology after 100 Years

265

and passim) and Simon (2010 and 2013). On cuneiform #V-VC- spellings see Weeden (2011: 61–68), and on initial a- vs. á- in Hieroglyphic Luvian Melchert (2010c). The contrast of CLuvian a-an-na-an ‘below’ as a freestanding a­ dverb vs. an-na-a-an ti-iš-ša-a-an ‘prepared/ ready below’ (preverb) and an-na-a-an pa-a-ta-an-za ‘under the feet’ (preposition) argues decisively for synchronic lengthening under the accent: /ánnan/ > [á:nnan] vs. /annán/ > [anná:n]. ii. Vocalism a. Proto-Anatolian did have a long vowel distinct from */e:/ and */a:/, conventionally /æ:/, that leads to /e:/ in Palaic and Hittite, but /a:/ in Luvian, Lycian, and Lydian (Melchert 1994a: 56). b. Accented short *ó lengthened in closed syllables in Hittite, but not *á, different from the development in Palaic and Luvian (Melchert 1994a: 146). C Case Study: The PIE Syllabic Sonorants in Anatolian and Hittite There are (at least) four issues to be dealt with concerning the reflexes of the PIE syllabic sonorants in Anatolian and Hittite. First, how late was *R̥ preserved? Second, what vowel was inserted in the change of *R̥ > *VR? Third, what was the result of *wR̥ (and *KwR̥ ) between consonants (with word boundary counting as C)? Fourth, does *N̥ always yield Hittite aN? I will address these questions in reverse order, since the answers to the last two questions constrain those to the first two. 1. The last question may now be definitively answered as: no. Already Hrozný (1917: 187) gave the unmarked result as an (anz- ‘us’ < *n̥ s-), but he also entertained a as the outcome before voiceless stop in katta ‘down’ < *k̂m̥ta (1917: 32 and 187) and akk- ‘die’ < *ŋ̥k̂-´ (1917: 174 with reservations). The demonstration by Goedegebuure (2010: 301–312) that CLuvian zanta means ‘down’ ­confirms that Hittite katta reflects *k̂m̥ta. For support for the derivation of akk‘die’ < *ŋ̥k̂-´ based on the weak stem see Melchert 2012b: 180–182, with note 15. The conditioning given there for prehistoric syllabic nasals homorganic with following stops is that they yield simple aN when accented and a when unaccented. This is compatible with what little relevant evidence is available, but examples are so few that this formulation must be regarded as merely provisional. However, the accompanying account for prehistoric sequences of syllabic nasal followed by nonhomorganic stop is problematic in two respects. First, *ń̥-mh2yent- > amiyant- ‘immature’ (see Kloekhorst 2008: 172 for this shape as the regular outcome) cannot be cited as an example, since it surely passed through a stage *Vn-myant-, and the loss of the *n is part of a broader deletion

266

Melchert

of the dental nasal before *m and *w in Hittite: NB especially Hitt. kuemi ‘I kill’ < *gwhénmi, where no syllabic nasal is involved (see Melchert 1994a: 168 for further examples). Second, despite my blithe statement in the footnote cited, it is hardly satisfying to suppose that syllabic nasals heterorganic with following stops lead to precisely the opposite results posited for their homorganic counterparts, namely a when accented, aN when not. With the irrelevant example of *ń̥-mh2yent- > amiyant removed, we are left only with cases involving putative *m̥ before nonlabial stop: katta ‘down’ < *k̂m̥ta/ō (~ Gr. κατά or κάτω), katta ‘beside’ < *kḿ̥tV[+back] (with allomorph *kḿ̥ti in katti- ~ Gr. κασι- and Middle Welsh gennyf ‘with me’),4 and antarā- ‘blue’ < *m̥ dhró- (thus with Kloekhorst 2008: 186 contra Melchert 1994a: 1215). The preforms of katta ‘down’ and katta ‘beside’ must have been accented as given, since an accent on the final syllable could only lead to †kattā. To derive the adverbs from unaccented variants (Kloekhorst 2008: 604) is egregiously ad hoc, since other local adverbs clearly reflect accented forms, e.g. Hitt. š(a)rā from accented *sr-ó (thus also Kloekhorst 2008: 730). In any case, this would not account for the difference between the adverbs and antarā-, where the syllabic *m̥ would also have been unaccented. What does condition the different outcome in the two adverbs versus the adjective is an open question, but it is far from assured that it is to be attributed to the accent. The inherited syllabic *m̥ in the adverbs is assured by the Greek and Celtic cognates, but that in andarā- is merely inferred. Dare we suppose that Hittite reflects rather *modhró- like its Slavic cognates and that a prehistoric syncope led only secondarily to an *m̥ d- sequence? The ad hoc nature of this suggestion is manifest, and I cite the alternative merely to show that we do not by any means control the details of the development of *m̥ plus heterorganic stop. Nevertheless, the different result in Hitt. katta and CLuvian zanta does assure that syllabic nasals were preserved into the prehistory of the individual languages at least in nonfinal syllables and that the Hittite result of *N̥ is in some instances merely a. 2. Per Melchert (1994a: 126–127) following Schindler, *wR̥ resyllabified between consonants or consonant and word boundary as uR. Such a change is possible, but unmotivated. Further, it is now clear that u-ur-ki- ‘track, trail’ < *wŕ̥gi- is /ó:rgi-/, so that we are not dealing with a simple resyllabification *wr̥ > ur. There is other evidence for the lowering of *u > /o/ before r in Hittite:

4 Contra Dunkel 2014: 2.424 and 426, the clear association with katta would have surely blocked assibilation in *kḿ̥ti. There is no justification for doubting that the Greek, Welsh, and Hittite forms are cognate.

Hittite Historical Phonology after 100 Years

267

contrast Hitt. iš-nu-u-ra- ‘kneading tray’ < *isn-úro- with Pal. ta-šu-ú-ra‘sacrificial table’ < *dhh1s-úro-. Note also ku-u-ur-ka- ‘foal’ /ko:rka-/ < *kúrko- ~ Gr. κύρνος (Forssman 1980) and see Rieken 2005: 540–542 and Kloekhorst 2008: 55–56. However, we also find pít-tu-u-la- ‘snare, loop’ /pit:o:la-/ hypostasized from *peth2-wl ̥ ‘thing spread’ (after Rieken 1999: 471–472 and Puhvel 1979: 211 and 2011: 71). Thus instead of resyllabification, we should assume rather *CwR̥ > *CwoR > CoR (under the accent Co:R). That is, *R̥ > *oR, and in the presence of a preceding labial glide the *o is continued as the new vowel /o(:)/. The conditioning *w is then lost by dissimilation. Likewise in Hittite *KwR̥ > KwoR > KoR: hence the weak stem *kwl ̥s- > ‘incise’ > Hitt. gulš- /ko:ls-/ (etymology after Eichner 1974: 67–68).5 There is, however, a conditioned exception to the rule just given before two consonants, where the result is CwaR and KwaR with regular lowering of *o to a, as shown by Kloekhorst (2007): duwarni/a- ‘break’ < *dhwr̥-né-h1-, dhwr̥-n-h1-´ (with leveling of the phonologically regular result *dornizzi, *dwarnanzi to d(u)warnizzi, d(u)warnanzi), k(u)war(a)ške- ‘cut’ < *kwr̥-sk̂e-, k(u)waške- ‘slay’ < *gwhn̥ -sk̂e-. Support for the assumption that it was *o that was inserted comes from occasional for simple *R̥ instead of aR in Luvian and Palaic: note LuvoHittite gurta- /kort/da-/ ‘citadel’ < *ĝhr̥-to- or *ĝhr̥dh-o- and HLuv. /tsornid-/ ‘horn’ < *k̂r̥ng-id- cognate with Hittite *karkid- in karkidant- ‘horned’, both to the base *k̂r̥ngo- of Skt. śŕ̥ṅga-. Also likely is Palaic ēšḫur ‘blood’ < *h1ḗsh2r̥ (see already Melchert 1994a: 260 and 214, but with no independent support). Lycian umlaut and syncope make it hard to determine whether *R̥ > oR in nonfinal syllables is Proto-Anatolian or a Hittite-Palaic-Luvian isogloss. But if we posit PIE *h1mé for the non-subject first person singular pronoun (see now Simon 2012: 488–491 for further arguments in favor of the initial laryngeal), then we would predict existence of a Lindeman variant *h1m̥ mé, whence with secondary u-vocalism from second singular *tŭ PA *h1m̥ mú. If we assume that this became *h1ommú with *m̥ > *om, this PA preform would lead regularly to Lyd. amu (unaccented ẽ > a, per Eichner 1986: 211–212), Hitt. ammug, and Lyc. e/ẽmu (also with umlaut amu). One should note that in absolute final position, Hittite athematic preterite first singular -un < *-m̥ vs. nom.-acc. sg. neuter n-stem ending -an < *-n̥ clearly contrasts with Lycian -ã from both *-m̥ (consonant stem acc. sg. -ã as in lãtã ‘dead’) and *-n̥ (hrm̃ mã ‘temenos, land section’ *oR, if it truly was 5 The absence of spellings with scriptio plena directly showing the long vowel is due to the exclusive spelling of the verb stem with the cvc sign . The word is unattested in OS.

