How to Find the Origins of a Dragon?

HOW TO FIND THE ORIGINS OF A DRAGON? A COGNITIVE LINGUISTIC APPROACH TOWARDS THE PROTOHISTORY OF THE GREEK HEXAMETER: ARCHAISMS AND INNOVATIONS IN THE COLOMETRY OF HOMERIC VERSE

Niels Schoubben Student number: 01400028

Supervisor: Prof. dr. Mark Janse Co-supervisor: Dr. Filip De Decker

Master thesis submitted in fulfilment of the requirements for the degree of Master of Arts in Linguistics and Literature: Main subject Latin and Greek.

Academical year: 2017-2018.

सह नाववतु सह नौ भनु क्तु सह वीयं करवावहै । तेजस्वव नावधीतमवतु मा स्वस्िषावहै ॥ ओ ंशास््तिः शास््तिः शस््तिः!

May the Lord protect us both, may the Lord nourish us both, may we accomplish an heroic deed, may our study be brilliant, may we not be hateful to each other! Om, peace, peace, peace!

(Opening invocation of Kaṭhopaniṣad, referring to a teacher and a student)

CONTENT PREFACE: How to express your gratitude? ...... 6 List of metrical symbols ...... 8 Samenvatting in het Nederlands ...... 9 INTRODUCTION: The search for a dragon ...... 12 CHAPTER 1: THE PROTOHISTORY OF GREEK EPIC ...... 17 1.1 Telling a story about storytellers: The oral preservation of an Indo-European past...... 17 1.2 The language of the Homeric poems: A synchronic approach...... 25 1.3 The language of the Homeric poems: A diachronic approach...... 31 CHAPTER 2: A COGNITIVE APPROACH TO GREEK METRE ...... 37 2.1 Some preliminary facts about Greek epic metre ...... 37 2.2 How to put a caesura? - The traditional theories ...... 42 2.3 How to put a caesura? - Cognitive problems ...... 49 2.4 How to put a caesura? - Cognitive solutions ...... 57 CHAPTER 3: THE PROTOHISTORY OF HOMERIC METRE ...... 63 3.1 A critical survey of previous attempts ...... 63 3.1.1 Preliminary remarks ...... 63 3.1.2 Contact metrics: Borrowing a metre? ...... 65 3.1.3 Splitting up the Ionic hexameter: the main theories ...... 67 3.1.4 The extension of a metre: Nagy's proposal ...... 75 3.1.5 Splitting up the Ionic hexameter: the minor theories ...... 77 3.1.6 Anaclastic hexameters: Kiparsky's vision ...... 79 3.1.7 The theory that received a damnatio memoriae: Kurt Witte ...... 82 3.2 A cognitive attempt towards the protohistory of Greek hexameter ...... 84 3.2.1 Methodological remarks ...... 84 3.2.2 The Homeric diaeresis ...... 85 3.2.3 A neglected caesura of Homeric verse ...... 90 3.2.4 Archaisms and innovations in Homeric caesurae ...... 95 3.2.5 The relevance of unmetrical verses ...... 99 CONCLUSION: Did we find the dragon? ...... 103 MAIN REFERENCES ...... 105 APPENDIX: Syntactic statistics for the bucolic diaeresis (Iliad I, XI, XVI)...... 115 1. Iliad I ...... 115 2. Iliad XI ...... 116 3. Iliad XVI ...... 117 4. Concluding table...... 119

Word count: 38.651 words, including the main text with citations, but without the ackowledgements, the list of symbols, the Dutch summary [samenvatting in het Nederlands], the footnotes, the bibliographical list and the appendix.

PREFACE: How to express your gratitude? At first sight, writing the preface of your MA thesis can seem the most easy part of the job. You do not have to read sources for it, you do not have to make clear references, you do not have to write in an academic way etc. However, this is a simplification of the situation. How can you adequately express your gratitude towards the many people who assisted, if only while standing on the sidelines, in bringing this massive project to favourable conclusions? First of all, I am grateful to my supervisor, prof. dr. Mark Janse. Since our first conversation, during a break in first year's course "European languages in contact", our paths never separated again. Due to my interest in Greek linguistics in general, and historical linguistics in particular, writing my BA paper and MA thesis under his supervision was in fact a logical choice. The only problem was to find the definitive subject. Last year, the quest resulted in a fascinating journey to the Hellenistic world and the contacts between Greeks and Indians. This year, the search resulted in an adventurous odyssey, sailing across Venetian Crete and its Italian influence, until Aeolus' sack with the different winds was opened and I landed in a pittoresque village in 19th century Cappadocia, Sinasos, but finally I reached my Ithaca: Homeric and Indo-European linguistics. Reason one to thank prof. Janse, for his patience and neverlasting support during this search for the right subject. When this was finally found, he encouraged me to do this as best as possible, tirelessly proposing new sources while finding my path through the immense literature about Homer and his language, patiently correcting my English etc. Our regular meetings were always equally enjoying and inspiring, always starting with a polite refusal of the offered cup of coffee ("No thank you, I do not drink coffee"). Secondly, I had the good luck to be supported by a second supervisor, dr. Filip De Decker. In fact, he was the positive outcome of some administrative problems. Only some two months before the deadline, he became officially my co-supervisor. However, during this short period he did as much as one would expect from a full-year one. In no time, I received the comments for each part of the dissertation I sent to him, comments which showed his profound interest in the topic, his extensive knowledge of Homeric linguistics and his irresistible urge to help me with it. Our meetings in the coffee room (it pursues me apparently…) were also a pleasure for me. Thank you for this support and also for the lift to Ghent after the conference in Katwijk! Thirdly, I would like to thank the people who, as the cliché declares, made my student life one of the best times of my life: my friends from the "Klassieke Kring". From the first day onwards, this student circle felt like home, one year later I started a three-year career in the praesidium, extending my social skills to such an extent, as I never dared to dream when I first entered the Mons Blandinus. Thank you to all the amazing people I encountered here, and especially to Inge, my "Limburg comrade", to Alexander, for his talent to be a good friend, to Roos, my "linguistic partner in crime", to Gaëtan, my "Homeric partner in crime", to Siebe & Tijl, who were there from the first day, to Shauni, the always equally charming

"Mother Superior", to Leanne, "my Dutch darling", to Sophie, or would you like "Vergilia"?, to Justine, "the granny driving with her Orvie", to Emma, the "glitter queen of the KK", to Jasper, the "stoic part of the furniture of the Blandijn" and to all the people, whose name I forgot to mention here. Indeed, writing a MA thesis is the end of a longer process, one which starts when you first enter university. Not only my fellow students were important in making this a great experience, but also my teachers, I want to thank them all for their rousing courses. To name only a few in particular, I would like to thank prof. dr. Marc De Groote, for his philological ἀκρίβεια (with the correct accent), prof. dr. Kristoffel Demoen, for his thought provoking questions, prof. dr. Eva De Clercq, for learning me my first words in Sanskrit, prof. dr. Giovanbattista Galdi, for his neverlasting smile when discussing Latin linguistics, prof. dr. Christophe Vielle, for our discussions about the meaning of the text, prof. dr. Gunnar De Boel, for his encyclopedic knowledge, dr. Leonid Kulikov, for being a living Vedic grammar and concordance and again prof. dr. Mark Janse, for his interesting digressions and funny word jokes. Thanks to all, you were inspiring professors! Remains one person at university who deserves to be praised in this preface: Els De Loor, for helping me to find the books which were hosted at the office of some researchers, for not applying the strict rules of the central library and certainly for the casual talks when I was at the library. Thank you! Last but not least, I would like to express my gratitude to my mum, Marina Schoubben and the rest of the family. Without them, I would not have been able to come to Ghent and study these beautiful, dead languages.

I cannot aptly express my gratitude to all these people, οὐδ᾽ εἴ μοι δϋκα μὲν γλῶςςαι, δϋκα δὲ ςτϐματ᾽ εἶεν, φωνὴ δ᾽ ἄρρηκτοσ, χϊλκεον δϋ μοι ἦτορ ἐνεύη1.

1 "Not even if I had ten tongues, not even if I had ten mouths, an unbreakable voice, if I had a bronze heart". (Il. II, 489-490).

List of metrical symbols

– long ∪ short ⨯ syllaba anceps (long or short) ∪∪ two shorts can be replaced with a long one (biceps procedure) ○○ two positions of which at least one must be long. | demarcation line between two feet || general sign for a caesura / main caesura (only in some theories) ⋮ supplementary caesura (only in some theories) # end of the metre

Samenvatting in het Nederlands

Deze thesis heeft betrekking op de voorgeschiedenis van de Homerische hexameter. Vanaf de negentiende eeuw hebben wetenschappers zich de vraag gesteld waar dit lange epische vers in het Grieks vandaan kwam. Waar men in de negentiende eeuw het vers probeerde terug te voeren tot een combinatie van kleine Aeolische maten of zelfs tot een Indogermaans Urvers wou herleiden, werd dit in de jaren '20 van de twintigste eeuw vervangen door een visie waarbij de hexameter gezien werd als een ontlening aan een pre- Griekse bevolking (Meister 1921; Meillet 1923). Deze visie bleef dominant tot in de jaren '70, wanneer verschillende nieuwe hypotheses het licht zagen, die opnieuw probeerden om de hexameter terug te voeren op Indo-Europese modellen, die in Aeolische verzen bleven voortbestaan. De belangrijkste hiervan waren West (1973a) die de hexameter herleidde de combinatie van een hemiepes (–∪∪–∪∪–) en een paroemiacus (∪̅∪–∪∪–∪∪–⨯), Nagy (1974) die sprak over een dactylische uitbreiding van een pherecrateus (○○–∪∪––) en Berg (1978) die de hexameter zag als een samenvoeging van een glyconeus (○○–∪∪–∪–) en een pherecrateus (○○–∪∪––). Elk voorstel stuit echter op onvolkomenheden. De vernieuwde analyse van vocalische /ṛ/ in Homerisch Grieks door van Beek (2013) biedt een sterk tegenargument tegen één van de fundamentele vertrekpunten van Bergs hypothese (cf. Tichy 1981). Daarnaast verklaart deze hypothese de karakteristieke cadens na de bucolische diërese als een innovatie, hetgeen onwaarschijnlijk is, gezien het veelvuldig voorkomen hiervan (Miller 2014: 86). Daarom wil deze dissertatie een ander voorstel doen in verband met de historische conditionering van de Homerische colometrie. Daarbij zal gebruik gemaakt worden van de hypothese van Kurt Witte (1913), die vertrekt van het grote belang dat de bucolische diërese heeft in de verssegmentatie van het Homerisch vers. Dit leidde ertoe dat hij de hexameter zag als de combinatie van een dactylische tetrameter en een dimeter (de adoneus –∪∪–⨯). Om deze hypothese te verifiëren maakt deze thesis gebruik van een corpus van drie zangen uit de Ilias (I, XI en XVI) die door middel van een cognitief model bestudeerd zullen worden. Deze cognitieve benadering gaat lijnrecht in tegen de bestaande metrische theorieën, die te weinig rekening houden met syntactische en inhoudelijke elementen binnen de Homerische colometrie, hetgeen onwaarschijnlijk is gezien de oral performance waaruit het Homerisch epos ontstaan is. In deze scriptie zal de verssegmentatie van Janse (1998; 2012) gebruikt worden, die zelf gebaseerd is op het taalkundig concept van intonation units door Chafe (e.g. 1982; 1985; 1987; 1994) en dat reeds door Bakker (e.g. 1990a; 1990b; 1997a; 1997b) op een succesvolle wijze is toegepast op het Homerische taaleigen. Deze cognitieve benadering maakt het mogelijk om het belang van de bucolische diërese verder te beklemtonen en als dusdanig de hypothese van Witte aan overtuigingskracht te doen toenemen. Hoe uit zich deze methodologie binnen de structuur van deze masterproef? In een eerste hoofdstuk geven we een algemene inleiding op het probleem van de voorgeschiedenis

van de Homerische epiek. Op basis van een uitgebreid literatuuronderzoek zal aangetoond worden dat door middel van een orale traditie versificatorische en poëtische elementen bewaard zijn gebleven in archaïsche Griekse poëzie die terug te voeren zijn op Indo- Europese prototypes. De belangrijkste ontwikkelingen in het onderzoek naar deze orale traditie zullen daarbij besproken worden. Deze poëtische kijk wordt in een tweede deel van het eerste hoofdstuk aangevuld met een linguïstische optiek. Er zal een beknopt overzicht gegeven worden van de Homerische Kunstsprache en in een laatste onderdeel zullen de verschillende historische verklaringen hiervoor gepresenteerd en geëvalueerd worden. Het tweede hoofdstuk biedt het methodologische kader van deze thesis en focust zich uitsluitend op het probleem van de Homerische metriek. Na een kort overzicht van de belangrijke prosodische en metrische noties die noodzakelijk zijn voor een goed begrip van de Homerische versificatie, wordt een overzicht gegeven van de traditionele opvattingen in verband met de colometrie van het epische vers. Daarbij zal de noodzakelijke kritiek gegeven worden, die vooral focust op de onwaarschijnlijkheid van een te rigide opsplitsing van verzen binnen een orale context. Als alternatief zal dan het reeds aangehaalde systeem van Janse voorgesteld worden, dat een meer dynamische visie op het Homerische vers toelaat. Een aantal verzen uit het geselecteerde corpus zal geanalyseerd worden volgens de verschillende mogelijke theorieën waarbij duidelijk moet worden dat het systeem van Janse te verkiezen valt. Het derde en laatste hoofdstuk focust uitsluitend op het probleem van de voorgeschiedenis van de Homerische hexameter. Eerst wordt er een overzicht geboden van de belangrijkste hypotheses die in de voorbije honderd jaar de revue zijn gepasseerd. Zij worden gegroepeerd naargelang het algemene conceptuele principe dat zij gebruiken. Elke hypothese wordt summier voorgesteld waarna een overzicht gegeven wordt van de problemen die aan elke hypothese verbonden zijn. Als laatste wordt de hypothese van Witte voorgesteld die de basis zal vormen voor de verdere discussie van de oorsprong van de hexameter. Op basis van de cognitieve analyse van het corpus en literatuuronderzoek van studies na Witte (1913), zal diens voorstel verder uitgewerkt worden. Daarbij zal het belang van de bucolische diërese beargumenteerd worden, waarbij er ook aandacht zal gaan naar het veelvuldig samen voorkomen van deze diërese met een cesuur vroeg in het vers. Dusdanig zal een model gepresenteerd worden dat een mogelijke verklaring biedt voor een versmelting van een oorspronkelijke tetrameter met een adoneus. Daarbij zal gebruik gemaakt worden van het principe van flexibele formules, zoals dat onder andere door Hainsworth (1968) ontwikkeld is. Op basis van verschillende voorbeelden zal beargumenteerd worden dat niet alleen de hephthemimerische cesuur, zoals Witte zelf al aangetoond heeft, maar (minstens ten dele) ook de middencesuren (de penthemimerische en de trochaïsche) innovaties zijn. Een verklaring wordt gezocht in de uitbreiding van formules die oorspronkelijk bij de bucolische diërese begonnen. Dit proces was gemotiveerd door het zoeken van de aoidoi naar een cesuur in de buurt van het midden van de nieuw ontstane hexameter. De trihemimerische cesuur zal daarbij ook als een archaïsme, met name als een oorspronkelijke cesuur van de tetrameter benaderd worden. Het laatste onderdeel

van hoofdstuk drie behandelt kort het probleem van onmetrische verzen in de Homerische metriek en de vraag in hoeverre deze gebruikt kunnen worden in analyses over de oorsprong van de Homerische hexameter. Het voordeel van deze analyse op andere voorstellen betreffende de oorsprong van de hexameter is dat alle cesuren en bruggen verklaard kunnen worden, waar andere hypotheses vaak maar één cesuur kunnen verklaren. Een ander voordeel is dat verschillende taalkundige aspecten, variërend van pure metriek, tot historische grammatica, formulariteit, syntaxis enzovoort gebruikt worden in onze analyse van de Homerische colometrie. Andere hypotheses vertrekken vaak uitsluitend van metrische overwegingen, zoals Nagy (1974; 1979; 1998) veelvuldig bekritiseert. Zijn eigen analyse blijft echter ook beperkt tot metriek en formulariteit. Op deze manier willen we benadrukken dat een cognitieve benadering van het Homerische vers niet alleen op het synchrone maar ook op het diachrone vlak interessante inzichten kan opleveren in verband met de colometrie van het Homerische vers. Het is niet onze pretentie dat deze thesis een sluitende oplossing biedt voor het probleem van de voorgeschiedenis van het Homerische vers. Dit vergt veel meer onderzoek, zoals in de conclusie besproken zal worden, en zelfs dan is het waarschijnlijk dat we nooit de volledige waarheid achter de oorsprong van de hexameter zullen vinden.

INTRODUCTION: The search for a dragon

"The origins of the Greek epic meter, the dactylic hexameter, are particularly challenging" (Watkins 1995: 21). The fact that one of the most distinguished Indo- Europeanists of the past century openly declares the difficulty which is inherent in the study of the protohistory of the Greek hexameter, implies that no consensus about its origins is reached until today2. The origins of this fascinating field of research lies in 19th century when scholars such as Bergk, Usener and others not only wanted to find the origins of the Greek hexameter, but were even hopeful to reconstruct a kind of Indo-European Urvers. Their ambition for doing so was certainly influenced by the manifest interest in comparative grammar culminating in the generation of the Junggrammatiker in late 19th century. However, it was not until Antoine Meillet, the founder of the French school of historical Indo-European linguistics, that a systematic study about the Indo-European origins of Greek and Vedic verse was published (Meillet 1923). Based on a thorough comparison between Vedic and - especially lyric - Greek verse, he proved beyond doubt that both metrical systems were "derived from some common source", as such implying an Indo- European metrical heritage3. His conclusions about the hexameter were less promising, for he was convinced that the Greek hexameter, having some properties which are not to be found in Aeolic lyrical verse, needs to be viewed as a borrowing from a pre-Greek, Aegean civilisation. This view remained dominant in scholarly literature, to such a degree that it was not until the seventies that new research concerning the protohistory of Greek hexameter was conducted4. But, this decade became a flourishing one concerning the origins of Greek verse. Several new hypotheses about the Greek hexameter were proposed, most notably by West (1973a), Nagy (1974) and Berg (1978). All three rejected an Aegean origin of the Greek hexameter, instead viewing the hexameter as a composite verse based on two shorter

2 In the words of Sicking (1993: 70): "Über die Herkunft des Hexameters gibt es verschiedene, notwendigerweise spekulative Hypothesen". 3 However, we need to be cautious not to confuse between elements which derive from a common Indo- European heritage and a Graeco-Aryan one, because these languages remained for some time together, after the diffusion of the Indo-European languages (cf. West 2007: 6; 20). Later research by Jakobson (1952) and Watkins (1963) on Slavic and Celtic metrics proved beyond doubt that it is possible to postulate a common Indo-European metrical heritage. Surveys can be found in West (1973b; 2007: 45-56) and Watkins (1995: 19- 21). Kurylowicz (1970) argues against an Indo-European origin of Greek and Vedic metre, because the similarities are too superficial and cross-linguistically too frequent to offer unrejectable proof for a common origin. "As for the question of a common IE origin of the Indic and the Greek verse, it must remain open" (Kurylowicz 1970: 429). This seems unlikely, because the metrical facts are accompanied by many linguistic similarities, e.g. the common use of tmesis in Greek and Vedic verse. Moreover, Kiparsky (2018: 7-8; pages refer to online edition) highlights the fact that anaclasis is also a common feature in Indo-European metrics which is not to be found in other metrical systems. 4 Apart from that, a general deficiency exists in Greek philology that the conclusions of comparative Indo- European metrics are mostly neglected in scholarship dealing with archaic poetry (Watkins 1995: 19).

metres (West and Berg) or as an internal development of a shorter Aeolic metre (Nagy)5. During the past decades, research focussed on criticisms against these hypotheses or on modifications and additional arguments in favour of these proposals. For instance, the theory of Berg was further promoted by the German scholar Tichy (1981; 2010) and by Bergs pupil Dag Haug (2000; 2001; 2002). We could in fact postulate a Norwegian school about the protohistory of Greek epic, which - generally spoken - is convinced that both the metre and the language of the Homeric epics are less archaic than normally assumed in Greek linguistics6. However, as I stressed in the beginning, the origins of Greek hexameter are not convincingly solved by these proposals. Every hypothesis has its own faults, as we will see in detail, and especially the hypothesis of the Norwegian school, which is nowadays the most widely accepted theory, lacks unrejectable proof. For example, the explanation of van Beek (2013 passim) that the supposed remnants of vocalic /ṛ/ in Homer need to be regarded as an inner-epic development, offers an important counter-argument against one of the most decisive elements in Tichy's argumentation (1981). Furthermore, Bergs hypothesis interprets the characteristic end of the Greek hexameter, the adonean (–∪∪–⨯), as a recent phenomenon which seems unlikely in light of its extremely high incidence (e.g. Miller 2014: 86). Therefore, this dissertation will focus on an alternative explanation concerning the origin of the Greek hexameter, which had already been put forward in 1913 by Kurt Witte. This hypothesis is based on the observation of the high importance of the bucolic diaeresis, segregating the first part of the hexameter from its characteristic adonean. As such, Witte explained the hexameter as a coalescence of an original dactylic tetrameter with a dimeter (the adonean). Although Witte argued for his hypothesis with strong arguments, both linguistic and metrical, his hypothesis was never accepted in scholarly literature. One of the purposes of this dissertation is to reconsider this theory, proving the validity of his arguments and even enlarging them. Furthermore, it will be necessary to draw a plausible picture of how an original tetrameter and dimeter could coalesce into one epic verse, resulting in the synchronic colometry of Homeric verses. I will use a non-traditional approach towards Greek metrics, based on cognitive principles. Most metricians consider the caesurae as demarcations between characteristic metrical phrases caused by regular word-end at a given place in the verse, most notably the penthimemeral caesura (3a) and the trochaic caesura (3b), as such neglecting the syntactic structure of the verse and the information structure of Homeric discourse (cf. West 1982: 6; 1987: 4)7. On the other hand, Bakker argued convincingly in several publications (e.g. 1990a;

5 West: hemiepes (–∪∪–∪∪–) and paroemiacus (∪̅∪–∪∪–∪∪–⨯); Berg: glyconeus (○○–∪∪–∪–) and pherecrateus (○○–∪∪––); Nagy: an inner expansion with three dactyls of an original pherecrateus (○○–∪∪––). 6 They conceptualize the hexameter, as we know it, as a post-Homeric development. In their opinion, the Aeolic phase lasted until the 9th century BC, only for a short period succeeded by an Ionian one. They also deny an Achaean phase which contributed to the formation of the Homeric epics (cf. Haug 2002: 35-62). 7 For the notation of caesurae and diaereses, I will make use of the notation proposed by Janse (1998: 138; 2003: 347; 2012: 18). The number refers to the number of the foot, the letter refers to the position within the

1990b; 1997a; 1997b), that Homeric language has to be interpreted as a form of spoken discourse. Making use of a theory initiated by Chafe (e.g. 1982; 1985; 1987; 1994), he demonstrated that Homeric verse lines are constructed as sequences of intonation/information units, short linguistic entities which match with the cognitive constraints of the human mind and are accompanied with prosodic demarcations. This interpretation of the Greek hexameter offers strong indications that the aoidoi composed their verses as a kind of "special speech" (Bakker 1997a: 146-150), which makes it highly improbable that they would pause in the middle of a grammatical or conceptual unit (cf. Janse 1998; 2012; Vergote 2011). A further advantage of this cognitive approach towards Greek epic verse is the fact that it further highlights the importance of the bucolic diaeresis, because it can be proven that a great many of the intonation units starts at this point in the verse and that a large amount of different syntactic structures start after the bucolic diaeresis (e.g. appositional noun-phrases, coordinated verb phrases, subordinate clauses etc.). As such, this cognitive approach could be an ideal opportunity to reconsider and possibly revive the proposal of Witte concerning the origins of Greek hexameter. In arguing for such a cognitive approach towards Greek metre and its relevance for research about the protohistory of the hexameter, I will make use of a corpus of three books from Homer's Iliad (I, XI, XVI), totalling 2326 out of 15693 lines in the modern vulgate edition. Using the Iliad for the reconstruction of Greek hexameter, deserves no surprise. It is generally accepted that it is the oldest literary work of Ancient Greece that has come down to us8. In view of Homer's massive number of lines and the limited number of pages of this thesis, I had to restrict myself to a limited corpus. The actual choice of the three books is determined by the observations of West (2011a: 48-55) about the composition of the Iliad. The three books I chose, are considered to belong to the oldest layer of the Homeric epics. One can ask whether such an analystic argumentation is not wholly arbitrary, mostly because it is rather based on observations regarding the content and not on linguistic ones. The fact is that it is very difficult to use linguistic data to establish a relative chronology within one work. Even between different works this can only be done with a very sophisticated methodology (Janko 1982). Therefore, I stick with my rather pragmatic choice,

foot, as such "a" refers to the first longum, "b" to the first breve and "c" to the second breve or to the second longum: –a∪b∪c. 8 West thinks Hesiod is to be regarded as the oldest poet of Hellas (e.g. West 2014: 33). He argues for this interpretation on the basis of some assumed intertextual passages between Hesiod and the Iliad and the Odyssey. However, this is a misleading methodology in an oral context. Similar phrasing can be derived from a common oral source, which is no longer attested. Moreover, we cannot be sure about the relative chronology of an intertextual allusion. Who adopts it from whom? (cf. Janko 1982: 9 for these problems or Kousoulini 2013: 428-429 with regard to Homer and lyrical poetry). Furthermore, linguistic arguments point towards the priority of Homer before Hesiod (Janko 1982 passim; Andersen & Haug 2012: 11). Therefore, West's proposal is seldomly followed by other researchers (Kirk 1985: 10).

but I will always be cautious not to overgeneralize some facts which are only based on a limited extract of the Homeric epics9. The methodology, which was outlined above, will be reflected in the structure of this dissertation. The first chapter will give an overview of the main research concerning the protohistory of the Homeric epics. Firstly, it will be useful to stress that by means of an oral tradition some elements of a remote Indo-European poetical heritage were preserved into archaic Greek poetry. The fact that the Homeric epics belong to an oral poetic tradition was argued on the basis of the formulaic character of noun-epithet formulae by Milman Parry in 192810. From that time onwards, this approach became gradually dominant in Homeric scholarship, resulting in publications dealing with comparative poetics, mostly between the Homeric epics and South Slavic poetry (Lord 1960), which was already initiated by Parry himself. Studies which exclusively focussed on Greek poetry further established the theory of the orality of archaic Greek poetry. Doing so, some adaptations were proposed with regard of Parry's theory, gradually underlining the flexibility of Homeric formulae (e.g. Russo 1963; 1966; Hoekstra 1965; Hainsworth 1968). In the last decades of 20th century this culminated in abstract compositional schemes of the Homeric epics (e.g. Visser 1987). A second section of the first chapter deals with the language of the Homeric epics, which is often referred to as a Kunstsprache, after Meister (1921). Firstly, I will give a short sketch of the main peculiarities of Homeric language, emphasizing the artificial character of the language. Afterwards, some possible explanations for the linguistic properties of the Kunstsprache will be given. The second chapter deals with the Homeric metre, which will be the core business of this thesis. After a short introduction, which will introduce the basic facts about Greek metre, I will focus on caesurae, being an important argument for the reconstruction of the hexameter. For a start, I give an account of the traditional approaches towards caesurae. However, as I already stressed above, important criticisms are to be objected against these conceptions. The theories of Chafe and Bakker will be introduced to argue for a cognitive segmentation of Homeric colometry. This will be further exemplified using some examples from the corpus of Homeric books. It will be useful to apply the different colometries which have been proposed, to a certain amount of verses in order to compare their relative usefulness and probability. The third and last chapter will be entirely devoted to the problem of the origin of the Greek hexameter. In a first section, I will re-examine the different proposals which were hypothesized during the last century. They will not be arranged in a chronological order but

9 Further decisive arguments for choosing these particular books were the fact that the phrase ἀνδροτῆτα καὶ ἥβην, one of the arguments used in favour of the Norwegian hypothesis is used at Il. XVI, 857. This book was furthermore used by Kirk (1966: 117ff.; 1976: 155ff.) in his study on enjambment and turned out to be representative for the whole Iliad (cf. Higbie 1990). Both XI and XVI belong to the longest in the Iliad, consisting of 848 and 867 verses respectively and book I has a fair amount of 611 verses as well. As such, they constitute a good representation for the whole Iliad. 10 His publications were collected in the volume by his son Adam Parry in 1971. References to Parry will be quoted from this volume.

proposals which use similar conceptual backgrounds will be placed together. I will present and criticize the proposal of a foreign origin of the Greek hexameter (Meister 1921; Meillet 1923), some proposals departing from a coalescence of two (Aeolic) metres (West 1973a; Peabody 1975; Vigorita 1977; Berg 1978), the proposal of Nagy (1974) concerning an inner dactylic development of a shorter pherecrateus and the most recent account by Kiparsky (2018) who argues for an iambic reconstruction of the protohexameter. Last but not least, I will present the hypothesis of Witte (1913), which will be further argued for in the last section of the dissertation. Using the cognitive analysis of the Homeric colometry, presented in chapter two, I will emphasize the importance of the bucolic diaeresis (4c) and the trithemimeral caesura (2a). They will be explained as archaisms in the colometry of Homeric verse. A next section will offer a possible explanation of a coalescence of an original dactylic tetrameter with an adonean, as such explaining the other common caesurae (3a, 3b and 4a) (partially) as innovations resulting from the coalescence of the original verses. The last section will briefly discuss the problem of unmetrical verses in Homer and their relevance for the origins of the Greek hexameter. The conclusion will briefly summarize the most important findings of this thesis and offer some directions for further research concerning the (origins of the) colometry of Homeric verse. As it was stressed in the beginning of this introduction, the question of the origins of the Greek hexameter are particularly challenging, so it would be rather audacious to pretend that this dissertation will offer a complete understanding. The aims of this dissertation are more modest. It will argue for the possibilities cognitive linguistics offer for the colometry of Homeric verse in general and especially for the history of the Greek hexameter, departing from Witte's proposal. Unfortunately, it is quite likely that Greek and Indo-European linguistics will never find the definitive origins of the hexameter. This dragon for scholarly research will possibly remain forever in his hole in Bronze Age Greece. Nevertheless, it remains very interesting to search where this cavern is situated.

CHAPTER 1: THE PROTOHISTORY OF GREEK EPIC

1.1 Telling a story about storytellers: The oral preservation of an Indo-European past. "How can I reach *ḱléṷos *´ṇdhgṷhitom ("unperishable fame")?" This question summarizes the ultimate aim of an Indo-European hero. He - female heroes are absent in ancient Indo-European epics - does his utmost to achieve immortal fame, mostly by dying young, but accomplishing heroic deeds during his short life. His life may be short, but his fame is not, people and especially poets will sing about his deeds and doing so, remember him forever. Adalbert Kuhn (1853: 467) was the first scholar who proved that this concept reflects an Indo-European heritage, comparing the poetic formulae of Greek κλέοσ ἄφθιτον and Vedic śrávaḥ ákṣitam / ákṣiti śrávaḥ, both meaning "unperishable fame" and being etymological cognates (e.g. Watkins 1995: 13; 19)11. This discovery meant the start of a new academic discipline, that of "comparative Indo-European poetics"12. The aim of this discipline is to compare poems in different ancient Indo-European languages and search for common phrases and themes which emerge in these and could in fact reflect an Indo- European poetic heritage, "genetic intertextuality" as it was coined by Watkins (1995: vii). Three domains offer important information concerning comparative poetics: formularity, metrics and stylistics (Watkins 1995: 12). This discipline has proven, for instance, that some prosodical features in Greek and Vedic verse reflect an Indo-European heritage, e.g. the quantitative metrics based on an alternation between heavy and light syllables (Meillet

11 The most exhaustive discussion of κλέοσ ἄφθιτον and Vedic śrávaḥ ákṣitam / ákṣiti śrávaḥ is offered by Nagy (1974), with criticism by West (1974) and Haslam (1976), cf. infra. There are some problems with his grammatical analysis of the Vedic forms (Nagy 1974: 159-160). He wants to interpret ákṣiti as a bahuvrīhi compound based on the "a privativum" and an abstract noun kṣití, but in such cases with "privative a", the rule that a BV is accentuated on its first member is not applied (cf. MacDonell 1916: 455). 12 Watkins (1995) and West (2007) are two monograph studies which can be used as introductory books, although they contain much more information than a normal introduction. Further references are also found there. A short survey of the most striking similarities between Indo-European poetical traditions (especially focussed on Greek and Indo-Aryan literature) is offered by West (1988: 152-156). For a summary of the epic tradition in Greece with reference to its Indo-European heritage, cf. Dowden (2004). Mahoney (2007) stresses the fact that a metapoetical system existed in Indo-European literature to refer to the technical aspects of making poetry (also concerning the metre itself). Cf. also Finkelberg (2011 vol. 2: 407-409 s.v. Indo-European background).

1923: 11)13. Therefore, we are able to postulate an Indo-European poetical tradition which is transmitted into the different daughter languages14. One question remains: how could this poetic tradition be maintained over a period of thousands of years? The answer was found by a young American scholar, who wrote his PhD in Paris and published it in 1928. In his main dissertation, entitled "l'Épithète traditionelle dans Homère", Milman Parry proved on the basis of a meticulous analysis of noun-epithet formulae in the Iliad that Homer belonged to an oral tradition15. One has to envisage hundreds of generations of poets who from the Indo-European age down to archaic Greece learned how to compose poems in a traditional manner and perform them orally, without the aid of writing16. In order to do so, they had to rely on a vast range of pre-made phrases, which were transmitted from teacher to pupil and which were coined "formulae" by Parry. His definition of a formula is possibly the most quoted definition in the history of Greek philology, because it heralded a new era in the study of Homeric scholarship. The old controversy between Analysts and Unitarians became substituted with a new, more realistic paradigm, "oral poetics"17. Parry characterized the formula as "a group of words which is regularly employed under the same metrical conditions to express a given essential idea" (Parry 1971: 13). When we apply this definition to a common noun-epithet formula, e.g. ποδάρκησ δῖοσ Ἀχιλλεύσ ("swift-footed divine Achilles"), this means that a group of two adjectives and a personal name with a particular metrical structure (∪–––∪∪–⨯) is regularly used after the trochaic caesura (3b) to express the essential idea "Achilles". To corroborate Parry's thesis, Antoine Meillet, his unofficial PhD-supervisor, introduced his pupil to Matija Murko, a specialist in Balkan poetics (De Lamberterie 1997: 14). In the Southern Balkans, a living tradition of orally performing poets still existed, named guslari. Therefore, Meillet thought it to be very interesting to compare Parry's findings about the traditionality of Homer with this living tradition18. Parry learnt Serbo-Croatian to conduct such comparative research and collected an extensive corpus of tape recordings. The comparison of this modern, but very traditional poetry with the Homeric epics, strengthened his main beliefs

13 Comparative Indo-European metrics will be further discussed in chapter three. 14 For an overview of the poetic traditions of the Indo-European daughter languages, reference is made to Watkins (1995: 50-67) and West (2007: 12-19). They offer a (non-exhaustive) list of possible sources for comparative research. For example, West does not mention the Old Indian purāṇās (long mythological texts, offering many religious information). 15 An interesting discussion of Parry's life and the impact of his research on Homeric scholarship is given by his son Adam Parry in the introduction of his edition (and English translation) of his father's works (Parry 1971: ix ff.). Cf. also Foley (1997: 146-151). 16 An overview of the social function of Indo-European poets is given by Watkins (1995: 68-93) and West (2007: 25-30). 17 For overviews of the Homeric question with the debate between "Analysts" and "Unitarians", cf. Davison (1962); Fowler (2004b); Finkelberg (2011 vol. 1: 47-50 s.v. Analysts; vol. 2: 362-364 s.v. Homeric question; vol. 3: 911-913 s.v. Unitarians). 18 Originally, Parry intended to prove that Homer was a traditional poet. The conception that he belongs to an oral tradition was gradually developed due to his comparative research, as will be discussed below (cf. De Lamberterie 1997: 13; Bakker 1997a: 10).

about Homer. For him, Homer was an anonymous author and the ultimate exponent of an oral tradition19, a tradition which Parry himself characterized with two main features: "simplicity" and "extension" (e.g. Parry 1971: 7). Oral poetry is "simple" because it is built on a relative straightforward system of traditional formulae and themes and "extensive" for every new generation of poets creates new formulae. Unfortunately, Parry died already in 1935 due to an accidental gunshot, so that he was not able to finish his comparative research. This was done by some of his pupils, most notably Albert Lord (cf. Lord 1960) and James Notopoulos. Whereas Lord extended the research of his teacher concerning the oral traditions in Serbo-Croatian language, Notopoulos opened a new domain of comparative research, by comparing the findings of his teacher about the Homeric epics with Modern Greek folk songs, mostly those that were still performed in Crete. Decades of research culminated in Lord's masterpiece, "The Singer of Tales" (Lord 1960). In the parryistic paradigm, Homer was no longer an author, he became a singer of tales, such as the modern guslari in former Yugoslavia. Amongst others, he developed a model for the education of an oral poet. Whoever wants to become a good poet, has to begin with listening to other poets. You need to incorporate the characteristic rhythm of oral poetry, so that you can start performing yourself in a second stage. You have to learn the formulae and the poetic language. Once the neophyte is able to perform a poem to an audience, he is considered a full-fledged member of the fraternity of poets. The last stage of his education starts here. During the rest of his professional life, he has to extend his repertoire and refine his techniques (Lord 1960: 21-26). Moreover, Lord was able to refine the notion of formula. Gradually, he stressed the important fact that an oral poet is not a slave of his traditional technique, but that a good one needs to deal creatively with this inherited system. Doing so, formulae are modernized by every new generation of poets (Lord 1960: 43-54). This new conception of a formula, namely as a creative technique - not a fixed idiom, but a flexible poetic tool - became gradually dominant in Homeric scholarship (cf. Graziosi & Haubold 2010: 15). Before discussing it, I want to highlight the fact that it took a considerable amount of time before Parry's thesis became actually accepted in the field of Greek literary studies (Greek linguistics was more inclined towards an acceptation of the

19 West (1999) went even so far as to use the term "the invention of Homer". In his opinion, the name Homer was given to a fictional poet whose text was preserved by a group of professionals called the "Homeridae". West's conception of Homer is somewhat paradoxical. On the one hand, he wants to stress the anonymity of the author, even to such a degree that he uses "author P" for the composer of the Iliad (West 2011a) and "author Q" for the poet of the Odyssey (West 2014). On the other hand, he denies the importance of the oral tradition (cf. West 2001: 3ff.: "Homer is a written text as any other ancient Greek author"). For further criticism on this approach, cf. Chadwick (1990: 174) and Nagy (2004: 40-108 - also about the idiosyncrasies in West's edition of the Iliad). Nagy is to be placed at the other end of the continuüm. He believes that different oral recensions of the epics existed until the Hellenistic period (e.g. Nagy 2004: 3-39). For criticism on this approach, cf. Andersen & Haug (2012: 7 - Nagy underestimates the difference between aoidoi and rhapsōidoi). A more modest approach towards the orality of the Homeric epics seems more appropriate.

theory)20. Parry was influenced by important German scholars such as Düntzer, Ellendt and Witte (cf. Parry 1971: 5; De Lamberterie 1997: 15), but parts of the classic Altertumswissenschaft sticked to an analytic approach of Homeric scholarship, even after Parry's publications. In the Anglo-Saxon world as well, where Parry himself originated, it lasted until the fifties before Homer was accepted as an oral poet. The sixties became a flourishing decade, when new studies concerning Homeric formularity were published. The conclusions of Russo (1963; 1966), Hoekstra (1965) and Hainsworth (1968) shared some similar points to those of Lord, although they based their research only on an internal analysis of the Homeric poems. They searched for a broader vision on formulae, stressing their flexibility. In fact, this is already found in the writings of Parry himself21, but later scholarship reduced his conception of a Homeric formula to "a regularly repeated Homeric phrase". Russo (1963: 237) argued for a definition that characterizes formulae as "localized phrases whose resemblance goes no further than the use of identical metrical word-types of the same grammatical and syntactic pattern, as truly representing certain more general types of formulaic systems"22. For example, ἄλγε' ἔθηκε (Il. I, 2) ("he caused sufferings") is in his opinion not only a formula because it is repeated in the Homeric corpus (Il. XXII, 422) but even more so because it reflects an abstract grammatical phrase, based on a direct object and a finite verb, that is metrically localized, namely after the bucolic diaeresis (4c) (cf. Russo 1963: 243). Therefore, his ideas about Homeric formulae can be summarized as a combination of exact repetitions and recurring formulaic patterns, both on a grammatical and on a metrical level (Russo 1966: 226-227). Similar conclusions were independently reached by Hoekstra (1965) and Hainsworth (1968). Hoekstra criticized some of the methodological approaches initiated by Parry, mainly his inconsistent definition of the Homeric formula. For example, Hoekstra stressed the fact that the Homeric poems also consists of formulae which are not regularly employed23 and that formularity in itself is not a decisive proof of the orality of a given text. Later authors can also imitate a formulaic style, without being themselves oral poets (Hoekstra 1965: 12-18)24. Therefore, Hoekstra shifted

20 Parry's findings were on the linguistic level already preceded by Witte's (1913; 1972) and Meister's (1921) conception of the artificiality of the Homeric language, as I will discuss in detail in the next section of this thesis. Meister (1921: 240) already pointed to the similarity between the Homeric Kunstsprache and the Serbo- Croatian epics. 21 Parry (1971: 301-304) offers a formulaic analysis of the openings of the Iliad and the Odyssey and stresses the fact that a formula is not necessarily a repeated phrase. 22 A thorough statistical study of the place where different words and phrases of different metrical schemes are placed was completed by O´Neill (1942) and Porter (1951). O´Neill (1942: 144) coined the term "localization" to refer to the tendency of certain metrical shapes to be restricted to limited positions in the verse. 23 However, we always have to realize that we only possess a limited corpus. Therefore, a phrase that is only used once by Homer, could in fact be used hundreds of times in works that were lost. 24 For a comparative study between Homer and later authors, cf. Parry (1971: 24-36) for Vergil and Apollonius Rhodius, or Visser (1987: 266ff.) for Quintus Smyrnaeus. Janko (1982: 19, cited mostly with approval by Higbie 1990: 80) offers three features for oral texts: a complex formulaic system based on extension and economy, a limited amount of necessary enjambment (cf. infra) and metrical anomalies caused by a "jeu des formules" (cf.

