<<

ROOTS OF EUROPE – LANGUAGE, CULTURE, AND MIGRATIONS UNIVERSITY OF COPENHAGEN

The sound of Indo-European Phonetics, phonemics, and morphophonemics 16–19 April 2009

Abstracts

photo: Mirjam Marti

ROOTS OF EUROPE – LANGUAGE, CULTURE, AND MIGRATIONS UNIVERSITY OF COPENHAGEN

The sound of Indo-European Phonetics, phonemics, and morphophonemics 16–19 April 2009

Abstracts

rootsofeurope.ku.dk

Contents

CONFERENCE PROGRAMME V Václav Blažek On Indo-European correspondences to Afro-Asiatic laryngeals 1 Lars Brink The etymology of Danishkone ‘wife etc.’ 6 ANDREW BYRD Predicting PIE syllabification through phonotactic analysis 7 Paul S. Cohen & Adam Hyllested A new sound law of PIE: initial **h3w > *h2w 9 Michael Frotscher On final‑r in Vedic 10 josé virgilio garcía trabazo Phonologische und morphologische Bemerkungen zu den hethitischen -i̯e/a-Verben 12 Piotr Gąsiorowski The Compass Conspiracy, a footnote to Verner’s Law and some further observations on the reflexes of PIE -* sr- 14 Aaron Griffith Raising before *µ in Old Irish: evidence and consequences 15 Irén Hegedűs Theruki -rule in Nuristani 17 Eugen Hill 19 Hidden sound laws in the inflectional morphology of Proto-Indo- European Adam Hyllested 21 The status and prevalence of Proto-Indo-European *a ANders Richardt Jørgensen The development of Proto‑Celtic *sk in Brythonic 23 II Contents

Joshua T. Katz Greek ὀπυίω and other words without a digamma 24 Götz Keydana Evidence for non-linear phonological structure in Indo-European: the case of clusters 25 The phonological interpretation of plene and non-plene spellede in Hittite 27 Agnes Korn Parthian ž 28 Guus Kroonen The rise of the Germanic iterative verbs: another case of ablaut being translated into gradation 29 Martin J. Kümmel Typology and reconstruction: the IE and 30 Rosemarie Lühr & Susanne Zeilfelder Optimale Onsets im Indogermanischen 31 Marek Majer The development of Proto-Slavic word-final-ī * , *-ě from PIE *-oi̯, ‘*-ai̯’ 32 Biliana Mihaylova Some cases of the converse of Sievers’ law in Greek 33 Paolo Milizia On the morphophonemics of Proto-Indo-European *-sḱe/o- suffixed presents 35 Kanehiro Nishimura Vowel reduction and deletion in Sabellic: history and chronology 37 Georges-Jean Pinault Remarks on the PIE amphikinetic nouns 39 Charles Prescott Germanic and the ruki dialects 41 Contents III

Tijmen Pronk The “Saussure Effect” in Indo-European languages, other than Greek 43 JENS ELMEGÅRD RASMUSSEN Rule ordering paradoxes in Indo-European: reintroducing phonetics 44 Giancarlo Schirru Laryngeal features in stops of Armenian dialects 45 Ilja Seržant Tocharisches Lautgesetz *-sl- > -ll- / -l- 47 Vitalij Shevoroshkin Indo-European laryngeals in Anatolian 48 Zsolt SImon PIE ‘me’ and a new Lydian sound law 50 Thomas Smitherman Some consequences of Kartvelian realisations of Indo-European borrowings for the reconstruction of the Proto-Indo-European phonological system 51 Dieter Steinbauer Proto-Indo-European *kakós ‘bad’ –another good example of unconditioned /a/ 54 David Stifter The sociophonemics of Late Gaulish 56 Lucien van Beek Greek evidence for the “Saussure Effect” reconsidered 57 Brent Vine PIE mobile accent in Italic: further evidence 58 Gordon Whittaker Aspects of Euphratic 60 Paul Widmer Types of ablaut class assignment in non-primary nominal derivation 61 IV Contents

ILYA YAKUBOVICH Phonetic interpretation of Hurrian sibilants in the light of Indo-European evidence 62 Nicholas Zair A new environment for laryngeal loss in Proto-Celtic 64 V

Programme

All lectures will take place in: Room 23.0.50 The Faculty of Humanities, Københavns Universitet Amager (KUA) Njalsgade 120 2300 Copenhagen S

Wednesday 15 April 2009

16:00–18:00 Registration and informal drinks Room 22.5.26 Department of Nordic Studies and Linguistics (KUA) 18:30 Informal dinner at participants’ own expense (details to be announced) Bryggens Spisehus Islands Brygge 18 2300 København S Phone: (+45) 32540082 www.bryggens.dk VI Thursday 16 April 2009

8:00–10:30 Registration (in front of room 23.0.50) 10:30–10:45 Welcome and practical information (room 23.0.50) 10:45–12:15 1st section. Chair: Paul Widmer 10:45–11:15 Rosemarie Lühr & Susanne Zeilfelder Optimale Onsets im Indogermanischen 11:15–11:45 Andrew Byrd Predicting PIE syllabification through phonotactic analysis 11:45–12:15 Götz Keydana Evidence for non-linear phonological structure in Indo-European: the case of fricative clusters 12:15–13:45 Lunch break 13:45–15:15 2nd section. Chair: Jens Elmegård Rasmussen 13:45–14:45 Eugen Hill Hidden sound laws in the inflectional morphology of Proto-Indo- European 14:45–15:15 Georges-Jean Pinault Remarks on the PIE amphikinetic nouns 15:15–15:30 Coffee break 15:30–17:00 3rd section. Chair: Jenny Helena Larsson 15:30–16:00 Jens Elmegård Rasmussen Rule ordering paradoxes in Indo-European morphophonemics: reintroducing phonetics 16:00–16:30 Joshua Katz Greek ὀπυίω and other words without a digamma 16:30–17:00 Biliana Mihaylova Some cases of the converse of Sievers’ law in Greek

17:00–17:15 Coffee break 17:15–18:15 4th section. Chair: Agnes Korn 17:15–17:45 Tijmen Pronk The “Saussure Effect” in Indo-European languages, other than Greek 17:45–18:15 Lucien van Beek Greek evidence for the “Saussure Effect” reconsidered Friday 17 April 2009 VII

8:00–9:30 Registration (in front of room 23.0.50) 9:30–10:30 5th section. Chair: Thomas Olander 9:30–10:00 Marek Majer The development of Proto-Slavic word-final *-ī, *-ĕ from PIE *‑oi, ‘*‑ai̯’ 10:00–10:30 Ina Vishogradska Vowel harmony in the Slavonic loanwords hosted in Hungarian (Phonol. characteristics of Slavonic loanwords in non-IE languages) 10:30–10:45 Coffee break 10:45–12.15 6th section. Chair: Rosemarie Lühr 10:45–11:15 Piotr Gąsiorowski The Compass Conspiracy, a footnote to Verner’s Law and some further observations on the reflexes of PIE *-sr- 11:15–11:45 Paolo Milizia On the morphophonemics of Proto-Indo-European *-sk̂e/o‑ suffixed presents 11:45–12:15 Ilja Seržant Ein tocharisches Lautgesetz *-sl- > -ll-/-l- 12.15–13:45 Lunch break 13:45–15:15 7th section. Chair: Birgit Anette Olsen 13:45–14:45 Martin J. Kümmel Typology and reconstruction: the IE consonants and 14:45–15:15 Paul Widmer Types of ablaut class assignment in non-primary nominal deriva‑ tion 15:15–15:30 Coffee break 15:30–16:30 8th section. Chair: Piotr Gąsiorowski 15:30–16:00 Adam Hyllested The status and prevalence of Proto-Indo-European *a 16:00–16:30 Dieter Steinbauer Proto-Indo-European *kakós ‘bad’ – another good example of unconditioned /a/ 17:30 Sightseeing: Copenhagen from the seaside (for registered participants) VIII Saturday 18 April 2009

8:00–9.00 Registration (in front of room 23.0.50) 9:00–10:30 9th section. Chair: Georges-Jean Pinault 9:00–9:30 Michael Frotscher On final -r in Vedic 9:30–10:00 Irén Hegedűs The ruki-rule in Nuristani 10:00–10:30 Agnes Korn Parthian ž 10:30–10:45 Coffee break 10:45–12:15 10th section. Chair: Irén Hegedűs 10:45–11:15 Václav Blažek On Indo-European correspondences to Afro-Asiatic laryngeals 11:15–11:45 Thomas Smitherman Some consequences of Kartvelian realisations of IE borrowings for the reconstruction of the PIE phonological system 11:45–12:15 Gordon Whittaker Aspects of Euphratic phonology 12:15–13:45 Lunch break 13:45–15:15 11th section. Chair: Benedicte Nielsen 13:45–14:45 Brent Vine PIE mobile accent in Italic: further evidence 14:45–15:15 Kanehiro Nishimura Vowel reduction and deletion in Sabellic: history and chronology 15:15–15:30 Coffee break 15:30–17:00 12th section. Chair: Joshua Katz 15:30–16:00 Paul S. Cohen & Adam Hyllested A new sound law of PIE: initial **h₃w > *h₂w 16:00–16:30 Vitalij Shevoroshkin Indo-European laryngeals in Anatolian 16:30–17:00 Giancarlo Schirru Laryngeal features in stops of Armenian dialects

19:00 Conference dinner (details to be announced) Madklubben Store Kongensgade 66, 1264 København K Phone: (+45) 33 32 32 34; www.madklubben.info Sunday 19 April 2009 IX

10:00–10:45 Registration (in front of room 23.0.50) 10:45–12:15 13th section. Chair: David Stifter 10:45–11:15 Aaron Griffith Raising before *µ in Old Irish: evidence and consequences 11:15–11:45 Anders Richardt Jørgensen The development of Proto-Celtic *sk in Brythonic 11:45–12:15 Nicholas Zair A new environment for laryngeal loss in Proto-Celtic 12.15–13:45 Lunch break 13:45–15:15 14th section. Chair: Hans Götzsche 13:45–14:15 David Stifter The sociophonemics of Late Gaulish 14:15–14:45 Charles Prescott Germanic and the ruki dialects 14:45–15:15 Guus Kroonen The rise of the Germanic iterative verbs: another case of ablaut being translated into consonant gradation 15:15–15:30 Coffee break 15:30–16:30 15th section. Chair: Susanne Zeilfelder 15:30–16:00 Lars Brink The etymology of Danish kone ‘wife etc.’ 16:00–16:30 Alwin Kloekhorst The phonological interpretation of plene and non-plene spelled e in Hittite 16:30–17:00 Zsolt Simon PIE ‘me’ and a new Lydian sound law 17:00–17:15 Coffee break 17:15–18:15 16th section. Chair: Gordon Whittaker 17:15–17:45 José Virgilio García Trabazo Phonologische und morphologische Bemerkungen zu den hethitischen -i̯e/a-Verben 17:45–18:15 Ilya Yakubovich Phonetic interpretation of Hurrian sibilants in the light of IE evidence 18:15 End of the conference

1

On Indo-European correspondences to Afro-Asiatic laryngeals

vÁCLAV bLAžеK Masaryk University, Brno

The present list of lexical comparisons between Afroasiatic and Indo-Europe- an was chosen according to two criteria: (i) The phonetic correspondences are in agreement with the phonetic rules established by V.M. Illič-Svityč and A. Dolgopolsky (with exception of laryngeals which represent the weakest point of the ‘Old’ Moscow Nostratic school); (ii) The cognates contain the larynge- als. Preliminary conclusions are based on these figures:

1 AA *’ ~ IE *H1 (9×). 2 AA *C ~ IE *H2 (12×), *H3 (4×), plus 3 cases of uncertain laryngeal in IE. 3 AA *ħ ~ IE *H2 (8×), *H3 (maybe 1×) plus 1–2 cases of uncertain laryngeal in IE. 4 AA *h ~ IE *H2 (3×), *H3 (2×), plus 1 case of uncertain laryngeal in IE.

It seems that AA *C, *ħ, *h merged in IE *H2 and *H3 is only a rarer variant of *H2.

AA *’

1 Semitic √’-b-y “to want, desire” ||| Egyptian 3bj “to wish, desire” ‡‡ IE *Hyebh-/*Heibh- to copulate”. 2 AA √’-m-: Egyptian (OK) 3mm “to catch by fist” ||| Berber: Shilha ämi “to contain ||| Chadic: (C) Tera oom “to catch, seize”; Musgu íma, ime id. || (E) Lele ōm “to seize, take”, Kabalai am “to catch” ‡‡ IE *H1em- “to take”. 3 AA √’-y-s: Semitic *’iš-: Old Akkadian inf. išû, in the suffix conjugation iš-āku “I have”, in the prefix conjugation tīšu “you have” < *ti-išu, lit. “tibi est”, īšu “he has” < *yi-išu, lit. “ei est” (→ “there is”); ||| Cushitic: (E) Afar ase, Saho as- “to spend the day” ||| Omotic: (N) Zayse yes- “esserci” ‡‡ IE *H1es- “to be”. 2 Václav Blažek

4 AA √’-t-y/w & √t-’-y/w: Semitic √t-’-w „to eat“ ||| ?Egyptian t „bread“ ||| Cushitic: (N) Beja tiyu „food“ || (E) Bayso eede “to eat”; Oromo ito “food”; Sidamo, Burji it- “to eat” ||| Chadic: (W) Fyer et; Tangale edi; Mburku ti || (C) Zime Batna tí || (E) Dangla tèè, Migama tíyáw; Mubi túwà/tìyá „to eat“ ||| Berber √w-t-t: habitative *tattaH id. ‡‡ IE *H1ed- “to eat, bite”. 5 Semitic *bi’r- ~ *bu’r- “well, spring” ||| Egyptian b3y(t) “Wasserloch” ‡‡ IE *bhréH1wr, gen. *bhruH1nós “well, spring”. 6 Semitic √r-’-y “to see, look at, understand, observe” ‡‡ IE *reH1- “berech- nen, meinen”. 7 Semitic √r-d-’ “to give a support to; dedicate, offer, protect “‡‡ IE *reH1dh- “beachten, sich kümmern”. 8 Semitic √w-d-’ “to achieve; complete, bring to a conclusion, make (an end)” ||| Egyptian wdj “setzen, stellen, legen; geben, darbringen; einsetzen, ernen- nen” ‡‡ IE *dheH1- “stellen, legen, setzen; machen”. 9 Semitic √w-r-’ “to be afraid” ‡‡ IE *wer(H1)- “to warn” (according to LIV 685 essive *wr-H1yé-, but an alternative morphological segmentation *wrH1-yé- is possible).