268

Melchert

­ roto-Anatolian, was limited to nonfinal syllables, while the differing results in P ­Hittite and Lycian require that syllabic nasals remained in at least word-final position into the prehistory of the individual languages. 3. The precise Hittite result of word-final *-m̥ # and the entire question of the animate accusative plural ending -uš present intractable problems. As the very thorough treatment by Kloekhorst (2008: 42–43, 56, and 609) makes abundantly clear, the evidence is limited and frustratingly contradictory. A fully satisfying solution is at present beyond reach, but at least a new attempt seems called for.6 We may begin with the result of final *-m̥ #. As outlined in Melchert (1994a: 181 with refs.), the preterite first singular ending -un in root verbs of the miconjugation can hardly represent anything but*-m̥ # (thus also most recently Kloekhorst 2008: 609). The athematic nominal animate accusative singular ending -an must then be analogical to the thematic ending -an < *´-om. Per above, following Kloekhorst (2008: 586), the regular outcome of final accented *-óm was /-o:n/, preserved in the demonstratives ku-u-un /kó:n/ ‘this’ and a-puu-un /abó:n/ ‘that’. Given the very sparse number of reflexes of oxytone o-stems attested in Hittite, it is unsurprising that the result -an of the unaccented ending was generalized to all o-stems, from which it was further spread to athematic stems. Harder to determine is whether graphic final represents /-on/ or /-un/. Kloekhorst (2008: 609) cites occasional New Hittite (NS) spellings of ‘I took’ as e-ep-pu-u-un in favor of reading /-on/ as the regular reflex of *-m̥ #. However, elsewhere (2008: 42–43), he acknowledges that the situation is more complicated: ‘I went’ is spelled a few times pa-a-ú-un in Middle Hittite manuscripts (MS), but pa-a-u-un in NS. There he interprets this as showing a change from OH /pá:un/ to NH /pá:on/, implying that the OH result of *-m̥ # was /-un/. However, we should ask ourselves in the first place why an unaccented vowel that was surely short was written with a ‘plene’ spelling at all. The dominant spelling of ‘I went’ is pa-a-un at all times: even in New Hittite compositions it occurs more than 100 times vs. only 9 instances of pa-a-u-un, all of which to my knowledge are in texts of Muršili ii and Hattušili iii. Likewise, the normal spelling of ‘I took’ is the expected e-ep-pu-un, including 19x in the New Hittite corpus vs. only 5x for e-ep-pu-u-un, all in texts of Hattušili iii. By any derivation, Hittite pāun represents a preform in which there was loss of an intervocalic *y and a resulting hiatus (see for one account of ‘go’ 6 The following remarks reflect further research undertaken since the oral presentation of this paper. They are presented as, and should be received as, merely tentative suggestions towards an ultimate solution!

Hittite Historical Phonology after 100 Years

269

­Kloekhorst 2008: 617–618).7 Kloekhorst (2012: 248–250) has argued persuasively that in sequences of Ca-e-eC (as well as Ci-e-eC and Cu-e-eC) the -e- is not a ‘plene’ spelling indicating length of the /e/ vowel, but rather a mark of a preceding /j/ glide. Thus a nominative plural t/da-lu-ga-e-eš ‘long’ spells [talugajes] with a new hiatus-filling [j] (see also Kloekhorst 2014: 136–144 with full evidence). When we also find in OS an accusative plural ta-lu-ga-ú-uš (KBo 17.22 iii 6) where a long unaccented vowel is unmotivated, I suggest that at least some Hittite speakers likewise filled the hiatus resulting from loss of *y in a sequence [a.u/o] with a glide [w] ([talugawu/os]), which OH scribes wrote with . Similarly, MS pa-a-ú-un spells [pá:wu/on].8 I underscore that in both cases there is no basis for supposing that the unaccented vowel of the final syllable is long. As noted, Kloekhorst (2008: 42–43) interprets the spelling change of pa-a-úun to pa-a-u-un as a real change of /-un/ to /-on/, but offers no motivation for such a change. One possibility for the orthographic change is that the spelling of intervocalic [w] here was adjusted to that in medial sequences of [awa]. At all periods of Hittite there were next to the normal spellings with also some with , whereas to my knowledge spellings with are vanishingly rare. However, the spelling e-ep-pu-u-un suggests another possibility. The ‘plene’ spelling here, which again cannot represent a true long vowel, is surely modeled on pa-a-u-un (with which it cooccurs in the same manuscripts), where the has replaced earlier in marking the hiatus-filling [w]. However, it is hard to see what would have led a scribe to imitate the spelling in /é:p:Vn/, which had no [w], unless he also associated the in pa-a-u-un with the vocalism of the ending. I therefore conclude with some reservation that the of both pa-a-u-un and e-ep-pu-u-un does indirectly tell us that the ending was /-on/, the regular result of *-m̥ #. I believe that the overall facts suggest a similar account for the animate accusative plural ending, but one must openly acknowledge that no reasonable scenario can explain all of the attested examples, so one must inevitably 7 I hereby explicitly withdraw my own derivation (1994a: 177) via a nonexistent preverb *pe. I would differ from Kloekhorst only in supposing that with an accented preverb the root was in the zero grade, thus leading to the consistently short secondary diphthong in the OH strong stem before consonant: virtual *póy-h1i-mi *[pój.Ɂi.mi] > *[pó.ji.mi]> *[pó.i.mi] > pa-imi ‘I go’ [pój.mi]. However, in the Pret1Sg the result was a long vowel from accented *ó in an open syllable: *póy-h1  y- [pój.ɁjVm] > *[pój.jVm] > *[pó.jVm] > *[pó:.jVm] > pa-a-un. 8 Kloekhorst (2014: 138–139) shows that the spelling Ca-e-eš is the dominant one, whereas spellings Ca-ú/u-uš are relatively rare (see further below). This difference may well reflect that the hiatus filling in [aje] was far more prevalent/regular than that in [awu/o], but this does not prove that the latter is not real.

270

Melchert

d­ ismiss some spellings as erroneous. The degree of arbitrariness in the latter procedure leaves any analysis less than fully satisfying. We must take into account no less than four possible preforms: accented *-óms, unaccented *´-oms, athematic *-m̥ s, and u-stem *-ums. We have seen that the first yields Hittite /-o:s/, as in the demonstratives ku-u-uš ‘these’ and a-pu-u-uš ‘those’, but the scarcity of oxytone o-stem nouns in Hittite makes this an unlikely source for the general ending spelled . We find a few spellings in both and , but not all have the same status:9 those that reflect athematic *-m̥ s can hardly show genuine ‘plene’ spellings indicating a long vowel. The distribution of and in athematic stems is highly suggestive. First of all, we find only in i-stems: an-na-al-li-ú-u[š] (KUB 51.47 Vo 4; ?/NS); a-ú-li-ú-uš (KBo 25.178 i 2; OH/NS & KUB 24.3 ii 11; MH/NS), a-ú-li-úš(a) (KUB 17.21 ii 18 MS/MS); NINDAḫar-ša-ú-uš (KBo 17.4 ii 17; OH/OS); kap-píú-uš (KBo 34.47 ii 8; MH/MS); ku-i-ú-uš (HKM 23:9; MH/MS; KBo 18.57a+57 Ro 2.Vo 42; MH/MS); ma-ši-ú-u[š]? (KBo 9.109 Vo 4; OH/NS); pu-u-ri-ú-uš (KBo 19.163 i 23.iv 4; OH/NS); šu-up-pí-ú-uš (KUB 33.41 ii 10; OH/NS); ta-lu-ga-ú-uš (KBo 17.22 iii 6; OH/OS). One may note that all of these examples are either in OS, MS, or NS copies of OH compositions. As argued above, I regard all of these as spelling a hiatus-filling [w]: [kwiwVs], [xarsawVs], etc. They therefore can tell us nothing about the quality of the vowel of the ending. The idea that the is spelling [w] is supported by the complete absence of any spellings for the accusative plural of athematic stems ending in a consonant (non-glide). For the diphthongal stem lingāi- ‘oath’ one may cite the spelling li-in-ga-u!uš at KBo 4.4 i 45 (see CHD L-N: 64 for the reading) in the Annals of Muršili ii, a New Hittite text. I tentatively regard this as showing the same replacement as in NH pa-a-u-un for earlier pa-a-ú-un. That is, it is fundamentally spelling the same hiatus-filling [w] as in the OS examples with -Ca-ú-uš, but also has been adjusted to the vocalism of the ending, thus implying [lingá:wos]. Second, for u-stems we find mostly . Most importantly, we find it in a-ku-u-uš-(ša) ‘(also) seashells’ to aku- (KBo 19.156 ii 17; OH/OS), where it may reflect a genuine plene spelling for an accented long vowel: [agó:s]. Less straightforward are the spellings [i-da]-a-la-mu-u-uš (KUB 8.67:14; MH/NS) and [i-da-(a)]-la-mu-u-š(a) (KBo 15.10 iii 54; OH/MS), and the faulty pár-ga-uuš (KBo 3.8 iii 22; OH/NS). The last example is likely to be modeled on an