the attention to Parry's concept of "extension". Formulae are not a static entity, but they could be extended and adapted to create new formulae. In his "collection of essays", as he denotes his book (Hoekstra 1965: 5), he discusses several techniques that were used by the aoidoi to keep their formulae useful for epic diction: the declension of a formula (e.g. from the genitive μερόπων ἀνθρώπων to the nominative μέροπεσ ἄνθρωποι (already Witte 1913: 2223), the conjugation of a verbal formula, the replacement of obsolete forms by more common ones, dividing an older formula by placing a word in the middle of it, to shift a formula to a new metrical position (cf. O´Neill’s localization) or by enjambment (Hoekstra 1965: 88-109 with examples). Hainsworth (1968) equally highlighted the flexibility of Homeric formulae. Parry's definition is too specific in his opinion. According to him, we have to generalize it to "a repeated word-group"25 (Hainsworth 1968: 35), always realizing that "formulae are moved, modified and seperated" (Hainsworth 1968: 32). In the same way as Hoekstra, he discusses different techniques how Homeric formulae could be altered throughout time: for example by means of inflection, the use of alternative suffixes, the adaptation of prefixes etc. (Hainsworth 1968: 36-37)26. During the sixties, research concerning the Homeric epics not only focussed on a more refined definition of a formula, but also wanted to study more thoroughly the relation of the Homeric epics towards broader cultural developments. This was certainly triggered by the decipherment of Linear B by Ventris and Chadwick in 1953, because these Mycenaean documents could offer a new insight in the protohistory of the Homeric epics (and of its language, as we will see below) (cf. Kirk 1962: 23-29; 1965: 45-54). Kirk (1962; 1965; 1976) developed such a comprehensive view on the Homeric epics. He created a template to conceptualize the development of oral traditions all over the world, but especially based on the Homeric epics. Traditionally, oral poetry passes four stages in his opinion, first an originative stage when short, simple poems are composed (probably from the Proto-Indo- European period onwards), culminating in a second stage, the creative one, when longer and more refined poems are composed. The Iliad and the Odyssey belong to this second stage. Afterwards, a reproductive stage appears, when the repertoire gradually stops to be extended, as is seen among the Serbo-Croatian guslari. The fourth and last stage is infra). In fact, we only have negative tests to prove if a text is orally composed or not, when we cannot prove that it is not one, we have to conclude that it is probably an oral text (Janko 1982: 40-41). 25 This definition is somewhat vague, for it is not clear if concrete word groups are meant or a more general grammatical and/or metrical structure (cf. Russo 1963; 1966). While reading his book, the reader realizes that indeed this last meaning is intended. Discussion of different processes with numerous examples: Hainsworth (1968: 46-108). His interpretation of Homeric formulae is briefly summarized in his introduction to the third volume of the Cambridge Commentary on the Iliad (Hainsworth 1993: 1-30). Cf. his conclusion (p. 30): "The language and diction of the epic tradition was never static; it was always an amalgam of old and new, and the old was constantly being eroded". 26 A similar survey of past research concerning Homeric formulae is found in Edwards (1986: 189-197; 1997: 264-267); Bakker (1990b: 386-389; 1995: 97-99); Nasta (1995: 202ff.); Russo (1997: 242-257); Clark (2004: 118-119; 123-134). For good discussions, cf. also Bowra (1962b); Hainsworth (1993: 1-31). More generally about "oral theory", cf. e.g. Foley (1997: 165-173); Finkelberg (2011 vol. 2: 604-610 s.v. Oral-Formulaic theory & Oral traditions).

degenerative, the oral poetic tradition declines and eventually disappears, the aoidos becomes a rapsoidos (Kirk 1962: 45-47; 1965: 27-29)27. Moreover, he argued for the possibility of a Mycenaean epic tradition, but equally highlights the fact that the Dark Ages must have provided enough opportunities to transmit the oral poetical tradition. The cultural flourishing of Athens could have provided the ideal context for doing so (Kirk 1976: 26-32)28. Apart from his research about the relationship between the Homeric epics, Kirk became equally renowned in Greek philology for his new classification of enjambments (Kirk 1966: 106-110). Because of the relatively high importance of this poetic figure for our understanding of the orality of the Homeric epics, it seems appropriate to include here a short excursus about this procedure29. Parry realized its importance and wrote a paper about enjambment in the Homeric epics. He differentiated between two kinds of enjambment: unperiodic and necessary. The first one refers to verses which were complete at the end of the verse, but for which additional information is given in the following verse. This a feature of the Homeric epics that Aristotle called λέξισ εἰρομένη, literally "stringing style" (cf. infra). The other one means that the syntactic structure was not finished at verse- end, so it had to be completed in the following one (Parry 1971: 251-263)30. Kirk deepened Parry's classification of enjambment, postulating four different possibilities: progressive, periodic, integral and violent. Progressive enjambment is an alternative name for Parry's unperiodic enjambment. Kirk's periodic enjambment encloses the cases when a protasis is followed by an apodosis in the next verse, whereas integral enjambment is similar to Parry's necessary enjambment. Kirk's last category, that of violent enjambment is in fact a sub- category of his integral enjambment and refers to the fact that a grammatical unit, such as

27 This template is too deterministic and circular. It contains striking similarities with e.g. Polybius' vision about the rise and (future) fall of Rome or with Schleicher's theory about the degeneration of languages. Janko (2012: 33) criticizes another deficiency in Kirk's conception of the oral tradition. Kirk believes that due to the oral tradition the Homeric epics are meticulously transmitted throughout time, as e.g. Ṛgveda. This holds indeed for Ṛgveda, which is a religious text, but not for the Homeric epics, which - their enormous importance notwithstanding - were never a real sacred text (Janko 2012: 33). Moreover, Kirk should have differentiated between aoidoi and rapsōidoi. 28 Due to the limited scope of this thesis, I confine myself to a short survey of the relationship between Homer and his cultural context. Cf. e.g. the relevant parts of Kirk (1962; 1965; 1976) or Bennet (1997). The latter minimizes the relationship between Homer and the Mycenaean age. 29 Cf. Edwards (1986: 223-228) for a (partially) out-dated status quaestionis of the concept. Allan (2009: 114) gives a short survey about the commonly used terms. Cf. also Bakker (1990a; 1997b: 302); Nasta (1995: 213ff.); Finkelberg (2011 vol. 1: 252-253 s.v. Enjambment). 30 A survey of the different grammatical possibilities with necessary enjambment is given by Edwards (1966: 123). A problem inherent in the study of enjambment is the question whether they add a kind of emphasis on the enjambed word. Bassett (1926: 142-148) totally denied a possible emphasis on the enjambed word, which seems unlikely because many interjections of the Homeric narrator (as the famous οὐλομένην ("wicked") in Il. I,2) are placed in an enjambed position (cf. Edwards 1966: 139). It seems more appropriate to discuss whether a certain enjambment causes emphasis or not for each particular example and with reference to the kind of enjambment, as will be presented infra.

for example the combination of an adjective and a noun, is divided by the verse-end31. Higbie (1990) devoted a monograph to the problem of enjambment and further refined the categorization of the phenomenon. She replaces progressive enjambment with adding enjambment and periodic enjambment is substituted in her classification with clausal enjambment. Instead of Kirk's integral enjambment she recovers Parry's term necessary enjambment, while she retains the phrasing violent enjambment (Higbie 1990: 29). Her theory makes a further distinction between internal and external expansion, the first one denoting the case when a clause itself is enlarged, the second one the attachment of a new clause (Higbie 1990: 32). Why are these subdivisions important in our study of the orality of the Homeric epics? They provide important evidence for distinguishing oral from written texts. Different statistical studies, culminating in the meticulous analysis by Higbie (1990: 66ff.; 1995), stress the fact that Homer and other oral texts mostly use verses which either do not make use of enjambment, or if they do, of "progressive/adding enjambment". This can be explained with reference to the oral performance. An oral poet does not have the time to think again and again about the verse he is going to produce. Therefore, he tries as much as possible to end his verse with a grammatically completed clause. In the process, more emphatic kinds of enjambment are avoided by Homer and his colleagues (Higbie 1990: 67)32. Let us return to the notion of a Homeric formula and the consequences it had for our understanding of the composition of archaic Greek epic. Stressing the fact that a formula needs to be viewed as a flexible unit, was one step forward to prove that, although Homer belonged to an oral tradition, he remained a creative genius33. Another problem concerned the meaning of the typical epitheta ornantia. Formalistic adherents of Parry's method denied the fact that such an epithet had in fact any meaning, which is a noxious conception for a literary appraisal of the Homeric poems. However, this is a again a misunderstanding of Parry's own writings, because he himself already made a distinction between ornamental and particularized epithets (Parry 1971: 21; but cf. Parry 1971: 118ff.). Vivante (1997: 1-4) stands at the other end of the spectrum. In his opinion, an epithet always has a meaning, because it offers a materialistic view within the Homeric world. This alternative view shows a lack of understanding of the oral tradition and merely endeavours to put Homer again on a pedestal as a literary genius. He is worthy of such an appraisal, but this can only be

31 This last category is very unusual (cf. Kirk's statistics (1976: 172-182) for book XVI of the Iliad). Edwards (1966: 124-135) terms it "harsh enjambment", emphasizes its rarity and discusses the different syntactic possibilities. 32 This was apparent from the statistical studies of Parry (1971: 251-265) himself onwards. Further refined and corroborated by Kirk (1976: 172-182), Barnes (1979: 7-10) and Clayman (1981: 115; 121-133 - she mainly focusses on sentence length basing her material on phonemes). They marked a tendency for longer clauses and a different use of enjambment in later, written poetry, as that of Apollonius and Vergil. Higbie (1995) applies the statistical method she used for the Iliad on Hesiod’s Works and Theogony. As such, she proved that the Works is quite similar in its use to the Iliad, but that the Theogony makes more use of adding enjambment (Higbie 1995: 71; 108-110). Higbie’s data for the Iliad (1990: 66) 75,53% verses without enjambment or with adding enjambment; 24,47% for the other types of enjambment. 33 Chafe (1982: 49) catches the problem of an oral interpretation of literary works: "The term "oral literature" seems, etymologically at least, to contain an internal contradiction".

accomplished by developing a correct understanding of what is traditional and what is created by Homer himself. Applying this to epithets, we have (again) to become aware of the particular context and decide if a specific epithet is purely ornamental, or has a particularized meaning34. The question remains how we have to envisage an oral Homer composing his verses orally. Did he use a kind of template? The Swiss school of Homeric studies, which was guided by Latacz, thought he did so. Some PhD's were devoted to the problem, most notably the one by Visser (1987). Based on a precise analysis of the battle scenes in the Iliad, Visser tried to make a distinction between ready-made determinants (the fixed elements) and "filler words" (the variable elements) (Visser 1987: 29; 1997: 165). For a good understanding of his template, we have to distinguish between determinants concerning both metrics and content (components which have a fixed metrical localization in O´Neills term and are necessary to express the "essential idea" of the verse), determinants with regard the content, but which are variable on a metrical level, and ingredients which are complementary with regard to metrics as well as content (Visser 1987: 194-195). For example, the express the idea "X answered Y", the poet can use the following template:

– δ' ἀπαμειβόμενοσ (-η) προςέφη ∪∪–∪∪–⨯ or ––∪∪–⨯

The determinant is already filled in and according to the situation, variables are added which suit both the content and the metrical structure, for instance resulting in the verse:

τὸν δ᾽ ἀπαμειβϐμενοσ προςϋφη πϐδασ ὠκὺσ Ἀχιλλεϑσ. (Il. I, 84). ("Answering him, the swift-footed Achilles said")35

This distinction between fixed and variable elements became also applied to the notion of a formula. Bakker (1990b) proposed a new method to avoid the problems which were inherent in Parry's definition of the formula (cf. supra). According to him, we have to distinguish between central elements, i.e. the fixed element of a formula, and the peripheral ones, which are additions and could be altered according to the context or throughout time. For example, when we return to the example of ποδάρκησ δῖοσ Ἀχιλλεύσ, we can describe Ἀχιλλεύσ as the centre of the formula, because it refers to the essential idea "Achilles". The two adjectives are peripheral elements which were added throughout time and their

34 Cf. Edwards (1966: 118; 1988: 24-37; 1997: 272-277); Bakker (1990b: 385); de Jong (2012: 27). 35 Cf. Visser (1987: 33ff.). Visser splendidly summarizes his concept at the end of his book: "In der Zusammenfassung ist somit festzuhalten: die homerische Technik der Verskomposition ist kein reines Addieren von Formeln, sondern in der meisten Versen ein immer wieder neues Zusammensetzen einerseits der metrischen Determinanten, die den individuellen Aussagewillen des Dichters Homer repräsentieren, anderseits der mit Blick auf die metrischen Voraussetzungen gewählten, von der epischen Tradition geprägten und vorgegebenen Füllelemente". (Visser 1987: 336). Short surveys of the theory are found in Bakker (1990b: 400- 403), Visser (1997) and Hajnal (2003: 226-228). All translations in this thesis are my own, unless otherwise mentioned. Quotations of Homer's text are from the OCT edition by Monro & Allen.

possible use by the oral poet depends on metrical reasons (Bakker 1990b: 389-390; 1995: 116; Bakker 2005: 5). Formulae are in fact a kind of "grammaticalization", which become pre-made phrases, because they are constantly used through time (Bakker 1995: 100; 104)36. During the last decades, Homeric scholarship became dominated by two Anglo-Saxon scholars: Martin L. West and Gregory Nagy37. As I already stressed, they are the polar ends in the spectrum of contemporary Homeric scholarship. West (e.g. 2011a: 3ff.) reacts against the Parry-Lord hypothesis, because it minimizes the originality of the poet. He acknowledges the fact that an oral tradition needs to be envisaged for Homers predecessors, but the Iliad, which West dates as late as 680-640 BC, was written down by an anonymous author, probably on Euboea (cf. infra). Nagy on the other hand, describes the whole archaic (and even partially the classical) period of Greek literature as oral literature. Both the epic and the lyrical traditions originated in different parts of the Greek cultural area at different moments and because of the cultural prosperity of Athens, a panhellenic awareness arose, which crystallized in a canonization procedure of archaic poetry (cf. Nagy 1990: 52ff.). With regard to Homer, he developed a template that stresses the multiformity of the Homeric epics. For in his opinion, the Homeric text was only fixed in the Hellenistic Period with the edition by Aristarchus (Nagy 1996: 108-109)38. Probably, the "truth" lies in the middle of this two extremes.

1.2 The language of the Homeric poems: A synchronic approach. The oral composition of the Homeric poems is also reflected in their traditional language. They are not "written" in the ordinary dialect of their composer, not even in a stylised form of such an everyday language. Instead, they are composed in a language which shows traces of diachronic and diatopic variation, consisting of elements of Achaean, Arcado-Cypriot, Aeolic, Ionic, Attic and (possibly) West-Greek origin39. Therefore, Meister

36 A good survey of his theory is given in Bakker (2005: 1-36). The term "grammaticalization" is Bakker's, but it seems preferable to use "lexicalization" or "idiomaticization". Epithets of heroes have a meaning according to Bakker, because they glorify and justify their heroic deeds (Bakker 1997a passim), but every individual case needs to be accounted for. This is also the main objection of de Jong (2012: 26-27) against the theories of Visser and Bakker. 37 Gregory Nagy was born in Hungary, but he spent almost his complete life in the USA. 38 Summary of his model: 1) most fluid stage: no written text, 2) formative, pan-Hellenic stage (8th-6th century BC), 3) definitive stage: transcription of the text in Athens, 4) standardizing: transcriptions in 4th century, most notably by Demetrius of Phalerum, 5) most rigid: Aristarchus' edition in 2th century. Criticism by Andersen & Haug (2012: 7). 39 In what follows I am not going to offer a full treatment of Homeric language. This would require a monograph, instead I offer a brief survey of the main peculiarities, stressing the composite and artificial nature and some differences with the Attic language. The standard grammars remain until today Monro's Grammar of the Homeric dialect (1891) and Chantraine's Grammaire Homérique (dating from the fifties but recently (2013;

(1921) in his seminal work on the artificiality of the Homeric poems coined it a Kunstsprache40. Certainly, he was influenced by the many writings of Witte (1913; 1972), who already discussed a huge amount of artificial forms, which in his account need to be interpreted as ein Gebilde des epischen Verses (Witte 1913: 2214). Due to the limited space, I will restrict this survey of Homeric language to the most salient features of phonology and morphology. Only features which will be used in the further discussions of this thesis, will be discussed in this section. The following brief outline of Homeric language will be arranged according to their diatopic origins and highlights the contrasts with the Attic dialect, for these regional differences are significant for our conception of the development of the epic tradition (cf. infra). Firstly, it needs to be stressed that Ancient Ionic constitutes the core dialect of the Homeric language. Already in Antiquity, grammarians denoted it as Παλαιά / Ἀρχαία Ἰάσ (Ruijgh 1995: 1; Janse 2012: 2). More specifically, the dialect used is the dialect of the Ionian colonies in Asia Minor. Some eye-catching features, mainly phonological, point towards this direction41. In the first place, we can think of the common Ionic vowel shift, where inherited /a:/ evolved into /ε:/, also after /i/, /e/, /r/ (e.g. χώρη vs. Att. χώρα). This feature is responsible for some typical Ionic genitives, which were affected by quantitative metathesis after the application of the above stated sound change, resulting in the gen. sg. of masculine nouns in -/a:/- (e.g. Ἀτρεΐδᾱο > *Ἀτρεΐδηο > Ἀτρεΐδεω). There exists yet another phonological process which results in a distinctive Ionic genitive: vocalis ante vocalem corripitur before /o:/ in the next syllable (e.g. *βουλα ́ ςων > βουλα ́ ων > *βουλήων > βουλέων (Ion.) vs. Att. βουλῶν)42. With this example we can proceed to a following characteristic of Ancient Ionic: the lack of contractions (cf. infra). Additionally, Eastern Ionic is normally distinguished by psilosis, the phonological phenomenomn when an initial /h/ disappears.

2015) corrected in a new edition). Good surveys of the Homeric language are found in Witte (1913); Palmer (1962); Ruijgh (1995); Horrocks (1997); Wachter (2000); Meier-Brügger (2003) and Hackstein (2010). Hackstein (2002) is a monograph study about different aspects of the Homeric language. This work has not been accepted universally and contains some problematic analyses. Miller (2014) places the Homeric language in the broader context of Greek dialects. For a thorough study of artificial forms, cf. Witte (1913; 1972) and Meister (1921). A very short introduction to the Homeric language is offered by de Jong (2012: 29-33) and other "Green and Yellows". Cf. also Finkelberg (2011 vol. 2: 448-449 s.v. Kunstsprache; vol. 2: 458-464 s.v. Language, Homeric). Due to the massive literature on the subject, references will be limited to the above mentioned standard accounts. 40 Hackstein (2002: 4) proposes to use the term Dichtersprache. This is, however, a problematic term, because it cannot adequately describe the peculiarities of the Homeric language. Vergil equally writes in a "poetic language", but his language is by far not such a linguistic amalgam as Homeric Greek. Van Beek (2013: 191) wants to consider it as a separate variety, with its own internal developments. Hackstein (2002: 47) already pointed in this direction when he postulated the existence of "Innersprachliche Konditionierung". 41 The following survey of Ionic phonological features is mainly based on Palmer (1962: 77-81); Ruijgh (1995: 13-17); Horrocks (1997: 212); Tribulato (2010: 390); Finkelberg (2011 vol. 2: 416-417 s.v. Ionic dialect); Chantraine (2013²: 15-74); Miller (2014: 139-182). 42 Tribulato (2010: 390) interprets this also as quantitative metathesis. This approach is problematic, because in this case the quantity of only one vocal is affected, not of both of them (cf. also Janse 2012: 9 fn. 20).

For instance, this is to be observed in the distinction between Homeric οὖλοσ vs. Attic ὅλοσ ("every")43. Some morphological endings further point in this direction. We encounter the athematic infinitive in -ναι, which is a salient feature of Ionic-Attic dialect. In addition, the modal particle ἄν is not found outside Ionic-Attic (Aeolic has κε, West-Greek κᾱ)44. Specifically Ionic is the contraction ἤν from ἐάν, where Attic has only ἄν with /a:/(Palmer 1962: 85). Generally speaking, the Ionic features in Homeric morphology are more archaic than Attic morphology, because contraction is mostly not applied45. This lack of contraction can also leave traces in the metrics of Homeric language (cf. infra), e.g. the gen. sg. of second declension -ου sometimes needs to be interpreted as -οο (> *ος(j)ο) to scan an hexameter properly (e.g. Palmer 1962: 108)46. Furthermore, Homer's Ionic preserves other salient archaisms. For instance, sometimes a petrified vocative case is used instead of a nominative, e.g. in the noun-epithetformula ἱππότα Νέςτωρ instead of *ἱππότησ Νέςτωρ (e.g. Palmer 1962: 108)47. In addition, Homer offers examples of an archaic formation of the subjunctive, with short instead of a long vowel. Originally, this formation was characteristic for athematic verbs, especially asigmatic aorists, e.g. πεποίθομεν (e.g. Palmer 1962: 126-127). Another archaism pertains the optional use of the augment. This preterite morpheme is not yet

43 Further discussion: e.g. Meister (1921: 196ff.); Chantraine (2013²: 184-188); Palmer (1962: 78); Miller (2014: 197ff.). On the other hand, there are some instances where psilosis is not applied. This is one of the arguments used by West (1988; 1992), Ruijgh (1995: 47-50) and Wathelet (1997: 47) to argue for a final edition of the Homeric epics in Euboea. For criticism, cf. e.g. Chadwick (1990); Crespo (1997: 135); Willi in Finkelberg (2011 vol. 2: 459 s.v. Language, Homeric). 44 There are some problems to relate the different dialectal forms with one another. An elegant solution is proposed by Palmer (1962: 91-92). Two important papers are Forbson (1958) and Colvin (2016), the former interprets the forms with /k/ as derivations of Proto-Greek *kṇ (cognate with Vedic kam) and ἄν as a Greek development. However, this cannot convincingly explain the long /a:/ in West-Greek. Colvin interprets ἄν as an inherited word and the forms with /k/ as re-analysed forms in Greek. For a summary, cf. Miller (2014: 328- 331). 45 We need to be cautious with terms as "archaisms" and "innovations" in the epic language. Some assumed archaisms could in fact be artificial formations by the poets, or "false archaisms" (Janko 1982: 76; 2012: 32) who due to later intropolations intruded in the language. Furthermore, not every feature of the Homeric language is archaic, cf. Hackstein (2002: 19-33; 46-91; 2010: 404-407) who stresses the importance of innovations in epic Greek (also Wachter 2012 passim; van Beek 2013). Crespo (1997) discusses the hierarchy between the different layers of the Homeric Kunstsprache. 46 There is discussion whether this genitive comes from PIE *-osjo, preserved e.g. in Sanskrit -asya or Archaic Latin -osio or if we need to posit a second gen. sg. in *-so, which is possibly preserved in Old Bulgarian. Cf. Haug (2002: 70-106), arguing for the first option. Cf. also Rix (1992²: 137ff.) preferring the second explanation. I also agree with the second one. Further discussion, cf. Monro (1891: 83); Palmer (1962: 106); Horrocks (1997: 207); Chantraine (2013²: 44-47; 165-170). 47 Hainsworth (1993: 24) attempted to explain these archaisms on the basis of phonotactic rules. According to him, a form as *νεφεληγερέτησ Ζεύσ, with the sequence (/s-zd/) would be difficult to pronounce. This proposal is not able to explain why these forms are also attested before other consonants, e.g. ἱππότα Πηλεύσ/Νέςτωρ. It is important that these forms are localized at the end of the verse, where many archaisms are preserved (cf. infra). Cf. also Wachter (2000: 93).

obligatory in Homeric Greek and can be omitted, an archaic feature inherited from Mycenaean (cf. infra) and as such, also attested in Vedic Sanskrit48. However, as already stressed in the beginning, the Homeric epics are by far not a pure exponent of Ancient Ionic, but a composite language based on different dialects. The most important other dialect which contributed to the Homeric poetic language, is the Aeolic one. This was already observed in Antiquity and became one of the most fertile grounds for Homeric research from the end of 19th century onwards49. We can think of Hinrichs (1875), who was the first to postulate an Aeolic phase in the genesis of the Greek epic tradition or Fick, who went as far as to try to recover an original Aeolic version of the Iliad which later on was transmitted in the Ionic dialect (cf. Andersen & Haug 2012b: 8). These endeavours directly emphasize the pervasiveness of Aeolic forms in the Homeric speech forms, enclosing all linguistic levels ranging from phonology to syntax and from lexicon to morphology. For the phonological part of the question, a first commonly noted aeolism in the Homeric epics are the sporadic labial outcomes of original labiovelars. The normal rule for Post-Mycenaean Greek is the evolution of these labiovelars to dentals before front vowels, but in Aeolic Greek they become always labial consonants, also in these "dental cases" (cf. Miller 2014: 312). We can think of πελώριον or πέλομαι, whereas an Ionic form τελέθω is also attested in Homer (e.g. Palmer 1962: 97; Wachter 2000: 69). A secure Aeolic trait of the composite language are the assimilations of a liquid with an /s/-sound, whereas Ionic makes use of the first compensatory lengthening. For example, from original *erebes-nós Aeolic has ἐρεβεννόσ vs. Ion. *ἐρεβεινόσ (e.g. Ruijgh 1995: 51ff.)50. A last phonological aeolism, I want to present, is the fact that original /a:/ is sometimes preserved in Homer's text. For instance, the old gen. sg. in -ᾱο is used alongside the Ionic form in -εω (cf. Ruijgh 1995: 52-53). As such, we remark the fact that Aeolic phonological traits can also result in different morphology (cf. the gen. sg. in -ᾱο). There are other morphological peculiarities of the Homeric language which can be accounted for in terms of such different phonological origins. One can think of the Aeolic forms of the personal pronouns in the epic language. Apart from the Ionic form ἡμέεσ (Att. ἡμεῖσ), we encounter the Aeolic form ἄμμεσ (e.g. Ruijgh 1995: 53). As such, both an Ionic and an Aeolic pronominal paradigm could be used by the epic singers, according to the needs of the hexameter (cf. Palmer 1962: 80). The above- mentioned preservation of double sigma can also affect morphological paradigms, e.g. in the sigmatic aorist. In order to create a metrically correct phrase, the aoidos could choose to use an Ionic form ἐπέταςα or an Aeolic one ἐπέταςςα (Ruijgh 1995: 54). Perhaps the most quoted morphological aeolism concerns the athematic infinitive ending -μεν(αι), the short form -μεν is a West-Aeolic trait attested in the Boeotian and Thessalic dialects and the most

48 For thorough discussions of its use and meaning, cf. Monro (1891: 60-62); Chantraine (2013²: 479-484); Bakker (2005: 114-135); De Decker (2016; 2017; 2019). 49 Until now, the existence or non-existence of an Aeolic phase remains an important issue in Homeric scholarship (cf. infra, where other possible models will be presented as well). 50 This is one of the forms which will be discussed as possible artificial forms in van Beek's current project "Unraveling Homer’s Language" (cf. http://www.ru.nl/oikos/anchoring-innovation/related-projects/wp-1- discourse-rhetoric/unraveling-homer-language-linguistic-relationship/).

archaic form (Ruijgh 1995: 55-54). This is reflected in the limited metrical localization of this morpheme in the Homeric verse: it is limited to the position before the bucolic diaeresis (Witte 1913: 2217; Wathelet 1997: 49-54; cf. infra). The longer form -μεναι is an innovation, caused by the conflation of the original form -μεν with the Ionic-Attic ending -ναι. Another commonly mentioned aeolism affects the ending of the dative plural of the third declension, where in Aeolic an ending -εςςι was generalized, e.g. κύνεςςι vs. the more archaic κυςί51. A last Aeolic morphological trait I want to mention, is the frequent use of apocope. For example, the preposition/prefix κατά can be reduced to simple κατ', which is strongly prohibited in Ionic-Attic (e.g. Palmer 1962: 85; 140)52 . Already in 19th century, researchers came to the conclusion that not every non-Ionic form in the Homeric language could be explained as an aeolism53. For instance, Fick, the great exponent of an Aeolic phase, came to the conclusion that some Arcado-Cypriot words, such as ἄναξ or αἶςα were preserved in the Kunstsprache. This caused some difficulties, how could these peripheral dialects be linked with the genesis of the Homeric epics? A sound explanation was first formulated by the great Antoine Meillet, who on the basis of an assumed genetic relationship between the Cypriots and the protohistoric Achaeans, hypothesized an Achaean phase in the genesis of the Homeric epics. It was not until the decipherment of the Linear B tablets that scholarship gained direct access to a Greek variant of the second millennium BC. The first one to dedicate a full monograph to the relationship of these newly deciphered documents with the Homeric language was the Dutch scholar Cornelis Ruijgh (1957; summary in 1995: 63ff.; 1997). He fortified the comments by Fick and Meillet that a genetic relationship existed between the Achaean and Arcado-Cypriot dialects. Furthermore, he was able to offer an extensive list of words which were attested both in Mycenaean Greek, in Arcado-Cypriot and in the Homeric epics (Ruijgh 1957: 89-97; 112- 167). He was even able to point to some structural elements, which were likely to be inherited from an Achaean protohistory of the Greek epic tradition. A first one is the contrastive particle αὐτάρ. The Homeric corpus contains examples of an alternative form ἀτάρ. Based on a comparative study of the formularity of both variants, Ruijgh reached the conclusion that the former is indeed a traditional element in the Homeric corpus and because of its occurrence in Arcado-Cypriot, it is likely to reflect an Achaean heritage (Ruijgh 1957: 29-55). The same methodology of searching for the formulaic character of assumed Achaean influences on the epic language offered other (possible) structural achaeisms: the conjunction ἰδέ, the particle νυ and the guttural aorist in forms such as ἀμβροτάξομεν etc.

51 There are different competing models: It is mostly explained with reference to dat pl. as εὐγενέςςι from which a new morpheme -εςςι was re-analyzed. In addition, there exists another proposal, initially proposed by Wackernagel, based on the equivalence between λύκοι: λύκοιςι and ῥήτορεσ: ῥητόρεςςι. Because in the first declension, the dative plural can be explained as nom. pl. with extra -ςι, this was extended to the third declension, instead of the normal form ῥήτορςι (cf. e.g. Witte 1972: 179). For a study of its relationship to the formularity of the Homeric epics, cf. Wathelet (1997: 52-54). 52 Short summary of aeolisms e.g. Ruijgh (1995: 50-57: 1997: 34-35); Horrocks (1997: 212ff.); Wachter (2000; 68-73); Hackstein (2010: 402). 53 The following survey is based on Ruijgh (1957: 1-7).

(Ruijgh 1957: 55ff.). The word ἀμβροτάξομεν is noteworthy in the discussion of a preservation of vocalic /ṛ/ in Mycenaean and Homeric Greek. It would lead us to far to discuss here the scholarly debate of this concept. To sum it up, there is discussion whether this Indo-European phoneme continued to exist until the Mycenaean period, because of the obscure orthographical rules of Linear B and whether it is preserved in some phrases in Homeric Greek54. Recently, van Beek (2013 passim) suggested that indeed traces of vocalic /ṛ/ could be preserved in epic Greek, although they were no longer used in the spoken language and therefore received a specific vowel colouring when they were eventually vocalized in epic Greek, namely to /-ra-/ or /-ro-/ in labial environments55. A similar problem is the concept of tmesis: the fact that in Homeric Greek the prefix mostly remains separated from the verb itself56. Because this process is not found in Mycenaean Greek, some scholars consider it to be a pre-Mycenaean archaism (e.g. Horrocks 1997: 201), which was inherited from Proto-Indo-European, because it is also attested in Vedic Sanskrit. The fact is that the Linear B corpus is limited to nine compound verbs, so we cannot sure whether tmesis still existed at that time in the spoken language (Haug 2012: 98-99)57. A last mycenaean feature in the Homeric language concerns the old instrumental ending -φι, inherited from PIE *-bhis, which was gradually lost in Ancient Greek, but was preserved in Linear B and the Homeric language. However, in Homer we have to look upon it as a petrified case, because it is also used instead of other noun cases, such as a genitive or a dative (Palmer 1962: 107)58. Ionic, Aeolic and Achaean are the three main dialects which are used in the Homeric Dichtersprache. Furthermore, the last important component of Homeric language is the so- called artificial word forms, which were never used as such in everyday language, but were

54 Before the decipherment of Linear B, this phoneme was generally understood as an aeolism in the diction of the Homeric epics. 55 Van Beek's proposal is based on a thorough study of the scansion of mutae cum liquidae in Homer's text. This can be seen as a result of this inner-epic vocalisation (cf. van Beek 2013: 158-230 for a meticulous analysis of the different forms). For other views: cf. Tichy (1981) interpreting is as a metrical archaism, followed inter alios by Hackstein (2002: 8-9). Hajnal (2003) and Meier-Brügger (2003: 237ff.), on the other hand, express their doubts about the preservation of this phoneme in Homeric Greek. Further discussion, cf. e.g. Janko (1992: 11); Miller (2014: 303-307). 56 The term tmesis (τμῆςισ) is borrowed from the grammatical tradition of Antiquity, but this term can cause chronological misconceptions. Because of its meaning as "separation", it can give the impression that the prefix and verb form are deliberately broken up from one another, but in fact this usage reflects the original function of the prefix as an adverb, which only gradually was fixed to the verb (cf. e.g. Finkelberg 2011 vol. 3: 884 s.v. Tmesis; de Jong 2012: 32). 57 The archaistic character of tmesis can also be verified internally in the epic language. There are 171 forms of ἐν in tmesis in the Homeric corpus, whereas εἰσ or ἐσ is used in tmesis only 13 times. Therefore, the probable chronology is that tmesis is an older phenomenon than the evolution of *ἐνσ to εἰσ. Haug believes tmesis to be a poetical technique, to give the tekst an archaic colouring (Haug 2012: 99-101). 58 It can be a "casus-indifferent form", used for different cases and both for singular and plural. According to Chantraine, it needs to be classified as an adverbial form. Further discussion: cf. e.g. Monro (1891: 104; 154- 158); Meister (1921: 135ff.); Ruijgh (1997: 36); Wachter (2000: 92); Chantraine (2013²: 234ff.); Miller (2014: 294ff.).

created by the epic singers in order to fit the metre of the dactylic hexameter. Some vowels were lengthened in order to fit a word in a hexameter (cf. e.g. Finkelberg 2011 vol. 2: 520 s.v. Metrical lenghtening), e.g. ἀθάνατοσ with its three short vowels could not be used in an hexameter. Therefore, the first /a/ was lengthened to /a:/. This technique became also utilized for other vowels, e.g. δολιχοδείρων became δουλιχοδείρων. Another phonological artificial construction concerns the so-called epic diectasis ("distension"), the fact that contracted verbs were again decontracted due to the limits of epic metre, by doubling the vowel quality which emerged after the contraction, e.g. original *ὁράω > ὁρῶ (as in Attic) > ὁρόω (e.g. Witte 1913: 2223-2224)59. Also some morphological forms were formed on the basis of analogy, e.g. on the basis of the genitive ἡνιόχοιο (scanning –∪∪–∪) an accusative ἡνιοχῆα was formed as if from a nominative *ἡνιοχεύσ, because the normal accusative ἡνίοχον would not fit at the same place in the verse (scanning –∪∪∪) (e.g. Wachter 2000: 82- 83; Hackstein 2010: 412)60. To name some syntactic examples, we can think of the choice between a singular or a plural form or an active versus a middle form according to the needs of the hexameter61.

1.3 The language of the Homeric poems: A diachronic approach. The preceding chapter offered a short overview of the composite nature of the Homeric language, which could partially be explained as a result of the overwhelming influence of the hexameter on the linguistic forms of Greek epic. Further explanations are necessary to draw a plausible picture of how different dialectal influences could be mixed in the epic language. How is it possible that the main colouring of the text is Ionic, but that a large amount of Aeolic forms is equally observable in the text? Firstly, one could think of a poet who lived in a region where a kind of mixed dialect was spoken at the time. This was inter alia the opinion of the great Altertumswissenschaftler von Willamowitz-Moellendorf, who pointed to the city of Smyrna, which is situated on the border between the Ionic and Aeolic dialect regions in Asia Minor. In that case, the traditional claim of this city to be the home city of the great poet could be corroborated.

59 Further discussion: Palmer (1962: 95); Horrocks (1997: 208); Wachter (2000: 82); Finkelberg (2011 vol. 1: 204f. s.v. Diectasis); Chantraine (2013²: 75-83); Miller (2014: 93). It was not discovered by Witte; already Wackernagel and other 19th century researchers were able to explain these forms. In the last decades the Achaean substrate of Homeric Greek seems to be minimized, already so by Palmer (1962: 75): "The impact of the new evidence [sc. Linear B] on Homeric studies has been much exaggerated". However, as I will discuss further below, there are strong indications that there existed an epic tradition in Mycenaean Greek (cf. e.g. Hackstein 2010: 403). 60 Cf. also Monro (1891: 91), but without explaining the form; Chantraine (2013²: 95). Similarly, a verb form ἡνιοχεύω (–∪∪–∪) was created instead of the normal vorm ἡνιοχέω. For a discussion, cf. Chantraine (2013²: 368). 61 Discussion of poetic plurals in Witte (1972: 1-8), morphological analogies due to metrical necessity, cf. Witte (1972: 34-41), active vs. middle forms (Witte 1972: 72-76).

However, it seems very implausible that a natural language could maintain such a grammatical variation on a diachronic and a diatopic level (cf. Palmer 1962: 98). There exist indeed so-called mixed languages, e.g. Cappadocian, but such a situation can only emerge after centuries of intensive language contact and bilingualism. This seems far-fetched for the contacts between Ionic and Aeolic in the eight century BC. Moreover, inscriptional evidence does not corroborate this hypothesis. Therefore, this possibility can be excluded. The most commonly accepted hypothesis is that of an Aeolic phase in the genesis of the Homeric epics, which from Antoine Meillet onwards, and certainly on the basis of the decipherment of the Linear B tablets, is extended with the postulation of an Achaean phase62. How do we have to understand such different phases in the development of the Greek epic tradition? This would mean that, due to the growing cultural and economic importance of Mycenae and other cities in Bronze Age Greece, the oral tradition which was inherited from Indo-European times could flourish and give birth to a native epic tradition, in order to glorify the deeds of the Greek aristocracy63. After the decay of the Mycenaean culture, this oral tradition continued during the notorious Dark Ages. Because of the West-Aeolic forms in the Homeric language, such as the already quoted infinitive form in -μεν or the lack of assibilation in some words, adherents of the phase model, are convinced that the oral tradition gradually extended to northern regions, to Boeotia and Thessaly, areas where West Aeolic was spoken at the time (11th - 10th century BC)64. In order to explain how such an epic tradition eventually reached the Ionic cities in Asia Minor, more migrations need to have occurred. This has to be conceptualized in the context of the colonization of West Asia Minor, both by Aeolic and Ionic migrants. To focus on the Aeolic ones, migration existed from Thessaly to the island of Lesbos and the neighbouring parts of the Anatolian mainland. In this new Heimat, the oral tradition could flourish and incorporate some East Aeolic forms, such as the infinitive in -μεναι, which is usually explained as a contamination of an Aeolic and an Ionic form. Due to such contacts, the Ionian population of the more southern parts of Asia Minor could borrow the oral tradition and adapt it to their own dialect. The most important linguistic arguments in favour of the phase model for the history of Greek epic were given above. In addition, this hypothesis is corroborated by some

62 Adherents of the phase model (alphabetically), inter alios the Norwegian school of Berg and Haug; Chantraine (various publications, especially 2013²: 495-513); the Italian school of Gentili and Giannini; Hackstein (2002; 2010 pointing to the fact that there remains some discussion (2010: 402)); Hoekstra (1965; 1981); Janko (1982; 1992: 15-19; 2012); Wathelet (1997); Janse (1998; 2012); Kirk (various publications); Meier-Brügger (2003); Meister (1921); Nasta (1994-1995; especially 1994: 116-117); Palmer (1962); Ruijgh (various publications); Tichy (1981; 2010); Tsopanakis (1981); West (various publications); Witte (1913; 1972). A good survey is provided in Finkelberg (2011 vol. 1: 9-10 s.v. Aeolic phase) 63 Cf. supra: there is some discussion about the importance of a Mycenaean phase, e.g. Andersen & Haug (2012: 9) believe that the forms listed by Ruijgh (1957) are simple archaisms. Extensive criticism against Mycenaean phase: see Haug (2002: 41-69). 64 Interesting in this case is the attestation of a pair of names Ὀρτίλοχοσ and Ὀρςίλοχοσ, the first one being the name of the grandfather and the latter the name of the grandson, so one could guess that the sound change occurred within these two generations (personal communication with prof. dr. Mark Janse). For a discussion of their persons, cf. Finkelberg (2011 vol. 2: 613 s.v. Orsilochus & Ortilochus).

arguments on a cultural and literary level65. I already stressed the relationship that could have existed between this phase model and the migrations due to changes in the political and economic situation of Bronze Age Greece. Moreover, this can also be observed in the importance of some protagonists in the Iliad. Let me bring into focus the two main characters of the Iliad: Agamemnon and Achilles. The first one could be the ideal reflection of an Achaean phase in the development of Greek epic. He was the king of Mycenae and large parts of the Peloponnese and therefore it seems very likely that there could have existed an epic tradition during Mycenaean times, to glorify the deeds of such great kings. The (possibly) historical fact of a Trojan war would certainly contribute to the flourishing of such a tradition, which would emerge in a period when also other important myths were created, e.g. the Argonauts. Achilles on the other hand, does not have a relationship with the old Mycenaean world. In the epic world, he is the king of Phthia, a region in Thessaly. Therefore, it would make sense that Achilles is a later addition to the mythological story, in order to stress the importance of Thessaly as the new home land of the oral tradition. This is corroborated by the fact that Achilles plays a crucial role at the end of the story of the Trojan war. He is not one of the military commanders who depart from Aulis at the beginning of the war. Moreover, his function in the epic suggests Oriental influence. His bromance with Patroclus for instance, could be built on the model of the friendship between Gilgameš and Enkiddu in the Gilgameš-epic, which was also translated into Hittite and could therefore have influenced the story of Achilles66. Probably, this happened during the East Aeolic phase of the epic tradition, when contacts existed between Greek and Anatolian populations. This East Aeolic phase can also account for the importance of the Trojan war in the epic tradition. Because of the vicinity of Troy, the Greek victory over the Trojans could be interpreted as a justification for the colonization of western Asia Minor. The island of Lesbos would be an ideal place for doing so, as Janko (1982: 90) points out67. During the eighties of the previous century some criticism arosed against the phase model, explaining the Aeolic forms instead as borrowings from a neighbouring tradition in Asia Minor, possibly influenced by the manifest interest in contact linguistics from that time onwards (cf. Horrocks 1997: 214). This diffusionist model came into being in the writings of Miller (1982; 2014) and Horrocks (1987; 1997), and was recently also vindicated by van

65 A good survey is provided by West (2011: 38ff.): Mycenaean elements are e.g. also corroborated by the reference in Hittite cuneiform texts of a city Wiluša and a country Aḫḫiawa or by names as Alaksandus (Paris is named Alexandros) (cf. Finkelberg 2011 vol. 1: 18-19 s.v. Ahhiawa). He further offers a stage model about the intrusion of the figure of Achilles in the epic tradition (cf. infra). More concise: West (1973a: 189-192; 1988: 161-162). Haug (2002: 151-152) underlines also the importance of Philoctetes as an indication in favour of the Aeolic hypothesis. Miller (1982: 9ff.) also offers a good overview, but argues instead for a diffusionist model (cf. infra). 66 For a discussion, cf. e.g. Miller (1982: 16-21); Dowden (2004: 192-193); Finkelberg (2011 vol. 1: 313-314 s.v. Gilgāmesh; vol. 2: 559-562 s.v. Near East and Homer); West (2014: 31). 67 Moreover, this could explain the rarity of the dative in -οισ, because it could be confused with the accusative plural in -οισ of the Lesbian dialect. He further points to the importance of Troy in the Aeolic region and notes that Agamemnon can also be linked with Cyme and Mytilene (Janko 1982: 89-91). Cf. also West (1988: 163- 164).