AA *C

10 Semitic √C-d-w “to cross, pass over, move, march” ‡‡ IE *weH2dh- “durch- schreiten”. 11 Semitic √C-k-w “to rise, be big and fat” ‡‡ IE *H2eug- “to be strong; grow”. 12 Semitic √C-l-y “to mound up, ascend” ||| Egyptian Cr ~ jCr “to mount up, ascend” ||| Berber √H-l-y: “to mount, rise” ||| Chadic: (W) Angas yaal “to get up, rise”, Tangale ile “to stand up, rise, start”, Geji hilya “to stand up” | (E) Kwang aalé-, Ngam alé; Sumray ’àyl- “to climb” ||| Cushitic: (E) *Cal- “mountain” ‡‡ IE *H2el- “to grow, nourish”; *H2el-to- “high, old”. 13 Semitic: Ugaritic Cl “next” or “second” ‡‡ IE **H2el- in the derivatives *alyo-, *alno-, *al-tero- “other, second”. 14 Semitic *Ca/ing-(at-) > „neck“ and / or ħanğarat & ħunğūr “larynx” ||| Egyptian (18Dyn) ħngg “Schlund” ‡‡ IE **H2enĝhu- “neck”. 15 Semitic: Ugaritic Caq “eyeball”; Hebrew Cāqā id. ‡‡ IE *okʷ- (**H3ekʷ-) “e y e”. 16 Semitic √C-q-q “to make the cloud to rain”, √C-q-y “to give to drink” ||| Cushitic (C) *’aqʷ “water” ||| Omotic: (N) Yemsa akà id. ‡‡ IE *H2ekʷ- “water”. Václav Blažek 3

17 Egyptian (OK) Cr “Binse”, C3C “Schilfrohr” ‡‡ IE *H2er-/*H2or- > Greek áron “Art Schilfrohr, Natterwurz”; harundō “Rohr”; ?Hittite arisanda- “Art Rohr”. 18 Semitic √C-r-k “to prepare; lay out, set in rows” ‡‡ IE *H3reĝ- “gerade rich- ten, ausstrecken”. 19 Egyptian Cw.t “Kleinvieh” ||| Cushitic: (N) Beja eewu “capricorn” ‡‡ IE *H3ewi- „sheep“. 20 Semitic *Cayan- “eye” ||| Egyptian Cn *”eye” (reconstructed on the basis of the hieroglyph depicting “eye”) ||| Cushitic: Dahalo Ceen-aad- “to see from afar” ||| Berber √n-H-y “to see” ||| Chadic: (W) Bole ìnnáa- “to see, think”; Daffo-Buturayen “to see”; Geji yenî, yèni “to see” || (E) Jegu ’inn- “to know, can” ‡‡ IE *neiH- > Old Indic nayana-/ā- “eye”, Latin re-nīdeō “erglänzen”, niteō, -ēre “glänze, strahle”; Middle Irish níam “Glanz, Schönheit”. 21 Semitic *√d-r-C “to spread out, stretch arms” ‡‡ IE *sterH3- “ausbreiten”. 22 Semitic *naCar- “young man; lad, fellow” ‡‡ IE *H2ner- “m a n”. 23 Egyptian pCw “fire” ||| Berber (S) *ē-fiHiw, pl. *ī-fiHiw-ān “fire” ||| Chadic: (C) Gidar afá; Logone fo; Musgu afu “fire” ‡‡ IE *peH2w-(r/n-) “fire”. 24 Semitic √š-m-C “to hear, listen, understand” ‡‡ IE *sH2em- “song” . 25 Semitic √w-C-k: Arabic waCaka “être trés chaud (jour ďété) ‡‡ IE *H2eug- “Glanz, Strahl, Tageslicht”. 26 Semitic: Syriac yaCCā „avis quadem, pterocles al. coturnix“; Tigre wiCe „sor- te de passereau“ || Egyptian CwC “ein Vogel”, Cjw “Graukranich-Jungvogel” ‡‡ IE *H2woi-s nom. : *H2wei-s gen. “bird” (Schindler). 27 Semitic √w/y-d-C “to know” ||| Egyptian ydC “klug” ||| Cushitic: (E) Saho diC- & deC- “to know, understand, can”; Somali daC- “may, understand” ‡‡ IE **dheiH-/*dhyeH- “to see, know, think”. 28 Semitic √w-l-C “to grow frivolous, greedy for, desire” ||| Chadic (W) Sura wal, Mupun wāl “to love” ‡‡ IE *welH- “to want, wish”.

AA *h

29 Semitic: Biblical Aramaic hāk in yəhāk “he goes”; Arabic dial. (Maghreb) hāk “être bon marcheur, rapide (cheval)” ‡‡ IE **Haeĝ- > *aĝ- “to drive, lead, guide, travel”. 30 Semitic *harar- “mountain” ||| Egyptian hr “Waldgebirge” ‡‡ IE *H3er- /*H3?or-: Hittite aru- “high” : arai- “to rise, lift, raise”; Greek óros “Berg, Anhöhe”, orésteros “auf den Bergen lebend”; Middle Irish or, acc. pl. uru m. 4 Václav Blažek

“coast, bank”, airer “coast, haven” < air- + or, Welsh ôr “limit, edge, brim, margin”, Old Breton or “edge” 31 Egyptian (OK) hrw “day” ||| Berber: Ahaggar tarut, pl. tirutîn, Taneslemt tărăhut “heures du milieu du jour” < *tā-ruHūH-t & *tā-raHūH-t respec- tively ||| Chadic: (C) Buduma yīrow, yēráu, īrau “day” || (E) Dangla ’eri‑ yo “noon”; cf. Semitic √w-h-r & √n-h-r “to shine”: Mandaic ywr’ “light, brightness, brilliance”, Hebrew nəhārā “light of day”; Arabic wāhir “shining, white” & nahār “clear day, morning” ‡‡ IE *H2reu-: Hittite harwanai- “to be clear; dawn”, Old Indic ravi- “ sun”; Armenian arew id. 32 Semitic √b-h-w/y “to be beautiful, shine with beauty” ‡‡ IE *bheH2- “to shine”. 33 Semitic √k-h-n “to foretell the future” ||| ?Cushitic: (N) Beja kehun “to like, have affection for” ‡‡ IE *ĝneH3- “to know”. 34 Semitic *nahar- “river” ||| Egyptian nhr “laufen, rennen” (*hinwegströ- men)” ‡‡ IE *nerH- “untertauchen”; cf. Old Indic pl. nārās “waters”; Lithu- anian narà “rivulet, brook”.

AA *ħ

35 Egyptian ħCpj “Nil; Überschwemmung”; maybe is related Arabic ħaffat “edge, border, rim, side” (cf. Latin rīpa “Ufer” : Spanish ribera “Ufer, Bach”, French rivière “Au”) ‡‡ IE *H2e(H)p- “water, river”, *H2ep-H3on- > *abon- “having the flowing water”. 36 Semitic √ħ-m-r- “to be red” ‡‡ IE *H2mr-u- /*H2mrwo-: Luwian marway(a)- “rot”, cf. the hydronym Marassantiya-, written also ÍD.SA5 “Red River”, cor- responding to the river Kızıl Irmak, in Turkish “Red River”; Greek amaurós “dunkel, trübe, schwach”. 37 Semitic √ħ-ś-w “to be dry (foliage)”, √ħ-ś/š-ś/š “to mow”, Hebrew ħăšaš “dried grass or foliage” ‡‡ IE *H2seus- < *H2s-(H)eus- “to become dry”. 38 Semitic √’-n-h/ħ/x “to breath” ||| Egyptian Cnx “to live” ‡‡ IE *H2enH1- “to breath”. Note: The etymons are compatible in the case of of the larynge- als in one of the families. 39 Semitic: Akkadian puxālu “male animal of ram, bull, stallion”; Ugaritic pħl “stallion, ass”, Arabic faħl “stallion” ‡‡ IE *poHl-/*peH3l- “foal, filly, mare”. 40 Semitic √g-ħ-f “to gather, take out, carry off” ‡‡ IE *(ĝ)heH2b- “to take, catch”. Václav Blažek 5

41 Semitic √n-ħ-y/w “to lead; move in a certain direction; across” ‡‡ IE *neiH- “to guide, lead”. 42 Semitic √l-p-ħ “to burn, scorch” ‡‡ IE *leH2p- “to burn, shine”. 43 Semitic √p-t-ħ “to open” ‡‡ IE *petH2- “to open”. 44 Semitic √r-w-ħ “to extend, spread oneself out, be wide” ||| ?Egyptian (MK) w3ħ “dauern” ‡‡ IE *reuH- “to open a space”.

AA *x

45 Semitic: Arabic naxšūš “nostril”; Mehri naxśîś “nose” ‡‡ IE *neHs-/*Hnos-? “nose”.

Abbreviations: AA Afroasiatic, C Central, IE Indo-European, N North, S South, W West.

Basic Sources

Brunner, Linus. 1969. Die gemeinsamen Wurzeln des semitischen und indogermanischen Wort- schatzes. Versuch eine Etymologie. Bern-München: Francke. Dictionnaire des racines sémitiques by David Cohen et al. Paris-La Haye: Mouton 1970f. Illič-Svityč, Vladislav M. 1967. Materialy k sravniteľnomu slovarju nostratičeskix jazykov. Ėtimologija 1965: 321–373. Illič-Svityč, Vladislav M. 1968. Sootvetstvija smyčnyx v nostratičeskix jazykax. Ėtimologija 1966: 304–355. Illič-Svityč, Vladislav M. 1971/76/84. Opyt sravnenija nostratičeskix jazykov, I-III. Moskva: Nau- ka. LIV Lexikon der indogermanischen Verben. Die Wurzeln und ihre Primärstammbildungen, by Helmut Rix, Martin Kümmel, Thomas Zehnder, Reiner Lipp and Brigitte Schirmer. Wies- baden: Reichert. Møller, Hermann. 1909. Indoeuropæisk-semitisk sammenlignende glossarium. Kjøbenhavn: Schultz. Møller, Hermann. 1917. Die semitisch-vorindogermanischen laryngalen Konsonanten. Køben- havn: Høst & søn. Pokorny, Julius. 1959. Indogermanisches etymologisches Wörterbuch. Bern-München: Francke. 6

The etymology of Danish kone ‘wife etc.’

Lars Brink Copenhagen

Our dictionaries tell us that behind Goth. qino, Eng. queen, Nord. kuna etc. are two ablaut Proto-Germanic forms: *kwenōn and *kunōn. The former leading to East and West Germanic, the latter and the former leading to Nord. kuna– kunu–kunu(r)–kunum–kwenna/kwinna. I show that the very double forms of the Nord. gen.pl. are lethal to *kunōn. And I support this by some Viking forms from the British Isles showing kw- outside the gen.pl. But how, then, do we come from kwe- to ku- in Nordic? By nothing less than a sound law that has never been acknowledged: we > u. Having established that sound law by numerous examples, I go deeper into it trying to explain some of its nature. I find that it consists of 2 reducive ten- dencies: non-fortis full vowel > ə, and -assimilation, the former being universal, the latter being semi-universal. But of course we shall never be able to explain why a sound law is active when, in fact, it is active. 7

Predicting PIE syllabification through phonotactic analysis

ANDREW BYRD University of California, Los Angeles

Gernot Schmidt (1973) was the first to address the two variants of the PIE word for ‘daughter’ attested in the Indo-European languages: *dhugh2ter- (Skt. duhitár-, Gk. thugáter-, etc.) and *dhukter- (Iranian duxtar-, Arm. dustr). He explains the loss of the laryngeal in certain reflexes of the PIE word for ‘daugh- ter’, by setting up a phonological rule for PIE (1973:54), whereby a laryngeal is lost in second position in a sequence of four consonants (CHCC > CCC). Pre- sumably this deletion would have occurred in the oblique stem of ‘daughter’, which is uncontroversially reconstructed as *dhugh2tr-. According to Schmidt, this process of laryngeal deletion also operates in initial clusters. For example, the PIE word for ‘father’ is uncontroversially re- constructed as *ph2té:r, with an oblique stem *ph2tr-. Almost all of the IE lan- guages that continue this root show a vocalic reflex of*h 2 in the initial throughout the paradigm (Skt. pitá:, pitr-, etc) while GAv. continues two dif- ferent stems pitar- (< *ph2ter-) and fǝδr- (< *ph2tr-, cf. dat.sg. fǝδrōi). Schmidt claims that the oblique stem fǝδr- is the direct reflex of *ptr-, from an earlier PIE *ph2tr- ‘father (dat.sg.)’. He further links YAv. tūriia- ‘father’s brother’ and Proto-Slavic *struyo- ‘id.’ as outcomes of the same process, where *h2 was de- leted in the sequence CHCC: *ptruya- < *ph2truya-. However, the loss of a laryngeal in the initial sequence *#CHCR- would have resulted in the sequence *#CCR-, which is the classic environment for the epenthesis of ‘schwa secundum’ (e.g. *mǵnó- > Lat. magnus ‘great’; Mayrhofer 1986: 175–6; Vine 1999: 11ff.) and therefore would have produced a phonotacti- cally illegal sequence in PIE. Moreover, is the only language to exhibit paradigmatic alternation between *ph2ter- and *ptr- and so it is conceivable that this allomorphy is a secondary development. Lastly, and most seriously, there are a number of counterexamples to Schmidt’s rule as provided by Hack- stein (2002); cf. *dhh1s-nó- > Lat. fānum ‘temple’ (via *fasno-) and related cog- nate Skt. dhiṣṇya-. 8 Andrew Byrd

The focus of my paper will be twofold. First, I will address these deficien- cies in Schmidt’s hypothesis and will attempt to generate a model that will better explain the data and also predict other instances of laryngeal deletion. My approach incorporates an idea that has been floating around the gen- eral phonological literature for some time now (cf. Hammond 1999: 68–9): the Decomposition Theorem (DT). The DT states that all medial consonant clusters should consist of an occurring word-final cluster and an occurring word-initial cluster. This means that a thorough phonotactic survey of all PIE word-initial and word-final clusters will allow us to predict which types of medial clusters were legal for the proto-language and which were not. Thus, the reason that *dhugh2trés ‘daughter (gen.sg.)’ simplified to*d huktrés was that the medial cluster did not consist of a legal coda plus a legal onset. An added benefit to such a phonotactically-based analysis of laryngeal deletion is that we gain further insight into the process of syllabification in PIE, a topic which we currently know little to nothing about.

References

Hackstein, Olav. 2002. “Uridg. CH.CC > C.CC,” HS 115. Hammond, Michael. 1999. The phonology of English: a prosodic optimality-theoretic approach. Oxford & New York: Oxford University Press. Mayrhofer, Manfred. 1986. Indogermanische Grammatik: I/2 Lautlehre. Heidelberg: Winter. Schmidt, Gernot. 1973. “Die iranischen Wörter für ,,Tochter“ und ,,Vater“ und die Reflexe des interkonsonantischen H (ǝ) in den idg. Sprachen,” KZ 87. 36–83. Vine, Brent. 1999. “Greek ῥίζα ‘root’ and “Schwa Secundum”,” in UCLA Indo-European Studies, Volume 1, ed. V. Ivanov and B. Vine, 5–30. 9

A new sound law of PIE: initial **h3w > *h2w

Paul S. Cohen & Adam Hyllested New York & University of Copenhagen

We formulate and give phonetic and phonological evidence for a new sound law of Proto-Indo-European, whereby an earlier initial sequence of **h3w > *h2w; in the process, we give arguments that bear on the phonetic makeup of *h2 and *h3. The PIE sound law has the benefit of explaining the infrequency (in fact, as we will show, absence) of initial PIE *h3w, and obviates any need for positing an ablauting a in the PIE root often given as h* 3wath2- [vel sim.] “to wound”. And, together with an Anatolian sound law we presented and sup- ported in Cohen & Hyllested (2006), the new sound law allows us to nail down the etyma of the PIE roots for “sheep”, “bird”, and “egg”, with no need for an ablauting a in “bird”.