9 Forms cited are taken from Kloekhorst (2008: 56), supplemented by further examples from my own files.

Hittite Historical Phonology after 100 Years

271

i-stem plural of the type of li-in-ga-u-uš. Likewise the aberrant plurals ḫal-luwa-u-uš (KBo 26.135:2; OH/NS) and [ḫal]-lu-ú-wa-u-uš (KBo 3.8 iii 4; OH/NS) to ḫalluwa- ‘deep’.10 It is less clear to me whether the instances with -Ca-mu-uuš are analogical to a-ku-u-uš or to the i-stem plurals. What evidence we have suggests then that *-ums led to Hittite /-os/. I take the solitary example of ḫe-e-mu-ú-uš to ḫēu- ‘rain’ (KBo 43.137:7; ?/NS) as erroneous. The most confusing picture is that presented by the oxytone a-stems: al-puú-uš (KUB 28.5 Vo 7; OH/NS); ir-ḫu-ú-š(a) (KUB 31.128 i 3; pre-NH/NS); MUNUS.MEŠ kat-ru-ú-uš (KUB 54.66 Vo? 13; OH/NS); and iš-ḫu-u-uš (KBo 15.31 i 14; OH/NS). The last is nonprobative because after ḫ Hittite regularly has only /o/, even from prehistoric *u (see Rieken 2005: 539 and Kloekhorst 2008: 51). This also means, however, that ir-ḫu-ú-š(a) must be an erroneous spelling (compare the hapax ḫu-ú-ni-ik-zi at KBo 6.2 i 16 OH/OS cited by Kloekhorst). Since we also find rare exceptions even for the accusative plural of the demonstratives (a-pu-ú-uš at KUB 14.14 Ro 21; NH), we can hardly put much weight on the hapax legomena al-pu-ú-uš and kat-ru-ú-uš. We seem thus to find /-os/ as the result of *-ums and probably also of *-m̥ s. The first is compatible with other evidence for lowering of prehistoric *u to /o/ before tautosyllabic nasal (see Rieken 2005: 540 and compare Kloekhorst 2008: 54), while the latter is at least consistent with the outcome of absolute wordfinal *-m̥ discussed above. Despite some spellings of oxytone a-stem accusative plurals with -Cu-ú-uš, the evidence of the demonstratives argues that *-óms also led to /-o:s/. We are left only with the question of the outcome of unaccented final *´-oms. The regular ending in Hittite is of course , as for all other stems. Is this the phonologically regular result, and does it also represent /-os/ or could it be /-us/? The latter seems unlikely: it would suppose that while *-ums was lowered to *-oms and accented *-óms was maintained, just unaccented *´-oms was raised to *-ums. Whether /-os/ was also the regular result of *´-oms is harder to answer, but I would like to suggest that there may be relevant evidence. For the word derived from wag- ‘bite’ that probably refers to a bite-sized kind of bread (thus ‘roll’ or similar), there is clear evidence for an animate a-stem NINDAwagāta- (NSg wagātaš OH/OS, ASg wagatan OH/NS, CollPl wagāta OH/NS), as correctly identified by Hoffner (1974: 188). However, Neu (1983: 208), Rieken (1999: 196– 197), and Kloekhorst (2008: 940) all assume a primary s-stem wagātaš- from which a New Hittite a-stem was somehow abstracted. 10

We would have expected *ḫallumuš for *ḫalluw-uš; compare nemuš for *new-us to newa- ‘new’.

272

Melchert

An s-stem is highly improbable. First of all, the formation of collective plurals to animate nouns is found only in Old Hittite texts (see Melchert 2000: 62–65 with note 36). The plural wagāta cannot therefore be a New Hittite creation. Second, while Rieken (1999: 197) concedes that there is no other support for a supposed formation in *-eh2-tos, Hittite and Luvian provide ample evidence for secondary stems in *-o-to- (Melchert 1999: 368–372). Since the singular may be used in Hittite with numbers higher than one, examples such as KUB 25.9 iii 21–22 (OH/NS) are completely ambiguous between an a-stem and an s-stem (pace Neu loc. cit.): 4 wagataš=ššan kitta ‘four w.-breads are lying on it’. The entire basis for an s-stem consists of examples such as KBo 20.33 Ro 12 (OH/OS): [LÚKA]Š4.E taruḫzi kuiš 1 MA.NA KÙ.BABBAR U 2 NINDAwagadaš pianzi ‘To the runner who wins they give one mina of silver and two w.-breads’. While Hittite may use the nominative case as the ‘default’ case in lists (Hoffner and Melchert 2008: 243), that usage cannot apply here, where both nouns are clearly the direct object of the verb. However, such examples do not justify the ad hoc assumption of an s-stem that is morphologically unparalleled in Hittite instead of the well supported a-stem. We need only suppose an animate accusative plural with an archaic ending -aš < *´-oms: the plene spelling in wagāta- suggests that the accent was on the penultimate syllable.11 In the absence of other examples for an OH accusative plural ending -aš I certainly do not insist on this analysis, but it eliminates the unmotivated s-stem and is not contradicted by any facts.12 In conclusion, we may admire the impressive first step in delineating Hittite historical phonology achieved by Hrozný in 1917, based on a very limited and still imperfectly understood Hittite text corpus. We must also humbly concede that by no means have all of the issues yet been resolved after a full century, and much work remains to be done. References Adiego Lajara, I.-J. (2001) Lenición y acento en protoanatólico, in: O. Carruba and W. Meid (eds.), Anatolisch und Indogermanisch – Anatolico e indoeuropeo [IBS 100], 11 12

However, since a preceding accented short *-ó- would not have ‘lenited’ a following *-t(see above), one should likely follow Rieken (1999: 197) in supposing that the immediate base of the *-to- stem was a collective *we/og-eh2-. This putative archaic thematic APlC ending -aš < *-oms may be compared with the equally rare thematic NPlC ending -aš < *-ōs attested once in [ḫante]zziyaš (KBo 22.2 Ro 18; OH/OS) and likely also in gaenaš=šeš ‘his inlaws’ (Telipinu Edict passim; OH/NS), on which see Otten (1973: 34–35), following for the latter Kammenhuber.

Hittite Historical Phonology after 100 Years

273

Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck, 11–18. Adiego Lajara, I.-J. (2007) The Carian Language [HdO I/86], Leiden, Boston: Brill. Čop, B. (1970) Eine luwische orthographisch-phonetische Regel, IF 75, 85–96. Dunkel, G. (2014) Lexikon der indogermanischen Partikeln und Pronominalstämme, 2 vols., Heidelberg: Winter. Eichner, H. (1974) Untersuchungen zur hethitischen Deklination (Dissertation Teildruck), Erlangen. Eichner, H. (1980) Phonetik und Lautgesetze des Hethitischen – Ein Weg zu ihrer Entschlüsselung, in: M. Mayrhofer, M. Peters and O.E. Pfeiffer (eds.), Lautgeschichte und Etymologie: Akten der vi. Fachtagung der Indogermanischen Gesellschaft. Wien, 24.–29. September 2008, Wiesbaden: Reichert, 120–165. Eichner, H. (1986) Neue Wege im Lydischen I: Vokalnasalität vor Nasalkonsonanten, ZVS 99, 203–219. Eichner, H. (1992) Anatolian, in: J. Gvozdanović (ed.), Indo-European Numerals, Berlin, New York: Mouton de Gruyter, 29–96. Forssman, B. (1980) Hethitisch kurka- Comm. ‘Fohlen’, ZVS 94, 70–74. Frantíková, D. (2015) Spatial cases in Hittite: 100 years since Bedřich Hrozný’s ­deciphering of the Hittite language, Slovo a Slovesnost 76/3, 215–231. Goedegebuure, P. (2010) The Luwian adverbs zanta “down” and *ānni “with, for, against”, in: A. Süel (ed.), Acts of the viith International Congress of Hittitology, ­Çorum, August 25–31, 2008, Ankara: T.C. Çorum Valiliği, 299–318. Hart, G. (1983) Problems of writing and phonology in cuneiform Hittite, TPhS 81, 100–154. Held, W. and Schmalstieg, W. (1969) Some comments on the Hittite phonological system, General Linguistics 9, 93–110. Hoffner, H.A., Jr. (1974) Alimenta Hethaeorum, New Haven: American Oriental Society. Hoffner, H.A., Jr. and Melchert, H.C. (2008) A Grammar of the Hittite Language, Part I: Reference Grammar, Winona Lake, IN: Eisenbrauns. van den Hout, T.P.J. (2009) A century of Hittite text dating and the origins of the Hittite cuneiform script, IncLing 32, 11–35. Hrozný, B. (1917 [2002]) Die Sprache der Hethiter [BoSt 1–2], Leipzig: Hinrichs [reprinted as DBH 3, Dresden: Verlag der TU Dresden]. Kimball, S. (1999) Hittite Historical Phonology [IBS 95], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Kloekhorst, A. (2004) The preservation of *h1 in Hieroglyphic Luwian: two separate a-signs, HS 117, 26–49. Kloekhorst, A. (2006a) Initial laryngeals in Anatolian, HS 119, 77–108. Kloekhorst, A. (2006b) Čop’s Law in Luwian revisited, Die Sprache 46, 131–136.