Beek (2013)68. According to this point of view, the Ionic tradition is interpreted as a direct descendant from the (South) Mycenaean tradition, which during the Dark Ages headed gradually for Asia Minor. They are supported in this case, by Kirk's (1962; 1965; 1976) point of view that the Dark Ages were a good breeding ground for the continuation and flourishing of the oral tradition. Linguistically, they establish their theory on diachronically distinct Ionic forms in the text. Let us consider the following example (Horrocks 1997: 215).

*νᾱϝ-όσ > νη-όσ (e.g. Il. I, 476) > νε-όσ (e.g. Il. XV, 423)

The first form represents the common Greek form, which was preserved as such in the Aeolic dialects, but which disappeared from Ionic after the vowel shift /a:/ > /ε:/. This Ionic form is found in the second example, which after a shortening of the /ε:/ became short /e/. In the diffusionist hypothesis, this sequence is explained as follows: the lack of attestation of the first example denies the existence of an Aeolic phase in the composition of the Homeric epics: why is this form not attested? Furthermore, because two different Ionic forms are found, this points in their opinion to a longer Ionic phase than generally assumed. However, the theory contains some shortcomings. Firstly, because the form νεόσ is a relatively late Ionic form, which gradually came into being due to analogy with the later form νεῶν, where the shortening is caused by a regular sound change69. Furthermore, this genitive is only attested three times in the Iliad, always in book XV (423; 693; 704), two times after the trochaic caesura (cf. infra; 423; 693) and once after the penthemimeral caesura (704). It seems possible that this form is influenced by the genitive plural νεῶν which is also used after the trochaic caesura only some verses after the first instance of the singular form: e.g. τεϑχεα ςυλόςωςι ||3b νεῶν ἐν ἀγῶνι πεςϐντα ("They will stripp him of his arms, fallen in the battle about the ships"; Il. XV, 428). Furthermore, due to the lateness of the form (it is the normal form in Herodotus), it could be a later intrusion in the text. Apart from this, a serious flaw remains in the diffusionist model, which was first observed by Meister (1921: 146-171) namely the fact that no Ionic genitives in -ηο and -ηων are found in the Homeric corpus, although the older Aeolic forms in -ᾱο and -ᾱων and the later Ionic forms with quantitative metathesis and synizesis (cf. infra) in -εω or -εων are both attested70. No convincing explanation of this distribution is provided by the adherents of the

68 The following survey is based on Horrocks (1997: 214-217). A similar overview can be found in Berg & Haug (2000: 16-18), but arguing the other way around. Jones (2012: 46) summarizes the theories with a graphic representation. Adherents of this model: Horrocks (1997); Jones (2012); Miller (1982; 2014: 336ff.). Crespo (1997: 135) seems to be in dubio, for he declares that a borrowing model could be possible. Van Beek (2013) also shows adherence to this theory (cf. infra). I will confine myself to a short overview of this theory. Many points of discussion remain, which would lead us too far. 69 There are some problems with Meister's assumption (1921: 146) that the shortening was also normal before a short velar vowel. In this case, more examples were to be expected. 70 According to Nagy (2014: 146): "Such a borrowing from Aeolic morphology into Ionic morphology could be truly systematic only if formulae involving genitives in -ηο and -όων had already existed in Ionic Greek, which could then be replaced by the genitives -α ο and -α ́ ων of Aeolic Greek to fill the metrical frames that could no

diffusionist model (cf. Haug 2002: 153; Janko 2012: 40)71. Furthermore, there are other shortcomings with the theory. Firstly, scholars such as van Beek (2013: 327) cast doubt on the plausibility of a poetic tradition which pass from one dialect into the other. Even in the diffusionist model, you need to explain how West Aeolic forms can be attested in the epic language. In addition, some superficial aeolisms could have intruded due to borrowings, but this seems less plausible for endings like the genitives with <ᾱ> /a:/ (cf. Janko 1982: 89- 91)72. In this case, a shift from West to East Aeolic needs to have taken place as well. Furthermore, his criticism that no traces of an Aeolic tradition have come down to us, is by no means an argument against the Aeolic phase, because the diffusionist model also needs to postulate the existence of an Aeolic phase, for otherwise no borrowings could have occurred. Van Beek's proposal that the epic language is a variety of its own, is surely a valuable contribution to our understanding of its artificial character, but it seems far-fetched to try to explain everything as Mycenaean, Ionic or artificial constructions. The Aeolic elements are too prominent to argue for it in a convincing manner. An interesting hypothesis is put forward by Nagy (2014)73, who refers to the linguistic concept of a Sprachbund, a linguistic area as it was defined by Jakobson (1931)74. In his opinion, we have to differentiate between obligatory aeolisms which came into being due to an Aeolic phase of the epic and optional aeolisms which can be explained as borrowings from an Ionic phase, which derived from Mycenaean times and came into contact with the Aeolic tradition in Asia Minor. As such, he attempts to explain the presence of Ionic as the dominant dialectal component, Aeolic as the recessive component and Mycenaean as the residual dialectal component (Nagy 2014: 148-149). In fact, this theory can overcome some of the problems inherent in the two hypotheses. For instance, it can longer be filled by -ηο and -όων at the time when quantitative metathesis became a phonological rule in Ionic. I say it this way because the borrowing here simply cannot be a matter of phonemics only. It is a matter of morphophonemics." (cf. infra for a discussion of Nagy's argument). A similar problem concerns the fact that the form ποτί could not have been the older form of πρόσ in the Homeric language, without the assumption of an Aeolic phase, or ποτί needs to be postulated as an archaic relic in Lesbian, which cannot be verified (Janko 2012: 40). 71 Jones (2012: 48ff.) denies the problem because there are other examples in the Homeric corpus with the sequence -ηο, as ἔκηα or αἰζηόσ (criticism: cf. Haug 2002: 157). Furthermore, the athematic duals as ςυλήτην, which are normally explained as Aeolic forms are not a valid argument in his opinion. If they were borrowed, we would expect *ςυλα ́ την. This is a problematic interpretation, because it seems unlikely that Ionic singers borrowed forms without altering anything. Because of the existence of normal Ionic verbal forms with /e:/ as e.g. ἐςύληςαν, they can be adapted to look more as an Ionic form. Such processes can also be observed in contact linguistics: forms are borrowed but adapted to the recipient language. Furthermore, the absence of the dual in Ionic renders a borrowing from Aeolic plausible in this case. Miller (1982: 57-69) points to phonic patterns with the Ionic vowel /e:/. 72 A similar problem is the lack of assibilation in some words (cf. supra). Miller (1982: 139ff.) interprets this as a mere archaism in the epic tradition, but this is impossible, because Mycenaean already assibilated. Only contacts with a West Aeolic (or theoretically West Greek) tradition can account for such forms. 73 Wachter (2000; 2012) shares some similar points with Nagy. 74 Stricto sensu, the term Sprachbund does not exactly cover Nagy's interpretation, because it refers in fact to a long term contact situation between three or more dialects or languages, as e.g. the Balkan Sprachbund.

explain every dialectal residu in the Homeric poems or the sequences of diachronically different Ionic forms. Moreover, it seems plausible that the oral tradition was far more scattered during the Dark Ages between different dialects. The assertion that in period X only dialect Y exploited the oral tradition, won't wash. The protohistory of the Greek epic remains speculative and needs further research in the future decades. Doing so, more complex theories need to be developed. It looks probable that oral traditions existed all over Greece, which influenced one another, both in a thematic, a linguistic and a metrical way and at different periods75. For example, it seems possible that a West-Greek tradition existed as well which could explain the few Doric forms in the Homeric poems (cf. West 1988: 167; Tsopanakis 1981: 66). If this would be true, then the Greek poetic tradition would indeed be a Sprachbund in the narrow sense of the word.

75 The problem of Euboean forms could also be explained with reference to such a multidialectal model (cf. West (1973a: 189; 1988: 166ff.; 1992; 2014: 90-91) and Chadwick (1990). Ruijgh (1995: 49f.) accepts West's hypothesis: it is possible that Homer encountered a rich person on the island to bestow him with papyrus to write down his work.

CHAPTER 2: A COGNITIVE APPROACH TO GREEK METRE

2.1 Some preliminary facts about Greek epic metre The preceding chapter stressed the importance of the oral tradition and its influence on the composite language of the Homeric epics. Two models to explain this language were discussed at some length, namely a phase model with different diatopic and diachronic developments and a diffusion model, explaining the Aeolic forms in the Homeric poems as borrowings, due to cultural contacts between Ionians and Aeolians in Asia Minor. Furthermore, it was already highlighted that the hexameter played an important role in the realization of an artificial epic language. The present chapter will deal extensively with this metrical aspect of the Homeric epics. Firstly, the basic facts about prosody will be discussed76. Afterwards, a short survey will be presented about the main theories which are used in secondary literature with regard to the caesurae of the hexameter. They will be criticized and instead I will propose, following Janse (1998; 2012), that we need to use a cognitive approach towards the colometry of the Greek hexameter. Therefore, the concept of intonation units, initiated by Chafe (1982; 1985; 1987; 1994), will be shortly discussed and applied to the Homeric language. In order to illustrate the usefulness of this approach, some verses of the corpus will be analyzed using the different colometries, to compare their applicability. Every extant Ancient Greek epic fragment uses the same basic metre, the dactylic hexameter, consisting of six feet of dactyls (–∪∪), which can be substituted with a spondee (––). The last foot is constructed either with a spondee, either with a trochee (–∪). This last option can be interpreted as a catalectic variant of a dactyl77. Schematically, this results in the following basic structure of an hexameter:

–∪∪ | –∪∪ | –∪∪ | –∪∪ | –∪∪ |–⨯

"–" is used to refer to a long syllable, "∪" to a short one. The sequence "∪∪" means that an original dactyl can be substituted with a spondee, a process coined "biceps" by Maas (1962:

76 I will make use of the standard metrical handbooks by Koster (1936), Maas (1962), Dain (1965), Korzeniewski (1968), West (1982; 1987) and Sicking (1993). Good surveys of the prosodical rules and their application to the Greek hexameter are offered by Bowra (1962a); Ruijgh (1995: 7-13), West (1997); Nünlist (2000), Graziosi & Haubold (2010: 10-13) and de Jong (2012: 33-35). Janse (2012) is an introductory syllabus for Homeric metre (based on Janse 1998; cf. 1998: 128-135 for basic facts about Homeric prosody). The basic references in the text will be to West (1982). Some other references will be quoted in footnotes, but they will be limited to the above mentioned sources. 77 Catalexis is "the truncation of colon- or metron-end in comparison with others" (quoted from the glossary in West (1982: 192). Cf. also Koster (1936: 17); West (1987: 5); Sicking (1993: 51).

7)78. This is also possible in the fifth foot, but its instances are so rare (3%) that it needs to be interpreted as an exception to the normal rule (cf. e.g. West 1982: 37; Sicking 1993: 73- 74). Such a spondaic verse can give more weight to the emotional or tragic character of a verse. The fact that they are rather uncommon, is also determined by the fact that the recurring adonean ending (–∪∪–⨯) reminds the listener that we are at the end of the verse (Sicking 1993: 69). The symbol "⨯" refers to the syllaba anceps at the end of the line, which can be realized both with a short or a long syllable. The first syllable of a foot is called "princeps" by West (1982), and traditionally it is accentuated in schematic structures of the hexameter. This accent is called the ictus, a kind of stress beginning every foot. In Greek, it was called the ςημαςία (Sicking 1993: 10). This ictus is a later interpretation, resulting in an inaccurate conception about Ancient Greek poetry, according to which every foot starts with a stressed syllable. It is more likely that this ictus was, as I said, a light pulse, clearly secondary to the Greek pitch accent or possibly did not exist at all (cf. infra about colometry). Therefore, I did not add this accent on my schematic representation (agreeing e.g. with Nünlist 2000: 111). Moreover, the two parts of a foot are commonly denoted as the thesis and arsis. However, there is no agreement if thesis refers to the first or the second part and vice versa79. According to the first view, thesis is interpreted as the placing down (τίθημι) of the first long syllable, which is lifted up (αἴρω) in the second part of the foot. According to the second view, arsis is the first part of the foot, because the first part is stressed, but in the second part laid down again. In order not to confuse the reader, I will avoid these terms alltogether, which were already ambiguous in Antiquity. Instead, I use the numerical system of Janse (2003), as mentioned above. Caesurae will be discussed in the following section. The prosodic rules of Greek metre are inherited from the Proto-Indo-European age, because they are the same in some other ancient Indo-European languages, such as Sanskrit. West (1982: 8) defines it as follows:

"A syllable is long if it is "closed" (i.e. ends with a consonant), or if it contains a long vowel or diphthong. Otherwise it is short."80

Four remarks need to be made. Firstly, it has to be apparent that the basic distinction is made between long and short syllables, not between long and short vowels. A syllable containing a short vowel is closed, hence measured long if it ends in at least one consonant. Secondly, a Greek verse has to be considered a continuüm, to such a degree that word ends are in itself not important to determine whether a syllable is short or long. For instance, the sequence τι ́ςειαν Δαναοί (Il. I, 42) has to be scanned as ––|–∪∪|–, because the syllable /an/

78 Maas (1962) used it also for a resolvable princeps in other metres than the hexameter, whereas West (1982 passim) restricts it to the substitution procedure mentioned above. 79 For example, Koster (1936: 25) uses arsis for the first and thesis for the second part, whereas Tsopanakis (1982 passim) uses thesis for the first and arsis for the second part. 80 Other formulations, e.g. Koster (1936: 27); Maas (1962: 75); Dain (1965: 18); Korzeniewski (1968: 20); West (1987: 12; 1997: 219); Nünlist (2000: 109).

is closed by the /d/ of the following word Δαναοί. Third, we need to remember that the graphemes <ξ> /ks/, <ψ> /ps/, <ζ> /dz/, count for two consonants and therefore close a preceding syllable (West 1982: 8). This does not affect the aspirated consonants <φ> /ph/, <θ> /th/, <χ> /kh/, because they represent only one phoneme81. Fourth, the normal euphonic rules of the Greek language are also applied in verse, therefore, when two vowels meet each other, the first one is normally elided. For instance, original ἠρᾶτο ὁ γεραιόσ ("the old man prayed") becomes ἠρᾶθ᾽ ὁ γεραιόσ, with aspiration of the voiceless dental stop /t/ to /th/ due to the following aspirate (Il. I, 35). Because the epic verse was somewhat rigid for later aoidoi, some licences were created in order to facilitate the construction of thousands of verses by the epic poets. Some of them are metrical anomalies, others have to do with rapid speech phenomena. Following West (1982: 10-18), I will first focus on the licences with regard to vowels and afterwards on the freedoms which were created concerning consonants. In the first place, we can think of the so-called epic correption (West 1982: 11-12)82. This process offers the possibility for long vowels, diphthongs and triphthongs (<εῃ> /eε:i/, <εῳ> /eo:i/), to be shortened before a vowel in the following word. It is called epic correption, because of its preponderance in the Greek epic language. For example, we can cite ἔπει ὤνηςασ (∪∪|––|–) (Il. I, 395), where the diphthong /e:i/ becomes a short syllable, due to the following /o:/. The next licence I want to present, concerns synizesis (West 1982: 12; 1987: 14; 1997: 220) uses the term "synecphonesis"). This means that "two or more vowels are slurred together to make one long syllable" (West 1982: 12)83. In most cases, an <ε> /e/, is pronounced as a glide /j/ before the following vowel, as for instance in the above quoted example. Occasionally, it can also be found with other vowels (cf. West 1982: 12 for examples). However, we need to be cautious not to muddle it with the consonantalization of the vowels <ι> /i/ and <υ> /u/ (West 1982: 14)84. In order to fit a word in the dactylic rhythm, mostly personal and geographical names, they can be pronounced as semivowels. For example, in the Catalogue (Il. II, 537) Ἱςτίαιαν needs to be read with a semivowel /j/. This process can also have had its influence on the epic correption. Greek diphthongs end in the semivowels /j/ or /w/ and therefore, correptions of the kind ἔπει ὤνηςασ can also be explained as a consonantalization of original /i/ to /j/ which is pronounced with the following syllable and therefore, does not close the preceding one. A last "vocalic licence" I want to introduce, is the so-called "hiatus" (e.g. Maas 1962: 89; West 1982: 14-15; 1987: 15-16). In fact, this is the opposite of the normal elision in Greek prosody and refers to the possibility that two vowels meet, but are not elided in the text. For example, in Il. I, 7 ἈτρεϏδησ τε ἄναξ ἀνδρῶν καὶ δῖοσ Ἀχιλλεϑσ

81 Cf. West (1987: 12); Sicking (1993: 62); Nünlist (2000: 109). 82 Cf. Korzeniewski (1968: 25); West (1987: 14); Sicking (1993: 65); West (1997: 220); Nünlist (2000: 110 s.v. "Hiatkürzung"); Finkelberg (2011 vol. 1: 180 s.v. Correption); Chantraine (2013²: 88-89). 83 Cf. Monro (1891: 55; 378); Korzeniewski (1968: 25); West (1987: 14-15; 1997: 220); Nünlist (2000: 110); Finkelberg (2011 vol. 3: 835 s.v. Synizesis); Chantraine (2013²: 84-85). 84 See West (1987: 15; 1997: 220).

("The son of Atreus, lord of men and the divine Achilles"), there is no elision between the <ε> /e/ of τε and the following <α> /a/. To explain why hiatus can exist in the epic metre, we need to proceed to the "consonantal licences". In fact, different factors can account for the presence of hiatus in the text, but one of the most prevailing reasons is the loss of an original consonant, especially the digamma <ϝ> /w/, a consonant, which, as I explained in the first chapter, was inherited from PIE, but was gradually lost in the different Greek dialects, quite early in the Ionic dialect (e.g. West 1982: 15-16). Wachter (2012: 70) uses the appropriate term "ghost digamma"85. This fact can explain the lack of elision in the sequence τε ἄναξ, because ἄναξ is a word which originally began with the semivowel /w/, as is also proved by the existence of Mycenaean wanaks (Linear B: ). The loss of original consonants can not only cause the existence of hiatus, but also a so-called brevis in longo, the fact that a syllable needs to be counted as a long one, although in the surviving text of the Homeric epics, it is only followed by one consonant. With regard to digamma, I can quote from the first book of the Iliad (I, 33) ἔδειςεν δ᾽ ὁ γϋρων ("The old man feared"), where the first syllable needs to be scanned as a long one because the original form was *ἔδϝειςεν, with a digamma after the <δ> /d/86. Other anomalies concern the fricative /s/, which not only disappeared at the beginning of a word before a vowel or between two vowels, but also near a liquid or a nasal. As we observed in the previous chapter, here a distinction between Ionic and Aeolic needs to be made. In Aeolic, we encounter an assimilation of the original /s/ with the following liquid /l, r/ or nasal /m, n/, as in the Aeolic infinitive ἔμμεναι (from *ἔςμεναι), as such always closing the preceding syllable. In Ionic on the other hand, this /s/ disappeared, usually with compensatory lengthening. However, it can leave traces in the prosody of the Homeric hexameter (West 1982: 16)87, to such a degree that epic singers generalized the rule that it is possible to lengthen a preceding syllable before a liquid, even in instances where etymologically no /s/ needs to be reconstructed, e.g. Il. I, 283 reads Ἀχιλλῆώ μεθϋμεν (∪|––|– ∪∪|–), with a long /ī/ due to following /m/, although the prefix μετά does not derive from *ςμετά88. However, it needs to be underlined that these "rules" are by no means always followed. Not every original digamma leaves his trace in the prosody of Homeric verse, nor do the other lost consonants. For instance, in Il. I, 232 ὕςτατα λωβόςαιο (–∪∪|––|–∪), the final /a/ of ὕςτατα remains short although λωβέω derives from original *ςλωβέω (cf. Lithuanian slėgti "oppress") (Chantraine 1999²: 653). An original /s/ at the beginning of a word can also cause metrical anomalies, e.g. Il. I, 51 βϋλοσ ἐχεπευκὲσ (∪|–∪∪|–∪) with lengthening of /os/ because ἐχε(πευκὲσ) derives from *ςεχε- > *ἑχε- (s>h /#_V) > ἐχε- (with Grasmann’s law) (cf. Ruijgh 1997: 39). Another example is to be observed in Il. I, 226

85 For a thorough discussion: cf. Monro (1891: 361-383); Finkelberg (2011 vol. 1: 205 s.v. Digamma); Chantraine (2013²: 116-157). 86 Other editions than the OCT (except for Leaf) write ἔδειςεν with <δδ> on the basis of some papyri, to account for the metrical lengthening. This implies that /w/ assimilated to /d/, which is not the case. 87 Cf. Bowra (1962a: 24); Maas (1962: 80-81); West (1987: 17; 1997: 221; 228-229); Nünlist (2000: 110). 88 Cf. the fact that is a possible cognate of Dutch "met" or German "mit", without an original sibilant at the beginning of the word (Chantraine 1999²: 689-690).

πϐλεμον ἅμα (∪∪|–∪∪), where /on/ is lengthened before ἅμα < *ςάμα (cf. Skr. sama, Dutch samen < PIE *sṃ-)89. For an example with hiatus, reference can be made to Il. I, 532 εἰσ ἅλα ἆλτο ("she [sc. Thetis] jumped into the sea" from original * ἐνσ ςάλα ςᾶλτο), where the final /a/ of ἅλα is not elided (cf. Ruijgh 1997: 40). Apart from the digamma, the semivowel /j/ also disappeared in Greek, in Anlaut resulting either in /dz/ or in /h/. This last option is relevant for the prosody of Homeric verse, because this one can cause hiatus or brevis in longo as well. An example of hiatus is furnished by κατὰ φρένα ὡσ Ἀχιλῆα ("in his mind how Achilles") (Il. II, 3) where the /a/ of φρένα is not elided because ὡσ derives from the PIE relative stem *yo (cf. Skr. relative yaḥ)90. Mette (1956: 5) also suggests the possibility of the loss of an original consonant in monosyllabic words, e.g. hiatus after τί (coming from *τίδ, cf. Lat. quid). This possibility is mostly not mentioned by other metricians, although this seems to be feasible, because we cannot be sure whether final consonants were already dropped in Mycenaean, as final consonants are not written in Linear B. As I already put forward in the preceding chapter, there are other anomalies in the Homeric language which we can explain with reference to Mycenaean. Firstly, we can think of the original dative ending -ει which was later on monophthongized to long -ῑ, e.g. the original *Διϝεί is still felt in the formula Διι ̀ μῆτιν ἀτάλαντον ("equal in intelligence to Zeus"; e.g. Il. II, 169)91. Secondly, let us not forget the problem of vocalic /ṛ/, which we will encounter ad nauseam. In addition, I already pointed to the existence of epic diectasis and metrical lengthening in the previous chapter (cf. also West 1982: 38). Apart from that, there are also metrical anomalies which defy an historical explanation. Some of them can be caused by a following caesura (cf. infra) or due to the combination of different formulae, but there are also examples which we cannot explain up to now on historical grounds (West 1982: 38)92. A last couple of metrical licencies needs to be mentioned here: Attic correption or correptio attica and the lack of metrical lengthening before some consonant clusters. Let us start with the former. We already encountered epic correption, but Attic correption, is

89 The fact that it is placed before the penthemimeral caesura (3a), will also have played a role. Such examples are an argument for a post-Mycenaean date for Grassmann's law (cf. De Decker 2015: 9-12). 90 The anomaly can also be explained by the fact that it is placed before the bucolic diaeresis. 91 Mῆτιν also ends in a long syllable, whereas the /i/ is normally short. This was probably caused by the heavy aspiration in the following word ἀτάλαντοσ < *hατάλαντοσ (from PIE. *sṃ-) during the Mycenaean period (Tichy 2010: 62 fn. 106 denies this explanation but is not able to offer a better one). Cf. Ruijgh (1995: 77f.: 1997: 39) for discussion. A similar example is provided by Il. I, 86 Διι ̀ φύλον from original *Διϝεὶ φύλον. 92 Tsopanakis (1981) is a monograph study about metrical anomalies in the Homeric verse. He uses the terms ChL (chasmodic long) when there is no epic correption, ChSh (chasmodic short) when a short vowel is not elided and PsL (pseudo-long) for a brevis in longo. He believes that historical grammar and formulae are unable to explain all the instances and therefore investigated, the anomalies in the lost syllables of the words in Homer. Doing so, he collected a very valuable collection with all the instances (Tsopanakis 1981: 121-149; 172- 174; 201-210; 215-219; 228-240). However, he is also not able to explain every instance. Furthermore, he lacks a good knowledge of historical grammar. McLennan (1974) stresses the fact that some words are more prone to offer (inexplicable) hiatus. He gives a list who points to some tendencies (McLennan 1974: 134). Furthermore, he criticizes Tsopanakis, who wanted to refer to correlatives as more prone to hiatus.

named so, because of its preponderance in Attic drama. It refers to the fact that it is possible, but not obligatory, before the combination of a stop and a liquid or a nasal (muta cum liquida) that the preceding syllable remains short (West 1982: 16f.). For example, in Il. I, 113 ῥα Κλυταιμνόςτρησ (∪∪|––|–) the /a/ in ῥα is not lengthened because of the following cluster comprising unvoiced velar stop /k/ plus liquid /l/93. A similar procedure is used in some clusters, which would cause unmetrical verses, mostly concerning personal and geographical names, especially <ςκ> /sk/ and <ζ> /dz/ (West 1982: 17)94. For example, the Trojan river Σκάμανδροσ and the corresponding adjective Σκαμάνδριοσ could not have been used by Homer, if he was not permitted to use this metrical licence (e.g. in Il. II, 465).

2.2 How to put a caesura? - The traditional theories The preceding section offered a survey of the main peculiarities of the Homeric hexameter. One important aspect was left out because it deserves to be treated separately and comprehensively in this section: the problem of the hexametrical caesurae. As Bassett (1919: 344) remarks: "The hexameter is too long for one syntactic unit"95. Like any verse, the hexameter is divided, in his opinion, in different parts of 3+3 or 2+2+2 (for discussion cf. infra). The points at which these different structural segments of the hexameter are divided, are normally called caesurae, a Latin word which is a literal translation of the original Greek word τομή. It has to be stressed right away that the first Greek definition goes back to late Antiquity, in third century AD, in the writings of Aristides Quintilianus (Bassett 1919: 348)96. Originally, the word τομή did not refer to the place of cutting (cf. τέμνω/caedere) the verse in different units, but to the different parts itselves, which are generally called "cola". Gradually, due to the etymological meaning of the word, τομή became to denote the place of cutting (cf. e.g. Janse 2012: 18). Formally, a caesura is terminologically distinguished in metrical theory from the similar diaeresis. Caesurae refer to divisions within a foot whereas diaereses refer to separations between different feet of the hexameter. The problem is how to define a caesura. This difficulty became already apparent in the writings of Gottfried Hermann, the founding father of the modern study of Ancient Greek metre in the early 19th century, for he gives three different meanings of this problematic term: 1) the places where a word ends within a verse, 2) where a rhythmical phrase ends with a word and 3) where completion of thought appears in the verse (cited inter alios in

93 Explanation also Dain (1965: 18); Korzeniewski (1968: 21); West (1987: 18); Sicking (1993: 63); West (1997: 220-221). 94 Cf. Maas (1962: 78); West (1987: 18; 1997: 221); Nünlist (2000: 110). 95 Sicking (1993: 52) points to the cognitive problem that "the span of immediate memory" cannot grasp a whole hexameter at once. The hexameter needs to be divided in "übersehbare Teilsequenzen" (cf. Bakker 1997a: 140; 148). 96 This is a fundamental problem with ancient metrical theory. They represent only a later interpretation, centuries after the composition of the Homeric epics. They have no idea about the original oral performance of these epics. Therefore, I will not consider ancient metricians in my further discussion of Greek metre and its protohistory. A very useful survey of ancient metrical theory is given by Bassett (1919).

Bassett 1919: 345). In this section I will present the most important modern "colometries" of Homeric verse, the purely metrical caesura, which is put forward inter alios by West (1982; 1987; 1997), the rhetorical caesura of Fraenkel (1968³) and the dynamic caesura of Kirk (1985)97. They will be introduced and afterwards, in the next section, be criticized from a cognitive point of view. Martin West (1937-2015) was not only one of the most influential Hellenists of his time, but also the authority in Anglo-Saxon metrical studies. He was not only asked to publish a new English metrical handbook to replace the English translation (1962) of Maas' short introduction from 1923, but afterwards he also published an abridged version of his book (1987) and was commissioned to write the chapter on Homeric metrics in the new and influential companion to Homer (Morris & Powell 1997). His colometric theory is based on pure metrical statistics98. Based on the frequency of word end in the third foot of the Homeric hexameter, he only accepts the penthemimeral or masculine caesura (3a) and the trochaic or feminine caesura (3b) as common caesurae of Homeric verse. In his opinion, one exception is possible, when the middle caesurae are bridged with a long word, namely by putting a caesura after the princeps of the fourth foot, the hephthemimeral caesura (4a). Such bridging of the third foot is quite rare, because in only 1,4% of the verses of the Iliad and 0,9% of the Odyssey, there is no word break in the third foot (West 1982: 36). Elision does not affect the colometry of the verse. Choosing the place of a caesura, therefore, only depends on regular word end at 3a/b or 4a and not on syntactic or semantic considerations99. This is clearly indicated by West himself, because he admits that this approach can cause the separation of (unstressed) prepositives or postpositives from the words to which they are attached (for criticism cf. infra)100. Let me give some examples to complete this survey of West's metrical colometry. Firstly, consider the following verse (Il. XVI, 19):

ἐξαϑδα, μὴ κεῦθε ||3b νϐῳ, ἵνα εἴδομεν ἄμφω. ("Speak loudly, do not hide it || in your mind, in order that we know it both.")

97 The terms are borrowed from Vergote (2011: 9-17), where she gives a similar account of the different modern colometries, with criticism. 98 Survey based on West (1982: 35-37; 1987: 19; 1997: 222-224). Bowra (1962a) and Sicking (1993) agree with West. Lehrs (1860) was the first modern scholar to argue for this statistic colometry. He also offers a list with the instances where the middle caesurae are bridged in the Homeric epics (Lehrs 1860: 514-521). 99 This is definitely not an innovation by West. Already in Antiquity, it was widely assumed that caesurae were not dependent of completions in thought (Bassett 1919: 356). 100 Sicking (1993: 53) equally sees no problem in splitting a prepositive from the word it governs, e.g. in Il. I, 53 ἐννῆμαρ μὲν ἀνὰ ||3a ςτρατὸν ᾤχετο κῆλα θεοῖο ("nine days long throughout || the army the missiles of the god raged") he would place a penthemimeral caesura after ἀνά, thus separating it from the noun ςτρατόν which it governs. But his view on the issue is rather paradoxical, because earlier in his book (1993: 19) he declares: "Als "Wort" bei erstrebten oder vermiedenen Wortgrenzen gilt nur das "Gesamtbild eines bedeutenden Redeteils", zusammen mit den zugehörigen Präpositiva und Postpositiva" (boldface added.). Similarly Maas (1962: 86). Sicking agrees with West that the caesura normally lies after 3a or 3b, but he does not accept 4a (1993: 75).

West would place a trochaic caesura after κεῦθε (3b), simply because there is word end in the third foot. An example with a prepositive is furnished by Il. I, 132:

κλϋπτε νϐῳ, ἐπεὶ οὐ ||3a παρελεϑςεαι οὐδϋ με πεύςεισ. ("Do hide it in your mind, because you are not || going to elude me, you are not going to persuade me".)

West would place a penthemimeral caesura after οὐ (3a), again because there is word end in the third foot101. The next theory I want to present, is the rhetorical colometry initiated by Hermann Fraenkel. He originally formulated his theory in 1926, only two years before Parry's thesis. Parry did not mention his article, although their conclusions share some similar points. This indicates the partial lack of interest for Fraenkel's analysis, which only gained some minor influence after the revised version in 1955 in his book "Wege und Formen frühgriechischen Denkens: Literarische und philosophiegeschichtliche Studien" (new editions in 1960 and 1968)102. The purpose of his study is to prove that "die Sinnesgliederung der Rede und die rhythmische Folge der langen und kurzen Silben aufeinander abgestimmt sind" (Fraenkel 1968³: 103). He believes that an hexameter is divided in four principal parts and that every part has some possible caesura places. Caesurae which can end the first colon of a verse are marked with A, those of the second colon B and those between the third and fourth colon are called C. The following scheme (adapted from Fraenkel 19683: 104; 111) offers an overview of his theory:

–A1∪A2∪A3 | –A4∪A!∪A! | –B1∪B2∪ | –C1∪∪C2 | –C!∪C!∪ | –⨯

It becomes immediately apparent that the A-part has the most varying positions of caesurae. Every position until Α4 (trithemimeral = 2a) is a possible caesura according to Fraenkel. In addition, he stresses throughout his paper that caesurae can be bridged in the hexameter. The caesurae characterized with an exclamation mark are caesurae which are acceptable

101 This is impossible, as will be further discussed below. Summary of the of other metrical handbooks: Koster (1936: 52): 2a, 3a, 3b, 4a, 4c (bucolic diaeresis); Maas (1962: 59-60): 3a, 3b and 4a; Dain (1965: 53-54): also preponderance of 3a and 3b, but he equally stresses the possibility of caesurae at 2a (trithemimeral) and 4a and of the bucolic diaeresis; Korzeniewski (1968: 31) follows Fraenkel; Sicking (1993: 75): 3a and 3b (cf. supra). Cf. also Finkelberg (2011 vol. 1: 149 s.v. Caesurae), where also 2a and 4c are acknowledged. 102 Summaries of his theory can be found inter alios in Kirk (1966: 76-88); Korzeniewski (1968: 30-31); Ingalls (1970: 2-3); Kirk (1985: 19); Edwards (1986: 176-178); Sicking (1993: 76); Russo (1997: 240-242); Edwards (1997: 265); Janse (1998: 140-141). Vivante (1997: 5) deplores the fact that Fraenkel's theory was not given much attention in the scholarly literature. Similar regrets are to be found in Ingalls (1970: 1). Korzeniewski (1968) is the only metrical handbook which accepts Fraenkel's theory. For other enthousiastic supporters, cf. Russo (1997: 240) and Edwards (1997: 265) (cf. Janse 2012: 23). Beekes (1972: 2) tries to reduce the Homeric colometry to six basic rules, which in his opinion can grasp the problem more fully than Fraenkel's theory.

only when the normal caesurae are bridged by a long word103. The B-caesurae are limited to B1 (penthemimeral = 3a) and B2 (trochaic = 3b). The C-caesurae are C1 (hephthemimeral = 4a) and C2 (bucolic diaeresis = 4c), sometimes bridged until 5a or 5b (C! in Fraenkel’s notation). As I stressed in the beginning, Fraenkel not only uses metrical arguments for his colometry, but he also pays attention to so-called "Sinneeinschnitte" (Fraenkel 1968³: 105). Therefore, he defines caesurae as "Die zur Binnengliederung des Verses benutzten Sinnesfugen" (Fraenkel 1968³: 111). He adds two important remarks later in his paper. Firstly, he underlines the fact that the Homeric formulae (cf. the similarities with Parry) are geared to this colometry of the Homeric verse (Fraenkel 1968³: 115). Secondly, he believes that caesurae do not imply a pause while reciting the verses (cf. infra for criticism) (Fraenkel 1968³: 149). His schematic structure of the Homeric hexameter is based on a statistical study of the places where "Sinneeinschnitte" are most prevalent in the Homeric hexameter. Fraenkel (1968³: 105) gives a short overview of the statistics, which stresses the importance of the bucolic diaeresis as the most preeminent place for syntactic and semantic breaks in the Homeric hexameter. To conclude this presentation of Fraenkel's paper, I want to discuss the same examples cited in the survey of West's theory, now using the colometry proposed by Fraenkel104.

ἐξαϑδα, ||A4 μὴ κεῦθε ||B1 νϐῳ, ||C1 ἵνα εἴδομεν ἄμφω. (Il. XVI, 19)

Fraenkel would place a trithemimeral caesura at A4 (2a), because the first word ἐξαϑδα ends here and the following word μή, a prepositive negative particle, belongs to the following clause. Next, he would place a penthemimeral caesura at B1 (3a), like West, because this is the place in the third foot where word end occurs at the end of the second clause. Thirdly, he would accept a hephthemimeral caesura at C1 (4a), because a new syntactic unit begins there. There is also word break at C2 (4c), but that would break up the syntactic structure, which Fraenkel would not do. Consider the other example:

κλϋπτε νϐῳ, ||A4 ἐπεὶ οὐ παρελεϑςεαι ||C2 οὐδϋ με πεύςεισ (Il. I, 132).

In this case, Fraenkel would put a trithemimeral caesura at A4 (2a), because of the word end of νϐῳ and the end of the syntactic structure: imperative plus complement. Placing a B- caesura causes some problems. If he had to place in the middle of the verse at B1 (3a), he would separate the prepositive negative particle οὐ from the verb, which is impossible according to Fraenkel (cf. 1968³: 142-143). The C-caesura is easier: there is room for a bucolic diaeresis at C2 (4c), as a new clause begins there with the coordinated prepositive

103 Fraenkel (1968³: 132) discusses the verses where the A-caesura is bridged. In these cases, we get an harmonic tripartition. Analogous with Kirk's "rising threefolders", which will be discussed below, we can use in this case the term "symmetrical threefolders". The instances of A! have to do with Meyer's law (cf. infra for discussion). 104 Further examples of his colometry are given in Fraenkel (1968³: 112-113).

negative particle οὐδϋ. The only way to have a fourth caesura, is to place an extra one before πεύςεισ (C!), which seems unlikely. The paper by Porter (1951) extends the statistical analyses of Fraenkel’s model (Porter 1951: 3). His elaborate statistical research culminated in very useful tables, which are attached at the end of his paper (Porter 1951: 51-63). A further contribution of his study to our conception of Homeric colometry is that he shifted the attention from the caesura to the cola, which are demarcated by the latter. One of his goals was to explain the localizations of O'Neill (1942) by paying more attention to cola than to separated words (Porter 1951: 9)105. Porter also proposes two adaptations to Fraenkel's theory, which actually weakens his analysis. Firstly, Porter rejects the notion of a colon as a unit of meaning, which Fraenkel clearly underlined as being "Sinneeinschnitte" (Porter 1951: 16 passim)106. Secondly, he reduces the places where caesurae normally occur to six positions, resulting in the following scheme (cf. also Ingalls 1970: 4):

–∪∪A2| –A1∪∪ | –B2∪B1∪ | –∪∪C1 | –C2∪∪ | –⨯

To start with, we observe that the general marking system is adapted, because the numbers do no longer refer to their chronological place in the sequence of the verse, but to their relative importance. Whereas the trochaic caesura was referred to as B2 by Fraenkel, it is called B1 by Porter because it is slightly more popular than the penthemimeral caesura. Secondly, the possible positions in the first portion of the hexameter are limited to the locations A2 (1c) and A1 (2b), the two most popular positions here (cf. already Fraenkel's own statistics). Finally, in the closing colon the importance of the hephthemimeral caesura is denied by Porter, for he does not include it as a normal position107. Instead, he upgrades position C2 (5a) to the second possible position in the last colon. I agree with Ingalls (1970: 5) that Fraenkel's choice for the hephthemimeral caesura clearly deserves our preference108.