References

Cohen, Paul S. & Adam Hyllested. 2006. “Initial h3 in Anatolian: Regularity in ostensible chaos”. Paper presented at the 18th Annual UCLA Indo-European Conference, Los Angeles, No- vember 2006. 10

On final ‑r in Vedic

Michael Frotscher University of Munich

Although word-internal sonantic -r̥- is – contrary to the other IE. languages – preserved as such in Ved., there is no evidence of ‑r̥ in word final position. Either originally final ‑r̥ is somehow extended by ‑t as in yákr̥-t- ‘liver’, śákr̥-t- ‘faeces’, agentive root-compound such as *-bhr̥-t etc. or ‑k as in ásr̥-k- ‘blood’, or formations which would expectedly exhibit final ‑r̥ are simply avoided as is the case of agentive root-nouns that do not show the t-extension, where we find the masculine form instead of the neutral one, cf. Acc.Sg. m.(!)bhójanam túram ‘vordringenden Genuß’ (RV 5, 82, 1d), gotrám hari-śríyam ‘den gold­ glänzenden Kuhstall’ (RV 8, 50, 10d). The only instances left to show the normal representation of final‑ r̥ are the few neutral r/n-heteroclitics (other than yákr̥t-, śákr̥t- and ásr̥k-) and the 3.pl. act. perf. ending ‑úr. Schindler, BSL 70 (1975), 8 argued for Ved. ‑ar as being the normal represen- tation of ‑r̥: ū́dhar ‘udder’, áhar ‘day’ < *-r̥. The objections made by Nussbaum that ‑ar is rather the continuant of a full-grade *-er, as in Lat. ūber, originally taken over from the oblique cases, Lat. feminis < *-en-es, is easily dismissed by the fact that the r/n-stems do not show any fullgrade suffix in the oblique cases, cf. Ved. g.sg. ū́dhnas. The- in- of the Lat. form is due to anaptyxis. But this leads to the question where ‑er in Lat. r/n-heteroclitics comes from. The normal representation of *‑r̥ in Lat. is ‑or > -ur. I would, however, reckon a double rep- resentation of PIE. final -r̥* showing up in Lat. as either ‑ur after labials iecur( < *i̯ekwr̥, femur) or ‑er after any other consonant iter( , aser, perh. acer ‘acorn’ if < *ák-r̥). This soundlaw only accounts for the absolute final. Problematic with regard to this sondlaw is ūber with its Lat. labial coming from Ital. dental *-þ-. It is therefore necessary to examine the relative chronology of the two develop- ments: Ital. *-þ- > Lat. ‑b- and *-r̥ > -or or ‑er resp. Is the double representation of PIE. *-r̥ already Proto-Ital. or only Lat.? Apart from that consideration, I will try to show that Lat. ūber is not at all a direct parallel to Ved. ū́dhar but rather a substantivised adjective Ital. *ūþeros ‘udder-like, fertile (metaph.)’ with sec- ondary (although rarely attested) transeferring to the 3rd inflection. The adj. Michael Frotscher 11

ūber ‘fertile’ is much earlier attested (Plaut., Ter.), the subst. ūber being only relativley late (Cic.+). Consequently, I think Schindler’s claim of ‑ar as being the normal represen- tation of *-r̥ still is sustained, the more so, as it perfectly parallels the develop- ment of final *-m̥ in Ved. I would, however, like to establish a special sound- law, according to which only unaccented final -r̥* > Ved. ‑ar whereas accented final -ŕ̥* > -úr. The only (problematic) evidence for the latter development is, admittedly, the 3.pl.act. perf. ending ‑úr. If there was, originally, no final ‑s in the ending (not *-r̥s), the conditioning factor for the development can only be the accent. In my paper, I will try to show that final ‑s is not to be expected. Perhaps, even the g.sg. ending of r-stems in ‑ur or ‑úr can be attributed to the effect of the accent, as most of the gen.-forms are oxytonic, so that the devel- opment can be understood as *-ŕ̥s > *-úr-s *-úr rather than *-ŕ̥s > *-úr due to the finals . If final -r̥* > -ar and *-ŕ̥ > -ur the t-extension in the above mentioned forma- tions cannot be due to phonological factors but must be explained morpho- logically. In my paper, I would like to present a morphological (analogical) so- lution to those t-extensions, which are also found in other daughter languages (cf. lat. ped-i-t- etc.). 12

Phonologische und morphologische Bemerkungen zu den hethitischen -i̯e/a-Verben

josé virgilio garcía trabazo Universidade de Santiago de Compostela

Ziel des vorliegenden Beiträges ist es, bei den Hethitischen sa-Verben mög- liche Reflexe der sogenannten “Essiva” (m.E. besser als “Fientiva” zu verstehen) auf *-h1-i̯é/ó- zu finden, die im Indoiranischen als Passiv grammatikalisiert worden sind. Morphologische Reflexe solcher Präsentien hätten wir z.B. bei ai. tāyáte ‘ausgedehnt werden’ < *tn̥-h1i̯é- (vgl. Stativ *tn̥-éh1- > lat. teneo, -ēre ‘halten’ usw.) und vielleicht auch bei der ai. Dehnung von kurzem i und u an der Morphemgrenze vor einem yod-Suffix, wenn die Regel doch aus dem Pas- sivum entstand (z.B. sruyáte ‘wird gehört’ < *k̂lu-h1-i̯é-; vgl. lat. cluēre ‘genannt werden, heißen’ usw.). Um die Hypothese zu überprüfen, verfügen wir über zwei Wege, und zwar am erster Stelle eine “morphologisch-vergleichende”, und dann eine zweite “phonologisch-synchronische” (das ist, die Überprüfung der Heth. Lenitionsregeln bei prähistorischen Akzentschwankungen). Das würde natürlich bedeuten, daß man mit zwei (morphologisch sowohl als diachronisch) verschiedenen Heth. Fientiv-Suffixe rechnen sollte: das ‘alte’ auf -se/a- (< *-h1-i̯é/ó-) und das ‘neue’ auf -ēšš- (< *-éh1-sh1-), obwohl doch mit derselben ‘Stativ’-Basis auf *-(e)h1-. Einige der möglichen Kandidaten zu ‘alten’ Hethitischen Fientiva wären, z.B. (<< bedeutet “indirekte” oder “umge- stalte” Herleitung) die folgende: arai-i / ari- ‘to (a)rise, to lift; to raise’ << IEh * 3r̥-h1i̯é/ó- (lat. orior [EDH (= Etymological Dictionary of the Hittite Inherited Lexicon, A. Kloek- horst 2008) 200]) 3pl.pres.act. a-ri-sa-an-zi (KUB 2.3 i 44, ii 29), usw. ẖarkii̯e/a-i ‘to get lost, to loose oneself, to disappear, to perish’ < IE [*h3erg- /*h3rg-,] *h3r̥g-h1i̯é/ó-(OIr. orgaid ‘to kill, to ravage, to devastate’, con·oirg ‘to smite’; Arm. (?) harkanem ‘to smite, to smash’) [EDH 306, HW2 (= Hethiti‑ sches Wörterbuch, J. Friedrich & A. Kammenhuber 1975ff.) e 297f.] 3pl.pres.act. ẖar-ki-i̯a-an-zi (OH/MS), 3sg.imp.act. [ẖar-]ki-e-ed-du (MH/ NS), usw. josé virgilio garcía trabazo 13

Als Kausativbildungen stehen die Ableitungen ẖarni(n)k-i und ẖarknu-i ‘to destroy, to ruin’. parkii̯e/a-i [, park-tta(ri)] ‘to raise, to lift, to elevate, to grow (trans.); to rise, to go up, to grow (intr.); to take away, to remove’ < IE [(midd.) *bherĝh-], (act.) *bhr̥ĝh-h1i̯é/ó- (Cluw. parrai̯a- (adj.) ‘high’ [< *bhr̥ĝh-ei̯-o-]; Arm. baṝnam, aor. ebarj ‘to raise’, barjr ‘high’, Skt. barh- ‘to make strong’, bṛhánt- ‘high’, TochB pärk- ‘to arise’, ON bjarg, berg ‘mountain’) [EDH 636f., CHD (= Chi‑ cago Hittite Dictionary, 1983ff.) P: 155f.]. Interessant ist doch die Bemer- kung bei CHD 157: “There appears to be no correlation between and transitivity”, was m.E. ein Indiz dafür ist, dass es sich um alte Aktionsart- Oppositionen handele (dazu auch die Kausativbildungen als alte Aktions- art-Oppositionselemente). 3sg.pres.act. pár-ki-i̯a-az-zi (MS), 3pl.pres.act. pár-ki-i̯a-an-zi (MH/MS), 1sg.pret.act. [pá]r-ki-i̯a-nu-un (NH), 3sg.pret.act. pár-ki-i̯a-at (OH/MS), usw. Kausativbildungen: parknu-i ‘to make high, to raise, to elevate’, parkii̯anu-i ‘to rise, to make rise’. Möglicher i-Adj. parkī (pár-ki-i) < IE *bhr̥ĝh-i- ?

Zu den Akzentschwankungen sollte man zuerst als wichtig – und echt alt – die Opposition zwischen **pék-i̯e- > ai. pácyate ‘wird reif’ und *pk-h1i̯é- > ai. pacyáte ‘wird reif / gekocht’ betrachten (vgl. gr. πέσσω, att. πέττω ‘lasse rei- fen, verdaue, koche’, ein Faktitives Oppositionsaktiv zu einem nicht erhalte- nen Medium, LIV 468, Fn. 4). So entsteht eine mögliche Erklärung dafür, daß z.B. Heth. ẖatuk-i ‘to be terrible’ (< IE *h2téu̯g-ti / *h2tug-énti; vgl. ved. tujyáte ‘erschrickt, flieht’; gr. ἀτυζόμενος ‘erschreckt [fliehend]’) nicht die erwartete Lesung **ẖa-at-tV- erweist (wie z.B. bei *h2p-ésr > ẖappeššar ‘Lippe’), sondern ẖa-tV- (wofür schlägt Kloekhorst in EDH 337 eine zu komplizierte Lösung vor): es handele sich um die erwartete Entwicklung aus *h2tug-h1i̯é/ó- mit Le- nition *t > d als Resultat der Akzentverschiebung bei den alten (Stativ-)Fientiv auf *-h1i̯é/ó-, der seinerseits später durch den ‘neuen’ Fientiv ẖatukēšš- ersetzt worden wäre. 14

The Compass Conspiracy, a footnote to Verner’s Law and some further observations on the reflexes of PIE *-sr-

Piotr Gąsiorowski Adam Mickiewicz University, Poznań

The paper reconsiders the fate of medial *-sr- in Germanic, especially in the context of Verner’s Law. It is argued that the epenthesis of *-t- took place later than the Vernerian voicing of *-s-, and that the further development of *-zr- involved no stop insertion. Several new etymologies are offered to sup- port this proposal, and some of its ramifications are explored, including the derivation of Lat. vēr ‘spring’ and the structure of the Germanic words for the four cardinal directions. Parallel developments in a few other branches of the Indo-European family are also discussed. 15

Raising before *µ in Old Irish: evidence and consequences

Aaron Griffith Universität Wien

The general conditions for raising and lowering of vowels in Old Irish are well- known. Although there is some disagreement (Schrijver 1995: 50–2), the most accepted formulation of the raising and lowering rules are: accented mid-vow- els raise to high vowels before a following high vowel, and high vowels lower to mid-vowels before a following non-high back vowel (McCone 1996: 109ff). The raising is more restricted than the lowering, both because the raised vowel must be accented and because the raising is blocked by voiceless and most consonant clusters. I would like to propose a further restriction on the raising rule: *μ blocks the raising of *e to *i. The data supporting this restriction are limited but rea- sonably compelling (reimir “as thick as”, leime “impotence, folly”, neimi “poi- sons”, do·eim “you protect”). The restriction is also phonetically plausible, since it seems likely that the highly marked sound *μ nasalized the preceding vowel, and nasalized vowels have fewer contrasts cross-linguistically than oral ones. Independent evidence further suggests that the contrast *ẽ / *ĩ was neutralized as a mid-vowel (Schrijver 1993). Though this is a very minor sound law, it has an outsize influence on our understanding of Irish historical grammar due to its interaction with the con- tested Insular Celtic *es > *is. The raising of unaccented*es to *is before vowel has been championed by McCone (1996: 99–100, 2006: 110–15) as the most straightforward means of accounting for a large number of verbal forms (e.g. *beresi > *berisi > biri). Schumacher (2004: 138–53), however, cit- ing OIr temel “darkness” and OW timuil < IC *temeselo- (not from *teμiselo-), has argued that the raising cannot have taken place in Insular Celtic or even in Pre-Irish. As a result, he must explain the various verbal forms differently and, I would argue, less plausibly. The restriction on raising before*μ as proposed here allows us essentially to compromise between McCone’s and Schumacher’s positions. While the Brit- ish evidence clearly militates against an Insular Celtic rule *es > *is, an early 16 Aaron Griffith

Pre-Irish raising is still possible. Inherited Pre-Irish *teμeselo- gave *teμiselo-, which regularly became OIr temel. At the same time, verbs forms like OIr biri are still derivable much as McCone argued (< *birisi (apocope) < *berisi (nor- mal raising) < Pre-OIr *beresi (via special raising of *es to *is)). Under the ac- count proposed here, some analogy is necessary to account for all of the forms McCone explains through his Insular Celtic *es > *is rule, but the analogy can be easily motivated. In summary, I suggest that the lack of raising of *e to *i before *μ allows us to explain a number of data directly. In addition, through the interaction of this rule with a Pre-Old Irish raising of *es to *is, a number of further forms fall out in a manner much more straightforward than would otherwise be pos- sible.