274

Melchert

Kloekhorst, A. (2007) The Hittite syllabification of PIE *CuR and *K u̯ R, in: D. Groddek and M. Zorman (eds.), Tabula Hethaeorum: Hethitologische Beiträge Silvin Košak zum 65. Geburtstag, Wiesbaden: Harrassowitz, 455–457. Kloekhorst, A. (2008) Etymological Dictionary of the Hittite Inherited Lexicon [LIEEDS 5], Leiden, Boston: Brill. Kloekhorst, A. (2010) Initial stops in Hittite (with an excursus on the spelling of stops in Alalaḫ Akkadian), ZA 100, 197–241. Kloekhorst, A. (2012) The phonological interpretation of plene and non-plene spelled e in Hittite, in: B. Nielsen Whitehead et al. (eds.), The Sound of Indo-European: Phonetics, Phonemics and Morphophonemics, Copenhagen: Museum Tusculanum, 243–261. Kloekhorst, A. (2013) The signs TA and DA in Old Hittite: evidence for a phonetic difference, AoF 40, 125–141. Kloekhorst, A. (2014) Accent in Hittite: A Study in Plene Spelling, Consonant Gradation, Clitics, and Metrics [StBoT 56], Wiesbaden: Harrassowitz. Melchert, H.C. (1983) The second singular personal pronoun in Anatolian, MSS 42, 151–165. Melchert, H.C. (1984) Studies in Hittite Historical Phonology, Göttingen: Vandenhoeck & Ruprecht. Melchert, H.C. (1994a) Anatolian Historical Phonology [LSIE 3], Leiden, Boston: Brill. Melchert, H.C. (1994b) “Čop’s Law” in Common Anatolian, in: J. Rasmussen (ed.), In honorem Holger Pedersen: Kolloquium der indogermanischen Gesellschaft vom 26. bis 28. März 1993 in Kopenhagen, Wiesbaden: Reichert, 297–306. Melchert, H.C. (1999) Two problems of Anatolian nominal derivation, in: H. Eichner and H.C. Luschützky (eds.), Compositiones indogermanicae in memoriam Jochem Schindler, Prague: Enigma, 365–375. Melchert, H.C. (2000) Tocharian plurals in -nt- and related phenomena, TIES 9, 53–75. Melchert, H.C. (2008) Problems in Hittite pronominal inflection, in: A. Lubotsky, J. Schaeken and J. Wiedenhof (eds.), Evidence and Counter-Evidence: Essays in Honour of Frederik Kortlandt, Volume 1: Balto-Slavic and Indo-European Linguistics, Amsterdam, New York: Rodopi, 367–375. Melchert, H.C. (2010a) The verbal prefix “u-” and vs. spellings in Anatolian cuneiform (paper presented at the 29th East Coast Indo-European Conference, Cornell University, June 19, 2010). Melchert, H.C. (2010b) On Hittite mūgā(i)-, in: R. Francia and G. Torri (eds.), Studi di ittitologia in onore di Alfonso Archi (= Orientalia 79/2), 207–215. Melchert, H.C. (2010c) Spelling of initial /a-/ in Hieroglyphic Luwian, in: I. Singer (ed.), ipamati kistamati pari tumatimis. Luwian and Hittite Studies Presented to J. David Hawkins on the Occasion of His 70th Birthday, Tel Aviv: Tel Aviv University, 147–158.

Hittite Historical Phonology after 100 Years

275

Melchert, H.C. (2011) The PIE verb for ‘to pour’ and medial *h3 in Anatolian, in: S.W. Jamison, H.C. Melchert and B. Vine (eds.), Proceedings of the 22nd Annual UCLA Indo-European Conference, Bremen: Hempen, 127–132. Melchert, H.C. (2012a) Luvo-Lycian stops revisited, in: R. Sukač and O. Šefčík (eds.), The Sound of Indo-European 2: Papers on Indo-European Phonetics, Phonemics, and Morphophonemics, Munich: Lincom Europa, 206–218. Melchert, H.C. (2012b) Hittite ḫi-verbs of the type -āC1i, -aC1C1anzi, IF 117, 173–186. Melchert, H.C. (forthcoming) Hittite and Luvian uppa- and Hittite uiya-, in: A. Süel (ed.), Acts of the ixth International Congress of Hittitology, Çorum, 1.–7. September, 2014. Neu, Erich (1983) Glossar zu den althethitischen Ritualtexten, Wiesbaden: Harrassowitz. Oettinger, N. (2004) Die Entwicklung von *h3 im Anatolischen und hethitisch arāi „­erhebt sich‟, in: A. Hyllested et al. (eds.), Per aspera ad asteriscos: Studia indogermanica in honorem Jens Elmegård Rasmussen sexagenarii Idibus Martiis anno MMIV [IBS 112], Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck, 397–405. Otten, H. (1973) Eine althethitische Erzählung um die Stadt Zalpa, Wiesbaden: Harrassowitz. Puhvel, J. (1979) Hittite words with the initial pít/pát sign, in: W. Meid and E. Neu (eds.), Hethitisch und Indogermanisch [IBS 25], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck, 209–217. Puhvel, J. (2011) Elliptical genitives and hypostatic nouns in Hittite, Aramazd 6/2, 68–72. Rieken, E. (1999) Untersuchungen zur nominalen Stammbildung des Hethitischen [StBoT 44], Wiesbaden: Harrassowitz. Rieken, E. (2005) Zur Wiedergabe von hethitisch /o/, in: G. Meiser and O. Hackstein (eds.), Sprachkontakt und Sprachwandel: Akten der xi. Fachtagung der Indogermanischen Gesellschaft, 17.–23. September 2001, Halle an der Saale, Wiesbaden: Reichert, 537–549. Rieken, E. (2008) The origin of the -l genitive and the history of the stems in -īl- and -ūl- in Hittite, in: K. Jones-Bley et al. (eds.), Proceedings of the 19th UCLA Indo-­ European Conference Los Angeles, November 3–4, 2007, Washington, DC: Institute for the Study of Man, 239–256. Rieken, E. (2010a) Das Zeichen im Hieroglyphen-Luwischen, in: Y. Cohen, A. Gilan, and J.L. Miller (eds.), Pax Hethitica. Studies on the Hittites and Their Neighbours in honour of Itamar Singer, Wiesbaden: Harrassowitz, 301–310. Rieken, E. (2010b) Das Zeichen im Hieroglyphen-Luwischen, in: A. Süel (ed.), Acts of the viith International Congress of Hittitology. Çorum, August 25–31, 2008, Ankara: T.C. Çorum Valiliği, 651–660.

276

Melchert

Rieken, E. (2015) Bemerkungen zum Ursprung einiger Merkmale der anatolischen ­Hieroglyphenschrift, WdO 45, 216–231. Simon, Z. (2010) Das Problem der phonetischen Interpretation der anlautenden scriptio plene im Keilschriftluwischen, in: L.E. Kogan et al. (eds.), Language in the Ancient Near East: Proceedings of the 53e Rencontre Assyriologique Internationale, Winona Lake, IN: Eisenbrauns, 249–265. Simon, Z. (2012) PIE ‘me’ and a new Lydian sound law, in: B. Nielsen Whitehead et al. (eds.), The Sound of Indo-European: Phonetics, Phonemics and Morphophonemics, Copenhagen: Museum Tusculanum, 485–499. Simon, Z. (2013) Once again on the Hieroglyphic Luwian sign *19 , IF 118, 1–21. Simon, Z. (2018) Das Hethitische und der grundsprachliche Vokalismus des Personalpronomens der 1. Sg., in E. Rieken (ed.), 100 Jahre Entzifferung des Hethitischen: ­Morphosyntaktische Kategorie in Sprachgeschichte und Forschung. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 21. bis 23 September 2015 in Marburg, Wiesbaden: Reichert, 355–362. Starke, F. (1980) Keilschriftluwisch manā-tti ‘sehen’, mammanna-i ‘schauen’, Kadmos 19, 142–148. Weeden, M. (2011) Review of Kloekhorst 2008, BSOAS 74, 59–76. Yakubovich, I. (2010) Sociolinguistics of the Luvian Language [BSIELL 2], Leiden, ­Boston: Brill. Yates, A. (2015) Anatolian default accentuation and its diachronic consequences, IndoEuropean Linguistics 3, 145–187.

Chapter 13

duttarii ̯ata/i- and Some Other Indo-European Maidens MUNUS/f

Veronika Milanova Abstract The word MUNUSduttarii̯ata/i- discovered in three Hittite texts (KUB 22.40 iii 18, Bo 4120 r. col. 4, and KBo 24.126 obv. 28) is a valuable and fascinating attestation because it may be a reflex of *dhugh2tḗr, one of the few PIE kinship terms preserved in Anatolian. However, the exact meaning, function and origin of this word in Hittite are questions of debate. The determinative MUNUS indicates that the word cannot be interpreted simply as an appellative ‘daughter’ but must be either a profession or a personal name of a woman. Admitting that duttarii̯ata/i- could be a PN, I argue that this word could also be a sacred title designating a maiden/virgin priestess, a post for well-born girls, which existed in some IE cultures.