105 Porter (1951: 17-18) stresses the advantages of the four-colon theory. It can explain the localizations of O'Neill and as such describe the hexameter ex positivo. Furthermore, it offers the possibility to compare different poets. Based on personal experience, Porter also believes his approach can facilitate the teaching of the hexameter to beginning students. Personally, I think that a cognitive approach would be even better for educational purposes (cf. infra). 106 Porter's formulations are rather vague and insecure. In subsequent pages, he seems to nuance this view, e.g. "In the hexameter a colon is frequently a short unit of meaning but need not be." (Porter 1951: 17) or "There are, of course, a great many more caesurae than punctuation marks in any given passage. Nonetheless, the evidence of punctuation, crude as it is, supports the contention that the colon is a unit of meaning, for, apart from the well-defined exceptions mentioned above, punctuation falls at a caesura" (Porter 1951: 25). 107 The hephthemimeral caesura is indeed by far not the most important caesura in Homeric verse, as was clearly proven in the paper by Bassett (1917). This problem will be further discussed in chapter three. 108 I do not want to imply that C2 (5a) is an impossible caesura. Using our cognitive approach to the hexameter, it will become clear that this is a position that can alternate with the bucolic diaeresis. In addition, Ingalls' own colometry contains some shortcomings. He has a clear preference to separate individual words, which results in rather staccato colometries (cf. Ingalls 1970: 8-11).

Porter is not the only scholar who became influenced by Fraenkel's colometry. Two other followers I briefly want to discuss, are Rossi and Barnes. Rossi (cited here in a later edition from 1995) uses Fraenkel's theory in a study about the aesthetics of long and short cola in Homeric verses. Like already Fraenkel (1968³: 113), Rossi denotes the hexameter as "une strofa in miniatura", consisting of four cola (Rossi 1995: 272)109. He further agrees with Porter that cola are no syntactic units; he believes they can only play a role in analyzing the colometry, when the rhythm alone is not decisive in pointing towards the caesura110. Disagreeing with Fraenkel, Rossi wants to highlight the importance of short cola in the colometry of Homeric verse, because they can bear more emphasis than long ones (Rossi 1995: 288). Barnes (1986) offers inter alia a reassessment of both the theories of Fraenkel and Porter. He believes that the penthemimeral and the trochaic caesurae (3a & 3b) are the only ones to be considered as fixed caesurae, as such agreeing with West (Barnes 1986: 138). This conclusion is based on his presupposition that sense-pauses do not affect the colometric structure of Homeric verse. Furthermore, he seems to have a love-hate relationship with Fraenkel's theory, because on the one hand he agrees with Fraenkel that the hexameter is a four-colon structure, but on the other hand, he finds it a less attractive hypothesis, because some unanswered questions still remain (Barnes 1986: 149). A colometry which shares some similar points with Fraenkel's theory, is the one proposed by Cornelis Ruijgh (1995: 8-11). He makes a distinction between main caesurae, which are either the penthemimeral either the trochaic caesura in the middle of the verse, as such following West et alii, and secondary caesurae which further divide the first and second hemistich of the hexameter. Doing so, he equally divides the hexameter in four principal parts, but unlike Fraenkel he adds an hierarchical structure in the different caesurae. According to Ruijgh, the first hemistich can be further demarcated into two parts by caesurae at 1a, 1b, 1c and 2a and the second hemistich by 4a, 4c and 5a, as such entirely agreeing with Fraenkel on the common places for caesurae, except for the fact that Ruijgh gives more attention to the ennehemimeral caesura (5a) which Fraenkel considered an exceptional place. Ruijgh would therefore analyse the verse ἐξαϑδα, μὴ κεῦθε νϐῳ, ἵνα εἴδομεν ἄμφω. (Il. XVI, 19) with 3a as the main caesura and the caesurae at 2a and 4a as secondary caesurae, the main one indicating with a double vertical line (||) and the secondary ones with a single line (|). The last theory I want to present, before proceeding to possible criticisms and the presentation of a cognitive approach, is Kirk's colometry. He offers a short survey of his theory in the introduction to first volume of the monumental Cambridge Commentary to the Iliad (1985-1993), for which he was appointed as the principal editor111. Generally spoken,

109 While presenting a schematic overview of Fraenkel's theory, he uses a system to refer to the different places of the hexameter, which is quite similar to the proposal of Janse (2003). Position 1a becomes 1lg [sc. longum], 1b equals 1tr [sc. trochaicum] and 1c agrees with 1bc [sc. brachys]. The problem with Rossi's notation is that it cannot refer adequately to spondees. 110 "La considerazione sintattica potrà avere ruolo decisivo solo in casi ritmicamente indifferenti, quando ci sia possibilità di scelta ritmica" (Rossi 1995: 277, cf. also 1995: 279). 111 To be found in Kirk (1985: 17-24).

his theory offers some striking similarities with Ruijgh's colometry, because it is also based on a distinction between main and supplementary caesurae. Furthermore, some parallels are found with the four-colon theory of Fraenkel and Porter (cf. Kirk 1985: 20). Schematically, his colometry can be summarized as follows:

–∪∪⋮A1 | –⋮A2∪∪ | –/M∪/F∪ | –⋮R∪∪⋮B | –∪∪ | –⨯

Firstly, some remarks need to be given about the used symbols. Kirk's distinction between main and supplementary caesurae is indicated by the difference between /, referring to a main caesura and ⋮, indicating a supplementary caesura. The terms "A1" and "A2" were certainly borrowed from Porter, but Kirk again uses Fraenkel's numerical system, where the lowest number refers to the first position in the verse and not to their relative frequency. "M" refers to the masculine caesura, "F" to the feminine one. "R" is used for the hephthemimeral caesura, because that is the place where Kirk's "Rising threefolders" have their second caesura, as will be discussed below. "B" points to the bucolic diaeresis. His theory starts, as most theories, from the observance of prevalent word end in the middle of the verse, either at 3a or at 3b. In addition, he also signals the tendency of these two cola to be further divided in two "kommata", as such resulting in a quadripartite structure of the verse. This is the ideal structure of a Homeric hexameter in his opinion. Some of the examples he uses, show that his verse segmentation is based on metrical grounds instead of syntactic ones112. Let us consider for instance the following case (the example is taken verbatim from Kirk 1985: 18):

τὴν δ᾽ ἐγὼ ⋮A1 οὐ λϑςω· /M πρύν μιν καὶ ⋮B γῆρασ ἔπειςιν. (Il. I, 29) ("I will not set her free, before also the old age will come upon her".)

To begin with, Kirk would place a main trithemimeral caesura (M), between λϑςω and πρύν, because there is word end in this place. Secondly, he will search for a further division of the first and the second colon of the verse. In the first colon, he will place a supplementary caesura (A1) after ἐγώ, because there is word end at this position and it is more natural to keep the prepositive negative particle οὐ with the verb λϑςω in the following komma. The second colon will be divided at the bucolic diaeresis, after the prepositive conjunction καί, a decisive proof that metrical structure is superior to syntactic units in Kirk's theory of Homeric colometry (cf. Kirk 1985: 18). However, as Vergote (2011: 15) rightly notes, Kirk's colometry is a dynamic one. Firstly, because he draws attention to the existence of "rising threefolders" or "rising threefold verses" (Kirk 1985: 20). In these cases, the main caesurae are bridged, either by the lack of word end in these positions, or by a syntactic unit. Consequently, there are only

112 "The cola are not therefore units of meaning, although they tend to comprise organic word-groups." (Kirk 1985: 19). This is probably the main inconsistency in Kirk's theory, as will be discussed below. Moreover, because he uses semantic units in his argumentation for some alternative segmentations (cf. infra).

two caesurae, one in the first part of the verse, the other one at the end of the verse, most notably at the hephthemimeral caesura, which is therefore coined "R" in Kirk's notation. This results in a tripartite division of the verse, a tricolon crescens as it is called in ancient stylistics113. We can look at the following examples:

διογενὲσ ⋮A2 Λαερτιϊδη /M πολυμόχαν᾽ Ὀδυςςεῦ (e.g. Il. II, 173). ("Sprung from Zeus, son of Laërtes, inventive Odysseus")

ἐξαϑδα, ⋮A2 μὴ κεῦθε νϐῳ, /M ἵνα εἴδομεν ἄμφω (again Il. XVI, 19).

The first example is one of the rare cases where the middle caesura is bridged for want of word end, the second example is syntactic in nature. Both verses can be segmented with a trithemimeral and a hephthemimeral caesura, resulting in a rising threefolder. Remark also that this division conforms with the punctuation proposed for the second example. A further advantage of this segmentation is that the imperative κεῦθε needs not to be seperated from its locative complement νϐῳ, as most other metricians would do. Furthermore, Kirk does not agree with Porter that the final conjunction ἵνα needs to be separated from the rest of this subordinate clause. Rising threefolders are not the only alternative colometries which are allowed by Kirk. In the rest of his survey he also admits inter alia the existence of threefolders which are not rising, but in some cases "symmetrical", a term which I already used above. The arguments he uses for such segmentations are based on phonetic structures in the verse, e.g. the use of alliteration at the beginning of different cola, etc. Moreover, he pays attention to semantic units in order not to divide them by a too mechanical placing of caesurae in the verse (cf. Kirk 1985: 21-24).

2.3 How to put a caesura? - Cognitive problems The above section presented the conventional theories about the colometry of Homeric verse, beginning with the statistical, and hence static vision of West, advancing towards the four-colon theory of the German scholar Fraenkel with his followers Porter, Barnes, Ruijgh and especially Kirk. Their theories will be criticized in the following section, in order to clear the way for a full-dynamic vision about Homeric colometry: a cognitive solution. Firstly, let us look again at the verse "ἐξαϑδα, μὴ κεῦθε νϐῳ, ἵνα εἴδομεν ἄμφω". (Il. XVI, 19). We saw that West would place only one caesura, namely after the word κεῦθε. This is an emblematic example for the criticisms one can object against his theory. West's

113 This is in fact a common procedure in Indo-European stylistics, coined by Behaghel as the "Gesetz der wachsenden Glieder" (cf. e.g. Korzeniewski 1968: 34). In fact this was already used by Mette (1956: 7-9) in his discussion about the colometry of Homeric verses: he admits the existence of different possible caesura places: 1c, 2a, 3a, 3b, 4a, 4b, 4c and 5a. Kirk's theory is criticized by Barnes (1986: 134-137) as too speculative. But, as will be further argued below, this theory can offer some valuable alternatives against too mechanical colometries and shares some common points with the cognitive approach to Greek colometry.

colometry regularly results in a counterintuitive division of syntactic and semantic units. In the above quoted example, a division is made between an imperative and a locative complement which is necessary to understand the imperative (where will he hide it?). Other inconsistencies caused by West's theory are e.g. the separation of an adjective and its noun, or even worse, a division between a prepositive or postpositive word and the word to which it is attached. This misrepresents the nature of these specific words in the Greek language, which need to be viewed as one unit with their determinant. This is e.g. proven by accentual practices, such as οὗτοί εἰςιν (the first word bears two accents, because the enclitic verb form εἰςιν is pronounced in one breath with the preceding demonstrative pronoun οὗτοι), a procedure which in many cases is preserved in Modern Greek. Secondly, some problems are inherent in West's choice for word end as the decisive argument for placing a caesura. Greek verse is characterized by the feature of "synapheia": all words are slurred together as a continuous stream of sounds. This accounts for the fact that, e.g., a syllable can also be closed by the beginning consonant(s) of the following word. When we return to the verse from book XVI, the synapheia can be represented as follows:

ἐξαϑδα, μὴ κεῦθε νϐῳ, ἵνα εἴδομεν ἄμφω. ek-sau-dā-mē-keu-the-no-ō-j(h)i-na-ei-do-me-nam-phō.

Word end is not relevant for synapheia, but West uses it as the only argument for placing a caesura. However, West is by far not the only metrician who neglects the fact of synapheia in his metrical theory. Obviously, I do not want to exaggerate this phenomenon, because in that case we would not be able to place any caesura at all, or it could be argued that words have no importance at all in placing a caesura, so that caesurae in the middle of a word would be possible. However, it remains important to bear this in mind, especially as Vergote (2011: 11) demonstrated that if we take synapheia into account, the percentage of word end in the third foot drops from 98% to 53% for the first 100 verses of book XIV of the Iliad114. Equally problematical are elisions within the verse. West sees no problem in placing a caesura after an elision, and he is partially right. For example, an elision before 3a could indicate that the poet actually paused at 3b, but resumed as if he had paused at 3a, this would not be heard by his listeners (Janse 2012: 32f.; cf. infra). However, when another segmentation of the verse is possible, which can avoid elision, this one can in some cases deserve our preference. Furthermore, West's proposal is too limited in its scope, because important breaks such as the trithemimeral caesura and the bucolic diaeresis are (almost) completely neglected in his theory. This is very remarkable because of the commonness of word end at the bucolic diaeresis (ca. 60%). West wants to base his theory on word end, but neglects the position which after the middle caesurae most commonly exhibits word end. Apart from that, the trithemimeral caesura is an important break, due to enjambments that regularly end there. Both places have a strong tendency to show a sense break (cf. statistics of

114 Furthermore, on the basis of word end alone, the penthemimeral caesura is clearly less preferable than the trithemimeral caesura and the bucolic diaeresis (cf. statistics in Steinrück 1995: 135f.).

Fraenkel et al.; cf. infra). The fact that they are not included by West is caused by his opinion that we have to distinguish metrical phrases from sense-pauses.

"Sentence- and phrase-structure is not closely tied to the verse-structure, but not altogether independent of it. The strongest sense-pauses […] occur at the end of a verse […]. After that, the commonest places for a sense-pause are at the caesura or at the end of the fourth foot. For the rest, sense-pauses are practically confined to the beginning of the line, in the first foot or at latest after the first syllable of the second" (West 1997: 224; boldface added).

Here, he admits the existence of important sense-pauses at the beginning and the end of the line, as distinguished from the main caesurae. He also accepts the fact that in most cases there is a correspondance between metrical phrases and sense-pauses, but he refuses to draw the logical conclusion, namely that metrical cola are to be reconciled with those sense- pauses. This is again caused by his mistaken conception of the oral nature of the Homeric epics. They are not normal written texts, but West treats them as if they are. A rather static conception of caesurae can be argued for in later written poetry, such as Hellenistic Greek verse, or a more artificial literature, e.g. the Latin poetry of the golden and silver age. It seems very improbable that an oral singer who gives an impromptu performance, would think in metrical phrases which are not connected to semantic and syntactic units. Could it be possible that during the performance a caesura is placed before a prepositive, which in daily speech is uttered in one breath with the preceding word(s)? Does it seems probable that an aoidos always placed a caesura somewhere in the middle of his verse, without regarding the syntactic, semantic and information structure of his utterance? We may not forget the fact that the Homeric language is composed as a λέξισ εἰρομένη or stringing style (Aristotle). The epic singer does not compose long sentences, but he strings together smaller intonation units, which do not always begin and end somewhere near the middle of the verse. This can lead us also to another objection against West's theory. He believes the alternation between the penthemimeral and the trochaic caesura to be a poetic technique used by the poets to differentiate between a falling rhythm in the first part of the verse and a rising rhythm in the second part of the verse, as such resulting in a pleasant variation (e.g. West 1997: 223). West appears to forget that this can also cause a strong monotony. If every verse stops at the same place, this will become very inadequate to captivate the listeners of the poem. Therefore, a more dynamic division of the verse would lead towards an augmentation of the artistic licence of the poet115. We can conclude that West's colometry is too static, which is very unlikely in the oral context the Homeric poems were composed in. Fraenkel's theory on the other hand, has some distinct advantages when compared with West's. Amongst other things, it catches the importance of syntactic and semantic units in the division of Homeric verses and it also acknowledges the importance of caesurae near

115 Cf. Bassett (1919: 343): "It is perfectly natural, and even desirable, that the constituent cola should be of varying length and that occasionally a hexameter should not contain two, but three, cola".

the beginning and the end of the verse, especially at the bucolic diaeresis. There are also some major criticisms to be raised against his theory, although some of the rejections proposed by previous scholarship are not convincing. Let me start with the (negative) evaluation of the theory by Lukikovich and Steinrück (2004: xi). One of their objections is the fact that ancient metricians did not believe in the existence of the trithemimeral caesura (2a) or a system similar to Fraenkel's (cf. Steinrück 2010-2011: 274). However, as already stressed above, we need to be cautious to pay much attention to ancient metrical theory, because it is only attested long after the composition of the Homeric poems and because the material is very fragmentary, very late and very ambiguous in nature. Let us further not forget that enjambments point towards the existence of caesurae early in the verse (cf. infra). Moreover, Lukikovich and Steinrück deny the importance of the frequency of word end in Fraenkel's colometry. They believe this is only caused by the use of a poetic vocabulary, which influences the places where many word ends occur. Probably, they reverse the causal order of the facts. It seems more likely that the choice of poetic words was caused by the colometry of the verse instead of vice versa. This will become apparent when I will discuss the influence of the verse on the creation of formulae beginning and ending at the common caesura positions. Remember, for example, Witte's (1913: 2214) statement: the Homeric language (and hence its lexicon) is "ein Gebilde des epischen Verses". A more important problem, which is also mentioned by Lukinovich and Steinrück, concerns the applicability of Fraenkel’s theory to actual hexameters116. Indeed, I agree that it constitutes a problem that not every hexameter can easily be divided into four meaningful components. Does it stand up to scrutiny that single words, which bear no special emphasis, can be separated as a single colon? In this case, the caesura seems to be restricted to a break between single words. This steers us towards the problem as to whether a caesura is a real pause in the performance of a verse. In my opinion, the answer has to be (at least partially) affirmative. Remember the remark by Bassett (1919: 344) that the hexameter is too long to contain one single grammatical unit. I would like to reformulate this remark: "a hexameter is too long to be uttered within one single breath"117. At least one pause needs to be postulated in order to utter an hexameter in a comfortable manner, both for the production by the speaker and the reception by the listener. On the other hand, it seems quite unlikely that every verse should have three pauses during its production. This can be related to Nagy's (1998: 497-499) criticism of Fraenkel's theory118. Firstly, Nagy rightly highlights the fact that Fraenkel does not pay sufficient attention to the Sinneseinschnit at the end of the verse. As

116 Cf. also Edwards (1986: 180). Such a segmentation can result in rather odd cola. 117 Cf. Beekes (1972: 3): "As half a hexameter is about the normal length to be accepted, and as the ratio of the caesura is to present units of this length, there are no more caesurae in the hexameter than one". He rightly points to the problem, that there needs to be a pause for the listener, but on the other hand sticks to a static colometry with only one caesura, normally towards the middle of the verse. Cf. also Sicking's remark, quoted supra. 118 Nagy does not believe that syntax has any influence on the colometry of Homeric verse: "from the synchronic point of view, metrical pause is independent of syntax" (Nagy 1998: 498). His criticism is also found in Nagy (2010: 385).

West (e.g. 1997: 224) points out, 63% of the Homeric hexameters end with a sense-pause119. Some problems arise with regard to Nagy's other objection that Fraenkel's theory cannot explain the Homeric formulaic system. It is precisely interesting that also formulae exist which are confined to the beginning of the verse or to the end of the verse and therefore can be interpreted as consequences of frequent word break in these positions. A more fundamental problem is the fact that although Fraenkel postulates the existence of Sinneseinschnitte, he does not pay corresponding attention to the syntactic and semantic structure of the verse120. Let us not forget that in the continuously quoted verse from Iliad XVI, he would also split the imperative from its locative complement, due to the fact that his colometry remains somewhat static. There needs to be a caesura in the middle of the verse, even if this seems counterintuitive. Probably, more attention needs to be paid to semantic and syntactic units, a criticism a fortiori valuable for Porter's pure metrical verse segmentation. Furthermore, Porter's denying of 1a and 1b as possible caesura places, is refused by the fact that important sense-pauses can end there. The most convincing colometry I have presented up to this point is Kirk’s. Firstly, I agree with his distinction between main and supplementary caesurae (cf. also Ruijgh). In doing so, we are able to differentiate between strong and lighter pauses during the performance, say between a full stop and a comma. In some cases, pauses can sometimes be hardly detectable and only marked by prosodic features like intonation, loudness etc. Furthermore, he is the only metrician who points towards the fact that a tripartite verse structure is possible as well, as an alternative to bipartite and quadripartite divisions. West accepts only bipartite and Fraenkel only quadripartite divisions, unless the exceptional cases where the middle caesurae are bridged, resulting in a rather monotonous system, whereas Kirk's proposal offers the possibility for a more varied verse segmentation. However, some important shortcomings remain with his theory. In the first place, he is not consistent about the question whether or not semantic and syntactic considerations should play a role in determining the colometry. In one case, he does not hesitate to separate the prepositive

119 Daitz (1991) discusses the problem of placing pauses during the performance of Homeric verses. He agrees with Nagy that there needs to have been a pause at the end of a verse, even if there was no sense boundary. This is indicated by the existence of the syllaba anceps at the end of the metrical structure and the lack of elision (Daitz 1991: 151-152). On the other hand, he does not believe in pauses in the middle of the verse. Firstly, because of ancient testimonies who state that this was not the case. He refers to Latin authors such as Cicero and Quintilian who lived almost thousand years after the putative composition of the Homeric poems, so are they to be trusted? He further argues for it on the basis of elision before a full stop, but as I already stressed, it seems possible that a vowel was pronounced furtively, in order not to influence the metre. We should not forget that we are dealing with a written text for which the earliest complete manuscripts are more than 1500 years older than the putative composition of the poems (Daitz 1991: 153-156). Cf. Daitz (1991) for more discussion, but I am not convinced by his arguments. For similar ideas see Barnes (1986: 135): "It need not be assumed, however, that a metrical caesura must create a pause (or any other effect) in speech" (cf. Irigoin 2004: 3f.). 120 For this reason I agree with Vivante (1997: 8) that Fraenkel's theory has some important advantages, amongst others because it highlights the simplicity and unity of the verse form. But he rightly adds the following criticism: "He does not elaborate on any further cognitive significance" (loc. cit.).

conjunction καί from the syntactic construction which it introduces (Il. I, 29), but on the other hand, he argues that we have to be aware of semantic units, stylistics etc. with regard to the colometric structure of the verse. This results in a rather paradoxical view. In addition, he continues to pay too much attention to word breaks towards the middle of the verse. Although he rightly demonstrates the existence of alternative colometries, he always places a main caesura where there is word end towards the middle of the verse. As we will see below, it is sometimes better to interpret caesurae at the beginning or the end of the verse as main instead of supplementary caesurae. In order to argue for this in a convincing manner, it will be necessary to offer a well-founded methodological ground for doing so. Barnes (1986: 137) offers some solid arguments to show that Kirk's theory remains somewhat speculative. Therefore, let us proceed to the theory of Wallace Chafe about intonation units and demonstrate how this can be applied to a less-speculative, but more dynamic segmentation of Homeric verses. Chafe is a cognitive linguist who became interested in the differences between spoken and written language. Although homo sapiens has been talking for more than ten or maybe even hundreds of thousands of years, writing systems are only attested from the beginning of the third millennium BC onwards (cf. Chafe 1985: 113-114). Chafe's research contributed to a more profound interest in and understanding of spoken language121. He did so by making use of a cognitive approach towards spoken language. He demonstrated that human cognition can only focus on one thing at a time (Chafe 1994: 28-30), but this focus is constantly shifted from one concept to the other. This is related to the point of view you take at a certain moment, and to several factors, such as time, place, the society you live in etc. The fact that only one concept can be active at a time in our mind, is called "the one-new- idea-constraint" in Chafe's theory (1994: 108)122. Because of this cognitive constraint, Chafe argues that spoken language is composed on the basis of a joining together of what he calls "idea units/intonation units"123. By this, it is meant that a speaker thinks in shorter units, which are uttered as such during the production of speech. There is some concept active in your head and you phrase it as a short intonation unit, which is also related to the prosody of your speech. The pitch alternates between different intonation units, just as the volume, the timing etc. (Chafe 1994: 53). During a spoken utterance, different intonation units are joined together, with a short pause between them (Chafe 1985: 111). There is some discussion about the length of such intonation units: in his various publications about the problem,

121 Two papers by Chafe (1982; 1985) are almost entirely devoted to the oppositions between spoken and written language. A good survey is found in Chafe (1994: 41-48), some examples include the contrast between the spontaneity of spoken discourse, versus the careful thought in written language, or the difference between evanescence in oral language versus permanence in written discourse (compare the proverbial saying Verba volant, scripta manent and Homer’s formular ἔπεα πτερόεντα). 122 However, other concepts may remain semi-active in a peripheral zone so they can be easily re-activated when you need them (for discussion cf. Chafe 1987: 25f.). 123 Some synonyms are given in Chafe (1994: 53): tone unit, intonation group, intonation(al) phrase or intermediate phrase.

Chafe's numbers vary between four and seven words124. It needs to be stressed that these numbers are based on English and Native American languages, so they could be different from a synthetic language such as Ancient Greek. In addition, one can ask whether grammatical words have to be counted as separate words and if so, whether there are hierarchical differences between articles and, e.g., auxiliaries. More research needs to be done, comparing different kinds of languages. Let me illustrate Chafe's theory with a short example, which shows that spoken language is a concatenation of short linguistic utterances, which do not even have to be joined with conjunctions such as and, or etc. (cf. Chafe 1982: 38; 1985: 111):

"And my room was small # it was like # nine by twelve or something # it seemed spacious at the time # I came home # I was really exhausted # I was eating a popsicle # I was sitting there in my chair #" (example from Chafe 1982: 38)125.

The above example shows that in spoken discourse the speaker makes use of relatively short intonation units. In the first one, we receive new information: the room was small, so we want to know how small. In order to do so, the speaker needs to find the information in some peripheral activity. "It was like" is used to bridge this search for the information, followed by the intonation unit which actually gives the new information. Afterwards, the proportions of the room remind the speaker of his youth, the room was quite spacious at that time. The following intonation units are in fact short sentences, each time beginning with a first personal reference. They are not linked as in written language: the listener has to establish the logical coherence of the "sentence" him/herself. In fact, it is very difficult to use the term "sentences" in spoken discourse, due to this preponderance of coordinate concatenation of different intonation units (cf. Chafe 1987: 45-46)126. How can this theory be applied to the Homeric epics? In fact, Chafe himself already pointed towards the possibility, for he stressed that an oral performance needs to be interpreted as an intermediate stage between normal spoken discourse and written language (cf. Chafe 1982: 52). Bakker (1990a; 1990b; 1997a; 1997b; 2005) laid the foundations for linguistic analyses of the Homeric poems as a special form of oral discourse: "It was a matter of speech and voice, and of the consciousness of the performer and its audience" (Bakker 1997a: 1). The Homeric language needs in this case to be viewed as a special, poetic form of spoken discourse. Because of this peculiar status, it is wrong in his opinion to interpret the poems merely as a written transcription of an oral performance: both orality and literacy were important in the composition of the Homeric poems (Bakker 1997a: 15-18; 1997b: 285-288; 2005: 44-45). This can be argued for on the basis of the

124 Chafe (1982: 37): six words; Chafe (1985: 106): seven words; Chafe (1987: 22): between five and six words; Chafe (1994: 65): only four words. Is is not entirely clear how he defines "word" in these cases. 125 Due to the special character of spoken discourse, I did not include punctuation marks in this example. # is used to refer to the end of an intonation unit. 126 For a brief summary of Chafe's theory, cf. Bakker (1990a: 3-4: 1997b: 288-291).

already mentioned λέξισ εἰρομένη of the Homeric poems, which we do not have to interpret as a more primitive form of Greek, as was regularly done in the past (e.g. Chantraine 2015²: 351-364), but rather as another kind of discourse than the periodic style we are accustomed to with regard to the written literature of Classical Antiquity127. Just as spoken discourse, we can analyze it as a concatenation of intonation units, which last for about two to three seconds and give some piece of new information (Bakker 1997a: 36-41; 1997b: 291-293). In fact, Homeric speech is composed of "spurts of vocalizations" (Bakker 1997a: 47)128. Therefore, it seems appropriate to minimize the use of the concept "sentence" in discussions about epic Greek and to replace it by the intonation units (Bakker 1990a: 7; 1997a: 49; Janse 1998; 2012: 17). Let us consider the following example, taken from Janse (2012: 25f.):

Κϑκλωψ, εἰρωτᾷσ μ᾽ ὄνομα κλυτϐν, αὐτὰρ ἐγώ τοι ἐξερϋω: ςὺ δϋ μοι δὸσ ξεύνιον, ὥσ περ ὑπϋςτησ (Od. IX, 364-365). ("Cyclops, you ask me for my known name, but I will say it to you: you, give me a friendly gift, like you promised").

We can analyse this as a string of intonation units, just like the English example above129:

Κϑκλωψ # Cyclops # εἰρωτᾷσ μ᾽ ὄνομα κλυτϐν # you ask me for my known name # αὐτὰρ ἐγώ τοι # But I, you know, # ἐξερϋω # I will say it to you # ςὺ δϋ μοι δὸσ ξεύνιον # you, give me a friendly gift # ὥσ περ ὑπϋςτησ # like you promised #

I hope it becomes apparent that there are striking similarities between this example from the Iliad and the example of English spoken discourse above, both are made from some short linguistic utterances (sometimes words), which are joined together, often without an overt connector. Every intonation unit offers a new piece of information, but only one ("the one- new-idea-constraint") (cf. Bakker 1997a: 99). Therefore, this modern linguistic theory can offer valuable new insights in the colometry of Homeric verse and be of great help as a well-founded theoretical background in placing caesurae. Bakker (1997a: 49-50) stresses the fact that the boundaries between the different intonation units (the # in my notation) can be interpreted as the caesurae of the

127 Cf. Chafe (1985: 111): intonation units are joined together in a coordinate way, very much like the λέξισ εἰρομένη of Homer. For a short discussion of parataxis in Homer cf. Finkelberg (2011 vol. 2: 627 s.v. Parataxis). 128 Cf. Bakker (1997a: 148; 2005: 47-48): the hexameter is too long to be captured as one entity by the human mind, so it has to be divided in shorter parts. Intonation units are related to the psychological limits of human cognition. Bakker (2005: 50-52) underlines the fact that Homer uses stylized forms of intonation units, because they are not only linguistic, but also metrical and rhythmical entities. 129 The arguments for this colometry will be discussed at length in the following section.

Homeric verses130. Metrical cola, therefore, should be equated with intonation units. How this can be done on the basis of Ancient Greek information structure and word order will be discussed in the following section, further discussing some examples in order to prove the validity of this cognitive approach towards Greek hexametrical verses131.

2.4 How to put a caesura? - Cognitive solutions How can we use the elements of Homeric language and versification in order to segment the verses into intonation units? In fact, Porter (1951: 7) gives quite an adequate answer to this question: "In short all syntax, the functional organizations of sound, becomes proper subject matter for metrical study and must be analyzed as an element in the poet's realization of his form". Unfortunately, he did not follow his own advice, as was discussed above. In this last section, I will briefly present the internal criteria of Homeric verses which can be used in order to offer a solid argumentation for a dynamic colometry which is less tentative than Kirk's proposal132. Firstly, because in this approach the importance of semantic and syntactic units is stressed, it might be useful to look at punctuation practices in critical editions. They could give a first indication where important breaks occur in the sentence. However, this punctuation constitutes only a modern phenomenon and is by no means older than the first modern edition of the text133. Furthermore, there are some differences to be found between divergent editions, caused by the fact that every editor has his/her own interpretation of the text and can further be influenced by the common punctuation practices in his/her mother tongue (Janse 1998: 135; 2012: 15). Therefore, we need to proceed to more secure arguments which can be detracted from the Greek text itself. An important aspect in this case concerns the problem of word order in Greek. It is generally assumed that Ancient Greek exhibits a free word order. This is true until a certain point, but Ancient Greek word order is by no means random or arbitrary (Dik 2007: 8). Some

130 Bakker was not the first to relate metrical phrases with syntactic and semantic units, but he was the first to argue for it with a strong linguistic methodology. Apart from Fraenkel, cf. inter alios Korzeniewski (1968: 31): "Sinneeinschnitte, die von unterschiedlicher Stärke sein können, und Metrische Gliederung bilden also eine Einheit. Satz und Rhythmus werden gegliedert". Or Edwards (1966: 117): "There is a close relationship between the sense-units of the sentence and the metrical cola, or, putting it another way, between the pauses in sense and the caesurae of the verse" (cf. Edwards 1997: 265). 131 For a summary with appraisal of Bakker's theory, cf. e.g. Russo (1997: 257-258); Allan (2009), focussing on the relationship between intonation units and caesurae, Janse (1998: 135; 2012: 16), Vergote (2011: 20-22) and Hajnal (2003: 226-228). Slings (1994) focuses on the implementation of Bakker's theory on the grammar of the Homeric poems. Some grammatical phenomena corroborate Bakker's thesis that Homer's language is a stylized form of spoken discourse, e.g. the existence of anacolutha, parataxis, short sentences, avoidance of AcI or indirect speech in general, limited use of participles, frequent repetition of topical information etc. 132 The following survey is based on Fraenkel (1968³: 143); Dover (1960); Janse (1991: xiv); Bakker (1997: 51); Janse (1998: 142-144); Vergote (2011: 27-32); Janse (2012: 23-28). 133 We always need to be cautious that we are dealing with manuscripts which are very recent, when we compare them with the original composition date of the poems!

words are more likely to be placed at fixed positions in the sentence. This can be best illustrated with so-called prepositive and postpositive words, which refer to words which are obligatorily placed before or after another word and cannot be used on their own. The former are placed before another word and can never be placed at the end of (a part of) the sentence (Fraenkel 1968³: 142f.; Janse 1998: 142f.; 2012: 23f.). They cannot be used on their own, as independent intonation units, but they depend on the preceding word with which they form an accentual unit. Therefore, it is impossible that a caesura is placed immediately after them (cf. Fraenkel 1968³: 145). Some examples include the unaccented prepositions ἐν, ἐσ, εἰσ or the negative particle οὐ (the proclitica stricto sensu) or co-ordinate and subordinate conjunctions, e.g. καί, ἤ, ἵνα or ὥςτε (Dover 1960: 13f.; Dik 2007: 18)134. Moreover, certain words have a natural tendency to bear some additional emphasis, hence they are regularly placed at the beginning of the (part of the) sentence: orthotonic pronouns like ἐγώ, demonstrative pronouns, interrogative pronouns, negations or quantifiers such as πᾶσ or ὅλοσ (Janse 1998: 143; 2012: 25). Dover (1960: 20ff.; 26ff.) has called these words "preferential", they are more likely to be placed at the beginning of an intonation unit. Hence, they can be an indication that we have to place a caesura in front of them, they are a kind of "border signal", although this never constitutes a strict law (Janse 1998: 143; 2012: 25). Two categories of prepositive words need some more discussion: imperatives and vocatives. Both categories are normally intended to catch the attendance of the addressee, more of less emphatically according to the position in the word and the situation (Janse 1998: 144; 2012: 26; Dik 2007: 206-207). Consider e.g. the following English example: "John, did you forget to send me a postal card although you were on a holiday? Send me one the next time!" In order to immediately stress his/her indignation, the speaker starts his/her sentence with an emphatic addressing, in the first sentence with a vocative and in the second one with an imperative. By placing both the imperative and the vocative at the beginning of the sentence, they bear considerable emphasis (generally on emphasis, cf. Dover 1960: 32-34). Hence, they can be used as an argument that a new intonation unit starts (Janse 1998: 144; 2012: 26). It is equally possible, according to the context, that they constitute a single linguistic unit. However, we find in Greek also examples where vocatives and imperatives are placed in the middle of the sentence, so it can be assumed that they bear less emphasis in these cases. Let us consider the opening verse of the Iliad (I, 1). μῆνιν ἄειδε θεά, ||3a Πηληώϊδεω Ἀχιλῆοσ. ("The wrath, sing of it, goddess, || of the son of Peleus, Achilles".) In this case, both the imperative (ἄειδε) and the vocative (θεά) are placed in the middle of the sentence after the emphatic word μῆνιν. Of course, the opposite word order, e.g. θεά, ἄειδε μῆνιν… would not scan as an hexameter (∪–∪–∪––), but it would also have another information structure, with much less emphasis on the word μῆνιν. "The wrath of Achilles" is the general theme of the epic, therefore it is placed in an emphatic position at the beginning

134 I do not distinguish here between proclitics and enclitics which bear no accent and prepositives and postpositives which bear an accent, this can be a later editorial practice. For similar thoughts: cf. Fraenkel (1968³: 145), Ruijgh (1990: 214).

of the verse and even of the whole epic. Hence, it can be considered as a kind of title for the work itself. Addressing a muse, on the other hand, is a common characteristic of the epic tradition. Therefore, this is expected by the listeners, who are not surprised that the imperative and the vocative bear less emphasis than the theme of the epic which they will be listening to. The imperative and vocative lose in fact their preferential position to a word which is even more emphatic in this particular context. In this case, Schwyzer (1950: 60) explains such vocatives as enclitics. Postpositive words on the other hand, are never placed at the beginning of a sentence or intonation unit (e.g. Fraenkel 1968³: 142f.; Janse 1998: 142f.; 2012: 24f.; Dik 2007: 11). They include inter alia unemphatic personal pronouns, the connector τε, the modal particles κε and ἄν or discourse particles such as δέ, δή, μέν or γάρ (Dover 1960: 12f.). This propensity for the second place was discovered by the Swiss Indo-Europeanist Jacob Wackernagel (1892), who proved it to be a common characteristic of ancient Indo-European languages135. Later decipherments of the Hittite language and the Mycenaean documents corroborated his law (Krisch 1990: 64). Wackernagel himself (1892 passim) defined his law as "enclitic and postpositive words are placed at the second place of a sentence in ancient Indo-European languages", but it seems better to refine his definition: in fact these words are not so much placed at the second place of a sentence, but at the second place of an intonation unit (Fraenkel 1968³: 142f.; Janse 1991: xv; 2012: 24)136. We can agree with the definition proposed by Janse (1991: xv): "Les enclitiques se placent volontiers après le premier mot accentué d'un segment phrastique, surtout quand il s'agit d'un mot focalisé." Such focalised words can even be accompanied with a whole string of postpositive words. The "preferential" words, which are emphasized "by their very nature" (cf. Janse 1998: 143; 2012: 25) and hence placed in the first position, attract postpositives into the second place as a result of Wackernagel's law, by which they become even more emphasized. Concluding, postpositives can offer indications that a caesura has to be placed in front of the word they are attached to. A last important argument to segment Homeric verses in a cognitively plausible manner are the syntactic structures of the sentence. Some structures belong together to such a degree that a caesura in the middle of them seems unlikely. For example, a combination of

135 It was already observed for the Indo-Iranian languages (cf. Wackernagel 1892: 402), but Wackernagel extended it to a thorough analysis of the Greek language and even supplemented it with examples from German, Celtic and Latin. His article discusses hundreds of examples in these different languages. The canonical order of postpositive words in Greek, focussing on Homer, is provided by Ruijgh (1990: 223-224). 136 There is some deficiency in Krisch' (1990) opinion about Wackernagel's law. For example, he tried to prove that in some cases the enclitic only puts some additional emphasis on the word before it, but if we define the law not in terms of sentences, but of intonation units, this makes no sense. Let us consider a Vedic example: RV 4,5,12: kiṃ no asyá dráviṇaṃ kád dha rátnam ("Which one therefrom is our wealth, which our goods?"), according to Krisch the particle ha (dha is a sandhi-form) only puts additional emphasis on the interrogative pronoun kad (Krisch 1990: 65). However, it makes more sense to interpret this verse as the combination of two intonation units, which are both introduced by an interrogative pronoun directly followed by an enclitic in Wackernagel-position: no and (d)ha. Doing so, a caesura after the seventh syllable of the triṣṭubh, can divide the verse in two almost symmetrical phrases.

a noun and a genitive which is attached to it, constitutes a syntactic unity, hence it cannot be separated by a caesura if we assume that semantic and syntactic considerations are important for the colometric segmentation of Homeric verses. The same can be said about other syntactic structures which clearly belong together, e.g., the combination of a preposition and a noun or the combination of an adjective and a noun. The last category needs two important remarks. Firstly, when placing caesurae one has to pay attention to the combination of nouns and adjectives. An adjective which is placed before a noun normally bears more emphasis, because when it is placed after a noun, it can be somehow guessed (Dik 2007: 39f.; Bakker 2009: 17), although discussion remains to which extent word position is decisive for the semantics of the adjective (Bakker 2009: 71)137. This is a specific instance of what was called by Bolinger "Linear Modification", when a speaker starts an utterance, the possibilities are almost endless, but while proceeding, every word restricts the possibilities (Bolinger 1952: 1117-1119). At a micro level, the same can be postulated for adjectives (Bolinger 1952: 1121ff.). For example, if a poet first uses a word for "elephant", you will not expect that an adjective follows, which describes this elephant as flying through the sky or something. Therefore, when an adjective is placed before a noun, one needs to be careful whether it bears some additional emphasis. In this case, it could be possible that a caesura needs to be placed in front of it. In addition, Greek has the advantage that due to the existence of cases, an adjective can be separated from its governing noun (hyperbaton), this put some extra emphasis on the adjective and may therefore be another argument in favour of a caesura (Dik 2007: 24f.; cf. Bolinger 1952: 1122f. for alternatives in English). The same can be said about superlatives (cf. Dik 2007: 75; 231 "inherently salient"). Secondly, we need to differentiate between cases where the adjective clearly belongs to the noun and cases where it has a certain independency and can be interpreted as an apposition. Such appositions can also be an argument for the colometry, due to their independent character (Janse 1998: 141f.; 146ff.; 2012: 23; 28f.). Apart from adjectives, other word classes can also used as appositives, e.g. nouns, participles etc. We can illustrate these remarks with the following example (cf. Janse 1998: 139; 141f.; 2012: 20; 23):

ἈτρεϏδησ τε ἄναξ ἀνδρῶν καὶ δῖοσ Ἀχιλλεϑσ (Il. I, 7). ("The son of Atreus, lord of men and the divine Achilles").