References

McCone, Kim. 1996. Towards a relative chronology of ancient and medieval Celtic sound change. Maynooth: Department of Old Irish. McCone, Kim. 2006. The origins and development of the Insular Celtic verbal complex. May- nooth: Department of Old Irish. Schrijver, Peter. 1993. “On the development of vowels before tautosyllabic nasals in Primitive Irish.” Ériu, 44: 33–52. Schrijver, Peter. 1995. Studies in British Celtic historical phonology. Amsterdam: Rodopi. Schumacher, Stefan. 2004. Die keltische Primärverben. Innsbruck: IBS. 17

The ruki-rule in Nuristani

Irén Hegedűs University of Pécs

Preliminaries

When dealing with the analysis of Nuristani etymological data it may be impor- tant to clarify some basic points of departure. a The classification of . Studies on Nuristani languages by G. Morgenstierne, D.I. Edelman, and R. Strand have accumulated sufficient and persuasive arguments for treating Nuristani as an early diverging, third Aryan branch, so in our treatment Nuristani is not part of Indo-Iranian, though it is obvious that Nuristani lan- guages have been influenced both by Iranian and Indo-Aryan languages; b The establishment of the relative chronology of the major phonological changes in the divergence of Aryan languages from Proto-Indo-European. The evaluation of the Nuristani etyma possibly affected by theruki -rule will be based on the following concept of the relative chronology of phonological evolution from PIE to Aryan: 1 the vocalization (or loss) of PIE laryngeals 2 the operation of Bartholomae’s Law 3 PIE *l > PIIr. *r 4 the operation of RUKI rule 5 satem palatalization

Research question

The operation of theruki -rule in Nuristani languages is a question that has led to the emergence of conflicting opinions varying between denying its operation and allowing the idea of its partial operation. Is it really true that Nuristani languages were only partially or not at all affected by the operation of the ruki-rule? 18 Irén Hegedűs

Structure of presentation

In an attempt to answer the research question the following points will be high- lighted:

1 the importance of the difference of the Nuristani phonological context from that of the Indic and in the Iranian branch in the treatment of PIE *s, 2 the conditions in which Nuristani š < *s irrespective of the ruki-rule (quasi ruki-forms), 3 the preservation of PIE *s after u (hypothesis to be promoted: the late loss of preconsonantal laryngeals in Nuristani).

Conclusion

– The limited amount of data available from Nuristani languages (especially of those that would be relevant for the examination of the ruki-rule) makes it difficult to formulate a strong conclusion, however, – our final evaluation leans towards the acceptance of the operation of the ruki-rule in Nuristani with the rider that 1 the true picture is often obscured by some internal phonological chang- es operating in the individual development of Nuristani languages, and 2 the ruki-rule may have been inoperative in a position after u* simply due to the inapplicability of the phonotactic context. 19

Hidden sound laws in the inflectional morphology of Proto-Indo-European

Eugen Hill Ludwig-Maximilians-Universität München

The lecture is aimed to show that solutions to morphological problems in re- constructed languages are often to be found not in the domain of morphol- ogy itself but rather in phonological developments once operative in the lan- guage in question. The particular problem to be discussed is the widely known asymmetry in the inflectional paradigm of the thematic present stems of PIE: while in the 2sg. and 3sg. active the desinences of the present indicative are constructed by simple adding an *-i to the corresponding desinences of the injunctive mood (ind. 2sg. *-es-i, 3sg. *-et-i ← inj. 2sg. *-es, 3sg. *-et), in the 1sg. the relation between the present indicative and the corresponding injunctive form is markedly different (ind. 1sg. -ō* or, possibly, *-oH ~ inj. 1sg. *-om). This asymmetry is even more embarassing if one considers the inflection of the athematic present stems where the 1sg. does not deviate (ind. 1sg. *-m-i, 2sg. *-s-i, 3sg. *-t-i ← inj. 1sg. *-m, 2sg. -s, 3sg. -t). The unexpected 1sg. present indicative of the thematic present stems in PIE *-ō or *-oH has been topic of many morphological hypotheses in which this desinence is mostly interpreted as a remnant of a more ancient inflection system which is assumed to have been originally entirely different from that of the athematic verbs. Hypoth- eses of this kind necessarily postulate an unusual amount of morphological restructuring in the inflection of the thematic present stems where only one inflectional form, the 1.sg., is believed to have retained its original shape. An entirely different solution to the problem was proposed in 1983 by W. Cowgill in his influencial paper on the thematic conjugation in PIE. Accord- ing to his view the unexspected shape of the 1sg. present indicative of the the- matic present stems is a result of a purely phonological development of an older *-om-i which is to be expected beside the corresponding injunctive des- inence *-om. Cowgill saw a further instance of the proposed sound change in the inflection of the thematic nouns where the instrumental pl. ends in PIE *-ōi̯s while *-o-mis should be expected due to the evidence of Baltic, Slavic and Germanic. However, Cowgill’s solution to the problem of the 1sg. present 20 Eugen Hill

indicative of the thematic present stems was widely ignored in the field due to the following obvious difficulties. The assumed sound change- * omi > *-ōi̯ > *-ō presupposes a syllabification which deviates from the syllabification rules known to have operated in PIE. The m-endings of the instrumental case – the only other instance of the proposed sound law – are traditionally viewed as a recent innovation of Baltic, Slavic and Germanic whereas more original bh‑endings are believed to be preserved in Indoiranian, Armenian, Greek, Ital- ic and Celtic. The lecture intends to show that both objections do not in fact invalidate Cowgill’s hypothesis. It can be demonstrated that the syllabification rules which are traditionally assumed for the latest stage of PIE did not operate in earlier times. Cf. the development known as Stang’s law in the acc.sg. forms such as *gʷóu̯m > *gʷṓm > Skr gā́m ‘cow’ but also in the 1pl. of nasal presents such as *k̂l̥nu-mé-s > *k̂l̥n-mé-s > Skr śr̥n-más ‘we hear’. The PIE morphology contains more independent evidence for a loss of *m before *i̯, cf. the oblique case forms of the thematic pronouns such as dat.sg. m. *tó-smo-ei̯ > Skr tásmai but f. *tó-si̯eh2-ei̯ > Skr tásyai where *tó-smi̯eh2-ei̯ is actually to be expected. As for the instrumental case of the thematic nouns, it can be shown that the m-endings of Baltic, Slavic and Germanic directly reflect the original situation whereas the bh-endings emerged secondarily by a sound change which has not been previously described in the literature. 21

The status and prevalence of Proto-Indo-European *a

Adam Hyllested University of Copenhagen

The basic root-vowel in PIE is universally reconstructed as e* , while *a is as- signed a marginal role. Many Indo-Europeanists deny the existence of a pho- neme *a altogether. This has severely challenged the importance of typological plausibility in linguistic reconstruction. Do languages with a single vowel pho- neme actually exist, and, if yes, isn’t this always /a/? The following traditional arguments against a phonemic a* might just as well be used in favour of it:

1 It is often claimed that a* does not participate in ablaut alternations. In fact it does (*Hi̯áĝ-e-ti ‘sacrifices’ ~ ppp. *Hi̯ĝ-tó-; *sā́l-s ‘salt’ ~ ASg. *sál-m̥; *-k̂as ‘in a row’ ~ *k̂os-mo-s ‘order’). Ascribing all instances of ablauting *a (and not those of *e) to “hidden” laryngeals is based upon a circular way of reasoning. And even if *a did not undergo ablaut, this would merely reveal that it behaved differently from e* , indicating that it was actually distinct. 2 *a is very rare in suffixes and endings. This is not typologically uncommon, and, again, a difference in occurrence need not indicate anything else than a difference in quality. 3 The evidence for a* is restricted to a few of the daughter-languages, espe- cially Celtic and Italic, known for having developed many cases of second- ary *a. That these languages are precisely the ones where a* is distinct from *e and *o in the first place should make one suspicious. 4 Many *a’s occur in words of expressive character. Even in such words, how- ever, few languages use vowels that do not already form part of their pho- nemic system. 5 Many lexemes with *a seem to have been borrowed from non-IE languages (e.g. Proto-Semitic *ϑawr > PIE *táu̯ros ‘bull’). It is noteworthy, however, that the allegedly unfamiliar vowel was not substituted with a domestic one (e.g. the way *ϑ- was replaced by *t-). 22 Adam Hyllested

6 *a appears in a rather limited set of positions, e.g. following *k. What has been overlooked is that *e rarely occurs in the same position. This speaks for a case of allophonic variation.

My paper will examine whether we can eat our cake and have it. Is there in fact a way to accept the reconstruction of *a without violating basic principles of phonological typology? It will be argued that *e and *a originally occurred in complementary dis- tribution, the timbre depending on the consonantal surroundings. Perhaps quite surprisingly, *e turns out to have been the marked vowel, while *a was the “elsewhere-”, meaning that the sequences in which it occurred had nothing particular in common, but simply constituted what was left when the original vowel came to be realized as *e in certain positions. The key point here is that, even though e* outnumbered *a in terms of pure occurrences, *a was the predictable outcome in a larger set of positions. In fact, the phonemic quality of the vowel might as well be interpreted as /a/ in early PIE. One of the corollaries of my analysis is that while *táu̯ros is indeed a Se- mitic loanword, we should still expect a-vocalism, had it been inherited. 23

The development of Proto‑Celtic *sk in Brythonic

ANders Richardt Jørgensen University of Copenhagen

The question of the development of sk* in Brythonic, more specifically why we sometimes find initial *xu̯‑ and medial *‑x‑ (instead of preserved *sk), is still a matter of some debate. It will be argued that the conditioning factor is a following front vowel. It will furthermore be argued that we are dealing with a form of palatalization, a phonetic parallel being provided by Swedish [xʷ‑] from sk‑ + front vowel. This explanation is extended to medial position, where the conditioning factor is quite similar (*‑sk‑ + *-i̯-). As a consequence, the various outcomes seen in Brythonic, namely initial *sk‑, *xu̯‑, medial *‑sk‑, *‑x‑ can all be derived from a simple PCelt. *sk (which is preserved in Old Irish). This has implications for the reconstruction of PIE *‑sk̂e/o‑presents to roots in final velar, where Brythonic does not, as has tradi- tionally been assumed, provide evidence for the preservation of the root final stop. 24

Greek ὀπυίω and other words without a digamma

Joshua T. Katz Princeton University

The Greek word ὀπυίω, conventionally translated as ‘marry, take as wife’, lacks an etymology on which everyone agrees. One view, given with a question mark in ²LIV (M. Kümmel) is that it goes back to *h₃pus-i̯é/ó- (the PIE root would be ?*h₃peu̯s- ‘sich mehren, reich werden an’) and thus makes an equation with Ved. púṣyati ‘flourishes’. My goal in this paper is to make plausible the al- ternative – hitherto unconsidered, as far as I know, except independently by Michael Janda (p.c.) – that the Vedic word with which ὀπυίω is cognate is in- stead vapuṣyo ‘gaze in wonder at’. As for vapuṣyo, though evidently a derivative of vápuṣ- ‘wonderful (adj.); beauty (noun)’, its root etymology is unfortunately not clear either, but I propose to connect it to √VAP¹ ‘strew’, i.e., PIE *u̯ep-, which 2LIV presents with a question mark and states is found only in Indo- Iranian. The morphology of this suggestion is relatively straightforward, but the se- mantics and phonology evidently are not: for one thing, ‘marry’ would not seem to be close in meaning to ‘strew’; and for another, the Gortyn Law Code offers prima facie evidence against the existence of a digamma in the Greek verb. But let us see … 25

Evidence for non-linear phonological structure in Indo-European: the case of fricative clusters

Götz Keydana Georg-August-Universität Göttingen

In my talk I argue for the neccessity of reconstructing hierarchical phonologi- cal structure for onset clusters containing in ancient IE languages and PIE. In the first part I discuss/s/ in (O) clusters. I start out with a short sketch of reduplication in ancient IE languages and PIE: Reduplication was a process of copying melodical specifications from the root into the reduplicant, a prefix which was phonologically specified as a CV-template with melodical information linked only to the V-slot specifying the vowel as /e/. The C-slot was filled by a copy of the melodical speci_cation of the least segment in the onset of the root to which the prefix attached. As is well known, s-clusters behave in different, but always deviant, ways with resepect to reduplication throughout the ancient IE languages. I show that this behaviour can only be explained by assuming hierarchical structure in the organization of phonological segments: /s/ in onset clusters was subseg- mental in languages like Germanic and Latin (the latter showing traces of an older pattern with extrasyllabic /s/). In Greek and Indo-Iranian /s/ was extra- syllabic. In other words, these languages allowed for non-moraic (the so-called semisyllables or Nebensilben of Sievers’) constituted by a sibilant and licensed (at least) in the left margin of the prosodic word. External reconstruc- tion leads to the proposal that PIE */s/ in sO-clusters was extrasyllabic, too. Next I demonstrate that semisyllables aid in explaining further peculiari- ties of sO-clusters: sO-clusters violate the sonority sequencing principle: In IE syllable structure there is always an increase in sonority from the left syllable margin to the nucleus; only sO-clusters show a decrease. They also infringe upon a constraint on onsets with more than two segments otherwise undomi- nated in PIE. Both sets of data are easily explained under the assumption of semisyllabicity in PIE. In the second part of my talk I extend the investigation to laryngeals (H) in onset clusters. I show that they behave similar to the sibilant in reduplica- 26 Götz Keydana

tion: With roots beginning in HO, the C-slot of the reduplicant is always filled by a copy of the melodic specification of the . With roots beginning in H + sonorant, on the other hand, the melodic specification of the laryngeal is copied. Therefore laryngeals must be more sonorous than , but less sonorous than : With respect to the sonority hierarchy they pattern with /s/. The similarity between laryngeals and the sibilant extends to the way they are treated in the process of syllable structure building. Again reduplication serves to illustrate this point: The attested patterns show that laryngeals were subsegmental in any type of onset cluster in Pre-Greek, while in Indo-Iranian they were part of the syllable onset when followed by sonorants and extrasyl- labic when followed by obstruents. The Indo-Iranian pattern can be recon- structed for PIE. We may therefore conclude that laryngeals formed a natural class with /s/ in PIE. Concerning syllable structure our observations lead to the following generalization: PIE allowed for non-moraic semisyllables just in case they contained a fricative (the maximal structure of PIE semisyllables remains to be established, the type *h2ster- may – but must not neccessarily – point to onset clusters). This PIE pattern can be backed by ample typological evidence. In the last part of my talk I return to s-clusters once again. I will tentatively argue that s-mobile can be explained if we take into account the special sta- tus of /s/ in clusters with respect to syllable structure. Far from being able to present a thorough explanation for s-mobile, I show that the traditional ap- proaches to this phenomenon suffer from being highly improbable in the light of our knowledge of PIE syntax. Semisyllables, on the other hand, are cross- linguistically highly marked. The instability of/s/ in onsets of roots is therefore easily accounted for, if we assume hierarchical structure in PIE phonology. 27

The phonological interpretation of plene and non-plene spelled e in Hittite

Alwin Kloekhorst Leiden University

Within Hittite linguistics, plene writing, i.e. the extra writing of the vowel of a CV or VC sign by its separate V-sign, e.g. me-e-ni, te-e-ez-zi, e-eš-zi, har- ga-e-eš, has always been a hotly debated subject. Nowadays, it seems to have become more and more consensus that plene writing should be interpreted as indicating , mēni, tēzzi, ēšzi, hargaēš, and that this vowel length is caused by accentuation. But why do we then also find words without any plene written vowel (e.g. me-er-zi, še-ep-pí-it), or words with multiple plene vowels (e.g. te-e-pa-u-e-eš)? In the case of -e-, it has been suggested that the plene vowel is also used for disambiguation of ambiguous signs (i.e. CV and VC signs that can be read with an e- as well as an i-vowel: CE/I, E/IC), and that in these cases the plene vowel does not indicate vowel length. Does this mean that e-EŠ-zi represents ēšzi with a long vowel, but e-E/IP-zi = epzi with a short vowel? Moreover, it has often been stated that the practice of plene spelling is decreasing in time because of a change in scribal habits. Why does this indeed seem to work for e.g. pé-e-eh-hi (which is attested with plene spelling in OS texts only, but is spelled pé-eh-hi, without plene -e-, in MS and NS texts), but not for ke-e-el (which is consistently spelled with plene -e-, in texts of all periods)? Is this really due to scribal peculiarities, or must these differences rather be regarded as betraying diachronic phonological processes? As we see, there are still many questions that can be asked when it comes to plene spelling in Hittite. In this paper, I will present the results of my investigations into the plene and non-plene writing of the vowel -e-, in which all forms with an e-vowel, whether written plene or not, have been taken into account. I will discuss the different uses of plene and non-plene writing of -e-, assess their inner-Hittite diachronic developments and explain how they should be phonologically interpreted and how they can subsequently be used for etymological purposes. 28

Parthian ž

Agnes Korn University of Frankfurt a.M.