Keywords Hittite – Luwian – Greek – female priesthood – kinship terms – age grades – semantic change

1 Morphology According to Starke (1987: 253–254), the base of the borrowed Hittite form duttarii̯ata/i- must be Luwian (*)duttar(i)1 ( kinship term improper’12 is fairly rare13 11 12 13

Taggar-Cohen (2006, 364) suggests interpreting the element AMA (= anna) in this compound not as ‘mother’ but as ‘mature/fertile female’. This term can refer to such words as ‘boy’, ‘girl’, ‘man’, ‘woman’, ‘bride’, ‘bridegroom’ i.e., words that can potentially become kinship terms or can be used as such in some contexts (Dahl & Koptjevskaja-Tamm 2001: 3–4). Semantic change in kinship terms and their interaction with words from neighbouring semantic fields is still a research gap in linguistics. That is why I have almost nothing to

284

Milanova

and usually takes place in languages whose natural, i.e. internal development has been heavily disturbed by some external influence. NPers. (doxt)ar ‘girl/ daughter’ and pesar ‘boy/son’ (Junker and Alavi 1967: 301–302) is one example of this semantic extension,14 which most likely took place due to hybridisation of New Persian, its massive extensive and profound contact with neighbouring languages (Gernot Windfuhr, p.c.) such as Arabic (bint and ibna ‘girl/daughter’, Cameron 1979: 28), Turkic (Azeri qiz ‘girl/daughter’, oğlan ‘boy/son’, Rahmati 1999: 444 and 501) and Mongol (oxin ‘girl/daughter’, Vietze 1981: 234 and 352). It is noteworthy that such a linguistic situation is typologically similar to the one in the Hittite kingdom. For languages which have not undergone such extensive contact, a semantic shift / meaning extension ‘kinship term improper > kinship term proper’ is more common. The most well-known and almost universal example is the shift ‘(adult) woman > wife’ (Zalizniak 2008: 228). The development ‘girl > daughter’ seems less obvious from the viewpoint of most modern cultures. However, in the context of a traditional (androcentric) society, in which a woman was associated either with her husband or with her father, it is not surprising that a woman’s age grade could also be used as a kinship term and a designation of her marital status.15 Therefore, the third possible answer to the question formulated above could be that the Anatolian continuant of *dhugh2ter- was primarily used as an agegrade term (or had a hybrid meaning ‘girl/daughter’ just like NPers. doxtar). HLuw. tuwatri- (TELL AHMAR §§24, 29) and Lyc. kbatra (Hawkins 1978) cannot be viewed as counterexamples here. In the texts in which they occur they without doubt mean ‘daughter’, but there are so few attestations that one can only guess what meaning these words could have had in lost texts and in the everyday speech of the Luwians and the Lycians. This third assumption, which would imply that Anatolian inherited the form *dhugh2ter- in a slightly different meaning from the other IE branches, would be especially fascinating for

14 15

quote here apart from Dahl & Koptjevskaja-Tamm (2001), Zalizniak (2008), Hentschel (2012) and my own forthcoming article ‘Semantic change in terms of social relation’ (working title), in which I analyse this phenomenon from the point of view of both linguistics and social anthropology, including empirical data. The assumptions about the direction of semantic change (in the main body of the text above) reflect the preliminary results of my research. In Pahlavi we find kanīk/g ‘girl’ and rēdak ‘boy’ (Rastorguyeva 1966: 30, 46, 54, 69; MacKenzie 1986 [1971]: 71). As was also correctly and eloquently expressed in B.J. Collins’ talk ‘Virginity in Hittite ritual’, ‘She was someone’s daughter because she was not yet someone’s wife’.

MUNUS/f

duttarii̯ata/i- and Some Other Indo-European Maidens

285

those scholars who believe in the early split-off of Anatolian and instrumental for determining the internal etymology of this PIE word. 6

PIE Kinship Terms in *-ter- (Morphology)

There are many and various hypotheses about the morphological structure of the kinship terms in *-ter-. In my work, I follow the scenario originally suggested by Lohmann (1965: 217) and Szemerényi (1977: 10) that has recently been developed in detail by Pinault (2005, 2007, 2012). Pinault views the suffix *-ter- as an instance of delocatival derivation (see Nussbaum 1986: 235–247). Formally, this is the same derivational pattern as we see in the agentive nouns (Tichy 1995: 61), but semantically it is closer to the contrastive suffix, which is used to form the comparative grade of some adjectives (Ved. tavástara- ‘stronger’, NPers. behtar ‘better’, Gr. παυρότερος ‘smaller’) and adverbs from prepositions (*en → *en-ter → Lat. inter, OIr. eter, etc.). The element *-(e)h2- reconstructed for all kinship terms in *-ter- is analysed by Pinault as a collective or abstract suffix. The general meaning of the complex reconstruction *-(e)h2- + *-ter- should be ‘belonging by contrast to the group of…’ (Pinault 2007: 276–277; 2012: 2–3), or ‘someone within a group of similar people’, or simply ‘one of…’.16 Pinault (ibid.) puts forth his own etymological suggestions for each of the kinship terms involved. In this article I will only mention his etymologies for female kinship terms in *-ter- because the male ones—*ph2tér- ‘father’17 and *bhréh2ter- ‘brother’18—represent a separate and complex topic. 16

17 18

If Pinault’s analysis is correct, we could compare the semantics of ‘*(e)h2 + *ter’ with that of the Slavic possessive suffix -ov-, which is still actively used in some West and South Slavic languages. If, for example, we have to use Marina Zorman’s surname in an oblique case, e.g. in the dative (‘to go to Zorman’), we cannot say iti k Zormanu (because the ending -u is masculine, and it would mean that Zorman is a man), nor can we say *iti k Zormani (because grammatically Zorman is a masculine noun and the usage of the feminine ending would be incorrect). Slovenian has one way out of this dilemma by using the suffix *-ov-: iti k Zorman-ov-i, lit. ‘to go to a woman who belongs to the family Zorman’. In this case, Zorman plays the role of a collective noun, not grammatically but semantically, ­because a surname refers to a group of people, and the suffix -ov- has a contrastive / distributive role: ‘one of the Zormans’. E.g., Ved. pitár-, Av. pitar-, Arm. hayr, Gr. πατήρ, Lat. pater, OIr. athair, Goth. fadar, Toch. A pācar, Toch. B pācer (cf. IEW 829; EIEC 194–195; NIL 554–562). E.g., Ved. bhrā́tar-, Av., OPers. brātar-, Arm. ełbayr, Gr. φρᾱ́τηρ ‘member of a phratry’, Lat. frāter, Goth. broþar, OIr. bráth(a)ir, OCS brat(r)ŭ, OPrus. brāti, Toch. A pracar, Toch. B procer (cf. IEW 163–164; EIEC 84; NIL 38–41).

286

Milanova

For *méh2ter- ‘mother’,19 Pinault offers a simple and elegant solution. He derives it from the nursery term *ma- (mama), and *ma-h2-ter is, according to him, ‘one of the mothers’. *Hi̯énh2ter- ‘husband’s brother’s wife’20 (containing **h1i̯-én ‘acquisition, gift’ from the root *h1ai̯- ‘give, take (exchange)’ per LIV 2: 229) is a person ‘within the group of acquired/given women’ (who takes part in an alliance relationship between two families), hence ‘one of the brides’. Finally, *dhugh2tér- is ‘one of the (female) children’ (the core of the word **dhug‘(female) child’ will be touched upon in section 8 below). The question that arises at this point is: what is the contrast between a female child, a bride and a mother? Pinault (2012: 2) suggests that we should apply here a ‘privative opposition with marking of one term only by the contrastive suffix’, e.g. Greek δεξιóς vs. ἀριστερóς, σκαιóς vs. δεξιτερóς, ὑμóς vs. ἡμέτερος. Accordingly: *dhugh2tér- vs. *suH-nu-/*suH-i̯u-21/*putló-22 *Hi̯énh2ter- vs. *sṷésor-23 (a woman taken from another family vs. one’s own woman) *ph2tér- vs. *máh2ter- (exception: remade after *ph2tér-). However, I believe that it would be more natural to assume that all terms in *-ter- were built from a single system of their own like other words containing this suffix: δεξιτερóς vs. ἀριστερóς (‘right’ vs. ‘left’), ἡμέτερος vs. ὑμέτερος (‘ours’ vs. ‘yours (pl.)’), Latin superus vs. inferus (‘upper’ vs. ‘lower’), etc. This would mean that *dhugh2tér- is to be contrasted with *Hi̯énh2ter- and with *méh2ter-. The most obvious criteria here would be, of course, age and/or marital status (cf. also Kullanda 2001).

19 20 21 22 23

E.g., Ved. mātár-, Av. mātar-, Arm. mayr, Phrygian matar, Gr. (Attic) μήτηρ, Lat. māter, ON móðer, OIr. máthair, OLith. mótė, (gen.sg.) móters, ORus. mati, (gen.sg.) matere, Toch. A mācar, Toch. B mācer, Alb. motër ‘sister’ (!) (IEW 700–701; EIEC 385–386; NIL 457–461). E.g., Ved. yā́tar-, Pashto yōr, NPers. (Isfahan) yād, Arm. nēr, (gen.sg.) niri, Gr. (Homeric, pl.) εἰνάτερες, Lat. (pl.) ianitrīcēs, OLith. jéntė, (gen.sg.) jenter͂s, Latv. (Courlandish) jentere, OCS jętry, (gen.sg.) jętrŭve (cf. IEW 505–506; EIEC 522; NIL 204–207). With the suffix *-nu- (e.g., Ved. sūnús, Av. hunuš, Goth. sunus, OCS synŭ, OPrus. souns) or the suffix *-i̯u- (e.g., Gr. υἱός, Toch. A se, gen.sg. seyo, Toch. B soy) (NIL 686–690). E.g., Ved. putrá-, OPers. puça, Av. puθra-; Lat. (Plautus) putillus ‘young boy’; Osc. puklum (EIEC 533; Szemerényi 1977:18). E.g., Ved. svásar-, Av. xvaŋhar-, Arm. k‘oyr, Lat. soror, OIr. siur, Goth. swistar, OPrus. swestro, Toch. A ṣar, Toch. B ṣer (< acc.sg. *su̯ ésrm̥ per EIEC 521) (IEW 1051; EIEC 521; NIL 680– 683), Gr. ἔορες ‘female relatives, daughters, nieces’, and probably also ὄαρ ‘wife’ (Janda 1999).