Traditional metricians would place a caesura in the middle of the verse (3a), between ἄναξ and ἀνδρῶν, but as was remarked above, this would require the separation of a syntactic unit, consisting of a noun and a genitive belonging to it. Therefore, it is preferable to search

137 Cf. Dik (2007: 39): "In a paper on the position of attributive adjectives in Herodotus (Dik 1997a) I first explored this terrain. There I proposed that for purposes of word order it was unnecessary to distinguish different semantic types of adjectives, but that a more general rule held for all nouns and their attributive adjectives. I proposed that by default, adjectives follow their nouns, and this will also be the preferred order if the noun is the most salient element in the noun phrase; if, however, the modifier is contrastive, or otherwise the most salient element of a noun phrase, it will precede the noun.” For further discussion, cf. also Dik (2007: 101ff.) and the monograph of Bakker (2009).

for another segmentation. First, we remark the enclitic position of τε after ἈτρεϏδησ, placing a caesura between them is therefore impossible. On the other hand, the prepositive conjunction καί is used, which is a strong indication that a caesura (4a) needs to be placed in front of it, because it is the place where the two main Greek characters of this verse and the epic in general are combined. In addition, the formulaic sequence ἄναξ ἀνδρῶν (e.g. Il. II, 234) can be interpreted as an apposition with ἈτρεϏδησ, which can account for an additional caesura at 2b in this verse. Formulaic sequences can indeed also be used as an argument for the verse segmentation. As was discussed above, they came into being because they were regularly used by the poets, hence they can be interpreted as "stylized intonation units" (Bakker 1997a: 53), which makes it probable that they can be separated by a caesura (Janse 1998: 136; 2012: 17). In addition, enjambments can be used as an indication for a caesura early in the verse, as will be further discussed in chapter three (cf. Janse 1998: 145; 2012: 27). The above presented criteria can thus be used for a segmentation of Homeric verses, although it needs to be stressed that Greek word order remains a free one. Therefore, these criteria are by no means "laws", like the sound changes, which were described as such by the Junggrammatiker. We have to interpret them as tendencies, which have to be defined with regard to the particular context they occur in, as will become apparent during the discussion of the following examples.

ἐξαϑδα, ||2a μὴ κεῦθε νϐῳ, ||4a ἵνα εἴδομεν ἄμφω. (Il. XVI, 19).

I cited this verse several times, now are we able to give a dynamic argumentation that this verse is indeed a "rising threefolder". Firstly, this can be argued for by means of the emphatic position of the imperative ἐξαϑδα at the beginning of the verse, which is a separated intonation unit in this case. It is followed by a prepositive negation μή, which is placed before another imperative which loses its preferential position here in favour of the prepositive negation, which cannot be placed after it. As was stressed above, it is unlikely to place here a caesura in the middle of the verse (3b), because in this case the syntactic unity between the imperative and its nominal complement would be broken. The bucolic diaeresis can be explained on the basis of the conjunction ἵνα, which is a prepositive word, that joins the following subordinate purpose clause to the preceding intonation units.

Κϑκλωψ, ||1c εἰρωτᾷσ μ᾽ ὄνομα κλυτϐν, ||4c αὐτὰρ ἐγώ τοι ἐξερϋω: ||2a ςὺ δϋ μοι (||3a) δὸσ ξεύνιον, ||4c ὥσ περ ὑπϋςτησ (Od. IX, 364-365).

A representation of these verses as a sequence of intonation units was already presented in the previous section. Now, we can proceed to a discussion of the arguments we have in favour of it (cf. also Janse 2012: 25ff.). We can start with the opening word Κϑκλωψ, which is a vocative placed at the first position of the verse and hence bears some emphasis. We can imagine that Odysseus calls the cyclop with a loud voice, in order to draw his attention immediately. This can be a first indication to place a caesura (1c) after it. This is

corroborated by the enclitic position of μ᾽ after the verb εἰρωτᾷσ, by which it is indicated as the beginning of a new intonation unit. A caesura in the middle of the verse (3a) is unlikely, because this would separate the direct object ὄνομα κλυτϐν from the governing verb. In addition, the sequence [name - known] is an illustration of the "linear modification" which was presented above. Because the noun is named first, the adjective can somewhat be guessed and therefore bears no additional emphasis, hence placing a caesura before it seems unlikely. On the other hand, important arguments can account for a bucolic diaeresis in this verse. The conjunction αὐτάρ is a prepositive word which joins this part of the sentence with the previous one. In addition, it is followed by a word in Wackernagel position, namely τοι. ἐγώ is normally a preferential word due to its ortothonic character (cf. supra), but is this case the conjunction αὐτάρ needs to be placed in front of it, for the joining with the previous intonation unit. In the second verse can the enjambment be used as a first indication that we can opt for a trithemimeral caesura (2a). Moreover, the emphatic position of the pronoun ςύ, which is followed by a string of postpositive words (δϋ μοι) are clear indications that a new intonation unit begins in this position. For the bucolic diaeresis, we have equally important arguments at our disposal. It begins again with a prepositive word (ὥσ) which is followed by a postpositive in Wackernagel position (περ), by which a subordinate construction begins, lasting until the end of the verse. The question remains whether we also have to admit a caesura before δόσ (3a). This can be argued for on the basis of the imperative as a "preferential" word. In addition, the sequence ςὺ δϋ μοι would be similar to αὐτὰρ ἐγώ τοι in the previous verse. However, on the other hand, the emphatic position of ςύ is an indication for the listener (both the cyclops and the audience of the poem) that it is likely that an imperative will follow, hence this one does not bear much emphasis. In addition, hospitality is one of the central themes of the Odyssey, hence the content of the imperative is also not very surprising. This makes it clear that one has to argue comprehensively for a dynamic segmentation of Homeric verses. Greek word order is by no means at random, but not fixed. Therefore, one always have to look at the specific context while applying the above stated "rules" to concrete examples. We have to keep this into mind, while proceeding to the following chapter, which discusses the origins of the hexameter.

CHAPTER 3: THE PROTOHISTORY OF HOMERIC METRE

3.1 A critical survey of previous attempts

3.1.1 Preliminary remarks "Les coïncidences signalées ici, entre les types grecs et les types védiques, ne sont pas fortuites: elles sont trop complètes, trop précises dans le détail pour qu'on y voie de purs accidents, ou pour qu'on les explique toutes par le parallélisme des types linguistiques" (Meillet 1923: vii). The founding father of modern comparative Indo-European metrics is convinced in his seminal study about Greek and Vedic verse: there are too many similarities to be found between these two metrical traditions in order not to draw the logical conclusion that they derive from a proto-model, which must be situated in Indo-European times138. Methodologically, he argued for the importance of Vedic and Aeolic verse for the reconstruction of an original IE metrical system, because they seem to be the most archaic forms. Their main features are in fact the same: metrical patterns are based on the opposition between long and short syllables and they neglect the pitch accent of the language itself139. Furthermore, the places where word end can occur, are limited to a restricted number of positions in the verse (Meillet 1923: 11; 31). Verses are based on a fixed number of syllables, some of them obligatorily long, others obligatorily short and, mostly at the beginning of the verse, the poet has the choice between long or short (e.g. West 2007: 48). They are put together in stanzas of three up to five verses. Based on Vedic and Aeolic strophes, we can postulate the following metrical patterns in Indo-European. The basic pattern is based on a verse of eight syllables, which evolved into the gāyatrī and anuṣṭubh in Vedic and the iambic and choriambic dimeters and glyconeus in Greek. The basic form can be represented with the following scheme: ⨯⨯⨯⨯∪–∪⨯||. In addition, there are stanzas which are based on longer verses of eleven and twelve syllables, which end either with a trochaic or iambic cadence and have a caesura after the fourth or fifth syllable (Tichy 2010: 2).

a) ⨯⨯⨯⨯|∪∪–|–∪–⨯#

138 Cf. supra for the methodological problems concerning this inference. 139 This point was criticized by Nagy (2010: 384), who believes that accentual patterns can play a role in the composition of the hexameter. This inspired Abritta (2015) to devote a paper to the problem. However, I am not fully convinced by their proposals. I certainly do not want to declare that the pitch accent was lost in poetry, this would be totally illogical. But I think, it had only a function in the rhythmical pattern as to cause some melody in the pronunciation which was accompanied with the underlying bass of long versus short syllables. This is also to be observed in the traditional recitation of Vedic literature.

b) ⨯⨯⨯⨯⨯|∪∪|–∪–⨯# (hendecasyllabe, base of triṣṭubh and Sapphic strophe). a) ⨯⨯⨯⨯|∪∪–|–∪–∪⨯# b) ⨯⨯⨯⨯⨯|∪∪|–∪–∪⨯# (dodecasyllable, base of jagatī, iambic trimeter etc.)140.

As I already put forward in the introduction, these hypothetical models could also be corroborated on the basis of extended researches in Slavic and Celtic metrics (cf. e.g. Nagy 1974: 2; West 2007: 46). However, one important metre remains which cannot be directly related to this models, because of three important reasons. Firstly, the dactylic rhythm which is characteristic of the hexameter, cannot be easily reconciled with the iambic and trochaic sequences to be found in the Indo-European models. When we consider for example the characteristic cadence of the hexameter (–∪∪|–⨯), it cannot be easily related to the cadence of the Indo-European prototypes (Tichy 2010: 6). Secondly, the possibility to replace two short syllables with a long one is an innovation which is not to be found in Vedic and Aeolic metres141. In doing so, the original syllable-counting principle of Indo-European verse got lost in the hexameter. This is a third important difference, moreover the average number of 15 2/3 syllables is far too much to be original (e.g. Tichy 2010: 3). The present chapter will first give an overview of the different theories which were proposed during the past century. As was stressed in the introduction, they will be placed together on the basis of their underlying principles: borrowing, the coalescence of shorter verses, internal expansion or the principle of anaclasis ("syncopation"). I will bring into focus important criticisms which can be raised against these proposals. It will be argued that they cannot adequately explain the synchronic colometry of Homeric verse and are based on insufficient methodological grounds. They start from one eye-catching feature and base a whole theory on it142. Afterwards, I will shortly summarize Kurt Witte's theory, who on the

140 A valuable study of Indo-European metrics is provided by West's (1973b) paper. He does not only focus on Greek and Indo-Iranian but discusses metres from almost every branch of Indo-European. Short discussions: cf. e.g. Watkins (1995: 19-21); West (2007: 45-54); Tichy (2010: 1-3). One can still benefit from the short book by Meillet (1923). Nagy (1974: 27ff.) introduces the basic notions which are similar in Greek and Indic metre. 141 The Norwegian school uses this as an argument to delay the invention of the hexameter to a period long after Mycenaean times. Cf. Berg & Haug (2000: 11): "If the equivalence between one long and two short syllables had been established already in an assumed pre-Mycenaean hexameter, it is rather strange that it is introduced into Aeolic metres only after the time of the Lesbian poets who stick to syllable-counting". This argumentum ex silentio is not valid. The preservation of this archaism in the Aeolic tradition can be motivated by its content. Lyric poetry is more apt to be composed in stanzic strophes with a fixed number of syllables (cf. Meillet 1923: 45). Their later forms can be influenced by the growing popularity of the Greek epic and the increasing contacts between different traditions. 142 For interesting discussions of the different proposals, cf. e.g. Sicking (1993: 70-71); Hackstein (2010: 413- 414); Miller (1982: 49-55; 2014: 80-87). Higbie (1990: 105) expresses general criticism while discussing the different proposals: "None of these views is entirely satisfactory. They run into difficulties in the choice of one break over another, because it is hard to understand why points of division which occur frequently are not given the same weight that the one selected - either in the third foot or after the fourth - has acquired". She regrets the fact that the bridges of the hexameter are not adequately understood in the different proposals, I will try to do this. Fantuzzi (1984) and Magnelli (1995) offer a critical status quaestionis of the problem until their times. Other studies will be referred to in the course of this paper.

basis of the existence of archaisms at the bucolic diaeresis argued for an original tetrameter, which was extended by the adonean cadence. His theory will be further extended during my own analysis in the last section of this chapter. Before beginning with it, I would like to briefly discuss some of the methodological points I will make use of in this final chapter. It may be apparent by now that the protohistory of the Greek epic and especially of his metre remains in darkness and that one has to be cautious to bear in mind that we are dealing with hypothetical theories. Therefore, it can only be put forward that one proposal is more plausible than the other. This caused some scepticism, especially in the Netherlands, where certain scholars totally denied the possibility to find a plausible reconstruction of the protohexameter. However, I can only agree with Bakker's remark (1995: 105 fn. 23) that "in itself the diachronic conception of the hexameter as resulting from smaller units seems highly probable, if not inevitable"143. However, some remarks need to be made. Firstly, a coalescence of smaller units can only be reasonably argued for if these smaller units are also attested in other forms of Greek poetry. Moreover, in order to be plausible, a reconstruction must be able to argue in a convincing manner how these shorter elements could be joined together into one longer verse. Nagy (1998; 2004) points to the problem that most theories about the origins of the hexameter are exclusively based on metrical facts. He extends this to the formulaic language, but even this is not enough. A plausible reconstruction needs to be able to explain the important facts of the hexameter, namely all important caesurae, the bridges and the whole colometric structure. Therefore, attention also needs to be paid to the formulae which are used by the epic singers and to the peculiarities of their composite language. The present analysis will make use of such a multilayered approach towards the protohistory of the Homeric hexameter144. As will be put forward at the end of this thesis, this needs to be extended even further in future research, e.g. combined with a broader study of the epic tradition before Homer and its relation to the historical context. This will be left out here and therefore, I will not give a definite answer to the question when the dactylic hexameter as we know it came into being.

3.1.2 Contact metrics: Borrowing a metre? We saw above that Meillet (1923) was the first to offer a well-founded comparison between the Greek and Vedic metrical tradition, but due to the above sketched peculiarities of the hexameter, he was not able to take the hexameter back to Indo-European prototypes. In addition, one of his students once asked him why the Greek language is in fact not that well suited for the epic metre, because artificial licencies such as metrical lengthening need

143 In addition, the caesurae of the hexameter are an indication if not direct evidence of these smaller units. Therefore, Fraenkel's interpretation (1968³: 117-119) of the caesurae as an innovation within the colometry of Homeric verse, partially lacks evidence. Some of them can be interpreted as innovations as will be discussed below, but others are prone to be original. 144 I agree with Peabody (1975: 19) that the origins of the Greek hexameter need to be conceptualized within the framework of "oral poetics" and that a functional approach to metre is indispensable for doing so.

to be permitted in order to be able to put every Greek word into the metre (Peabody 1975: 22)145. Therefore, Meillet reached the conclusion that the hexameter cannot be inherited from Proto-Indo-European times, but that it is in fact a borrowing from a pre-Greek culture (Meillet 1923: 63)146. In this way, the biceps procedure can be explained as a Greek innovation, which came into being due to the contractions which gradually arose in Greek grammar. Furthermore, the quantitative character of the epic metre could be emphasized in this way (Meillet 1923: 44-47; also Peabody 1975: 44 referring to Sanskrit sandhi-rules). Meister (1921: 56-58) independently reached the same conclusions. The hexameter remained so unchanged during the historical period that a reconstruction of its protohistory cannot convince him completely. In addition, he underlines the non-Indo-European etymologies of important metrical and musical terms, e.g. ἴαμβοσ, κίθαρισ, μέλοσ or ἔλεγοσ (cf. Hoekstra 1981: 34)147. The sceptical vision of these two authorities dominated the scholarly literature for half a century (cf. Fantuzzi 1984: 39). From the seventies onwards, new studies were conducted with regard to the Indo-European origins of the hexameter, but important Homerists such as Hoekstra (1981) or Ruijgh (various publications) remained convinced of its foreign origin148. Nevertheless, I can only agree with the Norwegian school that this hypothesis is totally ad hoc (Berg 1978: 12; Berg & Haug 2000: 10). There is nothing known about the metrical tradition of pre-Greek societies, we do not even know what kind of language they spoke, therefore this theory cannot be verified. In neighbouring societies, for instance in Asia Minor or the Levant, similar poetic traditions existed but none of them exhibit a similar metrical pattern such as the Greek hexameter (West 1997: 235). In addition, there needs to be some reluctance to accept the statement that dactyls are not well suited for the Greek language, because they did so in Mycenaean Greek (e.g. Chadwick 1990: 176). Typologically, the biceps procedure is also not that unusual, because it gradually developed in the Sanskrit metrical tradition as well. A further problem with the theory is mentioned by Sicking (1993: 70), who rightly asks why the Indo-European metrical heritage, the opposition between long and short syllables should be used in a borrowed metre. This criticism can even be extended to the use of inherited poetic formulae, themes, motives etc. Moreover, I hesitate to accept the principle behind this whole theory. Contact linguistics have shown that every linguistic unit, ranging from a phoneme to a syntactic structure, can

145 It was only after the independent conclusion made by Meister (1921) that Meillet accepted this hypothesis (Meillet 1923: viii). 146 For some recent discussions of the borrowing theory, cf. e.g. Bowra (1962a: 22-24); Fantuzzi (1984: 38-39); Ritoók (1987: 3); Chadwick (1990); Sicking (1993: 70); Nasta (1994: 110); Watkins (1995: 21); West (1997: 234-237); Janse (1998: 127; 2012: 5); Tichy (2010: 5). Bennet (1997: 526) seems to sympathize with it. 147 He seems to forget that a whole tradition exists of common Indo-European motives, themes etc., cf. also the paper by Mahoney (2007) who proved the existence of a common metapoetical tradition. 148 Cf. e.g. Ruijgh (1995: 8). His reconstruction of Mycenaean formulae is used as an argument in favour of it, but this in fact can only be used as an argument that the metre existed by that time, not that it was borrowed. Even this is not certain, because the reconstructions are too hypothetical and too few to offer unrejectable proof that the hexameter existed by that time. Remember also that the reconstructed verse (Ruijgh 1995: 85- 88) is holodactylic, so questions remain about the contraction procedure. Ruijgh thought the protohexameter was more dactylic than in historical times.

be borrowed as a result of bilingualism and/or language contact, but does this mean that also an entire metrical tradition could be borrowed? We must assume that at least some interaction must have occurred, judging from Meister's examples of metrical words without a clear and plausible etymology in Indo-European terms, but a complete mixture of the poetic traditions seems too far-fetched. Moreover, it does not stand up to scrutiny that the Aeolic metres can be derived from Indo-European times, but the hexameter not149. With regard to the content and their traditional context, this does not seem plausible. For example, Nagy (1974: 103ff.) points to the similar localization of the poetic syntagm κλέοσ ἄφθιτον both in hexametrical and Aeolic verse150. In conclusion, we have to reject the borrowing hypothesis.

3.1.3 Splitting up the Ionic hexameter: the main theories Again Martin West (1973a) was the first scholar who dared to propose a new model for the development of the Homeric hexameter from Indo-European prototypes, which could replace the borrowing hypothesis from the twenties. Doing so, he produced a model which combined a purely metrical analysis with a broader socio-cultural development of Greek epic151. He postulated a separation between the poetical tradition of the northern and southern part of the Mycenaean world, which forms the basis of the later Greek dialects. The northern part evolved into the Aeolic dialects and retained a more conservative poetical tradition. The southern part on the other hand, from which the Ionic and Doric tradition would evolve, created dactylic rhythms with the Greek innovation to alternate two shorts with a long syllable. Doing so, the following metres were created (West 1973a: 185):

–∪∪–∪∪–∪∪–∪∪ (dactylic tetrameter version 1) –∪∪–∪∪–∪∪–– (dactylic tetrameter version 2) ⨯–∪∪–∪∪–– (paroemiacus) –∪∪–∪∪– (hemiepes)

From these new traditions, the hexameter could eventually be created from the combination of a hemiepes followed by a paroemiacus, coming into being around 1100 BC152. West does

149 Cf. also Peabody (1975: 21), who finds it implausible that a metre would be borrowed in the context of a traditional and oral style. 150 Cf. also Whitman's foreword to Nagy's PhD (1974: xii): "Given the known Greek tendency towards the quantitative fixation of elements in the inherited verse-forms of Indo-European, (…), it seems reasonable to suppose that the hexameter arose in Greece from extension and fixation of elements already given in Indo- European prosody". 151 For a discussion of the theory, cf. e.g. Berg (1978: 17ff.); Hoekstra (1981: 34ff.); Fantuzzi (1984: 40-42); Edwards (1986: 174-175); Ritoók (1987: 4); Nagy (1998: 500-502; 2004: 151-152). Haslam (1976: 202) seems to accept the theory. West (1997: 236-237) briefly summarizes his theory. 152 For a similar point of view cf. Kirk (1985: 16) "The heroic poetry of the eleventh century B.C., probably already expressed in comparatively developed dactylic hexameters, may well have consisted, as in most other oral heroic poetry, of short and primarily whole-verse sentences; and similes and speeches are likely to have

not want to pretend that the hexameter received its definite form by that time, for the structure was rather loose, which can be observed in the remaining unmetrical verses in Homer (cf. West 2018). The dactylic poetry written by Stesichorus retained such a loose tradition even after the creation of Homer's monumental poems (West 1973a: 188). One must admit that West is the only scholar who could offer a plausible sociocultural context for a presumed origin of the Greek hexameter. Nonetheless, there are serious flaws in his theory, which is in fact an adaptation of two old theories initiated by Bergk (enoplius + paroemiacus) and Usener (two paroemiaci). Therefore, Hoekstra (1981: 34ff.) devotes some pages to these different theories in his book about "Epic verse before Homer". Some of them are valuable, although Hoekstra seems to adhere to the borrowing hypothesis, which is equally untenable as we saw above153. For instance, he questions the possibility of a division between North and South Mycenaean, because we cannot observe real differences between them in visual art (Hoekstra 1981: 35). In this case, I hesitate to pay much attention to art, as it is of a different nature than language and literature. Moreover, somewhere a distinction needs to be made between the different dialects and poetical traditions. Equally questionable is his criticism that there is no further regularization of the hexameter after Homer and that we therefore cannot take it for granted that it did so before. This is not true, for the hexameter further regularized afterwards, e.g. some tendencies such as Meyer's bridge (cf. infra) were further regularized in later authors. Another example is the fact that whereas Homer made use of every possible variation of the dactylic hexameter, the possibilities were largely reduced in later authors such as Callimachus, Apollonius, Nonnus etc. (e.g. West 1982: 152-157; 177-180). On the other hand, I acknowledge that Hoekstra (1981: 36) is right in his doubts about the validity of the unmetrical verses which are cited by West to underline his theory. Let us consider the following example:

πολλὸν ὑπεκπροθϋει, φθϊνει δϋ τε πᾶςαν ἐπ᾽ αἶαν (Il. IX, 506). ("many she [sc. Atè] outstripts, she is first over whole the earth").

West explains this as an unmetrical verse, because φθϊνει normally has a short <α> /a/, which would lead to an unmetrical verse with a trochee in the third foot of the hexameter, a ςτίχοσ λαγαρόσ as it was named in Antiquity (cf. e.g. Meister 1921: 42-44). However, Hoekstra rightly utters the possibility that due to an original digamma the alpha became long after the third compensatory lengthening in the Ionic dialect (confirmed by Mehler & Mehler

been similarly rudimentary." In earlier publications, West interpreted it as a combination of a pherecrateus and a reizianus (cf. Ritoók 1987: 4). 153 This becomes apparent, because he regrets the fact that the Mediterranean hypothesis is not mentioned nor refused in West's paper. He further underlines the importance of the Pre-Greek etymology of musical instruments etc. and he expresses his doubts about West's presupposition that epic poetry would have evolved from balladic poetry. Hoekstra thinks that there must have been already full epic poetry around 1200 BC (Hoekstra 1981: 34). On the other hand, he needs to admit that he cannot completely refuse that the hexameter could have evolved from the coalescence of two shorter metres (Hoekstra 1981: 38).

196813: 887)154. Moreover, this specific form is attested only twice in the corpus, the other example being Il. XXI, 262 (χώρῳ ἔνι προαλεῖ, φθϊνει δϋ τε καὶ τὸν ἄγοντα.). As can be observed, φθϊνει is placed here exactly at the same place in the verse and is again followed by the postpositive words δϋ τε155. Therefore, formularity will also have played an important role in this lengthening which can also put doubts on West's chronology. He interprets verses which exhibit a metrical flaw in the middle of the verse as archaisms in the colometry of Homeric verse, but this is mainly caused by his static vision that the caesura can only be located in the middle of the verse. When we interpret them as (partial) innovations, as I will argue for in the next section, these unmetrical verses can be interpreted the other way around156. Middle caesurae (especially 3b) became more prevalent in the verse, which leads to the fact that the poet begins to think in cola which end and begin in that position and doing so, some unmetrical syllables can arise (cf. infra for more discussion). These middle caesurae are the main problem with West's theory, because he has to be able to explain them both. In fact, his theory can only explain the penthemimeral caesura and to be able to equally explain the trochaic caesura, he has to postulate different forms of his cola, which makes it a theory ad hoc (cf. Berg 1978: 21; Hoekstra 1981: 39)157. Moreover, caused by his static colometry, he does not pay any attention to the prevalence of other caesurae in the verse, such as the bucolic diaeresis nor to the bridges. The Italian metrical school of Gentili, Giannini and their followers, attempted to surpass the general problem of West's theory: how can we explain the different caesurae in the middle of the verse? In order to do so, they produced what they call a "polygenetic model". In their opinion, it is not possible to develop one model for the origin of the hexameter which can fully convince anyone, therefore it seems more plausible that different combinations of shorter verses have contributed to the development of the epic verse158. The publication of new fragments of Stesichorus (7th-6th century BC) in 1976 by Ancher, Boyaval and Meillier, formed the impulse for their publication, because they can prove, in their opinion, that the hexameter developed from dactylo-epitrite metres (Gentili & Giannini 1977: 7; 1986: 45; 1995: 12). Hence, the hexameter needs to characterized as the combination of two cola, as was already observed by Aristotle (Metaphysica 1093a26),

154 Also in Frisk (1970: 1011-1012 s.v. φθάνω) and Chantraine (1999²: 1197 s.v. φθάνω). 155 They are a strong argument that there needs to be a caesura in this verse at 3a. 156 On this point, I agree with Berg's (1978: 21) theory that the middle caesurae are later innovations. 157 The variants are mentioned in West (1982: 35). Some further criticisms: Berg (1978: 19): He is not able to explain the preference for spondees in the beginning of the verse and Wernicke's law. Nagy (1998: 500-502): West is not able to explain how the two cola could be joined to one unit. Weilo & Haug (2001: 130): West's hypothesis resembles too much the real hexameter, by which almost every hexameter can be interpreted as a protohexameter. Steinrück (2005: 482): There is a contradiction in West's theory, because he interprets the hexameter on the one hand as a combination of two cola, but on the other hand wants to divide it in the traditional six feet. 158 For a short discussion of their theory, cf. Fantuzzi (1984: 53-56); Edwards (1986: 175-176); Ritoók (1987: 6). Their theory was first described in their joint paper in Italian (1977), which was reprinted in the volume edited by Fantuzzi & Pretagostini (1995). They also published an English translation of their article (1986). In order to aid the reader, references will be given to the three editions.

shorter units which can also be observed in archaic inscriptions which are the product of a diffusive, oral context (1977: 18; 20-23; 1986: 55; 58-61; 1995: 22; 24-27)159. In doing so, their theories show many similar points with West's theory, or the earlier theories of Bergk and Usener, for instance because they also point to the importance of the hemiepes and paroemiacus as important contributing factors in the development of the Homeric metre. Moreover, Giannini's input in the article (1977: 38ff.; 1986: 76ff.; 1995: 42ff.) focuses on the linguistic arguments in favour of this theory, mainly underlining the fact that hiatus and brevis in longo occur at the traditional main caesurae, which can account for their theory to interpret the archaic hexameter as the joining together of two shorter metrical cola. Because of the similarities between the Italian theory and West's proposal, the criticism one can object against it, is also of the same kind. I can again refer to the fact that only the medial caesurae are adequately explained by it, but not the bucolic diaeresis, the bridges etc. Likewise, their inscriptional evidence contains the same uncertainty as West's unmetrical verses: are they necessarily an archaism or can they be caused due to the growing importance of the medial caesura over time? Interesting in this case, is the remark made by Ritoók (1987: 12):

"Der Umstand, daß in den Formen Enoplios, Paroemiakos, Hemiepes, T1 und H1 keine archaischen Formeln zu finden sind (oder nur durch Erweiterung entstandene), daß diese also nicht nur in metrischer Hinsicht, sondern auch von den Formeln her gesehen nicht als elementare Gebilde betrachtet werden können"160.

Both West, the Italian school and, as we will see, the Norwegian school depart from metrical cola which are not the most archaic one and this fact casts doubt on their theories. Moreover, as Nagy (1979: 627; 1998: 502; 2004: 151) points out, the Italian polygenetic model is too vague, because it can yield many more possibilities than are actually found in real hexameters. The theory tries to explain everything, but on the other hand it explains nothing. To counter the criticisms one can object against West's theory and the polygenetic model, Berg (1978) proposed a new hypothesis which does not depart from the prevalence of word end at the middle caesurae, but from the hephthemimeral caesura, explaining the hexameter as the combination of a glyconeus and a pherecrateus (Berg 1978: 21)161. Doing

159 I agree with their approach towards the hexameter as the combination of different cola, which is indeed plausible with regard to the oral context they are composed in. A similar conclusion is reached by Steinrück (2005), who equally pays attention to the inscriptional evidence and the unmetrical examples they offer. However, cf. infra for criticism. 160 Witte's theory and my further development will offer an alternative approach that can resist this important criticism. In fact, it is quite remarkable that Ritoók (1987) does not cite Witte's proposal, which would agree with his own conception of the hexameter. 161 The middle caesurae are later developments in his opinion. He interprets the remaining part after the bucolic diaeresis as too short to be an original separate verse, but this argument is not valid, because the adonean is attested as a separate verse in the Aeolic tradition (cf. e.g. West 1982: 30; 177). His hypothesis of a coalescence of a glyconeus and a pherecrateus is corroborated by the fact that this combination is actually

so, he explains the hexameter as an Ionic specialisation of more archaic Aeolic metres, especially because of the biceps procedure, which needs to be interpreted as an Ionic contribution to the Greek metrical tradition162. As said above, the Norwegian school interprets the completion of the hexameter as we know it as a post-Homeric development (cf. Berg 1978: 20). To explain how an original distich of a glyconeus and a pherecrateus could eventually evolve into the Greek epic verse, Berg (1978: 26-27) needs to postulate different intermediate stages, which he coined as the "catametrionising" procedure, and which can be summarized with the following diagram:

a) ⨯⨯–∪∪–∪– b) –∪∪–⨯⨯⨯⨯ ⨯⨯–∪∪–– (3 different versions of original distich) c) ⨯⨯⨯⨯–∪∪– ⨯⨯⨯⨯–∪∪–∪∪–∪∪–– (dactylic fixation at the end) –––––∪∪–∪∪–∪∪–– (spondees arose at the beginning) –∪∪ | –∪∪ | –∪∪ | –∪∪ | –∪∪ | –⨯ (definitive hexameter with biceps)

As can be observed, Berg encounters a similar problem as West's proposal. The Norwegian scholar wanted to avoid the problem of the medial caesurae, he did not want to postulate two different versions for the origin of the hexameter, but by using the glyconeus as the beginning of the hexameter, he has to postulate even three original distichs! Moreover, is it really realistic to suppose that four intermediate stages need to be presupposed to explain how the hexameter eventually arose163? A further problem with the theory lies in the ancipitia which Berg hypothesizes. In the second version of his distich, he has to place them at the end of the glyconeus, although it is a common characteristic of Indo-European verse that the beginning of the verse may be somewhat loose, but the end is clearly fixed. Similarly, in Byzantine dodecasyllables and decapentasyllables the stress is regulated only before the caesura and at verse end (e.g. West 1982: 182-185). In addition, it does not seem very plausible that at the end of the second millennium BC, Greek has preserved metres with four ancipitia one after the other. Berg (1978: 29) assumes them because the prevalence of spondaic sequences at the beginning of the verse (he refers to O'Neill 1942: 159; cf. also

attested in Greek literature and the similar amount of syllables (15 syllables and the hexameter has an average of 15 2/3 syllables) (Berg 1978: 23-24). However, starting from the hephthemimeral caesura constitutes a first problem with the theory, because Bassett (1917: 85-86) convincingly argued for the minor importance of this caesura in Homeric colometry (cf. also Magnelli 1995: 122-123; Kiparsky 2018: 23). In addition, he does not pay sufficient attention to the formularity of the Homeric language. 162 For a discussion of his theory, cf. e.g. Ritoók (1987: 4); Berg & Lindeman (1992: 191); Janko (1992: 10); Magnelli (1995: 118-124); Haug (2002: 26-28); Hajnal (2003: 220-225); Hackstein (2010: 413-414); Tichy (2010: 6-17); Weilo & Haug (2001); West (2011b: 157); Miller (2014: 334-335); Kiparsky (2018: 23). Interestingly, Fantuzzi (1984) does not pay attention to the theory in his status quaestionis about the proposals of the seventies. 163 Cf. Kiparsky's (2018: 23) criticisms: the big problem is how to account for their coalescence, there are too many arbitrary intermediate stages.

Meister 1921: 7), which cannot be explained on the grounds of the other proposals, but his own explanation is totally ad hoc. The same can be said about his proposal for a historical interpretation of Wernicke's law and Hermann's bridge (Berg 1978: 29-30)164. He explains the first one by stating that two shorts are longer than a long syllable by position. Therefore, only a naturally long one could substitute them in his opinion, an argument which is somewhat denied by the general biceps procedure of the hexameter, although there is a general tendency of spondees to be long by nature in the second part of the foot. Hermann's bridge on the other hand, can be caused by the fact that word end after the first syllable of the pherecrateus would be somewhat surprising. This cannot be completely rejected, but let us not forget that there are some important word-breaks after the first syllable of the hexameter (1a). Berg equally tries to explain the metrical lengthenings of the Homeric Kunstsprache as results of the original combination of a glyconeus and a pherecrateus, but this cannot account for all the cases. Moreover, Vedic also lengthens some of its syllables only metri causa, because of the necessity of the metre, the so-called plūti-forms. In addition, he does not believe in the influence of the contraction processes of the Greek language on the biceps procedure, because this does normally not result in spondees. Here, Berg (1978: 33) makes a common mistake, for he forgets the importance of the synapheia for the Greek verse. These newly contracted forms may not evolve into spondees in daily Greek speech, but they do so when they are combined with other words in the continuous flow of Greek verse. Although it may be apparent by now that Berg's hypothesis about the origins of the epic hexameter has to be rejected, his theory gained some influence at the end of 20th century and the beginning of the present one. First of all, this can be observed in the subsequent writings of the Norwegian school. Berg and Lindeman (1992: 186-193) searched for a morphological argument in favour of the theory, referring to the dactylic paradigm of ἀνήρ in some plural forms, e.g. ἄνερεσ with /a:/ (cf. also Berg & Haug 2000: 14). The regular nominative plural ἄνδρεσ would be perfectly possible due to the synapheia of the dactylic hexameter, so why is this artificial paradigm created? Because of its preponderance in the first and the fourth foot (37,9%; 46,3%), they can be interpreted as caused by the ancipitia which Berg postulates in his theory. As I will discuss below, they can also be explained with reference to Witte's theory, their preponderance for the fourth foot being a strong argument for the coalescence of an original tetrameter and an adonean. The subsequent paper by Berg and his pupil Haug (2000) repeats their arguments and argues for the postulation of a later dating of the hexameter, after Homer165. Therefore, this verse is not a medium of oral poetry in their opinion, which seems highly unlikely in view of the traditional character of its content and linguistic form (cf. supra) (Berg & Haug 2000: 12-13). After Berg's death in

164 Wernicke's law states that a spondee in the fourth foot can only be produced by a syllable which is already long in itself. The strong avoidance of word end at 4b of a dactylic fourth foot was discovered by Hermann and named after him Hermann's bridge (Oswald 2014: 421; cf. infra for more discussion). 165 Cf. also Berg & Lindeman (1992: 185) "A fundamental flaw in Lubotsky's line of argumentation, which he regrettably shares with many others, is finally the unfounded presupposition of an unchangeable epic verse (i.e. hexametric poetry in the Mycenaean age - or even before)".

2000, his pupils dared to utter some criticism against it, pointing to the inadequacy of his "Katametrionisierung" and the minor importance of the hephthemimeral caesura (Weilo & Haug 2001: 131-132). Secondly, Berg's hypothesis found a follower in the person of the German scholar Eva Tichy166. Only three years after the publication of his theory, she published a paper that on the one hand was intended to corroborate his thesis, but on the other hand also proposed some adaptations to the original theory. In doing so, she focused on the supposed remnants of vocalic /ṛ/ in the Homeric corpus and most notably on the scansion of the word ἀνδροτῆτα in the Iliad, which occurs in the following verses:

ὃν πϐτμον γοϐωςα λιποῦς᾽ ἀνδροτῆτα καὶ ἥβην. (Il. XVI, 857; Il. XXII, 363) ("bemoaning his destiny, leaving his manliness and his youth") Πατρϐκλου ποθϋων ἀνδροτῆτϊ τε καὶ μϋνοσ ἠΰ. (Il. XXIV, 6) ("longing for Patroclus, his manliness and his valorous might")

Due to limitations of space, I will limit myself to a concise overview of the problem167. When one tries to scan the above verses, one will encounter some problems, because if we would scan ἀνδροτῆτα in the normal way, namely –∪–∪, this would not fit the hexameter as we know it. Therefore, scholars tried to explain this form as a preservation of a vocalic /ṛ/ from Mycenaean times, because it derives from /*h2nṛtātṃ/ (cf. Tichy 1981: 46). In that case, ἀνδροτῆτα can be scanned as ∪∪–∪, which fits the hexameter. Her own explanation is ad hoc, for she thinks that we need to preserve the trochaic sequence of ἀνδρο-, as the correct reading which could corroborate Berg's thesis that the hexameter is a late creation on the basis of a glyconeus and a pherecrateus (Tichy 1981: 59). However, if this were correct, than Berg's principle of "Katametronisierung" collapses, because it was his idea that the fourth foot was the first one to be stabilized in the formation of the hexameter. In this case however, the fourth foot contains a trochaic sequence and can therefore not be the place that was first stabilized in the formation of the hexameter. Hence, she proposes a kind of polygenetic model of Berg's theory with different pairs of glyconei and pherecratei which could explain the different forms of the hexameter and by means of which she does not need to postulate his catametrionising (Tichy 1981: 60ff.). Doing so, she makes the same mistake as her Italian colleagues: you can explain everything but on the other hand nothing with such models. Moreover, her theory is based on a counterintuitive colometric analysis because she pays too much attention to the existence of word break at the hephthemimeral caesura in her theory. When we apply the cognitive principles which I presented in the second chapter,

166 Cf. her admiration (Tichy 2010: 6): "Für Bergs Hypothese über den Ursprung des Hexameters spricht vor allem, dass es die Verse, von denen er ausgeht, wirklich gibt, und nicht nur das: Sie sind teil des von Wilamowitz aufgedeckten Systems der choriambischen Dimeter, gehören damit in die älteste Schicht der griechischen Metrik und erscheinen in Zeugnissen griechischer Dichtung zu aller Zeit und in weitester Verbreitung". 167 There are also problems with the derivation of this word and the accent, interested readers are referred to Tichy (1981) and Barnes (2011).

it seems more natural to posit only a trochaic caesura which divides the verse in two chiastic parts with a participle and a direct object168. Thus, the verse consists of two cola divided by the trochaic caesura and therefore does not adduce any evidence in favour of the importance of the hephthemimeral caesura and hence the Norwegian hypothesis. Moreover, in the other verse (Il. XXIV, 6) ἀνδροτῆτα is placed in the middle of the verse, which does not have any explanatory force for Berg's proposal at all. Last but not least, when we accept van Beek's (2013: 192-230) proposal that /ṛ/ needs to be interpreted as an inner-epic development (cf. supra), her argument loses every probability169. Nonetheless, she remains convinced by the theory until now and she even tried to reconstruct some "original" parts of the Iliad (Tichy 2010), reducing the hexameters to fifteen syllable verses, which is, as West (2011b: 163) rightly declares, "wasted ingenuity"170. Her methodology is totally at random, she deletes some short words from the text or adds some in order to have a verse of fifteen syllables, which fits into Berg's proposal (cf. Tichy 2010: 18ff.). Let us consider for instance, the following two verses from the Glaucus-episode in Iliad XI:

Ζεὺσ δὲ πατὴρ Αἴανθ᾽ ὑψύζυγοσ ἐν φϐβον ὦρςε. ςτῆ δὲ ταφών, ὄπιθεν δὲ ςϊκοσ βϊλεν ἑπταβϐειον. (Il. XI, 544-545) ("Father Zeus, sitting high, incited fear for Aias. Astonished he [sc. Aias] stood up and he threw his shield of seven-bulls hides behind.")

These are the original verses from the Iliad, but in order to make them look as a fifteen syllable verse, Tichy alters them towards:

Ζεὺσ δὲ πατὴρ Αἴαντι ὑψίζυγοσ φόβον ὄρςε· ςτῆδὲ ταφών, ὄπιςθε δὲ ςάκοσ ἀμβάλετ' εὐρύ.

Αἴαντι with hiatus replaces original Αἴανθ᾽, the tmesis with ἐν is deleted, ὦρςε loses its augment, although there is no metrical necessity for it, the anapestic sequence ὄπιθεν is replaced by the amphibrach ὄπιςθε and the Aeolic apocopated form ἀμβάλετ' εὐρύ is put in place of βϊλεν ἑπταβϐειον. Such endeavours are nothing more than a hypothetical play for

168 She admits that the hephthemimeral caesura is in this case clearly secondary to the trochaic caesura herself (Tichy 1981: 58). 169 Another hypothesis was promoted by Barnes (2011), who is not convinced by Tichy's proposal that they are metrical archaisms, because she does not argue sufficiently why they need to be (Barnes 2011: 8). However, his own thesis that the formula is based on a remake of ἀμβροτῆτα and as such related to an Avestan formula hauruuātā amərətātā ("wholeness (and) not-dying") is equally unsatisfying. The many intermediate stages he has to postulate make it a non-economic theory and there are some semantic problems (cf. van Beek 2013: 211, fn. 815). Van Beek also criticizes Tichy's approach throughout his PhD. 170 West (2011b) focuses on the methodological problems in his review of the monograph. It is not done to distinguish between old and modern parts of the diction. In fact, she does the same as Fick, who wanted to reconstruct an original Aeolic Iliad. Such attempts have never been convincing.