It has generally been assumed that Old Iranian ž, j and postvocalic č all come out as /ž/ in Parthian (or generally in Middle Northwestern Iranian) while they yield /z/ in Middle Persian, and that these processes are parallel. This paper will attempt to show that the developments of OIr. j and postvo- calic č in Middle Persian and Parthian cannot have been parallel, and will ar- gue that Parthian preserves two distinct /j/ ( [j], [ž]) < j, ž) and /č/ in all positions of the word. The development in Parthian is also relevant for the grouping of West Ira- nian languages as some contemporary languages point towards a preservation of OIr. j as an affricate in Middle Northwestern Iranian. 29

The rise of the Germanic iterative verbs: another case of ablaut being translated into consonant gradation

Guus Kroonen Leiden University

The origin of the Germanic class two iterative verbs has been debated for over a century. The controversy primarily seems to be the result of the eccentric poly- morphism displayed by the iterative verbs, e.g. OHG zogōn ‘to drag’ < *tugōn-, ME toggen ‘to tug’ < *tuggōn-, OHG zockōn ‘to jerk’ < *tukkōn-, LG tuken ‘to tear, pluck’ < *tukōn-. This polymorphism has been ascribed to “expressivity” or to the influence of an unknown substrate language. Still, its systematic mor- phological patterning, its strict limitation to class 2 verbs, and its unambiguous semantics indicate that Osthoff’s old linkage of the Germanic iteratives with the Indo-European n-presents must be correct. 30

Typology and reconstruction: the IE consonants and vowel

Martin J. Kümmel Albert-Ludwigs-Universität Freiburg

When we reconstruct a proto-language, we produce a hypothesis about a non- attested synchronic stage. That the probability of this hypothesis should be compatible with typological knowledge about synchronic systems is certainly reasonable. But first of all, reconstruction is a hypothesis about non-attested changes, and its primary goal is to explain attested language data by assuming these changes. Therefore, the changes themselves should also be compatible with typological knowledge, i.e., knowledge about diachronic typology. Unfor- tunately, this kind of knowledge is less easily accessible than information about synchronic typology. In this paper I shall try to look at some problems of phonological recon- struction in PIE and what typological considerations might tell us about them together with internal evidence. Of course, the well-known controversial topics of PIE consonantism will be addressed: The stop system and the “glottal” mod- els, the Centum–Satem problem, and the “laryngeals” (especially their various effects and what they really might tell). These problems will be discussed from a diachronic typological viewpoint, but also some recent claims about later IE reflexes of these sounds are critically examined. In contrast to this, the reconstruction of vowels and their prehistory seems to be rather less controversial, but quite some problems have remained unclear in this field. In the light of typological parallels I shall try to hint at possible new solutions that might allow a better understanding of PIE phonological rules and some reflexes of PIE vowels in the extant IE languages. 31

Optimale Onsets im Indogermanischen

Rosemarie Lühr & Susanne Zeilfelder Friedrich-Schiller-Universität Jena

Innerhalb der Optimalitätstheorie sind Reduplikationssilben Beispiele für das Prinzip „Emergence of the Unmarked“. Da man annimmt, dass Redu­plika­ tions­silben keinen Input haben, sollten nur unmarkierte lautliche Strukturen im Output erscheinen. Wenn dies ein allgemeines Prinzip ist, erwartet man dessen Geltung auch in den indogermanischen Sprachen. In der Tat trifft es z.B. für den Reduplikationssilbenvokal ai im Gotischen zu, der sicher für ein kurzes offenese steht. In ähnlicher Weise würde man bei den Konsonanten un- markierte Strukturen erwarten. Insbesondere sollten Anlautcluster der Wur- zelsilbe in der Reduplikationssilbe vereinfacht sein. Doch gibt es ganz unter- schiedliche Repräsentationen, wie Fortsetzer des indogermanischen Perfekts im Gotischen, Griechischen, Lateinischen und Altindischen zeigen. Besonders die Behandlung von Clustern aus s + Verschlusslaut bildet hier Probleme. Im Vortrag werden im Rahmen der Optimalitätstheorie die einschlägigen Cons- traints und deren Ranking vorgestellt. 32

The development of Proto-Slavic word- final *-ī, *-ě from PIE *-oi̯, ‘*-ai̯’

Marek Majer University of Łódź

In Slavic, there seems to be an unpredictable distribution of final *-ě and *-ī that appear in the positions where the older models of reconstructed Proto- Indo-European had the diphthongs *-oi̯ and *-ai̯ (the latter sequence must be rewritten as *-h2ei̯ etc. in today’s terms). Numerous solutions to the problem have been offered, none of them completely satisfactory. The paper suggests a framework of the Proto-Slavic development of the above-mentioned PIE word-final diphthongs in which it is assumed that PIE ‘*-ai̯’ and *-oi̯, while having merged medially and initially, remained distinct in word-final position (a development parallel e.g. to the treatment of ā* and *ō in Celtic). Two simple phonological rules are argued for:

PIE ‘*-ai̯’ > PSl. *-ě PIE *-oi̯ > PSl. *-ī

Apparent exceptions (limited to two cases seemingly violating the second rule) may actually be plausibly accounted for on morphophonemic grounds. No such explanation seems to be available for the forms that confirm the rule PIE *-oi̯ > PSl. *-ī, which strongly points to the fact that this was the only regular, phonological development. By explaining the two apparent cases of PSl. *-ě where *-ī would be expected (a development which may, in fact, roughly be codified under a separate sound law), a coherent and complete picture of the development of PIE word-final diphthongs in Slavic may be arrived at. 33

Some cases of the converse of Sievers’ law in Greek

Biliana Mihaylova Sofia University

The converse of Sievers’ law is a very debatable phenomenon in Indo-Europe- an linguistics. In Greek we find some examples of it mentioned by M. Peters (1980: 290, fn. 243). The aim of this paper is to provide evidence for the action of the converse of Sievers’ law in Greek analyzing four cases where it appears in the groups iy, uw. The application of this rule in etymological research could shed more light on the origin and word formation of Greek words considered until now obscure or ascribed to the Pre-Greek substratum languages. Three of the instances examined in the paper are derived from the IE root *tewh2- ‘to swell’:

1 Gk. σάμος ‘height’, cognate to Lat. tama ‘tumour, swelling’. Persson (1912: 471) derives it from IE *twəmos. However the sequence *twh2mos would result in *tuh2mos > *tūmos. This problem can be solved by assuming the action of the converse of Sievers’ law: *tuh2-em-o-s > *tuwamos > *twamos > σάμος. 2 Gk. σάλος ‘rolling swell of the sea’ for which Persson (ibidem 484) poses *twəlos, but we should reconstruct a protoform *tuh2-el-o-s > *tuwalos and after the action of the converse of Sievers’ law *twalos > σάλος. For the se- mantic development, cf. Lat. tullius ‘flush, torrent’, OIcel. fimbul-þul ‘myth- ical river’, OIr. ton ‘wave’. 3 Gk. σάϑη ‘membrum virile’ goes back to *twadhā which results from *tuwadhā < *tuh2-edh-eh2 after the action of the converse of Sievers’ law. The semantics could be supported for example by Gk. φαλλός, φαλής ‘membrum virile’ from the IE root *bhel- ‘to swell, grow’. 4 Gk. ζάλη ‘squall, storm, driving rain’ may be derived from the IE root *deyh2-, *dyeh2- ‘to swing, move’. In this case we should reconstruct *dih2‑el‑eh2 > *diyaleh2 and by the action of the converse of Sievers’ law *dyaleh2 > ζάλη. From semantic point of view the association of the notion for ‘storm’ to a root signifying ‘to swing, move, and precipitate’ seems very probable. 34 Biliana Mihaylova

From the point of view of word formation, it is obvious that the four lex- emes analyzed above present a parallel structure. The root shows zero grade, while the suffix full grade. This fact leads us to the conclusion that the forms belong originally to the hysterodynamic type of inflection in Indo-European. The forms presenting the structure u/iH-eR/C-C like the four lexemes ana- lyzed above result from the accusative forms which replaced those of the old nominative Cew/yH-R/C-. This process of substitution is well known, cf. for example Gk. παλάμη ‘palm’ < *pl̥h2-em-.

References

Persson, P. 1912. Beiträge zur indogermanischen Wortforschung. (Arbeten utgivna met understöd af V. Ekmans Universitetsfond 12). Uppsala. Peters, M. 1980. Untersuchungen zur Vertretung der indogermanischen Laryngale im Griechischen (Sitzungberichte der Österr. Akademie der Wissenschaften, Phil.-hist. Kl., 377). Wien. 35

On the morphophonemics of Proto-Indo- European *-sḱe/o- suffixed presents

Paolo Milizia Università degli Studi di Cassino

The LIV records 612 roots ending with a stop out of a total of 1182 (51.78%); if we consider the subset of the 74 recorded roots that have an (attested) de- rived present stem in *-ske/o-, we find that those ending with a stop are only 26 (35.14%). As the chi-square test shows, this discrepancy is hardly due to chance. It is thinkable that the various phenomena of consonantal simplification that the -Csk- clusters may have undergone in the pre-history of the different IE languages affected the morphotactic transparency of the inherited stems in *°C-+-ske/o-, thus weakening their relation with the rest of the paradigm and, as a result of that, undermining their diachronic stability. On the other hand, it is indisputable that in PIE the suffix sḱe/o*- - was applicable to stop-final roots, as *pr̥k-ské- (from *perk- ‘ask’, cf. OI pr̥ccháti, Av. pərəsā, Lat. poscō; LIV 490f.; IEW 821f.) testifies. Then the question arises whether in the proto-system theCsk - - clusters underlyingly contained in such stems were allowed to surface intact or were subject to the action of (morpho-)phonemic constraints. Synchronic deletion of root-final stops, or at least of dorsal root-final stops, seems to be suggested by the historic continuations of *pr̥ḱ-sḱé-, which point to a simplified surface form *pr̥sḱé-. In addition to that, some attested -sḱé- stems derived from roots ending with a voiced or voiced aspirate stop are noteworthy:

1 *bher̥ǵ- ‘fry, roast’: *bhr̥ǵ- + -sḱé- → ?*bhr̥zǵé- > OInd bhr̥jjáti ‘fries, roasts’ (LIV 78; IEW 137); 2 *bhlei̯g- ‘shine’: *bhlig- + -sḱé- → ?*bhlizǵé- > Lith blizgù ‘flicker’ LIV( 89, sv. ?*bhlei̯g-, n. 4; IEW 156f.; LitEW I: 46, sv. blaikštytis); 3 *bhreh1ǵ- ‘(be) white’: *bhr̥h1ǵ- + -sḱé- → ?*bhr̥h1zǵé-, cf. OChSl probrězgnǫti ‘dawn’ (LIV 92, sv. *bhreh1ǵ-, n. 4; IEW 139f.); 4 *kʷendh- ‘suffer’: kʷn̥d* h- + -sḱé-→ ?*kʷn̥(dh)zǵhé- > Grk πάσχω ‘suffer’ LIV( 390; IEW 641); 36 Paolo Milizia

5 *mei̯ḱ-/mei̯ǵ- ‘mix’: *miǵ- + -sḱé-→ ?*mizǵé- > Grk μίσγω ‘mix’ (LIV 428; IEW 714); 6 *regh- ‘sich aufrichten’ (LIV): *r̥gh- + -sḱé- → ?*r̥zghé- > Grk ἄρχω ‘be first’(LIV 498; IEW 854, 863); 7 *tu̯engh- ‘oppress’: *tu̯engh- (with second. e gr.) + -sḱé- → ?*tu̯enzǵhé- > Av ϑβązjaiti ‘fall in distress’ (LIV 655 sv. *tu̯enk-; IEW 1099f.); 8 *u̯eig- ‘in Bewegung geraten’ (LIV): *u̯ig- + -sḱé- → ?*uizǵé- > Lith vizgù ‘tremble’ (LIV 667f.; IEW 1130f.; LitEW 1264f. sv. viskė́ti).

Leaving aside minor problems such as the reflex of the dorsal stop in Av ϑβązjaiti (depalatalization of ǵ(h) after z – see AirWb. 798, or replacement of *tu̯enzǵh by *tu̯enzgh by analogy with the base *tu̯engh-), each of these stems is straightforwardly explainable by assuming a progressive voice assimila- tion accompanied by the same root-final stop deletion phenomenon shown by *pr̥sḱé-. Although being usually dealt with as outcomes of branch-specific sound changes – especially in the cases of Grk πάσχω and μίσγω (cf. Rix 1976: 94f.) and of Av. ϑβązjaiti (extension of Bartholomae’s law to ghsḱ clusters?) – these developments, if taken as a whole, may be interpreted as evidence for a particular tendency of the suffix sḱé-*- to take on the laryngeal features of an (underlyingly) preceding obstruent. A further step could consist in taking into consideration the possibility that, already at the PIE stage, the affixsḱé- - was associated to a specific suffixation rule requiring root-final stops – or at least dorsal root-final stops – to be deleted and their laryngeal features to be linked to the obstruents of the suffix. From a typological point of view, instances of progressive voice assimila- tion such as those hypothesized here are interpretable as the product of the interaction between the constraint dispreferring obstruent clusters that do not agree in voicing and the constraint according to which root faithfulness may, at least under certain conditions, dominate affix faithfulness (s. Beckman 1999).