MUNUS/f

7

duttarii̯ata/i- and Some Other Indo-European Maidens

287

PIE Kinship Terms in *-ter- as Stages of Life

The assumption that the terms in *-ter might have been improper kinship terms does not mean by itself that PIE lacked words for ‘father’, ‘mother’, ‘brother’, ‘daughter’, and ‘sister-in-law’ before the terms in *-ter- were reanalysed as kinship terms proper. It would rather mean that other words had been used instead: e.g., *atta- ‘father’ and *anna- ‘mother’, both preserved in Anatolian in exactly this meaning. *su̯ ésor- could originally have meant ‘(young) female relative’, i.e., it could also refer to daughters like the Greek ἔορ (Benveniste 1969: 214; Janda 1999: 320–324, esp. 322). Such words as *suH-nu-/suH-i̯u- and *putló- (because their basic meaning is ‘a born one’24 and ‘a little one’25 respectively) could originally have referred to both boys/sons and girls/daughters like Gr. παῖς (see also n. 10 above concerning Hitt. pulla-). Finally, *snusó-26 could have been applied as a generic term for all kinds of daughters- and sistersin-law, like Russian nevestka.27 This assumption thus does not contradict the general hypothesis of IE linguistics about PIE as a language that was once actively spoken. Any living language is dynamic, and its vocabulary is constantly undergoing semantic changes and innovations. A vivid typological example of similar semantic change in terms of social relations can be found in the Greek language. The Ancient Greeks viewed the active period of a woman’s life as divided into three stages: adolescence, family life (including motherhood), with a short transitional period between them (cf. Versnel 1994: 276–283; Clark 1998: 13–22, esp. 14; Cole 1998: 32–35).28 Girls between the ages of 7 and 15 (i.e., between the sacred posts ἀρρηφόρος and κανηφόρος) had the special designation παρθένος.29 A synonym of this 24 25 26 27

28

29

< *seu̯ H- ‘to give birth’ (LIV 2: 538). < *pau̯ - ‘small’ with the addition of the diminutive suffix *-tlo- (cf. EIEC 533). E.g., Ved. snuṣā́, Arm. now (gen.sg. nowoy), Lat. nurus, OHG snur, snora, ORus., Serbian CS snŭxa, Alb. núse ‘bride, young woman, younger daughter-in-law’ (IEW 978–979; EIEC 148; NIL 625–626) 1) ‘brother’s wife’; 2) ‘son’s wife in relation to the mother (more rarely: in respect to the father)’; 3) ‘husband’s brother’s wife’ (= yatrov’ (obs.) < *Hi̯énh2ter-). This word is an innovation (a derivative of nevesta ‘bride’) and is a rival of the inherited PIE word snoxa (< *snusó-), which is now viewed as specifically ‘son’s wife in relation to the father’ (Fedosyuk 1998: 24). Modern speakers regularly confuse these two words or even replace them with descriptive terms. Of course, this set of life stages is not unique and is characteristic for most traditional societies. I use the Greek material as an illustration because the long-term literary tradition of this culture permits one to trace how the meanings have evolved over the millennia. < *pr̥-steno- ‘having protruding breasts’, the meaning ‘virgin’ being attested only in the youngest texts (Klingenschmitt 1974: 273–278); alternatively: *pr̥-dhh1-nó- ‘ausgesetzter

288

Milanova

word, which is more important for the present work, was κόρη.30 The primary meaning of the latter in Ancient Greek was ‘young unmarried woman’. Secondarily, this word could be used as a synonym of θυγάτηρ, but only in combination with the genitive or a possessive pronoun (‘my girl’ = ‘my daughter’). Compare: μὰ τὴν Καλυψὼ τὰς τε Νηρέως (gen.) κόρας “by Calypso and Nereus’ daughters” euripides, Cyclops 264;

ὁ δ᾿ οὐ θέλων τε καὶ θέλων οἴκτῳ κόρης τέμνει σιδήρῳ πνεύματος διαρροάς· “And he, for pity of the girl both willing and reluctant, cut the breath’s passageways with his sword”. euripides, Hecuba 566–567

In Modern Greek this word still has both meanings, but it has almost replaced θυγατέρα and has become the only stylistically unmarked term for ‘daughter’ (Gates 1971: 39).31 At the age of 15 or even earlier, a Greek girl became engaged and subsequently married. A new transitional phase in her life began, and she bacame a νύμφη ‘young woman’, ‘bride/young wife’,32 ‘woman who is married but has not yet had her first child’ (Clark 1998: 14). Later, in Hellenistic times, this word replaced νυός (< *snusó-) and attained a kinship meaning ‘daughter-in-law’ (Gates 1971: 39): ἦλθον γὰρ διχάσαι ἄνθρωπον κατὰ τοῦ πατρὸς αὐτοῦ καὶ θυγατέρα κατὰ τῆς μητρὸς αὐτῆς καὶ νύμφην κατὰ τῆς πενθερᾶς αὐτῆς “for I came to set a man at variance against his father, and a daughter against her mother, and a daughter-in-law against her mother-in-law”. matthew 10:35

30 31 32

Siegerpreis’ (Janda 2014: 545–554). The etymology of this word is unfortunately rather vague. Beekes (2010: 752–753, 764) suggests two variants: κόρη as a derivative of the root *k̑erh3- ‘grow’ or somehow associated with κοῦρος ‘loppings, twigs lopped from a tree’. Georgios Kostopulos, a doctoral candidate at the University of Vienna and native speaker of Modern Greek, informs me that today only elderly people and nationalists use the word θυγατέρα instead of κόρη in a stylistically unmarked context. daughter/sister-in-law’ (parallel to ‘bridegroom > son/brother-in-law’) is not sporadic either, but reoccurs in other languages, e.g. NPers. Ɂarus ‘bride, daughter-in-law’ and dāmād ‘bridegroom, son-in-law’ (Tawakkulī 2003: 85, 425), Alb. núse ‘bride’, ‘daughter-in-law’, ‘young woman’ and dhëndërr ‘bridegroom, son-in-law’ (Orel 1998:82, 302–303). The same is true for Rus. nevesta ‘fiancée, bride’ > nevestka ‘daughter/sister-in-law’ (see above). Finally, after the first child was born, a woman gained the status of γυνή and μήτηρ. The word *gu̯ en(h2)- (> γυνή) was apparently used in both kinship and non-kinship meanings already in PIE. I suggest that a similar set of life stages might have existed in PIE and might have undergone a similar semantic change. PIE

Ancient Greek

*mé/áh2ter*Hi̯énh2ter*dhugh2tér-

~ γυνή / μήτηρ ~ νύμφη ~ κόρη

8 *dhugh2ter-: Etymology The internal etymology of *dhugh2ter- is still a major challenge and cannot be determined with complete certainty. If my hypothesis about stages of life is correct and the primary meaning of this word in PIE was ‘young unmarried woman, girl’, the core semantics must imply either her young age (as Gr. κανηφόρος is referred to παῖς καλὴ ‘beautiful child’) or her maturation (­etymology for παρθένος according to Klingenschmitt 1974: 273–278). The first alternative would mean that the first stage of a PIE girl’s life began ­immediately after her birth and lasted until her engagement (or wedding).33 The second alternative would only include the period after the onset of puberty before her transition into her husband’s family.34 A similar problem emerges when one tries to determine the etymology of Proto-Slav. *děva (< *doi-wā < *dhoh1i-u- per Snoj 1997: 88). It has the same

33 34

In social anthropology this type of age stratification is referred to as the New Guinea type, i.e. the first stage of life begins after the biological birth (Jeffreys 1950: 159). This is the Hamitic type of age stratification, in which the first stage of life begins with the onset of puberty/initiatory rite, etc. (social birth; Jeffreys 1950: 159).