Homerists who want to display their knowledge of historical grammar. There are no concrete arguments why one would delete one word and not the other, every example is in fact a circular argument. She starts from the transmitted Homeric text, makes a version of her own and tries to explain the transmitted text with her own proposition. Moreover, she does not pay any attention to the colometry of her verses. The most striking example is when she tries to recover a lyrical version of a dialogue between Achilles and Apollo (Il. XXI, 1-32) and "restores the original distichs". Doing so, she must end her verses many times in the middle of a syntactic or semantic unit, although it is a common characteristic of Indo- European verse that verses mostly coincides with a sentence (cf. e.g. West 2007: 47). Remark for instance the following example with Tichy's German translation (2010: 127- 129):

ἔβλαψάσ με, Φοῖβε, θεῶν "Getäuscht hast du mich, Phoibos, ὀλοϝώτατα πάντων, von allen Göttern der grausamste,"

Her distich needs to separate θεῶν from its syntagm, which is unlikely on a cognitive level. In addition, it is quite remarkable that, when we look at Tichy's German translation, this problem is suddenly solved. When she writes in her native language, she is not inclined towards such counterintuitive cola. The same can be said about her remark that her attempts can be compared to the dialogues of Ṛgveda, in these Indian texts, almost every verse is a single colon (intonation unit) or a single sentence.

3.1.4 The extension of a metre: Nagy's proposal The above presented theories shared some similar shortcomings, which were mostly caused by the fact that they need to divide the hexameter in different parts. How can the two middle caesurae be explained if you think they are the original breaking places or does it seems plausible that the hephthemimeral caesura, which is not one of the most prevalent stops, would be the original breaking point of the Greek epic verse? In order to avoid these problems, Nagy (1974), in his PhD about Greek and Indic metre, made use of another basic principle of Greek verse to explain the origins of the Greek hexameter as a development of Indo-European prototypes, namely the internal expansion of a shorter verse into a longer verse171. Using this approach, he argued for a revival of an old theory of Wilamowitz (cf. Nagy 1974: 7; 10), in other words that the hexameter came into being as the regularization

171 The principle of internal expansion in described at some length in Nagy (1974: 37-48). A short summary is for instance also provided in West's (1982: 32) metrical handbook. Based on the criticisms which were objected against his PhD, Nagy further discussed the problems in Nagy (1979; 1998). The most recent article was also reprinted in his book "Homer's Text and Language" (Nagy 2004: 144ff.). Nagy (1974: 57) refers to the central problem for West's proposal. For some discussions of his theory, cf. e.g. Berg (1978: 17-18); Fantuzzi (1984: 42-46); Gentili & Giannini (1977: 29-32; 1986: 68-70; 1995: 33-36); Edwards (1986: 175); Nasta (1994: 110-111). His whole book is reviewed by West (1974) and Haslam (1976).

of a pherecrateus which was expanded by three dactyls (pher3d). This can be summarized in the following scheme (based on Nagy 1974: 47):

⨯⨯–∪∪ –⨯ (simple pher) ⨯⨯–∪∪ –∪∪ –⨯ (pherd) ⨯⨯–∪∪ –∪∪ –∪∪ –⨯ (pher2d) ⨯⨯–∪∪ –∪∪ –∪∪ –∪∪ –⨯ (pher3d)

This last one can be interpreted as the basic model of the hexameter, when we accept two important adaptations (Nagy 1974: 49-50)172. Firstly, the biceps procedure must have come into being, so two short syllables could be substituted with a long one. This might have arisen as a result of the development of contracted forms in Ionic. Secondly, the Aeolic basis at the beginning of the verse could be changed into a spondee, which later on could be alternated with a dactyl. This would explain the preponderance of spondees in the first foot of the verse173. In the rest of his contribution, he takes the caesurae into consideration and tries to explain them on the basis of his theory. In order to do so, he profits from Parry's theory about caesurae, for Parry (1971 passim) had already proven that there is a close relationship between the caesurae of the hexameter and the places where formulae begin and end (Nagy 1974: 14). Important to note in this case is the fact that Nagy interprets the formulae on the one hand, as a diachronic element which generated the metre they are used in, but on the other hand, emphasizes that formulae are also synchronically motivated by the metre they once shaped (Nagy 1979: 617). In order to explain the caesurae, he points to the fact that many formulae can be explained as the expansion of shorter formulae which once began at the bucolic diaeresis and the hephthemimeral caesura and that these metrical cola were also copied in the beginning of the verse (Nagy 1974: 61-64)174. In this manner, he wants to explain the importance of the bucolic diaeresis, the existence of Hermann's bridge and the preference for dactylic structures in the fourth foot. Although Nagy's proposal was (rightly) criticized in the reviews of West (1974) and Haslam (1976), I first want to stress the important methodological advantage of his theory,

172 Some further arguments in favour of it, is the number of syllables, because sixteen syllables is also the number of the Sanskrit epic śloka and close to the average of 15 2/3 syllables of the definitive hexameter. Moreover, this kind of pherecrateus is actually attested in Alcaeus (7th-6th century BC) (Nagy 1974: 49). The adaptations are also mentioned in Nagy (1979: 619). In this second paper, he also pays attention to the new Stesichorus fragments (cf. supra). 173 Cf. Nagy (1974: 55): "These facts suggest that the 1st foot of the epic hexameter is not by origin a dactyl (– ∪∪) but a spondee (––)". 174 This issue is discussed in Nagy (1974: 56ff.). I am satisfied that he pays attention to the importance of the trithemimeral caesura and to his acknowledgement of flexible formulae as a contributing factor in the development of the metre. Remark the similarities with my own theory below.

when we compare it with the other ones175. Who is better suited to declare this than Nagy (1998: 495; 2004: 145) himself?

"The methods of those who restrict their perspectives to quantitative metrics cannot succeed, I argue, in any attempt to arrive at a complete picture of the hexameter in its full diachrony".

The advantage of his theory is the fact that he gives some attention to formulae and their importance for the history of the epic form (cf. Haslam 1976). On the other hand, he does not go far enough with this approach, mainly because he does not include arguments from other fields of linguistics, although he declares himself (1998: 503; 2004: 153) that this is equally important. Moreover, Berg (1978: 17-18) is right in his criticism that Nagy's proposal runs into chronological problems. He wants the hexameter to be an old metre, but this is impossible while referring to internal expansion, which is a later phenomenon of Aeolic metres, caused by the influence of the stichic Ionic metres. Another important objection is mentioned by West (1974: 457) who asks why the formulae would cause these caesurae in the dactylic hexameter, although they are also used in similar lyric metre but do not cause the existence of caesurae there. In fact, the same can be said about the bridges. Nagy tries to explain Hermann's bridge with his theory, but this would only stand up to reason if this also existed in the corresponding lyric metre, but this is not the case. In addition, Nagy's hypothesis tries to explain the preponderance of spondees in the first foot, but he cannot account for the fact that they are also prevalent in the second foot and occur even a little more in that position (cf. Meister 1921: 7). In addition, there are some doubts about the whole concept of internal expansion. It does not seem probable that epic literature could have started from a simple pherecrateus that only contains seven syllables. Moreover, how do we have to conceptualize an epic tradition which several times extended its metre?

3.1.5 Splitting up the Ionic hexameter: the minor theories The proposals of West, the Italian school, the Norwegian school and Nagy are the main contributions, which arose during the second half of 20th century, to our knowledge about the protohistory of the epic hexameter. In this short section, I will briefly discuss the theories of Peabody (1975), Vigorita (1977) and a rather casual remark of Irigoin (2004) about the origins of the Greek hexameter. Peabody conducted some research about Indo- European metrics in his study about the oral tradition of Hesiod, called "The Winged Word"176. He mainly focuses on the Indo-Iranian and Greek tradition, underlining the development which the Greek tradition has undergone when compared to the rather

175 Apart from Nagy's proposal for the origins of the hexameter, they concentrate on his discussion about the localization of the poetic formula for "imperishable fame" (cf. supra). West mentions two rather odd criticisms, first that Nagy's bibliography would be too large and that his discussion of ἄφθιτα μήδεα εἰδώσ (Nagy 1974: 265-278) is influenced by the "American propensity for the obscene". 176 For a discussion, cf. e.g. Fantuzzi (1984: 46-49); Edwards (1986: 175-176).

"primitive" form of early Indo-Iranian poetry. His theory shares some similar points with the Italian school because he also makes use of a kind of "polygenetic" model. He claims that both dimeter and trimeter couplets contributed to the formation of Greek hexameter. In this case, the hexameter can be divided into parts of 11+12, 10+13 (dimeter) or 7+7+9 (trimeter) (Peabody 1975: 47)177. It is his attempt purpose to explain all the caesurae of Homeric verse and therefore he approaches the problem with a polygenetic model. The different caesurae are to be interpreted as original breaking points, but he does not want to attribute more importance to one caesura than to the other. To further argue for it, he underlines the existence of hiatus and brevis in longo before the caesurae, which is again a striking similarity with the approach of the Italian school (Peabody 1975: 50-52). His theory therefore, merits the same criticism as his Italian colleagues. He wants to explain all the caesurae, but in fact he does nothing more than to postulate different origins of the hexameter. It looks as if he does not know himself where the hexameter could come from. Moreover, he does not pay attention to the important bridges or the colometry in his discussion about the origins of the metre. He equally neglects the importance of the Homeric formulae (Fantuzzi 1984: 48-49)178. Equally problematical is the theory initiated by Vigorita (1977), who aims to revive a theory of Bergk (1854) that the hexameter originated from a "heroic couplet" (cf. also Ritoók 1987: 4). In order to do so, she starts from a holodactylic hexameter which represents the original form in her opinion and argues on the basis of the penthemimeral caesura that the hexameter originated from the coalescence of a heptasyllabic verse with a decasyllable179. Because of its preponderance in the traditional poetry of Hesiod, she thinks this penthemimeral caesura to be more original than the trochaic one, although she must admit that the latter is preferred in the extant corpus of Greek hexametrical verses (Vigorita 1977: 289). In addition, her argument focuses on comparative material, most notably the decasyllables found in Vedic and Serbo-Croatian verse180. These verses normally have a caesura after the fourth or the fifth syllable (Vigorita 1977: 291-292), which is represented in the following scheme: ⨯⨯⨯⨯||⨯||∪∪–⨯. From this model, she derives the end of the hexameter, first by referring to the comment made by Meillet (1923: 23-24) that a succession of three light syllables is strongly avoided in Indo-European metrics. Therefore, the sixth syllaba anceps naturally becomes a long one, and results therefore into a dactylic rhythm, which can be generalized in regressive direction. Doing so, we arrive at the following sequence: ⨯⨯–∪|∪|–∪∪–⨯, which can explain the quantity of the syllables but not

177 The numbers refer to a mora, which counts for a short syllable in his theory. A holodactylic hexameter varied between 22 and 24 morae and due to the contractions, spondees came into being. For a schematic overview of the possibilities, cf. Peabody (1975: 50). 178 Fantuzzi (1984: 48-49) regrets also Peabody's neglect of ancient metrical theory, but cf. supra. 179 Magnelli (1995: 113) casts some doubt about her presupposition of an original holodactylic hexameter, but as Vigorita (1977: 288) rightly says, this is more probable because of the almost completely lack of holospondaic verses and the fact that it is easier to explain how spondees arose from dactyls than the other way around. 180 She adopts this from Jakobson's study, cf. also Nagy (1974: 6).

the hephthemimeral caesura and Hermann's bridge. In Vigorita's opinion, this can be explained as a re-analysis of the sequence as a pherecrateus with dactylic rhythm which would make it implausible that a caesura at 4b could exist, moreover because this would imply the sequence of two monosyllables. Therefore, the caesura was changed to 4a and Hermann's bridge came into being (Vigorita 1977: 293). As the attentive reader can observe, there are some positive ingredients in her study: her hypothesis can account for the importance of the bucolic diaeresis at the end of the verse and she pays attention to Hermann's bridge. On the other hand, her study contains striking shortcomings. Firstly, she is not able to explain how the trochaic caesura came into being and why the penthemimeral caesura needs to be more archaic. Additionally, her comparison with Vedic and Serbo- Croatian verses is not valid. We saw above that decasyllabic verses are not part of the original Indo-European prototypes. They are later evolutions, so we cannot be sure if Greek could have inherited them in this form and with these particular caesurae. There are furthermore problems with the fact that she pays much attention to the elegiac distich in her study about the origins of the epic metre. She interprets this distich as a quatrain of seven, ten, seven and seven syllables. However, as Magnelli (1995: 114-115) points out, most elegiac poets, except for Xenophanes, make use of the trochaic caesura in the hexameter line. Moreover, the pentameter is more likely to be a later development based on the combination of two half hexameters than part of an ancient couplet. These criticisms could even be extended, and therefore I fully agree with Magnelli (1995: 111-112) that her paper offers an ingenious hypothesis but "lasciano tuttavia adito a numerosi dubbi". This is a fortiori valid for Irigoin's (2004: 9) remark that the hexameter could be the combination of two tripodies, for which he can give only one argument: they are also attested in choral lyric. One can ask why then there is no caesura at 3c. The Indian śloka- verse arose from two shorter verses of eight syllables and therefore, a caesura remained at this place. Needless to say that this hypothesis cannot account for all the caesurae of the verse, the bridges etc.

3.1.6 Anaclastic hexameters: Kiparsky's vision After the different proposals of the seventies, scholarship mostly focused on criticisms against these theories and the search for further arguments for some of them. Therefore, it lasted a considerable time until a completely new hypothesis about the origin of the Greek hexameter was put forward. The Finnish Indo-Europeanist Paul Kiparsky presented a new one during the Munich colloquium about Indo-European metrics, which was held in 2013181.

181 It took some time before the editorial volume was finally published, namely only in April 2018. Due to the late publication, I was not able to consult the definitive edition. Kiparsky's paper can also be found in an online edition at the following website: https://web.stanford.edu/~kiparsky/Papers/hexameter.pdf. References will be made to the pages of this online edition. The definitive version can be found in Hackstein & Gunkel (2018: 77-128). I have no knowledge of a recent discussion of his theory, so all criticisms will be my own.

Kiparsky focuses on a neglected characteristic of the common Indo-European metrical tradition, the anaclasis technique which was presented in the preceding chapter as a feature of Greek verse182. In the first part of his paper, the author highlights its Indo- European character, because it is not only found in the Greek tradition, but also in Vedic, Classical Sanskrit and Middle Persian. Interesting in this case, is the almost complete absence of this aspect in other metrical traditions (Kiparsky 2018: 6-21)183. Afterwards, he tries to explain the dactylic hexameter as the evolution of an iambic metre, which would reveal the hexameter as a development from the most characteristic rhythm in ancient Indo- European poetry. He starts from a distich of two iambic tetrameters which due to the anaclasis procedure can produce a dactylic rhythm which evolved into the re-analysis as a metre of its own184. This can be observed in the following scheme (based on Kiparsky 2018: 25-26):

∪–∪–⨯–∪C / ⨯–∪–⨯–∪– (distich of iambic tetrameters, "C" refers to catalexis) –∪∪–∪∪–– / –∪∪–∪∪–∪ (creation of dactylic rhythm due to anaclasis)

As can be observed, long syllables have the possibility to evolve into short ones and short syllables into long ones (they can also preserve their original quantity). The catalectic first part is optional and therefore it can become a long one after the anaclasis occurred. In order to become a hexameter, the distich evolved into one stichic metre. Kiparsky then had to show how the caesurae and the bridges came into being and how they can be accounted for in terms of their historical evolution. Kiparsky refers to the correspondence that can be observed between the breaks of the iambic tetrameter and the hexameter. First of all, the penthemimeral caesura can be accounted for on the basis of the fusion of the iambic distich, because it corresponds to the end of the catalectic tetrameter. Other common caesurae of the hexameter can be explained with regard to the internal sense-pauses of the iambic tetrameters, as can be visualized with the following scheme185:

(∪) –∪||3–||17⨯–||9∪|5– // ⨯||23–∪ ||2–⨯–∪–||49 (sense-pauses of iambic distich) –||6∪||2∪||6–||7∪∪–||12∪||9∪–||3∪∪||11–∪∪–⨯||63 (hex., correspondences in bold)

182 Kiparsky avoids the term anaclasis in his paper, because this term is only common in the Greek metrical tradition, replacing it with "syncopation". Due to our focus on the Greek metrical tradition, I continue to use the term anaclasis. Moreover, the term "syncopation" can cause some confusion: the long and short syllables are not syncopated but only "metathesized". 183 It can be found in Classical Arabic literature, but there a Persian influence can be a sound explanation. 184 In this case, he refers to the existence of combinations of dactylic and iambic sequences in Archilochus' poetry (7th century BC), the Nestor cup and the Vedic tradition (Kiparsky 2018: 25-28). 185 I agree with Kiparsky that sense-pauses played an important role in the formation of the hexameter as we know it. Moreover, he is one of the few scholars who recognizes the importance of the trithemimeral caesura. The percentages are borrowed from West (1982: 36; 41).

Thus, some common breaks (2a, 3a, 4a, 4c) are explained by their corresponding sense- pauses in the iambic rhythm. I am not convinced by this argumentation, because there is too much difference between the matching percentages. For example, the corresponding place for the hepthemimeral caesura exhibits sense-pauses in 23% of all cases, but only 3% in the hexameter (cf. supra for our criticism against Berg et alii)186. The bucolic diaeresis on the other hand, is one of the most prevalent positions for sense-pauses in the Homeric hexameter (11%), but it is only a minor place in the corresponding iambic tetrameter. In addition, Kiparsky attempts to give a similar explanation for Hermann's bridge. In his opinion, this can be justified on the basis of the lack of sense-pauses at the corresponding position of the iambic tetrameter, but I do not see the reason why this would result in such a decisive bridge for the hexameter which is only broken once in every 550 lines (cf. West 1982: 38). Moreover, why did this come into being only at this place and not at other places where sense-pauses were limited in the iambic tetrameters? Another serious flaw with the theory concerns the trochaic caesura, which cannot be clarified adequately with Kiparsky's theory, although it is more prevalent than the penthemimeral one. There is no corresponding place in the iambic rhythm that can account for its importance. Moreover, his proposal of a distich would suggest a caesura at 3c, because that is the place where the non- catalectic variants would have merged together, but this place is strongly avoided in the hexameter (cf. supra with regard to Irigoin's theory; some examples include Il. XV, 18; Od. X, 58, HH. 2, 2 cf. De Decker 2017: 62). This is related to another shortcoming with the proposal. When one looks back at Kiparsky's protohexameter, it can be noted that a spondaic sequence can be found in the third foot. This does not seem very plausible due to the prevalence of a dactylic rhythm in this part of the Homeric hexameter (ca. 85% according to Meister 1921: 7). The Finnish scholar also deals with some unmetrical verses to corroborate his proposal, for remaining trochaic sequences in the Homeric corpus can be explained by his theory as relics of the original iambic-trochaic rhythm of the verse, moreover because they are chiefly found at the beginning of the verse and in the fourth foot, the two places where the original tetrameters began in his opinion. As will be argued below, these unmetrical places are better explained by Witte's theory187. To conclude, we can evaluate Kiparsky's conjecture as an interesting new proposal, but there are again serious shortcomings with the theory. Apart from the above listed imperfections, Kiparsky does not explain how the contraction procedure came into being, he pays no attention to extrametrical arguments etc. Moreover, does it stands up to scrutiny that iambic rhythms which are normally not used in distichs, evolved directly into the dactylic hexameter,

186 As was already discussed, these "sense-pauses" are based on interpunction. It seems better to define them as (potential) caesurae which demarcate intonation units. 187 We should not forget that it is a common feature of Indo-European verse that the beginning of the verse exhibits more deviations. Witte (1915) explained all the unmetrical verses as artificial metrical lengthenings, a theory which in fact cannot completely be refused, although I must confess it is somewhat ad hoc (cf. Berg 1978: 18; cf. infra).

although, as Kiparsky (2018: 18) declares himself, "Ionic syncopation is on the whole less common"?

3.1.7 The theory that received a damnatio memoriae: Kurt Witte None of the presented proposals about the origin of the Homeric hexameter could offer a credible account about its development in protohistoric times. They always run into problems, because they are too much based on a single metrical fact and cannot account for the whole colometry of Homeric verse. This section will therefore shortly present the proposal made by Witte (1913), issued both in Pauly’s Realenzyklopädie and in the journal Glotta (republished in a collection in 1972) that the hexameter resulted from the combination of a dactylic tetrameter and an adonean188. Firstly, I will give a brief account of Witte's arguments and in the last section of this chapter, I will elaborate on them with my own analyses, in order to prove it is until now the only proposal that can explain the different caesurae and bridges of Homeric verse in an economical way. His theory is based on the observation of the prevalence of the bucolic diaeresis in the colometry of Homeric verse, because ca. 60% of all Homeric verses have one (Witte 1913: 2245; 1972: 83). In addition, he notes that the bucolic diaeresis is a gold mine for archaic formulae in the Homeric diction. For instance, the archaic genitives ending in -ᾱο are regularly placed after the bucolic diaeresis, but there is more. Due to the strong unity of the cadence after the diaeresis, many artificial forms are created in this position so they could fit the characteristic rhythm. For instance, based on the genitive Ἀντιφάτᾱο an artificial accusative singular Ἀντιφατῆα was formed with a similar prosodic structure to be placed after the bucolic diaeresis (Witte 1913: 2225; Meister 1921: 30)189. More arguments for his theory are found in the fourth foot of the hexameter, because this was the place were the supposed tetrameter and dimeter would merge together. The dactylic structure of this foot had to be combined with the following adonean, therefore some archaic forms with a dactylic sequence were preserved at this place, most notably the West-Aeolic infinitive forms in -έμεν. On the other hand, new dactylic sequences were created in this place. One can think of the combination of a disyllabic preposition with the following adonean sequence, e.g. κατὰ δάκρυον εἴβεισ (Il. XVI, 11), as such creating the hephthemimeral caesura as a new break for the newly created hexameter (Witte 1913: 2245; 1972: 88-89). This can be further argued for because of the inclination towards a syntactic unity from the hephthemimeral caesura to the end of the verse. Rather casually, Witte (1913: 2246) also mentions the fact that Hermann's bridge can be accounted for with his theory, as I will discuss in more detail below. He also suggested

188 References will be made to the Pauly as (1913: X). As for the journal paper, reference is made to the reprint of 1972. The best overview of the theory is provided by Bassett (1917: 90). The theory was not completely new, because it was already put forward as the basis of the so-called "bucolic hexameter" by Bergk in his seminal study of 1854 (cf. Ritoók 1987: 2). Interestingly, even Berg (1978: 28) has to confess that the hexameter can be interpreted as the combination of dactylic tetrameter with an adonean: "Der auf solche Weise (…) entstandene Normalvers läßt sich als katal. daktylischer Tetrameter (…) auffassen". 189 Tsopanakis (1983: 67) casts some doubt about the relevance of the artificial forms in the fourth foot, but without giving interesting arguments against it.

that the medial caesurae could in this case be interpreted as innovations in the colometry of Homeric verse, which will be verified in the subsequent parts of this master thesis. Before proceeding to our further analysis of this theory, it may useful to review some of the criticisms which were raised against Witte's hypothesis. Bassett (1917) is one of the few scholars who pays some attention to the proposal, although he is not convinced by it. First of all, he denies the importance of the fourth foot and its preference for dactyls (Bassett 1917: 91). However, he is unable to offer a useful explanation himself. Neither should we forget that the first two feet have a preference for spondees and that Witte's theory can explain Hermann's bridge. He further casts doubt upon the relevance of the prevalence of verb forms before the bucolic diaeresis: it is quite normal that a verb stands at the end of the sentence (Bassett 1917: 93). This is in fact more an argument in favour of Witte's theory, because he declares himself that new sentences regularly start after the bucolic diaeresis and we saw above that Indo-European verse has a tendency to equate a verse with a sentence. Moreover, when we apply Chafe's concept of intonation units to the Homeric language, the concept "sentence" loses its importance and it can be proven that many intonation units end before and start after the bucolic diaeresis. Equally untenable is the criticism proposed by Porter (1951: 19) that Witte pays too much attention to metrical feet in his theory and not to cola. In this case, Porter seems to forget the difference between a caesura which divides a foot in two parts and a diaeresis, which separates feet. Witte’s is in fact the only theory which does not have to take feet into account for the protohistory of the Homeric hexameter. They existed already for the tetrameter and the dimeter is placed after it and as such became the fifth and sixth foot of the newly formed hexameter. In addition, Porter rejects the classic criticism that reconstructions of the hexameter can only explain one caesura, but as will be argued below, this objection is by no means valid for Witte's theory. I want to bring it again into the focus of scholarly research in the next section, because for too long it has been neglected, as if Witte had undergone a damnatio memoriae, in view of it seldomly being referred to in research dealing with the Homeric hexameter190.

190 For example, he is not mentioned in the critical survey of theories by Fantuzzi (1984) or by Ritoók (1987), although the latter shares some similar points with him.

3.2 A cognitive attempt towards the protohistory of Greek hexameter

3.2.1 Methodological remarks "Witte's theory breaks down because it fails to explain satisfactorily the frequency of one or the other of the caesurae of the third foot, just as the theories of Bergk, Usener, and others neglect the fondness of Homer for B" (Bassett 1917: 91)191. The aim of this last section is to prove the opposite, for, as we will see in detail, Witte's proposal is in fact the only hypothesis which offers the possibility to explain the different caesurae, from the trithemimeral caesura until the bucolic diaeresis, as well as the important bridges. An important methodological point is the comment by Watkins (1995: 152) that we need to distinguish between diachronic and synchronic approaches towards formulaics. This builds up the important methodological advantage of Witte's theory, which will be emphasized during the following analysis. Different approaches to the problem will be used, ranging from a statistical study of information units after the bucolic diaeresis over archaisms at the beginning of the verse to flexible formulae. A cognitive point of view and a modern theory will be superadded to Witte's own arguments, in order to explain the coalescence of a tetrameter and a dimeter. Due to the limited scope of this thesis, we will have to pass over some other important arguments, such as a thorough comparison between cola used in Aeolic metra and the hexameter or between the sense-pauses of the hexameter and the iambic metres192. In addition, the cognitive principles underlying the Byzantine metres could offer typological parallels. As will be repeated in the conclusion, these elements provide some ground for further research concerning the colometry of Homeric verse. Moreover, one could object against Witte's theory that it fails to offer a complete understanding of the development of the hexameter from Proto-Indo-European times down to Archaic Greece, because in this case the dactylic tetrameter equally needs to be explained. Because of the very hypothetical character of such a reconstruction, this aspect lies outside our scope. It suffices to bring into mind that dactylic tetrameters indeed came into being during the second millennium BC, as was noted by West (1973a: 185)193. The hexameter clausula, the adonean, is also known as a separate colon in the Aeolic tradition (cf. West 1982: 30). Therefore, Witte's theory is corroborated by the individual attestation of both parts of it. With regard to the chronology, the second part of the second millennium BC seems a suitable moment for the coalescence

191 Generally, Bassett is not convinced by attempts to divide the Homeric hexameter into original parts, cf. also Bassett (1919: 345). 192 This would have some similar points with Kiparsky's theory, but avoiding the criticisms which were raised above. 193 It is mainly used in West-Greek poetry, especially in the writings of the Spartan poet Alcman (7th century BC) (West 1982: 43-44; Ritoók 1987: 2).

into one metre. As was noticed above, no concrete proposal will be made concerning its date, for this would require a much broader linguistic and sociocultural study.

3.2.2 The Homeric diaeresis As Witte (1972: 83) himself had done, I would like to begin my analysis with pointing to the prevalence of the bucolic diaeresis, where circa 60% of all Homeric verses have an important word end. One can ask in fact why it received the name "bucolic diaeresis", if it is already so prevalent in the Homeric colometry. The term "bucolic" was coined by the Flemish-Dutch humanistic scholar Heinsius, who in his edition of Theocritus' poems (1604), observed the preference for word break at 4c (74%) (cf. Oswald 2014: 421)194. However, as Bassett (1905: 112) aptly demonstrated, this is by far not an innovation, because some parts of the Iliad, as for instance Andromache's lament for Hector (XXIV, 725-745) (95%!), exhibit more examples of this diaeresis. "Hence the epithet "bucolic" is not justified on this ground", as Bassett (1905: 112) remarks. Maybe therefore, the term "Homeric diaeresis" would be more adequate, because the later preference for it is only a slight intensification of a tendency which is already found in the Homeric epics. Nevertheless, I follow the tradition and will use the term "bucolic diaeresis". Not only the regular occurrence of this diaeresis in the Homeric colometry points towards its particular importance. As Bassett (1905: 115) emphasized, this caesura is similar to the middle caesurae, because it also offers examples of otherwise unexplainable hiatus, or unexpected metrical lengthenings etc. In addition, Witte (1913; 1972) offered irrefutable evidence of its archaic nature, pointing to phonological and morphosyntactic archaisms which are located just before or after the bucolic diaeresis, "a conspicuous receptacle of formular phrases" (Kirk 1985: 29)195. Edwards (1966: 168) made the relevant comment that instances of anaphora occur between the beginning of the verse and the bucolic diaeresis. Equally important is the fact that many sense-pauses occur at 4c, while many new sentences begin after it. This was already noted by previous scholars, for instance Bassett (1905) in his important study about the bucolic diaeresis and Edwards (1966: 167) emphasized this peculiarity at the end of Homeric verse196. They also gave an overview of different constituents which are used after the bucolic diaeresis, for example co-ordinated constituents, appositional phrases, brief similes, participial clauses etc. (Bassett 1905: 116- 122; Edwards 1966: 168-175). Because an important tendency existed in archaic Indo-

194 Therefore, it is sometimes named "Heinsius' law", introduced by O'Neill (1942: 166-167). 195 Or in the words of Rossi (1995: 304): "sede più comune del materiale formulare". 196 Already in Antiquity, this preference of the bucolic diaeresis for important sense-pauses was noticed (Bassett 1919: 354-355). For its importance, cf. Edwards' statements: "The C caesura [sc. 4c] is the most obvious of all breaks in the Homeric verse. More sentences or new phrases begin here than at any other place within the verse, including a fair number of cases where the break in sense at C is strong" (Edwards 1966: 167); "Homer's liking for a fresh start to a sentence or clause at the bucolic diaeresis must not be ignored" (Edwards 1986: 228). Cf. also Finkelberg (2011 vol. 1: 146-147 s.v. Bucolic diaeresis).

European poetry to equate a verse with a sentence (e.g. West 2007: 47)197, this observation is an additional, strong argument in favour of Witte's proposal. If it can be further confirmed that important sense-pauses occur in that position, an original distich of a tetrameter and a dimeter becomes even more plausible. In order to prove this, statistics were collected about different kinds of constituents which are used after the bucolic diaeresis in our corpus of three Homeric books, being 2326 verses in the Vulgate edition. The main findings can be summarized in the following table, which I will discuss afterwards with discussions of specific verses198.

Syntactic structure Number of examples Percentage Combination of noun and verb (VO/OV) 388 26,4 Coordinate clauses and noun-phrases 273 18,6 Simple adjective (apposition), noun or adjective 251 17,1 or noun accompanied with a genitive, both pronominal and nominal Noun phrase, consisting of adjective and noun 225 15,3 Simple verb without noun 115 7,8 Constructions with an adverb, a preposition or a 100 6,8 negation and a verb or a noun Subordinate clauses 87 5,9 Varia 30 2 Total 1469 99,9% (100%)

Firstly, it can be observed that combinations of nouns and verbs constitute the most frequently attested group of constituents. This can be explained at the synchronic level of Homeric colometry and syntax. Because the epic verse is mostly considered a grammatical unit (cf. Higbie 1990: 90), it should not cause any surprise that this combination is found at the end of it, because it is a general characteristic of Greek (and Indo-European) that the sentence ends with the conjugated verb (Dover 1960: 25; Dressler 1969: 3; 20-21; 1971: 18; Dik 2007: 38). This explanation is also possible for the smaller group (7,8%) of verses where a simple verb form is placed after the bucolic diaeresis. For example, reference can be made to the following verse:

197 Cf. also Watkins (1995: 39) or Nagy (1974: 143): "I posit that the Rig-Vedic verse evolved from an idealized grammatical phrase". 198 The full statistics are found in the appendix at the end of the thesis. The methodology is also explained there. West (1982: 36) gives an overview of sense-pauses in the hexameter, which makes it clear that the bucolic diaeresis is the second most important position for it (11%), only slightly less than the penthemimeral caesura (12%). His percentages are based on interpunction, but as was discussed in chapter 2, it is more objective to look at internal criteria in the Greek verses, which increase the percentages. Higbie (1990: 92) also emphasizes the importance of the bucolic diaeresis as a place for inter-sentential boundary (ISB), 22,15% in her statistics.

ςτῆςεν (||1b) ἐῢ κρύνασ, ||3a κρατερὸν δ᾽ ἐπὶ μῦθον ἔτελλε. (Il. XVI, 199). ("he stood, (||) dividing the troops || and he laid upon them a stern command".)

This verse consists of a first part with a conjugated verb and a participle, which is followed by the main clause. Therefore, the penthemimeral caesura is in this case the most likely one on a cognitive level. The last part is formulaic in nature and exhibit the traditional word order in Greek: the verb is placed at the end. Moreover, this verse offers an example of Witte's theory that a syntagm after the bucolic diaeresis can be enlarged with an prefix in tmesis with word end at 4a and an adjective at 3a (cf. infra). Secondly and more important for our discussion, is the prevalence of coordinate clauses and noun phrases coordinate constructions which begin at the bucolic diaeresis (18,6%). This corroborates the thesis that it constitutes an original breakpoint in the verse. In addition, a considerable amount of subordinate constructions (5,9%) is found at this position in the verse (incomplete ones, with enjambment in the following verses are accompanied). The lower occurrence of subordinate constructions when compared to coordinate can be explained by two important considerations. First of all, as was discussed at length above, the Homeric epics are composed in a stringing style, joining shorter information units, mostly in a coordinate way. Hypotaxis is generally rare in the Homeric epics (cf. Chantraine 2015²: 351-364), so it may not surprise us that the percentages are limited for this construction. Secondly, if we accept Witte's proposal, it is more natural that a coordinate construction could be used as a separate shorter verse than a subordinate one. A discussion of some examples will clarify the arguments:

ὣσ ἔον, ||1c εἴ ποτ᾽ ἔον γε, ||3b μετ᾽ ἀνδρϊςιν. ||4c αὐτὰρ Ἀχιλλεὺσ οἶοσ ||1c τῆσ ἀρετῆσ ἀπονόςεται· ||4c ἦ τϋ μιν οἴω πολλὰ μετακλαϑςεςθαι ||3b ἐπεύ κ᾽ ἀπὸ λαὸσ ὄληται (Il. XI, 762-764)199. ("So was I, || if indeed I ever was, || in the presence of men. || Achilles however, alone || he will have the enjoyment of that bravery. || I think, in fact, that he will weep for many things, || when the people will die".)

The first verse would perfectly fit into Fraenkel's colometry, because of its division in four, meaningful components. What is interesting for our discussion, is the important break at the end of the verse. Nestor is speaking in this passage, hence he is the subject of the first part, but at the end of the verse the topic switches to Achilles. This is emphasized by the use of the contrastive prepositive coordinator/coordinating conjunction αὐτάρ, by means of

199 Concerning verse 763, one can consider to place secondary caesurae at 1c, to emphasize that Achilles will be the only one, or at the penthemimeral caesura to stress the meaning of the verb. In addition, it is followed by τῆσ which can have a demonstrative meaning in this case and is placed before the verb, which suggests focus (cf. Dik 2007: 42). However, it needs to be stressed that it is less important than the important break at the bucolic diaeresis.

which a new clause begins at this point of the verse200. A similar procedure is to be observed in the second verse. The string of enclitics in the Wackernagel position emphasizes the beginning of the new sentence there. Comparable examples exist with regard to subordinate construction, for instance:

τριπλῇ ||1c τετραπλῇ τ᾽ ἀποτεύςομεν, ||4c αἴ κϋ ποθι Ζεὺσ δῷςι πϐλιν ||2a Τρούην εὐτεύχεον ||4c ἐξαλαπϊξαι (Il. I, 128-129). ("Three times || even four times we [sc. the Greeks] will refund him,|| when Zeus once give us the city, || Troy with its strong embankments || to destroy it".)201

In the first verse a conditional clause starts at the bucolic diaeresis, again strongly marked by the two enclitics in second position. In addition, the use of the Aeolic conditional conjunction and particle αἴ κε are archaisms in the diction of Homeric verse, corroborating both the assumption of an Aeolic phase in the development of the epic tradition and Witte's remarks that archaisms are found after the bucolic diaeresis. The second verse offers again an example of a verb form which ends the sentence, although this is an infinitive and not a "normal" conjugated one. These few examples make clear that the bucolic diaeresis is not only metrically, but also syntactically an important breakpoint which supports Witte's proposal. In addition, some coordinate constructions begin at the hephthemimeral caesura which can also be explained by his theory.

ΠηλεϏδησ μὲν ||2b ἐπὶ κλιςύασ ||4a καὶ νῆασ ἐϏςασ ἤώε ||1c ςϑν τε Μενοιτιϊδῃ ||4a καὶ οἷσ ἑτϊροιςιν (Il. I, 306-307). ("The son of Peleus [sc. Achilles] || to the tents || and the well-balanced ships he went, || with the son of Menoitios [sc. Patroclus] || and his comrades".)

Traditional colometrists would place a penthemimeral caesura in verse 306 or possibly an hephthemimeral, but doing so, they would have to neglect the strong connection between the preposition ἐπί and its complement. Therefore, a caesura before the preposition seems preferable, as such ending a first colon of the verse which contains the subject of the following actions: Achilles will do something. Verse 307 is one of the few examples in the Homeric corpus where the middle caesura is bridged by a long, semantically important word

200 In addition, the word αὐτάρ is archaic in nature and hence corroborates Witte's statement that the archaisms are placed after the bucolic diaeresis, for a thorough discussion of its Mycenaean character, cf. Ruijgh (1957: 29-55). 201 Theoretically, it would also be possible to decide in favour of a penthemimeral caesura. In that case, the translation would rather be "give us to destroy the city of Troy, with its strong embankments". There is however an important argument in favour of a trithemimeral caesura, namely the attestation of the similar verse: ἱϋμενοσ Τρούην εὐτεύχεον ἐξαλαπϊξαι (Il. VIII, 241) ("desiring to destroy Troy with its strong embankments"). As can be observed, the formulaic verse is adapted by replacing δῷςι πϐλιν with ἱϋμενοσ, pointing towards the strong connection between Τρούην εὐτεύχεον. On the other hand, I was not able to find another verse in the Iliad, where πϐλιν and Τρούην were placed together.

and an hephthemimeral caesura becomes irrefutable (cf. Lehrs 1860: 514). I agree, but I would also place a diaeresis (1c) in the beginning of the verse, because of the enclitic τε which is used after the preposition ςϑν and marks this one as the beginning of a new information unit, similarly emphasized as the following καί202. What mostly concerns us here is the double use of the coordinate conjunction καί after the hephthemimeral caesura. With Witte's theory in mind, these hephthemimeral caesurae can be explained as backformations from the bucolic diaeresis onwards, as was explained above. Observe in this case also the existence of word break at the bucolic diaeresis. On a cognitive level, the bucolic diaeresis is no longer a strong sense-pause, but the remaining word-end in this position is an argument that the hephthemimeral caesura is built to combine the tetrameter with the dimeter. The coordinate conjunction καί is an ideal candidate for doing so, because of its length and because it does not break Hermann's bridge or Wernicke's law (cf. infra). If Ruijgh's hypothesis (cited in Chantraine 1999²: 479 s.v. καί) that καί derives from disyllabic *καςί is correct, this would fit with the preference for dactylic word sequences in the fourth foot. As such, the fact that coordinate constructions (and even new sentences) begin at the hephthemimeral caesura, can also be used as an argument in favour of Witte's theory. Finally, I would like to discuss the figures for simple adjectives and nouns (17,1%) or the combination of adjective and noun (15,3%). Both are important groups as can be gathered from the statistics. This can be explained by their formulaic nature. For example, the end of the verse is a place where a considerable amount of noun-epithet formulae are found, e.g. ὄβριμοσ Ἕκτωρ (Il. XI, 347) or φαίδιμοσ Αἴασ (Il. XI, 496). In this case, they are a fundamental part of the sentence, as they are the subject, which mostly coincides with the metrical structure of the hexameter. However, more interesting for our discussion is the observation that many appositions are also placed after the bucolic diaeresis. Let us consider the subsequent instances:

Πϊτροκλοσ δ᾽ ||2a Ἀχιλῆώ παρύςτατο ||4c ποιμϋνι λαῶν (Il. XVI, 2) ("Patroclus || stood near Achilles, || the herdsman of the troops".)

τύπτε ||1b δεδϊκρυςαι Πατρϐκλεεσ, ||4c ἠΰτε κοϑρη (Il. XVI, 7) ("Why || are you weeping Patroclus, || like a young girl?")

αἰναρϋτη· ||2a τύ ςευ ἄλλοσ ὀνόςεται ||4c ὀψύγονϐσ περ (Il. XVI, 31) ("Terribly brave one, || in which case will someone other have profit of you, ||even a late-born".)203

In the first example an apposition to Ἀχιλῆώ is placed after the diaeresis, the second one has a short simile which is introduced by the prepositive conjunction ἠΰτε and the last one exhibits an apposition to ἄλλοσ, which is further accounted for on the basis of the

202 Kirk (1985: 84) seems to prefer the same colometry, for he calls the verse a rising threefolder. 203 For a discussion of the hapax αἰναρϋτη, cf. Janko (1992: 319).

Wackernagel position of the enclitic particle περ. Important for Witte's theory is the fact that such appositions or similes can be left out without damaging the sentence they occur in204. The first four feet make up an entity, just as the dimeter at the end. This can further point towards an original distich, which can be represented as follows:

Πϊτροκλοσ δ᾽ Ἀχιλῆώ παρύςτατο, Patroclus stood near Achilles, ποιμϋνι λαῶν the herdsman of the troops.

In conclusion, our statistics further emphasize the importance of the bucolic diaeresis: 24,5% of the verses in our corpus with bucolic diaeresis begin a coordinate or subordinate construction in that position. In addition, many appositions are placed there which are not an integral part of the first tetrameter, if they were counted separately, they would further increase the percentages which point towards a syntactic break at the end of the fourth foot. The considerable amount of verb-noun constructions at the end of the verse were explained as innovations in the colometry of Homeric verse, because they agree with the synchronic conception of the hexameter as a syntactic unit.