Bibliographical abbreviations

AirWb = Christian Bartholomae, Altiranisches Wörterbuch, Strassburg, 1904. IEW = Julius Pokorny, Indogermanisches etymologisches Wörterbuch, vol. 1, Bern, 1959. LIV = Helmut Rix et al., Lexikon der indogermanischen Verben, 2nd ed., Wiesbaden, 2001. LitEW = Ernst Fraenkel, Litauisches etymologisches Wörterbuch, Heidelberg, 1962–65. Rix, Helmut (1976), Historische Grammatik des Griechischen, Darmstadt. Beckman, Jill N. (1999), Positional Faithfulness, New York. Mayrhofer, Manfred. 1986. Indogermanische Grammatik: I/2 Lautlehre. Heidelberg: Winter. 37

Vowel reduction and deletion in Sabellic: history and chronology

Kanehiro Nishimura Kyoto University / SJPS

This paper focuses on both vowel reduction and deletion in Sabellic. Despite the traditional view (see, e.g., Solmsen 1894: 153) that vowel reduction was a rather restricted phenomenon there (which is partly true), I conclude that this vocalic change was already operative in early Sabellic, such as Pre-Samnite (e.g., ϝολαισυμος [personal name?]), Paleo-Umbrian (e.g., setums [personal name]), and South Picene (e.g., estuf and estufk‘here’), and that it was particu- larly noticeable in labial context. The application of vowel reduction to non-in- itial syllables, both word-medial and word-final, is construed as reflecting that the operation driven by initial stress was already active in these early dialects. In regard to syncope, scholars have traditionally assumed that final syncope predated medial syncope, which was dated as late as the fifth century b.c.e. However, I present several instances of medial syncope that could go back to the sixth century: e.g., Pre-Samnite imperatives ο(vac.)fρι[-]qτο[δ, (σ)τα[.] ιοσqτοδ, South Picene aitúpas (cf. Umbrian eitipes ‘they resolved’), múfqlúm ‘monument(?)’, mefiín ‘middle’, uelaimes [personal name]. The antiquity of medial syncope in these forms leads us to reconsider the alleged chronological gap between syncope in medial and final syllables. Oscan túvtíks ‘belonging to the community’ < *toutikos, for instance, preserved the vowel (*-i-) in the medial syllable, while it lost one (*-o-) in the final syllable. This case appears to justify the claim (see Benediktsson 1960: 197, et passim) that syncope first removed the vowel in final syllable (*-o-), closing the preceding syllable, and thus impeded syncope there (i.e., -í- was preserved). But I tentatively suggest that syncope in Sabellic targeted both medial and final syllables synchroni- cally (as vowel reduction; see above) and that for túvtíks the heavy consonant cluster **-tks was avoided, which would have occurred if both vowels had been syncopated. Further, this antiquity of vowel reduction and deletion in Sabellic, as early as 550–500 b.c.e. or earlier, is taken to reflect that Sabellic probably developed the initial-stress system on its own, independently from Etruscan 38 Kanehiro Nishimura

(cf. Skutsch 1913: 190, Sommer 1914: 86, and Prosdocimi 1986: 612, 614), where vowel reduction only came into effect around that period of time.

References

Benediktsson, H. 1960. The vowel syncope in Oscan-Umbrian.Norsk Tidsskrift for Sprogvidens‑ kap 19: 157‑297. Prosdocimi, A. L. 1986. Sull’accento latino e italico. In o-o-pe-ro-si: Festschrift für Ernst Risch zum 75. Geburtstag, ed. Annemarie Etter, 601–618. /New York: Walter de Gruyter. Skutsch, F. 1913. Der lateinische Accent. Glotta 4: 187–200. Solmsen, F. 1894. Studien zur lateinischen Lautgeschichte. Strassburg: Verlag von Karl. J. Trüb- ner. Sommer, F. 1914. Handbuch der lateinischen Laut- und Formenlehre: eine Einführung in das sprach­wis­sen­schaft­liche Studium des Lateins. 2nd and 3rd ed. Heidelberg: Carl Winter. 39

Remarks on the PIE amphikinetic nouns

Georges-Jean Pinault École Pratique des Hautes Études

Leaving aside differences of terminology, one may define in the system of PIE nominal stem formation an accent-ablaut class characterized by accent on the root in the strong cases, accent on the ending in the weak cases, with a special pattern of the locative singular, featuring accent on the suffix and zero grade of the root; in addition, the suffix has the o-grade in the strong cases. This inflec- tional type has been called amphikinetic (or holokinetic, or amphidynamic). It is exemplified by nouns that can be reconstructed for the proto-language, as ‘earth’, ‘dawn’, ‘path’, etc. If one follows the system proposed by Jochem Schin- dler, this type has several common features with the hysterokinetic class:

1 they share the same pattern for the weak cases and the locative singular; 2 both hysterokinetic and amphikinetic nouns are internally derived; 3 both classes serve to build derivatives with the same functions, to wit ab- stracts/collectives, agent nouns and animate action nouns.

Given the morphological and phonological possibilities, the relationships of the three first members of an internal derivational chain (acrostatic → proter- okinetic → hysterokinetic) are governed by a simple rule: the strong stem of the next class is based on the weak stem of the preceding class. This rule stops with the amphikinetic class, since its strong stem features the normal e-grade of the proterokinetic class, and it has the same weak stem as the “preceding”, i.e. hysterokinetic class. Therefore, the amphikinetic class looks as an “addition” to a system of internal derivation that would be ideally complete with the hyster- okinetic class. One may even go so far as to assume that there never existed a pure amphikinetic pattern, and that it is a mixed category, or a variant of the hysterokinetic type. It is anchored in the PIE derivational system, however, as shown by the formation of numerous stems. The paper will try to explore the motivation for the formation of the am- phikinetic type by analyzing the functions of the representatives that can be attributed to the proto-language. The e-grade of the strong stem can be related to a more general morphophonemic device, that makes derivatives based on 40 Georges-Jean Pinault

nouns with zero grade of the root. The o-grade of the suffix in the strong cases can be explained by an independent rule, that presupposes the accent on the preceding syllable. The hypothesis would be that the amphikinetic type served originally to build animate substantives as opposed to neuter substantives and to adjectives. The picture was blurred by two later trends: 1 a part of these animate substantives were transferred as collectives, while remaining substantives, as opposed to different singular stems; 2 some of them remained animate, but became adjectives through the use of these substantives to qualify other animate nouns. Accordingly, the patterns of internal derivation for the amphikinetic stems were restructured. This case will be exemplified by the fate of several catego- ries. The scenario may explain, among other things, the secondary differentia- tion between the two types of *-te/or- agent nouns (according to Tichy, 1995), that are both internally derived, as parallel to other pairs of hysterokinetic vs. amphikinetic types that do not show the same functional contrast as these agent nouns. 41

Germanic and the ruki dialects

Charles Prescott University of Sussex

A new relationship is proposed between Germanic, which shows lowering and retraction of vowels before /r/, /w/, /x/, and the ruki dialects of Indo-European, Baltic, Slavic and Indo-Iranian, which have retraction of consonants following /r/, /u/, /k/, /i/. There are two aspects to the relationship: the Germanic seg- ments correspond directly to three of the four ruki segments and the effects on vowels in Germanic and on consonants in the ruki dialects may be attributed to the spread of a common phonological feature, Retracted Tongue-root (RTR). The relationship may be illustrated by related words in Gothic and San- skrit:

• for the effect on vowels in Germanic, Gothic aúhsa, ‘ox,’ is taken to show a lowered allophone of /u/, aú, before /x/, h, the Germanic reflex of IE /k/; • for the effect on consonants in theruki dialects, ukṣā, ‘bull’, shows a retracted allophone of /s/ after /k/, assimilated to a retroflex phoneme, ṣ.

The association of Germanic with the ruki group is shown in bold in Figure 1 as Northern Indo-European (nIE). Though /r/ may often have the feature RTR, its addition to /i/, /u/ (making them lower) and to /k/ in the ruki group is seen as an arbitrary dialect marker. The posited lower realisation of /u/ may be a reason, by compressing the vowel space below it, for the widespread merger /o/ and /a/ in ruki dialects. Germanic has evidence for RTR in /r/, /w/ (consonantal /u/) and /x/ (the reflex of IE /k/) but not in /i/ or vocalic /u/. That it also has merger of /o/ and /a/ may suggest that the earliest Germanic shared more fully in the ruki dialect marking. 42 Charles Prescott

Figure 1 Core IE dialects (ie. excluding Anatolian and Tocharian)

centum satem, ruki nIE Germanic Baltic / Slavic Indo-Iranian ruki Celtic Albanian Armenian satem sIE Italic Greek centum wIE eIE 43

The “Saussure Effect” in Indo-European languages, other than Greek

Tijmen Pronk Leiden University

In 1905, observed that a root-final vowel was dropped if the root had o-grade. The focus of de Saussure’s formulation is of course Greek, since that was the language which provided evidence for the root-final vowels which we now know to be reflexes of the laryngeals. In laryngealist terms, de Saussure’s observation was described by Nussbaum as follows: “*H shows a vo- calic outcome in neither the environment #_Ro nor in the environment oR_C.” (1997: 182). Nussbaum coined the term “Saussure effect” for this phenomenon. Since Meillet it has been observed that the rule might apply to other Indo- European languages than Greek as well. Today the “Saussure effect” is applied throughout Indo-European whenever we expect a trace of a laryngeal in the vicinity of an o-grade but do not find one. The phonetic improbability of the effect, however, invokes skepticism about its existence. I will show that there is an acceptable alternative for all adduced examples of the “Saussure effect” outside Greek. I will briefly discuss all ex- amples which appear to be most convincing, with a special focus on the Latin examples. 44

Rule ordering paradoxes in Indo-European: reintroducing phonetics

Jens Elmegård Rasmussen University of Copenhagen

Why prs.ptc. *-n̥t-s of reduplicated verbs? The rules make us expect **-ent-z > *-ont-z > *-ōnt-z > *-ont-z, so what is wrong? Why *dóh₃-tōr, gen. dóh₃-tr̥-s (habitual type)? The rules would predict nom. *dēh₃-tér-z > *dēh₃-tḗr-z > *deh₃-tḗr-z > *déh₃-tēr-z > PIE *dóh₃-tēr. What has to be adjusted? The gen. is fine: dēh** ₃-ter-ós > *deh₃-tr-ós > *déh₃-tr-os > *déh₃-tr-s > PIE *dóh₃-tr̥-s. Why ablaut ḗ/é in the s-aorist? The z-induced lengthening works in nomi- natives after the weakening of unaccented -e- to -o-, so why not ḗ/ó ? Or does that perhaps exist? These and other examples will be addressed in the paper, and solutions suggsted. The most important seems to be: Sharpen the phonetics: there may be room around, and rules may have to be formulated in subphonemic terms. 45

Laryngeal features in stops of Armenian dialects

Giancarlo Schirru University of Cassino

Modern Eastern Armenian has a system of 15 stops: they can be classified in five classes, contrasting in and kind of release features, and three series with different laryngeal features. In the standard variety, based on the Erevan dialect, the three series are generally described as following: 1) voiced: /b, d, dz, dʒ, g/; 2) voiceless: /p, t, ts, tʃ, k/; 3) voiceless aspirate: /ph, th, tsh, tʃh, kh/ (s. Vaux 1998: 12–13; Hacopian 2003). Already in the end of 19th century, scholars have revealed the existence of eastern Armenian dialects where the 1st series of stops, which is voiced plain in Erevan and Tbilisi varieties, is described as “voiced aspirate” (Sievers 1901; Adjarian 1909). Moving from these observations, some scholars questioned the traditional reconstruction of Old Armenian historical phonology, in which is described a complete chain shift of the three main series of Indo-European stops; instead they argue the conservation of the I.-E. “voiced aspirates” in the oldest historical stage of Armenian, and in a few modern dialects (s. the refer- ences in Belardi 2006: 205–16). Along this hypothesis, Armenian would be the only other linguistic group of Indo-European, besides the Indo-Aryan one, preserving the so-called “voiced aspirates”. This paper is based on a field research on three eastern varieties of Arme- nian (Erevan, Aštarak, Gavar), and a western one (Gyumri). The research is based on the acoustic analysis of data coming from the reading of a list of 34 minimal pairs (by a total amount of 27 speakers). We can summarize the fol- lowing results: in all the eastern varieties, voiceless aspirate stops (3rd series) are characterized, in comparison to the plain voiceless ones (2nd series), by a longer Voice Onset Time (VOT). In Aštarak, Gavar and Gyumri dialects, VOT is irrelevant for the distinction between stops belonging to the 1st and 2nd se- ries, since in initial position the formers display a very weak (or no) sonority bar: the relevant difference seems to be the use either of murmured vs. modal voice, or of slack vs. , in the first section of the following vowel (s. Ladefoged & Maddieson 1996: 55–66 for these distinctions and their acoustic cues); therefore the 1st series of consonants can be phonetically described as murmured, or as slack (cf. Pisowicz 1998). These data can support a recon- 46 Giancarlo Schirru

struction of proto-Armenian consonant system which makes use of the fol- lowing two distinctive laryngeal features: [±spread glottis], responsible of the distinction between 2nd and 3rd series, and [±slack vocal folds] distinguishing 1st and 2nd series. A similar reconstruction seems to be consistent with data, concerning the phonetic realization of I.-E. so-called “voiced aspirates”, coming from Indo- Aryan languages (for ex. Hindi; see also the tonogenesis process in Panjabi).

References

Adjarian, H. (1909), Classification des dialectes arméniens, Paris, Champion. Belardi, W. (2006), Elementi di armeno aureo, II. Le origini indoeuropee del sistema fonologico dell’armeno aureo, Roma, Il Calamo. Ladefoged, P. & Maddieson, I. (1996), The sounds of world’s languages, Malden (Ma), Blackwell. Hacopian, N. (2003), “A three-way VOT contrast in final position: Data from Armenian”, Jour‑ nal of the International Phonetic Association 33: 51–80. Pisowicz, A. (1998), “What did Hratchia Adjarian mean by ‘voiced aspirates’ in Armenian dia- lects?”, Annual of Armenian Linguistics 19: 43–55. Sievers, E. (19015), Grundzüge der Phonetik zur Einführung in das Studium der Lautlehre der indogermanischen Sprachen, Leipzig, Bretikopf & Härtel. Vaux, B. (1998), The phonology of Armenian, Oxford, Clarendon Press. 47

Tocharisches Lautgesetz *-sl- > -ll- / -l-

Ilja Seržant Universität zu Köln1, Universitetet i Bergen

Obwohl die Beleglage relativ begrenzt ist, glaube ich zeigen zu können, dass im Urtocharischen eine Assimilation vor-urtoch. *sl > urtoch. *-ll- > TB -ll-, TA ‑l stattgefunden hat. Folgende Beispiele sollen diese Hypothese stützen:

1 TB /ṣəllə́ -/ (z. B. in ṣällatsi 412 b2 (M)) „to throw off smth.“) kann nun unmittelbar auf den reduplizierten idg. Stamm *si-sl̥- zurückgeführt wer- den. Das doppelte -ll- findet damit eine einfache und analogievermeidende Erklärung; 2 ferner im Anlaut (wo natürlich auch mit s-mobile gerechnet werden kann) idg. *sleubh- „gleiten, schlüpfen“ (ae. slūpan) (LIV2: 567) erscheint als toch. TB lup-, laup- / TA lop- (Adams 1999: 558); idg. *sleĝ/g- erschlaffen“ (an. slakr „schlaff“) erscheint als toch. ABläṅk- „hängen“ (aus dem Präsens idg. *sl-n̥ĝ/g-, vgl. lat. langueō „schlaff sein“, gr. λαγγών EM „Feigling“ (LIV2: 565); 3 Innerhalb des Tocharischen wird die Herleitung von TB yälloñ „Sinne“ als aus *yəs-lo- von der Wurzel TB yäs-/yās-2 „erregen, berühren“ ermöglicht.