290

Milanova

meaning as the meaning of *dhugh2ter- postulated above, ‘young unmarried woman’, and is most probably derived from √*dheh1(i)- ‘suck, suckle’ (LIV 2: 138). However, *děva is usually viewed as an adolescent girl. Therefore, she is neither the one who sucks nor the one who suckles. It is debated what kind of shift took place here. Berneker (apud IEW 242) prefers the shift ‘breastfeeding mother’ > ‘adolescent girl’ (‘potential breastfeeding mother’). Nevertheless, a shift ‘suckling (little girl)’ > ‘adolescent girl’ is also quite plausible. Pinault (2007: 276) derives **dhug-, the core of *dhugh2ter-, from the same verbal root √*dheh1(i)- ‘suck, suckle’ and analyses it as *dhh1-u-g-, a u-stem with a suffix/root extension *-g-.35 In spite of morphological difficulties, this suggestion looks attractive because it would connect this protoform with Slavic *děva and with the Latin terms fīlia ‘daughter’ and fēmina ‘woman’. However, the question of the etymology and semantic shift of *děva would still remain open. 9 Conclusion The word MUNUSduttarii̯ata/i- attested in Hittite ritual texts is obviously a Luwian borrowing. It is either a personal name or, more likely, a sacred title of wellborn (adolescent) girls who performed various ritual duties. Similar posts for girls existed, e.g., in Ancient Greece and had an educational and socialising function. DUMU.MUNUS (šuppeššara) might have been the corresponding logogram for this word, although this cannot be proven until/unless a text is discovered in which they both occur. The reconstruction of the semantic development of MUNUSduttarii̯ata/i- is especially complex. In the given context, the word cannot be translated as specifically ‘daughter’, but should be interpreted more broadly as ‘girl/maiden’. This latter meaning may simply be an occasional usage. However, it is also possible that in Anatolian this word could have a hybrid meaning ‘girl/daughter’, which is a characteristic feature of many languages of the Near East. It might show the semantic extension ‘daughter > girl’ as in NPers. doxtar. However, the opposite shift, which is typologically more frequent and natural, would be more fascinating because it could offer a hint as to the original meaning of *dhugh2tér- in PIE.

35

In his formulation, *-g- is ‘an expressive suffix referring to females’ as in Hitt. ne-g-a‘sister’, Lith. mer-g-à ‘girl, maiden’ (Pinault 2007: 276 n. 17; 2012: 2). It may probably be viewed as a sort of combinatory athematic variant of the individualising suffix *-(i)k- or *-(a)k- (Lat. sene-k-s, cf. Oettinger 2004: 169–170, Jamison 2009: 312–329).

MUNUS/f



duttarii̯ata/i- and Some Other Indo-European Maidens

291

Acknowledgments

The work on this paper has been possible thanks to a DOC-Stipendium from the Austrian Academy of Sciences. I would also like to express my special gratitude to Marinka Zorman and Heiner Eichner for sharing their knowledge of Hittite, Ilya Yakubovich for valuable remarks about morphological structure and possible semantics of the word MUNUS/fduttariia̯ ta/i-, Billie Jean Collins for consultation about the character DUMU.MUNUS šuppeššara-, Gernot Windfuhr for his valuable advice about semantic change in New Persian and my thesis supervisor Melanie Malzahn for supervising and directing me in the right way. The usual disclaimers apply. References Beekes, R.S.P. (2010) Etymological Dictionary of Greek [LIEEDS 10], Leiden: Brill. Benveniste, É. (1969) Le vocabulaire des institutions indo-européennes, I: Économie, parenté, société. Paris: Les Éditions de Minuit. Burkert, W. (1977) Griechische Religion der archaischen und klassischen Epoche, Stuttgart, Berlin, Köln, Mainz: Kohlhammer. Cameron, D.A. (1979) An Arabic-English Dictionary, Beirut: Librairie du Liban. Christiansen, B. (2006). Die Ritualtradition der Ambazzi. Eine philologische Bearbeitung und entstehungsgeschichtliche Analyse der Ritualtexte CTH 391, CTH 429 und CTH 463 [StBoT 48], Wiesbaden: Harrassowitz. Clark, I. (1998) The gamos of Hera, in: S. Blandell and M. Williamson (eds.), The Sacred and the Feminine in Ancient Greece, London, New York: Routledge, 13–26. Cole, S.G. (1998) Artemis at the borders of the polis, in: S. Blundell and M. Williamson (eds.), The Sacred and the Feminine in Ancient Greece, London, New York: Routledge, 27–43. Collins, B.J. (forthcoming) Paskuwatti’s ritual “When a man is not a man with respect to a woman” (CTH 406). Anniwiyani’s rituals “When I perform the ritual of the tutelary deity lulimi” and “When they invoke the protective deity kurša” (CTH 393), in: Hittite Rituals from Arzawa and the Lower Land, Atlanta: Scholars Press. Dahl, Ö. and Koptjevskaja-Tamm, M. (2001) Kinship in grammar, in: I. Baroh, M. Helsund and F. Sørensen (eds.), Dimensions of Possession, Amsterdam, Philadelphia: John Benjamins, 1–16. Dowden, K. (1989) Death and the Maiden: Girls’ Initiation Rites in Greek Mythology, London, New York: Routledge. Fedosyuk, Yu. (1998) Čto neponjatno u klassikov ili ėnciklopedija russkogo byta xix veka, Moskva: Flinta (2nd ed.).

292

Milanova

Gates, H.Ph. (1971) The Kinship Terminology of Homeric Greek [Indiana University Publications in Anthropology and Linguistics 27], Baltimore: Waverly Press. van Gennep, A. (1909) Les rites de passage, Paris: Nourry. Haas, V. (1994) Geschichte der hethitischen Religion, Leiden: Brill. Hawkins, J.D. (1978) The Luwian word for “daughter”, ZVS 92, 112–160. Hawkins, J.D. (2003) Scripts and texts, in: H.C. Melchert (ed.), 128–169. Hentschel, E. (2012) All men become brothers. The use of kinship terms for non-related persons as a sign of respect or disrespect, Linguistics Online 51/1, 29–42. Hoffman, J.M. (2010) And God Said: How Translations Conceal the Bible’s Meaning, New York: St. Martin’s Press. Hoffmann, I. (1992) Das hethitische Wort für “Sohn”, in: H. Otten et al. (eds.), Hittite and Other Anatolian and Near Eastern Studies in Honour of Sedat Alp, Ankara: Türk Tarih Kurumu Basımevi, 289–293. Hoffner, H. (1993) Hittite iu̯ ar and related modes of expressing comparison, Istanbuler Mitteilungen 43, 39–51. van den Hout, Theo (1995) Der Ulmitešub-Vertrag, Wiesbaden: Harrassowitz. Hutter, M. (2003) Aspects of Luwian religion, in: H.C. Melchert (ed.), 211–280. Jamison, S.W. (2009) Sociolinguistic remarks on the Indo-Iranian *-ka-suffix: a marker of colloquial register, Indo-Iranian Journal 52, 311–329. Janda, M. (1999) Zur Herkunft von homerisch ὄαρ, in: S. Deger-Jalkotzy, et al. (eds.). Floreant Studia Mycenaea: Akten des X. International Mykenologischen Colloquiums in Salzburg vom 1.–5. Mai 1995, Vol. 2, Vienna: Verlag der Österreichischen Akademie der Wissenschaften, 315–324. Janda, M. (2014) Purpurnes Meer: Sprache und Kultur der homerischen Welt [IBK 7], Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Jeffreys, M.D.W. (1950) Age-groups among the Ika and kindred people, African Studies 9/4, 157–166. Junker, H.F.J. and Alavi, B. (1967) Persisch-deutsches Wörterbuch, Leipzig: Verlag Enzyklopädie. Klingenschmitt, G. (1974) Griechisch παρθένος, in: M. Mayrhofer and W. Meid (eds.), Antiquitates Indogermanicae: Studien zur indogermanischen Altertumskunde und zur Sprach- und Kulturgeschichte der indogermanischen Völker. Gedenkschrift für Hermann Güntert zur 25. Wiederkehr seines Todestages [IBS 12], Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck, 273–278. Kloekhorst, A. (2008) Etymological Dictionary of the Hittite Inherited Lexicon [LIEEDS 5], Leiden: Brill. Kullanda, S. (2001) Indo-European “kinship terms” revisited, Current Anthropology 43/1, 89–111. Lohmann, J. (1965) Philosophie und Sprachwissenschaft, Berlin: Duncker & Humblot.