3.2.3 A neglected caesura of Homeric verse The theories which were presented concerning the roots of Greek epic metre, mostly build their plan on the medial caesurae or, in Berg's case, on the hephthemimeral caesura. Doing so, they neglect the possibility of caesurae early in the verse, most notably the so- called trithemimeral caesura. Two strong arguments should suffice to prove its existence in the Homeric colometry. The first one is of a comparative nature, because it is a common characteristic of Indo-European verse to have a caesura at the beginning of the verse. Meillet (1923: 66) first pointed to the similarity between the existence of early caesurae in the Homeric hexameter and the fact that in Vedic verse there is normally a caesura after the fourth or the fifth syllable205. Similarly, there is a caesura after the fourth syllable in Serbo- Croatian verse (e.g. West 1973b: 171) and a caesura after the fifth and seventh syllable of the Byzantine dodecasyllable, coming from the iambic trimeter (e.g. West 1982: 40; 182-185; 1987: 25). When the hexameter is truly based on Indo-European prototypes, a caesura in the first two feet seems probable. In addition, the existence of enjambments is a strong indication that caesurae in the beginning of the verse need to be admitted, especially when the colometry is based on syntactic and semantic considerations. Moreover, an enjambment can only get some emphasis when an audible pause is heard after it. So the question remains which is the ultimate caesura in the beginning of the verse? Traditionally, the trithemimeral

204 Cf. the statement by Bassett (1905: 122): "The second characteristic of the bucolic diaeresis which marks it as similar in kind to the main caesura in its influence on the connection of thought is the fact that for successive verses it is possible to omit the last two feet without disturbing the narrative". 205 This striking similarity is another argument against his borrowing theory: the variation between different places is also comparable between the hexameter and the Vedic verse forms. For further discussion, cf. Peabody (1975: 34).

caesura (2a) is the only one which is acknowledged in metrical overviews (e.g. Koster 1936: 52; Dain 1965: 53-54)206. However, as Fraenkel (1968³) demonstrated with his four-colon theory, this is not the only position where important breaks occur after the start of the verse, he counted in fact four standard positions (1a, 1b, 1c and 2a). It needs to be stressed however that only 1c and 2a occur frequently enough to be considered as a possible "basic caesura" (cf. Porter 1951). Some further arguments need to be developed in order to choose one of the two. Firstly, West's statistics (1982: 36) show that the trithemimeral caesura exhibits slightly more cases of punctuation (7% instead of 6%). This is corroborated by Higbie (1990: 133), who collected statistics about the relationship between inter-sentential boundaries and types of enjambment. These figures emphasize the importance of the trithemimeral caesura, because it is in almost every kind of enjambment the most prevalent position207. Finally, I would like to draw attention to the difference between 1c as a diaeresis and 2a as a "real" caesura. It is more likely that a break after the second princeps would be the original one. In addition, this would match with the Vedic and Serbo-Croatian caesurae after the fourth syllable. Furthermore, the trithemimeral caesura produces a choriambic sequence (–∪∪–), a metre which is recurringly used in Greek verse, also for the expansion of shorter metres and what is most interesting here, is regularly accompanied by a caesura (cf. West 1982: 32). Therefore, we can assume the trithemimeral caesura to be the most important one in the beginning of the hexameter, but not without keeping in mind the relevance of the other positions. Referring back to Witte's proposal, it can be argued that the trithemimeral caesura is an archaism, as an original caesura of the presumed tetrameter. Some arguments further point in this direction. First of all, the beginning of the verse is, after the end of the verse, the second place where many formulaic phrases are found. Αs Kirk (1985: 28) aptly declares:

"The fact is that the second colon is not very conspicuous for its formular content, and is a part of the verse in which the singer makes less use of pre-formed phraseology than in the first and the fourth. Much the same is true of the third foot".

This is a further indication that the trithemimeral caesura needs to be of considerable antiquity and that the middle parts of the hexameter were modernized due to the coalescence of a tetrameter and a dimeter. Due to limits of time, I was not able to collect full statistical details as for the end of the verse, but for our argument it suffices to discuss here some verses which clearly show that archaisms are found at the starting point of the verse.

ἀνςτότην, ||2a λῦςαν δ᾽ ἀγορὴν ||4a παρὰ νηυςὶν Ἀχαιῶν. (Il. I, 305)

206 As discussed above, it is also regularly neglected, e.g. by West (1982; 1987; 1997) or Sicking (1993). 207 To summarize the most important aspects: when there is no enjambment, the trithemimeral caesura coincides with a sense-boundary in 141 instances, and 1c only in 81 instances. On the other hand, when enjambment occurs, the figurs are as follows: 1c has 137 instances, 2a 176 examples in the whole Iliadic corpus.

("Both stood up, || they dissolved the assembly || near the ships of the Greeks".)

The enclitic position of δ᾽ after the verb form λῦςαν and the necessary enjambment when compared with the preceding verse are clear indications to place here a trithemimeral caesura. Interesting in this case is the fact that both ἀνςτότην and λῦςαν are morphosyntactic archaisms. Firstly, both verb forms lack an augment. Additionally, the form ἀνςτότην has two further peculiarities, first because it is in the dual number, which was no longer used in the Ionic language, but could be an Aeolic archaism208 (e.g. Palmer 1962: 128). Secondly, the form is syncopated from *ἀναςτήτην, which is also an Aeolic characteristic (Palmer 1962: 140). The lack of an augment in the beginning of the verse is found more than once (cf. De Decker 2019: 8 - page refers to the manuscript), as can be seen in the following examples as well:

… ||4c οὐδ᾽ Ἀγαμϋμνων λῆγ᾽ ἔριδοσ ||2a τὴν πρῶτον ||3b ἐπηπεύλης᾽ Ἀχιλῆώ, (Il. I, 318b-319)209. ("… and not Agamemnon calmed down from the quarrel, || by which he first || threatened Achilles".)

ὣσ φάτο, ||1c Πϊτροκλοσ δὲ ||3b φύλῳ ἐπεπεύθεθ᾽ ἑταύρῳ, ἐκ δ᾽ ἄγαγε κλιςύησ ||3a ΒριςηϏδα ||4c καλλιπϊρῃον, δῶκε δ᾽ ἄγειν: ||2a τὼ δ᾽ αὖτισ ἴτην ||4a παρὰ νῆασ Ἀχαιῶν. (Il. I, 345-347)210. ("So spoke he [sc. Achilles], || Patroclus || obeyed his dear friend, he led forth from the tent || Briseïs || with the beautiful cheeks, he gave her to take her away. || They [sc. the heralds] went ||to the ships of the Greeks".) (boldface added.)

In addition, verse 346 offers an interesting example of another archaism, namely the use of tmesis in ἐκ δ᾽ ἄγαγε, a construction which has a genitive complement after the trithemimeral caesura. The use of the adversative, formulaic conjunction αὐτάρ, which we

208 The lack of an augment is metrically secure in both cases, cf. De Decker (2017: 113-114). Discussion of the dual in this case, cf. e.g. De Decker (2017: 133). For a general discussion of grammatical number in Homeric Greek, cf. Monro (1891: 158-162); Chantraine (2015²: 22-34), for the dual cf. Finkelberg (2011 vol. 1: 223 s.v. Dual). 209 The prepositive relative pronoun τήν is a clear indication for a trithemimeral caesura. Remark also that the enjambment is accompanied with a bucolic diaeresis in the preceding verse; this will be discussed below. 210 Augment use in this passage is discussed in detail by De Decker (2017: 134). Short overview of arguments for the printed colometry: verse 345: formulaic character of ὣσ φϊτο, postpositive δέ after the subject Patroclus, hyperbaton and first position of the adjective φύλῳ; verse 346: most important caesura in the middle of the verse, 4a is secondary caesura, καλλιπϊρῃον can be taken as an apposition; verse 347: end of a syntactic structure at 2a, postpositive δ᾽ after the subject τώ (demonstrative), formulaic complement with preposition after the hephthemimeral caesura.

encountered after the bucolic diaeresis is also found at the beginning of verses, e.g. Il. I, 333 (αὐτὰρ ὅ ἔγνω...; "but he realized..."). Similarly, the Aeolic conjunction with double consonant ὁππότε, followed by the archaic enclitic personal pronoun μιν, can be found in Il. I, 399, while Il. I, 408 starts with the Aeolic conditional αἴ κέν πωσ, which we also encountered after the bucolic diaeresis. Hence, this short overview of some archaisms which are preserved at the beginning of the verse, corroborates our thesis that the trithemimeral caesura can be interpreted as an archaism in the colometry of Homeric verse. Secondly, this is borne out by the observation that early caesurae in the verse are mostly combined with a caesura at the end of the verse, the hephthemimeral caesura, the bucolic diaeresis or rarely the ennehemimeral caesura (5a) (Janse 2012: 29). A first case of this distribution is generally observed in the secondary literature, namely the fact that most instances of enjambment start after the bucolic diaeresis. Bassett (1926: 118) explained this with reference to the compact space which is available after the diaeresis. It is not always long enough to complete a whole thought at the end of the hexameter, hence it is continued in the next verse211. This is certainly a valuable explanation on a synchronic level, but with Witte's proposal a diachronic explanation seems possible as well. When we accept the hypothesis that an original tetrameter and dimeter were combined, it can be imagined that enjambment came into being in a similar way as the hephthemimeral caesura, as an expansion of the original adonean. Let me discuss the following example from Iliad XVI:

Πατρϐκλου δ᾽ ||2a ὑπὲρ ὦμον ἀριςτερὸν ||4c ἤλυθ᾽ ἀκωκὴ ἔγχεοσ, ||1c οὐδ᾽ ἔβαλ᾽ αὐτϐν· ||3b ὃ δ᾽ ὕςτεροσ ||4c ὤρνυτο χαλκῷ Πϊτροκλοσ· ||1c τοῦ δ᾽ οὐχ ἅλιον βϋλοσ ||4c ἔκφυγε χειρϐσ. (Il. XVI, 478-480)212. ("of Patroclus || over the left shoulder || went the point of the lance, || and it did not hit him. || Afterwards, || he rushed on with the bronze, Patroclus. || Not in vain the missile || fled from his hand".)

In the first verse, the syntagm after the bucolic diaeresis consists of the combination of a verb and a subject, which constitutes in fact an acceptable short clause. The same can be said about ὤρνυτο χαλκῷ: because Ancient Greek is a pro-drop language, this is a full sentence (ἔκφυγε χειρϐσ is a similar example). Therefore, they could have been a separate verse in an earlier phase of the tradition. In this case however, they are both accompanied by a periodical enjambment, ἤλυθ᾽ ἀκωκὴ is supplemented with a genitive ἔγχεοσ and ὤρνυτο χαλκῷ is followed by Πϊτροκλοσ, which can in fact be interpreted as an apposition to the subject ὅ, which was expressed before. On a diachronic level, such sequences could arise

211 Cf. Kirk (1966: 129): "Strong overrunning of the verse-end is often caused by stops at the bucolic caesura". 212 Arguments for the colometry: verse 478: enclitic position of δ᾽, syntactic unity from 2a to 4c, adonean can be interpreted as short verse; verse 479: enjambment, prepositive word οὐδ᾽, interpunction in the middle of the verse, corroborated by postpositive δ᾽, adonean can be short verse, parallel to previous verse; verse 480: enjambment, enclitic after τοῦ, unity until bucolic diaeresis. In this case, I followed the Teubner edition by West (1998-2000) in printing ὤρνυτο, because this one is found in most manuscripts, although the OCT prints ὄρνυτο.

when both parts of the hexameter were combined. In particular, it was felt that the adonean as a separate part of the distich was too short to contain full information units, for as Higbie (1990: 104) demonstrates, "there is a metrical pressure for a sentence to be at least half a verse in length". Τhe sentence was thus expanded both to the left to create the hephthemimeral caesura (as was discussed above), but also to the right, because by using enjambment the sentence (and verse) could also be connected to the following verse, thus strengthening the unity of the newly composed verse. Probably, there was already an early caesura in the original tetrameter and by combining both parts, this became an interesting place to extend the adonean to. Important in this case is the fact that the most exceptional kind of enjambment, the violent enjambment nearly always is accompanied by a bucolic diaeresis at the end of the preceding verse (Higbie 1995: 73). A second type of combination between early and late caesurae occurs within the same verse. When one makes use of a cognitive approach towards Greek metre, many verses exhibit a caesura in the beginning of the verse and at the end, both of which are generally more important breaks than a caesura in the middle of the verse. When we look back to the above discussed examples, it can be observed that they indeed present such patterns. For instance in Il. XVI, 478, traditional colometrists would place a trochaic caesura after ὦμον because of the word break in that position, but that would separate it from the adjective ἀριςτερόν, which is unlikely because this is not an autonomous apposition. Therefore, it is more interesting to divide it into three parts which are meaningful components and as such retain the archaic colometric structure by a tetrameter and his trithemimeral caesura and a dimeter which begins after the bucolic diaeresis. In addition, this sequence is not limited to this "archaic" combination of trithemimeral caesura and bucolic diaeresis, but over time, also combinations of the early diaeresis (1c) with the hephthemimeral caesura (4a) and the ennehemimeral caesura (5a) came into being213. It suffices to cite one example:

χϊζεο ||1c διογενὲσ Πατρϐκλεεσ· ||4c οὔ νϑ τοι αἶςα (Il. XVI, 200). ("Recoil || godlike Patroclus || not your fate…".)

In this case, an early diaeresis (1c), which can be accounted for on the basis of the emphatic vocative at the beginning of the verse, followed by an imperative with an adjective before it. The bucolic diaeresis can be accounted for on the basis of the prepositive οὔ, followed by a string of postpositives (νϑ τοι). The supposition that 1c and 4a are innovations was discussed at length above. 5a is the last position in the verse where an important break can occur in the verse, although it is quite rare (Higbie 1995: 86)214. Because of this scarcity, it can only be an innovation when compared to the bucolic diaeresis.

213 Bassett (1917: 89) already observed the fact that a trithemimeral caesura is regularly accompanied with a hephthemimeral caesura. Cf. also Kirk (1985: 18-24). 214 This is sometimes named Gerhard's law and was already noted in Antiquity (cf. Oswald 2014: 421).

3.2.4 Archaisms and innovations in Homeric caesurae The previous sections discussed the bucolic diaeresis as a likely place to be an original verse end and the possibility that the trithemimeral caesura can be interpreted as an authentic breakpoint of the tetrameter, which would match with the Indo-European tendency for early caesurae. This section will focus on the medial caesurae: are they archaisms or innovations in the colometry of Homeric verse? Firstly, I want to reiterate that Witte was able to explain the hephthemimeral caesura as an expansion from the bucolic diaeresis backwards. This argument can be (at least partially) extended to the medial caesurae as well. Consider the following verses:

τοῖςι δ᾽ ἄφαρ πϐλεμοσ ||3a γλυκύων γϋνετ᾽ ||4c ἠὲ νϋεςθαι ἐν νηυςὶ γλαφυρῇςι ||3b φύλην ἐσ πατρύδα γαῖαν. (Il. XI, 13-14). ("Suddenly, war || became sweeter for them || than to return in hollow ships || to their beloved homeland".)

Verse 14 shows one of the most recurring formulae which are used after the trochaic caesura. When we look in detail how this formula is built, it should be noted that word end occurs both at the hephthemimeral caesura and at the bucolic diaeresis. When we put caesurae on the basis of information units, it is not preferable to place them in that position, but the observance of word end at both positions can be important diachronically. When we apply Hainsworth's (1968) concept of flexible formulae to this sequence, it can be explained as an extension of a shorter formula. The adonean sequence πατρύδα γαῖαν could in fact be the original formula consisting of the combination of an adjective and a noun, just like a considerable amount of formulae at the end of the verse (cf. statistics above). It could for example be used as an apposition to a noun in a previous tetrameter, which ends for instance with a word for Greece. When both parts were combined with each other, the preposition ἐσ could be used to create a hephthemimeral caesura and transform the adonean sequence to a locative complement in the newly created hexameter. From that point backwards, it became relatively easy to add an adjective φύλην before it, thus creating a trochaic caesura. The tendency to search for break points in the middle of the hexameter will certainly have influenced this process considerably. When we look at verse 13, a similar pattern is found. The penthemimeral caesura is accompanied both by word end at 4a and 4c. Other conspicuous examples are provided in Iliad XI, 443-444:

ςοὶ δ᾽ ἐγὼ ἐνθϊδε φημὶ ||3b φϐνον καὶ κῆρα μϋλαιναν ἤματι τῷδ᾽ ἔςςεςθαι, ||3b ἐμῷ δ᾽ ὑπὸ δουρὶ δαμϋντα (Il. XI, 443-444). ("Here, I declare that for you || murder and black death will be on this day, || overpowered by my spear".)

The first verse is an example of what is called by Hainsworth (1968: 82) "expansion by means of a co-ordinated synonym". Again, both at 3b, 4a and 4c word breaks occur. The syntagm κῆρα μϋλαιναν has therefore the potential to be considered as the starting point of expansion, first with καί, a procedure which was discussed above. As such, a coordinate construction is developed at 4a, which offers the possibility to put a co-ordinated noun φϐνον before it, which can be further triggered by other collocations of φϐνον and κῆρα (six times according to Hainsworth (1968: 83)). The second verse exhibits a similar pattern, but this time the adonean sequence is extended to the hephthemimeral caesura by means of an adverb/preposition ὑπό215. Doing so, it could be further enlarged to the trochaic caesura by means of the possessive pronoun ἐμῷ. As can be noted, such sequences are a strong argument that the middle caesurae are (at least partially) innovations which arose as extensions of original shorter formulae after the bucolic diaeresis. This can be clearly observed in the following scheme, based on Horrocks (1997: 201).

δάκρυα λείβων (after bucolic diaeresis) κατὰ δάκρυ χέοντα (hephthemimeral caesura) τέρεν κατὰ δάκρυ χέοντα (trochaic caesura) θαλερὸν κατὰ δάκρυ χέοντα (penthemimeral caesura)

The Homeric formulaic system in the second part of the verse is based on four caesura positions: the bucolic diaeresis, the hephthemimeral caesura and the two middle caesurae. When we accept Witte's proposal, they can be explained as departing from the bucolic diaeresis backwards. Important in this case is the remark made by Hainsworth (1968: 79) that the expansion of formulae is typical for the second part of the hexameter and not for the beginning. This is an extra indication that the above sketched chronological sequence is plausible to be correct. Miller's (1982: 35) scheme of the colometric structure of the hexameter underlines this supposition:

To conclude, Hainsworth's theory of flexible formulae seems to suggest that the medial caesurae came into being as innovations when the tetrameter and dimeter were

215 Adverb when you consider this case as a tmesis form, otherwise it would be a preposition with the following dative case.

combined. The reality seems more complicated. Witte (1972: 103) also shortly discussed an example of a strophe of dactylic tetrameters as it is found in Alcman (7th century BC) with regard to the medial caesurae.

Μῶς’ ἄγε Καλλιϐπα, θϑγατερ Διϐσ, "Muse, go ahead, Calliope, daughter of Zeus, ἄρχ’ ἐρατῶν ϝεπϋων, ἐπὶ δ’ ἵμερον start your lovely words, place desire on ὕμνωι καὶ χαρύεντα τύθη χορϐν216. this hymn and make this dance charming".

Witte's statement about these verses is straightforward: they are the ultimate evidence that his proposal is corroborated by actually attested verses. In addition, the existence of a penthemimeral caesura in the first two verses and a trochaic one at the end, are interpreted by Witte as proof of their originality in the colometry of the tetrameter217. However, that is jumping to conclusions, for at least the possibility needs to be considered that dactylic tetrameters adopted the penthemimeral and the trochaic caesura by influence of the Homeric hexameter. The surviving examples of this metre were composed after the Homeric epics, hence it is very likely that they were influenced by its metre and its diction. This seems very plausible for the first verse, because it resembles the last verse of the prooemium of the Odyssey:

τῶν ἁμϐθεν γε, θεϊ, ||3a θϑγατερ Διϐσ, ||4c εἰπὲ καὶ ἡμῖν. (Od. I, 10) ("From some place, goddess, || daughter of Zeus, || speak also about them to us".)

As can be observed, the phrase θϑγατερ Διϐσ is localized at the same metrical place in both instances, after the penthemimeral caesura218. Due to the immense familiarity with the epic tradition, it can be imagined that Alcman composed this verse as a lyric alternative to one of the well-known verses of the heroic poems, including the same kind of caesura. Therefore, it seems a wise decision not to base absolute conclusions on these small amount of verses. The discussion of flexible formulae above, clearly pointed to their influence on the prevalence of

216 The text is based on the edition by Page (1962) in his Poetae Melici Graeci, fragment number 27. As can be observed, early caesurae are indeed found in these verses: 1c in verse 1 and 3 and at least word end at 2a? in verse 2. In the first two verses they are secondary to the main penthemimeral caesura, but in the last verse it is probably the most important one because it is placed before the prepositive καί. 217 This is followed in metrical handbooks, cf. e.g. West (1982: 43). 218 Note the similar localization of ἵμερον in the following verse from the Iliad: ὣσ φϊτο, τῷ δ᾽ ἄρα πατρὸσ ὑφ᾽ ἵμερον ὦρςε γϐοιο ("As such he [sc. Priam] spoke, he aroused for him [sc. Achilles] the desire to weep for his father") (Il. XXIV, 507). Interesting is also that both verses from the Iliad have a bucolic diaeresis. Stifler (1924: 353) equally emphasizes the influence of the hexameter on these verses, because verse 3 begins with a spondee, although the biceps procedure was avoided in lyric poetry. However, West (1982: 43) applies the biceps procedure to his printed scheme of the dactylic tetrameter. Interestingly, according to West's statistics this is mostly applied to the first two feet and only seldomly to the third. If this represents the original structure of the tetrameter, this could explain why in the hexameter spondees are more frequent at the beginning of the verse. Hence, one of the last arguments in favour of the Norwegian theory would be countered. For a discussion of Alcman in a broader oral context and with the question of Homeric influence, cf. Kousoulini (2013).

medial caesurae in the Homeric hexameter (already suggested by Witte (1913: 2246)). Further research on the colometry of dactylic tetrameters will have to determine in what degree they are influenced by the epic hexameter and which parts are likely to be original. Doing so, it may be clarified whether the medial caesurae are full innovations in the colometry of Homeric verse or intensifications of a tendency which existed already in the original tetrameter. A possible scenario would be that the penthemimeral caesura was already common in the tetrameter, and that the trochaic caesura arose due to the combination of both verses. This could explain why both caesurae exist and why more sense- pauses are found at the penthemimeral caesura than at the trochaic one219. For the time being, however, this remains mere speculation. A clearer picture can be obtained with regard to the important bridges/laws in the Homeric hexameter, for they can also be explained with Witte's proposal. Let me first bring them into mind. The most important one, Hermann's bridge was already discovered in 1805 by the German scholar Gottfried Hermann in his edition of the Orphic hymns (Hermann 1805: 692-699; cf. e.g. Oswald 2014: 421) and states that word break after the trochee in the fourth foot of the hexameter is strongly avoided. Only once in every 550 verses this bridge is broken (West 1982: 38). Wernicke's law is related to it and states that when a spondaic sequence ends at the fourth syllable, the latter is long by nature (cf. e.g. Stifler 1924: 323)220. Finally, concerning the second foot, one has to refer to Meyer's bridges, which were developed with regard to Hellenistic poetry, but of which clear tendencies are already to be observed in the Homeric poems. Meyer's first bridge can be summarized as the avoidance to end a word, which began in the first foot of the hexameter, at positions 2b or 2c221.

219 In West's (1982: 36) statistics: 12% for the penthemimeral caesura and 9% for the trochaic caesura. Higbie (1990: 92): 28,74% of ISB at the penthemimeral caesura and only 17% at the trochaic caesura. Cf. also Higbie (1995: 81) concerning internal clause boundary: "While the bucolic diaeresis (8) is an important place for ICB, the trochaic caesura (5 1/2) is relatively unimportant in the Catalogue of Ships, Theogony and Erga". These statistics are based on syntactic units, when we only look at word end, the trochaic caesura is more prevalent, but not in Hesiod (cf. e.g. Vigorita 1977: 289 based on O'Neill 1942 and Porter 1951), where the penthemimeral one is preferred. 220 Stifler (1924 passim) tries to deny the existence of Wernicke's law in the Homeric epics, but she is not able to offer convincing arguments. For example (1924: 334), she refers to the fact that most of the spondees which are found in the fourth foot exhibit a strong syntactic unity with the following adonean, but as we saw above, they can be explained with regard to the innovation of the hephthemimeral caesura. She tries to counter Witte's proposal by reversing the chronology of the formulaic argumentation. The bucolic diaeresis can also be an innovation in her opinion. For arguments the other way around, cf. supra. 221 Meyer's second law states that iambic words are avoided before the penthemimeral caesura, but this will not be discussed here (cf. Meyer 1884: 980). There is some discussion if they are already valid for Homeric poetry or not. Meyer himself (1884: 980-983) developed them with regard to Hellenistic poetry, but it can be argued that already a clear tendency existed in Homeric poetry. In fact, this can be inferred from Meyer's formulations, e.g. "Diese Regeln, (…), sind vielleicht von ihm [sc. Callimachus] festgesetzt" (1884: 983; boldface added). They can only be fixed by the Hellenistic poets if there was already a tendency, otherwise, they would have invented them. Cantilena (1995: 38ff.) is an in-depth study of the problem, he also does not believe them to be relevant in Homer, for 6 until 7% of all verses violate the "law" (cf. infra; only referring to 2b), followed inter alios by Oswald (2014: 422). I agree with De Decker (2016: 42-44; 2017: 62-66; 2019: 4) that the

How can Witte's proposal account for their existence? When we take a tetrameter as the original starting point, both Hermann and Wernicke can be explained. Firstly, because a word break at the trochee of the fourth foot would mean that the tetrameter needs to end with a monosyllabic orthotone word, a structure which is avoided in Indo-European languages, as was first observed by Wackernagel (cf. Rossi 1995: 294). In addition, such monosyllabic words are not likely to be used for the regressive combination of the adonean with the tetrameter, which results in the hephthemimeral caesura (cf. supra)222. Secondly, when a tetrameter ends with a spondee, this must be long by nature, because there is no following syllable which can make it long by position. This is similar to the later hexameter which can only end with a spondee, when the last syllable is long by nature and not by position. Meyer's bridges on the other hand, are concerned with the second foot, therefore the chances are that they arose within the original tetrameter. They are situated between the trithemimeral caesura and the penthemimeral caesura, of which probably the first and possibly also the second can be explained as original caesurae in the tetrameter. Therefore, Meyer's bridge could arise as an attempt not to confuse between these two places by putting too frequently a caesura in the middle of them. Equally interesting in this case is the difference in percentages between violations of Meyer's and Hermann's bridge. The latter, which can be explained by the end of the tetrameter, is violated in only 0,3%, whereas Meyer's bridge for 2b is prohibited in 7% of the verses in the Iliad and 6% of the verses in the Odyssey223. The difference is significant because this makes the supposition of an original verse end at 4c even more probable. Why would this otherwise almost never be violated?

3.2.5 The relevance of unmetrical verses When searching for the origins of the Homeric hexameter, some scholars pay considerable attention to the existence of unmetrical verses in the extant text of the Homeric epics. In their opinion, such verses can preserve an earlier stage of the tradition when more irregularities existed in the hexameter. This last section of my thesis will very briefly discuss the concept and argue that their relevance for our knowledge of the prehistory of Greek epic is limited, for most examples can be explained on a synchronic level224. Traditionally, such verses are divided into three categories, depending on the question whether they occur at

tendency was already present, because the percentages are not excessive. For an attempt to explain these laws, cf. Steinrück (2010-2011). 222 Synchronic explanations by Fraenkel (1968³: 121-123): it is not expected to have a monosyllabic word between the normal caesura places at 4a and 4c; Irigoin (2004: 6): the characteristic cadence of the adonean would be obscured otherwise. Berg & Lindeman (1992: 192 fn. 40) have to admit that both Wernicke and Hermann can be explained with the dactylising of the fourth foot (= Witte's proposal), but they do not draw the logical conclusion. 223 I would like to express my gratitude to Dr. Filip De Decker for discussing this point with me. 224 With "unmetrical verses" I do not intend here the metrical anomalies, like digamma, the original scanning /- oo/ of the genitive -ου etc. The problem of vocalic /ṛ/ was already discussed above. I will not concern myself here with ςτίχοι μείουροι (cf. infra for definition), because they are not relevant in the discussion about the origins of the hexameter.

the beginning, in the middle or at the end of the verse. Verses which have a short syllable in the thesis of the first foot or exhibit a trochee in that position, were called by the ancient grammarians ςτίχοι ἀκέφαλοι, "verses without a head". On the other hand, the terms ςτίχοι λαγαροί and ςτίχοι μείουροι respectively refer to verses with a metrical irregularity in the middle or at the end of the verse (e.g. Meister 1921: 42-44). The ἀκέφαλοι are explained in Berg's theory as remnants of the original ancipitia at the beginning of his presupposed glyconeus, but as we argued above, this is an ad hoc explanation in order to account for the spondaic sequences in the first two feet and hence also the limited amount of ἀκέφαλοι. The number of unmetrical verses is indeed very limited, for only 0,1% of Homeric verses have them (cf. Meister 1921: 44). This casts serious doubt on their relevance for the history of the epic tradition. They can simply be understood as remnants of the Indo-European metrical tendency that the first foot is somewhat more loose than the other ones (Meillet 1923: 37; cf. West 2018: 10)225. Because they do not constitute the final cadence of the verse, the hearer will have no problem if there would be some small anomaly in the beginning, probably he/she would even not note it226. Let me briefly discuss the following "unmetrical" verse:

δαϏζων ἵππουσ τε ||3b καὶ ἀνϋρασ· ||4c οὐδϋ πω Ἕκτωρ (Il. XI, 497) ("killing horses || and men; || and not Hector…")

Normally, the /a/ in δαϏζων needs to be short, but this is not possible in this case, because it is placed in the princeps of the first foot, hence a ςτίχοσ ἀκέφαλοσ. However, could it not be imagined that the beginning /a/ was somewhat lengthened by the singer? This occurred also in other positions of the verse, as we saw above with the example ἀθάνατοσ. In addition, it can be important that only verses, where the metrical licency is not expressed in the writing, are considered as "unmetrical" by the ancient commentators in the scholia (Meister 1921: 43). This is a strong indication that such unmetrical verses were only considered as problematic by later readers of the epic, but not by the earlier listeners of the aoidoi. Let me proceed to the λαγαροί, which were used by West and the Italian school as an argument that the hexameter came into being as the combination of two shorter cola which were merged at the middle caesurae of the hexameter. As was discussed above, such attempts fail to explain why two caesurae exist in the middle of the verse. Therefore, it is better to interpret these unmetrical verses as later developments, which arose because epic poets thinked in metrical cola when they were composing their verses227. Dactylic

225 West wrote an article about "unmetrical verses" in Homer. It is published in the same recent volume (Hackstein & Gunkel 2018: 362-379) as Kiparsky's paper. Hence, I was not able to consult the definitive edition. I refer to the manuscript edition, which was given to me by prof. dr. Mark Janse, whom I want to thank for this. 226 As was noted above, Witte (1915) explained them as ordinary metrical lengthenings. This is somewhat ad hoc, I confess, but certainly not impossible. Cf. also Watkins (1995: 29). 227 Cf. Steinrück (2005: 401): "The history of irregular verses, such as the so-called lagaroi, offers some evidence that there were more irregular, i.e. colon-based, verses in Homer before the long scholarly tradition

100

hexameters have a strong preference for a caesura in the middle of the verse, hence the poets used formulaic phrases which began and end in that position. Sometimes, this resulted in an extra syllable or a short syllable where a long one needs to be placed. Again, is it likely that a gifted epic singer could obscure such problems in order that the listener would not bother about them. Have a look at this verse from Iliad I:

μότε ςὺ Πηλεύδη ||3a ἔθελ᾽ ἐριζϋμεναι βαςιλῆώ (Il. I, 277). ("You, son of Peleus, || do not wish to quarrel with a king"!)

The sequence ἔθελ᾽ ἐριζϋμεναι offers some problems to scan, because three short /e/'s follow one another. This is most likely to be explained because of the formulaic character, for the sequence ἐριζϋμεναι βαςιλεῦςιν (with the plural form 'kings') is two times attested in book II of the Iliad228.

μϊψ, ||1a ἀτὰρ οὐ κατὰ κϐςμον, ||3b ἐριζέμεναι βασιλεῦσιν (Il. II, 214) ("In vain is it, || not duly || to quarrel with kings".) ἴςχεο, ||1c μηδ᾽ ἔθελ᾽ οἶοσ ||3b ἐριζέμεναι βασιλεῦσιν (Il. II, 247) ("Restrain yourself, || do not want alone || to quarrel with kings".) (boldface added)

In addition, it can be observed that the imperative ἔθελ᾽, accompanied by a negation is also used in verse 247. Therefore, it seems best to interpret this one as the source of the λαγαρόσ in Iliad I. Can one obtain interesting information about the history of the epic verse as such? This does not seem very probable. In addition, such irregularities also occur at other positions in the verse, for example in a formulaic verse of the Odyssey (this example is taken from Meister (1921: 43).

Τηλϋμαχε, ||2a ποῖϐν ςε ἔποσ φϑγεν ||4c ἕρκοσ ὀδϐντων (Od. III, 230). ("Telemachus, || what kind of word fled || from the hedge of your teeth?")

In this case, the vocative ending /e/ is lengthened in the word Τηλϋμαχε, although normally this is a short one, as can also be observed from the proparoxytone accentuation of the word. This example is in fact further evidence for some of the assumptions we presented in the course of this thesis. Firstly, this offers an extra indication that the trithemimeral caesura indeed needs to be admitted as an important caesura in the colometry of Homeric verse (cf. again the enclitic position of ςε). Secondly, such examples make it likely that caesurae are

connected them into metron-based schemas." (boldface added). However, it needs to be stressed that Steinrück's conclusions resemble those of West and the Italian school. 228 In totality, there are four instances of the infinitive ἐριζϋμεναι in the Iliad. The fourth instance is placed in the first part of the hexameter and hence does not bother us here (Il. XXI, 185). The Gesamtkommentar (vol. 1.2: 110-111) proposes to read with synizesis of the last /ε:/ in Πηλεύδη with the first /e/ of ἔθελ᾽. This seems unlikely due to the caesura which occurs in the middle of them.

101

indeed a real pause in the recitation of the Homeric poems. Due to a short pause, the listener will not have any troubles with a lengthening of the vocative ending229. To conclude, unmetrical verses are the outsiders in the corpus of the epic verse and they can be explained as a licency permitted to the oral poets. It is possible that there were more in the earlier phases of the epic tradition, but the indications they can give for the prehistory of the metre itself are very scarce, if not non-existent.

229 There is some discussion about verses which are unmetrical in the fourth foot. According to Stifler (1924: 345) the secure instances are limited. Kiparsky (2018: 32) on the other hand, considers this foot to be an important place for unmetrical verses, which cannot be explained by Berg's theory. Inter alia, he refers to the artificial dactylizing of ὦρτο to ὤρετο (Il. XXII, 102) in the fourth feet. Such examples are in fact an additional argument in favour of Witte's theory as we saw above. In that case, the original dactylic tetrameter could have been catalectic, ending with a trochee, just as the later hexameter. Or, when one does not accept this, they can be interpreted as artificial formations in order not to violate Wernicke's law.

102

CONCLUSION: Did we find the dragon? After making a comprehensive flight through different parts of Homeric linguistics and metrics, we now reached the point to answer the question whether some indications could be found where the origins of the hexameter, this dragon for scholarly research, are located. Firstly, it was shown that oral poetics can give an interesting point of view regarding research on the Homeric epics. Parry's dissertations were certainly one of the most important discoveries concerning the Greek epic tradition and they can offer a plausible framework, both for the style and the language of the Homeric epics. A brief survey of this composite language was presented in the remaining part of the first chapter, with an overview of possible diachronic explanations. Secondly, I focused on the metre which is used in the Greek epics, the dactylic hexameter. After giving an introduction to the basic facts, the discussion concentrated on the colometry of this long verse. Which is the best method for placing caesurae in the hexameter? Doing so, extensive criticism was objected against the statistical method of Martin West et alii, who only accept the penthemimeral, trochaic and (in exceptional cases) the hephthemimeral caesura. The four-colon proposal of Fraenkel and its adaptations by Kirk were characterized as valuable alternatives to the standard vision, but it was also noted that they did not go far enough to offer a colometry which can be reasonably argued for in an oral context. Therefore, this paper suggested that a more rewarding colometric method needs to be based on cognitive principles. Doing so, it was highlighted that Bakker's and Janse's attempt to apply Chafe's information units to the Homeric poems and their scansion can offer a more dynamic approach, which can account for the stringing style of their composition and the oral performances they were composed in. Thirdly, having this cognitive approach towards the colometry of Greek verse in mind, different proposals concerning the origins of the hexameter were discussed at some length. Each proposal has some arguments in favour of it, but each also runs into serious problems, when one tries to explain the whole colometry of the verse with it. Therefore, it was brought to mind that the old theory of Witte (1913), interpreting the "Ionic long verse" as a combination of a dactylic tetrameter and a dimeter (the adonean), could be more effective than the other theories. It is the only one that can clarify how the different caesurae and bridges of the hexameter came into being. In addition, it has the advantage that it uses not only metrical arguments but also linguistic ones. More specifically, when we pay attention to information units, we can observe that many different syntactic structures are used after the bucolic diaeresis and, even more important, that it constitutes a place where a considerable amount of new clauses begin. It was also emphasized that a caesura at the end of the verse is recurrently accompanied with one at the beginning of the verse. This can be a further argument that this early caesura, particularly the trithemimeral caesura, has to be interpreted as an original caesura of the supposed tetrameter at the beginning of the verse. This inner logic of Witte's proposal is maybe the major argument in favour of it. Searching for the origins of the hexameter remains somewhat hypothetical, but theories which are

103

built on such an inner logic and which can explain important aspects of the hexameter, are more likely to be interpreted as a plausible solution of the problem (cf. Visser 1987: 124 about his own formulaic theory). While undertaking our journey to the Bronze Age origins of the dactylic hexameter, some possibilities for future research were discovered. Concerning the overall development of the Homeric language, more research needs to be conducted with regard the problems of the phase model and the diffusionist model. More complex representations of the development with attention to criss-cross patterns of diffusion and influence between different traditions could offer resistance to some of the problems which are inherent in the traditional explanation models for the Homeric epics. Investigations with regard to Homeric syntax and its oral character can thus throw additional light on its evolution. The oral nature of Homeric syntax also needs to be kept in mind for further research on the cognitive character of Greek colometry. Following the example of Vergote's thesis (2011), comprehensive statistical studies about the synapheia of Homeric verse and the relation between metrical anomalies and caesurae can offer interesting information. In addition, one can study the cognitive cola of Aeolic metrics and their relation to the hexameter, keeping an eye on some common patterns of localization. For comparative research in Indo-European linguistics, Chafe's concept of information units can shed new light on the colometry of Indo-Iranian verse, because the Vedic and Avestan strophes came also into being in a traditional, oral context. Finally, some further research needs to be done with regard to Witte's proposal about the origins of Greek epic verse. One can examine more fully the colometry of dactylic tetrameters as they are attested in Greek literature. Which caesurae are likely to be original ones and which one were later on influenced by the hexameter? In addition, statistical studies can be conducted concerning the prevalence of archaisms and particular types of information units at the beginning of the verse. The statistics about the information units after the bucolic diaeresis need to be extended to the whole Homeric corpus. Moreover, as was mentioned above, comparative research on iambic and Byzantine metres can offer some typological parallels. With regard to the coalescence of an original tetrameter and a dimeter, attention needs to be paid to diachronic, morphosyntactic analyses of formulaic phrases which could explain how the two shorter metres could be combined into one, longer verse. Ending our odyssey through time to find some indications about the provenance of Homer's verse, we can be rather positive about the results. It could be stressed many times that the use of modern, cognitive linguistic theory can enlarge our understanding of Homeric linguistics, both on a synchronic and a diachronic level. Certainly, the origins of the hexameter remain hypothetical in nature and difficult to deal with, but the application of a refined methodology can point in the right direction. Maybe this paper did not find the cavern of the dragon itself, but at least a basic map of the environments could be created. The uncharted areas need to be filled in with future research.

104

MAIN REFERENCES230

Abritta, A. (2015). On the role of accent in Ancient Greek poetry: Pitch patterns in the Homeric hexameter. QUDC, 111, 11-27. Allen, R. (2009). Orale elementen in de Homerische grammatica: Intonatie-eenheid en enjambement. Lampas, 42, 136-151. Andersen, Ø. & Haug, D. (2012a). Relative chronology in early Greek epic poetry. Cambridge: Cambridge University Press. Andersen, Ø. & Haug, D. (2012b). Introduction. In Idem (Edd.), Relative chronology in early Greek epic poetry, 1-19, Cambridge: Cambridge University Press. Bakker, E. J. (1990a). Homeric discourse and enjambment: A cognitive approach. TAPA, 70, 1- 21. Bakker, E. J. (1990b). Homerus als orale poëzie: de recente ontwikkelingen. Lampas, 23, 384- 405. Bakker, E. J. (1995). Noun-epithet formulas, Milman Parry, and the grammar of poetry. In Crielaard, J.P. (Ed.), Homeric Questions. Essays in philology, ancient history and archaeology, 97-126, Amsterdam: Amsterdam University Press. Bakker, E. J. (1997a). Poetry in speech: orality and Homeric discourse. Ithaca (N.Y.): Cornell University Press. Bakker, E. J. (1997b). The study of Homeric discourse. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 284-304, Leiden: Brill. Bakker, E. J. (2005). Pointing at the past: From formula to performance in Homeric poetics. Cambridge (Mass.): Harvard University Press. Bakker, E. J. (Ed.) (2010). A companion to the Ancient Greek language. Chichester, West Sussex, U.K.: Wiley-Blackwell. Bakker, S. J. (2009). The noun phrase in ancient Greek: A functional analysis of the order and articulation of NP constituents in Herodotus. Leiden: Brill. Barnes, H. R. (1979). Enjambment and oral composition. TAPA, 109, 1-10. Barnes, H. R. (1986). The colometric structure of Homeric hexameter. GRBS, 27, 125-150. Barnes, T. (2011). Homeric ἀνδροτῆτα καὶ ἥβην. JHS, 131, 1-13. Bassett, S. E. (1905). Notes on the bucolic diaeresis. TAPA, 36, 111-124. Bassett, S. E. (1917). The hephtemimeral caesura in Greek hexameter poetry. TAPA, 48, 85- 110. Bassett, S. E. (1919). The theory of the Homeric caesura according to the extant remains of the ancient doctrine, AJPh, 40, 343-372. Bassett S. E. (1926). The so-called emphatic position of the runover word in the Homeric hexameter. TAPA, 57, 116-148.