References

LIV2: Lexicon der Indogermanischen Verben. Die Wurzeln und ihre Primärstammbildungen. Un- ter Leitung von Helmut Rix und der Mitarbeit vieler anderer bearb. von M. Kümmel, Th. Zehnder, R. Lipp, B. Schirmer. Zweite, erweiterte und verbesserte Auflage, bearb. von Mar- tin Kümmel und Helmut Rix. Wiesbaden 2001. Adams, Douglas Q. 1999: A Dictionary of Tocharian B, Amsterdam / Atlanta. Krause, Wolfgang 1952: Westtocharische Grammatik, Heidelberg.

1 Die jetztige Affilation ist die Universitetet i Bergen, diese Arbeit entstand jedoch an der Universität zu Köln unter Leitung und mit Hilfe von Prof. J. L. García Ramón. 2 Bei Krause (1952: 273, 275) fälschlich als zwei Wurzeln eingetragen, wobei yās- nur das PP Prt. I yayāsau bildet, das aber formal und semantisch zum Prt. I yasāte (Š) /yasá-/ (eingetra- gen unter der Wz. yäs-) gehört. 48

Indo-European laryngeals in Anatolian

Vitalij Shevoroshkin University of Michigan, Ann Arbor

The standard 3- assumes that Ht [Hittite]hant- ‘front’, Gr[eek] ἀντί- ‘opposite’ (etc.) originate from IE *h2ent- [*h2e > α], and that, Gr ὄρν-ις ‘bird’ (etc.) originates from IE *h3er-on- (where *h3 is the alleged IE o-coloring laryngeal). But, as stated (among others) by A. Lehrman in his book Indo-Hittite redux (1988, 257), there is no reason to ascribe any vowel-“coloring” power to the un- derlying IE laryngeals. Indeed, those IE languages which preserved IE vowels a and o (like Gr above) show no traces of an underlying IE [e]; they just show a and o no matter if the word contains an underlying laryngeal or not (the same happens in non-IE languages which relate to IE). Ht, Luw[ian], Hierogl. Luw, Palaic show [a] both for IE *a (IE *Hant- > Ht, Luw hant- ‘front’, above) and *o (IE *Hor-on- > Ht haran- ‘eagle’). In Late Anat[olian] (= [HtLuw]) languages of the 4th c. BC, – Lyc[ian] and Mil[yan], – we find x for Ht h- / -hh-, and both x and g for Ht -h(h)- (Cxx- < CVx-): 8 exx.

1 (a) Lyc-Mil xnt- vs Ht hant- (above) // (b) xnna- ‘grandmother’ vs Ht han‑ na- id. // (c) xuga- ‘grandfather’ vs Ht huhha- id., but note lenition in Luw huha- // (d) Lyc xaha- ‘altar’ [Lyc h = Mil s] vs Ht hassa- ‘hearth’ // (e) Lyc xba(i)- ‘irrigate’, Mil xbadiz (pl.) ‘(river) vallies; Lycians’ vs Ht hapa- [haba] ‘river’ // (f) Lyc xawa- ‘sheep’ vs Luw hawi- id. < IE *Howi- id. (not ‘*h3ewi-’) // (g) Lyc stem xdda- ‘to hurry’ vs Luw hu-da- ‘haste, agility’, Ht huta- ‘readiness’, hutak ‘readily, quickly’ < IE *Hew-dho- (:Lyc verb xuwa- ‘follow (closely)’ ?), cf. Lyc stem xudr- ː Luw hudr- (hudar-la- ‘servant’); see next // (h) Mil xu-s-ti- ‘to rush (smth.)’, n. xust(t)i- ‘raid (< rushing)’ (:xu- z-r- ‘very fast’ in xuzr-nta-, xuzruweti-, attr. to [god] Trqqiz / ‘warriors’, or: Ht hu(e)sa- ‘spindle’ < *Hʷeis-). 2 Lyc verb epi-(e)rije- ‘sell’ (?) can not match Ht happar- ‘price’ (cf. rather Lyc PN Xpparã-ma), there is no shift Lyc ZERO < IE ‘*h3-’. On the other hand, Lyc PN Xere-i matches Ht haran- ‘eagle’, IE (‘*h2/*h3’) > HtLuw *h- > Lyc x-. Vitalij Shevoroshkin 49

Recent studies of Mil texts (2 relatively long inscriptions) confirm J. Rasmus- sen’s & D. Schurr’s identification of Lyc-Mil q as mostly representing HtLuw [hʷ] < IE *Hʷ, e.g.: 3 (a) Lyc Trqqas, Mil Trqqiz < *t(a)rHʷents (god Tarhunt) // (b) Mil ute m̃qr- ‘to time / ration [some produce] for a year (ut-e)’ : Ht witt-i mehur ‘time in the year’; Ht mehur- ‘time, period’ (-ehV- vs -ahhV-) < IE *meHʷr or *meiHʷr // (c) Mil qidri- laxadi ‘dash with fight (toward …)’, qidrala ‘for raids/spoils’ : Ht huedar, huitar ‘(wild) animals, game’ (=Luw) // (d) Mil qtti- ‘remove’ : Ht huittiya- ‘draw, drag’ // (e) Mil qla-, qelene- (suff. *-anna-) ‘accumulate, harvest’, qi-qleni-re/i- ‘collection’, qelelija ‘wealth’ or ‘harvest’, q(e)lei (d.-l. sg.) ‘at/to the collection-place’ ː Ht hula- ‘wind, twist’ < IE *Hʷel- id. // (f) Mil qirze- (-z- < * ts or *ty) ‘obligation, share’ : (?) Ht hu(wa)rt- / hurz- ‘curse’ < IE *Hʷert- ‘vow’ // (g) Mil qre- ‘sprinkle’ (during an offering rite) : Hthu(wa)rai- id., etc. 50

PIE ‘me’ and a new Lydian sound law

Zsolt SImon Hungarian Academy of Sciences

Kloekhorst 2004 established that the phonetic value of the Hieroglyphic Lu- wian sign *19 <á> is the (with or without a following short or long a), i.e. [ʔ(ā̆)] (though this is debated, Kloekhorst’s explanation remains by far the most convincing one, cf. Melchert forthcoming, Rasmussen 2007, Simon in preparation). Kloekhorst’s interpretation has important ramifications not only for the history of Luwian, but both for PIE and for the history of other . Since one of these PIE ramifications is the corroboration of the reconstruction *h1mé ‘me’, first the validity of this reconstruction will be discussed where a new morphological analysis will be offered. This provides the basis of the second part of the paper, a discussion of the Lydian reflexes of *h1 in initial position based on the Lydian words amu ‘m e’, dẽt- ‘goods’ and asfã- ‘goods’. It will be suggested that though the *h1 disappears before a vowel it has been vocalized before a consonant. Incidentally, the ad hoc Lydian sound change of *y > d/#_ also will be integrated into this suggested sound law.

Selected references

Kloekhorst, Alwin (2004): “The preservation of h* 1 in Hieroglyphic Luwian. Two separate a- Signs”. HS 117: 26–49. Melchert, H. Craig (forthcoming): “The spelling of initial /a-/ in Hieroglyphic Luvian”. In: Fest‑ schrift for N. N. Rasmussen, Jens E. (2007): “A reflex of H* 1 in Hieroglyphic Luvian?” In: Karlene Joney-Bley et al. (eds.): Proceedings of the 18th Annual UCLA Indo-European Conference, Los Angeles, November 3–4, 2006. Washington, 161–167. Simon, Zsolt (forthcoming): “Once again on the Hieroglyphic Luwian sign *19 <á>”. 51

Some consequences of Kartvelian realisations of Indo-European borrowings for the reconstruction of the Proto- Indo-European phonological system

Thomas Smitherman University of Bergen

Indo-Europeanisms in the are occasionally the field for serious etymological studies (Klimov 1994, Furnée 1979, Gamqrelidze/Ivanov 1984). A number of suspected IE loanwords into the Proto-Kartvelian as well as Proto-Georgian-Zan layers of vocabulary have already been established. Some of these bear certain features of antiquity, not resembling their direct Greek, Armenian, Anatolian, or Iranian neighbours, while others appear to stem from an ancient layer of one or more such branches. Comparisons are relatively eas- ily made due to the large number of systemic phonological features shared by PIE and PK. Using a searchable database collected from the latest edition of Das ety‑ mologische Wörterbuch der kartwelischen Sprachen (Fähnrich 2006) and the The Oxford Introduction to Proto-Indo-European and the Proto-Indo-European World (Mallory & Adams 2006), useful for its division of the lexicon by se- mantic fields, some new possible borrowings can be proposed by direct lexical comparison. Examining the list of likely correspondences, one can compare the frequency of phonetic features of PIE initial and final consonants with their Kartvelian counterparts. The compiled data show that the PIE voiced aspirates have by far the great- est tendency to be rendered as voiced stops in Kartvelian. Unsurprisingly, the PIE unvoiced stops can rarely be said to lend themselves to Kartvelian voiced stops. However, perhaps more surprising is that PIE unvoiced stops are most rarely lax (aspirated or fully fricative) in Kartvelian, the exception to this being the labials where PIE *p usually corresponds with PK *ph, due to the relative infrequency of PIE *b and PK *p’. Indo-European borrowings are assimilated into the Kartvelian phonologi- cal systems with certain marked tendencies. The general tendency is for tradi- 52 Thomas Smitherman

tional PIE voiced stops to be tense and voiced in syllable-final position, while unvoiced PIE stops are more likely to be tense and voiced in initial position. Additionally, the great majority of dentals and labials are rendered as voiced, whereas velars are often voiced only in final position (with a common correla- tion of initial PIE *gh- ~ PK *ɣ). Therefore, roots of the type *beg- or *kherth-, with consonantal harmony in regard to the manner of articulation, are com- mon in the data. The great majority of roots of the type CeC-, where C is a plo- sive, show this harmony. The most common non-harmonic syllable structure in both PIE and PK is *keg-. This speaks against the , perhaps ironically, given Gamqrelidze’s native knowledge of the Kartvelian lexicon and the close proximity to the homeland of a taihun language (Armenian). Howev- er, the distribution of aspirated and non-aspirated plosives may serve as some evidence for allophony in the stage of Late PIE or early stages of the formation of the IE branches.

Some Examples of Relatively Likely Shared Vocabulary: PIE PK Gamq/Ivanov *kerp- *k’reph- Gather (note, not *khreph-) *g’e/on-h3- *ge/on- Hear/under- (preserved ablaut) stand *legh- *lag-, *rg- Set, plant *dheg’h-o:m- *diqha- Earth, clay *tep- *t’eph- warm (note, not *theph-) *bherg’h- *b(e)rg- old Klimov *gʷebh-u- *j1webu- frog, toad *gʷor- *gor- hill (must relate to satem languages) *g(h)rund(h)- *ɣrunt’- grunt *ghergh- *ɣerɣ- hammer grk κερχ- arm gergzh- *g’hor- *ɣor- pig ie dialectal, but note alb derr- *skendh- *c’k’e(n)d- plant, divide pie *sk- ~ pk *c’k’-, *c’q’- common Current *gʷes- *k’wes- extinguish (Mallory/Adams for pie form) *kʷel/r- *gwel/r- twist/circle *gel- *j1al- ability, power arm reflex, palatisation in pk? *ghel-u(H)-s *ɣl-az1- amphibian stem-extension unusual for pk *kle(n)g- *k’lak’- bend Thomas Smitherman 53

*kon-to- *k’unt’- crooked, bent kart syllabic structure unusual *g’en-h1- *c1hen- become, grow arm cn-aw but zan chen- *g’hwer- *zwer- beast, steer kart close to Slavic zver *rek- *rek’-, *rex- call, voice *reg’- *rg-, *rJ- rule, order, direct *(s)kel- *khel- bend *reup- *roph- smash, crush *dhg’h-(i)es- *deɣ-, *deg- day, burn PGmc *daɣ- gk χϑ-ες *prk’- *pharchx- harvest secondary palatisation in kart *meg’h(hx)- *mec1h- old Laz, Svan mech’ = not late Arm loan

References

Fähnrich, Heinz. Etymologisches Wörterbuch der kartwelischen Sprachen, 2nd ed. Amsterdam: Brill, 1994. Furnée, E.J. Studium zum ostmediterranen Subtrat nebst einem Versuch zu einer neuen pelasgi‑ schen Theorie. Louvain: 1979. Gamqrelidze, T. & Ivanov, V. V. Indoevropejskij jazyk i indoevropejcy. Tbilisi, 1984. Klimov, G. A. Drevnejšie indoevropeizmy kartvel’skix jazykov. Moscow: Nauka, 1994. Lexikon der indogermanischen Verbum, 2nd ed. Wiesbaden, 2001. Mallory, J. P. & Adams, D. Q. The Oxford Introduction to Proto-Indo-European and the Proto- Indo-European world. 54

Proto-Indo-European *kakós ‘bad’ – another good example of unconditioned /a/

Dieter Steinbauer

The first part of my paper discusses some problems of establishing the inven- tory of vowel phonemes in the proto-Indo-European phonemic system. The short vowels were a, e, i, o, u + shwa. The latter had no partner among the five long vowels ā, ē, ī, ō, ū. Shwa was the anaptyctic vowel that developed next to consonants under various conditions. The so-called coloring of pre-proto-Indo-European e/ / by adjacent laryn- geal should be seen as a persistent rule. Morphonological rules (reduplication of syllable onset; vriddhi: insertion of /e/ after first root consonant) can prove this, e.g. *vwo/ey- ‘bird’, *vowyom ‘egg’. The three laryngeals are seen as fricatives/spirants: h1 = /h/, h2 = /x/, h3 = /v/. There are some well-known rules that free /a, ā/ from the conditioning la- ryngeal. Accusative singular *-axm > PIE *-ām, cp. *dyewm > PIE *diēm ‘day’. Vocative singular *-ax > PIE *-a, cp. *dyew > PIE *dye; OI adyá ‘today’. Ablaut patterns with primary /a/ and /ā/ also exist (“long” *mak’rós : mā́k’os), but still some words will ever (?) be “beech”-argument-lexemes, i.e., no deci- sion can be made, if we have to part from a long /ā/ or vowel plus laryngeal. Relations between vowels of different quantity may be illustrated by ‘you = thou’ and ‘now’. Beside *tu, *tū and *nu, nū we find Latin (emphatic) tute /tū́te/ < *tūtu. Gothic nunu, Palaeo-Sicilian nunu (without hint to quantity). These lexemes are not mere iterations, but the combinations of the long and short form. The usually reconstructed *nūn must have lost its second vowel by elision (?). The phonological system of proto-Indo-European is reconstructed by gath- ering cognate sets, i.e. by comparing and evaluating etymologies. Many cases are crucial not allowing straightforward reconstructions. In the second part of my paper, a host of superficially unrelated problems must be tackled: validity of regular sound change, even with lexemes originat- ing from sound symbolism, or acceptance of obvious irregularities (violation Dieter Steinbauer 55

of root-structure, complete dissimilation), subgrouping, loan words within languages of the same family, etc. The proto-I.-E. word for ‘bad’ was kakós* . The cognate set:

• Greek (up to now!) kakós ‘bad, evil, ugly’, Phrygian kako- ‘evil’, Albanian keq (< *kakio-) ‘bad, evil’; OI ákam ‘pain’, Avestan aka- ‘bad, evil’ (<*kaka-, loss of k- by dissimilation!)