MUNUS/f

duttarii̯ata/i- and Some Other Indo-European Maidens

293

Lynn, B. (1978) Sitting in a cave: An analysis of ritual seclusion at menarche among Brahmans and Chhetris in Nepal, Contribution to Nepalese Studies 6/1, 31–45. MacKenzie, D.N. (1986 [1971]). A Concise Pahlavi Dictionary, London: Oxford University Press. Melchert, H.C. (1980) The Hittite word for “son”, IF 85, 90–95. Melchert, H.C. (1993) Cuneiform Luvian Lexicon, Chapel Hill, NC [self-published]. Melchert, H.C. (2003a) Prehistory, in: H.C. Melchert (ed.), 8–26. Melchert, H.C. (ed.) (2003b) The Luwians [HdO I/68], Leiden, Boston: Brill. Nussbaum, A. (1986) Head and Horn in Indo-European [UISK 2], Berlin, New York: de Gruyter. Oettinger, N. (2004) Griechisch γυνή „Frau“ und der Ursprung des „i-Kollektivums“, in: T. Kirsch and T. Lindner and U. Müller (eds.), Analecta homini universali dicata (Vol. 1), Stuttgart: Heinz. Orel, V. (1998) Albanian Etymological Dictionary, Leiden, Boston, Köln: Brill. Pinault, G.-J. (2005) Analyse étymologique d’un nom de parenté indo-européen, in: G. Schweiger (ed.), Indogermanica: Festschrift Gert Klingenschmitt. Indische, iranische und indogermanische Studien, Taimering: Schweiger, 465–485. Pinault, G.-J. (2007) A star is born: a “new” PIE *-ter-suffix, in: Alan J. Nussbaum (ed.), Verba Docenti: Studies in Historical and Indo-European Linguistics Presented to Jay H. Jasanoff, Ann Arbor, New York: Beech Stave, 271–279. Pinault, G.-J. (2012) Distance, proximity, exchange and kinship (semantics and etymology of Proto-Indo-European terms) (handout read at Lund University, 2012). Plöchl, R. (2003) Einführung ins Hieroglyphen-Luwische [DBH 8], Dresden: Verlag der TU Dresden. Poulkou, M. (2006) Arkteia. Überlegungen zu den nackten ‘Bärinnen’ in Brauron, in: G. Koiner et al. (eds.), Akten des 10. Österreichischen Archäologentages in Graz, 7–9 N ­ ovember 2003, Wien: Phoibos. Rahmati, N. (1999) Aserbaidschanisch-deutsches Wörterbuch unter Berücksichtigung der Besonderheiten des Nord- und Südaserbaidschanischen, Engelschoff: auf dem Ruffel. Rastorguyeva, G.S. (1966) Sredne-persidskij jazyk, Moscow: Nauka. Rieken, E. (2015) A Luwian crown-prince (handout from the 34th East Coast IndoEuropean Conference, Universität Wien, 04-07.06.2015). Snoj, M. (1997) Slovenski etimološki slovar, Ljubljana: Mladinska knjiga. Sommerstein, A.H. (ed., tr.) (1990) Lysistrata. The Comedies of Aristophanes, Warminster: Aris & Phillips Ltd. Sourvinou-Inwood, Ch. (1988) Studies in Girls’ Transition: Aspects of the Arkteia and Age Representation in Attic Iconography, Athens: Kardamitsa. Starke, F. (1980) Das luwische Wort für „Frau“, ZVS 94, 74–86.

294

Milanova

Starke, F. (1987) Die Vertretung von uridg. *dhugh2tér- „Tochter“ in den luwischen Sprachen und ihre Stammbildung, ZVS 100, 243–269. Szemerényi, O. (1977) Studies in the Kinship Terminology of the Indo-European Languages [Acta Iranica 17], Leiden: Brill; Teheran, Liège: Bibliothèque Pahlavi. Taggar-Cohen A. (2006) Hittite Priesthood [THeth 26], Heidelberg: Winter. Tawakkulī, Ḥ. (2003) Farhang-i ālmānī-fārsī, fārsī-ālmānī, Aachen: Tavakkoly; Tihrān: Intišārāt-i Tarma. Tichy, E. (1995) Die Nomina agentis auf -tar- im Vedischen, Heidelberg: Winter. de Vaan, M. (2008) Etymological Dictionary of Latin and the Other Italic Languages [LIEEDS 7], Leiden: Brill. Versnel, H.S. (1994) Inconsistencies in Greek and Roman Religion, 2: Transition and Reversal in Myth and Ritual, Leiden, New York: Brill. Vietze, H.-P. (1981) Wörterbuch Deutsch-Mongolisch, Leipzig: Verlag Enzyklopädie. Weeden, M. (2011) Hittite Logograms and Hittite Scholarship, Wiesbaden: Harrassowitz. Yakubovich, I.S. (2010) Sociolinguistics of the Luvian Language, Leiden, Boston: Brill. Yakubovich, I.S. (2015) The Luwian Language, Oxford Handbooks Online, http://www .oxfordhandbooks.com/view/10.1093/oxfordhb/9780199935345.001.0001/ oxfordhb-9780199935345-e-18. Zalizniak, A.A. (2008) A catalogue of semantic shifts: towards a typology of semantic derivation, in: M. Vanhove (ed.), From Polysemy to Semantic Change: Towards a Typology of Lexical Semantic Associations [Studies in Language Companion Series 106], Amsterdam, Philadelphia: John Benjamins. Zehnder, T. (2010) Die hethitischen Frauennamen. Katalog und Interpretation, Wiesbaden: Harrassowitz.

Chapter 14

One Century of Heteroclitic Inflection Georges-Jean Pinault Abstract The so-called heteroclitic Indo-European nouns, such as ‘water’, played a significant role in the identification of Hittite as an Indo-European language, because this type of inflection appeared to be characteristic of this language family. Nouns featuring heteroclitic inflection are known mostly as relics in other Indo-European languages, including Tocharian. Several of them were recorded long before the discoveries of Hittite and Tocharian, but the decipherment of Hittite in 1915 has inspired an enduring interest in the origin of the heteroclitic stems, which would open windows into the remote past of Proto-Indo-European. Hittite has well preserved the complete inflection of ‘water’, in both singular and plural, and has moreover developed the heteroclitic inflectional type extensively, especially in complex suffixes, so that it appears in a new light. The best characterized type combines in its inflection an r-stem in the nominativeaccusative of the singular and of the collective with an n-stem in the weak stem allomorph of both the singular and the collective. The paper reviews the most significant solutions that have been proposed for the origin and the development of this association of two stems in a single paradigm: 1) the phonological solution; 2) the scenario of mutual substitution of suffixes, by projecting the ‘heteroclitic’ system back to derivation, at a pre-inflectional stage; 3) the scenario of hypostasis, which accounts for the alternating inflectional stems through extensions from case forms marked by alternative endings. The preliminary question of the formal extension of the heteroclitic inflection is also addressed. Are there other such alternations of stems, for instance with i-stem or l-stem or even stem with zero suffix instead of r-stem? Which subtypes of complex suffixes of the type *-Cer/n-, which have become productive in Anatolian, can be safely attributed to Proto-IndoEuropean? As for the origin of the heteroclitic inflection, the paper maps out a purely morphological solution which would be connected with the process of internal derivation, and especially with the formation of collectives. It is argued that the heteroclitic inflection began as the reshaping of a previous inflectional type of plain r-stems.

© koninklijke brill nv, leiden, ��20 | doi:10.1163/9789004413122_016

296

Pinault

Keywords Proto-Indo-European – nominal inflection – nominal derivation – heteroclitic stems – allomorphy – collectives

1 The heteroclitic neuter stems were bound to the decipherment of Hittite by Bedřich Hrozný, which succeeded in the summer of 1914, before his military enrollment at the beginning of World War i. In the first account of his discovery, he made the following point: Sehr wichtig war die Feststellung einer Deklination, die gerade für die indogermanischen Sprachen und nur für sie besonders charakteristisch ist. Dem Verfasser / gelang es zunächst in dem Worte wa-a-tar mit ziemlicher Wahrscheinlichkeit das hethitische Wort für “Wasser” festzustellen, das natürlich mit altsächs. watar, ahd. waʓʓar, gr. ὕδωρ usw. “Wasser” identisch ist. Es gelang aber weiter festzustellen, daß von diesem Worte der Gen. sg. nicht etwas *wa-a-tar-aš o. ä., sondern überraschenderweise wohl ú-e-te-na-áš, der Abl./Instr. ú-e-te-ni-it/d usw. lautet; statt des -r des Nom. und Akk. (vgl. die Identität des Nom. und Akk. bei dem indogermanischen Neutrum!) bieten die übrigen Kasus des Sg. ein -n-. Denselben Wechsel zwischen -r und -n- weist aber das entsprechende indogermanische Wort auch z.B. im Griechischen auf, wo zu ὕδωρ der Gen. ὕδατος lautet, wobei das α dieser Form bekanntlich aus n̥ entstanden ist! Es ist die bekannte höchst eigenartige Deklination, die auch z.B. im lat. femur, Gen. feminis vorliegt. Einen stärkeren Beweis für den Indogermanismus des Hethitischen kann man sich wohl kaum wünschen.1 Other shibboleths for the identification of Hittite as an Indo-European (henceforth IE) language were the evidence for participles in -ant-, the major nominal endings, the personal pronouns, the interrogative and relative pronoun, the inflection of the present active, several adverbs, and some striking correspondences in the lexicon. 1 Hrozný 1915: 24–25. Later on (1915: 33–35), the author gives the philological analysis of the Hittite sentences which contain the words in question. Additional abbreviations used in the text: Att. = Attic; AV = Atharva Veda; CToch. = Common Tocharian; Dor. = Doric; Hom. = Homeric; Ion. = Ionic; Lesb. = Lesbian; RV = Rigveda; ŚB = Śatapatha Brāhmaṇa.

One Century of Heteroclitic Inflection

297

In his first description of the Hittite language, Hrozný devotes several pages to the heteroclitic stems.2 He discusses the ablaut of the nom.-acc. forms of ‘water’ in -ar and -ār. He traces back the type of Hitt. u̯ idār, plural, to IE *-ōr, as reflected also by Gr. ὕδωρ and Av. aiiārǝ ‘days’.3 He refers to the theory of J. Schmidt (1889: 191–218), according to which the stem allomorphs with final lengthened grade had collective meaning. In addition to ‘water’, Hrozný identifies further stems of the same type: uttar ‘word, thing’, pl. uddār (ud-da-a-ar); paḫḫur ‘fire’, dat. sg. paḫḫueni. For the latter, he gives4 the correct etymological connection (with Gr. πῦρ, etc.), and proposes to recognize in the weak stem allomorph a formant *-u̯ en-, *-un-, to be compared with that of Gr. πεῖραρ ‘limit’