230 Text editions are not listed under "main references".

105

Beekes, R. P. (1972). On the structure of the Greek hexameter. O'Neill interpreted. Glotta, 50, 1-10. Bennet, J. (1997). Homer and the Bronze Age. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 509-534, Leiden: Brill. Berg, N. (1978). Parergon metricum: der Ursprung des griechischen Hexameters. MSS, 37, 11-36. Berg, N. & Lindeman, F. O. (1992). The etymology of Greek αὖοσ and Od. 19.327 αὐςταλϋοσ: Homeric metrics and linguistics - A question of priority. Glotta, 70, 181-196. Berg, N. & Haug, D. (2000). Dividing Homer. 2: Innovation vs. tradition in Homer. SymbOsl, 75, 5-23. Bolinger, D. L. (1952). Linear Modification. PMLA, 67(7), 1117-1144. Bowra, M. (1962a). Metre. In Wace, A. J. B. & Stubbings, F. H. (Edd.), A companion to Homer, 19-25, London: Macmillan. Bowra, M. (1962b). Style. In Wace, A. J. B. & Stubbings, F. H. (Edd.), A companion to Homer, 26-37, London: Macmillan. Cantilena, M. (1995). Il ponte di Nicanore. In Fantuzzi, M., & Pretagostini, R. (Edd.). Struttura e storia dell'esametro greco (volume 1), 9-67, Roma: Gruppo editoriale internazionale. Chadwick, J. (1990). The descent of Greek epic. JHS, 110, 174-177. Chafe, W. L. (1982). Integration and involvement in speaking, writing, and oral literature. In Tannen, D. (Ed.), Spoken and written language. Exploring orality and literacy, 35-53, Norwood, NJ: Ablex. Chafe, W. L. (1985). Linguistic differences produced by differences between speaking and writing. In Olson, D. R., Torrance, N. & Hildyard, A. (Edd.), Literacy, language and learning. The nature and consequences of reading and writing, 105-123, Cambridge: Cambridge University Press. Chafe, W. L. (1987). Cognitive constraints on information flow. In Tomlin, R.S. (Ed.), Coherence and grounding in discourse, 21-51, Amsterdam: Benjamins. Chafe, W. L. (1994). Discourse, consciousness and time. The flow and displacement of conscious experience in speaking and writing. Cambridge: Cambridge University Press. Chantraine, P. (1999²). Dictionnaire étymologique de la langue grecque: histoire des mots. Paris: Klincksieck. Chantraine, P. (2013²). Grammaire homérique: Tome I Phonétique et morphologie. Nouvelle édition revue et corrigée par Michel Casevitz. Paris: Klincksieck. Chantraine, P. (2015²). Grammaire homérique: Tome II Syntaxe Nouvelle édition revue et corrigée par Michel Casevitz. Paris: Klincksieck. Clark, M. (2004). Formulas, metre and type-scenes. In Fowler, R. (Ed.), The Cambridge companion to Homer, 117-138, Cambridge: Cambridge University Press. Clayman, D. L. (1981). Sentence length in Greek hexameter poetry. In Grothjahn, R. (Ed.), Hexameter studies, 107-136, Bochum: Brockmeyer.

106

Colvin, S. (2016). The modal particle in Greek. CCJ, 62, 65-84. Crespo, E. (1997). L'ordre de préférence des éléments linguistiques de l'épopée. In Létoublon, F. (Ed.) (1997). Hommage à Milman Parry: Le style formulaire de l'épopée homérique et la théorie de l'oralité poétique., 129-136, Amsterdam: Gieben. Dain, A. (1965). Traité du métrique grecque. Paris: Klinksieck. Daitz, S. G. (1991). On reading Homer aloud: To pause or not to pause? AJPh, 112, 149-160. Davison, J. A. (1962). The Homeric question. In Wace, A. J. B. & Stubbings, F. H. (Edd.), A companion to Homer, 234-266, London: Macmillan. De Decker, F. (2015). Another attempt at a chronology of Grassmann's Law in Greek. JIES, 43, 140-177. De Decker, F. (2016). The augment use in Iliad 6: an evidential marker. LEC, 84(4), 259-317. De Decker, F. (2017). Ὅμηρον ἐξ Ὁμήρου ςαφηνύζειν: An analysis of the augment use in Iliad 1. JIES, 45, 58-171. De Decker, F. (2019). The augment use in the Homeric Hymn to Demeter (HH2). Glotta, 95. In Press. de Jong, I. J. F. (Ed.) (2012). Iliad: book XXII (CGLC). Cambridge: Cambridge University Press. De Lamberterie, C. (1997). Milman Parry et Antoine Meillet. In Létoublon, F. (Ed.) (1997). Hommage à Milman Parry: Le style formulaire de l'épopée homérique et la théorie de l'oralité poétique., 9-22, Amsterdam: Gieben. Dik, H. J. M. (2007). Word order in Greek tragic dialogue. Oxford: Oxford University Press. Dover, K. J. (1960). Greek word order. Cambridge: Cambridge University Press. Dowden, K. (2004). The epic tradition in Greece. In Fowler, R. (Ed.), The Cambridge companion to Homer, 188-205, Cambridge: Cambridge University Press. Dressler, W. (1969). Eine textsyntaktische Regel der idg. Wortstellung. KZ, 83(1), 1-25. Dressler, W. (1971). Über die Rekonstruktion der indogermanischen Syntax. KZ, 85(1), 5-22. Edwards, M. W. (1966). Some features of Homeric craftsmanship. TAPA, 97, 115-179. Edwards, M. W. (1986). Homer and oral tradition. The formula. Part I. Oral Tradition, 1-2, 171-230. Edwards, M. W. (1988). Homer and oral tradition. The formula. Part II. Oral Tradition, 3, 11- 60. Edwards, M. W. (1997). Homeric style and oral poetics. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 261-283, Leiden: Brill. Fantuzzi, M. (1984). Preistoria dell' esametro e storia della cultura greca arcaica: a proposito di alcuni studi recenti. MD, 12, 35-60. Fantuzzi, M., & Pretagostini, R. (Edd.) (1995). Struttura e storia dell'esametro greco (Two volumes). Roma: Gruppo editoriale internazionale. Finkelberg, M. (Ed.) (2011). The Homer encyclopedia (three volumes). Chichester: Wiley- Blackwell.

107

Foley, J. M. (1997). Oral tradition and its implications. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 146-173, Leiden: Brill. Forbes, K. (1958). The relations of the particle ἄν with κε(ν), κα, καν. Glotta, 37, 179-182. Fowler, R. L. (Ed.) (2004a). The Cambridge companion to Homer. Cambridge: Cambridge University Press. Fowler, R. L. (2004b). The Homeric question. In Idem (Ed.), The Cambridge companion to Homer, 220-233, Cambridge: Cambridge University Press. Fraenkel H. (1968³). Der homerische und der kallimachische Hexameter. In Idem (Ed.), Wege und Formen frühgriechischen Denkens. Literarische und philosophiegeschichtliche Studien, 100-156, München: Beck. Frisk, H. (1960). Griechisches etymologisches Wörterbuch (Teil I). Heidelberg: Winter. Frisk, H. (1970). Griechisches etymologisches Wörterbuch (Teil II). Heidelberg: Winter. Frisk, H. (1972). Griechisches etymologisches Wörterbuch (Supplementum). Heidelberg: Winter. Gentili, B. & Giannini, P. (1977). Preistoria e formazione dell'esametro. QUCC, 26, 7-51. Giannini, P. (1986). Pre-history and origin of the hexameter. MPhL, 7, 45-90. Gentili, B. & Giannini, P. (1995). Preistoria e formazione dell'esametro. In Fantuzzi, M., & Pretagostini, R. (Edd.), Struttura e storia dell'esametro greco (Volume 2), 11-62, Roma: Gruppo editoriale internazionale. Graziosi, B. & Haubold, J. (Edd.) (2010). Iliad: book VI (CGLC). Cambridge: Cambridge University Press. Hackstein, O. (2002). Die Sprachform der homerischen Epen: Faktoren morphologischer Variabilität in literarischen Frühformen, Tradition, Sprachwandel, sprachliche Anachronismen. Wiesbaden: Reichert. Hackstein, O. (2010). The Greek of epic. In Bakker, E. J. (Ed.), A companion to the Ancient Greek language, 401-423, Chichester, West Sussex, U.K.: Wiley-Blackwell. Hackstein, O. & Gunkel, D. (Edd.) (2018). Language and Meter. Leiden: Brill. Hainsworth, J. B. (1968). The flexibility of the Homeric formula. Oxford: Clarendon Press. Hainsworth, J. B. (1993). The Iliad: A commentary Volume III: books 9-12 (General editor: G.S.Kirk). Cambridge: Cambridge University Press. Hajnal, I. (2003). Der epische Hexameter im Rahmen der Homer-Troia-Debatte. In Ulf, Chr. (Ed.) Der neue Streit um Troia. Eine Bilanz, 217-231, Munich: Beck. Haslam, M. W. (1976). Review of Nagy (1974). JHS, 96, 202-203. Haug, D. (2002). Les phases de l'évolution de la langue épique: trois études de linguistique homérique. Göttingen: Vandenhoeck und Ruprecht. Haug, D. (2012). Tmesis in the epic tradition. In Andersen, Ø. & Haug, D. (Edd.), Relative chronology in early Greek epic poetry, 96-105, Cambridge: Cambridge University Press. Hermann, G. (Ed.) (1805). Orphica. Leipzig: Fritsch.

108

Higbie, C. (1990). Measure and music. Enjambment and sentence structure in the Iliad. Oxford: Oxford University Press. Higbie, C. (1995). Archaic hexameter: the Iliad, Theogony and Erga. In Fantuzzi, M., & Pretagostini, R. (Edd.). Struttura e storia dell'esametro greco (volume 1), 69-120, Roma: Gruppo editoriale internazionale. Hinrichs, G. (1875). De Homericae elocutionis vestigiis Aeolicis. Jena: Editores Fromannum. Hoekstra, A. (1965). Homeric modifications of formulaic prototypes: Studies in the development of Greek epic diction. Amsterdam: North-Holland Publishing Company. Hoekstra, A. (1981). Epic verse before Homer: Three studies. Amsterdam: North-Holland Publishing Company. Horrocks, G. (1997). Homer's dialect. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 191-217, Leiden: Brill. Ingalls, W. B. (1970). The structure of the Homeric hexameter. A review. Phoenix, 14, 1-12. Irigoin, J. (2004). Césure et diction du vers. Quelques réalités linguistiques à ne pas oublier. In Spaltenstein, F., Bianchi, O., Steinrück, M. & Lukinovich, A. (Edd.), Autour de la césure, 1-10, Berne: Lang. Janko, R. (1982). Homer, Hesiod, and the Hymns: Diachronic development in epic diction. Cambridge: Cambridge University Press. Janko, R. (1992). The Iliad: A commentary Volume IV: books 13-16 (General editor: G.S.Kirk). Cambridge: Cambridge University Press. Janko, R. (2012). πρῶτόν τε καὶ ὕςτατον αἰὲν ἀείδειν: relative chronology and the literary history of the early Greek epos. In Andersen, Ø. & Haug, D. (Edd.), Relative chronology in early Greek epic poetry, 20-43, Cambridge: Cambridge University Press. Janse, M. (1991). La phrase segmentée en grec ancien - le témoignage des enclitiques. BSL, 86, xiv-xvi. Janse, M. (1998). Prosodie en metriek van het Homerische epos: orale poëzie in de praktijk. DCG, 38, 125-151. Janse, M. (2003). The metrical schemes of the hexameter. Mnemosyne, 56, 343-348. Janse, M. (2012). Inleiding tot de Homerische taal en metriek. Ghent: Academia Press. Jones, B. (2012). Relative chronology and an 'Aeolic phase' of epic. In Andersen, Ø. & Haug, D. (Edd.), Relative chronology in early Greek epic poetry, 44-64, Cambridge: Cambridge University Press Kiparsky, P. (2018). Indo-European origins of the Greek hexameter. (online paper at https://web.stanford.edu/~kiparsky/Papers/hexameter.pdf; last time accessed at the 22th may 2018; final edition is published in Hackstein & Gunkel (2018)). Kirk, G. S. (1962). The songs of Homer. Cambridge: Cambridge University Press. Kirk, G. S. (1965). Homer and the epic: A shortened version of "The songs of Homer". Cambridge: Cambridge University Press. Kirk, G. S. (1966). Studies in some technical aspects of Homeric style. YCIS, 20, 73-152.

109

Kirk, G. S. (1976). Homer and the oral tradition. Cambridge: Cambridge University Press. Kirk, G. S. (1985). The Iliad: A commentary Volume I: books 1-4 (General editor: G.S.Kirk). Cambridge: Cambridge University Press. Korzeniewski, D. (1968). Griechische Metrik. Darmstadt: Wissenschaftliche Buchgesellschaft. Koster, W. J. W. (1936). Traité du métrique grecque. Leyde: Sijthoff. Kousoulini, V. (2013). Alcmanic hexameters and early hexametric poetry: Alcman’s poetry in its oral context. GRBS, 53, 420-440. Krisch, T. (1990). Das Wackernagelsche Gesetz aus heutiger Sicht. In Eichner, H. & Rix, H. (Edd.), Sprachwissenschaft und Philologie: Jacob Wackernagel und die Indogermanistik heute, 64-81, Wiesbaden: Reichert. Kuhn, A. (1853). Ueber die durch Nasale erweiterten Verbalstämme. KZ, 2, 455-471. Kurylowicz, J. (1970). The quantative meter of Indo-European. In Cardona, G., Hoenigswald, H. M., Senn, A. (Edd.) Indo-European and Indo-Europeans, 421-430, Philadelphia: University of Philadelphia Press. Latacz, J. (Ed.) (2000-). Homers Ilias: Gesamtkommentar (Basler Kommentar/BK). München: Saur. Lehrs, K. (1860). Einige Bemerkungen zur Caesur des Hexameters. JCPh, 6, 513-531. Létoublon, F. (Ed.) (1997). Hommage à Milman Parry: Le style formulaire de l'épopée homérique et la théorie de l'oralité poétique. Amsterdam: Gieben. Lord, A. B. (1960). The singer of tales. Cambridge: Harvard University Press Maas, P. (1962). Greek metre. Oxford: Clarendon Press. MacDonell, A. A. (1916). A Vedic grammar for students. Oxford: Clarendon Press. Magnelli, E. (1995). Studi recenti sull'origine dell'esametro: un profilo critico. In Fantuzzi, M., & Pretagostini, R. (Edd.). Struttura e storia dell'esametro greco (volume 2), 111-138, Roma: Gruppo editoriale internazionale. Mahoney, A. (2007). The feet of Greek and Sanskrit verse. CCJ, supp. 32, 179-187. McLennan, G., R. (1974). Hiatus in the Homeric hexameter. QUCC, 17, 131-135. Mehler, J. & Mehler, E. (196813). Mehler woordenboek op de gedichten van Homeros. 's- Gravenhage-Rotterdam: Nijgh en Van Ditmar. Meier-Brügger, M. (2003). Die homerische Kunstsprache. In Ulf, Chr. (Ed.) Der neue Streit um Troia. Eine Bilanz, 232-244, Munich: Beck. Meillet, A. (1923). Les origines indo-européennes des mètres grecs. Paris: Presses Universitaires de France. Meister, K. (1921). Die homerische Kunstsprache. Stuttgart: Teubner. Mette, H. J. (1956). Die Struktur des ältesten daktylischen Hexameters. Glotta, 35, 1-17. Meyer, W. (1884). Zur Geschichte des griechischen und lateinischen Hexameters. SBAW, 1884 (6), 979-1089.

110

Miller, D. G. (1982). Homer and the Ionian epic tradition: Some phonic and phonological evidence against an Aeolic "phase". Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Miller, D. G. (2014). Ancient Greek dialects and early authors: Introduction to the dialect mixture in Homer, with notes on lyric and Herodotus. Berlin: De Gruyter. Monro, D. B. (1891). A grammar of the Homeric dialect. Oxford: Clarendon Press. Morris, I. & Powell, B. B. (Edd.) (1997). A new companion to Homer. Leiden: Brill. Nagy, G. (1974). Comparative studies in Greek and Indic meter. Cambridge (Mass.): Harvard University Press. Nagy, G. (1979). On the origins of the Greek hexameter. Synchronic and diachronic perspectives. In Brogyani, B. (Ed.), Studies in diachronic, synchronic and typological linguistics. Festschrift for Oswald Szemerényi on the occasion of his 65th birthday, 611-631, Amsterdam: Benjamins. Nagy, G. (1990). Pindar's Homer: The lyric possession of an epic past. Baltimore: John Hopkins University Press. Nagy, G. (1996). Poetry as performance: Homer and beyond. Cambridge: Cambridge University Press. Nagy, G. (1998). Is there an etymology for the dactylic hexameter? In Melchert, H.C., Jasanoff, J.H., Oliver, L. (Edd.), Mir curad: studies in honor of Calvert Watkins, 495-508, Innsbruck: Insititut fur Sprachwissenschaft der Universität Innsbruck. Nagy, G. (2004). Homer's text and language. Urbana (Ill.): University of Illinois Press. Nagy, G. (2010). Language and meter. In Bakker, E. J. (Ed.) A companion to the Ancient Greek language, 370-387, Chichester, West Sussex, U.K.: Wiley-Blackwell. Nagy, G. (2011). The Aeolic component of Homeric diction. In Jamison, S. W. et al. (Edd.). Proceedings of the 22nd annual UCLA Indo-European conference., 133-179, Bremen: Ute Hempen Verlag. (online version at http://nrs.harvard.edu/urn3:hlnc.essay:Nagy.The_Aeolic_Component_of_Homeric_Diction.2 011.) Nasta, M. (1994). Principes d'analyse d'une versification orale. 1: Les factures spécifiques de l'hexamètre grec. LÉC, 62, 101-129. Nasta, M. (1995). Principes d'analyse d'une versification orale. 2: Caractères traditionnels de la versification épique. LÉC, 63, 197-223. O'Neill E. G., J. (1942). The localization of metrical word-types in the Greek hexameter. Homer, Hesiod and the Alexandrians. YCS, 8, 102-176. Oswald, S. (2014). Metrical laws. In Giannakis, G. K. (Ed.), Encyclopedia of Ancient Greek language and linguistics (Vol. 2: G-O), 419-423, Leiden: Brill. Palmer, L. R. (1962). The language of Homer. In Wace, A. J. B. & Stubbings, F. H. (Edd.), A Companion to Homer, 75-178, London: Macmillan.

111

Parry, M., & Parry, A. (1971). The making of Homeric verse: The collected papers of Milman Parry. Oxford: Clarendon Press. Peabody, B. (1975). The winged word: A study in the technique of Ancient Greek oral composition as seen principally through Hesiod's Works and Days. Albany: State University of New York Press. Porter, H. N. (1951). The early Greek hexameter. YCS, 12, 1-63. Ritoók, Z. (1987). Vermutungen zum Ursprung des griechischen Hexameters. Philologus, 131, 2-18. Rix, H. (1992²). Historische Grammatik des Griechischen : Laut- und Formenlehre. Darmstadt: Wissenschaftliche Buchgesellschaft. Rossi, L. E. (1995). Estenstione e valore del colon nell'esametro omerico. In Fantuzzi, M., & Pretagostini, R. (Edd.). Struttura e storia dell'esametro greco (Volume 2), 271-320, Roma: Gruppo editoriale internazionale. Ruijgh, C. J. (1957). L'élément achéen dans la langue epique. Assen: van Gorcum. Ruijgh, C. J. (1990). La place des enclitiques dans l'ordre des mots chez Homère d'après la loi de Wackernagel. In Eichner, H. & Rix, H. (Edd.), Sprachwissenschaft und Philologie: Jacob Wackernagel und die Indogermanistik heute, 213-233, Wiesbaden: Reichert. Ruijgh, C. J. (1995). D'Homère aux origines proto-mycéniennes de la tradition épique. Analyse dialectologique du langage homérique, avec un excursus sur la création de l'alphabet grec. In Crielaard, J.P. (Ed.), Homeric Questions. Essays in Philology, Ancient History and Archaeology, 1-96, Amsterdam: Amsterdam University Press. Ruijgh, C. J. (1997). Les origines proto-mycéniennes de la tradition épique. In Létoublon, F. (Ed.). Hommage à Milman Parry: Le style formulaire de l'épopée homérique et la théorie de l'oralité poétique., 33-46, Amsterdam: Gieben. Russo, J. (1963). A closer look at homeric formulas. TAPA, 94, 235-247. Russo, J. (1966). The structural formula in Homeric verse. YCS, 20, 219-240. Russo, J. (1997). The formula. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 238-260, Leiden: Brill. Schwyzer, E. (1950). Griechische Grammatik. Zweiter Band. Syntax und syntaktische Stilistik. München: Beck. Sicking, C. M. J. (1993). Griechische Verslehre. München: Beck. Slings, S.R. (1994). Een tandje lager: aanzetten voor een orale grammatica van Homerus. Lampas, 27, 411-427. Steinrück, M. (1995). Sprechpause und Wortende: zum Rhythmus des Hexameters. QUCC, 49, 135-140. Steinrück, M. (2005). Lagaroi. Le temps de la re-rhythmisation de l'hexamètre. Mnemosyne, 58, 481-498. Steinrück, M. (2010-2011). Remarques sur la loi de Meyer-Fränkel. IFC, 10, 273-278.

112

Steinrück, M., & Lukinovich, A. (2004). Avant-propos: la césure, un problème complexe. In Spaltenstein, F., Bianchi, O., Steinrück, M. & Lukinovich, A. (Edd.), Autour de la césure, IX- XXIII, Berne: Lang. Stifler, Th. (1924). Das Wernickesche Gesetz und die Bukolische Dihärese. Philologus, 79, 323-354. Tichy, E. (1981). Hom. ἀνδροτῆτα und die Vorgeschichte des daktylischen Hexameters. Glotta, 59, 28-67. Tichy, E. (2010). Älter als der Hexameter? Schiffskatalog, Troerkatalog und vier Einzelszenen der « Ilias ». Bremen: Hempen. Tsopanakis, A. G. (1983). Homeric researches: Irregularity and the construction of the verse. Θεςςαλονίκη: Αριςτοτέλειο Πανεπιςτήμιο Θεςςαλονίκησ. Tribulato, O. (2010). Literary Dialects. In Bakker, E. J. (Ed.). A companion to the Ancient Greek language, 388-400, Chichester, West Sussex, U.K.: Wiley-Blackwell. Ulf, Chr. (Ed.) (2003). Der neue Streit um Troia. Eine Bilanz. Munich: Beck. van Beek, L. (2013). The development of the Proto-Indo-European syllabic liquids in Greek. (Unpublished PhD-Thesis). Leyden University. Vergote, S. M. (2011). Denken in metrum: Metrische argumenten voor een cognitivistische verssegmentatie bij Homeros. (Unpublished MA.-Thesis). Ghent University. Vigorita J., F. (1977). The Indo-European origins of the Greek hexameter and distich. ZVS, 91, 288-299. Visser, E. (1987). Homerische Versifikationstechnik: Versuch einer Rekonstruktion. Frankfurt am Main: Lang. Visser, E. (1997). Die Formel als Resultat frühepischer Versifikationstechnik. Generative Versbildung und die Gestaltung von Katalogversen. In Létoublon, F. (Ed.) (1997). Hommage à Milman Parry: Le style formulaire de l'épopée homérique et la théorie de l'oralité poétique., 159-172, Amsterdam: Gieben. Vivante, P. (1997). Homeric rhythm. A philosophical study. Westport: Greenwood Press. Wace, A. J. B. & Stubbings, F. H. (Edd.) (1962). A companion to Homer. London: Macmillan. Wachter, R. (2000). Grammatik der homerischen Sprache. In Latacz, J. et al. (Edd.) Homer Ilias: Gesamtkommentar: Prolegomena, 61-108, München: Saur. Wachter, R. (2012). The other view: focus on linguistic innovations in the Homeric epics. In Andersen, Ø. & Haug, D. (Edd.), Relative chronology in early Greek epic poetry, 65-79, Cambridge: Cambridge University Press. Wackernagel, J. (1898). Über ein Gesetz der Indogermanischen Wortstellung. IF, 1, 333-446. Wathelet, P. (1997). L'œuvre de Milman Parry et l'analyse de la langue épique. In Létoublon, F. (Ed.) (1997). Hommage à Milman Parry: Le style formulaire de l'épopée homérique et la théorie de l'oralité poétique., 47-56, Amsterdam: Gieben. Watkins, C. (1995). How to kill a dragon? Aspects of Indo-European poetics. Oxford: Oxford University Press.

113

Weilo, E., & Haug, D. T. (2001). The proto-hexameter hypothesis: perspectives for further research. SymbOsl, 76, 130-136. West, M. L. (1973a). Greek poetry 2000-700 BC. CQ, 23, 179-192. West, M. L. (1973b). Indo-European metre. Glotta, 51, 161-187. West, M. L. (1974). Review of Nagy (1974). Phoenix, 28, 457-459. West, M. L. (1982). Greek metre. Oxford: Clarendon Press. West, M. L. (1987). Introduction to Greek metre. Oxford: Clarendon Press. West, M. L. (1988). The rise of the Greek epic. JHS, 108, 151-172. West, M. L. (1992). The descent of the Greek epic: a reply. JHS, 112, 173-175. West, M. (1997). Homer's meter. In Morris, I. & Powell, B. B. (Edd.), A new companion to Homer, 218-237, Leiden: Brill. West, M. L. (1999). The invention of Homer. CQ, 49, 364-382. West, M. L. (2001). Studies in the text and transmission of the Iliad. München: Saur. West, M. L. (2007). Indo-European poetry and myth. Oxford: Oxford University Press. West, M. L. (2011a). The making of the Iliad: Disquisition and analytical commentary. Oxford: Oxford University Press. West, M. L. (2011b). Review of Tichy (2010). Kratylos, 56, 156-163. West, M. L. (2014). The making of the Odyssey. Oxford: Oxford University Press. West, M. L. (2018). Unmetrical verses in Homer. (Unpublished manuscript consulted; published in Hackstein & Gunkel (2018)). Witte, K. (1913). 'Homeros B) Sprache', In Pauly-Wissowa (Edd.), Realencyclopädie der classichen Altertumswissenschaft, VIII: 2213-2247. Witte, K. (1915). Wortrhytmus bei Homer. RhM, 70, 481-523. Witte, K. (1972). Zur homerischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. (= based on papers published in Glotta, 1-5 (1909-1914)).

114

APPENDIX: Syntactic statistics for the bucolic diaeresis (Iliad I, XI, XVI).

In this appendix, I give an overview of my statistic research concerning the syntactic structures which are used after the bucolic diaeresis (cf. Section 3.2.2 in the core text). I collected all instances of a bucolic diaeresis in books I, XI and XVI of the Iliad. To be as objective as possible, I based it on the simple existence of word end. Cases where the bucolic diaeresis would be followed by a postpositive word were removed from the list. Different statistic structures were grouped together in broader categories. Some of them need some further remarks. Firstly, when different words were combined, e.g. noun + verb, word order does not bother us here. Coordinate constructions refer to instances where by means of a conjunction as καί, τε, ἤ etc. a parallel construction begins at the bucolic diaeresis or the beginning of a new clause. Instances which were difficult to place in these categories are classified as varia, being only a neglectable amount. The statistics below give all the instances per category in the individual books, with their percentages (rounded off to the nearest decimal). At the end, the concluding table for the three books together, which was discussed in the thesis itself, is reprinted here once again.

1. Iliad I

Totality of verses: 611 Verses with BD: 372 (without postpositives in verses 308, 492, 577, 585) Percentage of verses with BD: 60,9% Syntactic structures:

Structure Total % Combination of noun and verb (VO/OV) Instances: 2, 4, 18, 19, 23, 24, 25, 26, 29, 36, 38, 39, 40, 44, 47, 48, 63, 67, 77, 89, 97, 106, 110, 112, 114, 124, 126, 134, 137, 147, 151, 153, 166, 169, 171, 178, 197, 198, 200, 205, 210, 218, 224, 230, 233, 236, 239, 243, 244, 246, 254, 261, 269, 278, 279, 280, 286, 287, 297, 314, 324, 325, 326, 328, 337, 338, 341, 351, 356, 360, 362, 363, 365, 377, 378, 379, 381, 398, 399, 405, 409, 412, 413, 414, 424, 441, 445, 450, 452, 456, 464, 469, 480, 486, 495, 498, 507, 512, 528, 533, 535, 552, 563, 566, 567, 581, 583, 593, 610 109 29,3 Noun phrase, consisting of adjective and noun Instances: 7, 15, 42, 43, 58, 64, 72, 80, 84, 92, 111, 121, 145, 148, 175, 182, 213, 215, 219, 222, 292, 300, 306, 307, 312, 329, 357, 358, 364, 374, 393, 400, 402, 426, 433, 457, 462, 468, 482, 489, 508, 551, 568, 602 44 11,8 Simple adjective (apposition), simple noun or adjective/ noun accompanied with a genitive, both pronominal and nominal

115

Instances: 11, 12, 14, 21, 28, 53, 74, 103, 143, 149, 152, 155, 157, 162, 184, 203, 229, 237, 240, 242, 263, 265, 305, 310, 316, 323, 327, 346, 347, 366, 368, 369, 370, 371, 373, 383, 404, 419, 421, 423, 425, 438, 439, 454, 473, 484, 488, 490, 496, 499, 559, 570, 580, 584, 591, 605, 609 57 15,3 Varia 7 1,9 Instances: 54, 61, 234, 390, 513, 547, 596 Constructions with an adverb, a preposition or a negation and a verb or a noun 18 4,8 Instances: 27, 30, 71, 120, 141, 142, 156, 180, 290, 298, 336, 349, 354, 367, 418, 432, 494, 541 Simple verb without noun Instances: 31, 37, 46, 87, 88, 91, 107, 129, 189, 202, 216, 226, 228, 232, 241, 257, 258, 291, 31 8,3 353, 386, 446, 451, 453, 461, 472, 474, 571, 586, 587, 589, 600 Subordinate clauses Instances: 32, 78, 82, 83, 86, 116, 118, 128, 139, 158, 163, 186, 207, 211, 238, 261, 283, 32 8,6 294, 340, 394, 420, 509, 515, 522, 523, 524, 543, 554, 555, 558, 578, 603 Coordinate clauses and noun-phrases Instances: 50, 62, 65, 68, 73, 76, 93, 96, 101, 108, 115, 117, 119, 127, 132, 133, 138, 154, 160, 170, 173, 188, 193, 194, 196, 199, 209, 214, 217, 220, 247, 251, 253, 260, 262, 282, 295, 309, 318, 332, 333, 334, 335, 348, 359, 361, 380, 382, 384, 387, 395, 406, 407, 416, 428, 430, 481, 497, 500, 514, 526, 531, 534, 536, 542, 548, 550, 553, 74 19,9 557, 561, 562, 573, 575 TOTAL 372 99,9 (100)

2. Iliad XI

Totality of verses: 848 Verses with BD: 566 (without postpositives in verses 10, 98, 340, 352, 604, 606, 683) Percentage of verses with BD: 66,7% Syntactic structures:

Structure Total % Combination of noun and verb (VO/OV) Instances: 4, 11, 30, 33, 58, 69, 79, 81, 88, 89, 94, 115, 116, 131, 138, 142, 143, 144, 146, 151, 155, 158, 170, 176, 179, 180, 186, 189, 191, 194, 204, 206, 209, 211, 214, 218, 219, 225, 227, 231, 233, 235, 237, 243, 244, 288, 290, 298, 300, 306, 309, 310, 312, 320, 330, 334, 338, 342, 356, 357, 359, 375, 376, 381, 393, 394, 398, 402, 413, 426, 429, 432, 433, 440, 444, 454, 460, 466, 467, 477, 480, 487, 488, 489, 490, 498, 502, 515, 520, 522, 539, 544, 546, 547, 550, 569, 573, 579, 585, 594, 611, 613, 625, 632, 635, 640, 645, 652, 670, 671, 681, 688, 692, 693, 704, 716, 724, 737, 743, 745, 749, 750, 755, 756, 764, 781, 782, 137 24,2 784, 786, 792, 793, 796, 821, 825, 834, 839, 844 Noun phrase, consisting of adjective and noun Instances: 5, 8, 14, 16, 32, 62, 84, 95, 110, 111, 112, 113, 124, 127, 169, 174, 195, 196, 197,

116

199, 210, 213, 226, 229, 241, 247, 258, 278, 287, 289, 295, 297, 313, 322, 327, 347, 353, 355, 360, 363, 383, 387, 421, 428, 435, 441, 443, 449, 452, 455, 456, 472, 478, 484, 496, 504, 507, 510, 516, 566, 575, 577, 588, 599, 605, 607, 620, 644, 655, 674, 686, 698, 708, 710, 723, 729, 738, 741, 747, 752, 760, 767, 795, 814, 817, 818, 819, 828, 829, 830, 836, 94 16,6 837, 838, 845 Simple adjective (apposition), simple noun or adjective/ noun accompanied with a genitive, both pronominal and nominal Instances: 1, 3, 9, 25, 34, 41, 57, 60, 66, 67, 77, 82, 92, 96, 128, 135, 166, 173, 183, 187, 198, 202, 222, 224, 253, 259, 265, 268, 270, 272, 277, 286, 291, 292, 328, 364, 369, 370, 372, 373, 374, 385, 390, 412, 417, 420, 424, 431, 438, 450, 457, 459, 465, 501, 505, 506, 511, 513, 518, 541, 545, 548, 549, 557, 561, 578, 595, 598, 601, 614, 617, 619, 631, 651, 656, 672, , 675, 676, 678, 680, 694, 696, 712, 720, 726, 734, 739, 770, 773, 785, 800, 812, 823, 835, 840, 842 97 17,1 Varia 14 2,5 Instances: 6, 102, 118, 269, 351, 474, 583, 662, 707, 721, 758, 802, 810, 820 Constructions with an adverb, a preposition or a negation and a verb or a noun Instances: 44, 48, 50, 61, 80, 119, 129, 141, 212, 223, 263, 266, 282, 284, 285, 343, 346, 377, 378, 388, 392, 396, 415, 434, 447, 464, 486, 495, 559, 565, 570, 572, 580, 612, 634, 50 8,8 663, 666, 677, 689, 722, 730, 748, 765, 768, 771, 787, 789, 813, 822, 827 Simple verb without noun Instances: 36, 40, 49, 51, 53, 73, 74, 101, 103, 130, 145, 167, 185, 192, 201, 207, 230, 245, 246, 299, 335, 365, 368, 386, 419, 422, 423, 436, 479, 537, 596, 654, 673, 685, 695, 709, 715, 725, 753, 769, 809, 841 42 7,4 Subordinate clauses Instances: 21, 27, 54, 76, 104, 123, 228, 324, 329, 367, 404, 405, 475, 485, 499, 528, 535, 25 4,4 554, 626, 653, 691, 731, 757, 791, 847 Coordinate clauses and noun-phrases Instances: 2, 12, 13, 29, 38, 52, 65, 72, 75, 83, 85, 91, 97, 109, 116, 156, 161, 162, 163, 164, 178, 181, 216, 236, 239, 240, 260, 267, 274, 276, 293, 302, 304, 307, 314, 317, 319, 321, 326, 331, 339, 341, 354, 362, 380, 389, 395, 400, 401, 409, 410, 411, 430, 437, 451, 453, 458, 461, 473, 483, 491, 497, 512, 517, 524, 526, 527, 534, 538, 560, 582, 587, 589, 590, 592, 610, 622, 629, 633, 647, 648, 650, 657, 658, 659, 661, 664, 665, 668, 697, 703, 706, 714, 717, 759, 761, 762, 763, 776, 780, 788, 790, 803, 816, 826, 833, 848 107 18,9 TOTAL 566 99,9 (100)

3. Iliad XVI

Totality of verses: 867 Verses with BD: 531 (without postpositives in verses 63, 102, 136, 410, 716) Percentage of verses with BD: 61,3% Syntactic structures:

117

Structure Total % Combination of noun and verb (VO/OV) Instances: 11, 19, 23, 29, 32, 34, 35, 38, 47, 52, 64, 68, 69, 75, 80, 82, 83, 84, 86, 106, 112, 120, 121, 128, 129, 142, 152, 153, 166, 192, 199, 203, 206, 219, 221, 247, 255, 261, 264, 266, 274, 291, 293, 302, 309, 311, 316, 325, 339, 355, 365, 368, 369, 378, 383, 384, 405, 412, 424, 425, 426, 427, 440, 444, 449, 455, 457, 464, 467, 468, 469, 472, 478, 479, 480, 484, 497, 503, 504, 505, 516, 522, 523, 529, 534, 538, 546, 575, 578, 587, 595, 597, 599, 605, 608, 615, 624, 629, 634, 643, 646, 650, 652, 655, 656, 663, 667, 670, 675, 678, 680, 699, 704, 722, 730, 733, 734, 735, 736, 747, 753, 755, 772, 780, 782, 785, 799, 800, 810, 817, 821, 825, 828, 831, 836, 842, 846, 848, 849, 851, 861, 863 142 26,7 Noun phrase, consisting of adjective and noun Instances: 5, 33, 37, 48, 51, 103, 122, 130, 148, 157, 159, 185, 190, 226, 230, 241, 244, 249, 256, 265, 270, 278, 283, 284, 299, 300, 301, 307, 317, 318, 327, 334, 343, 345, 347, 357, 360, 370, 375, 380, 381, 399, 407, 408, 409, , 437, 439, 445, 454, 466, 477, 483, 501, 514, 527, 535, 539, 543, 561, 571, 577, 588, 603, 610, 613, 623, 626, 635, 645, 649, 658, 665, 673, 677, 683, 705, 712, 715, 720, 727, 739, 760, 761, 786, 793, 805, 806, 807, 819, 827, 832, 833, 853, 858, 859, 866, 867 87 16,4 Simple adjective (apposition), simple noun or adjective/ noun accompanied with a genitive, both pronominal and nominal Instances: 2, 14, 15, 21, 31, 42, 56, 77, 91, 113, 126, 134, 135, 140, 141, 144, 151, 154, 165, 167, 173, 181, 187, 189, 194, 197, 205, 210, 214, 217, 222, 232, 237, 240, 245, 275, 281, 286, 287, 315, 323, 329, 332, 342, 358, 361, 371, 392, 397, 401, 414, 418, 428, 431, 433, 436, 476, 486, 490, 507, 513, 528, 536, 541, 547, 549, 558, 570, 574, 580, 584, 585, 591, 593, 596, 635, 637, 654, 698, 702, 711, 717, 719, 729, 737, 765, 774, 776, 779, 803, 809, 815, 822, 839, 840, 854, 860, 865 97 18,3 Varia 9 1,7 Instances: 27, 44, 55, 351, 367, 519, 725, 743, 804 Constructions with an adverb, a preposition or a negation and a verb or a noun 32 6 Instances: 74, 95, 114, 115, 147, 208, 227, 233, 268, 288, 289, 324, 346, 364, 388, 461, 471, 481, 515, 619, 640, 709, 710, 713, 726, 741, 745, 749, 789, 794, 796, 862 Simple verb without noun Instances: 16, 28, 40, 66, 73, 76, 81, 92, 96, 111, 158, 176, 179, 182, 193, 218, 236, 257, 259, 279, 290, 350, 404, 435, 442, 485, 496, 506, 531, 533, 573, 647, 692, 714, 754, 756, 769, 777, 790, 847, 850 42 7,9 Subordinate clauses Instances: 10, 17, 30, 49, 50, 78, 87, 98, 168, 171, 242, 271, 314, 328, 337, 353, 354, 406, 30 5,6 423, 494, 499, 524, 604, 618, 621, 641, 672, 682, 835, 845 Coordinate clauses and noun-phrases Instances: 7, 12, 24, 26, 45, 46, 60, 61, 62, 105, 107, 109, 110, 117, 119, 124, 156, 162, 164, 169, 186, 191, 207, 215, 220, 225, 231, 234, 276, 294, 312, 322, 331, 333, 335, 338, 340, 349, 352, 356, 374, 376, 395, 396, 400, 403, 422, 465, 470, 474, 482, 492, 510, 517, 518, 520, 530, 542, 552, 553, 557, 560, 590, 601, 606, 616, 622, 631, 639, 644, 657, 659, 668, 685, 688, 695, 696, 707, 721, 728, 731, 740, 762, 763, 778, 791, 802, 813, 814, 820, 834, 844, 852 92 17,3

118

TOTAL 531 99,9 (100)

4. Concluding table

Totality of verses: 2326 Verses with BD: 1469 (without postpositives) Percentage of verses with BD: 63,2%

Concluding table with the highest percentage at the beginning and the lowest at the end of the table.

Syntactic structure Number of examples Percentage Combination of noun and verb (VO/OV) 388 26,4 Coordinate clauses and noun-phrases 273 18,6 Simple adjective (apposition), simple noun or 251 17,1 adjective/ noun accompanied with a genitive, both pronominal and nominal Noun phrase, consisting of adjective and noun 225 15,3 Simple verb without noun 115 7,8 Constructions with an adverb, a preposition or a 100 6,8 negation and a verb or a noun Subordinate clauses 87 5,9 Varia 30 2 Total 1469 99,9% (100%)

119