Until now among recognized etymologies for ‘bad’ only the prefix dus-* could be found. The adjective kakós* should be added! 56

The sociophonemics of Late Gaulish

David Stifter Universität Wien

In order to fully appreciate the diachronic developments of linguistic systems, for example the restructuring of the sound systems of languages, it is not suf- ficient to have a command of the linguistic facts only, but it is also important to have an understanding of the social and historical contexts in which those developments take place. When the contexts are known, other factors than abstract language-internal tendencies may gain explanational power, like, for example, sociolinguistic variation or contact influence. Under this hypothesis, I will look at documents of the Late Gaulish language, in particular Endli- cher’s Glossary and the inscription on the Châteaubleau tile, to assess how the diachronic-linguistic phenomena identifiable in those texts, especially in regard to phonology, can be integrated with what can be concluded from the scanty sources about the extralinguistic fate of Gaulish in the Roman provin- cial period. 57

Greek evidence for the “Saussure Effect” reconsidered

Lucien van Beek Leiden University

In this paper, I will reconsider the Greek evidence for the so-called Saussure Effect: the phenomenon in some of the daughter languages that a Proto-Indo- European laryngeal seems to be lost in the vicinity of an o-grade. The Greek evidence comprises forms of the following two types:

A Root-initial laryngeal, *HRoC(C)- (R any resonant or semivowel): e.g. ὀμείχω ‘urinate’ (Hes.) / μοιχός ‘adulterer’ (Ion.-Att.), B Root-final laryngeal before consonant, *(C)CoRH-C-: e.g. τελαμών ‘carry- ing strap’ (Hom.) / τόλμα ‘courage’ (Ion.-Att.).

The communis opinio seems to be that the laryngeal was lost in the environ- ments under (A) and (B) either in PIE or in a late dialectal stage of it (see the bibliography). However, if we accept the Saussure Effect as a regular sound change, the phonetics behind it are not easily understood. Moreover, it seems that evidence for it in Balto-Slavic, Latin, and Anatolian is slim. Therefore, my set-up will be to try how far we can get if we assume that the Saussure Effect did not operate in Greek in the way sketched above. The ques- tion in each individual case is whether a good alternative explanation can be provided.

Selected bibliography

H. C. Melchert, Anatolian Historical Phonology, Amsterdam/Atlanta: Rodopi (1994) 49ff. A. Nussbaum, The “Saussure Effect” in Latin and Italic. In: A. Lubotsky (ed.): Sound Law and Analogy (FS Beekes), Amsterdam/Atlanta: Rodopi (1997) 181ff. P. Schrijver, The development of the PIE laryngeals in Latin, Amsterdam/Atlanta: Rodopi (1991) 326ff. 58

PIE mobile accent in Italic: further evidence

Brent Vine University of California, Los Angeles

It is often assumed that PIE mobile accent was lost without a trace in Italic, hav- ing been converted to an initial-stress accent, witnessed by the Common Italic “period of initial stress” (inferable from Sabellic and Latin syncope, Middle/ Late Etruscan syncope, Latin vowel reduction, and Oscan/Latin “geminatio vocalium”). In recent years, however, several theories have proposed the main- tenance of limited forms of early Italic mobile accent, in some cases traceable to PIE accentual alternations. Thus Lat. palma ‘palm’ is said (e.g. Meiser 1998: 108f.) to display the accented treatment of CŔ̥HC (> Ital. *CaraC, i.e. *pĺ̥h2- meh2 > *palamā > Lat. palma), vs. unaccented CR̥HC (> Ital.*CRāC, e.g. *pl̥h2- nó- > Lat. plānus ‘flat’); cf. the analogous Greek pattern in παλάμη (again < **pĺ̥h2-meh2), vs. unaccented CR̥C > *CRāC (Rix 1992: 73). For a second type: according to a proposal of Rix (1996: 158n7, cf. Meiser 1998: 74), PIE unaccent- ed final vowels were subject to regular apocope in Latin (e.g. PIE *éti [: Ved. áti, Gk. ἔτι] > Lat. et), whereas accented final vowels were retained (e.g. loc. sg. *ped-í [cf. Gk. dat. ποδί] > Lat. pede). Both of these proposals, while intriguing and attractive in some respects, remain controversial. More recently, I have suggested that “Thurneysen-Havet’s Law”, which ac- counts for the a-vocalism of forms like Lat. cavēre ‘beware of’ (: κοέω ‘per- ceive’) < *kouh1‑éi̯e/o-, might also exhibit a trace of PIE mobile accent in early Italic. The extreme age of the process was demonstrated by Schrijver (1991: 436ff.), a conception that has been embraced more widely (cf. Meiser 1998: 85, Martzloff 2006: 125ff., de Vaan 2008: 8). The chronology can be further sup- ported by the possibility that Umbrian sauitu ‘she shall wound’ (cf. Lat. sau‑ cius ‘wounded’) provides a Sabellic example of the process (Vine 2004: 622ff.). Since forms with unaccented */ou̯/ (in PIE terms) undergo the change to /au̯/ (cf. *kouh1‑éi̯e/o- > Ital. *kau̯ē-) while forms with accented */óu̯/ (in PIE terms) do not (e.g. *h2óu̯-i- ‘sheep’ [acrostatic i‑stem] > Ital. *ou̯i-), I have proposed (2006) that Thurneysen-Havet’s Law was conditioned by what remained of PIE mobile accent at a very early phase of Italic. Thus, some instances may reflect still other PIE alternations, such asτομ ός vs. τόμος distinctions (e.g. Lat. cavus ‘hollow’ < *k̂ou̯H-ó- *‘swelling’ vs. Ibero-Romance *kou̯o- [Port. côvo Brent Vine 59

‘hollow, deep’, Span. cueva ‘cave’, etc.] < *k̂óu̯H-o-/-eh2, cf. κόοι· κοιλώματα, τὰ χάσματα τῆς γῆς Hsch.). The present paper explores two problematic morphological categories in Italic – acrostatic iterative-causatives (as in the notorious relic form Lat. sōpīre ‘put to sleep’) and o‑stem denominatives (with poorly-explained “4th-conjuga- tion” inflection: Lat.servus ‘slave’ → servīre ‘be a slave’, etc.) – and suggests that both can be explained on the basis of a phonological rule that was sensitive to the position of the accent, in PIE terms. The discussion of Lat. sōpīre begins with a critical reevaluation of the standard theory of acrostatic iterative-causa- tives (Klingenschmitt 1978, LIV), in particular the construct of a “zero-grade” version *-i̯e/o- of iterative-causative *-éi̯e/o-. A reconsideration of o‑stem de- nominatives points up the weaknesses of the prevailing view that these are based on i-stem pendants of thematic formations. The morphological make- up and phonological treatment of these categories, in the terms suggested here, may provide evidence that in addition to the accentual distinctions noted above, early Italic may also have distinguished among accentually contrasting patterns of verbal derivation, such as X́-ei̯e/o- (acrostatic iterative-causative) ~ X-éi̯e/o- (standard iterative-causative) ~ X-e-i̯é/ó- (o-stem denominative). 60

Aspects of Euphratic phonology

Gordon Whittaker Georg-August-Universität Göttingen

It has been argued in a series of recent publications that a apparently unmotivated sign values in the cuneiform writing system, b place names in Southern Mesopotamia long thought to be of substrate ori- gin, and c unsegmentable polysyllabic terminology in Sumerian may indicate prior contact with an Indo-European language.

Many of the terms in question exhibit sufficient regularity in their correspon- dences to permit tentative conclusions to be drawn as to the underlying pho- nology of the language in question, which has been dubbed Euphratic. The presentation will focus on several areas of phonological interest and discuss the potential and limitations of work on loanwords from a language for which no cuneiform texts are as yet accessible. 61

Types of ablaut class assignment in non-primary nominal derivation

Paul Widmer Philipps-University Marburg

Internally derived secondary nominal stems are usually assigned the ablaut class that follows the ablaut class of the derivational basis according to the hierarchically arranged derivational chain (acrostatic > proterokinetic > hys- terokinetic > holokinetic). Alternatively, the derived nominal stem may as well be assigned a holokinetic ablaut pattern regardless of the ablaut class of the derivational basis, a fact which is quite intriguing. It will be suggested that this assignment rule is closely interrelated with phenomena that occur in non- primary nominal derivation from case forms. 62

Phonetic interpretation of Hurrian sibilants in the light of Indo-European evidence

Ilya Yakubovich University of Chicago

The Hurrian language has three “sibilant” phonemes that are recorded with the help of the , <ša>, and series in the Mitanni Letter and conven- tionally labeled /S/, /Š/, and /TS/. In my presentation, I intend to argue that /Š/ of the traditional transcription should be rather interpreted as the alveolar fricative /s/ with two allophones [s] and [z]. The other two phonemes should be reconstructed as affricates /ts₁/ and /ts₂/, although their precise phonetic interpretation remains uncertain. The prima facie argument for the alveolar interpretation of /Š/ comes from Hurrian texts recorded in Hittite and Ugaritic orthographies. Hurrian /Š/ re- ceives the same graphic representation as Hitt. /Š/, which represents the regu- lar reflex of Indo-European s* . The nontrivial similarity between the rendition of Hitt. /Š/ as in the Ugaritic alphabet and the transcription of Hurr. /Š/ as ~ in the same system of writing further underscores the phonetic affinity between the Hittite and Hurrian sounds. The grapheme is most frequently used for rendering the reflexes of Proto-Semitic θ * ([þ]) in the inherited stratum of the Ugaritic lexicon. Yet, the assumption that the only Hittite sibilant was [þ] would typologically unlikely, while the change *θ > s can be accounted for as part of the chain shift in the history of Ugaritic and, in fact, increases the phonetic naturalness of the Ugaritic system of sibilants. Therefore, the alveolar fricative [s] emerges as the most economical phonetic interpretation for Hittite /Š/, Ugaritic , and the voiceless allophone of Hur- rian /Š/.

The voiceless allophones of the Hurrian affricates /ts1/ and /ts2/ were appar- ently rendered with the and <ṣ> signs in Ugaritic orthography, while was reserved for their voiceless allophone. The affricate articulation of Ugaritic and is accepted in modern Semitological research. In Hittite orthog- raphy, both Hurrian affricates were rendered with the series, which was also used for the Hittite affricate . The fact that Akk. ṣalmu- ‘statue’ was borrowed into Mitanni Hurrian as salamši further corroborates the hypothesis Ilya Yakubovich 63

that the Hurrian phoneme conventionally reconstructed as /S/ had an affricate pronunciation. An additional argument for the new reconstruction comes from the com- parison between Hurrian and Urartian. The close genetic relationship between these two languages prompts one to assume that their phonemes linked by regular correspondences were also phonetically similar, unless compelling evidence to the contrary can be presented. Numerous reliable cognates bear witness to the correspondence Hurr. /Š/ ~ Urart. /Š/: Hurr. šue- ~ Urart. šui-ni ‘all’, Hurr. šauri- ~ Urart. šuri- ‘weapon’, Hurr. taršuwani- ~ Urart. taršuwani- ‘human being’ etc. The Urartian borrowings into Armenian indicate, however, that the phonetic value of Urart. /Š/ was [s] rather than [ʃ]: Urart. šuri- ‘weap- on’ > Arm. sur ‘sword’, Urart. šani- ‘vessel’ ~ Arm. san ‘cauldron’ etc. Given that the Urartians borrowed their cuneiform writing from the Neo-Assyrian Empire, this conclusion is consistent with the likely pronunciation of /Š/ as [s] in Neo-Assyrian, which has been established on independent grounds. 64

A new environment for laryngeal loss in Proto-Celtic

Nicholas Zair Oxford University

It has long been assumed that the regular reflex of a laryngeal between two con- sonants in Proto-Celtic was *a (e.g. Schumacher 2004: 135–136). The purpose of this paper is to argue that in one environment the laryngeal was lost without reflex, on the basis of forms such as OIr. fo·ceird “throws” < *kerH-de/o-, ferc “anger” < *u̯erHg-, MIr. ferb “blister” < *u̯erH-bhā. Comparison of these forms with e.g. MIr. arathar “plough” < *H2erH3-tro-, OIr. talam “earth, ground” < *telH2-mō suggests a rule whereby laryngeals were lost in *CHT clusters but vocalised in *CHCC and *CHR clusters. The rule applied only to clusters in non-initial syllables (cf. MW. had “seeds” < *sH1-tV-). Different results in ini- tial and non-initial CHC clusters are of course well attested in other Indo- European languages, e.g. in Proto-Germanic (Ringe 2006: 79–80, 137–139). Recognition of this rule allows us to address several long-standing prob- lems of Proto-Celtic historical phonology. For example, OIr. do·cer “fell” re- flects kerH* 2-t but *kerat ought to have given *karat in Proto-Celtic by Joseph’s law (*eRa > *aRa; Joseph 1982: 55). According to the new rule, *kerH2-t > *kert > *ker > do·cer by regular sound change (for the loss of *-t cf. OIr. ·bé “may be” < *bu̯et). The same development also sheds light on the development of VIHC* clusters in Celtic: *H3rei̯H-trā > *rei̯atrā > W. rhaeadr “torrent”, *keu̯H2-ro- > *kauaro- > MW. cawr “giant”, Gaul. Καυαρος, but *kreu̯H-di- > *kreu̯di- > MIr. crúaid “hard, harsh”. The implications of this Proto-Celtic rule for a suggested Proto-Indo- European rule *CH.CC > *C.CC in non-initial syllables (Hackstein 2004) will be discussed.

References

Hackstein, Olav (2002). “Uridg. *CH.CC > *C.CC”, Historische Sprachforschung 115, 1–22 Joseph, Lionel S. (1982). “The treatment of *CRH- and the origin of CaRa- in Celtic”, Ériu 32: 31–57 Schumacher, Stefan (2004). Die keltischen Primärverben. Ein vergleichendes, etymologisches und morphologisches Lexikon. Innsbruck. Ringe, Don (2006). From Proto-Indo-European to Proto-Germanic. Oxford.